Show More
@@ -1,3925 +1,3928 b'' | |||
|
1 | 1 | # perf.py - performance test routines |
|
2 | 2 | '''helper extension to measure performance |
|
3 | 3 | |
|
4 | 4 | Configurations |
|
5 | 5 | ============== |
|
6 | 6 | |
|
7 | 7 | ``perf`` |
|
8 | 8 | -------- |
|
9 | 9 | |
|
10 | 10 | ``all-timing`` |
|
11 | 11 | When set, additional statistics will be reported for each benchmark: best, |
|
12 | 12 | worst, median average. If not set only the best timing is reported |
|
13 | 13 | (default: off). |
|
14 | 14 | |
|
15 | 15 | ``presleep`` |
|
16 | 16 | number of second to wait before any group of runs (default: 1) |
|
17 | 17 | |
|
18 | 18 | ``pre-run`` |
|
19 | 19 | number of run to perform before starting measurement. |
|
20 | 20 | |
|
21 | 21 | ``profile-benchmark`` |
|
22 | 22 | Enable profiling for the benchmarked section. |
|
23 | 23 | (The first iteration is benchmarked) |
|
24 | 24 | |
|
25 | 25 | ``run-limits`` |
|
26 | 26 | Control the number of runs each benchmark will perform. The option value |
|
27 | 27 | should be a list of `<time>-<numberofrun>` pairs. After each run the |
|
28 | 28 | conditions are considered in order with the following logic: |
|
29 | 29 | |
|
30 | 30 | If benchmark has been running for <time> seconds, and we have performed |
|
31 | 31 | <numberofrun> iterations, stop the benchmark, |
|
32 | 32 | |
|
33 | 33 | The default value is: `3.0-100, 10.0-3` |
|
34 | 34 | |
|
35 | 35 | ``stub`` |
|
36 | 36 | When set, benchmarks will only be run once, useful for testing |
|
37 | 37 | (default: off) |
|
38 | 38 | ''' |
|
39 | 39 | |
|
40 | 40 | # "historical portability" policy of perf.py: |
|
41 | 41 | # |
|
42 | 42 | # We have to do: |
|
43 | 43 | # - make perf.py "loadable" with as wide Mercurial version as possible |
|
44 | 44 | # This doesn't mean that perf commands work correctly with that Mercurial. |
|
45 | 45 | # BTW, perf.py itself has been available since 1.1 (or eb240755386d). |
|
46 | 46 | # - make historical perf command work correctly with as wide Mercurial |
|
47 | 47 | # version as possible |
|
48 | 48 | # |
|
49 | 49 | # We have to do, if possible with reasonable cost: |
|
50 | 50 | # - make recent perf command for historical feature work correctly |
|
51 | 51 | # with early Mercurial |
|
52 | 52 | # |
|
53 | 53 | # We don't have to do: |
|
54 | 54 | # - make perf command for recent feature work correctly with early |
|
55 | 55 | # Mercurial |
|
56 | 56 | |
|
57 | 57 | from __future__ import absolute_import |
|
58 | 58 | import contextlib |
|
59 | 59 | import functools |
|
60 | 60 | import gc |
|
61 | 61 | import os |
|
62 | 62 | import random |
|
63 | 63 | import shutil |
|
64 | 64 | import struct |
|
65 | 65 | import sys |
|
66 | 66 | import tempfile |
|
67 | 67 | import threading |
|
68 | 68 | import time |
|
69 | 69 | from mercurial import ( |
|
70 | 70 | changegroup, |
|
71 | 71 | cmdutil, |
|
72 | 72 | commands, |
|
73 | 73 | copies, |
|
74 | 74 | error, |
|
75 | 75 | extensions, |
|
76 | 76 | hg, |
|
77 | 77 | mdiff, |
|
78 | 78 | merge, |
|
79 | 79 | revlog, |
|
80 | 80 | util, |
|
81 | 81 | ) |
|
82 | 82 | |
|
83 | 83 | # for "historical portability": |
|
84 | 84 | # try to import modules separately (in dict order), and ignore |
|
85 | 85 | # failure, because these aren't available with early Mercurial |
|
86 | 86 | try: |
|
87 | 87 | from mercurial import branchmap # since 2.5 (or bcee63733aad) |
|
88 | 88 | except ImportError: |
|
89 | 89 | pass |
|
90 | 90 | try: |
|
91 | 91 | from mercurial import obsolete # since 2.3 (or ad0d6c2b3279) |
|
92 | 92 | except ImportError: |
|
93 | 93 | pass |
|
94 | 94 | try: |
|
95 | 95 | from mercurial import registrar # since 3.7 (or 37d50250b696) |
|
96 | 96 | |
|
97 | 97 | dir(registrar) # forcibly load it |
|
98 | 98 | except ImportError: |
|
99 | 99 | registrar = None |
|
100 | 100 | try: |
|
101 | 101 | from mercurial import repoview # since 2.5 (or 3a6ddacb7198) |
|
102 | 102 | except ImportError: |
|
103 | 103 | pass |
|
104 | 104 | try: |
|
105 | 105 | from mercurial.utils import repoviewutil # since 5.0 |
|
106 | 106 | except ImportError: |
|
107 | 107 | repoviewutil = None |
|
108 | 108 | try: |
|
109 | 109 | from mercurial import scmutil # since 1.9 (or 8b252e826c68) |
|
110 | 110 | except ImportError: |
|
111 | 111 | pass |
|
112 | 112 | try: |
|
113 | 113 | from mercurial import setdiscovery # since 1.9 (or cb98fed52495) |
|
114 | 114 | except ImportError: |
|
115 | 115 | pass |
|
116 | 116 | |
|
117 | 117 | try: |
|
118 | 118 | from mercurial import profiling |
|
119 | 119 | except ImportError: |
|
120 | 120 | profiling = None |
|
121 | 121 | |
|
122 | 122 | |
|
123 | 123 | def identity(a): |
|
124 | 124 | return a |
|
125 | 125 | |
|
126 | 126 | |
|
127 | 127 | try: |
|
128 | 128 | from mercurial import pycompat |
|
129 | 129 | |
|
130 | 130 | getargspec = pycompat.getargspec # added to module after 4.5 |
|
131 | 131 | _byteskwargs = pycompat.byteskwargs # since 4.1 (or fbc3f73dc802) |
|
132 | 132 | _sysstr = pycompat.sysstr # since 4.0 (or 2219f4f82ede) |
|
133 | 133 | _bytestr = pycompat.bytestr # since 4.2 (or b70407bd84d5) |
|
134 | 134 | _xrange = pycompat.xrange # since 4.8 (or 7eba8f83129b) |
|
135 | 135 | fsencode = pycompat.fsencode # since 3.9 (or f4a5e0e86a7e) |
|
136 | 136 | if pycompat.ispy3: |
|
137 | 137 | _maxint = sys.maxsize # per py3 docs for replacing maxint |
|
138 | 138 | else: |
|
139 | 139 | _maxint = sys.maxint |
|
140 | 140 | except (NameError, ImportError, AttributeError): |
|
141 | 141 | import inspect |
|
142 | 142 | |
|
143 | 143 | getargspec = inspect.getargspec |
|
144 | 144 | _byteskwargs = identity |
|
145 | 145 | _bytestr = str |
|
146 | 146 | fsencode = identity # no py3 support |
|
147 | 147 | _maxint = sys.maxint # no py3 support |
|
148 | 148 | _sysstr = lambda x: x # no py3 support |
|
149 | 149 | _xrange = xrange |
|
150 | 150 | |
|
151 | 151 | try: |
|
152 | 152 | # 4.7+ |
|
153 | 153 | queue = pycompat.queue.Queue |
|
154 | 154 | except (NameError, AttributeError, ImportError): |
|
155 | 155 | # <4.7. |
|
156 | 156 | try: |
|
157 | 157 | queue = pycompat.queue |
|
158 | 158 | except (NameError, AttributeError, ImportError): |
|
159 | 159 | import Queue as queue |
|
160 | 160 | |
|
161 | 161 | try: |
|
162 | 162 | from mercurial import logcmdutil |
|
163 | 163 | |
|
164 | 164 | makelogtemplater = logcmdutil.maketemplater |
|
165 | 165 | except (AttributeError, ImportError): |
|
166 | 166 | try: |
|
167 | 167 | makelogtemplater = cmdutil.makelogtemplater |
|
168 | 168 | except (AttributeError, ImportError): |
|
169 | 169 | makelogtemplater = None |
|
170 | 170 | |
|
171 | 171 | # for "historical portability": |
|
172 | 172 | # define util.safehasattr forcibly, because util.safehasattr has been |
|
173 | 173 | # available since 1.9.3 (or 94b200a11cf7) |
|
174 | 174 | _undefined = object() |
|
175 | 175 | |
|
176 | 176 | |
|
177 | 177 | def safehasattr(thing, attr): |
|
178 | 178 | return getattr(thing, _sysstr(attr), _undefined) is not _undefined |
|
179 | 179 | |
|
180 | 180 | |
|
181 | 181 | setattr(util, 'safehasattr', safehasattr) |
|
182 | 182 | |
|
183 | 183 | # for "historical portability": |
|
184 | 184 | # define util.timer forcibly, because util.timer has been available |
|
185 | 185 | # since ae5d60bb70c9 |
|
186 | 186 | if safehasattr(time, 'perf_counter'): |
|
187 | 187 | util.timer = time.perf_counter |
|
188 | 188 | elif os.name == b'nt': |
|
189 | 189 | util.timer = time.clock |
|
190 | 190 | else: |
|
191 | 191 | util.timer = time.time |
|
192 | 192 | |
|
193 | 193 | # for "historical portability": |
|
194 | 194 | # use locally defined empty option list, if formatteropts isn't |
|
195 | 195 | # available, because commands.formatteropts has been available since |
|
196 | 196 | # 3.2 (or 7a7eed5176a4), even though formatting itself has been |
|
197 | 197 | # available since 2.2 (or ae5f92e154d3) |
|
198 | 198 | formatteropts = getattr( |
|
199 | 199 | cmdutil, "formatteropts", getattr(commands, "formatteropts", []) |
|
200 | 200 | ) |
|
201 | 201 | |
|
202 | 202 | # for "historical portability": |
|
203 | 203 | # use locally defined option list, if debugrevlogopts isn't available, |
|
204 | 204 | # because commands.debugrevlogopts has been available since 3.7 (or |
|
205 | 205 | # 5606f7d0d063), even though cmdutil.openrevlog() has been available |
|
206 | 206 | # since 1.9 (or a79fea6b3e77). |
|
207 | 207 | revlogopts = getattr( |
|
208 | 208 | cmdutil, |
|
209 | 209 | "debugrevlogopts", |
|
210 | 210 | getattr( |
|
211 | 211 | commands, |
|
212 | 212 | "debugrevlogopts", |
|
213 | 213 | [ |
|
214 | 214 | (b'c', b'changelog', False, b'open changelog'), |
|
215 | 215 | (b'm', b'manifest', False, b'open manifest'), |
|
216 | 216 | (b'', b'dir', False, b'open directory manifest'), |
|
217 | 217 | ], |
|
218 | 218 | ), |
|
219 | 219 | ) |
|
220 | 220 | |
|
221 | 221 | cmdtable = {} |
|
222 | 222 | |
|
223 | 223 | # for "historical portability": |
|
224 | 224 | # define parsealiases locally, because cmdutil.parsealiases has been |
|
225 | 225 | # available since 1.5 (or 6252852b4332) |
|
226 | 226 | def parsealiases(cmd): |
|
227 | 227 | return cmd.split(b"|") |
|
228 | 228 | |
|
229 | 229 | |
|
230 | 230 | if safehasattr(registrar, 'command'): |
|
231 | 231 | command = registrar.command(cmdtable) |
|
232 | 232 | elif safehasattr(cmdutil, 'command'): |
|
233 | 233 | command = cmdutil.command(cmdtable) |
|
234 | 234 | if 'norepo' not in getargspec(command).args: |
|
235 | 235 | # for "historical portability": |
|
236 | 236 | # wrap original cmdutil.command, because "norepo" option has |
|
237 | 237 | # been available since 3.1 (or 75a96326cecb) |
|
238 | 238 | _command = command |
|
239 | 239 | |
|
240 | 240 | def command(name, options=(), synopsis=None, norepo=False): |
|
241 | 241 | if norepo: |
|
242 | 242 | commands.norepo += b' %s' % b' '.join(parsealiases(name)) |
|
243 | 243 | return _command(name, list(options), synopsis) |
|
244 | 244 | |
|
245 | 245 | |
|
246 | 246 | else: |
|
247 | 247 | # for "historical portability": |
|
248 | 248 | # define "@command" annotation locally, because cmdutil.command |
|
249 | 249 | # has been available since 1.9 (or 2daa5179e73f) |
|
250 | 250 | def command(name, options=(), synopsis=None, norepo=False): |
|
251 | 251 | def decorator(func): |
|
252 | 252 | if synopsis: |
|
253 | 253 | cmdtable[name] = func, list(options), synopsis |
|
254 | 254 | else: |
|
255 | 255 | cmdtable[name] = func, list(options) |
|
256 | 256 | if norepo: |
|
257 | 257 | commands.norepo += b' %s' % b' '.join(parsealiases(name)) |
|
258 | 258 | return func |
|
259 | 259 | |
|
260 | 260 | return decorator |
|
261 | 261 | |
|
262 | 262 | |
|
263 | 263 | try: |
|
264 | 264 | import mercurial.registrar |
|
265 | 265 | import mercurial.configitems |
|
266 | 266 | |
|
267 | 267 | configtable = {} |
|
268 | 268 | configitem = mercurial.registrar.configitem(configtable) |
|
269 | 269 | configitem( |
|
270 | 270 | b'perf', |
|
271 | 271 | b'presleep', |
|
272 | 272 | default=mercurial.configitems.dynamicdefault, |
|
273 | 273 | experimental=True, |
|
274 | 274 | ) |
|
275 | 275 | configitem( |
|
276 | 276 | b'perf', |
|
277 | 277 | b'stub', |
|
278 | 278 | default=mercurial.configitems.dynamicdefault, |
|
279 | 279 | experimental=True, |
|
280 | 280 | ) |
|
281 | 281 | configitem( |
|
282 | 282 | b'perf', |
|
283 | 283 | b'parentscount', |
|
284 | 284 | default=mercurial.configitems.dynamicdefault, |
|
285 | 285 | experimental=True, |
|
286 | 286 | ) |
|
287 | 287 | configitem( |
|
288 | 288 | b'perf', |
|
289 | 289 | b'all-timing', |
|
290 | 290 | default=mercurial.configitems.dynamicdefault, |
|
291 | 291 | experimental=True, |
|
292 | 292 | ) |
|
293 | 293 | configitem( |
|
294 | 294 | b'perf', |
|
295 | 295 | b'pre-run', |
|
296 | 296 | default=mercurial.configitems.dynamicdefault, |
|
297 | 297 | ) |
|
298 | 298 | configitem( |
|
299 | 299 | b'perf', |
|
300 | 300 | b'profile-benchmark', |
|
301 | 301 | default=mercurial.configitems.dynamicdefault, |
|
302 | 302 | ) |
|
303 | 303 | configitem( |
|
304 | 304 | b'perf', |
|
305 | 305 | b'run-limits', |
|
306 | 306 | default=mercurial.configitems.dynamicdefault, |
|
307 | 307 | experimental=True, |
|
308 | 308 | ) |
|
309 | 309 | except (ImportError, AttributeError): |
|
310 | 310 | pass |
|
311 | 311 | except TypeError: |
|
312 | 312 | # compatibility fix for a11fd395e83f |
|
313 | 313 | # hg version: 5.2 |
|
314 | 314 | configitem( |
|
315 | 315 | b'perf', |
|
316 | 316 | b'presleep', |
|
317 | 317 | default=mercurial.configitems.dynamicdefault, |
|
318 | 318 | ) |
|
319 | 319 | configitem( |
|
320 | 320 | b'perf', |
|
321 | 321 | b'stub', |
|
322 | 322 | default=mercurial.configitems.dynamicdefault, |
|
323 | 323 | ) |
|
324 | 324 | configitem( |
|
325 | 325 | b'perf', |
|
326 | 326 | b'parentscount', |
|
327 | 327 | default=mercurial.configitems.dynamicdefault, |
|
328 | 328 | ) |
|
329 | 329 | configitem( |
|
330 | 330 | b'perf', |
|
331 | 331 | b'all-timing', |
|
332 | 332 | default=mercurial.configitems.dynamicdefault, |
|
333 | 333 | ) |
|
334 | 334 | configitem( |
|
335 | 335 | b'perf', |
|
336 | 336 | b'pre-run', |
|
337 | 337 | default=mercurial.configitems.dynamicdefault, |
|
338 | 338 | ) |
|
339 | 339 | configitem( |
|
340 | 340 | b'perf', |
|
341 | 341 | b'profile-benchmark', |
|
342 | 342 | default=mercurial.configitems.dynamicdefault, |
|
343 | 343 | ) |
|
344 | 344 | configitem( |
|
345 | 345 | b'perf', |
|
346 | 346 | b'run-limits', |
|
347 | 347 | default=mercurial.configitems.dynamicdefault, |
|
348 | 348 | ) |
|
349 | 349 | |
|
350 | 350 | |
|
351 | 351 | def getlen(ui): |
|
352 | 352 | if ui.configbool(b"perf", b"stub", False): |
|
353 | 353 | return lambda x: 1 |
|
354 | 354 | return len |
|
355 | 355 | |
|
356 | 356 | |
|
357 | 357 | class noop(object): |
|
358 | 358 | """dummy context manager""" |
|
359 | 359 | |
|
360 | 360 | def __enter__(self): |
|
361 | 361 | pass |
|
362 | 362 | |
|
363 | 363 | def __exit__(self, *args): |
|
364 | 364 | pass |
|
365 | 365 | |
|
366 | 366 | |
|
367 | 367 | NOOPCTX = noop() |
|
368 | 368 | |
|
369 | 369 | |
|
370 | 370 | def gettimer(ui, opts=None): |
|
371 | 371 | """return a timer function and formatter: (timer, formatter) |
|
372 | 372 | |
|
373 | 373 | This function exists to gather the creation of formatter in a single |
|
374 | 374 | place instead of duplicating it in all performance commands.""" |
|
375 | 375 | |
|
376 | 376 | # enforce an idle period before execution to counteract power management |
|
377 | 377 | # experimental config: perf.presleep |
|
378 | 378 | time.sleep(getint(ui, b"perf", b"presleep", 1)) |
|
379 | 379 | |
|
380 | 380 | if opts is None: |
|
381 | 381 | opts = {} |
|
382 | 382 | # redirect all to stderr unless buffer api is in use |
|
383 | 383 | if not ui._buffers: |
|
384 | 384 | ui = ui.copy() |
|
385 | 385 | uifout = safeattrsetter(ui, b'fout', ignoremissing=True) |
|
386 | 386 | if uifout: |
|
387 | 387 | # for "historical portability": |
|
388 | 388 | # ui.fout/ferr have been available since 1.9 (or 4e1ccd4c2b6d) |
|
389 | 389 | uifout.set(ui.ferr) |
|
390 | 390 | |
|
391 | 391 | # get a formatter |
|
392 | 392 | uiformatter = getattr(ui, 'formatter', None) |
|
393 | 393 | if uiformatter: |
|
394 | 394 | fm = uiformatter(b'perf', opts) |
|
395 | 395 | else: |
|
396 | 396 | # for "historical portability": |
|
397 | 397 | # define formatter locally, because ui.formatter has been |
|
398 | 398 | # available since 2.2 (or ae5f92e154d3) |
|
399 | 399 | from mercurial import node |
|
400 | 400 | |
|
401 | 401 | class defaultformatter(object): |
|
402 | 402 | """Minimized composition of baseformatter and plainformatter""" |
|
403 | 403 | |
|
404 | 404 | def __init__(self, ui, topic, opts): |
|
405 | 405 | self._ui = ui |
|
406 | 406 | if ui.debugflag: |
|
407 | 407 | self.hexfunc = node.hex |
|
408 | 408 | else: |
|
409 | 409 | self.hexfunc = node.short |
|
410 | 410 | |
|
411 | 411 | def __nonzero__(self): |
|
412 | 412 | return False |
|
413 | 413 | |
|
414 | 414 | __bool__ = __nonzero__ |
|
415 | 415 | |
|
416 | 416 | def startitem(self): |
|
417 | 417 | pass |
|
418 | 418 | |
|
419 | 419 | def data(self, **data): |
|
420 | 420 | pass |
|
421 | 421 | |
|
422 | 422 | def write(self, fields, deftext, *fielddata, **opts): |
|
423 | 423 | self._ui.write(deftext % fielddata, **opts) |
|
424 | 424 | |
|
425 | 425 | def condwrite(self, cond, fields, deftext, *fielddata, **opts): |
|
426 | 426 | if cond: |
|
427 | 427 | self._ui.write(deftext % fielddata, **opts) |
|
428 | 428 | |
|
429 | 429 | def plain(self, text, **opts): |
|
430 | 430 | self._ui.write(text, **opts) |
|
431 | 431 | |
|
432 | 432 | def end(self): |
|
433 | 433 | pass |
|
434 | 434 | |
|
435 | 435 | fm = defaultformatter(ui, b'perf', opts) |
|
436 | 436 | |
|
437 | 437 | # stub function, runs code only once instead of in a loop |
|
438 | 438 | # experimental config: perf.stub |
|
439 | 439 | if ui.configbool(b"perf", b"stub", False): |
|
440 | 440 | return functools.partial(stub_timer, fm), fm |
|
441 | 441 | |
|
442 | 442 | # experimental config: perf.all-timing |
|
443 | 443 | displayall = ui.configbool(b"perf", b"all-timing", False) |
|
444 | 444 | |
|
445 | 445 | # experimental config: perf.run-limits |
|
446 | 446 | limitspec = ui.configlist(b"perf", b"run-limits", []) |
|
447 | 447 | limits = [] |
|
448 | 448 | for item in limitspec: |
|
449 | 449 | parts = item.split(b'-', 1) |
|
450 | 450 | if len(parts) < 2: |
|
451 | 451 | ui.warn((b'malformatted run limit entry, missing "-": %s\n' % item)) |
|
452 | 452 | continue |
|
453 | 453 | try: |
|
454 | 454 | time_limit = float(_sysstr(parts[0])) |
|
455 | 455 | except ValueError as e: |
|
456 | 456 | ui.warn( |
|
457 | 457 | ( |
|
458 | 458 | b'malformatted run limit entry, %s: %s\n' |
|
459 | 459 | % (_bytestr(e), item) |
|
460 | 460 | ) |
|
461 | 461 | ) |
|
462 | 462 | continue |
|
463 | 463 | try: |
|
464 | 464 | run_limit = int(_sysstr(parts[1])) |
|
465 | 465 | except ValueError as e: |
|
466 | 466 | ui.warn( |
|
467 | 467 | ( |
|
468 | 468 | b'malformatted run limit entry, %s: %s\n' |
|
469 | 469 | % (_bytestr(e), item) |
|
470 | 470 | ) |
|
471 | 471 | ) |
|
472 | 472 | continue |
|
473 | 473 | limits.append((time_limit, run_limit)) |
|
474 | 474 | if not limits: |
|
475 | 475 | limits = DEFAULTLIMITS |
|
476 | 476 | |
|
477 | 477 | profiler = None |
|
478 | 478 | if profiling is not None: |
|
479 | 479 | if ui.configbool(b"perf", b"profile-benchmark", False): |
|
480 | 480 | profiler = profiling.profile(ui) |
|
481 | 481 | |
|
482 | 482 | prerun = getint(ui, b"perf", b"pre-run", 0) |
|
483 | 483 | t = functools.partial( |
|
484 | 484 | _timer, |
|
485 | 485 | fm, |
|
486 | 486 | displayall=displayall, |
|
487 | 487 | limits=limits, |
|
488 | 488 | prerun=prerun, |
|
489 | 489 | profiler=profiler, |
|
490 | 490 | ) |
|
491 | 491 | return t, fm |
|
492 | 492 | |
|
493 | 493 | |
|
494 | 494 | def stub_timer(fm, func, setup=None, title=None): |
|
495 | 495 | if setup is not None: |
|
496 | 496 | setup() |
|
497 | 497 | func() |
|
498 | 498 | |
|
499 | 499 | |
|
500 | 500 | @contextlib.contextmanager |
|
501 | 501 | def timeone(): |
|
502 | 502 | r = [] |
|
503 | 503 | ostart = os.times() |
|
504 | 504 | cstart = util.timer() |
|
505 | 505 | yield r |
|
506 | 506 | cstop = util.timer() |
|
507 | 507 | ostop = os.times() |
|
508 | 508 | a, b = ostart, ostop |
|
509 | 509 | r.append((cstop - cstart, b[0] - a[0], b[1] - a[1])) |
|
510 | 510 | |
|
511 | 511 | |
|
512 | 512 | # list of stop condition (elapsed time, minimal run count) |
|
513 | 513 | DEFAULTLIMITS = ( |
|
514 | 514 | (3.0, 100), |
|
515 | 515 | (10.0, 3), |
|
516 | 516 | ) |
|
517 | 517 | |
|
518 | 518 | |
|
519 | 519 | def _timer( |
|
520 | 520 | fm, |
|
521 | 521 | func, |
|
522 | 522 | setup=None, |
|
523 | 523 | title=None, |
|
524 | 524 | displayall=False, |
|
525 | 525 | limits=DEFAULTLIMITS, |
|
526 | 526 | prerun=0, |
|
527 | 527 | profiler=None, |
|
528 | 528 | ): |
|
529 | 529 | gc.collect() |
|
530 | 530 | results = [] |
|
531 | 531 | begin = util.timer() |
|
532 | 532 | count = 0 |
|
533 | 533 | if profiler is None: |
|
534 | 534 | profiler = NOOPCTX |
|
535 | 535 | for i in range(prerun): |
|
536 | 536 | if setup is not None: |
|
537 | 537 | setup() |
|
538 | 538 | func() |
|
539 | 539 | keepgoing = True |
|
540 | 540 | while keepgoing: |
|
541 | 541 | if setup is not None: |
|
542 | 542 | setup() |
|
543 | 543 | with profiler: |
|
544 | 544 | with timeone() as item: |
|
545 | 545 | r = func() |
|
546 | 546 | profiler = NOOPCTX |
|
547 | 547 | count += 1 |
|
548 | 548 | results.append(item[0]) |
|
549 | 549 | cstop = util.timer() |
|
550 | 550 | # Look for a stop condition. |
|
551 | 551 | elapsed = cstop - begin |
|
552 | 552 | for t, mincount in limits: |
|
553 | 553 | if elapsed >= t and count >= mincount: |
|
554 | 554 | keepgoing = False |
|
555 | 555 | break |
|
556 | 556 | |
|
557 | 557 | formatone(fm, results, title=title, result=r, displayall=displayall) |
|
558 | 558 | |
|
559 | 559 | |
|
560 | 560 | def formatone(fm, timings, title=None, result=None, displayall=False): |
|
561 | 561 | |
|
562 | 562 | count = len(timings) |
|
563 | 563 | |
|
564 | 564 | fm.startitem() |
|
565 | 565 | |
|
566 | 566 | if title: |
|
567 | 567 | fm.write(b'title', b'! %s\n', title) |
|
568 | 568 | if result: |
|
569 | 569 | fm.write(b'result', b'! result: %s\n', result) |
|
570 | 570 | |
|
571 | 571 | def display(role, entry): |
|
572 | 572 | prefix = b'' |
|
573 | 573 | if role != b'best': |
|
574 | 574 | prefix = b'%s.' % role |
|
575 | 575 | fm.plain(b'!') |
|
576 | 576 | fm.write(prefix + b'wall', b' wall %f', entry[0]) |
|
577 | 577 | fm.write(prefix + b'comb', b' comb %f', entry[1] + entry[2]) |
|
578 | 578 | fm.write(prefix + b'user', b' user %f', entry[1]) |
|
579 | 579 | fm.write(prefix + b'sys', b' sys %f', entry[2]) |
|
580 | 580 | fm.write(prefix + b'count', b' (%s of %%d)' % role, count) |
|
581 | 581 | fm.plain(b'\n') |
|
582 | 582 | |
|
583 | 583 | timings.sort() |
|
584 | 584 | min_val = timings[0] |
|
585 | 585 | display(b'best', min_val) |
|
586 | 586 | if displayall: |
|
587 | 587 | max_val = timings[-1] |
|
588 | 588 | display(b'max', max_val) |
|
589 | 589 | avg = tuple([sum(x) / count for x in zip(*timings)]) |
|
590 | 590 | display(b'avg', avg) |
|
591 | 591 | median = timings[len(timings) // 2] |
|
592 | 592 | display(b'median', median) |
|
593 | 593 | |
|
594 | 594 | |
|
595 | 595 | # utilities for historical portability |
|
596 | 596 | |
|
597 | 597 | |
|
598 | 598 | def getint(ui, section, name, default): |
|
599 | 599 | # for "historical portability": |
|
600 | 600 | # ui.configint has been available since 1.9 (or fa2b596db182) |
|
601 | 601 | v = ui.config(section, name, None) |
|
602 | 602 | if v is None: |
|
603 | 603 | return default |
|
604 | 604 | try: |
|
605 | 605 | return int(v) |
|
606 | 606 | except ValueError: |
|
607 | 607 | raise error.ConfigError( |
|
608 | 608 | b"%s.%s is not an integer ('%s')" % (section, name, v) |
|
609 | 609 | ) |
|
610 | 610 | |
|
611 | 611 | |
|
612 | 612 | def safeattrsetter(obj, name, ignoremissing=False): |
|
613 | 613 | """Ensure that 'obj' has 'name' attribute before subsequent setattr |
|
614 | 614 | |
|
615 | 615 | This function is aborted, if 'obj' doesn't have 'name' attribute |
|
616 | 616 | at runtime. This avoids overlooking removal of an attribute, which |
|
617 | 617 | breaks assumption of performance measurement, in the future. |
|
618 | 618 | |
|
619 | 619 | This function returns the object to (1) assign a new value, and |
|
620 | 620 | (2) restore an original value to the attribute. |
|
621 | 621 | |
|
622 | 622 | If 'ignoremissing' is true, missing 'name' attribute doesn't cause |
|
623 | 623 | abortion, and this function returns None. This is useful to |
|
624 | 624 | examine an attribute, which isn't ensured in all Mercurial |
|
625 | 625 | versions. |
|
626 | 626 | """ |
|
627 | 627 | if not util.safehasattr(obj, name): |
|
628 | 628 | if ignoremissing: |
|
629 | 629 | return None |
|
630 | 630 | raise error.Abort( |
|
631 | 631 | ( |
|
632 | 632 | b"missing attribute %s of %s might break assumption" |
|
633 | 633 | b" of performance measurement" |
|
634 | 634 | ) |
|
635 | 635 | % (name, obj) |
|
636 | 636 | ) |
|
637 | 637 | |
|
638 | 638 | origvalue = getattr(obj, _sysstr(name)) |
|
639 | 639 | |
|
640 | 640 | class attrutil(object): |
|
641 | 641 | def set(self, newvalue): |
|
642 | 642 | setattr(obj, _sysstr(name), newvalue) |
|
643 | 643 | |
|
644 | 644 | def restore(self): |
|
645 | 645 | setattr(obj, _sysstr(name), origvalue) |
|
646 | 646 | |
|
647 | 647 | return attrutil() |
|
648 | 648 | |
|
649 | 649 | |
|
650 | 650 | # utilities to examine each internal API changes |
|
651 | 651 | |
|
652 | 652 | |
|
653 | 653 | def getbranchmapsubsettable(): |
|
654 | 654 | # for "historical portability": |
|
655 | 655 | # subsettable is defined in: |
|
656 | 656 | # - branchmap since 2.9 (or 175c6fd8cacc) |
|
657 | 657 | # - repoview since 2.5 (or 59a9f18d4587) |
|
658 | 658 | # - repoviewutil since 5.0 |
|
659 | 659 | for mod in (branchmap, repoview, repoviewutil): |
|
660 | 660 | subsettable = getattr(mod, 'subsettable', None) |
|
661 | 661 | if subsettable: |
|
662 | 662 | return subsettable |
|
663 | 663 | |
|
664 | 664 | # bisecting in bcee63733aad::59a9f18d4587 can reach here (both |
|
665 | 665 | # branchmap and repoview modules exist, but subsettable attribute |
|
666 | 666 | # doesn't) |
|
667 | 667 | raise error.Abort( |
|
668 | 668 | b"perfbranchmap not available with this Mercurial", |
|
669 | 669 | hint=b"use 2.5 or later", |
|
670 | 670 | ) |
|
671 | 671 | |
|
672 | 672 | |
|
673 | 673 | def getsvfs(repo): |
|
674 | 674 | """Return appropriate object to access files under .hg/store""" |
|
675 | 675 | # for "historical portability": |
|
676 | 676 | # repo.svfs has been available since 2.3 (or 7034365089bf) |
|
677 | 677 | svfs = getattr(repo, 'svfs', None) |
|
678 | 678 | if svfs: |
|
679 | 679 | return svfs |
|
680 | 680 | else: |
|
681 | 681 | return getattr(repo, 'sopener') |
|
682 | 682 | |
|
683 | 683 | |
|
684 | 684 | def getvfs(repo): |
|
685 | 685 | """Return appropriate object to access files under .hg""" |
|
686 | 686 | # for "historical portability": |
|
687 | 687 | # repo.vfs has been available since 2.3 (or 7034365089bf) |
|
688 | 688 | vfs = getattr(repo, 'vfs', None) |
|
689 | 689 | if vfs: |
|
690 | 690 | return vfs |
|
691 | 691 | else: |
|
692 | 692 | return getattr(repo, 'opener') |
|
693 | 693 | |
|
694 | 694 | |
|
695 | 695 | def repocleartagscachefunc(repo): |
|
696 | 696 | """Return the function to clear tags cache according to repo internal API""" |
|
697 | 697 | if util.safehasattr(repo, b'_tagscache'): # since 2.0 (or 9dca7653b525) |
|
698 | 698 | # in this case, setattr(repo, '_tagscache', None) or so isn't |
|
699 | 699 | # correct way to clear tags cache, because existing code paths |
|
700 | 700 | # expect _tagscache to be a structured object. |
|
701 | 701 | def clearcache(): |
|
702 | 702 | # _tagscache has been filteredpropertycache since 2.5 (or |
|
703 | 703 | # 98c867ac1330), and delattr() can't work in such case |
|
704 | 704 | if '_tagscache' in vars(repo): |
|
705 | 705 | del repo.__dict__['_tagscache'] |
|
706 | 706 | |
|
707 | 707 | return clearcache |
|
708 | 708 | |
|
709 | 709 | repotags = safeattrsetter(repo, b'_tags', ignoremissing=True) |
|
710 | 710 | if repotags: # since 1.4 (or 5614a628d173) |
|
711 | 711 | return lambda: repotags.set(None) |
|
712 | 712 | |
|
713 | 713 | repotagscache = safeattrsetter(repo, b'tagscache', ignoremissing=True) |
|
714 | 714 | if repotagscache: # since 0.6 (or d7df759d0e97) |
|
715 | 715 | return lambda: repotagscache.set(None) |
|
716 | 716 | |
|
717 | 717 | # Mercurial earlier than 0.6 (or d7df759d0e97) logically reaches |
|
718 | 718 | # this point, but it isn't so problematic, because: |
|
719 | 719 | # - repo.tags of such Mercurial isn't "callable", and repo.tags() |
|
720 | 720 | # in perftags() causes failure soon |
|
721 | 721 | # - perf.py itself has been available since 1.1 (or eb240755386d) |
|
722 | 722 | raise error.Abort(b"tags API of this hg command is unknown") |
|
723 | 723 | |
|
724 | 724 | |
|
725 | 725 | # utilities to clear cache |
|
726 | 726 | |
|
727 | 727 | |
|
728 | 728 | def clearfilecache(obj, attrname): |
|
729 | 729 | unfiltered = getattr(obj, 'unfiltered', None) |
|
730 | 730 | if unfiltered is not None: |
|
731 | 731 | obj = obj.unfiltered() |
|
732 | 732 | if attrname in vars(obj): |
|
733 | 733 | delattr(obj, attrname) |
|
734 | 734 | obj._filecache.pop(attrname, None) |
|
735 | 735 | |
|
736 | 736 | |
|
737 | 737 | def clearchangelog(repo): |
|
738 | 738 | if repo is not repo.unfiltered(): |
|
739 | 739 | object.__setattr__(repo, '_clcachekey', None) |
|
740 | 740 | object.__setattr__(repo, '_clcache', None) |
|
741 | 741 | clearfilecache(repo.unfiltered(), 'changelog') |
|
742 | 742 | |
|
743 | 743 | |
|
744 | 744 | # perf commands |
|
745 | 745 | |
|
746 | 746 | |
|
747 | 747 | @command(b'perf::walk|perfwalk', formatteropts) |
|
748 | 748 | def perfwalk(ui, repo, *pats, **opts): |
|
749 | 749 | opts = _byteskwargs(opts) |
|
750 | 750 | timer, fm = gettimer(ui, opts) |
|
751 | 751 | m = scmutil.match(repo[None], pats, {}) |
|
752 | 752 | timer( |
|
753 | 753 | lambda: len( |
|
754 | 754 | list( |
|
755 | 755 | repo.dirstate.walk(m, subrepos=[], unknown=True, ignored=False) |
|
756 | 756 | ) |
|
757 | 757 | ) |
|
758 | 758 | ) |
|
759 | 759 | fm.end() |
|
760 | 760 | |
|
761 | 761 | |
|
762 | 762 | @command(b'perf::annotate|perfannotate', formatteropts) |
|
763 | 763 | def perfannotate(ui, repo, f, **opts): |
|
764 | 764 | opts = _byteskwargs(opts) |
|
765 | 765 | timer, fm = gettimer(ui, opts) |
|
766 | 766 | fc = repo[b'.'][f] |
|
767 | 767 | timer(lambda: len(fc.annotate(True))) |
|
768 | 768 | fm.end() |
|
769 | 769 | |
|
770 | 770 | |
|
771 | 771 | @command( |
|
772 | 772 | b'perf::status|perfstatus', |
|
773 | 773 | [ |
|
774 | 774 | (b'u', b'unknown', False, b'ask status to look for unknown files'), |
|
775 | 775 | (b'', b'dirstate', False, b'benchmark the internal dirstate call'), |
|
776 | 776 | ] |
|
777 | 777 | + formatteropts, |
|
778 | 778 | ) |
|
779 | 779 | def perfstatus(ui, repo, **opts): |
|
780 | 780 | """benchmark the performance of a single status call |
|
781 | 781 | |
|
782 | 782 | The repository data are preserved between each call. |
|
783 | 783 | |
|
784 | 784 | By default, only the status of the tracked file are requested. If |
|
785 | 785 | `--unknown` is passed, the "unknown" files are also tracked. |
|
786 | 786 | """ |
|
787 | 787 | opts = _byteskwargs(opts) |
|
788 | 788 | # m = match.always(repo.root, repo.getcwd()) |
|
789 | 789 | # timer(lambda: sum(map(len, repo.dirstate.status(m, [], False, False, |
|
790 | 790 | # False)))) |
|
791 | 791 | timer, fm = gettimer(ui, opts) |
|
792 | 792 | if opts[b'dirstate']: |
|
793 | 793 | dirstate = repo.dirstate |
|
794 | 794 | m = scmutil.matchall(repo) |
|
795 | 795 | unknown = opts[b'unknown'] |
|
796 | 796 | |
|
797 | 797 | def status_dirstate(): |
|
798 | 798 | s = dirstate.status( |
|
799 | 799 | m, subrepos=[], ignored=False, clean=False, unknown=unknown |
|
800 | 800 | ) |
|
801 | 801 | sum(map(bool, s)) |
|
802 | 802 | |
|
803 | 803 | timer(status_dirstate) |
|
804 | 804 | else: |
|
805 | 805 | timer(lambda: sum(map(len, repo.status(unknown=opts[b'unknown'])))) |
|
806 | 806 | fm.end() |
|
807 | 807 | |
|
808 | 808 | |
|
809 | 809 | @command(b'perf::addremove|perfaddremove', formatteropts) |
|
810 | 810 | def perfaddremove(ui, repo, **opts): |
|
811 | 811 | opts = _byteskwargs(opts) |
|
812 | 812 | timer, fm = gettimer(ui, opts) |
|
813 | 813 | try: |
|
814 | 814 | oldquiet = repo.ui.quiet |
|
815 | 815 | repo.ui.quiet = True |
|
816 | 816 | matcher = scmutil.match(repo[None]) |
|
817 | 817 | opts[b'dry_run'] = True |
|
818 | 818 | if 'uipathfn' in getargspec(scmutil.addremove).args: |
|
819 | 819 | uipathfn = scmutil.getuipathfn(repo) |
|
820 | 820 | timer(lambda: scmutil.addremove(repo, matcher, b"", uipathfn, opts)) |
|
821 | 821 | else: |
|
822 | 822 | timer(lambda: scmutil.addremove(repo, matcher, b"", opts)) |
|
823 | 823 | finally: |
|
824 | 824 | repo.ui.quiet = oldquiet |
|
825 | 825 | fm.end() |
|
826 | 826 | |
|
827 | 827 | |
|
828 | 828 | def clearcaches(cl): |
|
829 | 829 | # behave somewhat consistently across internal API changes |
|
830 | 830 | if util.safehasattr(cl, b'clearcaches'): |
|
831 | 831 | cl.clearcaches() |
|
832 | 832 | elif util.safehasattr(cl, b'_nodecache'): |
|
833 | 833 | # <= hg-5.2 |
|
834 | 834 | from mercurial.node import nullid, nullrev |
|
835 | 835 | |
|
836 | 836 | cl._nodecache = {nullid: nullrev} |
|
837 | 837 | cl._nodepos = None |
|
838 | 838 | |
|
839 | 839 | |
|
840 | 840 | @command(b'perf::heads|perfheads', formatteropts) |
|
841 | 841 | def perfheads(ui, repo, **opts): |
|
842 | 842 | """benchmark the computation of a changelog heads""" |
|
843 | 843 | opts = _byteskwargs(opts) |
|
844 | 844 | timer, fm = gettimer(ui, opts) |
|
845 | 845 | cl = repo.changelog |
|
846 | 846 | |
|
847 | 847 | def s(): |
|
848 | 848 | clearcaches(cl) |
|
849 | 849 | |
|
850 | 850 | def d(): |
|
851 | 851 | len(cl.headrevs()) |
|
852 | 852 | |
|
853 | 853 | timer(d, setup=s) |
|
854 | 854 | fm.end() |
|
855 | 855 | |
|
856 | 856 | |
|
857 | 857 | @command( |
|
858 | 858 | b'perf::tags|perftags', |
|
859 | 859 | formatteropts |
|
860 | 860 | + [ |
|
861 | 861 | (b'', b'clear-revlogs', False, b'refresh changelog and manifest'), |
|
862 | 862 | ], |
|
863 | 863 | ) |
|
864 | 864 | def perftags(ui, repo, **opts): |
|
865 | 865 | opts = _byteskwargs(opts) |
|
866 | 866 | timer, fm = gettimer(ui, opts) |
|
867 | 867 | repocleartagscache = repocleartagscachefunc(repo) |
|
868 | 868 | clearrevlogs = opts[b'clear_revlogs'] |
|
869 | 869 | |
|
870 | 870 | def s(): |
|
871 | 871 | if clearrevlogs: |
|
872 | 872 | clearchangelog(repo) |
|
873 | 873 | clearfilecache(repo.unfiltered(), 'manifest') |
|
874 | 874 | repocleartagscache() |
|
875 | 875 | |
|
876 | 876 | def t(): |
|
877 | 877 | return len(repo.tags()) |
|
878 | 878 | |
|
879 | 879 | timer(t, setup=s) |
|
880 | 880 | fm.end() |
|
881 | 881 | |
|
882 | 882 | |
|
883 | 883 | @command(b'perf::ancestors|perfancestors', formatteropts) |
|
884 | 884 | def perfancestors(ui, repo, **opts): |
|
885 | 885 | opts = _byteskwargs(opts) |
|
886 | 886 | timer, fm = gettimer(ui, opts) |
|
887 | 887 | heads = repo.changelog.headrevs() |
|
888 | 888 | |
|
889 | 889 | def d(): |
|
890 | 890 | for a in repo.changelog.ancestors(heads): |
|
891 | 891 | pass |
|
892 | 892 | |
|
893 | 893 | timer(d) |
|
894 | 894 | fm.end() |
|
895 | 895 | |
|
896 | 896 | |
|
897 | 897 | @command(b'perf::ancestorset|perfancestorset', formatteropts) |
|
898 | 898 | def perfancestorset(ui, repo, revset, **opts): |
|
899 | 899 | opts = _byteskwargs(opts) |
|
900 | 900 | timer, fm = gettimer(ui, opts) |
|
901 | 901 | revs = repo.revs(revset) |
|
902 | 902 | heads = repo.changelog.headrevs() |
|
903 | 903 | |
|
904 | 904 | def d(): |
|
905 | 905 | s = repo.changelog.ancestors(heads) |
|
906 | 906 | for rev in revs: |
|
907 | 907 | rev in s |
|
908 | 908 | |
|
909 | 909 | timer(d) |
|
910 | 910 | fm.end() |
|
911 | 911 | |
|
912 | 912 | |
|
913 | 913 | @command(b'perf::discovery|perfdiscovery', formatteropts, b'PATH') |
|
914 | 914 | def perfdiscovery(ui, repo, path, **opts): |
|
915 | 915 | """benchmark discovery between local repo and the peer at given path""" |
|
916 | 916 | repos = [repo, None] |
|
917 | 917 | timer, fm = gettimer(ui, opts) |
|
918 | 918 | |
|
919 | 919 | try: |
|
920 | 920 | from mercurial.utils.urlutil import get_unique_pull_path |
|
921 | 921 | |
|
922 | 922 | path = get_unique_pull_path(b'perfdiscovery', repo, ui, path)[0] |
|
923 | 923 | except ImportError: |
|
924 | 924 | path = ui.expandpath(path) |
|
925 | 925 | |
|
926 | 926 | def s(): |
|
927 | 927 | repos[1] = hg.peer(ui, opts, path) |
|
928 | 928 | |
|
929 | 929 | def d(): |
|
930 | 930 | setdiscovery.findcommonheads(ui, *repos) |
|
931 | 931 | |
|
932 | 932 | timer(d, setup=s) |
|
933 | 933 | fm.end() |
|
934 | 934 | |
|
935 | 935 | |
|
936 | 936 | @command( |
|
937 | 937 | b'perf::bookmarks|perfbookmarks', |
|
938 | 938 | formatteropts |
|
939 | 939 | + [ |
|
940 | 940 | (b'', b'clear-revlogs', False, b'refresh changelog and manifest'), |
|
941 | 941 | ], |
|
942 | 942 | ) |
|
943 | 943 | def perfbookmarks(ui, repo, **opts): |
|
944 | 944 | """benchmark parsing bookmarks from disk to memory""" |
|
945 | 945 | opts = _byteskwargs(opts) |
|
946 | 946 | timer, fm = gettimer(ui, opts) |
|
947 | 947 | |
|
948 | 948 | clearrevlogs = opts[b'clear_revlogs'] |
|
949 | 949 | |
|
950 | 950 | def s(): |
|
951 | 951 | if clearrevlogs: |
|
952 | 952 | clearchangelog(repo) |
|
953 | 953 | clearfilecache(repo, b'_bookmarks') |
|
954 | 954 | |
|
955 | 955 | def d(): |
|
956 | 956 | repo._bookmarks |
|
957 | 957 | |
|
958 | 958 | timer(d, setup=s) |
|
959 | 959 | fm.end() |
|
960 | 960 | |
|
961 | 961 | |
|
962 | 962 | @command(b'perf::bundleread|perfbundleread', formatteropts, b'BUNDLE') |
|
963 | 963 | def perfbundleread(ui, repo, bundlepath, **opts): |
|
964 | 964 | """Benchmark reading of bundle files. |
|
965 | 965 | |
|
966 | 966 | This command is meant to isolate the I/O part of bundle reading as |
|
967 | 967 | much as possible. |
|
968 | 968 | """ |
|
969 | 969 | from mercurial import ( |
|
970 | 970 | bundle2, |
|
971 | 971 | exchange, |
|
972 | 972 | streamclone, |
|
973 | 973 | ) |
|
974 | 974 | |
|
975 | 975 | opts = _byteskwargs(opts) |
|
976 | 976 | |
|
977 | 977 | def makebench(fn): |
|
978 | 978 | def run(): |
|
979 | 979 | with open(bundlepath, b'rb') as fh: |
|
980 | 980 | bundle = exchange.readbundle(ui, fh, bundlepath) |
|
981 | 981 | fn(bundle) |
|
982 | 982 | |
|
983 | 983 | return run |
|
984 | 984 | |
|
985 | 985 | def makereadnbytes(size): |
|
986 | 986 | def run(): |
|
987 | 987 | with open(bundlepath, b'rb') as fh: |
|
988 | 988 | bundle = exchange.readbundle(ui, fh, bundlepath) |
|
989 | 989 | while bundle.read(size): |
|
990 | 990 | pass |
|
991 | 991 | |
|
992 | 992 | return run |
|
993 | 993 | |
|
994 | 994 | def makestdioread(size): |
|
995 | 995 | def run(): |
|
996 | 996 | with open(bundlepath, b'rb') as fh: |
|
997 | 997 | while fh.read(size): |
|
998 | 998 | pass |
|
999 | 999 | |
|
1000 | 1000 | return run |
|
1001 | 1001 | |
|
1002 | 1002 | # bundle1 |
|
1003 | 1003 | |
|
1004 | 1004 | def deltaiter(bundle): |
|
1005 | 1005 | for delta in bundle.deltaiter(): |
|
1006 | 1006 | pass |
|
1007 | 1007 | |
|
1008 | 1008 | def iterchunks(bundle): |
|
1009 | 1009 | for chunk in bundle.getchunks(): |
|
1010 | 1010 | pass |
|
1011 | 1011 | |
|
1012 | 1012 | # bundle2 |
|
1013 | 1013 | |
|
1014 | 1014 | def forwardchunks(bundle): |
|
1015 | 1015 | for chunk in bundle._forwardchunks(): |
|
1016 | 1016 | pass |
|
1017 | 1017 | |
|
1018 | 1018 | def iterparts(bundle): |
|
1019 | 1019 | for part in bundle.iterparts(): |
|
1020 | 1020 | pass |
|
1021 | 1021 | |
|
1022 | 1022 | def iterpartsseekable(bundle): |
|
1023 | 1023 | for part in bundle.iterparts(seekable=True): |
|
1024 | 1024 | pass |
|
1025 | 1025 | |
|
1026 | 1026 | def seek(bundle): |
|
1027 | 1027 | for part in bundle.iterparts(seekable=True): |
|
1028 | 1028 | part.seek(0, os.SEEK_END) |
|
1029 | 1029 | |
|
1030 | 1030 | def makepartreadnbytes(size): |
|
1031 | 1031 | def run(): |
|
1032 | 1032 | with open(bundlepath, b'rb') as fh: |
|
1033 | 1033 | bundle = exchange.readbundle(ui, fh, bundlepath) |
|
1034 | 1034 | for part in bundle.iterparts(): |
|
1035 | 1035 | while part.read(size): |
|
1036 | 1036 | pass |
|
1037 | 1037 | |
|
1038 | 1038 | return run |
|
1039 | 1039 | |
|
1040 | 1040 | benches = [ |
|
1041 | 1041 | (makestdioread(8192), b'read(8k)'), |
|
1042 | 1042 | (makestdioread(16384), b'read(16k)'), |
|
1043 | 1043 | (makestdioread(32768), b'read(32k)'), |
|
1044 | 1044 | (makestdioread(131072), b'read(128k)'), |
|
1045 | 1045 | ] |
|
1046 | 1046 | |
|
1047 | 1047 | with open(bundlepath, b'rb') as fh: |
|
1048 | 1048 | bundle = exchange.readbundle(ui, fh, bundlepath) |
|
1049 | 1049 | |
|
1050 | 1050 | if isinstance(bundle, changegroup.cg1unpacker): |
|
1051 | 1051 | benches.extend( |
|
1052 | 1052 | [ |
|
1053 | 1053 | (makebench(deltaiter), b'cg1 deltaiter()'), |
|
1054 | 1054 | (makebench(iterchunks), b'cg1 getchunks()'), |
|
1055 | 1055 | (makereadnbytes(8192), b'cg1 read(8k)'), |
|
1056 | 1056 | (makereadnbytes(16384), b'cg1 read(16k)'), |
|
1057 | 1057 | (makereadnbytes(32768), b'cg1 read(32k)'), |
|
1058 | 1058 | (makereadnbytes(131072), b'cg1 read(128k)'), |
|
1059 | 1059 | ] |
|
1060 | 1060 | ) |
|
1061 | 1061 | elif isinstance(bundle, bundle2.unbundle20): |
|
1062 | 1062 | benches.extend( |
|
1063 | 1063 | [ |
|
1064 | 1064 | (makebench(forwardchunks), b'bundle2 forwardchunks()'), |
|
1065 | 1065 | (makebench(iterparts), b'bundle2 iterparts()'), |
|
1066 | 1066 | ( |
|
1067 | 1067 | makebench(iterpartsseekable), |
|
1068 | 1068 | b'bundle2 iterparts() seekable', |
|
1069 | 1069 | ), |
|
1070 | 1070 | (makebench(seek), b'bundle2 part seek()'), |
|
1071 | 1071 | (makepartreadnbytes(8192), b'bundle2 part read(8k)'), |
|
1072 | 1072 | (makepartreadnbytes(16384), b'bundle2 part read(16k)'), |
|
1073 | 1073 | (makepartreadnbytes(32768), b'bundle2 part read(32k)'), |
|
1074 | 1074 | (makepartreadnbytes(131072), b'bundle2 part read(128k)'), |
|
1075 | 1075 | ] |
|
1076 | 1076 | ) |
|
1077 | 1077 | elif isinstance(bundle, streamclone.streamcloneapplier): |
|
1078 | 1078 | raise error.Abort(b'stream clone bundles not supported') |
|
1079 | 1079 | else: |
|
1080 | 1080 | raise error.Abort(b'unhandled bundle type: %s' % type(bundle)) |
|
1081 | 1081 | |
|
1082 | 1082 | for fn, title in benches: |
|
1083 | 1083 | timer, fm = gettimer(ui, opts) |
|
1084 | 1084 | timer(fn, title=title) |
|
1085 | 1085 | fm.end() |
|
1086 | 1086 | |
|
1087 | 1087 | |
|
1088 | 1088 | @command( |
|
1089 | 1089 | b'perf::changegroupchangelog|perfchangegroupchangelog', |
|
1090 | 1090 | formatteropts |
|
1091 | 1091 | + [ |
|
1092 | 1092 | (b'', b'cgversion', b'02', b'changegroup version'), |
|
1093 | 1093 | (b'r', b'rev', b'', b'revisions to add to changegroup'), |
|
1094 | 1094 | ], |
|
1095 | 1095 | ) |
|
1096 | 1096 | def perfchangegroupchangelog(ui, repo, cgversion=b'02', rev=None, **opts): |
|
1097 | 1097 | """Benchmark producing a changelog group for a changegroup. |
|
1098 | 1098 | |
|
1099 | 1099 | This measures the time spent processing the changelog during a |
|
1100 | 1100 | bundle operation. This occurs during `hg bundle` and on a server |
|
1101 | 1101 | processing a `getbundle` wire protocol request (handles clones |
|
1102 | 1102 | and pull requests). |
|
1103 | 1103 | |
|
1104 | 1104 | By default, all revisions are added to the changegroup. |
|
1105 | 1105 | """ |
|
1106 | 1106 | opts = _byteskwargs(opts) |
|
1107 | 1107 | cl = repo.changelog |
|
1108 | 1108 | nodes = [cl.lookup(r) for r in repo.revs(rev or b'all()')] |
|
1109 | 1109 | bundler = changegroup.getbundler(cgversion, repo) |
|
1110 | 1110 | |
|
1111 | 1111 | def d(): |
|
1112 | 1112 | state, chunks = bundler._generatechangelog(cl, nodes) |
|
1113 | 1113 | for chunk in chunks: |
|
1114 | 1114 | pass |
|
1115 | 1115 | |
|
1116 | 1116 | timer, fm = gettimer(ui, opts) |
|
1117 | 1117 | |
|
1118 | 1118 | # Terminal printing can interfere with timing. So disable it. |
|
1119 | 1119 | with ui.configoverride({(b'progress', b'disable'): True}): |
|
1120 | 1120 | timer(d) |
|
1121 | 1121 | |
|
1122 | 1122 | fm.end() |
|
1123 | 1123 | |
|
1124 | 1124 | |
|
1125 | 1125 | @command(b'perf::dirs|perfdirs', formatteropts) |
|
1126 | 1126 | def perfdirs(ui, repo, **opts): |
|
1127 | 1127 | opts = _byteskwargs(opts) |
|
1128 | 1128 | timer, fm = gettimer(ui, opts) |
|
1129 | 1129 | dirstate = repo.dirstate |
|
1130 | 1130 | b'a' in dirstate |
|
1131 | 1131 | |
|
1132 | 1132 | def d(): |
|
1133 | 1133 | dirstate.hasdir(b'a') |
|
1134 | 1134 | del dirstate._map._dirs |
|
1135 | 1135 | |
|
1136 | 1136 | timer(d) |
|
1137 | 1137 | fm.end() |
|
1138 | 1138 | |
|
1139 | 1139 | |
|
1140 | 1140 | @command( |
|
1141 | 1141 | b'perf::dirstate|perfdirstate', |
|
1142 | 1142 | [ |
|
1143 | 1143 | ( |
|
1144 | 1144 | b'', |
|
1145 | 1145 | b'iteration', |
|
1146 | 1146 | None, |
|
1147 | 1147 | b'benchmark a full iteration for the dirstate', |
|
1148 | 1148 | ), |
|
1149 | 1149 | ( |
|
1150 | 1150 | b'', |
|
1151 | 1151 | b'contains', |
|
1152 | 1152 | None, |
|
1153 | 1153 | b'benchmark a large amount of `nf in dirstate` calls', |
|
1154 | 1154 | ), |
|
1155 | 1155 | ] |
|
1156 | 1156 | + formatteropts, |
|
1157 | 1157 | ) |
|
1158 | 1158 | def perfdirstate(ui, repo, **opts): |
|
1159 | 1159 | """benchmap the time of various distate operations |
|
1160 | 1160 | |
|
1161 | 1161 | By default benchmark the time necessary to load a dirstate from scratch. |
|
1162 | 1162 | The dirstate is loaded to the point were a "contains" request can be |
|
1163 | 1163 | answered. |
|
1164 | 1164 | """ |
|
1165 | 1165 | opts = _byteskwargs(opts) |
|
1166 | 1166 | timer, fm = gettimer(ui, opts) |
|
1167 | 1167 | b"a" in repo.dirstate |
|
1168 | 1168 | |
|
1169 | 1169 | if opts[b'iteration'] and opts[b'contains']: |
|
1170 | 1170 | msg = b'only specify one of --iteration or --contains' |
|
1171 | 1171 | raise error.Abort(msg) |
|
1172 | 1172 | |
|
1173 | 1173 | if opts[b'iteration']: |
|
1174 | 1174 | setup = None |
|
1175 | 1175 | dirstate = repo.dirstate |
|
1176 | 1176 | |
|
1177 | 1177 | def d(): |
|
1178 | 1178 | for f in dirstate: |
|
1179 | 1179 | pass |
|
1180 | 1180 | |
|
1181 | 1181 | elif opts[b'contains']: |
|
1182 | 1182 | setup = None |
|
1183 | 1183 | dirstate = repo.dirstate |
|
1184 | 1184 | allfiles = list(dirstate) |
|
1185 | 1185 | # also add file path that will be "missing" from the dirstate |
|
1186 | 1186 | allfiles.extend([f[::-1] for f in allfiles]) |
|
1187 | 1187 | |
|
1188 | 1188 | def d(): |
|
1189 | 1189 | for f in allfiles: |
|
1190 | 1190 | f in dirstate |
|
1191 | 1191 | |
|
1192 | 1192 | else: |
|
1193 | 1193 | |
|
1194 | 1194 | def setup(): |
|
1195 | 1195 | repo.dirstate.invalidate() |
|
1196 | 1196 | |
|
1197 | 1197 | def d(): |
|
1198 | 1198 | b"a" in repo.dirstate |
|
1199 | 1199 | |
|
1200 | 1200 | timer(d, setup=setup) |
|
1201 | 1201 | fm.end() |
|
1202 | 1202 | |
|
1203 | 1203 | |
|
1204 | 1204 | @command(b'perf::dirstatedirs|perfdirstatedirs', formatteropts) |
|
1205 | 1205 | def perfdirstatedirs(ui, repo, **opts): |
|
1206 | 1206 | """benchmap a 'dirstate.hasdir' call from an empty `dirs` cache""" |
|
1207 | 1207 | opts = _byteskwargs(opts) |
|
1208 | 1208 | timer, fm = gettimer(ui, opts) |
|
1209 | 1209 | repo.dirstate.hasdir(b"a") |
|
1210 | 1210 | |
|
1211 | 1211 | def setup(): |
|
1212 | 1212 | del repo.dirstate._map._dirs |
|
1213 | 1213 | |
|
1214 | 1214 | def d(): |
|
1215 | 1215 | repo.dirstate.hasdir(b"a") |
|
1216 | 1216 | |
|
1217 | 1217 | timer(d, setup=setup) |
|
1218 | 1218 | fm.end() |
|
1219 | 1219 | |
|
1220 | 1220 | |
|
1221 | 1221 | @command(b'perf::dirstatefoldmap|perfdirstatefoldmap', formatteropts) |
|
1222 | 1222 | def perfdirstatefoldmap(ui, repo, **opts): |
|
1223 | 1223 | """benchmap a `dirstate._map.filefoldmap.get()` request |
|
1224 | 1224 | |
|
1225 | 1225 | The dirstate filefoldmap cache is dropped between every request. |
|
1226 | 1226 | """ |
|
1227 | 1227 | opts = _byteskwargs(opts) |
|
1228 | 1228 | timer, fm = gettimer(ui, opts) |
|
1229 | 1229 | dirstate = repo.dirstate |
|
1230 | 1230 | dirstate._map.filefoldmap.get(b'a') |
|
1231 | 1231 | |
|
1232 | 1232 | def setup(): |
|
1233 | 1233 | del dirstate._map.filefoldmap |
|
1234 | 1234 | |
|
1235 | 1235 | def d(): |
|
1236 | 1236 | dirstate._map.filefoldmap.get(b'a') |
|
1237 | 1237 | |
|
1238 | 1238 | timer(d, setup=setup) |
|
1239 | 1239 | fm.end() |
|
1240 | 1240 | |
|
1241 | 1241 | |
|
1242 | 1242 | @command(b'perf::dirfoldmap|perfdirfoldmap', formatteropts) |
|
1243 | 1243 | def perfdirfoldmap(ui, repo, **opts): |
|
1244 | 1244 | """benchmap a `dirstate._map.dirfoldmap.get()` request |
|
1245 | 1245 | |
|
1246 | 1246 | The dirstate dirfoldmap cache is dropped between every request. |
|
1247 | 1247 | """ |
|
1248 | 1248 | opts = _byteskwargs(opts) |
|
1249 | 1249 | timer, fm = gettimer(ui, opts) |
|
1250 | 1250 | dirstate = repo.dirstate |
|
1251 | 1251 | dirstate._map.dirfoldmap.get(b'a') |
|
1252 | 1252 | |
|
1253 | 1253 | def setup(): |
|
1254 | 1254 | del dirstate._map.dirfoldmap |
|
1255 | 1255 | del dirstate._map._dirs |
|
1256 | 1256 | |
|
1257 | 1257 | def d(): |
|
1258 | 1258 | dirstate._map.dirfoldmap.get(b'a') |
|
1259 | 1259 | |
|
1260 | 1260 | timer(d, setup=setup) |
|
1261 | 1261 | fm.end() |
|
1262 | 1262 | |
|
1263 | 1263 | |
|
1264 | 1264 | @command(b'perf::dirstatewrite|perfdirstatewrite', formatteropts) |
|
1265 | 1265 | def perfdirstatewrite(ui, repo, **opts): |
|
1266 | 1266 | """benchmap the time it take to write a dirstate on disk""" |
|
1267 | 1267 | opts = _byteskwargs(opts) |
|
1268 | 1268 | timer, fm = gettimer(ui, opts) |
|
1269 | 1269 | ds = repo.dirstate |
|
1270 | 1270 | b"a" in ds |
|
1271 | 1271 | |
|
1272 | 1272 | def setup(): |
|
1273 | 1273 | ds._dirty = True |
|
1274 | 1274 | |
|
1275 | 1275 | def d(): |
|
1276 | 1276 | ds.write(repo.currenttransaction()) |
|
1277 | 1277 | |
|
1278 | 1278 | timer(d, setup=setup) |
|
1279 | 1279 | fm.end() |
|
1280 | 1280 | |
|
1281 | 1281 | |
|
1282 | 1282 | def _getmergerevs(repo, opts): |
|
1283 | 1283 | """parse command argument to return rev involved in merge |
|
1284 | 1284 | |
|
1285 | 1285 | input: options dictionnary with `rev`, `from` and `bse` |
|
1286 | 1286 | output: (localctx, otherctx, basectx) |
|
1287 | 1287 | """ |
|
1288 | 1288 | if opts[b'from']: |
|
1289 | 1289 | fromrev = scmutil.revsingle(repo, opts[b'from']) |
|
1290 | 1290 | wctx = repo[fromrev] |
|
1291 | 1291 | else: |
|
1292 | 1292 | wctx = repo[None] |
|
1293 | 1293 | # we don't want working dir files to be stat'd in the benchmark, so |
|
1294 | 1294 | # prime that cache |
|
1295 | 1295 | wctx.dirty() |
|
1296 | 1296 | rctx = scmutil.revsingle(repo, opts[b'rev'], opts[b'rev']) |
|
1297 | 1297 | if opts[b'base']: |
|
1298 | 1298 | fromrev = scmutil.revsingle(repo, opts[b'base']) |
|
1299 | 1299 | ancestor = repo[fromrev] |
|
1300 | 1300 | else: |
|
1301 | 1301 | ancestor = wctx.ancestor(rctx) |
|
1302 | 1302 | return (wctx, rctx, ancestor) |
|
1303 | 1303 | |
|
1304 | 1304 | |
|
1305 | 1305 | @command( |
|
1306 | 1306 | b'perf::mergecalculate|perfmergecalculate', |
|
1307 | 1307 | [ |
|
1308 | 1308 | (b'r', b'rev', b'.', b'rev to merge against'), |
|
1309 | 1309 | (b'', b'from', b'', b'rev to merge from'), |
|
1310 | 1310 | (b'', b'base', b'', b'the revision to use as base'), |
|
1311 | 1311 | ] |
|
1312 | 1312 | + formatteropts, |
|
1313 | 1313 | ) |
|
1314 | 1314 | def perfmergecalculate(ui, repo, **opts): |
|
1315 | 1315 | opts = _byteskwargs(opts) |
|
1316 | 1316 | timer, fm = gettimer(ui, opts) |
|
1317 | 1317 | |
|
1318 | 1318 | wctx, rctx, ancestor = _getmergerevs(repo, opts) |
|
1319 | 1319 | |
|
1320 | 1320 | def d(): |
|
1321 | 1321 | # acceptremote is True because we don't want prompts in the middle of |
|
1322 | 1322 | # our benchmark |
|
1323 | 1323 | merge.calculateupdates( |
|
1324 | 1324 | repo, |
|
1325 | 1325 | wctx, |
|
1326 | 1326 | rctx, |
|
1327 | 1327 | [ancestor], |
|
1328 | 1328 | branchmerge=False, |
|
1329 | 1329 | force=False, |
|
1330 | 1330 | acceptremote=True, |
|
1331 | 1331 | followcopies=True, |
|
1332 | 1332 | ) |
|
1333 | 1333 | |
|
1334 | 1334 | timer(d) |
|
1335 | 1335 | fm.end() |
|
1336 | 1336 | |
|
1337 | 1337 | |
|
1338 | 1338 | @command( |
|
1339 | 1339 | b'perf::mergecopies|perfmergecopies', |
|
1340 | 1340 | [ |
|
1341 | 1341 | (b'r', b'rev', b'.', b'rev to merge against'), |
|
1342 | 1342 | (b'', b'from', b'', b'rev to merge from'), |
|
1343 | 1343 | (b'', b'base', b'', b'the revision to use as base'), |
|
1344 | 1344 | ] |
|
1345 | 1345 | + formatteropts, |
|
1346 | 1346 | ) |
|
1347 | 1347 | def perfmergecopies(ui, repo, **opts): |
|
1348 | 1348 | """measure runtime of `copies.mergecopies`""" |
|
1349 | 1349 | opts = _byteskwargs(opts) |
|
1350 | 1350 | timer, fm = gettimer(ui, opts) |
|
1351 | 1351 | wctx, rctx, ancestor = _getmergerevs(repo, opts) |
|
1352 | 1352 | |
|
1353 | 1353 | def d(): |
|
1354 | 1354 | # acceptremote is True because we don't want prompts in the middle of |
|
1355 | 1355 | # our benchmark |
|
1356 | 1356 | copies.mergecopies(repo, wctx, rctx, ancestor) |
|
1357 | 1357 | |
|
1358 | 1358 | timer(d) |
|
1359 | 1359 | fm.end() |
|
1360 | 1360 | |
|
1361 | 1361 | |
|
1362 | 1362 | @command(b'perf::pathcopies|perfpathcopies', [], b"REV REV") |
|
1363 | 1363 | def perfpathcopies(ui, repo, rev1, rev2, **opts): |
|
1364 | 1364 | """benchmark the copy tracing logic""" |
|
1365 | 1365 | opts = _byteskwargs(opts) |
|
1366 | 1366 | timer, fm = gettimer(ui, opts) |
|
1367 | 1367 | ctx1 = scmutil.revsingle(repo, rev1, rev1) |
|
1368 | 1368 | ctx2 = scmutil.revsingle(repo, rev2, rev2) |
|
1369 | 1369 | |
|
1370 | 1370 | def d(): |
|
1371 | 1371 | copies.pathcopies(ctx1, ctx2) |
|
1372 | 1372 | |
|
1373 | 1373 | timer(d) |
|
1374 | 1374 | fm.end() |
|
1375 | 1375 | |
|
1376 | 1376 | |
|
1377 | 1377 | @command( |
|
1378 | 1378 | b'perf::phases|perfphases', |
|
1379 | 1379 | [ |
|
1380 | 1380 | (b'', b'full', False, b'include file reading time too'), |
|
1381 | 1381 | ], |
|
1382 | 1382 | b"", |
|
1383 | 1383 | ) |
|
1384 | 1384 | def perfphases(ui, repo, **opts): |
|
1385 | 1385 | """benchmark phasesets computation""" |
|
1386 | 1386 | opts = _byteskwargs(opts) |
|
1387 | 1387 | timer, fm = gettimer(ui, opts) |
|
1388 | 1388 | _phases = repo._phasecache |
|
1389 | 1389 | full = opts.get(b'full') |
|
1390 | 1390 | |
|
1391 | 1391 | def d(): |
|
1392 | 1392 | phases = _phases |
|
1393 | 1393 | if full: |
|
1394 | 1394 | clearfilecache(repo, b'_phasecache') |
|
1395 | 1395 | phases = repo._phasecache |
|
1396 | 1396 | phases.invalidate() |
|
1397 | 1397 | phases.loadphaserevs(repo) |
|
1398 | 1398 | |
|
1399 | 1399 | timer(d) |
|
1400 | 1400 | fm.end() |
|
1401 | 1401 | |
|
1402 | 1402 | |
|
1403 | 1403 | @command(b'perf::phasesremote|perfphasesremote', [], b"[DEST]") |
|
1404 | 1404 | def perfphasesremote(ui, repo, dest=None, **opts): |
|
1405 | 1405 | """benchmark time needed to analyse phases of the remote server""" |
|
1406 | 1406 | from mercurial.node import bin |
|
1407 | 1407 | from mercurial import ( |
|
1408 | 1408 | exchange, |
|
1409 | 1409 | hg, |
|
1410 | 1410 | phases, |
|
1411 | 1411 | ) |
|
1412 | 1412 | |
|
1413 | 1413 | opts = _byteskwargs(opts) |
|
1414 | 1414 | timer, fm = gettimer(ui, opts) |
|
1415 | 1415 | |
|
1416 | 1416 | path = ui.getpath(dest, default=(b'default-push', b'default')) |
|
1417 | 1417 | if not path: |
|
1418 | 1418 | raise error.Abort( |
|
1419 | 1419 | b'default repository not configured!', |
|
1420 | 1420 | hint=b"see 'hg help config.paths'", |
|
1421 | 1421 | ) |
|
1422 | 1422 | dest = path.pushloc or path.loc |
|
1423 | 1423 | ui.statusnoi18n(b'analysing phase of %s\n' % util.hidepassword(dest)) |
|
1424 | 1424 | other = hg.peer(repo, opts, dest) |
|
1425 | 1425 | |
|
1426 | 1426 | # easier to perform discovery through the operation |
|
1427 | 1427 | op = exchange.pushoperation(repo, other) |
|
1428 | 1428 | exchange._pushdiscoverychangeset(op) |
|
1429 | 1429 | |
|
1430 | 1430 | remotesubset = op.fallbackheads |
|
1431 | 1431 | |
|
1432 | 1432 | with other.commandexecutor() as e: |
|
1433 | 1433 | remotephases = e.callcommand( |
|
1434 | 1434 | b'listkeys', {b'namespace': b'phases'} |
|
1435 | 1435 | ).result() |
|
1436 | 1436 | del other |
|
1437 | 1437 | publishing = remotephases.get(b'publishing', False) |
|
1438 | 1438 | if publishing: |
|
1439 | 1439 | ui.statusnoi18n(b'publishing: yes\n') |
|
1440 | 1440 | else: |
|
1441 | 1441 | ui.statusnoi18n(b'publishing: no\n') |
|
1442 | 1442 | |
|
1443 | 1443 | has_node = getattr(repo.changelog.index, 'has_node', None) |
|
1444 | 1444 | if has_node is None: |
|
1445 | 1445 | has_node = repo.changelog.nodemap.__contains__ |
|
1446 | 1446 | nonpublishroots = 0 |
|
1447 | 1447 | for nhex, phase in remotephases.iteritems(): |
|
1448 | 1448 | if nhex == b'publishing': # ignore data related to publish option |
|
1449 | 1449 | continue |
|
1450 | 1450 | node = bin(nhex) |
|
1451 | 1451 | if has_node(node) and int(phase): |
|
1452 | 1452 | nonpublishroots += 1 |
|
1453 | 1453 | ui.statusnoi18n(b'number of roots: %d\n' % len(remotephases)) |
|
1454 | 1454 | ui.statusnoi18n(b'number of known non public roots: %d\n' % nonpublishroots) |
|
1455 | 1455 | |
|
1456 | 1456 | def d(): |
|
1457 | 1457 | phases.remotephasessummary(repo, remotesubset, remotephases) |
|
1458 | 1458 | |
|
1459 | 1459 | timer(d) |
|
1460 | 1460 | fm.end() |
|
1461 | 1461 | |
|
1462 | 1462 | |
|
1463 | 1463 | @command( |
|
1464 | 1464 | b'perf::manifest|perfmanifest', |
|
1465 | 1465 | [ |
|
1466 | 1466 | (b'm', b'manifest-rev', False, b'Look up a manifest node revision'), |
|
1467 | 1467 | (b'', b'clear-disk', False, b'clear on-disk caches too'), |
|
1468 | 1468 | ] |
|
1469 | 1469 | + formatteropts, |
|
1470 | 1470 | b'REV|NODE', |
|
1471 | 1471 | ) |
|
1472 | 1472 | def perfmanifest(ui, repo, rev, manifest_rev=False, clear_disk=False, **opts): |
|
1473 | 1473 | """benchmark the time to read a manifest from disk and return a usable |
|
1474 | 1474 | dict-like object |
|
1475 | 1475 | |
|
1476 | 1476 | Manifest caches are cleared before retrieval.""" |
|
1477 | 1477 | opts = _byteskwargs(opts) |
|
1478 | 1478 | timer, fm = gettimer(ui, opts) |
|
1479 | 1479 | if not manifest_rev: |
|
1480 | 1480 | ctx = scmutil.revsingle(repo, rev, rev) |
|
1481 | 1481 | t = ctx.manifestnode() |
|
1482 | 1482 | else: |
|
1483 | 1483 | from mercurial.node import bin |
|
1484 | 1484 | |
|
1485 | 1485 | if len(rev) == 40: |
|
1486 | 1486 | t = bin(rev) |
|
1487 | 1487 | else: |
|
1488 | 1488 | try: |
|
1489 | 1489 | rev = int(rev) |
|
1490 | 1490 | |
|
1491 | 1491 | if util.safehasattr(repo.manifestlog, b'getstorage'): |
|
1492 | 1492 | t = repo.manifestlog.getstorage(b'').node(rev) |
|
1493 | 1493 | else: |
|
1494 | 1494 | t = repo.manifestlog._revlog.lookup(rev) |
|
1495 | 1495 | except ValueError: |
|
1496 | 1496 | raise error.Abort( |
|
1497 | 1497 | b'manifest revision must be integer or full node' |
|
1498 | 1498 | ) |
|
1499 | 1499 | |
|
1500 | 1500 | def d(): |
|
1501 | 1501 | repo.manifestlog.clearcaches(clear_persisted_data=clear_disk) |
|
1502 | 1502 | repo.manifestlog[t].read() |
|
1503 | 1503 | |
|
1504 | 1504 | timer(d) |
|
1505 | 1505 | fm.end() |
|
1506 | 1506 | |
|
1507 | 1507 | |
|
1508 | 1508 | @command(b'perf::changeset|perfchangeset', formatteropts) |
|
1509 | 1509 | def perfchangeset(ui, repo, rev, **opts): |
|
1510 | 1510 | opts = _byteskwargs(opts) |
|
1511 | 1511 | timer, fm = gettimer(ui, opts) |
|
1512 | 1512 | n = scmutil.revsingle(repo, rev).node() |
|
1513 | 1513 | |
|
1514 | 1514 | def d(): |
|
1515 | 1515 | repo.changelog.read(n) |
|
1516 | 1516 | # repo.changelog._cache = None |
|
1517 | 1517 | |
|
1518 | 1518 | timer(d) |
|
1519 | 1519 | fm.end() |
|
1520 | 1520 | |
|
1521 | 1521 | |
|
1522 | 1522 | @command(b'perf::ignore|perfignore', formatteropts) |
|
1523 | 1523 | def perfignore(ui, repo, **opts): |
|
1524 | 1524 | """benchmark operation related to computing ignore""" |
|
1525 | 1525 | opts = _byteskwargs(opts) |
|
1526 | 1526 | timer, fm = gettimer(ui, opts) |
|
1527 | 1527 | dirstate = repo.dirstate |
|
1528 | 1528 | |
|
1529 | 1529 | def setupone(): |
|
1530 | 1530 | dirstate.invalidate() |
|
1531 | 1531 | clearfilecache(dirstate, b'_ignore') |
|
1532 | 1532 | |
|
1533 | 1533 | def runone(): |
|
1534 | 1534 | dirstate._ignore |
|
1535 | 1535 | |
|
1536 | 1536 | timer(runone, setup=setupone, title=b"load") |
|
1537 | 1537 | fm.end() |
|
1538 | 1538 | |
|
1539 | 1539 | |
|
1540 | 1540 | @command( |
|
1541 | 1541 | b'perf::index|perfindex', |
|
1542 | 1542 | [ |
|
1543 | 1543 | (b'', b'rev', [], b'revision to be looked up (default tip)'), |
|
1544 | 1544 | (b'', b'no-lookup', None, b'do not revision lookup post creation'), |
|
1545 | 1545 | ] |
|
1546 | 1546 | + formatteropts, |
|
1547 | 1547 | ) |
|
1548 | 1548 | def perfindex(ui, repo, **opts): |
|
1549 | 1549 | """benchmark index creation time followed by a lookup |
|
1550 | 1550 | |
|
1551 | 1551 | The default is to look `tip` up. Depending on the index implementation, |
|
1552 | 1552 | the revision looked up can matters. For example, an implementation |
|
1553 | 1553 | scanning the index will have a faster lookup time for `--rev tip` than for |
|
1554 | 1554 | `--rev 0`. The number of looked up revisions and their order can also |
|
1555 | 1555 | matters. |
|
1556 | 1556 | |
|
1557 | 1557 | Example of useful set to test: |
|
1558 | 1558 | |
|
1559 | 1559 | * tip |
|
1560 | 1560 | * 0 |
|
1561 | 1561 | * -10: |
|
1562 | 1562 | * :10 |
|
1563 | 1563 | * -10: + :10 |
|
1564 | 1564 | * :10: + -10: |
|
1565 | 1565 | * -10000: |
|
1566 | 1566 | * -10000: + 0 |
|
1567 | 1567 | |
|
1568 | 1568 | It is not currently possible to check for lookup of a missing node. For |
|
1569 | 1569 | deeper lookup benchmarking, checkout the `perfnodemap` command.""" |
|
1570 | 1570 | import mercurial.revlog |
|
1571 | 1571 | |
|
1572 | 1572 | opts = _byteskwargs(opts) |
|
1573 | 1573 | timer, fm = gettimer(ui, opts) |
|
1574 | 1574 | mercurial.revlog._prereadsize = 2 ** 24 # disable lazy parser in old hg |
|
1575 | 1575 | if opts[b'no_lookup']: |
|
1576 | 1576 | if opts['rev']: |
|
1577 | 1577 | raise error.Abort('--no-lookup and --rev are mutually exclusive') |
|
1578 | 1578 | nodes = [] |
|
1579 | 1579 | elif not opts[b'rev']: |
|
1580 | 1580 | nodes = [repo[b"tip"].node()] |
|
1581 | 1581 | else: |
|
1582 | 1582 | revs = scmutil.revrange(repo, opts[b'rev']) |
|
1583 | 1583 | cl = repo.changelog |
|
1584 | 1584 | nodes = [cl.node(r) for r in revs] |
|
1585 | 1585 | |
|
1586 | 1586 | unfi = repo.unfiltered() |
|
1587 | 1587 | # find the filecache func directly |
|
1588 | 1588 | # This avoid polluting the benchmark with the filecache logic |
|
1589 | 1589 | makecl = unfi.__class__.changelog.func |
|
1590 | 1590 | |
|
1591 | 1591 | def setup(): |
|
1592 | 1592 | # probably not necessary, but for good measure |
|
1593 | 1593 | clearchangelog(unfi) |
|
1594 | 1594 | |
|
1595 | 1595 | def d(): |
|
1596 | 1596 | cl = makecl(unfi) |
|
1597 | 1597 | for n in nodes: |
|
1598 | 1598 | cl.rev(n) |
|
1599 | 1599 | |
|
1600 | 1600 | timer(d, setup=setup) |
|
1601 | 1601 | fm.end() |
|
1602 | 1602 | |
|
1603 | 1603 | |
|
1604 | 1604 | @command( |
|
1605 | 1605 | b'perf::nodemap|perfnodemap', |
|
1606 | 1606 | [ |
|
1607 | 1607 | (b'', b'rev', [], b'revision to be looked up (default tip)'), |
|
1608 | 1608 | (b'', b'clear-caches', True, b'clear revlog cache between calls'), |
|
1609 | 1609 | ] |
|
1610 | 1610 | + formatteropts, |
|
1611 | 1611 | ) |
|
1612 | 1612 | def perfnodemap(ui, repo, **opts): |
|
1613 | 1613 | """benchmark the time necessary to look up revision from a cold nodemap |
|
1614 | 1614 | |
|
1615 | 1615 | Depending on the implementation, the amount and order of revision we look |
|
1616 | 1616 | up can varies. Example of useful set to test: |
|
1617 | 1617 | * tip |
|
1618 | 1618 | * 0 |
|
1619 | 1619 | * -10: |
|
1620 | 1620 | * :10 |
|
1621 | 1621 | * -10: + :10 |
|
1622 | 1622 | * :10: + -10: |
|
1623 | 1623 | * -10000: |
|
1624 | 1624 | * -10000: + 0 |
|
1625 | 1625 | |
|
1626 | 1626 | The command currently focus on valid binary lookup. Benchmarking for |
|
1627 | 1627 | hexlookup, prefix lookup and missing lookup would also be valuable. |
|
1628 | 1628 | """ |
|
1629 | 1629 | import mercurial.revlog |
|
1630 | 1630 | |
|
1631 | 1631 | opts = _byteskwargs(opts) |
|
1632 | 1632 | timer, fm = gettimer(ui, opts) |
|
1633 | 1633 | mercurial.revlog._prereadsize = 2 ** 24 # disable lazy parser in old hg |
|
1634 | 1634 | |
|
1635 | 1635 | unfi = repo.unfiltered() |
|
1636 | 1636 | clearcaches = opts[b'clear_caches'] |
|
1637 | 1637 | # find the filecache func directly |
|
1638 | 1638 | # This avoid polluting the benchmark with the filecache logic |
|
1639 | 1639 | makecl = unfi.__class__.changelog.func |
|
1640 | 1640 | if not opts[b'rev']: |
|
1641 | 1641 | raise error.Abort(b'use --rev to specify revisions to look up') |
|
1642 | 1642 | revs = scmutil.revrange(repo, opts[b'rev']) |
|
1643 | 1643 | cl = repo.changelog |
|
1644 | 1644 | nodes = [cl.node(r) for r in revs] |
|
1645 | 1645 | |
|
1646 | 1646 | # use a list to pass reference to a nodemap from one closure to the next |
|
1647 | 1647 | nodeget = [None] |
|
1648 | 1648 | |
|
1649 | 1649 | def setnodeget(): |
|
1650 | 1650 | # probably not necessary, but for good measure |
|
1651 | 1651 | clearchangelog(unfi) |
|
1652 | 1652 | cl = makecl(unfi) |
|
1653 | 1653 | if util.safehasattr(cl.index, 'get_rev'): |
|
1654 | 1654 | nodeget[0] = cl.index.get_rev |
|
1655 | 1655 | else: |
|
1656 | 1656 | nodeget[0] = cl.nodemap.get |
|
1657 | 1657 | |
|
1658 | 1658 | def d(): |
|
1659 | 1659 | get = nodeget[0] |
|
1660 | 1660 | for n in nodes: |
|
1661 | 1661 | get(n) |
|
1662 | 1662 | |
|
1663 | 1663 | setup = None |
|
1664 | 1664 | if clearcaches: |
|
1665 | 1665 | |
|
1666 | 1666 | def setup(): |
|
1667 | 1667 | setnodeget() |
|
1668 | 1668 | |
|
1669 | 1669 | else: |
|
1670 | 1670 | setnodeget() |
|
1671 | 1671 | d() # prewarm the data structure |
|
1672 | 1672 | timer(d, setup=setup) |
|
1673 | 1673 | fm.end() |
|
1674 | 1674 | |
|
1675 | 1675 | |
|
1676 | 1676 | @command(b'perf::startup|perfstartup', formatteropts) |
|
1677 | 1677 | def perfstartup(ui, repo, **opts): |
|
1678 | 1678 | opts = _byteskwargs(opts) |
|
1679 | 1679 | timer, fm = gettimer(ui, opts) |
|
1680 | 1680 | |
|
1681 | 1681 | def d(): |
|
1682 | 1682 | if os.name != 'nt': |
|
1683 | 1683 | os.system( |
|
1684 | 1684 | b"HGRCPATH= %s version -q > /dev/null" % fsencode(sys.argv[0]) |
|
1685 | 1685 | ) |
|
1686 | 1686 | else: |
|
1687 | 1687 | os.environ['HGRCPATH'] = r' ' |
|
1688 | 1688 | os.system("%s version -q > NUL" % sys.argv[0]) |
|
1689 | 1689 | |
|
1690 | 1690 | timer(d) |
|
1691 | 1691 | fm.end() |
|
1692 | 1692 | |
|
1693 | 1693 | |
|
1694 | 1694 | @command(b'perf::parents|perfparents', formatteropts) |
|
1695 | 1695 | def perfparents(ui, repo, **opts): |
|
1696 | 1696 | """benchmark the time necessary to fetch one changeset's parents. |
|
1697 | 1697 | |
|
1698 | 1698 | The fetch is done using the `node identifier`, traversing all object layers |
|
1699 | 1699 | from the repository object. The first N revisions will be used for this |
|
1700 | 1700 | benchmark. N is controlled by the ``perf.parentscount`` config option |
|
1701 | 1701 | (default: 1000). |
|
1702 | 1702 | """ |
|
1703 | 1703 | opts = _byteskwargs(opts) |
|
1704 | 1704 | timer, fm = gettimer(ui, opts) |
|
1705 | 1705 | # control the number of commits perfparents iterates over |
|
1706 | 1706 | # experimental config: perf.parentscount |
|
1707 | 1707 | count = getint(ui, b"perf", b"parentscount", 1000) |
|
1708 | 1708 | if len(repo.changelog) < count: |
|
1709 | 1709 | raise error.Abort(b"repo needs %d commits for this test" % count) |
|
1710 | 1710 | repo = repo.unfiltered() |
|
1711 | 1711 | nl = [repo.changelog.node(i) for i in _xrange(count)] |
|
1712 | 1712 | |
|
1713 | 1713 | def d(): |
|
1714 | 1714 | for n in nl: |
|
1715 | 1715 | repo.changelog.parents(n) |
|
1716 | 1716 | |
|
1717 | 1717 | timer(d) |
|
1718 | 1718 | fm.end() |
|
1719 | 1719 | |
|
1720 | 1720 | |
|
1721 | 1721 | @command(b'perf::ctxfiles|perfctxfiles', formatteropts) |
|
1722 | 1722 | def perfctxfiles(ui, repo, x, **opts): |
|
1723 | 1723 | opts = _byteskwargs(opts) |
|
1724 | 1724 | x = int(x) |
|
1725 | 1725 | timer, fm = gettimer(ui, opts) |
|
1726 | 1726 | |
|
1727 | 1727 | def d(): |
|
1728 | 1728 | len(repo[x].files()) |
|
1729 | 1729 | |
|
1730 | 1730 | timer(d) |
|
1731 | 1731 | fm.end() |
|
1732 | 1732 | |
|
1733 | 1733 | |
|
1734 | 1734 | @command(b'perf::rawfiles|perfrawfiles', formatteropts) |
|
1735 | 1735 | def perfrawfiles(ui, repo, x, **opts): |
|
1736 | 1736 | opts = _byteskwargs(opts) |
|
1737 | 1737 | x = int(x) |
|
1738 | 1738 | timer, fm = gettimer(ui, opts) |
|
1739 | 1739 | cl = repo.changelog |
|
1740 | 1740 | |
|
1741 | 1741 | def d(): |
|
1742 | 1742 | len(cl.read(x)[3]) |
|
1743 | 1743 | |
|
1744 | 1744 | timer(d) |
|
1745 | 1745 | fm.end() |
|
1746 | 1746 | |
|
1747 | 1747 | |
|
1748 | 1748 | @command(b'perf::lookup|perflookup', formatteropts) |
|
1749 | 1749 | def perflookup(ui, repo, rev, **opts): |
|
1750 | 1750 | opts = _byteskwargs(opts) |
|
1751 | 1751 | timer, fm = gettimer(ui, opts) |
|
1752 | 1752 | timer(lambda: len(repo.lookup(rev))) |
|
1753 | 1753 | fm.end() |
|
1754 | 1754 | |
|
1755 | 1755 | |
|
1756 | 1756 | @command( |
|
1757 | 1757 | b'perf::linelogedits|perflinelogedits', |
|
1758 | 1758 | [ |
|
1759 | 1759 | (b'n', b'edits', 10000, b'number of edits'), |
|
1760 | 1760 | (b'', b'max-hunk-lines', 10, b'max lines in a hunk'), |
|
1761 | 1761 | ], |
|
1762 | 1762 | norepo=True, |
|
1763 | 1763 | ) |
|
1764 | 1764 | def perflinelogedits(ui, **opts): |
|
1765 | 1765 | from mercurial import linelog |
|
1766 | 1766 | |
|
1767 | 1767 | opts = _byteskwargs(opts) |
|
1768 | 1768 | |
|
1769 | 1769 | edits = opts[b'edits'] |
|
1770 | 1770 | maxhunklines = opts[b'max_hunk_lines'] |
|
1771 | 1771 | |
|
1772 | 1772 | maxb1 = 100000 |
|
1773 | 1773 | random.seed(0) |
|
1774 | 1774 | randint = random.randint |
|
1775 | 1775 | currentlines = 0 |
|
1776 | 1776 | arglist = [] |
|
1777 | 1777 | for rev in _xrange(edits): |
|
1778 | 1778 | a1 = randint(0, currentlines) |
|
1779 | 1779 | a2 = randint(a1, min(currentlines, a1 + maxhunklines)) |
|
1780 | 1780 | b1 = randint(0, maxb1) |
|
1781 | 1781 | b2 = randint(b1, b1 + maxhunklines) |
|
1782 | 1782 | currentlines += (b2 - b1) - (a2 - a1) |
|
1783 | 1783 | arglist.append((rev, a1, a2, b1, b2)) |
|
1784 | 1784 | |
|
1785 | 1785 | def d(): |
|
1786 | 1786 | ll = linelog.linelog() |
|
1787 | 1787 | for args in arglist: |
|
1788 | 1788 | ll.replacelines(*args) |
|
1789 | 1789 | |
|
1790 | 1790 | timer, fm = gettimer(ui, opts) |
|
1791 | 1791 | timer(d) |
|
1792 | 1792 | fm.end() |
|
1793 | 1793 | |
|
1794 | 1794 | |
|
1795 | 1795 | @command(b'perf::revrange|perfrevrange', formatteropts) |
|
1796 | 1796 | def perfrevrange(ui, repo, *specs, **opts): |
|
1797 | 1797 | opts = _byteskwargs(opts) |
|
1798 | 1798 | timer, fm = gettimer(ui, opts) |
|
1799 | 1799 | revrange = scmutil.revrange |
|
1800 | 1800 | timer(lambda: len(revrange(repo, specs))) |
|
1801 | 1801 | fm.end() |
|
1802 | 1802 | |
|
1803 | 1803 | |
|
1804 | 1804 | @command(b'perf::nodelookup|perfnodelookup', formatteropts) |
|
1805 | 1805 | def perfnodelookup(ui, repo, rev, **opts): |
|
1806 | 1806 | opts = _byteskwargs(opts) |
|
1807 | 1807 | timer, fm = gettimer(ui, opts) |
|
1808 | 1808 | import mercurial.revlog |
|
1809 | 1809 | |
|
1810 | 1810 | mercurial.revlog._prereadsize = 2 ** 24 # disable lazy parser in old hg |
|
1811 | 1811 | n = scmutil.revsingle(repo, rev).node() |
|
1812 | 1812 | cl = mercurial.revlog.revlog(getsvfs(repo), b"00changelog.i") |
|
1813 | 1813 | |
|
1814 | 1814 | def d(): |
|
1815 | 1815 | cl.rev(n) |
|
1816 | 1816 | clearcaches(cl) |
|
1817 | 1817 | |
|
1818 | 1818 | timer(d) |
|
1819 | 1819 | fm.end() |
|
1820 | 1820 | |
|
1821 | 1821 | |
|
1822 | 1822 | @command( |
|
1823 | 1823 | b'perf::log|perflog', |
|
1824 | 1824 | [(b'', b'rename', False, b'ask log to follow renames')] + formatteropts, |
|
1825 | 1825 | ) |
|
1826 | 1826 | def perflog(ui, repo, rev=None, **opts): |
|
1827 | 1827 | opts = _byteskwargs(opts) |
|
1828 | 1828 | if rev is None: |
|
1829 | 1829 | rev = [] |
|
1830 | 1830 | timer, fm = gettimer(ui, opts) |
|
1831 | 1831 | ui.pushbuffer() |
|
1832 | 1832 | timer( |
|
1833 | 1833 | lambda: commands.log( |
|
1834 | 1834 | ui, repo, rev=rev, date=b'', user=b'', copies=opts.get(b'rename') |
|
1835 | 1835 | ) |
|
1836 | 1836 | ) |
|
1837 | 1837 | ui.popbuffer() |
|
1838 | 1838 | fm.end() |
|
1839 | 1839 | |
|
1840 | 1840 | |
|
1841 | 1841 | @command(b'perf::moonwalk|perfmoonwalk', formatteropts) |
|
1842 | 1842 | def perfmoonwalk(ui, repo, **opts): |
|
1843 | 1843 | """benchmark walking the changelog backwards |
|
1844 | 1844 | |
|
1845 | 1845 | This also loads the changelog data for each revision in the changelog. |
|
1846 | 1846 | """ |
|
1847 | 1847 | opts = _byteskwargs(opts) |
|
1848 | 1848 | timer, fm = gettimer(ui, opts) |
|
1849 | 1849 | |
|
1850 | 1850 | def moonwalk(): |
|
1851 | 1851 | for i in repo.changelog.revs(start=(len(repo) - 1), stop=-1): |
|
1852 | 1852 | ctx = repo[i] |
|
1853 | 1853 | ctx.branch() # read changelog data (in addition to the index) |
|
1854 | 1854 | |
|
1855 | 1855 | timer(moonwalk) |
|
1856 | 1856 | fm.end() |
|
1857 | 1857 | |
|
1858 | 1858 | |
|
1859 | 1859 | @command( |
|
1860 | 1860 | b'perf::templating|perftemplating', |
|
1861 | 1861 | [ |
|
1862 | 1862 | (b'r', b'rev', [], b'revisions to run the template on'), |
|
1863 | 1863 | ] |
|
1864 | 1864 | + formatteropts, |
|
1865 | 1865 | ) |
|
1866 | 1866 | def perftemplating(ui, repo, testedtemplate=None, **opts): |
|
1867 | 1867 | """test the rendering time of a given template""" |
|
1868 | 1868 | if makelogtemplater is None: |
|
1869 | 1869 | raise error.Abort( |
|
1870 | 1870 | b"perftemplating not available with this Mercurial", |
|
1871 | 1871 | hint=b"use 4.3 or later", |
|
1872 | 1872 | ) |
|
1873 | 1873 | |
|
1874 | 1874 | opts = _byteskwargs(opts) |
|
1875 | 1875 | |
|
1876 | 1876 | nullui = ui.copy() |
|
1877 | 1877 | nullui.fout = open(os.devnull, 'wb') |
|
1878 | 1878 | nullui.disablepager() |
|
1879 | 1879 | revs = opts.get(b'rev') |
|
1880 | 1880 | if not revs: |
|
1881 | 1881 | revs = [b'all()'] |
|
1882 | 1882 | revs = list(scmutil.revrange(repo, revs)) |
|
1883 | 1883 | |
|
1884 | 1884 | defaulttemplate = ( |
|
1885 | 1885 | b'{date|shortdate} [{rev}:{node|short}]' |
|
1886 | 1886 | b' {author|person}: {desc|firstline}\n' |
|
1887 | 1887 | ) |
|
1888 | 1888 | if testedtemplate is None: |
|
1889 | 1889 | testedtemplate = defaulttemplate |
|
1890 | 1890 | displayer = makelogtemplater(nullui, repo, testedtemplate) |
|
1891 | 1891 | |
|
1892 | 1892 | def format(): |
|
1893 | 1893 | for r in revs: |
|
1894 | 1894 | ctx = repo[r] |
|
1895 | 1895 | displayer.show(ctx) |
|
1896 | 1896 | displayer.flush(ctx) |
|
1897 | 1897 | |
|
1898 | 1898 | timer, fm = gettimer(ui, opts) |
|
1899 | 1899 | timer(format) |
|
1900 | 1900 | fm.end() |
|
1901 | 1901 | |
|
1902 | 1902 | |
|
1903 | 1903 | def _displaystats(ui, opts, entries, data): |
|
1904 | 1904 | # use a second formatter because the data are quite different, not sure |
|
1905 | 1905 | # how it flies with the templater. |
|
1906 | 1906 | fm = ui.formatter(b'perf-stats', opts) |
|
1907 | 1907 | for key, title in entries: |
|
1908 | 1908 | values = data[key] |
|
1909 | 1909 | nbvalues = len(data) |
|
1910 | 1910 | values.sort() |
|
1911 | 1911 | stats = { |
|
1912 | 1912 | 'key': key, |
|
1913 | 1913 | 'title': title, |
|
1914 | 1914 | 'nbitems': len(values), |
|
1915 | 1915 | 'min': values[0][0], |
|
1916 | 1916 | '10%': values[(nbvalues * 10) // 100][0], |
|
1917 | 1917 | '25%': values[(nbvalues * 25) // 100][0], |
|
1918 | 1918 | '50%': values[(nbvalues * 50) // 100][0], |
|
1919 | 1919 | '75%': values[(nbvalues * 75) // 100][0], |
|
1920 | 1920 | '80%': values[(nbvalues * 80) // 100][0], |
|
1921 | 1921 | '85%': values[(nbvalues * 85) // 100][0], |
|
1922 | 1922 | '90%': values[(nbvalues * 90) // 100][0], |
|
1923 | 1923 | '95%': values[(nbvalues * 95) // 100][0], |
|
1924 | 1924 | '99%': values[(nbvalues * 99) // 100][0], |
|
1925 | 1925 | 'max': values[-1][0], |
|
1926 | 1926 | } |
|
1927 | 1927 | fm.startitem() |
|
1928 | 1928 | fm.data(**stats) |
|
1929 | 1929 | # make node pretty for the human output |
|
1930 | 1930 | fm.plain('### %s (%d items)\n' % (title, len(values))) |
|
1931 | 1931 | lines = [ |
|
1932 | 1932 | 'min', |
|
1933 | 1933 | '10%', |
|
1934 | 1934 | '25%', |
|
1935 | 1935 | '50%', |
|
1936 | 1936 | '75%', |
|
1937 | 1937 | '80%', |
|
1938 | 1938 | '85%', |
|
1939 | 1939 | '90%', |
|
1940 | 1940 | '95%', |
|
1941 | 1941 | '99%', |
|
1942 | 1942 | 'max', |
|
1943 | 1943 | ] |
|
1944 | 1944 | for l in lines: |
|
1945 | 1945 | fm.plain('%s: %s\n' % (l, stats[l])) |
|
1946 | 1946 | fm.end() |
|
1947 | 1947 | |
|
1948 | 1948 | |
|
1949 | 1949 | @command( |
|
1950 | 1950 | b'perf::helper-mergecopies|perfhelper-mergecopies', |
|
1951 | 1951 | formatteropts |
|
1952 | 1952 | + [ |
|
1953 | 1953 | (b'r', b'revs', [], b'restrict search to these revisions'), |
|
1954 | 1954 | (b'', b'timing', False, b'provides extra data (costly)'), |
|
1955 | 1955 | (b'', b'stats', False, b'provides statistic about the measured data'), |
|
1956 | 1956 | ], |
|
1957 | 1957 | ) |
|
1958 | 1958 | def perfhelpermergecopies(ui, repo, revs=[], **opts): |
|
1959 | 1959 | """find statistics about potential parameters for `perfmergecopies` |
|
1960 | 1960 | |
|
1961 | 1961 | This command find (base, p1, p2) triplet relevant for copytracing |
|
1962 | 1962 | benchmarking in the context of a merge. It reports values for some of the |
|
1963 | 1963 | parameters that impact merge copy tracing time during merge. |
|
1964 | 1964 | |
|
1965 | 1965 | If `--timing` is set, rename detection is run and the associated timing |
|
1966 | 1966 | will be reported. The extra details come at the cost of slower command |
|
1967 | 1967 | execution. |
|
1968 | 1968 | |
|
1969 | 1969 | Since rename detection is only run once, other factors might easily |
|
1970 | 1970 | affect the precision of the timing. However it should give a good |
|
1971 | 1971 | approximation of which revision triplets are very costly. |
|
1972 | 1972 | """ |
|
1973 | 1973 | opts = _byteskwargs(opts) |
|
1974 | 1974 | fm = ui.formatter(b'perf', opts) |
|
1975 | 1975 | dotiming = opts[b'timing'] |
|
1976 | 1976 | dostats = opts[b'stats'] |
|
1977 | 1977 | |
|
1978 | 1978 | output_template = [ |
|
1979 | 1979 | ("base", "%(base)12s"), |
|
1980 | 1980 | ("p1", "%(p1.node)12s"), |
|
1981 | 1981 | ("p2", "%(p2.node)12s"), |
|
1982 | 1982 | ("p1.nb-revs", "%(p1.nbrevs)12d"), |
|
1983 | 1983 | ("p1.nb-files", "%(p1.nbmissingfiles)12d"), |
|
1984 | 1984 | ("p1.renames", "%(p1.renamedfiles)12d"), |
|
1985 | 1985 | ("p1.time", "%(p1.time)12.3f"), |
|
1986 | 1986 | ("p2.nb-revs", "%(p2.nbrevs)12d"), |
|
1987 | 1987 | ("p2.nb-files", "%(p2.nbmissingfiles)12d"), |
|
1988 | 1988 | ("p2.renames", "%(p2.renamedfiles)12d"), |
|
1989 | 1989 | ("p2.time", "%(p2.time)12.3f"), |
|
1990 | 1990 | ("renames", "%(nbrenamedfiles)12d"), |
|
1991 | 1991 | ("total.time", "%(time)12.3f"), |
|
1992 | 1992 | ] |
|
1993 | 1993 | if not dotiming: |
|
1994 | 1994 | output_template = [ |
|
1995 | 1995 | i |
|
1996 | 1996 | for i in output_template |
|
1997 | 1997 | if not ('time' in i[0] or 'renames' in i[0]) |
|
1998 | 1998 | ] |
|
1999 | 1999 | header_names = [h for (h, v) in output_template] |
|
2000 | 2000 | output = ' '.join([v for (h, v) in output_template]) + '\n' |
|
2001 | 2001 | header = ' '.join(['%12s'] * len(header_names)) + '\n' |
|
2002 | 2002 | fm.plain(header % tuple(header_names)) |
|
2003 | 2003 | |
|
2004 | 2004 | if not revs: |
|
2005 | 2005 | revs = ['all()'] |
|
2006 | 2006 | revs = scmutil.revrange(repo, revs) |
|
2007 | 2007 | |
|
2008 | 2008 | if dostats: |
|
2009 | 2009 | alldata = { |
|
2010 | 2010 | 'nbrevs': [], |
|
2011 | 2011 | 'nbmissingfiles': [], |
|
2012 | 2012 | } |
|
2013 | 2013 | if dotiming: |
|
2014 | 2014 | alldata['parentnbrenames'] = [] |
|
2015 | 2015 | alldata['totalnbrenames'] = [] |
|
2016 | 2016 | alldata['parenttime'] = [] |
|
2017 | 2017 | alldata['totaltime'] = [] |
|
2018 | 2018 | |
|
2019 | 2019 | roi = repo.revs('merge() and %ld', revs) |
|
2020 | 2020 | for r in roi: |
|
2021 | 2021 | ctx = repo[r] |
|
2022 | 2022 | p1 = ctx.p1() |
|
2023 | 2023 | p2 = ctx.p2() |
|
2024 | 2024 | bases = repo.changelog._commonancestorsheads(p1.rev(), p2.rev()) |
|
2025 | 2025 | for b in bases: |
|
2026 | 2026 | b = repo[b] |
|
2027 | 2027 | p1missing = copies._computeforwardmissing(b, p1) |
|
2028 | 2028 | p2missing = copies._computeforwardmissing(b, p2) |
|
2029 | 2029 | data = { |
|
2030 | 2030 | b'base': b.hex(), |
|
2031 | 2031 | b'p1.node': p1.hex(), |
|
2032 | 2032 | b'p1.nbrevs': len(repo.revs('only(%d, %d)', p1.rev(), b.rev())), |
|
2033 | 2033 | b'p1.nbmissingfiles': len(p1missing), |
|
2034 | 2034 | b'p2.node': p2.hex(), |
|
2035 | 2035 | b'p2.nbrevs': len(repo.revs('only(%d, %d)', p2.rev(), b.rev())), |
|
2036 | 2036 | b'p2.nbmissingfiles': len(p2missing), |
|
2037 | 2037 | } |
|
2038 | 2038 | if dostats: |
|
2039 | 2039 | if p1missing: |
|
2040 | 2040 | alldata['nbrevs'].append( |
|
2041 | 2041 | (data['p1.nbrevs'], b.hex(), p1.hex()) |
|
2042 | 2042 | ) |
|
2043 | 2043 | alldata['nbmissingfiles'].append( |
|
2044 | 2044 | (data['p1.nbmissingfiles'], b.hex(), p1.hex()) |
|
2045 | 2045 | ) |
|
2046 | 2046 | if p2missing: |
|
2047 | 2047 | alldata['nbrevs'].append( |
|
2048 | 2048 | (data['p2.nbrevs'], b.hex(), p2.hex()) |
|
2049 | 2049 | ) |
|
2050 | 2050 | alldata['nbmissingfiles'].append( |
|
2051 | 2051 | (data['p2.nbmissingfiles'], b.hex(), p2.hex()) |
|
2052 | 2052 | ) |
|
2053 | 2053 | if dotiming: |
|
2054 | 2054 | begin = util.timer() |
|
2055 | 2055 | mergedata = copies.mergecopies(repo, p1, p2, b) |
|
2056 | 2056 | end = util.timer() |
|
2057 | 2057 | # not very stable timing since we did only one run |
|
2058 | 2058 | data['time'] = end - begin |
|
2059 | 2059 | # mergedata contains five dicts: "copy", "movewithdir", |
|
2060 | 2060 | # "diverge", "renamedelete" and "dirmove". |
|
2061 | 2061 | # The first 4 are about renamed file so lets count that. |
|
2062 | 2062 | renames = len(mergedata[0]) |
|
2063 | 2063 | renames += len(mergedata[1]) |
|
2064 | 2064 | renames += len(mergedata[2]) |
|
2065 | 2065 | renames += len(mergedata[3]) |
|
2066 | 2066 | data['nbrenamedfiles'] = renames |
|
2067 | 2067 | begin = util.timer() |
|
2068 | 2068 | p1renames = copies.pathcopies(b, p1) |
|
2069 | 2069 | end = util.timer() |
|
2070 | 2070 | data['p1.time'] = end - begin |
|
2071 | 2071 | begin = util.timer() |
|
2072 | 2072 | p2renames = copies.pathcopies(b, p2) |
|
2073 | 2073 | end = util.timer() |
|
2074 | 2074 | data['p2.time'] = end - begin |
|
2075 | 2075 | data['p1.renamedfiles'] = len(p1renames) |
|
2076 | 2076 | data['p2.renamedfiles'] = len(p2renames) |
|
2077 | 2077 | |
|
2078 | 2078 | if dostats: |
|
2079 | 2079 | if p1missing: |
|
2080 | 2080 | alldata['parentnbrenames'].append( |
|
2081 | 2081 | (data['p1.renamedfiles'], b.hex(), p1.hex()) |
|
2082 | 2082 | ) |
|
2083 | 2083 | alldata['parenttime'].append( |
|
2084 | 2084 | (data['p1.time'], b.hex(), p1.hex()) |
|
2085 | 2085 | ) |
|
2086 | 2086 | if p2missing: |
|
2087 | 2087 | alldata['parentnbrenames'].append( |
|
2088 | 2088 | (data['p2.renamedfiles'], b.hex(), p2.hex()) |
|
2089 | 2089 | ) |
|
2090 | 2090 | alldata['parenttime'].append( |
|
2091 | 2091 | (data['p2.time'], b.hex(), p2.hex()) |
|
2092 | 2092 | ) |
|
2093 | 2093 | if p1missing or p2missing: |
|
2094 | 2094 | alldata['totalnbrenames'].append( |
|
2095 | 2095 | ( |
|
2096 | 2096 | data['nbrenamedfiles'], |
|
2097 | 2097 | b.hex(), |
|
2098 | 2098 | p1.hex(), |
|
2099 | 2099 | p2.hex(), |
|
2100 | 2100 | ) |
|
2101 | 2101 | ) |
|
2102 | 2102 | alldata['totaltime'].append( |
|
2103 | 2103 | (data['time'], b.hex(), p1.hex(), p2.hex()) |
|
2104 | 2104 | ) |
|
2105 | 2105 | fm.startitem() |
|
2106 | 2106 | fm.data(**data) |
|
2107 | 2107 | # make node pretty for the human output |
|
2108 | 2108 | out = data.copy() |
|
2109 | 2109 | out['base'] = fm.hexfunc(b.node()) |
|
2110 | 2110 | out['p1.node'] = fm.hexfunc(p1.node()) |
|
2111 | 2111 | out['p2.node'] = fm.hexfunc(p2.node()) |
|
2112 | 2112 | fm.plain(output % out) |
|
2113 | 2113 | |
|
2114 | 2114 | fm.end() |
|
2115 | 2115 | if dostats: |
|
2116 | 2116 | # use a second formatter because the data are quite different, not sure |
|
2117 | 2117 | # how it flies with the templater. |
|
2118 | 2118 | entries = [ |
|
2119 | 2119 | ('nbrevs', 'number of revision covered'), |
|
2120 | 2120 | ('nbmissingfiles', 'number of missing files at head'), |
|
2121 | 2121 | ] |
|
2122 | 2122 | if dotiming: |
|
2123 | 2123 | entries.append( |
|
2124 | 2124 | ('parentnbrenames', 'rename from one parent to base') |
|
2125 | 2125 | ) |
|
2126 | 2126 | entries.append(('totalnbrenames', 'total number of renames')) |
|
2127 | 2127 | entries.append(('parenttime', 'time for one parent')) |
|
2128 | 2128 | entries.append(('totaltime', 'time for both parents')) |
|
2129 | 2129 | _displaystats(ui, opts, entries, alldata) |
|
2130 | 2130 | |
|
2131 | 2131 | |
|
2132 | 2132 | @command( |
|
2133 | 2133 | b'perf::helper-pathcopies|perfhelper-pathcopies', |
|
2134 | 2134 | formatteropts |
|
2135 | 2135 | + [ |
|
2136 | 2136 | (b'r', b'revs', [], b'restrict search to these revisions'), |
|
2137 | 2137 | (b'', b'timing', False, b'provides extra data (costly)'), |
|
2138 | 2138 | (b'', b'stats', False, b'provides statistic about the measured data'), |
|
2139 | 2139 | ], |
|
2140 | 2140 | ) |
|
2141 | 2141 | def perfhelperpathcopies(ui, repo, revs=[], **opts): |
|
2142 | 2142 | """find statistic about potential parameters for the `perftracecopies` |
|
2143 | 2143 | |
|
2144 | 2144 | This command find source-destination pair relevant for copytracing testing. |
|
2145 | 2145 | It report value for some of the parameters that impact copy tracing time. |
|
2146 | 2146 | |
|
2147 | 2147 | If `--timing` is set, rename detection is run and the associated timing |
|
2148 | 2148 | will be reported. The extra details comes at the cost of a slower command |
|
2149 | 2149 | execution. |
|
2150 | 2150 | |
|
2151 | 2151 | Since the rename detection is only run once, other factors might easily |
|
2152 | 2152 | affect the precision of the timing. However it should give a good |
|
2153 | 2153 | approximation of which revision pairs are very costly. |
|
2154 | 2154 | """ |
|
2155 | 2155 | opts = _byteskwargs(opts) |
|
2156 | 2156 | fm = ui.formatter(b'perf', opts) |
|
2157 | 2157 | dotiming = opts[b'timing'] |
|
2158 | 2158 | dostats = opts[b'stats'] |
|
2159 | 2159 | |
|
2160 | 2160 | if dotiming: |
|
2161 | 2161 | header = '%12s %12s %12s %12s %12s %12s\n' |
|
2162 | 2162 | output = ( |
|
2163 | 2163 | "%(source)12s %(destination)12s " |
|
2164 | 2164 | "%(nbrevs)12d %(nbmissingfiles)12d " |
|
2165 | 2165 | "%(nbrenamedfiles)12d %(time)18.5f\n" |
|
2166 | 2166 | ) |
|
2167 | 2167 | header_names = ( |
|
2168 | 2168 | "source", |
|
2169 | 2169 | "destination", |
|
2170 | 2170 | "nb-revs", |
|
2171 | 2171 | "nb-files", |
|
2172 | 2172 | "nb-renames", |
|
2173 | 2173 | "time", |
|
2174 | 2174 | ) |
|
2175 | 2175 | fm.plain(header % header_names) |
|
2176 | 2176 | else: |
|
2177 | 2177 | header = '%12s %12s %12s %12s\n' |
|
2178 | 2178 | output = ( |
|
2179 | 2179 | "%(source)12s %(destination)12s " |
|
2180 | 2180 | "%(nbrevs)12d %(nbmissingfiles)12d\n" |
|
2181 | 2181 | ) |
|
2182 | 2182 | fm.plain(header % ("source", "destination", "nb-revs", "nb-files")) |
|
2183 | 2183 | |
|
2184 | 2184 | if not revs: |
|
2185 | 2185 | revs = ['all()'] |
|
2186 | 2186 | revs = scmutil.revrange(repo, revs) |
|
2187 | 2187 | |
|
2188 | 2188 | if dostats: |
|
2189 | 2189 | alldata = { |
|
2190 | 2190 | 'nbrevs': [], |
|
2191 | 2191 | 'nbmissingfiles': [], |
|
2192 | 2192 | } |
|
2193 | 2193 | if dotiming: |
|
2194 | 2194 | alldata['nbrenames'] = [] |
|
2195 | 2195 | alldata['time'] = [] |
|
2196 | 2196 | |
|
2197 | 2197 | roi = repo.revs('merge() and %ld', revs) |
|
2198 | 2198 | for r in roi: |
|
2199 | 2199 | ctx = repo[r] |
|
2200 | 2200 | p1 = ctx.p1().rev() |
|
2201 | 2201 | p2 = ctx.p2().rev() |
|
2202 | 2202 | bases = repo.changelog._commonancestorsheads(p1, p2) |
|
2203 | 2203 | for p in (p1, p2): |
|
2204 | 2204 | for b in bases: |
|
2205 | 2205 | base = repo[b] |
|
2206 | 2206 | parent = repo[p] |
|
2207 | 2207 | missing = copies._computeforwardmissing(base, parent) |
|
2208 | 2208 | if not missing: |
|
2209 | 2209 | continue |
|
2210 | 2210 | data = { |
|
2211 | 2211 | b'source': base.hex(), |
|
2212 | 2212 | b'destination': parent.hex(), |
|
2213 | 2213 | b'nbrevs': len(repo.revs('only(%d, %d)', p, b)), |
|
2214 | 2214 | b'nbmissingfiles': len(missing), |
|
2215 | 2215 | } |
|
2216 | 2216 | if dostats: |
|
2217 | 2217 | alldata['nbrevs'].append( |
|
2218 | 2218 | ( |
|
2219 | 2219 | data['nbrevs'], |
|
2220 | 2220 | base.hex(), |
|
2221 | 2221 | parent.hex(), |
|
2222 | 2222 | ) |
|
2223 | 2223 | ) |
|
2224 | 2224 | alldata['nbmissingfiles'].append( |
|
2225 | 2225 | ( |
|
2226 | 2226 | data['nbmissingfiles'], |
|
2227 | 2227 | base.hex(), |
|
2228 | 2228 | parent.hex(), |
|
2229 | 2229 | ) |
|
2230 | 2230 | ) |
|
2231 | 2231 | if dotiming: |
|
2232 | 2232 | begin = util.timer() |
|
2233 | 2233 | renames = copies.pathcopies(base, parent) |
|
2234 | 2234 | end = util.timer() |
|
2235 | 2235 | # not very stable timing since we did only one run |
|
2236 | 2236 | data['time'] = end - begin |
|
2237 | 2237 | data['nbrenamedfiles'] = len(renames) |
|
2238 | 2238 | if dostats: |
|
2239 | 2239 | alldata['time'].append( |
|
2240 | 2240 | ( |
|
2241 | 2241 | data['time'], |
|
2242 | 2242 | base.hex(), |
|
2243 | 2243 | parent.hex(), |
|
2244 | 2244 | ) |
|
2245 | 2245 | ) |
|
2246 | 2246 | alldata['nbrenames'].append( |
|
2247 | 2247 | ( |
|
2248 | 2248 | data['nbrenamedfiles'], |
|
2249 | 2249 | base.hex(), |
|
2250 | 2250 | parent.hex(), |
|
2251 | 2251 | ) |
|
2252 | 2252 | ) |
|
2253 | 2253 | fm.startitem() |
|
2254 | 2254 | fm.data(**data) |
|
2255 | 2255 | out = data.copy() |
|
2256 | 2256 | out['source'] = fm.hexfunc(base.node()) |
|
2257 | 2257 | out['destination'] = fm.hexfunc(parent.node()) |
|
2258 | 2258 | fm.plain(output % out) |
|
2259 | 2259 | |
|
2260 | 2260 | fm.end() |
|
2261 | 2261 | if dostats: |
|
2262 | 2262 | entries = [ |
|
2263 | 2263 | ('nbrevs', 'number of revision covered'), |
|
2264 | 2264 | ('nbmissingfiles', 'number of missing files at head'), |
|
2265 | 2265 | ] |
|
2266 | 2266 | if dotiming: |
|
2267 | 2267 | entries.append(('nbrenames', 'renamed files')) |
|
2268 | 2268 | entries.append(('time', 'time')) |
|
2269 | 2269 | _displaystats(ui, opts, entries, alldata) |
|
2270 | 2270 | |
|
2271 | 2271 | |
|
2272 | 2272 | @command(b'perf::cca|perfcca', formatteropts) |
|
2273 | 2273 | def perfcca(ui, repo, **opts): |
|
2274 | 2274 | opts = _byteskwargs(opts) |
|
2275 | 2275 | timer, fm = gettimer(ui, opts) |
|
2276 | 2276 | timer(lambda: scmutil.casecollisionauditor(ui, False, repo.dirstate)) |
|
2277 | 2277 | fm.end() |
|
2278 | 2278 | |
|
2279 | 2279 | |
|
2280 | 2280 | @command(b'perf::fncacheload|perffncacheload', formatteropts) |
|
2281 | 2281 | def perffncacheload(ui, repo, **opts): |
|
2282 | 2282 | opts = _byteskwargs(opts) |
|
2283 | 2283 | timer, fm = gettimer(ui, opts) |
|
2284 | 2284 | s = repo.store |
|
2285 | 2285 | |
|
2286 | 2286 | def d(): |
|
2287 | 2287 | s.fncache._load() |
|
2288 | 2288 | |
|
2289 | 2289 | timer(d) |
|
2290 | 2290 | fm.end() |
|
2291 | 2291 | |
|
2292 | 2292 | |
|
2293 | 2293 | @command(b'perf::fncachewrite|perffncachewrite', formatteropts) |
|
2294 | 2294 | def perffncachewrite(ui, repo, **opts): |
|
2295 | 2295 | opts = _byteskwargs(opts) |
|
2296 | 2296 | timer, fm = gettimer(ui, opts) |
|
2297 | 2297 | s = repo.store |
|
2298 | 2298 | lock = repo.lock() |
|
2299 | 2299 | s.fncache._load() |
|
2300 | 2300 | tr = repo.transaction(b'perffncachewrite') |
|
2301 | 2301 | tr.addbackup(b'fncache') |
|
2302 | 2302 | |
|
2303 | 2303 | def d(): |
|
2304 | 2304 | s.fncache._dirty = True |
|
2305 | 2305 | s.fncache.write(tr) |
|
2306 | 2306 | |
|
2307 | 2307 | timer(d) |
|
2308 | 2308 | tr.close() |
|
2309 | 2309 | lock.release() |
|
2310 | 2310 | fm.end() |
|
2311 | 2311 | |
|
2312 | 2312 | |
|
2313 | 2313 | @command(b'perf::fncacheencode|perffncacheencode', formatteropts) |
|
2314 | 2314 | def perffncacheencode(ui, repo, **opts): |
|
2315 | 2315 | opts = _byteskwargs(opts) |
|
2316 | 2316 | timer, fm = gettimer(ui, opts) |
|
2317 | 2317 | s = repo.store |
|
2318 | 2318 | s.fncache._load() |
|
2319 | 2319 | |
|
2320 | 2320 | def d(): |
|
2321 | 2321 | for p in s.fncache.entries: |
|
2322 | 2322 | s.encode(p) |
|
2323 | 2323 | |
|
2324 | 2324 | timer(d) |
|
2325 | 2325 | fm.end() |
|
2326 | 2326 | |
|
2327 | 2327 | |
|
2328 | 2328 | def _bdiffworker(q, blocks, xdiff, ready, done): |
|
2329 | 2329 | while not done.is_set(): |
|
2330 | 2330 | pair = q.get() |
|
2331 | 2331 | while pair is not None: |
|
2332 | 2332 | if xdiff: |
|
2333 | 2333 | mdiff.bdiff.xdiffblocks(*pair) |
|
2334 | 2334 | elif blocks: |
|
2335 | 2335 | mdiff.bdiff.blocks(*pair) |
|
2336 | 2336 | else: |
|
2337 | 2337 | mdiff.textdiff(*pair) |
|
2338 | 2338 | q.task_done() |
|
2339 | 2339 | pair = q.get() |
|
2340 | 2340 | q.task_done() # for the None one |
|
2341 | 2341 | with ready: |
|
2342 | 2342 | ready.wait() |
|
2343 | 2343 | |
|
2344 | 2344 | |
|
2345 | 2345 | def _manifestrevision(repo, mnode): |
|
2346 | 2346 | ml = repo.manifestlog |
|
2347 | 2347 | |
|
2348 | 2348 | if util.safehasattr(ml, b'getstorage'): |
|
2349 | 2349 | store = ml.getstorage(b'') |
|
2350 | 2350 | else: |
|
2351 | 2351 | store = ml._revlog |
|
2352 | 2352 | |
|
2353 | 2353 | return store.revision(mnode) |
|
2354 | 2354 | |
|
2355 | 2355 | |
|
2356 | 2356 | @command( |
|
2357 | 2357 | b'perf::bdiff|perfbdiff', |
|
2358 | 2358 | revlogopts |
|
2359 | 2359 | + formatteropts |
|
2360 | 2360 | + [ |
|
2361 | 2361 | ( |
|
2362 | 2362 | b'', |
|
2363 | 2363 | b'count', |
|
2364 | 2364 | 1, |
|
2365 | 2365 | b'number of revisions to test (when using --startrev)', |
|
2366 | 2366 | ), |
|
2367 | 2367 | (b'', b'alldata', False, b'test bdiffs for all associated revisions'), |
|
2368 | 2368 | (b'', b'threads', 0, b'number of thread to use (disable with 0)'), |
|
2369 | 2369 | (b'', b'blocks', False, b'test computing diffs into blocks'), |
|
2370 | 2370 | (b'', b'xdiff', False, b'use xdiff algorithm'), |
|
2371 | 2371 | ], |
|
2372 | 2372 | b'-c|-m|FILE REV', |
|
2373 | 2373 | ) |
|
2374 | 2374 | def perfbdiff(ui, repo, file_, rev=None, count=None, threads=0, **opts): |
|
2375 | 2375 | """benchmark a bdiff between revisions |
|
2376 | 2376 | |
|
2377 | 2377 | By default, benchmark a bdiff between its delta parent and itself. |
|
2378 | 2378 | |
|
2379 | 2379 | With ``--count``, benchmark bdiffs between delta parents and self for N |
|
2380 | 2380 | revisions starting at the specified revision. |
|
2381 | 2381 | |
|
2382 | 2382 | With ``--alldata``, assume the requested revision is a changeset and |
|
2383 | 2383 | measure bdiffs for all changes related to that changeset (manifest |
|
2384 | 2384 | and filelogs). |
|
2385 | 2385 | """ |
|
2386 | 2386 | opts = _byteskwargs(opts) |
|
2387 | 2387 | |
|
2388 | 2388 | if opts[b'xdiff'] and not opts[b'blocks']: |
|
2389 | 2389 | raise error.CommandError(b'perfbdiff', b'--xdiff requires --blocks') |
|
2390 | 2390 | |
|
2391 | 2391 | if opts[b'alldata']: |
|
2392 | 2392 | opts[b'changelog'] = True |
|
2393 | 2393 | |
|
2394 | 2394 | if opts.get(b'changelog') or opts.get(b'manifest'): |
|
2395 | 2395 | file_, rev = None, file_ |
|
2396 | 2396 | elif rev is None: |
|
2397 | 2397 | raise error.CommandError(b'perfbdiff', b'invalid arguments') |
|
2398 | 2398 | |
|
2399 | 2399 | blocks = opts[b'blocks'] |
|
2400 | 2400 | xdiff = opts[b'xdiff'] |
|
2401 | 2401 | textpairs = [] |
|
2402 | 2402 | |
|
2403 | 2403 | r = cmdutil.openrevlog(repo, b'perfbdiff', file_, opts) |
|
2404 | 2404 | |
|
2405 | 2405 | startrev = r.rev(r.lookup(rev)) |
|
2406 | 2406 | for rev in range(startrev, min(startrev + count, len(r) - 1)): |
|
2407 | 2407 | if opts[b'alldata']: |
|
2408 | 2408 | # Load revisions associated with changeset. |
|
2409 | 2409 | ctx = repo[rev] |
|
2410 | 2410 | mtext = _manifestrevision(repo, ctx.manifestnode()) |
|
2411 | 2411 | for pctx in ctx.parents(): |
|
2412 | 2412 | pman = _manifestrevision(repo, pctx.manifestnode()) |
|
2413 | 2413 | textpairs.append((pman, mtext)) |
|
2414 | 2414 | |
|
2415 | 2415 | # Load filelog revisions by iterating manifest delta. |
|
2416 | 2416 | man = ctx.manifest() |
|
2417 | 2417 | pman = ctx.p1().manifest() |
|
2418 | 2418 | for filename, change in pman.diff(man).items(): |
|
2419 | 2419 | fctx = repo.file(filename) |
|
2420 | 2420 | f1 = fctx.revision(change[0][0] or -1) |
|
2421 | 2421 | f2 = fctx.revision(change[1][0] or -1) |
|
2422 | 2422 | textpairs.append((f1, f2)) |
|
2423 | 2423 | else: |
|
2424 | 2424 | dp = r.deltaparent(rev) |
|
2425 | 2425 | textpairs.append((r.revision(dp), r.revision(rev))) |
|
2426 | 2426 | |
|
2427 | 2427 | withthreads = threads > 0 |
|
2428 | 2428 | if not withthreads: |
|
2429 | 2429 | |
|
2430 | 2430 | def d(): |
|
2431 | 2431 | for pair in textpairs: |
|
2432 | 2432 | if xdiff: |
|
2433 | 2433 | mdiff.bdiff.xdiffblocks(*pair) |
|
2434 | 2434 | elif blocks: |
|
2435 | 2435 | mdiff.bdiff.blocks(*pair) |
|
2436 | 2436 | else: |
|
2437 | 2437 | mdiff.textdiff(*pair) |
|
2438 | 2438 | |
|
2439 | 2439 | else: |
|
2440 | 2440 | q = queue() |
|
2441 | 2441 | for i in _xrange(threads): |
|
2442 | 2442 | q.put(None) |
|
2443 | 2443 | ready = threading.Condition() |
|
2444 | 2444 | done = threading.Event() |
|
2445 | 2445 | for i in _xrange(threads): |
|
2446 | 2446 | threading.Thread( |
|
2447 | 2447 | target=_bdiffworker, args=(q, blocks, xdiff, ready, done) |
|
2448 | 2448 | ).start() |
|
2449 | 2449 | q.join() |
|
2450 | 2450 | |
|
2451 | 2451 | def d(): |
|
2452 | 2452 | for pair in textpairs: |
|
2453 | 2453 | q.put(pair) |
|
2454 | 2454 | for i in _xrange(threads): |
|
2455 | 2455 | q.put(None) |
|
2456 | 2456 | with ready: |
|
2457 | 2457 | ready.notify_all() |
|
2458 | 2458 | q.join() |
|
2459 | 2459 | |
|
2460 | 2460 | timer, fm = gettimer(ui, opts) |
|
2461 | 2461 | timer(d) |
|
2462 | 2462 | fm.end() |
|
2463 | 2463 | |
|
2464 | 2464 | if withthreads: |
|
2465 | 2465 | done.set() |
|
2466 | 2466 | for i in _xrange(threads): |
|
2467 | 2467 | q.put(None) |
|
2468 | 2468 | with ready: |
|
2469 | 2469 | ready.notify_all() |
|
2470 | 2470 | |
|
2471 | 2471 | |
|
2472 | 2472 | @command( |
|
2473 | 2473 | b'perf::unidiff|perfunidiff', |
|
2474 | 2474 | revlogopts |
|
2475 | 2475 | + formatteropts |
|
2476 | 2476 | + [ |
|
2477 | 2477 | ( |
|
2478 | 2478 | b'', |
|
2479 | 2479 | b'count', |
|
2480 | 2480 | 1, |
|
2481 | 2481 | b'number of revisions to test (when using --startrev)', |
|
2482 | 2482 | ), |
|
2483 | 2483 | (b'', b'alldata', False, b'test unidiffs for all associated revisions'), |
|
2484 | 2484 | ], |
|
2485 | 2485 | b'-c|-m|FILE REV', |
|
2486 | 2486 | ) |
|
2487 | 2487 | def perfunidiff(ui, repo, file_, rev=None, count=None, **opts): |
|
2488 | 2488 | """benchmark a unified diff between revisions |
|
2489 | 2489 | |
|
2490 | 2490 | This doesn't include any copy tracing - it's just a unified diff |
|
2491 | 2491 | of the texts. |
|
2492 | 2492 | |
|
2493 | 2493 | By default, benchmark a diff between its delta parent and itself. |
|
2494 | 2494 | |
|
2495 | 2495 | With ``--count``, benchmark diffs between delta parents and self for N |
|
2496 | 2496 | revisions starting at the specified revision. |
|
2497 | 2497 | |
|
2498 | 2498 | With ``--alldata``, assume the requested revision is a changeset and |
|
2499 | 2499 | measure diffs for all changes related to that changeset (manifest |
|
2500 | 2500 | and filelogs). |
|
2501 | 2501 | """ |
|
2502 | 2502 | opts = _byteskwargs(opts) |
|
2503 | 2503 | if opts[b'alldata']: |
|
2504 | 2504 | opts[b'changelog'] = True |
|
2505 | 2505 | |
|
2506 | 2506 | if opts.get(b'changelog') or opts.get(b'manifest'): |
|
2507 | 2507 | file_, rev = None, file_ |
|
2508 | 2508 | elif rev is None: |
|
2509 | 2509 | raise error.CommandError(b'perfunidiff', b'invalid arguments') |
|
2510 | 2510 | |
|
2511 | 2511 | textpairs = [] |
|
2512 | 2512 | |
|
2513 | 2513 | r = cmdutil.openrevlog(repo, b'perfunidiff', file_, opts) |
|
2514 | 2514 | |
|
2515 | 2515 | startrev = r.rev(r.lookup(rev)) |
|
2516 | 2516 | for rev in range(startrev, min(startrev + count, len(r) - 1)): |
|
2517 | 2517 | if opts[b'alldata']: |
|
2518 | 2518 | # Load revisions associated with changeset. |
|
2519 | 2519 | ctx = repo[rev] |
|
2520 | 2520 | mtext = _manifestrevision(repo, ctx.manifestnode()) |
|
2521 | 2521 | for pctx in ctx.parents(): |
|
2522 | 2522 | pman = _manifestrevision(repo, pctx.manifestnode()) |
|
2523 | 2523 | textpairs.append((pman, mtext)) |
|
2524 | 2524 | |
|
2525 | 2525 | # Load filelog revisions by iterating manifest delta. |
|
2526 | 2526 | man = ctx.manifest() |
|
2527 | 2527 | pman = ctx.p1().manifest() |
|
2528 | 2528 | for filename, change in pman.diff(man).items(): |
|
2529 | 2529 | fctx = repo.file(filename) |
|
2530 | 2530 | f1 = fctx.revision(change[0][0] or -1) |
|
2531 | 2531 | f2 = fctx.revision(change[1][0] or -1) |
|
2532 | 2532 | textpairs.append((f1, f2)) |
|
2533 | 2533 | else: |
|
2534 | 2534 | dp = r.deltaparent(rev) |
|
2535 | 2535 | textpairs.append((r.revision(dp), r.revision(rev))) |
|
2536 | 2536 | |
|
2537 | 2537 | def d(): |
|
2538 | 2538 | for left, right in textpairs: |
|
2539 | 2539 | # The date strings don't matter, so we pass empty strings. |
|
2540 | 2540 | headerlines, hunks = mdiff.unidiff( |
|
2541 | 2541 | left, b'', right, b'', b'left', b'right', binary=False |
|
2542 | 2542 | ) |
|
2543 | 2543 | # consume iterators in roughly the way patch.py does |
|
2544 | 2544 | b'\n'.join(headerlines) |
|
2545 | 2545 | b''.join(sum((list(hlines) for hrange, hlines in hunks), [])) |
|
2546 | 2546 | |
|
2547 | 2547 | timer, fm = gettimer(ui, opts) |
|
2548 | 2548 | timer(d) |
|
2549 | 2549 | fm.end() |
|
2550 | 2550 | |
|
2551 | 2551 | |
|
2552 | 2552 | @command(b'perf::diffwd|perfdiffwd', formatteropts) |
|
2553 | 2553 | def perfdiffwd(ui, repo, **opts): |
|
2554 | 2554 | """Profile diff of working directory changes""" |
|
2555 | 2555 | opts = _byteskwargs(opts) |
|
2556 | 2556 | timer, fm = gettimer(ui, opts) |
|
2557 | 2557 | options = { |
|
2558 | 2558 | 'w': 'ignore_all_space', |
|
2559 | 2559 | 'b': 'ignore_space_change', |
|
2560 | 2560 | 'B': 'ignore_blank_lines', |
|
2561 | 2561 | } |
|
2562 | 2562 | |
|
2563 | 2563 | for diffopt in ('', 'w', 'b', 'B', 'wB'): |
|
2564 | 2564 | opts = {options[c]: b'1' for c in diffopt} |
|
2565 | 2565 | |
|
2566 | 2566 | def d(): |
|
2567 | 2567 | ui.pushbuffer() |
|
2568 | 2568 | commands.diff(ui, repo, **opts) |
|
2569 | 2569 | ui.popbuffer() |
|
2570 | 2570 | |
|
2571 | 2571 | diffopt = diffopt.encode('ascii') |
|
2572 | 2572 | title = b'diffopts: %s' % (diffopt and (b'-' + diffopt) or b'none') |
|
2573 | 2573 | timer(d, title=title) |
|
2574 | 2574 | fm.end() |
|
2575 | 2575 | |
|
2576 | 2576 | |
|
2577 | 2577 | @command( |
|
2578 | 2578 | b'perf::revlogindex|perfrevlogindex', |
|
2579 | 2579 | revlogopts + formatteropts, |
|
2580 | 2580 | b'-c|-m|FILE', |
|
2581 | 2581 | ) |
|
2582 | 2582 | def perfrevlogindex(ui, repo, file_=None, **opts): |
|
2583 | 2583 | """Benchmark operations against a revlog index. |
|
2584 | 2584 | |
|
2585 | 2585 | This tests constructing a revlog instance, reading index data, |
|
2586 | 2586 | parsing index data, and performing various operations related to |
|
2587 | 2587 | index data. |
|
2588 | 2588 | """ |
|
2589 | 2589 | |
|
2590 | 2590 | opts = _byteskwargs(opts) |
|
2591 | 2591 | |
|
2592 | 2592 | rl = cmdutil.openrevlog(repo, b'perfrevlogindex', file_, opts) |
|
2593 | 2593 | |
|
2594 | 2594 | opener = getattr(rl, 'opener') # trick linter |
|
2595 | 2595 | indexfile = rl.indexfile |
|
2596 | 2596 | data = opener.read(indexfile) |
|
2597 | 2597 | |
|
2598 | 2598 | header = struct.unpack(b'>I', data[0:4])[0] |
|
2599 | 2599 | version = header & 0xFFFF |
|
2600 | 2600 | if version == 1: |
|
2601 | 2601 | revlogio = revlog.revlogio() |
|
2602 | 2602 | inline = header & (1 << 16) |
|
2603 | 2603 | else: |
|
2604 | 2604 | raise error.Abort(b'unsupported revlog version: %d' % version) |
|
2605 | 2605 | |
|
2606 | 2606 | rllen = len(rl) |
|
2607 | 2607 | |
|
2608 | 2608 | node0 = rl.node(0) |
|
2609 | 2609 | node25 = rl.node(rllen // 4) |
|
2610 | 2610 | node50 = rl.node(rllen // 2) |
|
2611 | 2611 | node75 = rl.node(rllen // 4 * 3) |
|
2612 | 2612 | node100 = rl.node(rllen - 1) |
|
2613 | 2613 | |
|
2614 | 2614 | allrevs = range(rllen) |
|
2615 | 2615 | allrevsrev = list(reversed(allrevs)) |
|
2616 | 2616 | allnodes = [rl.node(rev) for rev in range(rllen)] |
|
2617 | 2617 | allnodesrev = list(reversed(allnodes)) |
|
2618 | 2618 | |
|
2619 | 2619 | def constructor(): |
|
2620 | 2620 | revlog.revlog(opener, indexfile) |
|
2621 | 2621 | |
|
2622 | 2622 | def read(): |
|
2623 | 2623 | with opener(indexfile) as fh: |
|
2624 | 2624 | fh.read() |
|
2625 | 2625 | |
|
2626 | 2626 | def parseindex(): |
|
2627 | 2627 | revlogio.parseindex(data, inline) |
|
2628 | 2628 | |
|
2629 | 2629 | def getentry(revornode): |
|
2630 | 2630 | index = revlogio.parseindex(data, inline)[0] |
|
2631 | 2631 | index[revornode] |
|
2632 | 2632 | |
|
2633 | 2633 | def getentries(revs, count=1): |
|
2634 | 2634 | index = revlogio.parseindex(data, inline)[0] |
|
2635 | 2635 | |
|
2636 | 2636 | for i in range(count): |
|
2637 | 2637 | for rev in revs: |
|
2638 | 2638 | index[rev] |
|
2639 | 2639 | |
|
2640 | 2640 | def resolvenode(node): |
|
2641 | 2641 | index = revlogio.parseindex(data, inline)[0] |
|
2642 | 2642 | rev = getattr(index, 'rev', None) |
|
2643 | 2643 | if rev is None: |
|
2644 | 2644 | nodemap = getattr( |
|
2645 | 2645 | revlogio.parseindex(data, inline)[0], 'nodemap', None |
|
2646 | 2646 | ) |
|
2647 | 2647 | # This only works for the C code. |
|
2648 | 2648 | if nodemap is None: |
|
2649 | 2649 | return |
|
2650 | 2650 | rev = nodemap.__getitem__ |
|
2651 | 2651 | |
|
2652 | 2652 | try: |
|
2653 | 2653 | rev(node) |
|
2654 | 2654 | except error.RevlogError: |
|
2655 | 2655 | pass |
|
2656 | 2656 | |
|
2657 | 2657 | def resolvenodes(nodes, count=1): |
|
2658 | 2658 | index = revlogio.parseindex(data, inline)[0] |
|
2659 | 2659 | rev = getattr(index, 'rev', None) |
|
2660 | 2660 | if rev is None: |
|
2661 | 2661 | nodemap = getattr( |
|
2662 | 2662 | revlogio.parseindex(data, inline)[0], 'nodemap', None |
|
2663 | 2663 | ) |
|
2664 | 2664 | # This only works for the C code. |
|
2665 | 2665 | if nodemap is None: |
|
2666 | 2666 | return |
|
2667 | 2667 | rev = nodemap.__getitem__ |
|
2668 | 2668 | |
|
2669 | 2669 | for i in range(count): |
|
2670 | 2670 | for node in nodes: |
|
2671 | 2671 | try: |
|
2672 | 2672 | rev(node) |
|
2673 | 2673 | except error.RevlogError: |
|
2674 | 2674 | pass |
|
2675 | 2675 | |
|
2676 | 2676 | benches = [ |
|
2677 | 2677 | (constructor, b'revlog constructor'), |
|
2678 | 2678 | (read, b'read'), |
|
2679 | 2679 | (parseindex, b'create index object'), |
|
2680 | 2680 | (lambda: getentry(0), b'retrieve index entry for rev 0'), |
|
2681 | 2681 | (lambda: resolvenode(b'a' * 20), b'look up missing node'), |
|
2682 | 2682 | (lambda: resolvenode(node0), b'look up node at rev 0'), |
|
2683 | 2683 | (lambda: resolvenode(node25), b'look up node at 1/4 len'), |
|
2684 | 2684 | (lambda: resolvenode(node50), b'look up node at 1/2 len'), |
|
2685 | 2685 | (lambda: resolvenode(node75), b'look up node at 3/4 len'), |
|
2686 | 2686 | (lambda: resolvenode(node100), b'look up node at tip'), |
|
2687 | 2687 | # 2x variation is to measure caching impact. |
|
2688 | 2688 | (lambda: resolvenodes(allnodes), b'look up all nodes (forward)'), |
|
2689 | 2689 | (lambda: resolvenodes(allnodes, 2), b'look up all nodes 2x (forward)'), |
|
2690 | 2690 | (lambda: resolvenodes(allnodesrev), b'look up all nodes (reverse)'), |
|
2691 | 2691 | ( |
|
2692 | 2692 | lambda: resolvenodes(allnodesrev, 2), |
|
2693 | 2693 | b'look up all nodes 2x (reverse)', |
|
2694 | 2694 | ), |
|
2695 | 2695 | (lambda: getentries(allrevs), b'retrieve all index entries (forward)'), |
|
2696 | 2696 | ( |
|
2697 | 2697 | lambda: getentries(allrevs, 2), |
|
2698 | 2698 | b'retrieve all index entries 2x (forward)', |
|
2699 | 2699 | ), |
|
2700 | 2700 | ( |
|
2701 | 2701 | lambda: getentries(allrevsrev), |
|
2702 | 2702 | b'retrieve all index entries (reverse)', |
|
2703 | 2703 | ), |
|
2704 | 2704 | ( |
|
2705 | 2705 | lambda: getentries(allrevsrev, 2), |
|
2706 | 2706 | b'retrieve all index entries 2x (reverse)', |
|
2707 | 2707 | ), |
|
2708 | 2708 | ] |
|
2709 | 2709 | |
|
2710 | 2710 | for fn, title in benches: |
|
2711 | 2711 | timer, fm = gettimer(ui, opts) |
|
2712 | 2712 | timer(fn, title=title) |
|
2713 | 2713 | fm.end() |
|
2714 | 2714 | |
|
2715 | 2715 | |
|
2716 | 2716 | @command( |
|
2717 | 2717 | b'perf::revlogrevisions|perfrevlogrevisions', |
|
2718 | 2718 | revlogopts |
|
2719 | 2719 | + formatteropts |
|
2720 | 2720 | + [ |
|
2721 | 2721 | (b'd', b'dist', 100, b'distance between the revisions'), |
|
2722 | 2722 | (b's', b'startrev', 0, b'revision to start reading at'), |
|
2723 | 2723 | (b'', b'reverse', False, b'read in reverse'), |
|
2724 | 2724 | ], |
|
2725 | 2725 | b'-c|-m|FILE', |
|
2726 | 2726 | ) |
|
2727 | 2727 | def perfrevlogrevisions( |
|
2728 | 2728 | ui, repo, file_=None, startrev=0, reverse=False, **opts |
|
2729 | 2729 | ): |
|
2730 | 2730 | """Benchmark reading a series of revisions from a revlog. |
|
2731 | 2731 | |
|
2732 | 2732 | By default, we read every ``-d/--dist`` revision from 0 to tip of |
|
2733 | 2733 | the specified revlog. |
|
2734 | 2734 | |
|
2735 | 2735 | The start revision can be defined via ``-s/--startrev``. |
|
2736 | 2736 | """ |
|
2737 | 2737 | opts = _byteskwargs(opts) |
|
2738 | 2738 | |
|
2739 | 2739 | rl = cmdutil.openrevlog(repo, b'perfrevlogrevisions', file_, opts) |
|
2740 | 2740 | rllen = getlen(ui)(rl) |
|
2741 | 2741 | |
|
2742 | 2742 | if startrev < 0: |
|
2743 | 2743 | startrev = rllen + startrev |
|
2744 | 2744 | |
|
2745 | 2745 | def d(): |
|
2746 | 2746 | rl.clearcaches() |
|
2747 | 2747 | |
|
2748 | 2748 | beginrev = startrev |
|
2749 | 2749 | endrev = rllen |
|
2750 | 2750 | dist = opts[b'dist'] |
|
2751 | 2751 | |
|
2752 | 2752 | if reverse: |
|
2753 | 2753 | beginrev, endrev = endrev - 1, beginrev - 1 |
|
2754 | 2754 | dist = -1 * dist |
|
2755 | 2755 | |
|
2756 | 2756 | for x in _xrange(beginrev, endrev, dist): |
|
2757 | 2757 | # Old revisions don't support passing int. |
|
2758 | 2758 | n = rl.node(x) |
|
2759 | 2759 | rl.revision(n) |
|
2760 | 2760 | |
|
2761 | 2761 | timer, fm = gettimer(ui, opts) |
|
2762 | 2762 | timer(d) |
|
2763 | 2763 | fm.end() |
|
2764 | 2764 | |
|
2765 | 2765 | |
|
2766 | 2766 | @command( |
|
2767 | 2767 | b'perf::revlogwrite|perfrevlogwrite', |
|
2768 | 2768 | revlogopts |
|
2769 | 2769 | + formatteropts |
|
2770 | 2770 | + [ |
|
2771 | 2771 | (b's', b'startrev', 1000, b'revision to start writing at'), |
|
2772 | 2772 | (b'', b'stoprev', -1, b'last revision to write'), |
|
2773 | 2773 | (b'', b'count', 3, b'number of passes to perform'), |
|
2774 | 2774 | (b'', b'details', False, b'print timing for every revisions tested'), |
|
2775 | 2775 | (b'', b'source', b'full', b'the kind of data feed in the revlog'), |
|
2776 | 2776 | (b'', b'lazydeltabase', True, b'try the provided delta first'), |
|
2777 | 2777 | (b'', b'clear-caches', True, b'clear revlog cache between calls'), |
|
2778 | 2778 | ], |
|
2779 | 2779 | b'-c|-m|FILE', |
|
2780 | 2780 | ) |
|
2781 | 2781 | def perfrevlogwrite(ui, repo, file_=None, startrev=1000, stoprev=-1, **opts): |
|
2782 | 2782 | """Benchmark writing a series of revisions to a revlog. |
|
2783 | 2783 | |
|
2784 | 2784 | Possible source values are: |
|
2785 | 2785 | * `full`: add from a full text (default). |
|
2786 | 2786 | * `parent-1`: add from a delta to the first parent |
|
2787 | 2787 | * `parent-2`: add from a delta to the second parent if it exists |
|
2788 | 2788 | (use a delta from the first parent otherwise) |
|
2789 | 2789 | * `parent-smallest`: add from the smallest delta (either p1 or p2) |
|
2790 | 2790 | * `storage`: add from the existing precomputed deltas |
|
2791 | 2791 | |
|
2792 | 2792 | Note: This performance command measures performance in a custom way. As a |
|
2793 | 2793 | result some of the global configuration of the 'perf' command does not |
|
2794 | 2794 | apply to it: |
|
2795 | 2795 | |
|
2796 | 2796 | * ``pre-run``: disabled |
|
2797 | 2797 | |
|
2798 | 2798 | * ``profile-benchmark``: disabled |
|
2799 | 2799 | |
|
2800 | 2800 | * ``run-limits``: disabled use --count instead |
|
2801 | 2801 | """ |
|
2802 | 2802 | opts = _byteskwargs(opts) |
|
2803 | 2803 | |
|
2804 | 2804 | rl = cmdutil.openrevlog(repo, b'perfrevlogwrite', file_, opts) |
|
2805 | 2805 | rllen = getlen(ui)(rl) |
|
2806 | 2806 | if startrev < 0: |
|
2807 | 2807 | startrev = rllen + startrev |
|
2808 | 2808 | if stoprev < 0: |
|
2809 | 2809 | stoprev = rllen + stoprev |
|
2810 | 2810 | |
|
2811 | 2811 | lazydeltabase = opts['lazydeltabase'] |
|
2812 | 2812 | source = opts['source'] |
|
2813 | 2813 | clearcaches = opts['clear_caches'] |
|
2814 | 2814 | validsource = ( |
|
2815 | 2815 | b'full', |
|
2816 | 2816 | b'parent-1', |
|
2817 | 2817 | b'parent-2', |
|
2818 | 2818 | b'parent-smallest', |
|
2819 | 2819 | b'storage', |
|
2820 | 2820 | ) |
|
2821 | 2821 | if source not in validsource: |
|
2822 | 2822 | raise error.Abort('invalid source type: %s' % source) |
|
2823 | 2823 | |
|
2824 | 2824 | ### actually gather results |
|
2825 | 2825 | count = opts['count'] |
|
2826 | 2826 | if count <= 0: |
|
2827 | 2827 | raise error.Abort('invalide run count: %d' % count) |
|
2828 | 2828 | allresults = [] |
|
2829 | 2829 | for c in range(count): |
|
2830 | 2830 | timing = _timeonewrite( |
|
2831 | 2831 | ui, |
|
2832 | 2832 | rl, |
|
2833 | 2833 | source, |
|
2834 | 2834 | startrev, |
|
2835 | 2835 | stoprev, |
|
2836 | 2836 | c + 1, |
|
2837 | 2837 | lazydeltabase=lazydeltabase, |
|
2838 | 2838 | clearcaches=clearcaches, |
|
2839 | 2839 | ) |
|
2840 | 2840 | allresults.append(timing) |
|
2841 | 2841 | |
|
2842 | 2842 | ### consolidate the results in a single list |
|
2843 | 2843 | results = [] |
|
2844 | 2844 | for idx, (rev, t) in enumerate(allresults[0]): |
|
2845 | 2845 | ts = [t] |
|
2846 | 2846 | for other in allresults[1:]: |
|
2847 | 2847 | orev, ot = other[idx] |
|
2848 | 2848 | assert orev == rev |
|
2849 | 2849 | ts.append(ot) |
|
2850 | 2850 | results.append((rev, ts)) |
|
2851 | 2851 | resultcount = len(results) |
|
2852 | 2852 | |
|
2853 | 2853 | ### Compute and display relevant statistics |
|
2854 | 2854 | |
|
2855 | 2855 | # get a formatter |
|
2856 | 2856 | fm = ui.formatter(b'perf', opts) |
|
2857 | 2857 | displayall = ui.configbool(b"perf", b"all-timing", False) |
|
2858 | 2858 | |
|
2859 | 2859 | # print individual details if requested |
|
2860 | 2860 | if opts['details']: |
|
2861 | 2861 | for idx, item in enumerate(results, 1): |
|
2862 | 2862 | rev, data = item |
|
2863 | 2863 | title = 'revisions #%d of %d, rev %d' % (idx, resultcount, rev) |
|
2864 | 2864 | formatone(fm, data, title=title, displayall=displayall) |
|
2865 | 2865 | |
|
2866 | 2866 | # sorts results by median time |
|
2867 | 2867 | results.sort(key=lambda x: sorted(x[1])[len(x[1]) // 2]) |
|
2868 | 2868 | # list of (name, index) to display) |
|
2869 | 2869 | relevants = [ |
|
2870 | 2870 | ("min", 0), |
|
2871 | 2871 | ("10%", resultcount * 10 // 100), |
|
2872 | 2872 | ("25%", resultcount * 25 // 100), |
|
2873 | 2873 | ("50%", resultcount * 70 // 100), |
|
2874 | 2874 | ("75%", resultcount * 75 // 100), |
|
2875 | 2875 | ("90%", resultcount * 90 // 100), |
|
2876 | 2876 | ("95%", resultcount * 95 // 100), |
|
2877 | 2877 | ("99%", resultcount * 99 // 100), |
|
2878 | 2878 | ("99.9%", resultcount * 999 // 1000), |
|
2879 | 2879 | ("99.99%", resultcount * 9999 // 10000), |
|
2880 | 2880 | ("99.999%", resultcount * 99999 // 100000), |
|
2881 | 2881 | ("max", -1), |
|
2882 | 2882 | ] |
|
2883 | 2883 | if not ui.quiet: |
|
2884 | 2884 | for name, idx in relevants: |
|
2885 | 2885 | data = results[idx] |
|
2886 | 2886 | title = '%s of %d, rev %d' % (name, resultcount, data[0]) |
|
2887 | 2887 | formatone(fm, data[1], title=title, displayall=displayall) |
|
2888 | 2888 | |
|
2889 | 2889 | # XXX summing that many float will not be very precise, we ignore this fact |
|
2890 | 2890 | # for now |
|
2891 | 2891 | totaltime = [] |
|
2892 | 2892 | for item in allresults: |
|
2893 | 2893 | totaltime.append( |
|
2894 | 2894 | ( |
|
2895 | 2895 | sum(x[1][0] for x in item), |
|
2896 | 2896 | sum(x[1][1] for x in item), |
|
2897 | 2897 | sum(x[1][2] for x in item), |
|
2898 | 2898 | ) |
|
2899 | 2899 | ) |
|
2900 | 2900 | formatone( |
|
2901 | 2901 | fm, |
|
2902 | 2902 | totaltime, |
|
2903 | 2903 | title="total time (%d revs)" % resultcount, |
|
2904 | 2904 | displayall=displayall, |
|
2905 | 2905 | ) |
|
2906 | 2906 | fm.end() |
|
2907 | 2907 | |
|
2908 | 2908 | |
|
2909 | 2909 | class _faketr(object): |
|
2910 | 2910 | def add(s, x, y, z=None): |
|
2911 | 2911 | return None |
|
2912 | 2912 | |
|
2913 | 2913 | |
|
2914 | 2914 | def _timeonewrite( |
|
2915 | 2915 | ui, |
|
2916 | 2916 | orig, |
|
2917 | 2917 | source, |
|
2918 | 2918 | startrev, |
|
2919 | 2919 | stoprev, |
|
2920 | 2920 | runidx=None, |
|
2921 | 2921 | lazydeltabase=True, |
|
2922 | 2922 | clearcaches=True, |
|
2923 | 2923 | ): |
|
2924 | 2924 | timings = [] |
|
2925 | 2925 | tr = _faketr() |
|
2926 | 2926 | with _temprevlog(ui, orig, startrev) as dest: |
|
2927 | 2927 | dest._lazydeltabase = lazydeltabase |
|
2928 | 2928 | revs = list(orig.revs(startrev, stoprev)) |
|
2929 | 2929 | total = len(revs) |
|
2930 | 2930 | topic = 'adding' |
|
2931 | 2931 | if runidx is not None: |
|
2932 | 2932 | topic += ' (run #%d)' % runidx |
|
2933 | 2933 | # Support both old and new progress API |
|
2934 | 2934 | if util.safehasattr(ui, 'makeprogress'): |
|
2935 | 2935 | progress = ui.makeprogress(topic, unit='revs', total=total) |
|
2936 | 2936 | |
|
2937 | 2937 | def updateprogress(pos): |
|
2938 | 2938 | progress.update(pos) |
|
2939 | 2939 | |
|
2940 | 2940 | def completeprogress(): |
|
2941 | 2941 | progress.complete() |
|
2942 | 2942 | |
|
2943 | 2943 | else: |
|
2944 | 2944 | |
|
2945 | 2945 | def updateprogress(pos): |
|
2946 | 2946 | ui.progress(topic, pos, unit='revs', total=total) |
|
2947 | 2947 | |
|
2948 | 2948 | def completeprogress(): |
|
2949 | 2949 | ui.progress(topic, None, unit='revs', total=total) |
|
2950 | 2950 | |
|
2951 | 2951 | for idx, rev in enumerate(revs): |
|
2952 | 2952 | updateprogress(idx) |
|
2953 | 2953 | addargs, addkwargs = _getrevisionseed(orig, rev, tr, source) |
|
2954 | 2954 | if clearcaches: |
|
2955 | 2955 | dest.index.clearcaches() |
|
2956 | 2956 | dest.clearcaches() |
|
2957 | 2957 | with timeone() as r: |
|
2958 | 2958 | dest.addrawrevision(*addargs, **addkwargs) |
|
2959 | 2959 | timings.append((rev, r[0])) |
|
2960 | 2960 | updateprogress(total) |
|
2961 | 2961 | completeprogress() |
|
2962 | 2962 | return timings |
|
2963 | 2963 | |
|
2964 | 2964 | |
|
2965 | 2965 | def _getrevisionseed(orig, rev, tr, source): |
|
2966 | 2966 | from mercurial.node import nullid |
|
2967 | 2967 | |
|
2968 | 2968 | linkrev = orig.linkrev(rev) |
|
2969 | 2969 | node = orig.node(rev) |
|
2970 | 2970 | p1, p2 = orig.parents(node) |
|
2971 | 2971 | flags = orig.flags(rev) |
|
2972 | 2972 | cachedelta = None |
|
2973 | 2973 | text = None |
|
2974 | 2974 | |
|
2975 | 2975 | if source == b'full': |
|
2976 | 2976 | text = orig.revision(rev) |
|
2977 | 2977 | elif source == b'parent-1': |
|
2978 | 2978 | baserev = orig.rev(p1) |
|
2979 | 2979 | cachedelta = (baserev, orig.revdiff(p1, rev)) |
|
2980 | 2980 | elif source == b'parent-2': |
|
2981 | 2981 | parent = p2 |
|
2982 | 2982 | if p2 == nullid: |
|
2983 | 2983 | parent = p1 |
|
2984 | 2984 | baserev = orig.rev(parent) |
|
2985 | 2985 | cachedelta = (baserev, orig.revdiff(parent, rev)) |
|
2986 | 2986 | elif source == b'parent-smallest': |
|
2987 | 2987 | p1diff = orig.revdiff(p1, rev) |
|
2988 | 2988 | parent = p1 |
|
2989 | 2989 | diff = p1diff |
|
2990 | 2990 | if p2 != nullid: |
|
2991 | 2991 | p2diff = orig.revdiff(p2, rev) |
|
2992 | 2992 | if len(p1diff) > len(p2diff): |
|
2993 | 2993 | parent = p2 |
|
2994 | 2994 | diff = p2diff |
|
2995 | 2995 | baserev = orig.rev(parent) |
|
2996 | 2996 | cachedelta = (baserev, diff) |
|
2997 | 2997 | elif source == b'storage': |
|
2998 | 2998 | baserev = orig.deltaparent(rev) |
|
2999 | 2999 | cachedelta = (baserev, orig.revdiff(orig.node(baserev), rev)) |
|
3000 | 3000 | |
|
3001 | 3001 | return ( |
|
3002 | 3002 | (text, tr, linkrev, p1, p2), |
|
3003 | 3003 | {'node': node, 'flags': flags, 'cachedelta': cachedelta}, |
|
3004 | 3004 | ) |
|
3005 | 3005 | |
|
3006 | 3006 | |
|
3007 | 3007 | @contextlib.contextmanager |
|
3008 | 3008 | def _temprevlog(ui, orig, truncaterev): |
|
3009 | 3009 | from mercurial import vfs as vfsmod |
|
3010 | 3010 | |
|
3011 | 3011 | if orig._inline: |
|
3012 | 3012 | raise error.Abort('not supporting inline revlog (yet)') |
|
3013 | 3013 | revlogkwargs = {} |
|
3014 | 3014 | k = 'upperboundcomp' |
|
3015 | 3015 | if util.safehasattr(orig, k): |
|
3016 | 3016 | revlogkwargs[k] = getattr(orig, k) |
|
3017 | 3017 | |
|
3018 | 3018 | origindexpath = orig.opener.join(orig.indexfile) |
|
3019 | 3019 | origdatapath = orig.opener.join(orig.datafile) |
|
3020 | 3020 | indexname = 'revlog.i' |
|
3021 | 3021 | dataname = 'revlog.d' |
|
3022 | 3022 | |
|
3023 | 3023 | tmpdir = tempfile.mkdtemp(prefix='tmp-hgperf-') |
|
3024 | 3024 | try: |
|
3025 | 3025 | # copy the data file in a temporary directory |
|
3026 | 3026 | ui.debug('copying data in %s\n' % tmpdir) |
|
3027 | 3027 | destindexpath = os.path.join(tmpdir, 'revlog.i') |
|
3028 | 3028 | destdatapath = os.path.join(tmpdir, 'revlog.d') |
|
3029 | 3029 | shutil.copyfile(origindexpath, destindexpath) |
|
3030 | 3030 | shutil.copyfile(origdatapath, destdatapath) |
|
3031 | 3031 | |
|
3032 | 3032 | # remove the data we want to add again |
|
3033 | 3033 | ui.debug('truncating data to be rewritten\n') |
|
3034 | 3034 | with open(destindexpath, 'ab') as index: |
|
3035 | 3035 | index.seek(0) |
|
3036 | 3036 | index.truncate(truncaterev * orig._io.size) |
|
3037 | 3037 | with open(destdatapath, 'ab') as data: |
|
3038 | 3038 | data.seek(0) |
|
3039 | 3039 | data.truncate(orig.start(truncaterev)) |
|
3040 | 3040 | |
|
3041 | 3041 | # instantiate a new revlog from the temporary copy |
|
3042 | 3042 | ui.debug('truncating adding to be rewritten\n') |
|
3043 | 3043 | vfs = vfsmod.vfs(tmpdir) |
|
3044 | 3044 | vfs.options = getattr(orig.opener, 'options', None) |
|
3045 | 3045 | |
|
3046 | 3046 | dest = revlog.revlog( |
|
3047 | 3047 | vfs, indexfile=indexname, datafile=dataname, **revlogkwargs |
|
3048 | 3048 | ) |
|
3049 | 3049 | if dest._inline: |
|
3050 | 3050 | raise error.Abort('not supporting inline revlog (yet)') |
|
3051 | 3051 | # make sure internals are initialized |
|
3052 | 3052 | dest.revision(len(dest) - 1) |
|
3053 | 3053 | yield dest |
|
3054 | 3054 | del dest, vfs |
|
3055 | 3055 | finally: |
|
3056 | 3056 | shutil.rmtree(tmpdir, True) |
|
3057 | 3057 | |
|
3058 | 3058 | |
|
3059 | 3059 | @command( |
|
3060 | 3060 | b'perf::revlogchunks|perfrevlogchunks', |
|
3061 | 3061 | revlogopts |
|
3062 | 3062 | + formatteropts |
|
3063 | 3063 | + [ |
|
3064 | 3064 | (b'e', b'engines', b'', b'compression engines to use'), |
|
3065 | 3065 | (b's', b'startrev', 0, b'revision to start at'), |
|
3066 | 3066 | ], |
|
3067 | 3067 | b'-c|-m|FILE', |
|
3068 | 3068 | ) |
|
3069 | 3069 | def perfrevlogchunks(ui, repo, file_=None, engines=None, startrev=0, **opts): |
|
3070 | 3070 | """Benchmark operations on revlog chunks. |
|
3071 | 3071 | |
|
3072 | 3072 | Logically, each revlog is a collection of fulltext revisions. However, |
|
3073 | 3073 | stored within each revlog are "chunks" of possibly compressed data. This |
|
3074 | 3074 | data needs to be read and decompressed or compressed and written. |
|
3075 | 3075 | |
|
3076 | 3076 | This command measures the time it takes to read+decompress and recompress |
|
3077 | 3077 | chunks in a revlog. It effectively isolates I/O and compression performance. |
|
3078 | 3078 | For measurements of higher-level operations like resolving revisions, |
|
3079 | 3079 | see ``perfrevlogrevisions`` and ``perfrevlogrevision``. |
|
3080 | 3080 | """ |
|
3081 | 3081 | opts = _byteskwargs(opts) |
|
3082 | 3082 | |
|
3083 | 3083 | rl = cmdutil.openrevlog(repo, b'perfrevlogchunks', file_, opts) |
|
3084 | 3084 | |
|
3085 | 3085 | # _chunkraw was renamed to _getsegmentforrevs. |
|
3086 | 3086 | try: |
|
3087 | 3087 | segmentforrevs = rl._getsegmentforrevs |
|
3088 | 3088 | except AttributeError: |
|
3089 | 3089 | segmentforrevs = rl._chunkraw |
|
3090 | 3090 | |
|
3091 | 3091 | # Verify engines argument. |
|
3092 | 3092 | if engines: |
|
3093 | 3093 | engines = {e.strip() for e in engines.split(b',')} |
|
3094 | 3094 | for engine in engines: |
|
3095 | 3095 | try: |
|
3096 | 3096 | util.compressionengines[engine] |
|
3097 | 3097 | except KeyError: |
|
3098 | 3098 | raise error.Abort(b'unknown compression engine: %s' % engine) |
|
3099 | 3099 | else: |
|
3100 | 3100 | engines = [] |
|
3101 | 3101 | for e in util.compengines: |
|
3102 | 3102 | engine = util.compengines[e] |
|
3103 | 3103 | try: |
|
3104 | 3104 | if engine.available(): |
|
3105 | 3105 | engine.revlogcompressor().compress(b'dummy') |
|
3106 | 3106 | engines.append(e) |
|
3107 | 3107 | except NotImplementedError: |
|
3108 | 3108 | pass |
|
3109 | 3109 | |
|
3110 | 3110 | revs = list(rl.revs(startrev, len(rl) - 1)) |
|
3111 | 3111 | |
|
3112 | 3112 | def rlfh(rl): |
|
3113 | 3113 | if rl._inline: |
|
3114 | 3114 | return getsvfs(repo)(rl.indexfile) |
|
3115 | 3115 | else: |
|
3116 | 3116 | return getsvfs(repo)(rl.datafile) |
|
3117 | 3117 | |
|
3118 | 3118 | def doread(): |
|
3119 | 3119 | rl.clearcaches() |
|
3120 | 3120 | for rev in revs: |
|
3121 | 3121 | segmentforrevs(rev, rev) |
|
3122 | 3122 | |
|
3123 | 3123 | def doreadcachedfh(): |
|
3124 | 3124 | rl.clearcaches() |
|
3125 | 3125 | fh = rlfh(rl) |
|
3126 | 3126 | for rev in revs: |
|
3127 | 3127 | segmentforrevs(rev, rev, df=fh) |
|
3128 | 3128 | |
|
3129 | 3129 | def doreadbatch(): |
|
3130 | 3130 | rl.clearcaches() |
|
3131 | 3131 | segmentforrevs(revs[0], revs[-1]) |
|
3132 | 3132 | |
|
3133 | 3133 | def doreadbatchcachedfh(): |
|
3134 | 3134 | rl.clearcaches() |
|
3135 | 3135 | fh = rlfh(rl) |
|
3136 | 3136 | segmentforrevs(revs[0], revs[-1], df=fh) |
|
3137 | 3137 | |
|
3138 | 3138 | def dochunk(): |
|
3139 | 3139 | rl.clearcaches() |
|
3140 | 3140 | fh = rlfh(rl) |
|
3141 | 3141 | for rev in revs: |
|
3142 | 3142 | rl._chunk(rev, df=fh) |
|
3143 | 3143 | |
|
3144 | 3144 | chunks = [None] |
|
3145 | 3145 | |
|
3146 | 3146 | def dochunkbatch(): |
|
3147 | 3147 | rl.clearcaches() |
|
3148 | 3148 | fh = rlfh(rl) |
|
3149 | 3149 | # Save chunks as a side-effect. |
|
3150 | 3150 | chunks[0] = rl._chunks(revs, df=fh) |
|
3151 | 3151 | |
|
3152 | 3152 | def docompress(compressor): |
|
3153 | 3153 | rl.clearcaches() |
|
3154 | 3154 | |
|
3155 | 3155 | try: |
|
3156 | 3156 | # Swap in the requested compression engine. |
|
3157 | 3157 | oldcompressor = rl._compressor |
|
3158 | 3158 | rl._compressor = compressor |
|
3159 | 3159 | for chunk in chunks[0]: |
|
3160 | 3160 | rl.compress(chunk) |
|
3161 | 3161 | finally: |
|
3162 | 3162 | rl._compressor = oldcompressor |
|
3163 | 3163 | |
|
3164 | 3164 | benches = [ |
|
3165 | 3165 | (lambda: doread(), b'read'), |
|
3166 | 3166 | (lambda: doreadcachedfh(), b'read w/ reused fd'), |
|
3167 | 3167 | (lambda: doreadbatch(), b'read batch'), |
|
3168 | 3168 | (lambda: doreadbatchcachedfh(), b'read batch w/ reused fd'), |
|
3169 | 3169 | (lambda: dochunk(), b'chunk'), |
|
3170 | 3170 | (lambda: dochunkbatch(), b'chunk batch'), |
|
3171 | 3171 | ] |
|
3172 | 3172 | |
|
3173 | 3173 | for engine in sorted(engines): |
|
3174 | 3174 | compressor = util.compengines[engine].revlogcompressor() |
|
3175 | 3175 | benches.append( |
|
3176 | 3176 | ( |
|
3177 | 3177 | functools.partial(docompress, compressor), |
|
3178 | 3178 | b'compress w/ %s' % engine, |
|
3179 | 3179 | ) |
|
3180 | 3180 | ) |
|
3181 | 3181 | |
|
3182 | 3182 | for fn, title in benches: |
|
3183 | 3183 | timer, fm = gettimer(ui, opts) |
|
3184 | 3184 | timer(fn, title=title) |
|
3185 | 3185 | fm.end() |
|
3186 | 3186 | |
|
3187 | 3187 | |
|
3188 | 3188 | @command( |
|
3189 | 3189 | b'perf::revlogrevision|perfrevlogrevision', |
|
3190 | 3190 | revlogopts |
|
3191 | 3191 | + formatteropts |
|
3192 | 3192 | + [(b'', b'cache', False, b'use caches instead of clearing')], |
|
3193 | 3193 | b'-c|-m|FILE REV', |
|
3194 | 3194 | ) |
|
3195 | 3195 | def perfrevlogrevision(ui, repo, file_, rev=None, cache=None, **opts): |
|
3196 | 3196 | """Benchmark obtaining a revlog revision. |
|
3197 | 3197 | |
|
3198 | 3198 | Obtaining a revlog revision consists of roughly the following steps: |
|
3199 | 3199 | |
|
3200 | 3200 | 1. Compute the delta chain |
|
3201 | 3201 | 2. Slice the delta chain if applicable |
|
3202 | 3202 | 3. Obtain the raw chunks for that delta chain |
|
3203 | 3203 | 4. Decompress each raw chunk |
|
3204 | 3204 | 5. Apply binary patches to obtain fulltext |
|
3205 | 3205 | 6. Verify hash of fulltext |
|
3206 | 3206 | |
|
3207 | 3207 | This command measures the time spent in each of these phases. |
|
3208 | 3208 | """ |
|
3209 | 3209 | opts = _byteskwargs(opts) |
|
3210 | 3210 | |
|
3211 | 3211 | if opts.get(b'changelog') or opts.get(b'manifest'): |
|
3212 | 3212 | file_, rev = None, file_ |
|
3213 | 3213 | elif rev is None: |
|
3214 | 3214 | raise error.CommandError(b'perfrevlogrevision', b'invalid arguments') |
|
3215 | 3215 | |
|
3216 | 3216 | r = cmdutil.openrevlog(repo, b'perfrevlogrevision', file_, opts) |
|
3217 | 3217 | |
|
3218 | 3218 | # _chunkraw was renamed to _getsegmentforrevs. |
|
3219 | 3219 | try: |
|
3220 | 3220 | segmentforrevs = r._getsegmentforrevs |
|
3221 | 3221 | except AttributeError: |
|
3222 | 3222 | segmentforrevs = r._chunkraw |
|
3223 | 3223 | |
|
3224 | 3224 | node = r.lookup(rev) |
|
3225 | 3225 | rev = r.rev(node) |
|
3226 | 3226 | |
|
3227 | 3227 | def getrawchunks(data, chain): |
|
3228 | 3228 | start = r.start |
|
3229 | 3229 | length = r.length |
|
3230 | 3230 | inline = r._inline |
|
3231 | iosize = r._io.size | |
|
3231 | try: | |
|
3232 | iosize = r.index.entry_size | |
|
3233 | except AttributeError: | |
|
3234 | iosize = r._io.size | |
|
3232 | 3235 | buffer = util.buffer |
|
3233 | 3236 | |
|
3234 | 3237 | chunks = [] |
|
3235 | 3238 | ladd = chunks.append |
|
3236 | 3239 | for idx, item in enumerate(chain): |
|
3237 | 3240 | offset = start(item[0]) |
|
3238 | 3241 | bits = data[idx] |
|
3239 | 3242 | for rev in item: |
|
3240 | 3243 | chunkstart = start(rev) |
|
3241 | 3244 | if inline: |
|
3242 | 3245 | chunkstart += (rev + 1) * iosize |
|
3243 | 3246 | chunklength = length(rev) |
|
3244 | 3247 | ladd(buffer(bits, chunkstart - offset, chunklength)) |
|
3245 | 3248 | |
|
3246 | 3249 | return chunks |
|
3247 | 3250 | |
|
3248 | 3251 | def dodeltachain(rev): |
|
3249 | 3252 | if not cache: |
|
3250 | 3253 | r.clearcaches() |
|
3251 | 3254 | r._deltachain(rev) |
|
3252 | 3255 | |
|
3253 | 3256 | def doread(chain): |
|
3254 | 3257 | if not cache: |
|
3255 | 3258 | r.clearcaches() |
|
3256 | 3259 | for item in slicedchain: |
|
3257 | 3260 | segmentforrevs(item[0], item[-1]) |
|
3258 | 3261 | |
|
3259 | 3262 | def doslice(r, chain, size): |
|
3260 | 3263 | for s in slicechunk(r, chain, targetsize=size): |
|
3261 | 3264 | pass |
|
3262 | 3265 | |
|
3263 | 3266 | def dorawchunks(data, chain): |
|
3264 | 3267 | if not cache: |
|
3265 | 3268 | r.clearcaches() |
|
3266 | 3269 | getrawchunks(data, chain) |
|
3267 | 3270 | |
|
3268 | 3271 | def dodecompress(chunks): |
|
3269 | 3272 | decomp = r.decompress |
|
3270 | 3273 | for chunk in chunks: |
|
3271 | 3274 | decomp(chunk) |
|
3272 | 3275 | |
|
3273 | 3276 | def dopatch(text, bins): |
|
3274 | 3277 | if not cache: |
|
3275 | 3278 | r.clearcaches() |
|
3276 | 3279 | mdiff.patches(text, bins) |
|
3277 | 3280 | |
|
3278 | 3281 | def dohash(text): |
|
3279 | 3282 | if not cache: |
|
3280 | 3283 | r.clearcaches() |
|
3281 | 3284 | r.checkhash(text, node, rev=rev) |
|
3282 | 3285 | |
|
3283 | 3286 | def dorevision(): |
|
3284 | 3287 | if not cache: |
|
3285 | 3288 | r.clearcaches() |
|
3286 | 3289 | r.revision(node) |
|
3287 | 3290 | |
|
3288 | 3291 | try: |
|
3289 | 3292 | from mercurial.revlogutils.deltas import slicechunk |
|
3290 | 3293 | except ImportError: |
|
3291 | 3294 | slicechunk = getattr(revlog, '_slicechunk', None) |
|
3292 | 3295 | |
|
3293 | 3296 | size = r.length(rev) |
|
3294 | 3297 | chain = r._deltachain(rev)[0] |
|
3295 | 3298 | if not getattr(r, '_withsparseread', False): |
|
3296 | 3299 | slicedchain = (chain,) |
|
3297 | 3300 | else: |
|
3298 | 3301 | slicedchain = tuple(slicechunk(r, chain, targetsize=size)) |
|
3299 | 3302 | data = [segmentforrevs(seg[0], seg[-1])[1] for seg in slicedchain] |
|
3300 | 3303 | rawchunks = getrawchunks(data, slicedchain) |
|
3301 | 3304 | bins = r._chunks(chain) |
|
3302 | 3305 | text = bytes(bins[0]) |
|
3303 | 3306 | bins = bins[1:] |
|
3304 | 3307 | text = mdiff.patches(text, bins) |
|
3305 | 3308 | |
|
3306 | 3309 | benches = [ |
|
3307 | 3310 | (lambda: dorevision(), b'full'), |
|
3308 | 3311 | (lambda: dodeltachain(rev), b'deltachain'), |
|
3309 | 3312 | (lambda: doread(chain), b'read'), |
|
3310 | 3313 | ] |
|
3311 | 3314 | |
|
3312 | 3315 | if getattr(r, '_withsparseread', False): |
|
3313 | 3316 | slicing = (lambda: doslice(r, chain, size), b'slice-sparse-chain') |
|
3314 | 3317 | benches.append(slicing) |
|
3315 | 3318 | |
|
3316 | 3319 | benches.extend( |
|
3317 | 3320 | [ |
|
3318 | 3321 | (lambda: dorawchunks(data, slicedchain), b'rawchunks'), |
|
3319 | 3322 | (lambda: dodecompress(rawchunks), b'decompress'), |
|
3320 | 3323 | (lambda: dopatch(text, bins), b'patch'), |
|
3321 | 3324 | (lambda: dohash(text), b'hash'), |
|
3322 | 3325 | ] |
|
3323 | 3326 | ) |
|
3324 | 3327 | |
|
3325 | 3328 | timer, fm = gettimer(ui, opts) |
|
3326 | 3329 | for fn, title in benches: |
|
3327 | 3330 | timer(fn, title=title) |
|
3328 | 3331 | fm.end() |
|
3329 | 3332 | |
|
3330 | 3333 | |
|
3331 | 3334 | @command( |
|
3332 | 3335 | b'perf::revset|perfrevset', |
|
3333 | 3336 | [ |
|
3334 | 3337 | (b'C', b'clear', False, b'clear volatile cache between each call.'), |
|
3335 | 3338 | (b'', b'contexts', False, b'obtain changectx for each revision'), |
|
3336 | 3339 | ] |
|
3337 | 3340 | + formatteropts, |
|
3338 | 3341 | b"REVSET", |
|
3339 | 3342 | ) |
|
3340 | 3343 | def perfrevset(ui, repo, expr, clear=False, contexts=False, **opts): |
|
3341 | 3344 | """benchmark the execution time of a revset |
|
3342 | 3345 | |
|
3343 | 3346 | Use the --clean option if need to evaluate the impact of build volatile |
|
3344 | 3347 | revisions set cache on the revset execution. Volatile cache hold filtered |
|
3345 | 3348 | and obsolete related cache.""" |
|
3346 | 3349 | opts = _byteskwargs(opts) |
|
3347 | 3350 | |
|
3348 | 3351 | timer, fm = gettimer(ui, opts) |
|
3349 | 3352 | |
|
3350 | 3353 | def d(): |
|
3351 | 3354 | if clear: |
|
3352 | 3355 | repo.invalidatevolatilesets() |
|
3353 | 3356 | if contexts: |
|
3354 | 3357 | for ctx in repo.set(expr): |
|
3355 | 3358 | pass |
|
3356 | 3359 | else: |
|
3357 | 3360 | for r in repo.revs(expr): |
|
3358 | 3361 | pass |
|
3359 | 3362 | |
|
3360 | 3363 | timer(d) |
|
3361 | 3364 | fm.end() |
|
3362 | 3365 | |
|
3363 | 3366 | |
|
3364 | 3367 | @command( |
|
3365 | 3368 | b'perf::volatilesets|perfvolatilesets', |
|
3366 | 3369 | [ |
|
3367 | 3370 | (b'', b'clear-obsstore', False, b'drop obsstore between each call.'), |
|
3368 | 3371 | ] |
|
3369 | 3372 | + formatteropts, |
|
3370 | 3373 | ) |
|
3371 | 3374 | def perfvolatilesets(ui, repo, *names, **opts): |
|
3372 | 3375 | """benchmark the computation of various volatile set |
|
3373 | 3376 | |
|
3374 | 3377 | Volatile set computes element related to filtering and obsolescence.""" |
|
3375 | 3378 | opts = _byteskwargs(opts) |
|
3376 | 3379 | timer, fm = gettimer(ui, opts) |
|
3377 | 3380 | repo = repo.unfiltered() |
|
3378 | 3381 | |
|
3379 | 3382 | def getobs(name): |
|
3380 | 3383 | def d(): |
|
3381 | 3384 | repo.invalidatevolatilesets() |
|
3382 | 3385 | if opts[b'clear_obsstore']: |
|
3383 | 3386 | clearfilecache(repo, b'obsstore') |
|
3384 | 3387 | obsolete.getrevs(repo, name) |
|
3385 | 3388 | |
|
3386 | 3389 | return d |
|
3387 | 3390 | |
|
3388 | 3391 | allobs = sorted(obsolete.cachefuncs) |
|
3389 | 3392 | if names: |
|
3390 | 3393 | allobs = [n for n in allobs if n in names] |
|
3391 | 3394 | |
|
3392 | 3395 | for name in allobs: |
|
3393 | 3396 | timer(getobs(name), title=name) |
|
3394 | 3397 | |
|
3395 | 3398 | def getfiltered(name): |
|
3396 | 3399 | def d(): |
|
3397 | 3400 | repo.invalidatevolatilesets() |
|
3398 | 3401 | if opts[b'clear_obsstore']: |
|
3399 | 3402 | clearfilecache(repo, b'obsstore') |
|
3400 | 3403 | repoview.filterrevs(repo, name) |
|
3401 | 3404 | |
|
3402 | 3405 | return d |
|
3403 | 3406 | |
|
3404 | 3407 | allfilter = sorted(repoview.filtertable) |
|
3405 | 3408 | if names: |
|
3406 | 3409 | allfilter = [n for n in allfilter if n in names] |
|
3407 | 3410 | |
|
3408 | 3411 | for name in allfilter: |
|
3409 | 3412 | timer(getfiltered(name), title=name) |
|
3410 | 3413 | fm.end() |
|
3411 | 3414 | |
|
3412 | 3415 | |
|
3413 | 3416 | @command( |
|
3414 | 3417 | b'perf::branchmap|perfbranchmap', |
|
3415 | 3418 | [ |
|
3416 | 3419 | (b'f', b'full', False, b'Includes build time of subset'), |
|
3417 | 3420 | ( |
|
3418 | 3421 | b'', |
|
3419 | 3422 | b'clear-revbranch', |
|
3420 | 3423 | False, |
|
3421 | 3424 | b'purge the revbranch cache between computation', |
|
3422 | 3425 | ), |
|
3423 | 3426 | ] |
|
3424 | 3427 | + formatteropts, |
|
3425 | 3428 | ) |
|
3426 | 3429 | def perfbranchmap(ui, repo, *filternames, **opts): |
|
3427 | 3430 | """benchmark the update of a branchmap |
|
3428 | 3431 | |
|
3429 | 3432 | This benchmarks the full repo.branchmap() call with read and write disabled |
|
3430 | 3433 | """ |
|
3431 | 3434 | opts = _byteskwargs(opts) |
|
3432 | 3435 | full = opts.get(b"full", False) |
|
3433 | 3436 | clear_revbranch = opts.get(b"clear_revbranch", False) |
|
3434 | 3437 | timer, fm = gettimer(ui, opts) |
|
3435 | 3438 | |
|
3436 | 3439 | def getbranchmap(filtername): |
|
3437 | 3440 | """generate a benchmark function for the filtername""" |
|
3438 | 3441 | if filtername is None: |
|
3439 | 3442 | view = repo |
|
3440 | 3443 | else: |
|
3441 | 3444 | view = repo.filtered(filtername) |
|
3442 | 3445 | if util.safehasattr(view._branchcaches, '_per_filter'): |
|
3443 | 3446 | filtered = view._branchcaches._per_filter |
|
3444 | 3447 | else: |
|
3445 | 3448 | # older versions |
|
3446 | 3449 | filtered = view._branchcaches |
|
3447 | 3450 | |
|
3448 | 3451 | def d(): |
|
3449 | 3452 | if clear_revbranch: |
|
3450 | 3453 | repo.revbranchcache()._clear() |
|
3451 | 3454 | if full: |
|
3452 | 3455 | view._branchcaches.clear() |
|
3453 | 3456 | else: |
|
3454 | 3457 | filtered.pop(filtername, None) |
|
3455 | 3458 | view.branchmap() |
|
3456 | 3459 | |
|
3457 | 3460 | return d |
|
3458 | 3461 | |
|
3459 | 3462 | # add filter in smaller subset to bigger subset |
|
3460 | 3463 | possiblefilters = set(repoview.filtertable) |
|
3461 | 3464 | if filternames: |
|
3462 | 3465 | possiblefilters &= set(filternames) |
|
3463 | 3466 | subsettable = getbranchmapsubsettable() |
|
3464 | 3467 | allfilters = [] |
|
3465 | 3468 | while possiblefilters: |
|
3466 | 3469 | for name in possiblefilters: |
|
3467 | 3470 | subset = subsettable.get(name) |
|
3468 | 3471 | if subset not in possiblefilters: |
|
3469 | 3472 | break |
|
3470 | 3473 | else: |
|
3471 | 3474 | assert False, b'subset cycle %s!' % possiblefilters |
|
3472 | 3475 | allfilters.append(name) |
|
3473 | 3476 | possiblefilters.remove(name) |
|
3474 | 3477 | |
|
3475 | 3478 | # warm the cache |
|
3476 | 3479 | if not full: |
|
3477 | 3480 | for name in allfilters: |
|
3478 | 3481 | repo.filtered(name).branchmap() |
|
3479 | 3482 | if not filternames or b'unfiltered' in filternames: |
|
3480 | 3483 | # add unfiltered |
|
3481 | 3484 | allfilters.append(None) |
|
3482 | 3485 | |
|
3483 | 3486 | if util.safehasattr(branchmap.branchcache, 'fromfile'): |
|
3484 | 3487 | branchcacheread = safeattrsetter(branchmap.branchcache, b'fromfile') |
|
3485 | 3488 | branchcacheread.set(classmethod(lambda *args: None)) |
|
3486 | 3489 | else: |
|
3487 | 3490 | # older versions |
|
3488 | 3491 | branchcacheread = safeattrsetter(branchmap, b'read') |
|
3489 | 3492 | branchcacheread.set(lambda *args: None) |
|
3490 | 3493 | branchcachewrite = safeattrsetter(branchmap.branchcache, b'write') |
|
3491 | 3494 | branchcachewrite.set(lambda *args: None) |
|
3492 | 3495 | try: |
|
3493 | 3496 | for name in allfilters: |
|
3494 | 3497 | printname = name |
|
3495 | 3498 | if name is None: |
|
3496 | 3499 | printname = b'unfiltered' |
|
3497 | 3500 | timer(getbranchmap(name), title=printname) |
|
3498 | 3501 | finally: |
|
3499 | 3502 | branchcacheread.restore() |
|
3500 | 3503 | branchcachewrite.restore() |
|
3501 | 3504 | fm.end() |
|
3502 | 3505 | |
|
3503 | 3506 | |
|
3504 | 3507 | @command( |
|
3505 | 3508 | b'perf::branchmapupdate|perfbranchmapupdate', |
|
3506 | 3509 | [ |
|
3507 | 3510 | (b'', b'base', [], b'subset of revision to start from'), |
|
3508 | 3511 | (b'', b'target', [], b'subset of revision to end with'), |
|
3509 | 3512 | (b'', b'clear-caches', False, b'clear cache between each runs'), |
|
3510 | 3513 | ] |
|
3511 | 3514 | + formatteropts, |
|
3512 | 3515 | ) |
|
3513 | 3516 | def perfbranchmapupdate(ui, repo, base=(), target=(), **opts): |
|
3514 | 3517 | """benchmark branchmap update from for <base> revs to <target> revs |
|
3515 | 3518 | |
|
3516 | 3519 | If `--clear-caches` is passed, the following items will be reset before |
|
3517 | 3520 | each update: |
|
3518 | 3521 | * the changelog instance and associated indexes |
|
3519 | 3522 | * the rev-branch-cache instance |
|
3520 | 3523 | |
|
3521 | 3524 | Examples: |
|
3522 | 3525 | |
|
3523 | 3526 | # update for the one last revision |
|
3524 | 3527 | $ hg perfbranchmapupdate --base 'not tip' --target 'tip' |
|
3525 | 3528 | |
|
3526 | 3529 | $ update for change coming with a new branch |
|
3527 | 3530 | $ hg perfbranchmapupdate --base 'stable' --target 'default' |
|
3528 | 3531 | """ |
|
3529 | 3532 | from mercurial import branchmap |
|
3530 | 3533 | from mercurial import repoview |
|
3531 | 3534 | |
|
3532 | 3535 | opts = _byteskwargs(opts) |
|
3533 | 3536 | timer, fm = gettimer(ui, opts) |
|
3534 | 3537 | clearcaches = opts[b'clear_caches'] |
|
3535 | 3538 | unfi = repo.unfiltered() |
|
3536 | 3539 | x = [None] # used to pass data between closure |
|
3537 | 3540 | |
|
3538 | 3541 | # we use a `list` here to avoid possible side effect from smartset |
|
3539 | 3542 | baserevs = list(scmutil.revrange(repo, base)) |
|
3540 | 3543 | targetrevs = list(scmutil.revrange(repo, target)) |
|
3541 | 3544 | if not baserevs: |
|
3542 | 3545 | raise error.Abort(b'no revisions selected for --base') |
|
3543 | 3546 | if not targetrevs: |
|
3544 | 3547 | raise error.Abort(b'no revisions selected for --target') |
|
3545 | 3548 | |
|
3546 | 3549 | # make sure the target branchmap also contains the one in the base |
|
3547 | 3550 | targetrevs = list(set(baserevs) | set(targetrevs)) |
|
3548 | 3551 | targetrevs.sort() |
|
3549 | 3552 | |
|
3550 | 3553 | cl = repo.changelog |
|
3551 | 3554 | allbaserevs = list(cl.ancestors(baserevs, inclusive=True)) |
|
3552 | 3555 | allbaserevs.sort() |
|
3553 | 3556 | alltargetrevs = frozenset(cl.ancestors(targetrevs, inclusive=True)) |
|
3554 | 3557 | |
|
3555 | 3558 | newrevs = list(alltargetrevs.difference(allbaserevs)) |
|
3556 | 3559 | newrevs.sort() |
|
3557 | 3560 | |
|
3558 | 3561 | allrevs = frozenset(unfi.changelog.revs()) |
|
3559 | 3562 | basefilterrevs = frozenset(allrevs.difference(allbaserevs)) |
|
3560 | 3563 | targetfilterrevs = frozenset(allrevs.difference(alltargetrevs)) |
|
3561 | 3564 | |
|
3562 | 3565 | def basefilter(repo, visibilityexceptions=None): |
|
3563 | 3566 | return basefilterrevs |
|
3564 | 3567 | |
|
3565 | 3568 | def targetfilter(repo, visibilityexceptions=None): |
|
3566 | 3569 | return targetfilterrevs |
|
3567 | 3570 | |
|
3568 | 3571 | msg = b'benchmark of branchmap with %d revisions with %d new ones\n' |
|
3569 | 3572 | ui.status(msg % (len(allbaserevs), len(newrevs))) |
|
3570 | 3573 | if targetfilterrevs: |
|
3571 | 3574 | msg = b'(%d revisions still filtered)\n' |
|
3572 | 3575 | ui.status(msg % len(targetfilterrevs)) |
|
3573 | 3576 | |
|
3574 | 3577 | try: |
|
3575 | 3578 | repoview.filtertable[b'__perf_branchmap_update_base'] = basefilter |
|
3576 | 3579 | repoview.filtertable[b'__perf_branchmap_update_target'] = targetfilter |
|
3577 | 3580 | |
|
3578 | 3581 | baserepo = repo.filtered(b'__perf_branchmap_update_base') |
|
3579 | 3582 | targetrepo = repo.filtered(b'__perf_branchmap_update_target') |
|
3580 | 3583 | |
|
3581 | 3584 | # try to find an existing branchmap to reuse |
|
3582 | 3585 | subsettable = getbranchmapsubsettable() |
|
3583 | 3586 | candidatefilter = subsettable.get(None) |
|
3584 | 3587 | while candidatefilter is not None: |
|
3585 | 3588 | candidatebm = repo.filtered(candidatefilter).branchmap() |
|
3586 | 3589 | if candidatebm.validfor(baserepo): |
|
3587 | 3590 | filtered = repoview.filterrevs(repo, candidatefilter) |
|
3588 | 3591 | missing = [r for r in allbaserevs if r in filtered] |
|
3589 | 3592 | base = candidatebm.copy() |
|
3590 | 3593 | base.update(baserepo, missing) |
|
3591 | 3594 | break |
|
3592 | 3595 | candidatefilter = subsettable.get(candidatefilter) |
|
3593 | 3596 | else: |
|
3594 | 3597 | # no suitable subset where found |
|
3595 | 3598 | base = branchmap.branchcache() |
|
3596 | 3599 | base.update(baserepo, allbaserevs) |
|
3597 | 3600 | |
|
3598 | 3601 | def setup(): |
|
3599 | 3602 | x[0] = base.copy() |
|
3600 | 3603 | if clearcaches: |
|
3601 | 3604 | unfi._revbranchcache = None |
|
3602 | 3605 | clearchangelog(repo) |
|
3603 | 3606 | |
|
3604 | 3607 | def bench(): |
|
3605 | 3608 | x[0].update(targetrepo, newrevs) |
|
3606 | 3609 | |
|
3607 | 3610 | timer(bench, setup=setup) |
|
3608 | 3611 | fm.end() |
|
3609 | 3612 | finally: |
|
3610 | 3613 | repoview.filtertable.pop(b'__perf_branchmap_update_base', None) |
|
3611 | 3614 | repoview.filtertable.pop(b'__perf_branchmap_update_target', None) |
|
3612 | 3615 | |
|
3613 | 3616 | |
|
3614 | 3617 | @command( |
|
3615 | 3618 | b'perf::branchmapload|perfbranchmapload', |
|
3616 | 3619 | [ |
|
3617 | 3620 | (b'f', b'filter', b'', b'Specify repoview filter'), |
|
3618 | 3621 | (b'', b'list', False, b'List brachmap filter caches'), |
|
3619 | 3622 | (b'', b'clear-revlogs', False, b'refresh changelog and manifest'), |
|
3620 | 3623 | ] |
|
3621 | 3624 | + formatteropts, |
|
3622 | 3625 | ) |
|
3623 | 3626 | def perfbranchmapload(ui, repo, filter=b'', list=False, **opts): |
|
3624 | 3627 | """benchmark reading the branchmap""" |
|
3625 | 3628 | opts = _byteskwargs(opts) |
|
3626 | 3629 | clearrevlogs = opts[b'clear_revlogs'] |
|
3627 | 3630 | |
|
3628 | 3631 | if list: |
|
3629 | 3632 | for name, kind, st in repo.cachevfs.readdir(stat=True): |
|
3630 | 3633 | if name.startswith(b'branch2'): |
|
3631 | 3634 | filtername = name.partition(b'-')[2] or b'unfiltered' |
|
3632 | 3635 | ui.status( |
|
3633 | 3636 | b'%s - %s\n' % (filtername, util.bytecount(st.st_size)) |
|
3634 | 3637 | ) |
|
3635 | 3638 | return |
|
3636 | 3639 | if not filter: |
|
3637 | 3640 | filter = None |
|
3638 | 3641 | subsettable = getbranchmapsubsettable() |
|
3639 | 3642 | if filter is None: |
|
3640 | 3643 | repo = repo.unfiltered() |
|
3641 | 3644 | else: |
|
3642 | 3645 | repo = repoview.repoview(repo, filter) |
|
3643 | 3646 | |
|
3644 | 3647 | repo.branchmap() # make sure we have a relevant, up to date branchmap |
|
3645 | 3648 | |
|
3646 | 3649 | try: |
|
3647 | 3650 | fromfile = branchmap.branchcache.fromfile |
|
3648 | 3651 | except AttributeError: |
|
3649 | 3652 | # older versions |
|
3650 | 3653 | fromfile = branchmap.read |
|
3651 | 3654 | |
|
3652 | 3655 | currentfilter = filter |
|
3653 | 3656 | # try once without timer, the filter may not be cached |
|
3654 | 3657 | while fromfile(repo) is None: |
|
3655 | 3658 | currentfilter = subsettable.get(currentfilter) |
|
3656 | 3659 | if currentfilter is None: |
|
3657 | 3660 | raise error.Abort( |
|
3658 | 3661 | b'No branchmap cached for %s repo' % (filter or b'unfiltered') |
|
3659 | 3662 | ) |
|
3660 | 3663 | repo = repo.filtered(currentfilter) |
|
3661 | 3664 | timer, fm = gettimer(ui, opts) |
|
3662 | 3665 | |
|
3663 | 3666 | def setup(): |
|
3664 | 3667 | if clearrevlogs: |
|
3665 | 3668 | clearchangelog(repo) |
|
3666 | 3669 | |
|
3667 | 3670 | def bench(): |
|
3668 | 3671 | fromfile(repo) |
|
3669 | 3672 | |
|
3670 | 3673 | timer(bench, setup=setup) |
|
3671 | 3674 | fm.end() |
|
3672 | 3675 | |
|
3673 | 3676 | |
|
3674 | 3677 | @command(b'perf::loadmarkers|perfloadmarkers') |
|
3675 | 3678 | def perfloadmarkers(ui, repo): |
|
3676 | 3679 | """benchmark the time to parse the on-disk markers for a repo |
|
3677 | 3680 | |
|
3678 | 3681 | Result is the number of markers in the repo.""" |
|
3679 | 3682 | timer, fm = gettimer(ui) |
|
3680 | 3683 | svfs = getsvfs(repo) |
|
3681 | 3684 | timer(lambda: len(obsolete.obsstore(repo, svfs))) |
|
3682 | 3685 | fm.end() |
|
3683 | 3686 | |
|
3684 | 3687 | |
|
3685 | 3688 | @command( |
|
3686 | 3689 | b'perf::lrucachedict|perflrucachedict', |
|
3687 | 3690 | formatteropts |
|
3688 | 3691 | + [ |
|
3689 | 3692 | (b'', b'costlimit', 0, b'maximum total cost of items in cache'), |
|
3690 | 3693 | (b'', b'mincost', 0, b'smallest cost of items in cache'), |
|
3691 | 3694 | (b'', b'maxcost', 100, b'maximum cost of items in cache'), |
|
3692 | 3695 | (b'', b'size', 4, b'size of cache'), |
|
3693 | 3696 | (b'', b'gets', 10000, b'number of key lookups'), |
|
3694 | 3697 | (b'', b'sets', 10000, b'number of key sets'), |
|
3695 | 3698 | (b'', b'mixed', 10000, b'number of mixed mode operations'), |
|
3696 | 3699 | ( |
|
3697 | 3700 | b'', |
|
3698 | 3701 | b'mixedgetfreq', |
|
3699 | 3702 | 50, |
|
3700 | 3703 | b'frequency of get vs set ops in mixed mode', |
|
3701 | 3704 | ), |
|
3702 | 3705 | ], |
|
3703 | 3706 | norepo=True, |
|
3704 | 3707 | ) |
|
3705 | 3708 | def perflrucache( |
|
3706 | 3709 | ui, |
|
3707 | 3710 | mincost=0, |
|
3708 | 3711 | maxcost=100, |
|
3709 | 3712 | costlimit=0, |
|
3710 | 3713 | size=4, |
|
3711 | 3714 | gets=10000, |
|
3712 | 3715 | sets=10000, |
|
3713 | 3716 | mixed=10000, |
|
3714 | 3717 | mixedgetfreq=50, |
|
3715 | 3718 | **opts |
|
3716 | 3719 | ): |
|
3717 | 3720 | opts = _byteskwargs(opts) |
|
3718 | 3721 | |
|
3719 | 3722 | def doinit(): |
|
3720 | 3723 | for i in _xrange(10000): |
|
3721 | 3724 | util.lrucachedict(size) |
|
3722 | 3725 | |
|
3723 | 3726 | costrange = list(range(mincost, maxcost + 1)) |
|
3724 | 3727 | |
|
3725 | 3728 | values = [] |
|
3726 | 3729 | for i in _xrange(size): |
|
3727 | 3730 | values.append(random.randint(0, _maxint)) |
|
3728 | 3731 | |
|
3729 | 3732 | # Get mode fills the cache and tests raw lookup performance with no |
|
3730 | 3733 | # eviction. |
|
3731 | 3734 | getseq = [] |
|
3732 | 3735 | for i in _xrange(gets): |
|
3733 | 3736 | getseq.append(random.choice(values)) |
|
3734 | 3737 | |
|
3735 | 3738 | def dogets(): |
|
3736 | 3739 | d = util.lrucachedict(size) |
|
3737 | 3740 | for v in values: |
|
3738 | 3741 | d[v] = v |
|
3739 | 3742 | for key in getseq: |
|
3740 | 3743 | value = d[key] |
|
3741 | 3744 | value # silence pyflakes warning |
|
3742 | 3745 | |
|
3743 | 3746 | def dogetscost(): |
|
3744 | 3747 | d = util.lrucachedict(size, maxcost=costlimit) |
|
3745 | 3748 | for i, v in enumerate(values): |
|
3746 | 3749 | d.insert(v, v, cost=costs[i]) |
|
3747 | 3750 | for key in getseq: |
|
3748 | 3751 | try: |
|
3749 | 3752 | value = d[key] |
|
3750 | 3753 | value # silence pyflakes warning |
|
3751 | 3754 | except KeyError: |
|
3752 | 3755 | pass |
|
3753 | 3756 | |
|
3754 | 3757 | # Set mode tests insertion speed with cache eviction. |
|
3755 | 3758 | setseq = [] |
|
3756 | 3759 | costs = [] |
|
3757 | 3760 | for i in _xrange(sets): |
|
3758 | 3761 | setseq.append(random.randint(0, _maxint)) |
|
3759 | 3762 | costs.append(random.choice(costrange)) |
|
3760 | 3763 | |
|
3761 | 3764 | def doinserts(): |
|
3762 | 3765 | d = util.lrucachedict(size) |
|
3763 | 3766 | for v in setseq: |
|
3764 | 3767 | d.insert(v, v) |
|
3765 | 3768 | |
|
3766 | 3769 | def doinsertscost(): |
|
3767 | 3770 | d = util.lrucachedict(size, maxcost=costlimit) |
|
3768 | 3771 | for i, v in enumerate(setseq): |
|
3769 | 3772 | d.insert(v, v, cost=costs[i]) |
|
3770 | 3773 | |
|
3771 | 3774 | def dosets(): |
|
3772 | 3775 | d = util.lrucachedict(size) |
|
3773 | 3776 | for v in setseq: |
|
3774 | 3777 | d[v] = v |
|
3775 | 3778 | |
|
3776 | 3779 | # Mixed mode randomly performs gets and sets with eviction. |
|
3777 | 3780 | mixedops = [] |
|
3778 | 3781 | for i in _xrange(mixed): |
|
3779 | 3782 | r = random.randint(0, 100) |
|
3780 | 3783 | if r < mixedgetfreq: |
|
3781 | 3784 | op = 0 |
|
3782 | 3785 | else: |
|
3783 | 3786 | op = 1 |
|
3784 | 3787 | |
|
3785 | 3788 | mixedops.append( |
|
3786 | 3789 | (op, random.randint(0, size * 2), random.choice(costrange)) |
|
3787 | 3790 | ) |
|
3788 | 3791 | |
|
3789 | 3792 | def domixed(): |
|
3790 | 3793 | d = util.lrucachedict(size) |
|
3791 | 3794 | |
|
3792 | 3795 | for op, v, cost in mixedops: |
|
3793 | 3796 | if op == 0: |
|
3794 | 3797 | try: |
|
3795 | 3798 | d[v] |
|
3796 | 3799 | except KeyError: |
|
3797 | 3800 | pass |
|
3798 | 3801 | else: |
|
3799 | 3802 | d[v] = v |
|
3800 | 3803 | |
|
3801 | 3804 | def domixedcost(): |
|
3802 | 3805 | d = util.lrucachedict(size, maxcost=costlimit) |
|
3803 | 3806 | |
|
3804 | 3807 | for op, v, cost in mixedops: |
|
3805 | 3808 | if op == 0: |
|
3806 | 3809 | try: |
|
3807 | 3810 | d[v] |
|
3808 | 3811 | except KeyError: |
|
3809 | 3812 | pass |
|
3810 | 3813 | else: |
|
3811 | 3814 | d.insert(v, v, cost=cost) |
|
3812 | 3815 | |
|
3813 | 3816 | benches = [ |
|
3814 | 3817 | (doinit, b'init'), |
|
3815 | 3818 | ] |
|
3816 | 3819 | |
|
3817 | 3820 | if costlimit: |
|
3818 | 3821 | benches.extend( |
|
3819 | 3822 | [ |
|
3820 | 3823 | (dogetscost, b'gets w/ cost limit'), |
|
3821 | 3824 | (doinsertscost, b'inserts w/ cost limit'), |
|
3822 | 3825 | (domixedcost, b'mixed w/ cost limit'), |
|
3823 | 3826 | ] |
|
3824 | 3827 | ) |
|
3825 | 3828 | else: |
|
3826 | 3829 | benches.extend( |
|
3827 | 3830 | [ |
|
3828 | 3831 | (dogets, b'gets'), |
|
3829 | 3832 | (doinserts, b'inserts'), |
|
3830 | 3833 | (dosets, b'sets'), |
|
3831 | 3834 | (domixed, b'mixed'), |
|
3832 | 3835 | ] |
|
3833 | 3836 | ) |
|
3834 | 3837 | |
|
3835 | 3838 | for fn, title in benches: |
|
3836 | 3839 | timer, fm = gettimer(ui, opts) |
|
3837 | 3840 | timer(fn, title=title) |
|
3838 | 3841 | fm.end() |
|
3839 | 3842 | |
|
3840 | 3843 | |
|
3841 | 3844 | @command( |
|
3842 | 3845 | b'perf::write|perfwrite', |
|
3843 | 3846 | formatteropts |
|
3844 | 3847 | + [ |
|
3845 | 3848 | (b'', b'write-method', b'write', b'ui write method'), |
|
3846 | 3849 | (b'', b'nlines', 100, b'number of lines'), |
|
3847 | 3850 | (b'', b'nitems', 100, b'number of items (per line)'), |
|
3848 | 3851 | (b'', b'item', b'x', b'item that is written'), |
|
3849 | 3852 | (b'', b'batch-line', None, b'pass whole line to write method at once'), |
|
3850 | 3853 | (b'', b'flush-line', None, b'flush after each line'), |
|
3851 | 3854 | ], |
|
3852 | 3855 | ) |
|
3853 | 3856 | def perfwrite(ui, repo, **opts): |
|
3854 | 3857 | """microbenchmark ui.write (and others)""" |
|
3855 | 3858 | opts = _byteskwargs(opts) |
|
3856 | 3859 | |
|
3857 | 3860 | write = getattr(ui, _sysstr(opts[b'write_method'])) |
|
3858 | 3861 | nlines = int(opts[b'nlines']) |
|
3859 | 3862 | nitems = int(opts[b'nitems']) |
|
3860 | 3863 | item = opts[b'item'] |
|
3861 | 3864 | batch_line = opts.get(b'batch_line') |
|
3862 | 3865 | flush_line = opts.get(b'flush_line') |
|
3863 | 3866 | |
|
3864 | 3867 | if batch_line: |
|
3865 | 3868 | line = item * nitems + b'\n' |
|
3866 | 3869 | |
|
3867 | 3870 | def benchmark(): |
|
3868 | 3871 | for i in pycompat.xrange(nlines): |
|
3869 | 3872 | if batch_line: |
|
3870 | 3873 | write(line) |
|
3871 | 3874 | else: |
|
3872 | 3875 | for i in pycompat.xrange(nitems): |
|
3873 | 3876 | write(item) |
|
3874 | 3877 | write(b'\n') |
|
3875 | 3878 | if flush_line: |
|
3876 | 3879 | ui.flush() |
|
3877 | 3880 | ui.flush() |
|
3878 | 3881 | |
|
3879 | 3882 | timer, fm = gettimer(ui, opts) |
|
3880 | 3883 | timer(benchmark) |
|
3881 | 3884 | fm.end() |
|
3882 | 3885 | |
|
3883 | 3886 | |
|
3884 | 3887 | def uisetup(ui): |
|
3885 | 3888 | if util.safehasattr(cmdutil, b'openrevlog') and not util.safehasattr( |
|
3886 | 3889 | commands, b'debugrevlogopts' |
|
3887 | 3890 | ): |
|
3888 | 3891 | # for "historical portability": |
|
3889 | 3892 | # In this case, Mercurial should be 1.9 (or a79fea6b3e77) - |
|
3890 | 3893 | # 3.7 (or 5606f7d0d063). Therefore, '--dir' option for |
|
3891 | 3894 | # openrevlog() should cause failure, because it has been |
|
3892 | 3895 | # available since 3.5 (or 49c583ca48c4). |
|
3893 | 3896 | def openrevlog(orig, repo, cmd, file_, opts): |
|
3894 | 3897 | if opts.get(b'dir') and not util.safehasattr(repo, b'dirlog'): |
|
3895 | 3898 | raise error.Abort( |
|
3896 | 3899 | b"This version doesn't support --dir option", |
|
3897 | 3900 | hint=b"use 3.5 or later", |
|
3898 | 3901 | ) |
|
3899 | 3902 | return orig(repo, cmd, file_, opts) |
|
3900 | 3903 | |
|
3901 | 3904 | extensions.wrapfunction(cmdutil, b'openrevlog', openrevlog) |
|
3902 | 3905 | |
|
3903 | 3906 | |
|
3904 | 3907 | @command( |
|
3905 | 3908 | b'perf::progress|perfprogress', |
|
3906 | 3909 | formatteropts |
|
3907 | 3910 | + [ |
|
3908 | 3911 | (b'', b'topic', b'topic', b'topic for progress messages'), |
|
3909 | 3912 | (b'c', b'total', 1000000, b'total value we are progressing to'), |
|
3910 | 3913 | ], |
|
3911 | 3914 | norepo=True, |
|
3912 | 3915 | ) |
|
3913 | 3916 | def perfprogress(ui, topic=None, total=None, **opts): |
|
3914 | 3917 | """printing of progress bars""" |
|
3915 | 3918 | opts = _byteskwargs(opts) |
|
3916 | 3919 | |
|
3917 | 3920 | timer, fm = gettimer(ui, opts) |
|
3918 | 3921 | |
|
3919 | 3922 | def doprogress(): |
|
3920 | 3923 | with ui.makeprogress(topic, total=total) as progress: |
|
3921 | 3924 | for i in _xrange(total): |
|
3922 | 3925 | progress.increment() |
|
3923 | 3926 | |
|
3924 | 3927 | timer(doprogress) |
|
3925 | 3928 | fm.end() |
@@ -1,2977 +1,2984 b'' | |||
|
1 | 1 | /* |
|
2 | 2 | parsers.c - efficient content parsing |
|
3 | 3 | |
|
4 | 4 | Copyright 2008 Olivia Mackall <olivia@selenic.com> and others |
|
5 | 5 | |
|
6 | 6 | This software may be used and distributed according to the terms of |
|
7 | 7 | the GNU General Public License, incorporated herein by reference. |
|
8 | 8 | */ |
|
9 | 9 | |
|
10 | 10 | #define PY_SSIZE_T_CLEAN |
|
11 | 11 | #include <Python.h> |
|
12 | 12 | #include <assert.h> |
|
13 | 13 | #include <ctype.h> |
|
14 | 14 | #include <limits.h> |
|
15 | 15 | #include <stddef.h> |
|
16 | 16 | #include <stdlib.h> |
|
17 | 17 | #include <string.h> |
|
18 | #include <structmember.h> | |
|
18 | 19 | |
|
19 | 20 | #include "bitmanipulation.h" |
|
20 | 21 | #include "charencode.h" |
|
21 | 22 | #include "compat.h" |
|
22 | 23 | #include "revlog.h" |
|
23 | 24 | #include "util.h" |
|
24 | 25 | |
|
25 | 26 | #ifdef IS_PY3K |
|
26 | 27 | /* The mapping of Python types is meant to be temporary to get Python |
|
27 | 28 | * 3 to compile. We should remove this once Python 3 support is fully |
|
28 | 29 | * supported and proper types are used in the extensions themselves. */ |
|
29 | 30 | #define PyInt_Check PyLong_Check |
|
30 | 31 | #define PyInt_FromLong PyLong_FromLong |
|
31 | 32 | #define PyInt_FromSsize_t PyLong_FromSsize_t |
|
32 | 33 | #define PyInt_AsLong PyLong_AsLong |
|
33 | 34 | #endif |
|
34 | 35 | |
|
35 | 36 | typedef struct indexObjectStruct indexObject; |
|
36 | 37 | |
|
37 | 38 | typedef struct { |
|
38 | 39 | int children[16]; |
|
39 | 40 | } nodetreenode; |
|
40 | 41 | |
|
41 | 42 | typedef struct { |
|
42 | 43 | int abi_version; |
|
43 | 44 | Py_ssize_t (*index_length)(const indexObject *); |
|
44 | 45 | const char *(*index_node)(indexObject *, Py_ssize_t); |
|
45 | 46 | int (*index_parents)(PyObject *, int, int *); |
|
46 | 47 | } Revlog_CAPI; |
|
47 | 48 | |
|
48 | 49 | /* |
|
49 | 50 | * A base-16 trie for fast node->rev mapping. |
|
50 | 51 | * |
|
51 | 52 | * Positive value is index of the next node in the trie |
|
52 | 53 | * Negative value is a leaf: -(rev + 2) |
|
53 | 54 | * Zero is empty |
|
54 | 55 | */ |
|
55 | 56 | typedef struct { |
|
56 | 57 | indexObject *index; |
|
57 | 58 | nodetreenode *nodes; |
|
58 | 59 | Py_ssize_t nodelen; |
|
59 | 60 | size_t length; /* # nodes in use */ |
|
60 | 61 | size_t capacity; /* # nodes allocated */ |
|
61 | 62 | int depth; /* maximum depth of tree */ |
|
62 | 63 | int splits; /* # splits performed */ |
|
63 | 64 | } nodetree; |
|
64 | 65 | |
|
65 | 66 | typedef struct { |
|
66 | 67 | PyObject_HEAD /* ; */ |
|
67 | 68 | nodetree nt; |
|
68 | 69 | } nodetreeObject; |
|
69 | 70 | |
|
70 | 71 | /* |
|
71 | 72 | * This class has two behaviors. |
|
72 | 73 | * |
|
73 | 74 | * When used in a list-like way (with integer keys), we decode an |
|
74 | 75 | * entry in a RevlogNG index file on demand. We have limited support for |
|
75 | 76 | * integer-keyed insert and delete, only at elements right before the |
|
76 | 77 | * end. |
|
77 | 78 | * |
|
78 | 79 | * With string keys, we lazily perform a reverse mapping from node to |
|
79 | 80 | * rev, using a base-16 trie. |
|
80 | 81 | */ |
|
81 | 82 | struct indexObjectStruct { |
|
82 | 83 | PyObject_HEAD |
|
83 | 84 | /* Type-specific fields go here. */ |
|
84 | 85 | PyObject *data; /* raw bytes of index */ |
|
85 | 86 | Py_ssize_t nodelen; /* digest size of the hash, 20 for SHA-1 */ |
|
86 | 87 | PyObject *nullentry; /* fast path for references to null */ |
|
87 | 88 | Py_buffer buf; /* buffer of data */ |
|
88 | 89 | const char **offsets; /* populated on demand */ |
|
89 | 90 | Py_ssize_t length; /* current on-disk number of elements */ |
|
90 | 91 | unsigned new_length; /* number of added elements */ |
|
91 | 92 | unsigned added_length; /* space reserved for added elements */ |
|
92 | 93 | char *added; /* populated on demand */ |
|
93 | 94 | PyObject *headrevs; /* cache, invalidated on changes */ |
|
94 | 95 | PyObject *filteredrevs; /* filtered revs set */ |
|
95 | 96 | nodetree nt; /* base-16 trie */ |
|
96 | 97 | int ntinitialized; /* 0 or 1 */ |
|
97 | 98 | int ntrev; /* last rev scanned */ |
|
98 | 99 | int ntlookups; /* # lookups */ |
|
99 | 100 | int ntmisses; /* # lookups that miss the cache */ |
|
100 | 101 | int inlined; |
|
101 | 102 | long hdrsize; /* size of index headers. Differs in v1 v.s. v2 format */ |
|
102 | 103 | }; |
|
103 | 104 | |
|
104 | 105 | static Py_ssize_t index_length(const indexObject *self) |
|
105 | 106 | { |
|
106 | 107 | return self->length + self->new_length; |
|
107 | 108 | } |
|
108 | 109 | |
|
109 | 110 | static const char nullid[32] = {0}; |
|
110 | 111 | static const Py_ssize_t nullrev = -1; |
|
111 | 112 | |
|
112 | 113 | static Py_ssize_t inline_scan(indexObject *self, const char **offsets); |
|
113 | 114 | |
|
114 | 115 | static int index_find_node(indexObject *self, const char *node); |
|
115 | 116 | |
|
116 | 117 | #if LONG_MAX == 0x7fffffffL |
|
117 | 118 | static const char *const v1_tuple_format = PY23("Kiiiiiis#", "Kiiiiiiy#"); |
|
118 | 119 | static const char *const v2_tuple_format = PY23("Kiiiiiis#Ki", "Kiiiiiiy#Ki"); |
|
119 | 120 | #else |
|
120 | 121 | static const char *const v1_tuple_format = PY23("kiiiiiis#", "kiiiiiiy#"); |
|
121 | 122 | static const char *const v2_tuple_format = PY23("kiiiiiis#ki", "kiiiiiiy#ki"); |
|
122 | 123 | #endif |
|
123 | 124 | |
|
124 | 125 | /* A RevlogNG v1 index entry is 64 bytes long. */ |
|
125 | 126 | static const long v1_hdrsize = 64; |
|
126 | 127 | |
|
127 | 128 | /* A Revlogv2 index entry is 96 bytes long. */ |
|
128 | 129 | static const long v2_hdrsize = 96; |
|
129 | 130 | |
|
130 | 131 | static void raise_revlog_error(void) |
|
131 | 132 | { |
|
132 | 133 | PyObject *mod = NULL, *dict = NULL, *errclass = NULL; |
|
133 | 134 | |
|
134 | 135 | mod = PyImport_ImportModule("mercurial.error"); |
|
135 | 136 | if (mod == NULL) { |
|
136 | 137 | goto cleanup; |
|
137 | 138 | } |
|
138 | 139 | |
|
139 | 140 | dict = PyModule_GetDict(mod); |
|
140 | 141 | if (dict == NULL) { |
|
141 | 142 | goto cleanup; |
|
142 | 143 | } |
|
143 | 144 | Py_INCREF(dict); |
|
144 | 145 | |
|
145 | 146 | errclass = PyDict_GetItemString(dict, "RevlogError"); |
|
146 | 147 | if (errclass == NULL) { |
|
147 | 148 | PyErr_SetString(PyExc_SystemError, |
|
148 | 149 | "could not find RevlogError"); |
|
149 | 150 | goto cleanup; |
|
150 | 151 | } |
|
151 | 152 | |
|
152 | 153 | /* value of exception is ignored by callers */ |
|
153 | 154 | PyErr_SetString(errclass, "RevlogError"); |
|
154 | 155 | |
|
155 | 156 | cleanup: |
|
156 | 157 | Py_XDECREF(dict); |
|
157 | 158 | Py_XDECREF(mod); |
|
158 | 159 | } |
|
159 | 160 | |
|
160 | 161 | /* |
|
161 | 162 | * Return a pointer to the beginning of a RevlogNG record. |
|
162 | 163 | */ |
|
163 | 164 | static const char *index_deref(indexObject *self, Py_ssize_t pos) |
|
164 | 165 | { |
|
165 | 166 | if (pos >= self->length) |
|
166 | 167 | return self->added + (pos - self->length) * self->hdrsize; |
|
167 | 168 | |
|
168 | 169 | if (self->inlined && pos > 0) { |
|
169 | 170 | if (self->offsets == NULL) { |
|
170 | 171 | Py_ssize_t ret; |
|
171 | 172 | self->offsets = |
|
172 | 173 | PyMem_Malloc(self->length * sizeof(*self->offsets)); |
|
173 | 174 | if (self->offsets == NULL) |
|
174 | 175 | return (const char *)PyErr_NoMemory(); |
|
175 | 176 | ret = inline_scan(self, self->offsets); |
|
176 | 177 | if (ret == -1) { |
|
177 | 178 | return NULL; |
|
178 | 179 | }; |
|
179 | 180 | } |
|
180 | 181 | return self->offsets[pos]; |
|
181 | 182 | } |
|
182 | 183 | |
|
183 | 184 | return (const char *)(self->buf.buf) + pos * self->hdrsize; |
|
184 | 185 | } |
|
185 | 186 | |
|
186 | 187 | /* |
|
187 | 188 | * Get parents of the given rev. |
|
188 | 189 | * |
|
189 | 190 | * The specified rev must be valid and must not be nullrev. A returned |
|
190 | 191 | * parent revision may be nullrev, but is guaranteed to be in valid range. |
|
191 | 192 | */ |
|
192 | 193 | static inline int index_get_parents(indexObject *self, Py_ssize_t rev, int *ps, |
|
193 | 194 | int maxrev) |
|
194 | 195 | { |
|
195 | 196 | const char *data = index_deref(self, rev); |
|
196 | 197 | |
|
197 | 198 | ps[0] = getbe32(data + 24); |
|
198 | 199 | ps[1] = getbe32(data + 28); |
|
199 | 200 | |
|
200 | 201 | /* If index file is corrupted, ps[] may point to invalid revisions. So |
|
201 | 202 | * there is a risk of buffer overflow to trust them unconditionally. */ |
|
202 | 203 | if (ps[0] < -1 || ps[0] > maxrev || ps[1] < -1 || ps[1] > maxrev) { |
|
203 | 204 | PyErr_SetString(PyExc_ValueError, "parent out of range"); |
|
204 | 205 | return -1; |
|
205 | 206 | } |
|
206 | 207 | return 0; |
|
207 | 208 | } |
|
208 | 209 | |
|
209 | 210 | /* |
|
210 | 211 | * Get parents of the given rev. |
|
211 | 212 | * |
|
212 | 213 | * If the specified rev is out of range, IndexError will be raised. If the |
|
213 | 214 | * revlog entry is corrupted, ValueError may be raised. |
|
214 | 215 | * |
|
215 | 216 | * Returns 0 on success or -1 on failure. |
|
216 | 217 | */ |
|
217 | 218 | static int HgRevlogIndex_GetParents(PyObject *op, int rev, int *ps) |
|
218 | 219 | { |
|
219 | 220 | int tiprev; |
|
220 | 221 | if (!op || !HgRevlogIndex_Check(op) || !ps) { |
|
221 | 222 | PyErr_BadInternalCall(); |
|
222 | 223 | return -1; |
|
223 | 224 | } |
|
224 | 225 | tiprev = (int)index_length((indexObject *)op) - 1; |
|
225 | 226 | if (rev < -1 || rev > tiprev) { |
|
226 | 227 | PyErr_Format(PyExc_IndexError, "rev out of range: %d", rev); |
|
227 | 228 | return -1; |
|
228 | 229 | } else if (rev == -1) { |
|
229 | 230 | ps[0] = ps[1] = -1; |
|
230 | 231 | return 0; |
|
231 | 232 | } else { |
|
232 | 233 | return index_get_parents((indexObject *)op, rev, ps, tiprev); |
|
233 | 234 | } |
|
234 | 235 | } |
|
235 | 236 | |
|
236 | 237 | static inline int64_t index_get_start(indexObject *self, Py_ssize_t rev) |
|
237 | 238 | { |
|
238 | 239 | const char *data; |
|
239 | 240 | uint64_t offset; |
|
240 | 241 | |
|
241 | 242 | if (rev == nullrev) |
|
242 | 243 | return 0; |
|
243 | 244 | |
|
244 | 245 | data = index_deref(self, rev); |
|
245 | 246 | offset = getbe32(data + 4); |
|
246 | 247 | if (rev == 0) { |
|
247 | 248 | /* mask out version number for the first entry */ |
|
248 | 249 | offset &= 0xFFFF; |
|
249 | 250 | } else { |
|
250 | 251 | uint32_t offset_high = getbe32(data); |
|
251 | 252 | offset |= ((uint64_t)offset_high) << 32; |
|
252 | 253 | } |
|
253 | 254 | return (int64_t)(offset >> 16); |
|
254 | 255 | } |
|
255 | 256 | |
|
256 | 257 | static inline int index_get_length(indexObject *self, Py_ssize_t rev) |
|
257 | 258 | { |
|
258 | 259 | const char *data; |
|
259 | 260 | int tmp; |
|
260 | 261 | |
|
261 | 262 | if (rev == nullrev) |
|
262 | 263 | return 0; |
|
263 | 264 | |
|
264 | 265 | data = index_deref(self, rev); |
|
265 | 266 | |
|
266 | 267 | tmp = (int)getbe32(data + 8); |
|
267 | 268 | if (tmp < 0) { |
|
268 | 269 | PyErr_Format(PyExc_OverflowError, |
|
269 | 270 | "revlog entry size out of bound (%d)", tmp); |
|
270 | 271 | return -1; |
|
271 | 272 | } |
|
272 | 273 | return tmp; |
|
273 | 274 | } |
|
274 | 275 | |
|
275 | 276 | /* |
|
276 | 277 | * RevlogNG format (all in big endian, data may be inlined): |
|
277 | 278 | * 6 bytes: offset |
|
278 | 279 | * 2 bytes: flags |
|
279 | 280 | * 4 bytes: compressed length |
|
280 | 281 | * 4 bytes: uncompressed length |
|
281 | 282 | * 4 bytes: base revision |
|
282 | 283 | * 4 bytes: link revision |
|
283 | 284 | * 4 bytes: parent 1 revision |
|
284 | 285 | * 4 bytes: parent 2 revision |
|
285 | 286 | * 32 bytes: nodeid (only 20 bytes used with SHA-1) |
|
286 | 287 | */ |
|
287 | 288 | static PyObject *index_get(indexObject *self, Py_ssize_t pos) |
|
288 | 289 | { |
|
289 | 290 | uint64_t offset_flags, sidedata_offset; |
|
290 | 291 | int comp_len, uncomp_len, base_rev, link_rev, parent_1, parent_2, |
|
291 | 292 | sidedata_comp_len; |
|
292 | 293 | const char *c_node_id; |
|
293 | 294 | const char *data; |
|
294 | 295 | Py_ssize_t length = index_length(self); |
|
295 | 296 | |
|
296 | 297 | if (pos == nullrev) { |
|
297 | 298 | Py_INCREF(self->nullentry); |
|
298 | 299 | return self->nullentry; |
|
299 | 300 | } |
|
300 | 301 | |
|
301 | 302 | if (pos < 0 || pos >= length) { |
|
302 | 303 | PyErr_SetString(PyExc_IndexError, "revlog index out of range"); |
|
303 | 304 | return NULL; |
|
304 | 305 | } |
|
305 | 306 | |
|
306 | 307 | data = index_deref(self, pos); |
|
307 | 308 | if (data == NULL) |
|
308 | 309 | return NULL; |
|
309 | 310 | |
|
310 | 311 | offset_flags = getbe32(data + 4); |
|
311 | 312 | /* |
|
312 | 313 | * The first entry on-disk needs the version number masked out, |
|
313 | 314 | * but this doesn't apply if entries are added to an empty index. |
|
314 | 315 | */ |
|
315 | 316 | if (self->length && pos == 0) |
|
316 | 317 | offset_flags &= 0xFFFF; |
|
317 | 318 | else { |
|
318 | 319 | uint32_t offset_high = getbe32(data); |
|
319 | 320 | offset_flags |= ((uint64_t)offset_high) << 32; |
|
320 | 321 | } |
|
321 | 322 | |
|
322 | 323 | comp_len = getbe32(data + 8); |
|
323 | 324 | uncomp_len = getbe32(data + 12); |
|
324 | 325 | base_rev = getbe32(data + 16); |
|
325 | 326 | link_rev = getbe32(data + 20); |
|
326 | 327 | parent_1 = getbe32(data + 24); |
|
327 | 328 | parent_2 = getbe32(data + 28); |
|
328 | 329 | c_node_id = data + 32; |
|
329 | 330 | |
|
330 | 331 | if (self->hdrsize == v1_hdrsize) { |
|
331 | 332 | return Py_BuildValue(v1_tuple_format, offset_flags, comp_len, |
|
332 | 333 | uncomp_len, base_rev, link_rev, parent_1, |
|
333 | 334 | parent_2, c_node_id, self->nodelen); |
|
334 | 335 | } else { |
|
335 | 336 | sidedata_offset = getbe64(data + 64); |
|
336 | 337 | sidedata_comp_len = getbe32(data + 72); |
|
337 | 338 | |
|
338 | 339 | return Py_BuildValue(v2_tuple_format, offset_flags, comp_len, |
|
339 | 340 | uncomp_len, base_rev, link_rev, parent_1, |
|
340 | 341 | parent_2, c_node_id, self->nodelen, |
|
341 | 342 | sidedata_offset, sidedata_comp_len); |
|
342 | 343 | } |
|
343 | 344 | } |
|
344 | 345 | |
|
345 | 346 | /* |
|
346 | 347 | * Return the hash of node corresponding to the given rev. |
|
347 | 348 | */ |
|
348 | 349 | static const char *index_node(indexObject *self, Py_ssize_t pos) |
|
349 | 350 | { |
|
350 | 351 | Py_ssize_t length = index_length(self); |
|
351 | 352 | const char *data; |
|
352 | 353 | |
|
353 | 354 | if (pos == nullrev) |
|
354 | 355 | return nullid; |
|
355 | 356 | |
|
356 | 357 | if (pos >= length) |
|
357 | 358 | return NULL; |
|
358 | 359 | |
|
359 | 360 | data = index_deref(self, pos); |
|
360 | 361 | return data ? data + 32 : NULL; |
|
361 | 362 | } |
|
362 | 363 | |
|
363 | 364 | /* |
|
364 | 365 | * Return the hash of the node corresponding to the given rev. The |
|
365 | 366 | * rev is assumed to be existing. If not, an exception is set. |
|
366 | 367 | */ |
|
367 | 368 | static const char *index_node_existing(indexObject *self, Py_ssize_t pos) |
|
368 | 369 | { |
|
369 | 370 | const char *node = index_node(self, pos); |
|
370 | 371 | if (node == NULL) { |
|
371 | 372 | PyErr_Format(PyExc_IndexError, "could not access rev %d", |
|
372 | 373 | (int)pos); |
|
373 | 374 | } |
|
374 | 375 | return node; |
|
375 | 376 | } |
|
376 | 377 | |
|
377 | 378 | static int nt_insert(nodetree *self, const char *node, int rev); |
|
378 | 379 | |
|
379 | 380 | static int node_check(Py_ssize_t nodelen, PyObject *obj, char **node) |
|
380 | 381 | { |
|
381 | 382 | Py_ssize_t thisnodelen; |
|
382 | 383 | if (PyBytes_AsStringAndSize(obj, node, &thisnodelen) == -1) |
|
383 | 384 | return -1; |
|
384 | 385 | if (nodelen == thisnodelen) |
|
385 | 386 | return 0; |
|
386 | 387 | PyErr_Format(PyExc_ValueError, "node len %zd != expected node len %zd", |
|
387 | 388 | thisnodelen, nodelen); |
|
388 | 389 | return -1; |
|
389 | 390 | } |
|
390 | 391 | |
|
391 | 392 | static PyObject *index_append(indexObject *self, PyObject *obj) |
|
392 | 393 | { |
|
393 | 394 | uint64_t offset_flags, sidedata_offset; |
|
394 | 395 | int rev, comp_len, uncomp_len, base_rev, link_rev, parent_1, parent_2; |
|
395 | 396 | Py_ssize_t c_node_id_len, sidedata_comp_len; |
|
396 | 397 | const char *c_node_id; |
|
397 | 398 | char *data; |
|
398 | 399 | |
|
399 | 400 | if (self->hdrsize == v1_hdrsize) { |
|
400 | 401 | if (!PyArg_ParseTuple(obj, v1_tuple_format, &offset_flags, |
|
401 | 402 | &comp_len, &uncomp_len, &base_rev, |
|
402 | 403 | &link_rev, &parent_1, &parent_2, |
|
403 | 404 | &c_node_id, &c_node_id_len)) { |
|
404 | 405 | PyErr_SetString(PyExc_TypeError, "8-tuple required"); |
|
405 | 406 | return NULL; |
|
406 | 407 | } |
|
407 | 408 | } else { |
|
408 | 409 | if (!PyArg_ParseTuple(obj, v2_tuple_format, &offset_flags, |
|
409 | 410 | &comp_len, &uncomp_len, &base_rev, |
|
410 | 411 | &link_rev, &parent_1, &parent_2, |
|
411 | 412 | &c_node_id, &c_node_id_len, |
|
412 | 413 | &sidedata_offset, &sidedata_comp_len)) { |
|
413 | 414 | PyErr_SetString(PyExc_TypeError, "10-tuple required"); |
|
414 | 415 | return NULL; |
|
415 | 416 | } |
|
416 | 417 | } |
|
417 | 418 | |
|
418 | 419 | if (c_node_id_len != self->nodelen) { |
|
419 | 420 | PyErr_SetString(PyExc_TypeError, "invalid node"); |
|
420 | 421 | return NULL; |
|
421 | 422 | } |
|
422 | 423 | |
|
423 | 424 | if (self->new_length == self->added_length) { |
|
424 | 425 | size_t new_added_length = |
|
425 | 426 | self->added_length ? self->added_length * 2 : 4096; |
|
426 | 427 | void *new_added = PyMem_Realloc(self->added, new_added_length * |
|
427 | 428 | self->hdrsize); |
|
428 | 429 | if (!new_added) |
|
429 | 430 | return PyErr_NoMemory(); |
|
430 | 431 | self->added = new_added; |
|
431 | 432 | self->added_length = new_added_length; |
|
432 | 433 | } |
|
433 | 434 | rev = self->length + self->new_length; |
|
434 | 435 | data = self->added + self->hdrsize * self->new_length++; |
|
435 | 436 | putbe32(offset_flags >> 32, data); |
|
436 | 437 | putbe32(offset_flags & 0xffffffffU, data + 4); |
|
437 | 438 | putbe32(comp_len, data + 8); |
|
438 | 439 | putbe32(uncomp_len, data + 12); |
|
439 | 440 | putbe32(base_rev, data + 16); |
|
440 | 441 | putbe32(link_rev, data + 20); |
|
441 | 442 | putbe32(parent_1, data + 24); |
|
442 | 443 | putbe32(parent_2, data + 28); |
|
443 | 444 | memcpy(data + 32, c_node_id, c_node_id_len); |
|
444 | 445 | /* Padding since SHA-1 is only 20 bytes for now */ |
|
445 | 446 | memset(data + 32 + c_node_id_len, 0, 32 - c_node_id_len); |
|
446 | 447 | if (self->hdrsize != v1_hdrsize) { |
|
447 | 448 | putbe64(sidedata_offset, data + 64); |
|
448 | 449 | putbe32(sidedata_comp_len, data + 72); |
|
449 | 450 | /* Padding for 96 bytes alignment */ |
|
450 | 451 | memset(data + 76, 0, self->hdrsize - 76); |
|
451 | 452 | } |
|
452 | 453 | |
|
453 | 454 | if (self->ntinitialized) |
|
454 | 455 | nt_insert(&self->nt, c_node_id, rev); |
|
455 | 456 | |
|
456 | 457 | Py_CLEAR(self->headrevs); |
|
457 | 458 | Py_RETURN_NONE; |
|
458 | 459 | } |
|
459 | 460 | |
|
460 | 461 | /* Replace an existing index entry's sidedata offset and length with new ones. |
|
461 | 462 | This cannot be used outside of the context of sidedata rewriting, |
|
462 | 463 | inside the transaction that creates the given revision. */ |
|
463 | 464 | static PyObject *index_replace_sidedata_info(indexObject *self, PyObject *args) |
|
464 | 465 | { |
|
465 | 466 | uint64_t sidedata_offset; |
|
466 | 467 | int rev; |
|
467 | 468 | Py_ssize_t sidedata_comp_len; |
|
468 | 469 | char *data; |
|
469 | 470 | #if LONG_MAX == 0x7fffffffL |
|
470 | 471 | const char *const sidedata_format = PY23("nKi", "nKi"); |
|
471 | 472 | #else |
|
472 | 473 | const char *const sidedata_format = PY23("nki", "nki"); |
|
473 | 474 | #endif |
|
474 | 475 | |
|
475 | 476 | if (self->hdrsize == v1_hdrsize || self->inlined) { |
|
476 | 477 | /* |
|
477 | 478 | There is a bug in the transaction handling when going from an |
|
478 | 479 | inline revlog to a separate index and data file. Turn it off until |
|
479 | 480 | it's fixed, since v2 revlogs sometimes get rewritten on exchange. |
|
480 | 481 | See issue6485. |
|
481 | 482 | */ |
|
482 | 483 | raise_revlog_error(); |
|
483 | 484 | return NULL; |
|
484 | 485 | } |
|
485 | 486 | |
|
486 | 487 | if (!PyArg_ParseTuple(args, sidedata_format, &rev, &sidedata_offset, |
|
487 | 488 | &sidedata_comp_len)) |
|
488 | 489 | return NULL; |
|
489 | 490 | |
|
490 | 491 | if (rev < 0 || rev >= index_length(self)) { |
|
491 | 492 | PyErr_SetString(PyExc_IndexError, "revision outside index"); |
|
492 | 493 | return NULL; |
|
493 | 494 | } |
|
494 | 495 | if (rev < self->length) { |
|
495 | 496 | PyErr_SetString( |
|
496 | 497 | PyExc_IndexError, |
|
497 | 498 | "cannot rewrite entries outside of this transaction"); |
|
498 | 499 | return NULL; |
|
499 | 500 | } |
|
500 | 501 | |
|
501 | 502 | /* Find the newly added node, offset from the "already on-disk" length |
|
502 | 503 | */ |
|
503 | 504 | data = self->added + self->hdrsize * (rev - self->length); |
|
504 | 505 | putbe64(sidedata_offset, data + 64); |
|
505 | 506 | putbe32(sidedata_comp_len, data + 72); |
|
506 | 507 | |
|
507 | 508 | Py_RETURN_NONE; |
|
508 | 509 | } |
|
509 | 510 | |
|
510 | 511 | static PyObject *index_stats(indexObject *self) |
|
511 | 512 | { |
|
512 | 513 | PyObject *obj = PyDict_New(); |
|
513 | 514 | PyObject *s = NULL; |
|
514 | 515 | PyObject *t = NULL; |
|
515 | 516 | |
|
516 | 517 | if (obj == NULL) |
|
517 | 518 | return NULL; |
|
518 | 519 | |
|
519 | 520 | #define istat(__n, __d) \ |
|
520 | 521 | do { \ |
|
521 | 522 | s = PyBytes_FromString(__d); \ |
|
522 | 523 | t = PyInt_FromSsize_t(self->__n); \ |
|
523 | 524 | if (!s || !t) \ |
|
524 | 525 | goto bail; \ |
|
525 | 526 | if (PyDict_SetItem(obj, s, t) == -1) \ |
|
526 | 527 | goto bail; \ |
|
527 | 528 | Py_CLEAR(s); \ |
|
528 | 529 | Py_CLEAR(t); \ |
|
529 | 530 | } while (0) |
|
530 | 531 | |
|
531 | 532 | if (self->added_length) |
|
532 | 533 | istat(new_length, "index entries added"); |
|
533 | 534 | istat(length, "revs in memory"); |
|
534 | 535 | istat(ntlookups, "node trie lookups"); |
|
535 | 536 | istat(ntmisses, "node trie misses"); |
|
536 | 537 | istat(ntrev, "node trie last rev scanned"); |
|
537 | 538 | if (self->ntinitialized) { |
|
538 | 539 | istat(nt.capacity, "node trie capacity"); |
|
539 | 540 | istat(nt.depth, "node trie depth"); |
|
540 | 541 | istat(nt.length, "node trie count"); |
|
541 | 542 | istat(nt.splits, "node trie splits"); |
|
542 | 543 | } |
|
543 | 544 | |
|
544 | 545 | #undef istat |
|
545 | 546 | |
|
546 | 547 | return obj; |
|
547 | 548 | |
|
548 | 549 | bail: |
|
549 | 550 | Py_XDECREF(obj); |
|
550 | 551 | Py_XDECREF(s); |
|
551 | 552 | Py_XDECREF(t); |
|
552 | 553 | return NULL; |
|
553 | 554 | } |
|
554 | 555 | |
|
555 | 556 | /* |
|
556 | 557 | * When we cache a list, we want to be sure the caller can't mutate |
|
557 | 558 | * the cached copy. |
|
558 | 559 | */ |
|
559 | 560 | static PyObject *list_copy(PyObject *list) |
|
560 | 561 | { |
|
561 | 562 | Py_ssize_t len = PyList_GET_SIZE(list); |
|
562 | 563 | PyObject *newlist = PyList_New(len); |
|
563 | 564 | Py_ssize_t i; |
|
564 | 565 | |
|
565 | 566 | if (newlist == NULL) |
|
566 | 567 | return NULL; |
|
567 | 568 | |
|
568 | 569 | for (i = 0; i < len; i++) { |
|
569 | 570 | PyObject *obj = PyList_GET_ITEM(list, i); |
|
570 | 571 | Py_INCREF(obj); |
|
571 | 572 | PyList_SET_ITEM(newlist, i, obj); |
|
572 | 573 | } |
|
573 | 574 | |
|
574 | 575 | return newlist; |
|
575 | 576 | } |
|
576 | 577 | |
|
577 | 578 | static int check_filter(PyObject *filter, Py_ssize_t arg) |
|
578 | 579 | { |
|
579 | 580 | if (filter) { |
|
580 | 581 | PyObject *arglist, *result; |
|
581 | 582 | int isfiltered; |
|
582 | 583 | |
|
583 | 584 | arglist = Py_BuildValue("(n)", arg); |
|
584 | 585 | if (!arglist) { |
|
585 | 586 | return -1; |
|
586 | 587 | } |
|
587 | 588 | |
|
588 | 589 | result = PyObject_Call(filter, arglist, NULL); |
|
589 | 590 | Py_DECREF(arglist); |
|
590 | 591 | if (!result) { |
|
591 | 592 | return -1; |
|
592 | 593 | } |
|
593 | 594 | |
|
594 | 595 | /* PyObject_IsTrue returns 1 if true, 0 if false, -1 if error, |
|
595 | 596 | * same as this function, so we can just return it directly.*/ |
|
596 | 597 | isfiltered = PyObject_IsTrue(result); |
|
597 | 598 | Py_DECREF(result); |
|
598 | 599 | return isfiltered; |
|
599 | 600 | } else { |
|
600 | 601 | return 0; |
|
601 | 602 | } |
|
602 | 603 | } |
|
603 | 604 | |
|
604 | 605 | static inline void set_phase_from_parents(char *phases, int parent_1, |
|
605 | 606 | int parent_2, Py_ssize_t i) |
|
606 | 607 | { |
|
607 | 608 | if (parent_1 >= 0 && phases[parent_1] > phases[i]) |
|
608 | 609 | phases[i] = phases[parent_1]; |
|
609 | 610 | if (parent_2 >= 0 && phases[parent_2] > phases[i]) |
|
610 | 611 | phases[i] = phases[parent_2]; |
|
611 | 612 | } |
|
612 | 613 | |
|
613 | 614 | static PyObject *reachableroots2(indexObject *self, PyObject *args) |
|
614 | 615 | { |
|
615 | 616 | |
|
616 | 617 | /* Input */ |
|
617 | 618 | long minroot; |
|
618 | 619 | PyObject *includepatharg = NULL; |
|
619 | 620 | int includepath = 0; |
|
620 | 621 | /* heads and roots are lists */ |
|
621 | 622 | PyObject *heads = NULL; |
|
622 | 623 | PyObject *roots = NULL; |
|
623 | 624 | PyObject *reachable = NULL; |
|
624 | 625 | |
|
625 | 626 | PyObject *val; |
|
626 | 627 | Py_ssize_t len = index_length(self); |
|
627 | 628 | long revnum; |
|
628 | 629 | Py_ssize_t k; |
|
629 | 630 | Py_ssize_t i; |
|
630 | 631 | Py_ssize_t l; |
|
631 | 632 | int r; |
|
632 | 633 | int parents[2]; |
|
633 | 634 | |
|
634 | 635 | /* Internal data structure: |
|
635 | 636 | * tovisit: array of length len+1 (all revs + nullrev), filled upto |
|
636 | 637 | * lentovisit |
|
637 | 638 | * |
|
638 | 639 | * revstates: array of length len+1 (all revs + nullrev) */ |
|
639 | 640 | int *tovisit = NULL; |
|
640 | 641 | long lentovisit = 0; |
|
641 | 642 | enum { RS_SEEN = 1, RS_ROOT = 2, RS_REACHABLE = 4 }; |
|
642 | 643 | char *revstates = NULL; |
|
643 | 644 | |
|
644 | 645 | /* Get arguments */ |
|
645 | 646 | if (!PyArg_ParseTuple(args, "lO!O!O!", &minroot, &PyList_Type, &heads, |
|
646 | 647 | &PyList_Type, &roots, &PyBool_Type, |
|
647 | 648 | &includepatharg)) |
|
648 | 649 | goto bail; |
|
649 | 650 | |
|
650 | 651 | if (includepatharg == Py_True) |
|
651 | 652 | includepath = 1; |
|
652 | 653 | |
|
653 | 654 | /* Initialize return set */ |
|
654 | 655 | reachable = PyList_New(0); |
|
655 | 656 | if (reachable == NULL) |
|
656 | 657 | goto bail; |
|
657 | 658 | |
|
658 | 659 | /* Initialize internal datastructures */ |
|
659 | 660 | tovisit = (int *)malloc((len + 1) * sizeof(int)); |
|
660 | 661 | if (tovisit == NULL) { |
|
661 | 662 | PyErr_NoMemory(); |
|
662 | 663 | goto bail; |
|
663 | 664 | } |
|
664 | 665 | |
|
665 | 666 | revstates = (char *)calloc(len + 1, 1); |
|
666 | 667 | if (revstates == NULL) { |
|
667 | 668 | PyErr_NoMemory(); |
|
668 | 669 | goto bail; |
|
669 | 670 | } |
|
670 | 671 | |
|
671 | 672 | l = PyList_GET_SIZE(roots); |
|
672 | 673 | for (i = 0; i < l; i++) { |
|
673 | 674 | revnum = PyInt_AsLong(PyList_GET_ITEM(roots, i)); |
|
674 | 675 | if (revnum == -1 && PyErr_Occurred()) |
|
675 | 676 | goto bail; |
|
676 | 677 | /* If root is out of range, e.g. wdir(), it must be unreachable |
|
677 | 678 | * from heads. So we can just ignore it. */ |
|
678 | 679 | if (revnum + 1 < 0 || revnum + 1 >= len + 1) |
|
679 | 680 | continue; |
|
680 | 681 | revstates[revnum + 1] |= RS_ROOT; |
|
681 | 682 | } |
|
682 | 683 | |
|
683 | 684 | /* Populate tovisit with all the heads */ |
|
684 | 685 | l = PyList_GET_SIZE(heads); |
|
685 | 686 | for (i = 0; i < l; i++) { |
|
686 | 687 | revnum = PyInt_AsLong(PyList_GET_ITEM(heads, i)); |
|
687 | 688 | if (revnum == -1 && PyErr_Occurred()) |
|
688 | 689 | goto bail; |
|
689 | 690 | if (revnum + 1 < 0 || revnum + 1 >= len + 1) { |
|
690 | 691 | PyErr_SetString(PyExc_IndexError, "head out of range"); |
|
691 | 692 | goto bail; |
|
692 | 693 | } |
|
693 | 694 | if (!(revstates[revnum + 1] & RS_SEEN)) { |
|
694 | 695 | tovisit[lentovisit++] = (int)revnum; |
|
695 | 696 | revstates[revnum + 1] |= RS_SEEN; |
|
696 | 697 | } |
|
697 | 698 | } |
|
698 | 699 | |
|
699 | 700 | /* Visit the tovisit list and find the reachable roots */ |
|
700 | 701 | k = 0; |
|
701 | 702 | while (k < lentovisit) { |
|
702 | 703 | /* Add the node to reachable if it is a root*/ |
|
703 | 704 | revnum = tovisit[k++]; |
|
704 | 705 | if (revstates[revnum + 1] & RS_ROOT) { |
|
705 | 706 | revstates[revnum + 1] |= RS_REACHABLE; |
|
706 | 707 | val = PyInt_FromLong(revnum); |
|
707 | 708 | if (val == NULL) |
|
708 | 709 | goto bail; |
|
709 | 710 | r = PyList_Append(reachable, val); |
|
710 | 711 | Py_DECREF(val); |
|
711 | 712 | if (r < 0) |
|
712 | 713 | goto bail; |
|
713 | 714 | if (includepath == 0) |
|
714 | 715 | continue; |
|
715 | 716 | } |
|
716 | 717 | |
|
717 | 718 | /* Add its parents to the list of nodes to visit */ |
|
718 | 719 | if (revnum == nullrev) |
|
719 | 720 | continue; |
|
720 | 721 | r = index_get_parents(self, revnum, parents, (int)len - 1); |
|
721 | 722 | if (r < 0) |
|
722 | 723 | goto bail; |
|
723 | 724 | for (i = 0; i < 2; i++) { |
|
724 | 725 | if (!(revstates[parents[i] + 1] & RS_SEEN) && |
|
725 | 726 | parents[i] >= minroot) { |
|
726 | 727 | tovisit[lentovisit++] = parents[i]; |
|
727 | 728 | revstates[parents[i] + 1] |= RS_SEEN; |
|
728 | 729 | } |
|
729 | 730 | } |
|
730 | 731 | } |
|
731 | 732 | |
|
732 | 733 | /* Find all the nodes in between the roots we found and the heads |
|
733 | 734 | * and add them to the reachable set */ |
|
734 | 735 | if (includepath == 1) { |
|
735 | 736 | long minidx = minroot; |
|
736 | 737 | if (minidx < 0) |
|
737 | 738 | minidx = 0; |
|
738 | 739 | for (i = minidx; i < len; i++) { |
|
739 | 740 | if (!(revstates[i + 1] & RS_SEEN)) |
|
740 | 741 | continue; |
|
741 | 742 | r = index_get_parents(self, i, parents, (int)len - 1); |
|
742 | 743 | /* Corrupted index file, error is set from |
|
743 | 744 | * index_get_parents */ |
|
744 | 745 | if (r < 0) |
|
745 | 746 | goto bail; |
|
746 | 747 | if (((revstates[parents[0] + 1] | |
|
747 | 748 | revstates[parents[1] + 1]) & |
|
748 | 749 | RS_REACHABLE) && |
|
749 | 750 | !(revstates[i + 1] & RS_REACHABLE)) { |
|
750 | 751 | revstates[i + 1] |= RS_REACHABLE; |
|
751 | 752 | val = PyInt_FromSsize_t(i); |
|
752 | 753 | if (val == NULL) |
|
753 | 754 | goto bail; |
|
754 | 755 | r = PyList_Append(reachable, val); |
|
755 | 756 | Py_DECREF(val); |
|
756 | 757 | if (r < 0) |
|
757 | 758 | goto bail; |
|
758 | 759 | } |
|
759 | 760 | } |
|
760 | 761 | } |
|
761 | 762 | |
|
762 | 763 | free(revstates); |
|
763 | 764 | free(tovisit); |
|
764 | 765 | return reachable; |
|
765 | 766 | bail: |
|
766 | 767 | Py_XDECREF(reachable); |
|
767 | 768 | free(revstates); |
|
768 | 769 | free(tovisit); |
|
769 | 770 | return NULL; |
|
770 | 771 | } |
|
771 | 772 | |
|
772 | 773 | static int add_roots_get_min(indexObject *self, PyObject *roots, char *phases, |
|
773 | 774 | char phase) |
|
774 | 775 | { |
|
775 | 776 | Py_ssize_t len = index_length(self); |
|
776 | 777 | PyObject *item; |
|
777 | 778 | PyObject *iterator; |
|
778 | 779 | int rev, minrev = -1; |
|
779 | 780 | char *node; |
|
780 | 781 | |
|
781 | 782 | if (!PySet_Check(roots)) { |
|
782 | 783 | PyErr_SetString(PyExc_TypeError, |
|
783 | 784 | "roots must be a set of nodes"); |
|
784 | 785 | return -2; |
|
785 | 786 | } |
|
786 | 787 | iterator = PyObject_GetIter(roots); |
|
787 | 788 | if (iterator == NULL) |
|
788 | 789 | return -2; |
|
789 | 790 | while ((item = PyIter_Next(iterator))) { |
|
790 | 791 | if (node_check(self->nodelen, item, &node) == -1) |
|
791 | 792 | goto failed; |
|
792 | 793 | rev = index_find_node(self, node); |
|
793 | 794 | /* null is implicitly public, so negative is invalid */ |
|
794 | 795 | if (rev < 0 || rev >= len) |
|
795 | 796 | goto failed; |
|
796 | 797 | phases[rev] = phase; |
|
797 | 798 | if (minrev == -1 || minrev > rev) |
|
798 | 799 | minrev = rev; |
|
799 | 800 | Py_DECREF(item); |
|
800 | 801 | } |
|
801 | 802 | Py_DECREF(iterator); |
|
802 | 803 | return minrev; |
|
803 | 804 | failed: |
|
804 | 805 | Py_DECREF(iterator); |
|
805 | 806 | Py_DECREF(item); |
|
806 | 807 | return -2; |
|
807 | 808 | } |
|
808 | 809 | |
|
809 | 810 | static PyObject *compute_phases_map_sets(indexObject *self, PyObject *args) |
|
810 | 811 | { |
|
811 | 812 | /* 0: public (untracked), 1: draft, 2: secret, 32: archive, |
|
812 | 813 | 96: internal */ |
|
813 | 814 | static const char trackedphases[] = {1, 2, 32, 96}; |
|
814 | 815 | PyObject *roots = Py_None; |
|
815 | 816 | PyObject *phasesetsdict = NULL; |
|
816 | 817 | PyObject *phasesets[4] = {NULL, NULL, NULL, NULL}; |
|
817 | 818 | Py_ssize_t len = index_length(self); |
|
818 | 819 | char *phases = NULL; |
|
819 | 820 | int minphaserev = -1, rev, i; |
|
820 | 821 | const int numphases = (int)(sizeof(phasesets) / sizeof(phasesets[0])); |
|
821 | 822 | |
|
822 | 823 | if (!PyArg_ParseTuple(args, "O", &roots)) |
|
823 | 824 | return NULL; |
|
824 | 825 | if (roots == NULL || !PyDict_Check(roots)) { |
|
825 | 826 | PyErr_SetString(PyExc_TypeError, "roots must be a dictionary"); |
|
826 | 827 | return NULL; |
|
827 | 828 | } |
|
828 | 829 | |
|
829 | 830 | phases = calloc(len, 1); |
|
830 | 831 | if (phases == NULL) { |
|
831 | 832 | PyErr_NoMemory(); |
|
832 | 833 | return NULL; |
|
833 | 834 | } |
|
834 | 835 | |
|
835 | 836 | for (i = 0; i < numphases; ++i) { |
|
836 | 837 | PyObject *pyphase = PyInt_FromLong(trackedphases[i]); |
|
837 | 838 | PyObject *phaseroots = NULL; |
|
838 | 839 | if (pyphase == NULL) |
|
839 | 840 | goto release; |
|
840 | 841 | phaseroots = PyDict_GetItem(roots, pyphase); |
|
841 | 842 | Py_DECREF(pyphase); |
|
842 | 843 | if (phaseroots == NULL) |
|
843 | 844 | continue; |
|
844 | 845 | rev = add_roots_get_min(self, phaseroots, phases, |
|
845 | 846 | trackedphases[i]); |
|
846 | 847 | if (rev == -2) |
|
847 | 848 | goto release; |
|
848 | 849 | if (rev != -1 && (minphaserev == -1 || rev < minphaserev)) |
|
849 | 850 | minphaserev = rev; |
|
850 | 851 | } |
|
851 | 852 | |
|
852 | 853 | for (i = 0; i < numphases; ++i) { |
|
853 | 854 | phasesets[i] = PySet_New(NULL); |
|
854 | 855 | if (phasesets[i] == NULL) |
|
855 | 856 | goto release; |
|
856 | 857 | } |
|
857 | 858 | |
|
858 | 859 | if (minphaserev == -1) |
|
859 | 860 | minphaserev = len; |
|
860 | 861 | for (rev = minphaserev; rev < len; ++rev) { |
|
861 | 862 | PyObject *pyphase = NULL; |
|
862 | 863 | PyObject *pyrev = NULL; |
|
863 | 864 | int parents[2]; |
|
864 | 865 | /* |
|
865 | 866 | * The parent lookup could be skipped for phaseroots, but |
|
866 | 867 | * phase --force would historically not recompute them |
|
867 | 868 | * correctly, leaving descendents with a lower phase around. |
|
868 | 869 | * As such, unconditionally recompute the phase. |
|
869 | 870 | */ |
|
870 | 871 | if (index_get_parents(self, rev, parents, (int)len - 1) < 0) |
|
871 | 872 | goto release; |
|
872 | 873 | set_phase_from_parents(phases, parents[0], parents[1], rev); |
|
873 | 874 | switch (phases[rev]) { |
|
874 | 875 | case 0: |
|
875 | 876 | continue; |
|
876 | 877 | case 1: |
|
877 | 878 | pyphase = phasesets[0]; |
|
878 | 879 | break; |
|
879 | 880 | case 2: |
|
880 | 881 | pyphase = phasesets[1]; |
|
881 | 882 | break; |
|
882 | 883 | case 32: |
|
883 | 884 | pyphase = phasesets[2]; |
|
884 | 885 | break; |
|
885 | 886 | case 96: |
|
886 | 887 | pyphase = phasesets[3]; |
|
887 | 888 | break; |
|
888 | 889 | default: |
|
889 | 890 | /* this should never happen since the phase number is |
|
890 | 891 | * specified by this function. */ |
|
891 | 892 | PyErr_SetString(PyExc_SystemError, |
|
892 | 893 | "bad phase number in internal list"); |
|
893 | 894 | goto release; |
|
894 | 895 | } |
|
895 | 896 | pyrev = PyInt_FromLong(rev); |
|
896 | 897 | if (pyrev == NULL) |
|
897 | 898 | goto release; |
|
898 | 899 | if (PySet_Add(pyphase, pyrev) == -1) { |
|
899 | 900 | Py_DECREF(pyrev); |
|
900 | 901 | goto release; |
|
901 | 902 | } |
|
902 | 903 | Py_DECREF(pyrev); |
|
903 | 904 | } |
|
904 | 905 | |
|
905 | 906 | phasesetsdict = _dict_new_presized(numphases); |
|
906 | 907 | if (phasesetsdict == NULL) |
|
907 | 908 | goto release; |
|
908 | 909 | for (i = 0; i < numphases; ++i) { |
|
909 | 910 | PyObject *pyphase = PyInt_FromLong(trackedphases[i]); |
|
910 | 911 | if (pyphase == NULL) |
|
911 | 912 | goto release; |
|
912 | 913 | if (PyDict_SetItem(phasesetsdict, pyphase, phasesets[i]) == |
|
913 | 914 | -1) { |
|
914 | 915 | Py_DECREF(pyphase); |
|
915 | 916 | goto release; |
|
916 | 917 | } |
|
917 | 918 | Py_DECREF(phasesets[i]); |
|
918 | 919 | phasesets[i] = NULL; |
|
919 | 920 | } |
|
920 | 921 | |
|
921 | 922 | return Py_BuildValue("nN", len, phasesetsdict); |
|
922 | 923 | |
|
923 | 924 | release: |
|
924 | 925 | for (i = 0; i < numphases; ++i) |
|
925 | 926 | Py_XDECREF(phasesets[i]); |
|
926 | 927 | Py_XDECREF(phasesetsdict); |
|
927 | 928 | |
|
928 | 929 | free(phases); |
|
929 | 930 | return NULL; |
|
930 | 931 | } |
|
931 | 932 | |
|
932 | 933 | static PyObject *index_headrevs(indexObject *self, PyObject *args) |
|
933 | 934 | { |
|
934 | 935 | Py_ssize_t i, j, len; |
|
935 | 936 | char *nothead = NULL; |
|
936 | 937 | PyObject *heads = NULL; |
|
937 | 938 | PyObject *filter = NULL; |
|
938 | 939 | PyObject *filteredrevs = Py_None; |
|
939 | 940 | |
|
940 | 941 | if (!PyArg_ParseTuple(args, "|O", &filteredrevs)) { |
|
941 | 942 | return NULL; |
|
942 | 943 | } |
|
943 | 944 | |
|
944 | 945 | if (self->headrevs && filteredrevs == self->filteredrevs) |
|
945 | 946 | return list_copy(self->headrevs); |
|
946 | 947 | |
|
947 | 948 | Py_DECREF(self->filteredrevs); |
|
948 | 949 | self->filteredrevs = filteredrevs; |
|
949 | 950 | Py_INCREF(filteredrevs); |
|
950 | 951 | |
|
951 | 952 | if (filteredrevs != Py_None) { |
|
952 | 953 | filter = PyObject_GetAttrString(filteredrevs, "__contains__"); |
|
953 | 954 | if (!filter) { |
|
954 | 955 | PyErr_SetString( |
|
955 | 956 | PyExc_TypeError, |
|
956 | 957 | "filteredrevs has no attribute __contains__"); |
|
957 | 958 | goto bail; |
|
958 | 959 | } |
|
959 | 960 | } |
|
960 | 961 | |
|
961 | 962 | len = index_length(self); |
|
962 | 963 | heads = PyList_New(0); |
|
963 | 964 | if (heads == NULL) |
|
964 | 965 | goto bail; |
|
965 | 966 | if (len == 0) { |
|
966 | 967 | PyObject *nullid = PyInt_FromLong(-1); |
|
967 | 968 | if (nullid == NULL || PyList_Append(heads, nullid) == -1) { |
|
968 | 969 | Py_XDECREF(nullid); |
|
969 | 970 | goto bail; |
|
970 | 971 | } |
|
971 | 972 | goto done; |
|
972 | 973 | } |
|
973 | 974 | |
|
974 | 975 | nothead = calloc(len, 1); |
|
975 | 976 | if (nothead == NULL) { |
|
976 | 977 | PyErr_NoMemory(); |
|
977 | 978 | goto bail; |
|
978 | 979 | } |
|
979 | 980 | |
|
980 | 981 | for (i = len - 1; i >= 0; i--) { |
|
981 | 982 | int isfiltered; |
|
982 | 983 | int parents[2]; |
|
983 | 984 | |
|
984 | 985 | /* If nothead[i] == 1, it means we've seen an unfiltered child |
|
985 | 986 | * of this node already, and therefore this node is not |
|
986 | 987 | * filtered. So we can skip the expensive check_filter step. |
|
987 | 988 | */ |
|
988 | 989 | if (nothead[i] != 1) { |
|
989 | 990 | isfiltered = check_filter(filter, i); |
|
990 | 991 | if (isfiltered == -1) { |
|
991 | 992 | PyErr_SetString(PyExc_TypeError, |
|
992 | 993 | "unable to check filter"); |
|
993 | 994 | goto bail; |
|
994 | 995 | } |
|
995 | 996 | |
|
996 | 997 | if (isfiltered) { |
|
997 | 998 | nothead[i] = 1; |
|
998 | 999 | continue; |
|
999 | 1000 | } |
|
1000 | 1001 | } |
|
1001 | 1002 | |
|
1002 | 1003 | if (index_get_parents(self, i, parents, (int)len - 1) < 0) |
|
1003 | 1004 | goto bail; |
|
1004 | 1005 | for (j = 0; j < 2; j++) { |
|
1005 | 1006 | if (parents[j] >= 0) |
|
1006 | 1007 | nothead[parents[j]] = 1; |
|
1007 | 1008 | } |
|
1008 | 1009 | } |
|
1009 | 1010 | |
|
1010 | 1011 | for (i = 0; i < len; i++) { |
|
1011 | 1012 | PyObject *head; |
|
1012 | 1013 | |
|
1013 | 1014 | if (nothead[i]) |
|
1014 | 1015 | continue; |
|
1015 | 1016 | head = PyInt_FromSsize_t(i); |
|
1016 | 1017 | if (head == NULL || PyList_Append(heads, head) == -1) { |
|
1017 | 1018 | Py_XDECREF(head); |
|
1018 | 1019 | goto bail; |
|
1019 | 1020 | } |
|
1020 | 1021 | } |
|
1021 | 1022 | |
|
1022 | 1023 | done: |
|
1023 | 1024 | self->headrevs = heads; |
|
1024 | 1025 | Py_XDECREF(filter); |
|
1025 | 1026 | free(nothead); |
|
1026 | 1027 | return list_copy(self->headrevs); |
|
1027 | 1028 | bail: |
|
1028 | 1029 | Py_XDECREF(filter); |
|
1029 | 1030 | Py_XDECREF(heads); |
|
1030 | 1031 | free(nothead); |
|
1031 | 1032 | return NULL; |
|
1032 | 1033 | } |
|
1033 | 1034 | |
|
1034 | 1035 | /** |
|
1035 | 1036 | * Obtain the base revision index entry. |
|
1036 | 1037 | * |
|
1037 | 1038 | * Callers must ensure that rev >= 0 or illegal memory access may occur. |
|
1038 | 1039 | */ |
|
1039 | 1040 | static inline int index_baserev(indexObject *self, int rev) |
|
1040 | 1041 | { |
|
1041 | 1042 | const char *data; |
|
1042 | 1043 | int result; |
|
1043 | 1044 | |
|
1044 | 1045 | data = index_deref(self, rev); |
|
1045 | 1046 | if (data == NULL) |
|
1046 | 1047 | return -2; |
|
1047 | 1048 | result = getbe32(data + 16); |
|
1048 | 1049 | |
|
1049 | 1050 | if (result > rev) { |
|
1050 | 1051 | PyErr_Format( |
|
1051 | 1052 | PyExc_ValueError, |
|
1052 | 1053 | "corrupted revlog, revision base above revision: %d, %d", |
|
1053 | 1054 | rev, result); |
|
1054 | 1055 | return -2; |
|
1055 | 1056 | } |
|
1056 | 1057 | if (result < -1) { |
|
1057 | 1058 | PyErr_Format( |
|
1058 | 1059 | PyExc_ValueError, |
|
1059 | 1060 | "corrupted revlog, revision base out of range: %d, %d", rev, |
|
1060 | 1061 | result); |
|
1061 | 1062 | return -2; |
|
1062 | 1063 | } |
|
1063 | 1064 | return result; |
|
1064 | 1065 | } |
|
1065 | 1066 | |
|
1066 | 1067 | /** |
|
1067 | 1068 | * Find if a revision is a snapshot or not |
|
1068 | 1069 | * |
|
1069 | 1070 | * Only relevant for sparse-revlog case. |
|
1070 | 1071 | * Callers must ensure that rev is in a valid range. |
|
1071 | 1072 | */ |
|
1072 | 1073 | static int index_issnapshotrev(indexObject *self, Py_ssize_t rev) |
|
1073 | 1074 | { |
|
1074 | 1075 | int ps[2]; |
|
1075 | 1076 | Py_ssize_t base; |
|
1076 | 1077 | while (rev >= 0) { |
|
1077 | 1078 | base = (Py_ssize_t)index_baserev(self, rev); |
|
1078 | 1079 | if (base == rev) { |
|
1079 | 1080 | base = -1; |
|
1080 | 1081 | } |
|
1081 | 1082 | if (base == -2) { |
|
1082 | 1083 | assert(PyErr_Occurred()); |
|
1083 | 1084 | return -1; |
|
1084 | 1085 | } |
|
1085 | 1086 | if (base == -1) { |
|
1086 | 1087 | return 1; |
|
1087 | 1088 | } |
|
1088 | 1089 | if (index_get_parents(self, rev, ps, (int)rev) < 0) { |
|
1089 | 1090 | assert(PyErr_Occurred()); |
|
1090 | 1091 | return -1; |
|
1091 | 1092 | }; |
|
1092 | 1093 | if (base == ps[0] || base == ps[1]) { |
|
1093 | 1094 | return 0; |
|
1094 | 1095 | } |
|
1095 | 1096 | rev = base; |
|
1096 | 1097 | } |
|
1097 | 1098 | return rev == -1; |
|
1098 | 1099 | } |
|
1099 | 1100 | |
|
1100 | 1101 | static PyObject *index_issnapshot(indexObject *self, PyObject *value) |
|
1101 | 1102 | { |
|
1102 | 1103 | long rev; |
|
1103 | 1104 | int issnap; |
|
1104 | 1105 | Py_ssize_t length = index_length(self); |
|
1105 | 1106 | |
|
1106 | 1107 | if (!pylong_to_long(value, &rev)) { |
|
1107 | 1108 | return NULL; |
|
1108 | 1109 | } |
|
1109 | 1110 | if (rev < -1 || rev >= length) { |
|
1110 | 1111 | PyErr_Format(PyExc_ValueError, "revlog index out of range: %ld", |
|
1111 | 1112 | rev); |
|
1112 | 1113 | return NULL; |
|
1113 | 1114 | }; |
|
1114 | 1115 | issnap = index_issnapshotrev(self, (Py_ssize_t)rev); |
|
1115 | 1116 | if (issnap < 0) { |
|
1116 | 1117 | return NULL; |
|
1117 | 1118 | }; |
|
1118 | 1119 | return PyBool_FromLong((long)issnap); |
|
1119 | 1120 | } |
|
1120 | 1121 | |
|
1121 | 1122 | static PyObject *index_findsnapshots(indexObject *self, PyObject *args) |
|
1122 | 1123 | { |
|
1123 | 1124 | Py_ssize_t start_rev; |
|
1124 | 1125 | PyObject *cache; |
|
1125 | 1126 | Py_ssize_t base; |
|
1126 | 1127 | Py_ssize_t rev; |
|
1127 | 1128 | PyObject *key = NULL; |
|
1128 | 1129 | PyObject *value = NULL; |
|
1129 | 1130 | const Py_ssize_t length = index_length(self); |
|
1130 | 1131 | if (!PyArg_ParseTuple(args, "O!n", &PyDict_Type, &cache, &start_rev)) { |
|
1131 | 1132 | return NULL; |
|
1132 | 1133 | } |
|
1133 | 1134 | for (rev = start_rev; rev < length; rev++) { |
|
1134 | 1135 | int issnap; |
|
1135 | 1136 | PyObject *allvalues = NULL; |
|
1136 | 1137 | issnap = index_issnapshotrev(self, rev); |
|
1137 | 1138 | if (issnap < 0) { |
|
1138 | 1139 | goto bail; |
|
1139 | 1140 | } |
|
1140 | 1141 | if (issnap == 0) { |
|
1141 | 1142 | continue; |
|
1142 | 1143 | } |
|
1143 | 1144 | base = (Py_ssize_t)index_baserev(self, rev); |
|
1144 | 1145 | if (base == rev) { |
|
1145 | 1146 | base = -1; |
|
1146 | 1147 | } |
|
1147 | 1148 | if (base == -2) { |
|
1148 | 1149 | assert(PyErr_Occurred()); |
|
1149 | 1150 | goto bail; |
|
1150 | 1151 | } |
|
1151 | 1152 | key = PyInt_FromSsize_t(base); |
|
1152 | 1153 | allvalues = PyDict_GetItem(cache, key); |
|
1153 | 1154 | if (allvalues == NULL && PyErr_Occurred()) { |
|
1154 | 1155 | goto bail; |
|
1155 | 1156 | } |
|
1156 | 1157 | if (allvalues == NULL) { |
|
1157 | 1158 | int r; |
|
1158 | 1159 | allvalues = PyList_New(0); |
|
1159 | 1160 | if (!allvalues) { |
|
1160 | 1161 | goto bail; |
|
1161 | 1162 | } |
|
1162 | 1163 | r = PyDict_SetItem(cache, key, allvalues); |
|
1163 | 1164 | Py_DECREF(allvalues); |
|
1164 | 1165 | if (r < 0) { |
|
1165 | 1166 | goto bail; |
|
1166 | 1167 | } |
|
1167 | 1168 | } |
|
1168 | 1169 | value = PyInt_FromSsize_t(rev); |
|
1169 | 1170 | if (PyList_Append(allvalues, value)) { |
|
1170 | 1171 | goto bail; |
|
1171 | 1172 | } |
|
1172 | 1173 | Py_CLEAR(key); |
|
1173 | 1174 | Py_CLEAR(value); |
|
1174 | 1175 | } |
|
1175 | 1176 | Py_RETURN_NONE; |
|
1176 | 1177 | bail: |
|
1177 | 1178 | Py_XDECREF(key); |
|
1178 | 1179 | Py_XDECREF(value); |
|
1179 | 1180 | return NULL; |
|
1180 | 1181 | } |
|
1181 | 1182 | |
|
1182 | 1183 | static PyObject *index_deltachain(indexObject *self, PyObject *args) |
|
1183 | 1184 | { |
|
1184 | 1185 | int rev, generaldelta; |
|
1185 | 1186 | PyObject *stoparg; |
|
1186 | 1187 | int stoprev, iterrev, baserev = -1; |
|
1187 | 1188 | int stopped; |
|
1188 | 1189 | PyObject *chain = NULL, *result = NULL; |
|
1189 | 1190 | const Py_ssize_t length = index_length(self); |
|
1190 | 1191 | |
|
1191 | 1192 | if (!PyArg_ParseTuple(args, "iOi", &rev, &stoparg, &generaldelta)) { |
|
1192 | 1193 | return NULL; |
|
1193 | 1194 | } |
|
1194 | 1195 | |
|
1195 | 1196 | if (PyInt_Check(stoparg)) { |
|
1196 | 1197 | stoprev = (int)PyInt_AsLong(stoparg); |
|
1197 | 1198 | if (stoprev == -1 && PyErr_Occurred()) { |
|
1198 | 1199 | return NULL; |
|
1199 | 1200 | } |
|
1200 | 1201 | } else if (stoparg == Py_None) { |
|
1201 | 1202 | stoprev = -2; |
|
1202 | 1203 | } else { |
|
1203 | 1204 | PyErr_SetString(PyExc_ValueError, |
|
1204 | 1205 | "stoprev must be integer or None"); |
|
1205 | 1206 | return NULL; |
|
1206 | 1207 | } |
|
1207 | 1208 | |
|
1208 | 1209 | if (rev < 0 || rev >= length) { |
|
1209 | 1210 | PyErr_SetString(PyExc_ValueError, "revlog index out of range"); |
|
1210 | 1211 | return NULL; |
|
1211 | 1212 | } |
|
1212 | 1213 | |
|
1213 | 1214 | chain = PyList_New(0); |
|
1214 | 1215 | if (chain == NULL) { |
|
1215 | 1216 | return NULL; |
|
1216 | 1217 | } |
|
1217 | 1218 | |
|
1218 | 1219 | baserev = index_baserev(self, rev); |
|
1219 | 1220 | |
|
1220 | 1221 | /* This should never happen. */ |
|
1221 | 1222 | if (baserev <= -2) { |
|
1222 | 1223 | /* Error should be set by index_deref() */ |
|
1223 | 1224 | assert(PyErr_Occurred()); |
|
1224 | 1225 | goto bail; |
|
1225 | 1226 | } |
|
1226 | 1227 | |
|
1227 | 1228 | iterrev = rev; |
|
1228 | 1229 | |
|
1229 | 1230 | while (iterrev != baserev && iterrev != stoprev) { |
|
1230 | 1231 | PyObject *value = PyInt_FromLong(iterrev); |
|
1231 | 1232 | if (value == NULL) { |
|
1232 | 1233 | goto bail; |
|
1233 | 1234 | } |
|
1234 | 1235 | if (PyList_Append(chain, value)) { |
|
1235 | 1236 | Py_DECREF(value); |
|
1236 | 1237 | goto bail; |
|
1237 | 1238 | } |
|
1238 | 1239 | Py_DECREF(value); |
|
1239 | 1240 | |
|
1240 | 1241 | if (generaldelta) { |
|
1241 | 1242 | iterrev = baserev; |
|
1242 | 1243 | } else { |
|
1243 | 1244 | iterrev--; |
|
1244 | 1245 | } |
|
1245 | 1246 | |
|
1246 | 1247 | if (iterrev < 0) { |
|
1247 | 1248 | break; |
|
1248 | 1249 | } |
|
1249 | 1250 | |
|
1250 | 1251 | if (iterrev >= length) { |
|
1251 | 1252 | PyErr_SetString(PyExc_IndexError, |
|
1252 | 1253 | "revision outside index"); |
|
1253 | 1254 | return NULL; |
|
1254 | 1255 | } |
|
1255 | 1256 | |
|
1256 | 1257 | baserev = index_baserev(self, iterrev); |
|
1257 | 1258 | |
|
1258 | 1259 | /* This should never happen. */ |
|
1259 | 1260 | if (baserev <= -2) { |
|
1260 | 1261 | /* Error should be set by index_deref() */ |
|
1261 | 1262 | assert(PyErr_Occurred()); |
|
1262 | 1263 | goto bail; |
|
1263 | 1264 | } |
|
1264 | 1265 | } |
|
1265 | 1266 | |
|
1266 | 1267 | if (iterrev == stoprev) { |
|
1267 | 1268 | stopped = 1; |
|
1268 | 1269 | } else { |
|
1269 | 1270 | PyObject *value = PyInt_FromLong(iterrev); |
|
1270 | 1271 | if (value == NULL) { |
|
1271 | 1272 | goto bail; |
|
1272 | 1273 | } |
|
1273 | 1274 | if (PyList_Append(chain, value)) { |
|
1274 | 1275 | Py_DECREF(value); |
|
1275 | 1276 | goto bail; |
|
1276 | 1277 | } |
|
1277 | 1278 | Py_DECREF(value); |
|
1278 | 1279 | |
|
1279 | 1280 | stopped = 0; |
|
1280 | 1281 | } |
|
1281 | 1282 | |
|
1282 | 1283 | if (PyList_Reverse(chain)) { |
|
1283 | 1284 | goto bail; |
|
1284 | 1285 | } |
|
1285 | 1286 | |
|
1286 | 1287 | result = Py_BuildValue("OO", chain, stopped ? Py_True : Py_False); |
|
1287 | 1288 | Py_DECREF(chain); |
|
1288 | 1289 | return result; |
|
1289 | 1290 | |
|
1290 | 1291 | bail: |
|
1291 | 1292 | Py_DECREF(chain); |
|
1292 | 1293 | return NULL; |
|
1293 | 1294 | } |
|
1294 | 1295 | |
|
1295 | 1296 | static inline int64_t |
|
1296 | 1297 | index_segment_span(indexObject *self, Py_ssize_t start_rev, Py_ssize_t end_rev) |
|
1297 | 1298 | { |
|
1298 | 1299 | int64_t start_offset; |
|
1299 | 1300 | int64_t end_offset; |
|
1300 | 1301 | int end_size; |
|
1301 | 1302 | start_offset = index_get_start(self, start_rev); |
|
1302 | 1303 | if (start_offset < 0) { |
|
1303 | 1304 | return -1; |
|
1304 | 1305 | } |
|
1305 | 1306 | end_offset = index_get_start(self, end_rev); |
|
1306 | 1307 | if (end_offset < 0) { |
|
1307 | 1308 | return -1; |
|
1308 | 1309 | } |
|
1309 | 1310 | end_size = index_get_length(self, end_rev); |
|
1310 | 1311 | if (end_size < 0) { |
|
1311 | 1312 | return -1; |
|
1312 | 1313 | } |
|
1313 | 1314 | if (end_offset < start_offset) { |
|
1314 | 1315 | PyErr_Format(PyExc_ValueError, |
|
1315 | 1316 | "corrupted revlog index: inconsistent offset " |
|
1316 | 1317 | "between revisions (%zd) and (%zd)", |
|
1317 | 1318 | start_rev, end_rev); |
|
1318 | 1319 | return -1; |
|
1319 | 1320 | } |
|
1320 | 1321 | return (end_offset - start_offset) + (int64_t)end_size; |
|
1321 | 1322 | } |
|
1322 | 1323 | |
|
1323 | 1324 | /* returns endidx so that revs[startidx:endidx] has no empty trailing revs */ |
|
1324 | 1325 | static Py_ssize_t trim_endidx(indexObject *self, const Py_ssize_t *revs, |
|
1325 | 1326 | Py_ssize_t startidx, Py_ssize_t endidx) |
|
1326 | 1327 | { |
|
1327 | 1328 | int length; |
|
1328 | 1329 | while (endidx > 1 && endidx > startidx) { |
|
1329 | 1330 | length = index_get_length(self, revs[endidx - 1]); |
|
1330 | 1331 | if (length < 0) { |
|
1331 | 1332 | return -1; |
|
1332 | 1333 | } |
|
1333 | 1334 | if (length != 0) { |
|
1334 | 1335 | break; |
|
1335 | 1336 | } |
|
1336 | 1337 | endidx -= 1; |
|
1337 | 1338 | } |
|
1338 | 1339 | return endidx; |
|
1339 | 1340 | } |
|
1340 | 1341 | |
|
1341 | 1342 | struct Gap { |
|
1342 | 1343 | int64_t size; |
|
1343 | 1344 | Py_ssize_t idx; |
|
1344 | 1345 | }; |
|
1345 | 1346 | |
|
1346 | 1347 | static int gap_compare(const void *left, const void *right) |
|
1347 | 1348 | { |
|
1348 | 1349 | const struct Gap *l_left = ((const struct Gap *)left); |
|
1349 | 1350 | const struct Gap *l_right = ((const struct Gap *)right); |
|
1350 | 1351 | if (l_left->size < l_right->size) { |
|
1351 | 1352 | return -1; |
|
1352 | 1353 | } else if (l_left->size > l_right->size) { |
|
1353 | 1354 | return 1; |
|
1354 | 1355 | } |
|
1355 | 1356 | return 0; |
|
1356 | 1357 | } |
|
1357 | 1358 | static int Py_ssize_t_compare(const void *left, const void *right) |
|
1358 | 1359 | { |
|
1359 | 1360 | const Py_ssize_t l_left = *(const Py_ssize_t *)left; |
|
1360 | 1361 | const Py_ssize_t l_right = *(const Py_ssize_t *)right; |
|
1361 | 1362 | if (l_left < l_right) { |
|
1362 | 1363 | return -1; |
|
1363 | 1364 | } else if (l_left > l_right) { |
|
1364 | 1365 | return 1; |
|
1365 | 1366 | } |
|
1366 | 1367 | return 0; |
|
1367 | 1368 | } |
|
1368 | 1369 | |
|
1369 | 1370 | static PyObject *index_slicechunktodensity(indexObject *self, PyObject *args) |
|
1370 | 1371 | { |
|
1371 | 1372 | /* method arguments */ |
|
1372 | 1373 | PyObject *list_revs = NULL; /* revisions in the chain */ |
|
1373 | 1374 | double targetdensity = 0; /* min density to achieve */ |
|
1374 | 1375 | Py_ssize_t mingapsize = 0; /* threshold to ignore gaps */ |
|
1375 | 1376 | |
|
1376 | 1377 | /* other core variables */ |
|
1377 | 1378 | Py_ssize_t idxlen = index_length(self); |
|
1378 | 1379 | Py_ssize_t i; /* used for various iteration */ |
|
1379 | 1380 | PyObject *result = NULL; /* the final return of the function */ |
|
1380 | 1381 | |
|
1381 | 1382 | /* generic information about the delta chain being slice */ |
|
1382 | 1383 | Py_ssize_t num_revs = 0; /* size of the full delta chain */ |
|
1383 | 1384 | Py_ssize_t *revs = NULL; /* native array of revision in the chain */ |
|
1384 | 1385 | int64_t chainpayload = 0; /* sum of all delta in the chain */ |
|
1385 | 1386 | int64_t deltachainspan = 0; /* distance from first byte to last byte */ |
|
1386 | 1387 | |
|
1387 | 1388 | /* variable used for slicing the delta chain */ |
|
1388 | 1389 | int64_t readdata = 0; /* amount of data currently planned to be read */ |
|
1389 | 1390 | double density = 0; /* ration of payload data compared to read ones */ |
|
1390 | 1391 | int64_t previous_end; |
|
1391 | 1392 | struct Gap *gaps = NULL; /* array of notable gap in the chain */ |
|
1392 | 1393 | Py_ssize_t num_gaps = |
|
1393 | 1394 | 0; /* total number of notable gap recorded so far */ |
|
1394 | 1395 | Py_ssize_t *selected_indices = NULL; /* indices of gap skipped over */ |
|
1395 | 1396 | Py_ssize_t num_selected = 0; /* number of gaps skipped */ |
|
1396 | 1397 | PyObject *chunk = NULL; /* individual slice */ |
|
1397 | 1398 | PyObject *allchunks = NULL; /* all slices */ |
|
1398 | 1399 | Py_ssize_t previdx; |
|
1399 | 1400 | |
|
1400 | 1401 | /* parsing argument */ |
|
1401 | 1402 | if (!PyArg_ParseTuple(args, "O!dn", &PyList_Type, &list_revs, |
|
1402 | 1403 | &targetdensity, &mingapsize)) { |
|
1403 | 1404 | goto bail; |
|
1404 | 1405 | } |
|
1405 | 1406 | |
|
1406 | 1407 | /* If the delta chain contains a single element, we do not need slicing |
|
1407 | 1408 | */ |
|
1408 | 1409 | num_revs = PyList_GET_SIZE(list_revs); |
|
1409 | 1410 | if (num_revs <= 1) { |
|
1410 | 1411 | result = PyTuple_Pack(1, list_revs); |
|
1411 | 1412 | goto done; |
|
1412 | 1413 | } |
|
1413 | 1414 | |
|
1414 | 1415 | /* Turn the python list into a native integer array (for efficiency) */ |
|
1415 | 1416 | revs = (Py_ssize_t *)calloc(num_revs, sizeof(Py_ssize_t)); |
|
1416 | 1417 | if (revs == NULL) { |
|
1417 | 1418 | PyErr_NoMemory(); |
|
1418 | 1419 | goto bail; |
|
1419 | 1420 | } |
|
1420 | 1421 | for (i = 0; i < num_revs; i++) { |
|
1421 | 1422 | Py_ssize_t revnum = PyInt_AsLong(PyList_GET_ITEM(list_revs, i)); |
|
1422 | 1423 | if (revnum == -1 && PyErr_Occurred()) { |
|
1423 | 1424 | goto bail; |
|
1424 | 1425 | } |
|
1425 | 1426 | if (revnum < nullrev || revnum >= idxlen) { |
|
1426 | 1427 | PyErr_Format(PyExc_IndexError, |
|
1427 | 1428 | "index out of range: %zd", revnum); |
|
1428 | 1429 | goto bail; |
|
1429 | 1430 | } |
|
1430 | 1431 | revs[i] = revnum; |
|
1431 | 1432 | } |
|
1432 | 1433 | |
|
1433 | 1434 | /* Compute and check various property of the unsliced delta chain */ |
|
1434 | 1435 | deltachainspan = index_segment_span(self, revs[0], revs[num_revs - 1]); |
|
1435 | 1436 | if (deltachainspan < 0) { |
|
1436 | 1437 | goto bail; |
|
1437 | 1438 | } |
|
1438 | 1439 | |
|
1439 | 1440 | if (deltachainspan <= mingapsize) { |
|
1440 | 1441 | result = PyTuple_Pack(1, list_revs); |
|
1441 | 1442 | goto done; |
|
1442 | 1443 | } |
|
1443 | 1444 | chainpayload = 0; |
|
1444 | 1445 | for (i = 0; i < num_revs; i++) { |
|
1445 | 1446 | int tmp = index_get_length(self, revs[i]); |
|
1446 | 1447 | if (tmp < 0) { |
|
1447 | 1448 | goto bail; |
|
1448 | 1449 | } |
|
1449 | 1450 | chainpayload += tmp; |
|
1450 | 1451 | } |
|
1451 | 1452 | |
|
1452 | 1453 | readdata = deltachainspan; |
|
1453 | 1454 | density = 1.0; |
|
1454 | 1455 | |
|
1455 | 1456 | if (0 < deltachainspan) { |
|
1456 | 1457 | density = (double)chainpayload / (double)deltachainspan; |
|
1457 | 1458 | } |
|
1458 | 1459 | |
|
1459 | 1460 | if (density >= targetdensity) { |
|
1460 | 1461 | result = PyTuple_Pack(1, list_revs); |
|
1461 | 1462 | goto done; |
|
1462 | 1463 | } |
|
1463 | 1464 | |
|
1464 | 1465 | /* if chain is too sparse, look for relevant gaps */ |
|
1465 | 1466 | gaps = (struct Gap *)calloc(num_revs, sizeof(struct Gap)); |
|
1466 | 1467 | if (gaps == NULL) { |
|
1467 | 1468 | PyErr_NoMemory(); |
|
1468 | 1469 | goto bail; |
|
1469 | 1470 | } |
|
1470 | 1471 | |
|
1471 | 1472 | previous_end = -1; |
|
1472 | 1473 | for (i = 0; i < num_revs; i++) { |
|
1473 | 1474 | int64_t revstart; |
|
1474 | 1475 | int revsize; |
|
1475 | 1476 | revstart = index_get_start(self, revs[i]); |
|
1476 | 1477 | if (revstart < 0) { |
|
1477 | 1478 | goto bail; |
|
1478 | 1479 | }; |
|
1479 | 1480 | revsize = index_get_length(self, revs[i]); |
|
1480 | 1481 | if (revsize < 0) { |
|
1481 | 1482 | goto bail; |
|
1482 | 1483 | }; |
|
1483 | 1484 | if (revsize == 0) { |
|
1484 | 1485 | continue; |
|
1485 | 1486 | } |
|
1486 | 1487 | if (previous_end >= 0) { |
|
1487 | 1488 | int64_t gapsize = revstart - previous_end; |
|
1488 | 1489 | if (gapsize > mingapsize) { |
|
1489 | 1490 | gaps[num_gaps].size = gapsize; |
|
1490 | 1491 | gaps[num_gaps].idx = i; |
|
1491 | 1492 | num_gaps += 1; |
|
1492 | 1493 | } |
|
1493 | 1494 | } |
|
1494 | 1495 | previous_end = revstart + revsize; |
|
1495 | 1496 | } |
|
1496 | 1497 | if (num_gaps == 0) { |
|
1497 | 1498 | result = PyTuple_Pack(1, list_revs); |
|
1498 | 1499 | goto done; |
|
1499 | 1500 | } |
|
1500 | 1501 | qsort(gaps, num_gaps, sizeof(struct Gap), &gap_compare); |
|
1501 | 1502 | |
|
1502 | 1503 | /* Slice the largest gap first, they improve the density the most */ |
|
1503 | 1504 | selected_indices = |
|
1504 | 1505 | (Py_ssize_t *)malloc((num_gaps + 1) * sizeof(Py_ssize_t)); |
|
1505 | 1506 | if (selected_indices == NULL) { |
|
1506 | 1507 | PyErr_NoMemory(); |
|
1507 | 1508 | goto bail; |
|
1508 | 1509 | } |
|
1509 | 1510 | |
|
1510 | 1511 | for (i = num_gaps - 1; i >= 0; i--) { |
|
1511 | 1512 | selected_indices[num_selected] = gaps[i].idx; |
|
1512 | 1513 | readdata -= gaps[i].size; |
|
1513 | 1514 | num_selected += 1; |
|
1514 | 1515 | if (readdata <= 0) { |
|
1515 | 1516 | density = 1.0; |
|
1516 | 1517 | } else { |
|
1517 | 1518 | density = (double)chainpayload / (double)readdata; |
|
1518 | 1519 | } |
|
1519 | 1520 | if (density >= targetdensity) { |
|
1520 | 1521 | break; |
|
1521 | 1522 | } |
|
1522 | 1523 | } |
|
1523 | 1524 | qsort(selected_indices, num_selected, sizeof(Py_ssize_t), |
|
1524 | 1525 | &Py_ssize_t_compare); |
|
1525 | 1526 | |
|
1526 | 1527 | /* create the resulting slice */ |
|
1527 | 1528 | allchunks = PyList_New(0); |
|
1528 | 1529 | if (allchunks == NULL) { |
|
1529 | 1530 | goto bail; |
|
1530 | 1531 | } |
|
1531 | 1532 | previdx = 0; |
|
1532 | 1533 | selected_indices[num_selected] = num_revs; |
|
1533 | 1534 | for (i = 0; i <= num_selected; i++) { |
|
1534 | 1535 | Py_ssize_t idx = selected_indices[i]; |
|
1535 | 1536 | Py_ssize_t endidx = trim_endidx(self, revs, previdx, idx); |
|
1536 | 1537 | if (endidx < 0) { |
|
1537 | 1538 | goto bail; |
|
1538 | 1539 | } |
|
1539 | 1540 | if (previdx < endidx) { |
|
1540 | 1541 | chunk = PyList_GetSlice(list_revs, previdx, endidx); |
|
1541 | 1542 | if (chunk == NULL) { |
|
1542 | 1543 | goto bail; |
|
1543 | 1544 | } |
|
1544 | 1545 | if (PyList_Append(allchunks, chunk) == -1) { |
|
1545 | 1546 | goto bail; |
|
1546 | 1547 | } |
|
1547 | 1548 | Py_DECREF(chunk); |
|
1548 | 1549 | chunk = NULL; |
|
1549 | 1550 | } |
|
1550 | 1551 | previdx = idx; |
|
1551 | 1552 | } |
|
1552 | 1553 | result = allchunks; |
|
1553 | 1554 | goto done; |
|
1554 | 1555 | |
|
1555 | 1556 | bail: |
|
1556 | 1557 | Py_XDECREF(allchunks); |
|
1557 | 1558 | Py_XDECREF(chunk); |
|
1558 | 1559 | done: |
|
1559 | 1560 | free(revs); |
|
1560 | 1561 | free(gaps); |
|
1561 | 1562 | free(selected_indices); |
|
1562 | 1563 | return result; |
|
1563 | 1564 | } |
|
1564 | 1565 | |
|
1565 | 1566 | static inline int nt_level(const char *node, Py_ssize_t level) |
|
1566 | 1567 | { |
|
1567 | 1568 | int v = node[level >> 1]; |
|
1568 | 1569 | if (!(level & 1)) |
|
1569 | 1570 | v >>= 4; |
|
1570 | 1571 | return v & 0xf; |
|
1571 | 1572 | } |
|
1572 | 1573 | |
|
1573 | 1574 | /* |
|
1574 | 1575 | * Return values: |
|
1575 | 1576 | * |
|
1576 | 1577 | * -4: match is ambiguous (multiple candidates) |
|
1577 | 1578 | * -2: not found |
|
1578 | 1579 | * rest: valid rev |
|
1579 | 1580 | */ |
|
1580 | 1581 | static int nt_find(nodetree *self, const char *node, Py_ssize_t nodelen, |
|
1581 | 1582 | int hex) |
|
1582 | 1583 | { |
|
1583 | 1584 | int (*getnybble)(const char *, Py_ssize_t) = hex ? hexdigit : nt_level; |
|
1584 | 1585 | int level, maxlevel, off; |
|
1585 | 1586 | |
|
1586 | 1587 | /* If the input is binary, do a fast check for the nullid first. */ |
|
1587 | 1588 | if (!hex && nodelen == self->nodelen && node[0] == '\0' && |
|
1588 | 1589 | node[1] == '\0' && memcmp(node, nullid, self->nodelen) == 0) |
|
1589 | 1590 | return -1; |
|
1590 | 1591 | |
|
1591 | 1592 | if (hex) |
|
1592 | 1593 | maxlevel = nodelen; |
|
1593 | 1594 | else |
|
1594 | 1595 | maxlevel = 2 * nodelen; |
|
1595 | 1596 | if (maxlevel > 2 * self->nodelen) |
|
1596 | 1597 | maxlevel = 2 * self->nodelen; |
|
1597 | 1598 | |
|
1598 | 1599 | for (level = off = 0; level < maxlevel; level++) { |
|
1599 | 1600 | int k = getnybble(node, level); |
|
1600 | 1601 | nodetreenode *n = &self->nodes[off]; |
|
1601 | 1602 | int v = n->children[k]; |
|
1602 | 1603 | |
|
1603 | 1604 | if (v < 0) { |
|
1604 | 1605 | const char *n; |
|
1605 | 1606 | Py_ssize_t i; |
|
1606 | 1607 | |
|
1607 | 1608 | v = -(v + 2); |
|
1608 | 1609 | n = index_node(self->index, v); |
|
1609 | 1610 | if (n == NULL) |
|
1610 | 1611 | return -2; |
|
1611 | 1612 | for (i = level; i < maxlevel; i++) |
|
1612 | 1613 | if (getnybble(node, i) != nt_level(n, i)) |
|
1613 | 1614 | return -2; |
|
1614 | 1615 | return v; |
|
1615 | 1616 | } |
|
1616 | 1617 | if (v == 0) |
|
1617 | 1618 | return -2; |
|
1618 | 1619 | off = v; |
|
1619 | 1620 | } |
|
1620 | 1621 | /* multiple matches against an ambiguous prefix */ |
|
1621 | 1622 | return -4; |
|
1622 | 1623 | } |
|
1623 | 1624 | |
|
1624 | 1625 | static int nt_new(nodetree *self) |
|
1625 | 1626 | { |
|
1626 | 1627 | if (self->length == self->capacity) { |
|
1627 | 1628 | size_t newcapacity; |
|
1628 | 1629 | nodetreenode *newnodes; |
|
1629 | 1630 | newcapacity = self->capacity * 2; |
|
1630 | 1631 | if (newcapacity >= SIZE_MAX / sizeof(nodetreenode)) { |
|
1631 | 1632 | PyErr_SetString(PyExc_MemoryError, |
|
1632 | 1633 | "overflow in nt_new"); |
|
1633 | 1634 | return -1; |
|
1634 | 1635 | } |
|
1635 | 1636 | newnodes = |
|
1636 | 1637 | realloc(self->nodes, newcapacity * sizeof(nodetreenode)); |
|
1637 | 1638 | if (newnodes == NULL) { |
|
1638 | 1639 | PyErr_SetString(PyExc_MemoryError, "out of memory"); |
|
1639 | 1640 | return -1; |
|
1640 | 1641 | } |
|
1641 | 1642 | self->capacity = newcapacity; |
|
1642 | 1643 | self->nodes = newnodes; |
|
1643 | 1644 | memset(&self->nodes[self->length], 0, |
|
1644 | 1645 | sizeof(nodetreenode) * (self->capacity - self->length)); |
|
1645 | 1646 | } |
|
1646 | 1647 | return self->length++; |
|
1647 | 1648 | } |
|
1648 | 1649 | |
|
1649 | 1650 | static int nt_insert(nodetree *self, const char *node, int rev) |
|
1650 | 1651 | { |
|
1651 | 1652 | int level = 0; |
|
1652 | 1653 | int off = 0; |
|
1653 | 1654 | |
|
1654 | 1655 | while (level < 2 * self->nodelen) { |
|
1655 | 1656 | int k = nt_level(node, level); |
|
1656 | 1657 | nodetreenode *n; |
|
1657 | 1658 | int v; |
|
1658 | 1659 | |
|
1659 | 1660 | n = &self->nodes[off]; |
|
1660 | 1661 | v = n->children[k]; |
|
1661 | 1662 | |
|
1662 | 1663 | if (v == 0) { |
|
1663 | 1664 | n->children[k] = -rev - 2; |
|
1664 | 1665 | return 0; |
|
1665 | 1666 | } |
|
1666 | 1667 | if (v < 0) { |
|
1667 | 1668 | const char *oldnode = |
|
1668 | 1669 | index_node_existing(self->index, -(v + 2)); |
|
1669 | 1670 | int noff; |
|
1670 | 1671 | |
|
1671 | 1672 | if (oldnode == NULL) |
|
1672 | 1673 | return -1; |
|
1673 | 1674 | if (!memcmp(oldnode, node, self->nodelen)) { |
|
1674 | 1675 | n->children[k] = -rev - 2; |
|
1675 | 1676 | return 0; |
|
1676 | 1677 | } |
|
1677 | 1678 | noff = nt_new(self); |
|
1678 | 1679 | if (noff == -1) |
|
1679 | 1680 | return -1; |
|
1680 | 1681 | /* self->nodes may have been changed by realloc */ |
|
1681 | 1682 | self->nodes[off].children[k] = noff; |
|
1682 | 1683 | off = noff; |
|
1683 | 1684 | n = &self->nodes[off]; |
|
1684 | 1685 | n->children[nt_level(oldnode, ++level)] = v; |
|
1685 | 1686 | if (level > self->depth) |
|
1686 | 1687 | self->depth = level; |
|
1687 | 1688 | self->splits += 1; |
|
1688 | 1689 | } else { |
|
1689 | 1690 | level += 1; |
|
1690 | 1691 | off = v; |
|
1691 | 1692 | } |
|
1692 | 1693 | } |
|
1693 | 1694 | |
|
1694 | 1695 | return -1; |
|
1695 | 1696 | } |
|
1696 | 1697 | |
|
1697 | 1698 | static PyObject *ntobj_insert(nodetreeObject *self, PyObject *args) |
|
1698 | 1699 | { |
|
1699 | 1700 | Py_ssize_t rev; |
|
1700 | 1701 | const char *node; |
|
1701 | 1702 | Py_ssize_t length; |
|
1702 | 1703 | if (!PyArg_ParseTuple(args, "n", &rev)) |
|
1703 | 1704 | return NULL; |
|
1704 | 1705 | length = index_length(self->nt.index); |
|
1705 | 1706 | if (rev < 0 || rev >= length) { |
|
1706 | 1707 | PyErr_SetString(PyExc_ValueError, "revlog index out of range"); |
|
1707 | 1708 | return NULL; |
|
1708 | 1709 | } |
|
1709 | 1710 | node = index_node_existing(self->nt.index, rev); |
|
1710 | 1711 | if (nt_insert(&self->nt, node, (int)rev) == -1) |
|
1711 | 1712 | return NULL; |
|
1712 | 1713 | Py_RETURN_NONE; |
|
1713 | 1714 | } |
|
1714 | 1715 | |
|
1715 | 1716 | static int nt_delete_node(nodetree *self, const char *node) |
|
1716 | 1717 | { |
|
1717 | 1718 | /* rev==-2 happens to get encoded as 0, which is interpreted as not set |
|
1718 | 1719 | */ |
|
1719 | 1720 | return nt_insert(self, node, -2); |
|
1720 | 1721 | } |
|
1721 | 1722 | |
|
1722 | 1723 | static int nt_init(nodetree *self, indexObject *index, unsigned capacity) |
|
1723 | 1724 | { |
|
1724 | 1725 | /* Initialize before overflow-checking to avoid nt_dealloc() crash. */ |
|
1725 | 1726 | self->nodes = NULL; |
|
1726 | 1727 | |
|
1727 | 1728 | self->index = index; |
|
1728 | 1729 | /* The input capacity is in terms of revisions, while the field is in |
|
1729 | 1730 | * terms of nodetree nodes. */ |
|
1730 | 1731 | self->capacity = (capacity < 4 ? 4 : capacity / 2); |
|
1731 | 1732 | self->nodelen = index->nodelen; |
|
1732 | 1733 | self->depth = 0; |
|
1733 | 1734 | self->splits = 0; |
|
1734 | 1735 | if (self->capacity > SIZE_MAX / sizeof(nodetreenode)) { |
|
1735 | 1736 | PyErr_SetString(PyExc_ValueError, "overflow in init_nt"); |
|
1736 | 1737 | return -1; |
|
1737 | 1738 | } |
|
1738 | 1739 | self->nodes = calloc(self->capacity, sizeof(nodetreenode)); |
|
1739 | 1740 | if (self->nodes == NULL) { |
|
1740 | 1741 | PyErr_NoMemory(); |
|
1741 | 1742 | return -1; |
|
1742 | 1743 | } |
|
1743 | 1744 | self->length = 1; |
|
1744 | 1745 | return 0; |
|
1745 | 1746 | } |
|
1746 | 1747 | |
|
1747 | 1748 | static int ntobj_init(nodetreeObject *self, PyObject *args) |
|
1748 | 1749 | { |
|
1749 | 1750 | PyObject *index; |
|
1750 | 1751 | unsigned capacity; |
|
1751 | 1752 | if (!PyArg_ParseTuple(args, "O!I", &HgRevlogIndex_Type, &index, |
|
1752 | 1753 | &capacity)) |
|
1753 | 1754 | return -1; |
|
1754 | 1755 | Py_INCREF(index); |
|
1755 | 1756 | return nt_init(&self->nt, (indexObject *)index, capacity); |
|
1756 | 1757 | } |
|
1757 | 1758 | |
|
1758 | 1759 | static int nt_partialmatch(nodetree *self, const char *node, Py_ssize_t nodelen) |
|
1759 | 1760 | { |
|
1760 | 1761 | return nt_find(self, node, nodelen, 1); |
|
1761 | 1762 | } |
|
1762 | 1763 | |
|
1763 | 1764 | /* |
|
1764 | 1765 | * Find the length of the shortest unique prefix of node. |
|
1765 | 1766 | * |
|
1766 | 1767 | * Return values: |
|
1767 | 1768 | * |
|
1768 | 1769 | * -3: error (exception set) |
|
1769 | 1770 | * -2: not found (no exception set) |
|
1770 | 1771 | * rest: length of shortest prefix |
|
1771 | 1772 | */ |
|
1772 | 1773 | static int nt_shortest(nodetree *self, const char *node) |
|
1773 | 1774 | { |
|
1774 | 1775 | int level, off; |
|
1775 | 1776 | |
|
1776 | 1777 | for (level = off = 0; level < 2 * self->nodelen; level++) { |
|
1777 | 1778 | int k, v; |
|
1778 | 1779 | nodetreenode *n = &self->nodes[off]; |
|
1779 | 1780 | k = nt_level(node, level); |
|
1780 | 1781 | v = n->children[k]; |
|
1781 | 1782 | if (v < 0) { |
|
1782 | 1783 | const char *n; |
|
1783 | 1784 | v = -(v + 2); |
|
1784 | 1785 | n = index_node_existing(self->index, v); |
|
1785 | 1786 | if (n == NULL) |
|
1786 | 1787 | return -3; |
|
1787 | 1788 | if (memcmp(node, n, self->nodelen) != 0) |
|
1788 | 1789 | /* |
|
1789 | 1790 | * Found a unique prefix, but it wasn't for the |
|
1790 | 1791 | * requested node (i.e the requested node does |
|
1791 | 1792 | * not exist). |
|
1792 | 1793 | */ |
|
1793 | 1794 | return -2; |
|
1794 | 1795 | return level + 1; |
|
1795 | 1796 | } |
|
1796 | 1797 | if (v == 0) |
|
1797 | 1798 | return -2; |
|
1798 | 1799 | off = v; |
|
1799 | 1800 | } |
|
1800 | 1801 | /* |
|
1801 | 1802 | * The node was still not unique after 40 hex digits, so this won't |
|
1802 | 1803 | * happen. Also, if we get here, then there's a programming error in |
|
1803 | 1804 | * this file that made us insert a node longer than 40 hex digits. |
|
1804 | 1805 | */ |
|
1805 | 1806 | PyErr_SetString(PyExc_Exception, "broken node tree"); |
|
1806 | 1807 | return -3; |
|
1807 | 1808 | } |
|
1808 | 1809 | |
|
1809 | 1810 | static PyObject *ntobj_shortest(nodetreeObject *self, PyObject *args) |
|
1810 | 1811 | { |
|
1811 | 1812 | PyObject *val; |
|
1812 | 1813 | char *node; |
|
1813 | 1814 | int length; |
|
1814 | 1815 | |
|
1815 | 1816 | if (!PyArg_ParseTuple(args, "O", &val)) |
|
1816 | 1817 | return NULL; |
|
1817 | 1818 | if (node_check(self->nt.nodelen, val, &node) == -1) |
|
1818 | 1819 | return NULL; |
|
1819 | 1820 | |
|
1820 | 1821 | length = nt_shortest(&self->nt, node); |
|
1821 | 1822 | if (length == -3) |
|
1822 | 1823 | return NULL; |
|
1823 | 1824 | if (length == -2) { |
|
1824 | 1825 | raise_revlog_error(); |
|
1825 | 1826 | return NULL; |
|
1826 | 1827 | } |
|
1827 | 1828 | return PyInt_FromLong(length); |
|
1828 | 1829 | } |
|
1829 | 1830 | |
|
1830 | 1831 | static void nt_dealloc(nodetree *self) |
|
1831 | 1832 | { |
|
1832 | 1833 | free(self->nodes); |
|
1833 | 1834 | self->nodes = NULL; |
|
1834 | 1835 | } |
|
1835 | 1836 | |
|
1836 | 1837 | static void ntobj_dealloc(nodetreeObject *self) |
|
1837 | 1838 | { |
|
1838 | 1839 | Py_XDECREF(self->nt.index); |
|
1839 | 1840 | nt_dealloc(&self->nt); |
|
1840 | 1841 | PyObject_Del(self); |
|
1841 | 1842 | } |
|
1842 | 1843 | |
|
1843 | 1844 | static PyMethodDef ntobj_methods[] = { |
|
1844 | 1845 | {"insert", (PyCFunction)ntobj_insert, METH_VARARGS, |
|
1845 | 1846 | "insert an index entry"}, |
|
1846 | 1847 | {"shortest", (PyCFunction)ntobj_shortest, METH_VARARGS, |
|
1847 | 1848 | "find length of shortest hex nodeid of a binary ID"}, |
|
1848 | 1849 | {NULL} /* Sentinel */ |
|
1849 | 1850 | }; |
|
1850 | 1851 | |
|
1851 | 1852 | static PyTypeObject nodetreeType = { |
|
1852 | 1853 | PyVarObject_HEAD_INIT(NULL, 0) /* header */ |
|
1853 | 1854 | "parsers.nodetree", /* tp_name */ |
|
1854 | 1855 | sizeof(nodetreeObject), /* tp_basicsize */ |
|
1855 | 1856 | 0, /* tp_itemsize */ |
|
1856 | 1857 | (destructor)ntobj_dealloc, /* tp_dealloc */ |
|
1857 | 1858 | 0, /* tp_print */ |
|
1858 | 1859 | 0, /* tp_getattr */ |
|
1859 | 1860 | 0, /* tp_setattr */ |
|
1860 | 1861 | 0, /* tp_compare */ |
|
1861 | 1862 | 0, /* tp_repr */ |
|
1862 | 1863 | 0, /* tp_as_number */ |
|
1863 | 1864 | 0, /* tp_as_sequence */ |
|
1864 | 1865 | 0, /* tp_as_mapping */ |
|
1865 | 1866 | 0, /* tp_hash */ |
|
1866 | 1867 | 0, /* tp_call */ |
|
1867 | 1868 | 0, /* tp_str */ |
|
1868 | 1869 | 0, /* tp_getattro */ |
|
1869 | 1870 | 0, /* tp_setattro */ |
|
1870 | 1871 | 0, /* tp_as_buffer */ |
|
1871 | 1872 | Py_TPFLAGS_DEFAULT, /* tp_flags */ |
|
1872 | 1873 | "nodetree", /* tp_doc */ |
|
1873 | 1874 | 0, /* tp_traverse */ |
|
1874 | 1875 | 0, /* tp_clear */ |
|
1875 | 1876 | 0, /* tp_richcompare */ |
|
1876 | 1877 | 0, /* tp_weaklistoffset */ |
|
1877 | 1878 | 0, /* tp_iter */ |
|
1878 | 1879 | 0, /* tp_iternext */ |
|
1879 | 1880 | ntobj_methods, /* tp_methods */ |
|
1880 | 1881 | 0, /* tp_members */ |
|
1881 | 1882 | 0, /* tp_getset */ |
|
1882 | 1883 | 0, /* tp_base */ |
|
1883 | 1884 | 0, /* tp_dict */ |
|
1884 | 1885 | 0, /* tp_descr_get */ |
|
1885 | 1886 | 0, /* tp_descr_set */ |
|
1886 | 1887 | 0, /* tp_dictoffset */ |
|
1887 | 1888 | (initproc)ntobj_init, /* tp_init */ |
|
1888 | 1889 | 0, /* tp_alloc */ |
|
1889 | 1890 | }; |
|
1890 | 1891 | |
|
1891 | 1892 | static int index_init_nt(indexObject *self) |
|
1892 | 1893 | { |
|
1893 | 1894 | if (!self->ntinitialized) { |
|
1894 | 1895 | if (nt_init(&self->nt, self, (int)self->length) == -1) { |
|
1895 | 1896 | nt_dealloc(&self->nt); |
|
1896 | 1897 | return -1; |
|
1897 | 1898 | } |
|
1898 | 1899 | if (nt_insert(&self->nt, nullid, -1) == -1) { |
|
1899 | 1900 | nt_dealloc(&self->nt); |
|
1900 | 1901 | return -1; |
|
1901 | 1902 | } |
|
1902 | 1903 | self->ntinitialized = 1; |
|
1903 | 1904 | self->ntrev = (int)index_length(self); |
|
1904 | 1905 | self->ntlookups = 1; |
|
1905 | 1906 | self->ntmisses = 0; |
|
1906 | 1907 | } |
|
1907 | 1908 | return 0; |
|
1908 | 1909 | } |
|
1909 | 1910 | |
|
1910 | 1911 | /* |
|
1911 | 1912 | * Return values: |
|
1912 | 1913 | * |
|
1913 | 1914 | * -3: error (exception set) |
|
1914 | 1915 | * -2: not found (no exception set) |
|
1915 | 1916 | * rest: valid rev |
|
1916 | 1917 | */ |
|
1917 | 1918 | static int index_find_node(indexObject *self, const char *node) |
|
1918 | 1919 | { |
|
1919 | 1920 | int rev; |
|
1920 | 1921 | |
|
1921 | 1922 | if (index_init_nt(self) == -1) |
|
1922 | 1923 | return -3; |
|
1923 | 1924 | |
|
1924 | 1925 | self->ntlookups++; |
|
1925 | 1926 | rev = nt_find(&self->nt, node, self->nodelen, 0); |
|
1926 | 1927 | if (rev >= -1) |
|
1927 | 1928 | return rev; |
|
1928 | 1929 | |
|
1929 | 1930 | /* |
|
1930 | 1931 | * For the first handful of lookups, we scan the entire index, |
|
1931 | 1932 | * and cache only the matching nodes. This optimizes for cases |
|
1932 | 1933 | * like "hg tip", where only a few nodes are accessed. |
|
1933 | 1934 | * |
|
1934 | 1935 | * After that, we cache every node we visit, using a single |
|
1935 | 1936 | * scan amortized over multiple lookups. This gives the best |
|
1936 | 1937 | * bulk performance, e.g. for "hg log". |
|
1937 | 1938 | */ |
|
1938 | 1939 | if (self->ntmisses++ < 4) { |
|
1939 | 1940 | for (rev = self->ntrev - 1; rev >= 0; rev--) { |
|
1940 | 1941 | const char *n = index_node_existing(self, rev); |
|
1941 | 1942 | if (n == NULL) |
|
1942 | 1943 | return -3; |
|
1943 | 1944 | if (memcmp(node, n, self->nodelen) == 0) { |
|
1944 | 1945 | if (nt_insert(&self->nt, n, rev) == -1) |
|
1945 | 1946 | return -3; |
|
1946 | 1947 | break; |
|
1947 | 1948 | } |
|
1948 | 1949 | } |
|
1949 | 1950 | } else { |
|
1950 | 1951 | for (rev = self->ntrev - 1; rev >= 0; rev--) { |
|
1951 | 1952 | const char *n = index_node_existing(self, rev); |
|
1952 | 1953 | if (n == NULL) |
|
1953 | 1954 | return -3; |
|
1954 | 1955 | if (nt_insert(&self->nt, n, rev) == -1) { |
|
1955 | 1956 | self->ntrev = rev + 1; |
|
1956 | 1957 | return -3; |
|
1957 | 1958 | } |
|
1958 | 1959 | if (memcmp(node, n, self->nodelen) == 0) { |
|
1959 | 1960 | break; |
|
1960 | 1961 | } |
|
1961 | 1962 | } |
|
1962 | 1963 | self->ntrev = rev; |
|
1963 | 1964 | } |
|
1964 | 1965 | |
|
1965 | 1966 | if (rev >= 0) |
|
1966 | 1967 | return rev; |
|
1967 | 1968 | return -2; |
|
1968 | 1969 | } |
|
1969 | 1970 | |
|
1970 | 1971 | static PyObject *index_getitem(indexObject *self, PyObject *value) |
|
1971 | 1972 | { |
|
1972 | 1973 | char *node; |
|
1973 | 1974 | int rev; |
|
1974 | 1975 | |
|
1975 | 1976 | if (PyInt_Check(value)) { |
|
1976 | 1977 | long idx; |
|
1977 | 1978 | if (!pylong_to_long(value, &idx)) { |
|
1978 | 1979 | return NULL; |
|
1979 | 1980 | } |
|
1980 | 1981 | return index_get(self, idx); |
|
1981 | 1982 | } |
|
1982 | 1983 | |
|
1983 | 1984 | if (node_check(self->nodelen, value, &node) == -1) |
|
1984 | 1985 | return NULL; |
|
1985 | 1986 | rev = index_find_node(self, node); |
|
1986 | 1987 | if (rev >= -1) |
|
1987 | 1988 | return PyInt_FromLong(rev); |
|
1988 | 1989 | if (rev == -2) |
|
1989 | 1990 | raise_revlog_error(); |
|
1990 | 1991 | return NULL; |
|
1991 | 1992 | } |
|
1992 | 1993 | |
|
1993 | 1994 | /* |
|
1994 | 1995 | * Fully populate the radix tree. |
|
1995 | 1996 | */ |
|
1996 | 1997 | static int index_populate_nt(indexObject *self) |
|
1997 | 1998 | { |
|
1998 | 1999 | int rev; |
|
1999 | 2000 | if (self->ntrev > 0) { |
|
2000 | 2001 | for (rev = self->ntrev - 1; rev >= 0; rev--) { |
|
2001 | 2002 | const char *n = index_node_existing(self, rev); |
|
2002 | 2003 | if (n == NULL) |
|
2003 | 2004 | return -1; |
|
2004 | 2005 | if (nt_insert(&self->nt, n, rev) == -1) |
|
2005 | 2006 | return -1; |
|
2006 | 2007 | } |
|
2007 | 2008 | self->ntrev = -1; |
|
2008 | 2009 | } |
|
2009 | 2010 | return 0; |
|
2010 | 2011 | } |
|
2011 | 2012 | |
|
2012 | 2013 | static PyObject *index_partialmatch(indexObject *self, PyObject *args) |
|
2013 | 2014 | { |
|
2014 | 2015 | const char *fullnode; |
|
2015 | 2016 | Py_ssize_t nodelen; |
|
2016 | 2017 | char *node; |
|
2017 | 2018 | int rev, i; |
|
2018 | 2019 | |
|
2019 | 2020 | if (!PyArg_ParseTuple(args, PY23("s#", "y#"), &node, &nodelen)) |
|
2020 | 2021 | return NULL; |
|
2021 | 2022 | |
|
2022 | 2023 | if (nodelen < 1) { |
|
2023 | 2024 | PyErr_SetString(PyExc_ValueError, "key too short"); |
|
2024 | 2025 | return NULL; |
|
2025 | 2026 | } |
|
2026 | 2027 | |
|
2027 | 2028 | if (nodelen > 2 * self->nodelen) { |
|
2028 | 2029 | PyErr_SetString(PyExc_ValueError, "key too long"); |
|
2029 | 2030 | return NULL; |
|
2030 | 2031 | } |
|
2031 | 2032 | |
|
2032 | 2033 | for (i = 0; i < nodelen; i++) |
|
2033 | 2034 | hexdigit(node, i); |
|
2034 | 2035 | if (PyErr_Occurred()) { |
|
2035 | 2036 | /* input contains non-hex characters */ |
|
2036 | 2037 | PyErr_Clear(); |
|
2037 | 2038 | Py_RETURN_NONE; |
|
2038 | 2039 | } |
|
2039 | 2040 | |
|
2040 | 2041 | if (index_init_nt(self) == -1) |
|
2041 | 2042 | return NULL; |
|
2042 | 2043 | if (index_populate_nt(self) == -1) |
|
2043 | 2044 | return NULL; |
|
2044 | 2045 | rev = nt_partialmatch(&self->nt, node, nodelen); |
|
2045 | 2046 | |
|
2046 | 2047 | switch (rev) { |
|
2047 | 2048 | case -4: |
|
2048 | 2049 | raise_revlog_error(); |
|
2049 | 2050 | return NULL; |
|
2050 | 2051 | case -2: |
|
2051 | 2052 | Py_RETURN_NONE; |
|
2052 | 2053 | case -1: |
|
2053 | 2054 | return PyBytes_FromStringAndSize(nullid, self->nodelen); |
|
2054 | 2055 | } |
|
2055 | 2056 | |
|
2056 | 2057 | fullnode = index_node_existing(self, rev); |
|
2057 | 2058 | if (fullnode == NULL) { |
|
2058 | 2059 | return NULL; |
|
2059 | 2060 | } |
|
2060 | 2061 | return PyBytes_FromStringAndSize(fullnode, self->nodelen); |
|
2061 | 2062 | } |
|
2062 | 2063 | |
|
2063 | 2064 | static PyObject *index_shortest(indexObject *self, PyObject *args) |
|
2064 | 2065 | { |
|
2065 | 2066 | PyObject *val; |
|
2066 | 2067 | char *node; |
|
2067 | 2068 | int length; |
|
2068 | 2069 | |
|
2069 | 2070 | if (!PyArg_ParseTuple(args, "O", &val)) |
|
2070 | 2071 | return NULL; |
|
2071 | 2072 | if (node_check(self->nodelen, val, &node) == -1) |
|
2072 | 2073 | return NULL; |
|
2073 | 2074 | |
|
2074 | 2075 | self->ntlookups++; |
|
2075 | 2076 | if (index_init_nt(self) == -1) |
|
2076 | 2077 | return NULL; |
|
2077 | 2078 | if (index_populate_nt(self) == -1) |
|
2078 | 2079 | return NULL; |
|
2079 | 2080 | length = nt_shortest(&self->nt, node); |
|
2080 | 2081 | if (length == -3) |
|
2081 | 2082 | return NULL; |
|
2082 | 2083 | if (length == -2) { |
|
2083 | 2084 | raise_revlog_error(); |
|
2084 | 2085 | return NULL; |
|
2085 | 2086 | } |
|
2086 | 2087 | return PyInt_FromLong(length); |
|
2087 | 2088 | } |
|
2088 | 2089 | |
|
2089 | 2090 | static PyObject *index_m_get(indexObject *self, PyObject *args) |
|
2090 | 2091 | { |
|
2091 | 2092 | PyObject *val; |
|
2092 | 2093 | char *node; |
|
2093 | 2094 | int rev; |
|
2094 | 2095 | |
|
2095 | 2096 | if (!PyArg_ParseTuple(args, "O", &val)) |
|
2096 | 2097 | return NULL; |
|
2097 | 2098 | if (node_check(self->nodelen, val, &node) == -1) |
|
2098 | 2099 | return NULL; |
|
2099 | 2100 | rev = index_find_node(self, node); |
|
2100 | 2101 | if (rev == -3) |
|
2101 | 2102 | return NULL; |
|
2102 | 2103 | if (rev == -2) |
|
2103 | 2104 | Py_RETURN_NONE; |
|
2104 | 2105 | return PyInt_FromLong(rev); |
|
2105 | 2106 | } |
|
2106 | 2107 | |
|
2107 | 2108 | static int index_contains(indexObject *self, PyObject *value) |
|
2108 | 2109 | { |
|
2109 | 2110 | char *node; |
|
2110 | 2111 | |
|
2111 | 2112 | if (PyInt_Check(value)) { |
|
2112 | 2113 | long rev; |
|
2113 | 2114 | if (!pylong_to_long(value, &rev)) { |
|
2114 | 2115 | return -1; |
|
2115 | 2116 | } |
|
2116 | 2117 | return rev >= -1 && rev < index_length(self); |
|
2117 | 2118 | } |
|
2118 | 2119 | |
|
2119 | 2120 | if (node_check(self->nodelen, value, &node) == -1) |
|
2120 | 2121 | return -1; |
|
2121 | 2122 | |
|
2122 | 2123 | switch (index_find_node(self, node)) { |
|
2123 | 2124 | case -3: |
|
2124 | 2125 | return -1; |
|
2125 | 2126 | case -2: |
|
2126 | 2127 | return 0; |
|
2127 | 2128 | default: |
|
2128 | 2129 | return 1; |
|
2129 | 2130 | } |
|
2130 | 2131 | } |
|
2131 | 2132 | |
|
2132 | 2133 | static PyObject *index_m_has_node(indexObject *self, PyObject *args) |
|
2133 | 2134 | { |
|
2134 | 2135 | int ret = index_contains(self, args); |
|
2135 | 2136 | if (ret < 0) |
|
2136 | 2137 | return NULL; |
|
2137 | 2138 | return PyBool_FromLong((long)ret); |
|
2138 | 2139 | } |
|
2139 | 2140 | |
|
2140 | 2141 | static PyObject *index_m_rev(indexObject *self, PyObject *val) |
|
2141 | 2142 | { |
|
2142 | 2143 | char *node; |
|
2143 | 2144 | int rev; |
|
2144 | 2145 | |
|
2145 | 2146 | if (node_check(self->nodelen, val, &node) == -1) |
|
2146 | 2147 | return NULL; |
|
2147 | 2148 | rev = index_find_node(self, node); |
|
2148 | 2149 | if (rev >= -1) |
|
2149 | 2150 | return PyInt_FromLong(rev); |
|
2150 | 2151 | if (rev == -2) |
|
2151 | 2152 | raise_revlog_error(); |
|
2152 | 2153 | return NULL; |
|
2153 | 2154 | } |
|
2154 | 2155 | |
|
2155 | 2156 | typedef uint64_t bitmask; |
|
2156 | 2157 | |
|
2157 | 2158 | /* |
|
2158 | 2159 | * Given a disjoint set of revs, return all candidates for the |
|
2159 | 2160 | * greatest common ancestor. In revset notation, this is the set |
|
2160 | 2161 | * "heads(::a and ::b and ...)" |
|
2161 | 2162 | */ |
|
2162 | 2163 | static PyObject *find_gca_candidates(indexObject *self, const int *revs, |
|
2163 | 2164 | int revcount) |
|
2164 | 2165 | { |
|
2165 | 2166 | const bitmask allseen = (1ull << revcount) - 1; |
|
2166 | 2167 | const bitmask poison = 1ull << revcount; |
|
2167 | 2168 | PyObject *gca = PyList_New(0); |
|
2168 | 2169 | int i, v, interesting; |
|
2169 | 2170 | int maxrev = -1; |
|
2170 | 2171 | bitmask sp; |
|
2171 | 2172 | bitmask *seen; |
|
2172 | 2173 | |
|
2173 | 2174 | if (gca == NULL) |
|
2174 | 2175 | return PyErr_NoMemory(); |
|
2175 | 2176 | |
|
2176 | 2177 | for (i = 0; i < revcount; i++) { |
|
2177 | 2178 | if (revs[i] > maxrev) |
|
2178 | 2179 | maxrev = revs[i]; |
|
2179 | 2180 | } |
|
2180 | 2181 | |
|
2181 | 2182 | seen = calloc(sizeof(*seen), maxrev + 1); |
|
2182 | 2183 | if (seen == NULL) { |
|
2183 | 2184 | Py_DECREF(gca); |
|
2184 | 2185 | return PyErr_NoMemory(); |
|
2185 | 2186 | } |
|
2186 | 2187 | |
|
2187 | 2188 | for (i = 0; i < revcount; i++) |
|
2188 | 2189 | seen[revs[i]] = 1ull << i; |
|
2189 | 2190 | |
|
2190 | 2191 | interesting = revcount; |
|
2191 | 2192 | |
|
2192 | 2193 | for (v = maxrev; v >= 0 && interesting; v--) { |
|
2193 | 2194 | bitmask sv = seen[v]; |
|
2194 | 2195 | int parents[2]; |
|
2195 | 2196 | |
|
2196 | 2197 | if (!sv) |
|
2197 | 2198 | continue; |
|
2198 | 2199 | |
|
2199 | 2200 | if (sv < poison) { |
|
2200 | 2201 | interesting -= 1; |
|
2201 | 2202 | if (sv == allseen) { |
|
2202 | 2203 | PyObject *obj = PyInt_FromLong(v); |
|
2203 | 2204 | if (obj == NULL) |
|
2204 | 2205 | goto bail; |
|
2205 | 2206 | if (PyList_Append(gca, obj) == -1) { |
|
2206 | 2207 | Py_DECREF(obj); |
|
2207 | 2208 | goto bail; |
|
2208 | 2209 | } |
|
2209 | 2210 | sv |= poison; |
|
2210 | 2211 | for (i = 0; i < revcount; i++) { |
|
2211 | 2212 | if (revs[i] == v) |
|
2212 | 2213 | goto done; |
|
2213 | 2214 | } |
|
2214 | 2215 | } |
|
2215 | 2216 | } |
|
2216 | 2217 | if (index_get_parents(self, v, parents, maxrev) < 0) |
|
2217 | 2218 | goto bail; |
|
2218 | 2219 | |
|
2219 | 2220 | for (i = 0; i < 2; i++) { |
|
2220 | 2221 | int p = parents[i]; |
|
2221 | 2222 | if (p == -1) |
|
2222 | 2223 | continue; |
|
2223 | 2224 | sp = seen[p]; |
|
2224 | 2225 | if (sv < poison) { |
|
2225 | 2226 | if (sp == 0) { |
|
2226 | 2227 | seen[p] = sv; |
|
2227 | 2228 | interesting++; |
|
2228 | 2229 | } else if (sp != sv) |
|
2229 | 2230 | seen[p] |= sv; |
|
2230 | 2231 | } else { |
|
2231 | 2232 | if (sp && sp < poison) |
|
2232 | 2233 | interesting--; |
|
2233 | 2234 | seen[p] = sv; |
|
2234 | 2235 | } |
|
2235 | 2236 | } |
|
2236 | 2237 | } |
|
2237 | 2238 | |
|
2238 | 2239 | done: |
|
2239 | 2240 | free(seen); |
|
2240 | 2241 | return gca; |
|
2241 | 2242 | bail: |
|
2242 | 2243 | free(seen); |
|
2243 | 2244 | Py_XDECREF(gca); |
|
2244 | 2245 | return NULL; |
|
2245 | 2246 | } |
|
2246 | 2247 | |
|
2247 | 2248 | /* |
|
2248 | 2249 | * Given a disjoint set of revs, return the subset with the longest |
|
2249 | 2250 | * path to the root. |
|
2250 | 2251 | */ |
|
2251 | 2252 | static PyObject *find_deepest(indexObject *self, PyObject *revs) |
|
2252 | 2253 | { |
|
2253 | 2254 | const Py_ssize_t revcount = PyList_GET_SIZE(revs); |
|
2254 | 2255 | static const Py_ssize_t capacity = 24; |
|
2255 | 2256 | int *depth, *interesting = NULL; |
|
2256 | 2257 | int i, j, v, ninteresting; |
|
2257 | 2258 | PyObject *dict = NULL, *keys = NULL; |
|
2258 | 2259 | long *seen = NULL; |
|
2259 | 2260 | int maxrev = -1; |
|
2260 | 2261 | long final; |
|
2261 | 2262 | |
|
2262 | 2263 | if (revcount > capacity) { |
|
2263 | 2264 | PyErr_Format(PyExc_OverflowError, |
|
2264 | 2265 | "bitset size (%ld) > capacity (%ld)", |
|
2265 | 2266 | (long)revcount, (long)capacity); |
|
2266 | 2267 | return NULL; |
|
2267 | 2268 | } |
|
2268 | 2269 | |
|
2269 | 2270 | for (i = 0; i < revcount; i++) { |
|
2270 | 2271 | int n = (int)PyInt_AsLong(PyList_GET_ITEM(revs, i)); |
|
2271 | 2272 | if (n > maxrev) |
|
2272 | 2273 | maxrev = n; |
|
2273 | 2274 | } |
|
2274 | 2275 | |
|
2275 | 2276 | depth = calloc(sizeof(*depth), maxrev + 1); |
|
2276 | 2277 | if (depth == NULL) |
|
2277 | 2278 | return PyErr_NoMemory(); |
|
2278 | 2279 | |
|
2279 | 2280 | seen = calloc(sizeof(*seen), maxrev + 1); |
|
2280 | 2281 | if (seen == NULL) { |
|
2281 | 2282 | PyErr_NoMemory(); |
|
2282 | 2283 | goto bail; |
|
2283 | 2284 | } |
|
2284 | 2285 | |
|
2285 | 2286 | interesting = calloc(sizeof(*interesting), ((size_t)1) << revcount); |
|
2286 | 2287 | if (interesting == NULL) { |
|
2287 | 2288 | PyErr_NoMemory(); |
|
2288 | 2289 | goto bail; |
|
2289 | 2290 | } |
|
2290 | 2291 | |
|
2291 | 2292 | if (PyList_Sort(revs) == -1) |
|
2292 | 2293 | goto bail; |
|
2293 | 2294 | |
|
2294 | 2295 | for (i = 0; i < revcount; i++) { |
|
2295 | 2296 | int n = (int)PyInt_AsLong(PyList_GET_ITEM(revs, i)); |
|
2296 | 2297 | long b = 1l << i; |
|
2297 | 2298 | depth[n] = 1; |
|
2298 | 2299 | seen[n] = b; |
|
2299 | 2300 | interesting[b] = 1; |
|
2300 | 2301 | } |
|
2301 | 2302 | |
|
2302 | 2303 | /* invariant: ninteresting is the number of non-zero entries in |
|
2303 | 2304 | * interesting. */ |
|
2304 | 2305 | ninteresting = (int)revcount; |
|
2305 | 2306 | |
|
2306 | 2307 | for (v = maxrev; v >= 0 && ninteresting > 1; v--) { |
|
2307 | 2308 | int dv = depth[v]; |
|
2308 | 2309 | int parents[2]; |
|
2309 | 2310 | long sv; |
|
2310 | 2311 | |
|
2311 | 2312 | if (dv == 0) |
|
2312 | 2313 | continue; |
|
2313 | 2314 | |
|
2314 | 2315 | sv = seen[v]; |
|
2315 | 2316 | if (index_get_parents(self, v, parents, maxrev) < 0) |
|
2316 | 2317 | goto bail; |
|
2317 | 2318 | |
|
2318 | 2319 | for (i = 0; i < 2; i++) { |
|
2319 | 2320 | int p = parents[i]; |
|
2320 | 2321 | long sp; |
|
2321 | 2322 | int dp; |
|
2322 | 2323 | |
|
2323 | 2324 | if (p == -1) |
|
2324 | 2325 | continue; |
|
2325 | 2326 | |
|
2326 | 2327 | dp = depth[p]; |
|
2327 | 2328 | sp = seen[p]; |
|
2328 | 2329 | if (dp <= dv) { |
|
2329 | 2330 | depth[p] = dv + 1; |
|
2330 | 2331 | if (sp != sv) { |
|
2331 | 2332 | interesting[sv] += 1; |
|
2332 | 2333 | seen[p] = sv; |
|
2333 | 2334 | if (sp) { |
|
2334 | 2335 | interesting[sp] -= 1; |
|
2335 | 2336 | if (interesting[sp] == 0) |
|
2336 | 2337 | ninteresting -= 1; |
|
2337 | 2338 | } |
|
2338 | 2339 | } |
|
2339 | 2340 | } else if (dv == dp - 1) { |
|
2340 | 2341 | long nsp = sp | sv; |
|
2341 | 2342 | if (nsp == sp) |
|
2342 | 2343 | continue; |
|
2343 | 2344 | seen[p] = nsp; |
|
2344 | 2345 | interesting[sp] -= 1; |
|
2345 | 2346 | if (interesting[sp] == 0) |
|
2346 | 2347 | ninteresting -= 1; |
|
2347 | 2348 | if (interesting[nsp] == 0) |
|
2348 | 2349 | ninteresting += 1; |
|
2349 | 2350 | interesting[nsp] += 1; |
|
2350 | 2351 | } |
|
2351 | 2352 | } |
|
2352 | 2353 | interesting[sv] -= 1; |
|
2353 | 2354 | if (interesting[sv] == 0) |
|
2354 | 2355 | ninteresting -= 1; |
|
2355 | 2356 | } |
|
2356 | 2357 | |
|
2357 | 2358 | final = 0; |
|
2358 | 2359 | j = ninteresting; |
|
2359 | 2360 | for (i = 0; i < (int)(2 << revcount) && j > 0; i++) { |
|
2360 | 2361 | if (interesting[i] == 0) |
|
2361 | 2362 | continue; |
|
2362 | 2363 | final |= i; |
|
2363 | 2364 | j -= 1; |
|
2364 | 2365 | } |
|
2365 | 2366 | if (final == 0) { |
|
2366 | 2367 | keys = PyList_New(0); |
|
2367 | 2368 | goto bail; |
|
2368 | 2369 | } |
|
2369 | 2370 | |
|
2370 | 2371 | dict = PyDict_New(); |
|
2371 | 2372 | if (dict == NULL) |
|
2372 | 2373 | goto bail; |
|
2373 | 2374 | |
|
2374 | 2375 | for (i = 0; i < revcount; i++) { |
|
2375 | 2376 | PyObject *key; |
|
2376 | 2377 | |
|
2377 | 2378 | if ((final & (1 << i)) == 0) |
|
2378 | 2379 | continue; |
|
2379 | 2380 | |
|
2380 | 2381 | key = PyList_GET_ITEM(revs, i); |
|
2381 | 2382 | Py_INCREF(key); |
|
2382 | 2383 | Py_INCREF(Py_None); |
|
2383 | 2384 | if (PyDict_SetItem(dict, key, Py_None) == -1) { |
|
2384 | 2385 | Py_DECREF(key); |
|
2385 | 2386 | Py_DECREF(Py_None); |
|
2386 | 2387 | goto bail; |
|
2387 | 2388 | } |
|
2388 | 2389 | } |
|
2389 | 2390 | |
|
2390 | 2391 | keys = PyDict_Keys(dict); |
|
2391 | 2392 | |
|
2392 | 2393 | bail: |
|
2393 | 2394 | free(depth); |
|
2394 | 2395 | free(seen); |
|
2395 | 2396 | free(interesting); |
|
2396 | 2397 | Py_XDECREF(dict); |
|
2397 | 2398 | |
|
2398 | 2399 | return keys; |
|
2399 | 2400 | } |
|
2400 | 2401 | |
|
2401 | 2402 | /* |
|
2402 | 2403 | * Given a (possibly overlapping) set of revs, return all the |
|
2403 | 2404 | * common ancestors heads: heads(::args[0] and ::a[1] and ...) |
|
2404 | 2405 | */ |
|
2405 | 2406 | static PyObject *index_commonancestorsheads(indexObject *self, PyObject *args) |
|
2406 | 2407 | { |
|
2407 | 2408 | PyObject *ret = NULL; |
|
2408 | 2409 | Py_ssize_t argcount, i, len; |
|
2409 | 2410 | bitmask repeat = 0; |
|
2410 | 2411 | int revcount = 0; |
|
2411 | 2412 | int *revs; |
|
2412 | 2413 | |
|
2413 | 2414 | argcount = PySequence_Length(args); |
|
2414 | 2415 | revs = PyMem_Malloc(argcount * sizeof(*revs)); |
|
2415 | 2416 | if (argcount > 0 && revs == NULL) |
|
2416 | 2417 | return PyErr_NoMemory(); |
|
2417 | 2418 | len = index_length(self); |
|
2418 | 2419 | |
|
2419 | 2420 | for (i = 0; i < argcount; i++) { |
|
2420 | 2421 | static const int capacity = 24; |
|
2421 | 2422 | PyObject *obj = PySequence_GetItem(args, i); |
|
2422 | 2423 | bitmask x; |
|
2423 | 2424 | long val; |
|
2424 | 2425 | |
|
2425 | 2426 | if (!PyInt_Check(obj)) { |
|
2426 | 2427 | PyErr_SetString(PyExc_TypeError, |
|
2427 | 2428 | "arguments must all be ints"); |
|
2428 | 2429 | Py_DECREF(obj); |
|
2429 | 2430 | goto bail; |
|
2430 | 2431 | } |
|
2431 | 2432 | val = PyInt_AsLong(obj); |
|
2432 | 2433 | Py_DECREF(obj); |
|
2433 | 2434 | if (val == -1) { |
|
2434 | 2435 | ret = PyList_New(0); |
|
2435 | 2436 | goto done; |
|
2436 | 2437 | } |
|
2437 | 2438 | if (val < 0 || val >= len) { |
|
2438 | 2439 | PyErr_SetString(PyExc_IndexError, "index out of range"); |
|
2439 | 2440 | goto bail; |
|
2440 | 2441 | } |
|
2441 | 2442 | /* this cheesy bloom filter lets us avoid some more |
|
2442 | 2443 | * expensive duplicate checks in the common set-is-disjoint |
|
2443 | 2444 | * case */ |
|
2444 | 2445 | x = 1ull << (val & 0x3f); |
|
2445 | 2446 | if (repeat & x) { |
|
2446 | 2447 | int k; |
|
2447 | 2448 | for (k = 0; k < revcount; k++) { |
|
2448 | 2449 | if (val == revs[k]) |
|
2449 | 2450 | goto duplicate; |
|
2450 | 2451 | } |
|
2451 | 2452 | } else |
|
2452 | 2453 | repeat |= x; |
|
2453 | 2454 | if (revcount >= capacity) { |
|
2454 | 2455 | PyErr_Format(PyExc_OverflowError, |
|
2455 | 2456 | "bitset size (%d) > capacity (%d)", |
|
2456 | 2457 | revcount, capacity); |
|
2457 | 2458 | goto bail; |
|
2458 | 2459 | } |
|
2459 | 2460 | revs[revcount++] = (int)val; |
|
2460 | 2461 | duplicate:; |
|
2461 | 2462 | } |
|
2462 | 2463 | |
|
2463 | 2464 | if (revcount == 0) { |
|
2464 | 2465 | ret = PyList_New(0); |
|
2465 | 2466 | goto done; |
|
2466 | 2467 | } |
|
2467 | 2468 | if (revcount == 1) { |
|
2468 | 2469 | PyObject *obj; |
|
2469 | 2470 | ret = PyList_New(1); |
|
2470 | 2471 | if (ret == NULL) |
|
2471 | 2472 | goto bail; |
|
2472 | 2473 | obj = PyInt_FromLong(revs[0]); |
|
2473 | 2474 | if (obj == NULL) |
|
2474 | 2475 | goto bail; |
|
2475 | 2476 | PyList_SET_ITEM(ret, 0, obj); |
|
2476 | 2477 | goto done; |
|
2477 | 2478 | } |
|
2478 | 2479 | |
|
2479 | 2480 | ret = find_gca_candidates(self, revs, revcount); |
|
2480 | 2481 | if (ret == NULL) |
|
2481 | 2482 | goto bail; |
|
2482 | 2483 | |
|
2483 | 2484 | done: |
|
2484 | 2485 | PyMem_Free(revs); |
|
2485 | 2486 | return ret; |
|
2486 | 2487 | |
|
2487 | 2488 | bail: |
|
2488 | 2489 | PyMem_Free(revs); |
|
2489 | 2490 | Py_XDECREF(ret); |
|
2490 | 2491 | return NULL; |
|
2491 | 2492 | } |
|
2492 | 2493 | |
|
2493 | 2494 | /* |
|
2494 | 2495 | * Given a (possibly overlapping) set of revs, return the greatest |
|
2495 | 2496 | * common ancestors: those with the longest path to the root. |
|
2496 | 2497 | */ |
|
2497 | 2498 | static PyObject *index_ancestors(indexObject *self, PyObject *args) |
|
2498 | 2499 | { |
|
2499 | 2500 | PyObject *ret; |
|
2500 | 2501 | PyObject *gca = index_commonancestorsheads(self, args); |
|
2501 | 2502 | if (gca == NULL) |
|
2502 | 2503 | return NULL; |
|
2503 | 2504 | |
|
2504 | 2505 | if (PyList_GET_SIZE(gca) <= 1) { |
|
2505 | 2506 | return gca; |
|
2506 | 2507 | } |
|
2507 | 2508 | |
|
2508 | 2509 | ret = find_deepest(self, gca); |
|
2509 | 2510 | Py_DECREF(gca); |
|
2510 | 2511 | return ret; |
|
2511 | 2512 | } |
|
2512 | 2513 | |
|
2513 | 2514 | /* |
|
2514 | 2515 | * Invalidate any trie entries introduced by added revs. |
|
2515 | 2516 | */ |
|
2516 | 2517 | static void index_invalidate_added(indexObject *self, Py_ssize_t start) |
|
2517 | 2518 | { |
|
2518 | 2519 | Py_ssize_t i, len; |
|
2519 | 2520 | |
|
2520 | 2521 | len = self->length + self->new_length; |
|
2521 | 2522 | i = start - self->length; |
|
2522 | 2523 | if (i < 0) |
|
2523 | 2524 | return; |
|
2524 | 2525 | |
|
2525 | 2526 | for (i = start; i < len; i++) |
|
2526 | 2527 | nt_delete_node(&self->nt, index_deref(self, i) + 32); |
|
2527 | 2528 | |
|
2528 | 2529 | self->new_length = start - self->length; |
|
2529 | 2530 | } |
|
2530 | 2531 | |
|
2531 | 2532 | /* |
|
2532 | 2533 | * Delete a numeric range of revs, which must be at the end of the |
|
2533 | 2534 | * range. |
|
2534 | 2535 | */ |
|
2535 | 2536 | static int index_slice_del(indexObject *self, PyObject *item) |
|
2536 | 2537 | { |
|
2537 | 2538 | Py_ssize_t start, stop, step, slicelength; |
|
2538 | 2539 | Py_ssize_t length = index_length(self) + 1; |
|
2539 | 2540 | int ret = 0; |
|
2540 | 2541 | |
|
2541 | 2542 | /* Argument changed from PySliceObject* to PyObject* in Python 3. */ |
|
2542 | 2543 | #ifdef IS_PY3K |
|
2543 | 2544 | if (PySlice_GetIndicesEx(item, length, &start, &stop, &step, |
|
2544 | 2545 | &slicelength) < 0) |
|
2545 | 2546 | #else |
|
2546 | 2547 | if (PySlice_GetIndicesEx((PySliceObject *)item, length, &start, &stop, |
|
2547 | 2548 | &step, &slicelength) < 0) |
|
2548 | 2549 | #endif |
|
2549 | 2550 | return -1; |
|
2550 | 2551 | |
|
2551 | 2552 | if (slicelength <= 0) |
|
2552 | 2553 | return 0; |
|
2553 | 2554 | |
|
2554 | 2555 | if ((step < 0 && start < stop) || (step > 0 && start > stop)) |
|
2555 | 2556 | stop = start; |
|
2556 | 2557 | |
|
2557 | 2558 | if (step < 0) { |
|
2558 | 2559 | stop = start + 1; |
|
2559 | 2560 | start = stop + step * (slicelength - 1) - 1; |
|
2560 | 2561 | step = -step; |
|
2561 | 2562 | } |
|
2562 | 2563 | |
|
2563 | 2564 | if (step != 1) { |
|
2564 | 2565 | PyErr_SetString(PyExc_ValueError, |
|
2565 | 2566 | "revlog index delete requires step size of 1"); |
|
2566 | 2567 | return -1; |
|
2567 | 2568 | } |
|
2568 | 2569 | |
|
2569 | 2570 | if (stop != length - 1) { |
|
2570 | 2571 | PyErr_SetString(PyExc_IndexError, |
|
2571 | 2572 | "revlog index deletion indices are invalid"); |
|
2572 | 2573 | return -1; |
|
2573 | 2574 | } |
|
2574 | 2575 | |
|
2575 | 2576 | if (start < self->length) { |
|
2576 | 2577 | if (self->ntinitialized) { |
|
2577 | 2578 | Py_ssize_t i; |
|
2578 | 2579 | |
|
2579 | 2580 | for (i = start; i < self->length; i++) { |
|
2580 | 2581 | const char *node = index_node_existing(self, i); |
|
2581 | 2582 | if (node == NULL) |
|
2582 | 2583 | return -1; |
|
2583 | 2584 | |
|
2584 | 2585 | nt_delete_node(&self->nt, node); |
|
2585 | 2586 | } |
|
2586 | 2587 | if (self->new_length) |
|
2587 | 2588 | index_invalidate_added(self, self->length); |
|
2588 | 2589 | if (self->ntrev > start) |
|
2589 | 2590 | self->ntrev = (int)start; |
|
2590 | 2591 | } else if (self->new_length) { |
|
2591 | 2592 | self->new_length = 0; |
|
2592 | 2593 | } |
|
2593 | 2594 | |
|
2594 | 2595 | self->length = start; |
|
2595 | 2596 | goto done; |
|
2596 | 2597 | } |
|
2597 | 2598 | |
|
2598 | 2599 | if (self->ntinitialized) { |
|
2599 | 2600 | index_invalidate_added(self, start); |
|
2600 | 2601 | if (self->ntrev > start) |
|
2601 | 2602 | self->ntrev = (int)start; |
|
2602 | 2603 | } else { |
|
2603 | 2604 | self->new_length = start - self->length; |
|
2604 | 2605 | } |
|
2605 | 2606 | done: |
|
2606 | 2607 | Py_CLEAR(self->headrevs); |
|
2607 | 2608 | return ret; |
|
2608 | 2609 | } |
|
2609 | 2610 | |
|
2610 | 2611 | /* |
|
2611 | 2612 | * Supported ops: |
|
2612 | 2613 | * |
|
2613 | 2614 | * slice deletion |
|
2614 | 2615 | * string assignment (extend node->rev mapping) |
|
2615 | 2616 | * string deletion (shrink node->rev mapping) |
|
2616 | 2617 | */ |
|
2617 | 2618 | static int index_assign_subscript(indexObject *self, PyObject *item, |
|
2618 | 2619 | PyObject *value) |
|
2619 | 2620 | { |
|
2620 | 2621 | char *node; |
|
2621 | 2622 | long rev; |
|
2622 | 2623 | |
|
2623 | 2624 | if (PySlice_Check(item) && value == NULL) |
|
2624 | 2625 | return index_slice_del(self, item); |
|
2625 | 2626 | |
|
2626 | 2627 | if (node_check(self->nodelen, item, &node) == -1) |
|
2627 | 2628 | return -1; |
|
2628 | 2629 | |
|
2629 | 2630 | if (value == NULL) |
|
2630 | 2631 | return self->ntinitialized ? nt_delete_node(&self->nt, node) |
|
2631 | 2632 | : 0; |
|
2632 | 2633 | rev = PyInt_AsLong(value); |
|
2633 | 2634 | if (rev > INT_MAX || rev < 0) { |
|
2634 | 2635 | if (!PyErr_Occurred()) |
|
2635 | 2636 | PyErr_SetString(PyExc_ValueError, "rev out of range"); |
|
2636 | 2637 | return -1; |
|
2637 | 2638 | } |
|
2638 | 2639 | |
|
2639 | 2640 | if (index_init_nt(self) == -1) |
|
2640 | 2641 | return -1; |
|
2641 | 2642 | return nt_insert(&self->nt, node, (int)rev); |
|
2642 | 2643 | } |
|
2643 | 2644 | |
|
2644 | 2645 | /* |
|
2645 | 2646 | * Find all RevlogNG entries in an index that has inline data. Update |
|
2646 | 2647 | * the optional "offsets" table with those entries. |
|
2647 | 2648 | */ |
|
2648 | 2649 | static Py_ssize_t inline_scan(indexObject *self, const char **offsets) |
|
2649 | 2650 | { |
|
2650 | 2651 | const char *data = (const char *)self->buf.buf; |
|
2651 | 2652 | Py_ssize_t pos = 0; |
|
2652 | 2653 | Py_ssize_t end = self->buf.len; |
|
2653 | 2654 | long incr = self->hdrsize; |
|
2654 | 2655 | Py_ssize_t len = 0; |
|
2655 | 2656 | |
|
2656 | 2657 | while (pos + self->hdrsize <= end && pos >= 0) { |
|
2657 | 2658 | uint32_t comp_len, sidedata_comp_len = 0; |
|
2658 | 2659 | /* 3rd element of header is length of compressed inline data */ |
|
2659 | 2660 | comp_len = getbe32(data + pos + 8); |
|
2660 | 2661 | if (self->hdrsize == v2_hdrsize) { |
|
2661 | 2662 | sidedata_comp_len = getbe32(data + pos + 72); |
|
2662 | 2663 | } |
|
2663 | 2664 | incr = self->hdrsize + comp_len + sidedata_comp_len; |
|
2664 | 2665 | if (offsets) |
|
2665 | 2666 | offsets[len] = data + pos; |
|
2666 | 2667 | len++; |
|
2667 | 2668 | pos += incr; |
|
2668 | 2669 | } |
|
2669 | 2670 | |
|
2670 | 2671 | if (pos != end) { |
|
2671 | 2672 | if (!PyErr_Occurred()) |
|
2672 | 2673 | PyErr_SetString(PyExc_ValueError, "corrupt index file"); |
|
2673 | 2674 | return -1; |
|
2674 | 2675 | } |
|
2675 | 2676 | |
|
2676 | 2677 | return len; |
|
2677 | 2678 | } |
|
2678 | 2679 | |
|
2679 | 2680 | static int index_init(indexObject *self, PyObject *args, PyObject *kwargs) |
|
2680 | 2681 | { |
|
2681 | 2682 | PyObject *data_obj, *inlined_obj, *revlogv2; |
|
2682 | 2683 | Py_ssize_t size; |
|
2683 | 2684 | |
|
2684 | 2685 | static char *kwlist[] = {"data", "inlined", "revlogv2", NULL}; |
|
2685 | 2686 | |
|
2686 | 2687 | /* Initialize before argument-checking to avoid index_dealloc() crash. |
|
2687 | 2688 | */ |
|
2688 | 2689 | self->added = NULL; |
|
2689 | 2690 | self->new_length = 0; |
|
2690 | 2691 | self->added_length = 0; |
|
2691 | 2692 | self->data = NULL; |
|
2692 | 2693 | memset(&self->buf, 0, sizeof(self->buf)); |
|
2693 | 2694 | self->headrevs = NULL; |
|
2694 | 2695 | self->filteredrevs = Py_None; |
|
2695 | 2696 | Py_INCREF(Py_None); |
|
2696 | 2697 | self->ntinitialized = 0; |
|
2697 | 2698 | self->offsets = NULL; |
|
2698 | 2699 | self->nodelen = 20; |
|
2699 | 2700 | self->nullentry = NULL; |
|
2700 | 2701 | |
|
2701 | 2702 | revlogv2 = NULL; |
|
2702 | 2703 | if (!PyArg_ParseTupleAndKeywords(args, kwargs, "OO|O", kwlist, |
|
2703 | 2704 | &data_obj, &inlined_obj, &revlogv2)) |
|
2704 | 2705 | return -1; |
|
2705 | 2706 | if (!PyObject_CheckBuffer(data_obj)) { |
|
2706 | 2707 | PyErr_SetString(PyExc_TypeError, |
|
2707 | 2708 | "data does not support buffer interface"); |
|
2708 | 2709 | return -1; |
|
2709 | 2710 | } |
|
2710 | 2711 | if (self->nodelen < 20 || self->nodelen > (Py_ssize_t)sizeof(nullid)) { |
|
2711 | 2712 | PyErr_SetString(PyExc_RuntimeError, "unsupported node size"); |
|
2712 | 2713 | return -1; |
|
2713 | 2714 | } |
|
2714 | 2715 | |
|
2715 | 2716 | if (revlogv2 && PyObject_IsTrue(revlogv2)) { |
|
2716 | 2717 | self->hdrsize = v2_hdrsize; |
|
2717 | 2718 | } else { |
|
2718 | 2719 | self->hdrsize = v1_hdrsize; |
|
2719 | 2720 | } |
|
2720 | 2721 | |
|
2721 | 2722 | if (self->hdrsize == v1_hdrsize) { |
|
2722 | 2723 | self->nullentry = |
|
2723 | 2724 | Py_BuildValue(PY23("iiiiiiis#", "iiiiiiiy#"), 0, 0, 0, -1, |
|
2724 | 2725 | -1, -1, -1, nullid, self->nodelen); |
|
2725 | 2726 | } else { |
|
2726 | 2727 | self->nullentry = |
|
2727 | 2728 | Py_BuildValue(PY23("iiiiiiis#ii", "iiiiiiiy#ii"), 0, 0, 0, |
|
2728 | 2729 | -1, -1, -1, -1, nullid, self->nodelen, 0, 0); |
|
2729 | 2730 | } |
|
2730 | 2731 | |
|
2731 | 2732 | if (!self->nullentry) |
|
2732 | 2733 | return -1; |
|
2733 | 2734 | PyObject_GC_UnTrack(self->nullentry); |
|
2734 | 2735 | |
|
2735 | 2736 | if (PyObject_GetBuffer(data_obj, &self->buf, PyBUF_SIMPLE) == -1) |
|
2736 | 2737 | return -1; |
|
2737 | 2738 | size = self->buf.len; |
|
2738 | 2739 | |
|
2739 | 2740 | self->inlined = inlined_obj && PyObject_IsTrue(inlined_obj); |
|
2740 | 2741 | self->data = data_obj; |
|
2741 | 2742 | |
|
2742 | 2743 | self->ntlookups = self->ntmisses = 0; |
|
2743 | 2744 | self->ntrev = -1; |
|
2744 | 2745 | Py_INCREF(self->data); |
|
2745 | 2746 | |
|
2746 | 2747 | if (self->inlined) { |
|
2747 | 2748 | Py_ssize_t len = inline_scan(self, NULL); |
|
2748 | 2749 | if (len == -1) |
|
2749 | 2750 | goto bail; |
|
2750 | 2751 | self->length = len; |
|
2751 | 2752 | } else { |
|
2752 | 2753 | if (size % self->hdrsize) { |
|
2753 | 2754 | PyErr_SetString(PyExc_ValueError, "corrupt index file"); |
|
2754 | 2755 | goto bail; |
|
2755 | 2756 | } |
|
2756 | 2757 | self->length = size / self->hdrsize; |
|
2757 | 2758 | } |
|
2758 | 2759 | |
|
2759 | 2760 | return 0; |
|
2760 | 2761 | bail: |
|
2761 | 2762 | return -1; |
|
2762 | 2763 | } |
|
2763 | 2764 | |
|
2764 | 2765 | static PyObject *index_nodemap(indexObject *self) |
|
2765 | 2766 | { |
|
2766 | 2767 | Py_INCREF(self); |
|
2767 | 2768 | return (PyObject *)self; |
|
2768 | 2769 | } |
|
2769 | 2770 | |
|
2770 | 2771 | static void _index_clearcaches(indexObject *self) |
|
2771 | 2772 | { |
|
2772 | 2773 | if (self->offsets) { |
|
2773 | 2774 | PyMem_Free((void *)self->offsets); |
|
2774 | 2775 | self->offsets = NULL; |
|
2775 | 2776 | } |
|
2776 | 2777 | if (self->ntinitialized) { |
|
2777 | 2778 | nt_dealloc(&self->nt); |
|
2778 | 2779 | } |
|
2779 | 2780 | self->ntinitialized = 0; |
|
2780 | 2781 | Py_CLEAR(self->headrevs); |
|
2781 | 2782 | } |
|
2782 | 2783 | |
|
2783 | 2784 | static PyObject *index_clearcaches(indexObject *self) |
|
2784 | 2785 | { |
|
2785 | 2786 | _index_clearcaches(self); |
|
2786 | 2787 | self->ntrev = -1; |
|
2787 | 2788 | self->ntlookups = self->ntmisses = 0; |
|
2788 | 2789 | Py_RETURN_NONE; |
|
2789 | 2790 | } |
|
2790 | 2791 | |
|
2791 | 2792 | static void index_dealloc(indexObject *self) |
|
2792 | 2793 | { |
|
2793 | 2794 | _index_clearcaches(self); |
|
2794 | 2795 | Py_XDECREF(self->filteredrevs); |
|
2795 | 2796 | if (self->buf.buf) { |
|
2796 | 2797 | PyBuffer_Release(&self->buf); |
|
2797 | 2798 | memset(&self->buf, 0, sizeof(self->buf)); |
|
2798 | 2799 | } |
|
2799 | 2800 | Py_XDECREF(self->data); |
|
2800 | 2801 | PyMem_Free(self->added); |
|
2801 | 2802 | Py_XDECREF(self->nullentry); |
|
2802 | 2803 | PyObject_Del(self); |
|
2803 | 2804 | } |
|
2804 | 2805 | |
|
2805 | 2806 | static PySequenceMethods index_sequence_methods = { |
|
2806 | 2807 | (lenfunc)index_length, /* sq_length */ |
|
2807 | 2808 | 0, /* sq_concat */ |
|
2808 | 2809 | 0, /* sq_repeat */ |
|
2809 | 2810 | (ssizeargfunc)index_get, /* sq_item */ |
|
2810 | 2811 | 0, /* sq_slice */ |
|
2811 | 2812 | 0, /* sq_ass_item */ |
|
2812 | 2813 | 0, /* sq_ass_slice */ |
|
2813 | 2814 | (objobjproc)index_contains, /* sq_contains */ |
|
2814 | 2815 | }; |
|
2815 | 2816 | |
|
2816 | 2817 | static PyMappingMethods index_mapping_methods = { |
|
2817 | 2818 | (lenfunc)index_length, /* mp_length */ |
|
2818 | 2819 | (binaryfunc)index_getitem, /* mp_subscript */ |
|
2819 | 2820 | (objobjargproc)index_assign_subscript, /* mp_ass_subscript */ |
|
2820 | 2821 | }; |
|
2821 | 2822 | |
|
2822 | 2823 | static PyMethodDef index_methods[] = { |
|
2823 | 2824 | {"ancestors", (PyCFunction)index_ancestors, METH_VARARGS, |
|
2824 | 2825 | "return the gca set of the given revs"}, |
|
2825 | 2826 | {"commonancestorsheads", (PyCFunction)index_commonancestorsheads, |
|
2826 | 2827 | METH_VARARGS, |
|
2827 | 2828 | "return the heads of the common ancestors of the given revs"}, |
|
2828 | 2829 | {"clearcaches", (PyCFunction)index_clearcaches, METH_NOARGS, |
|
2829 | 2830 | "clear the index caches"}, |
|
2830 | 2831 | {"get", (PyCFunction)index_m_get, METH_VARARGS, "get an index entry"}, |
|
2831 | 2832 | {"get_rev", (PyCFunction)index_m_get, METH_VARARGS, |
|
2832 | 2833 | "return `rev` associated with a node or None"}, |
|
2833 | 2834 | {"has_node", (PyCFunction)index_m_has_node, METH_O, |
|
2834 | 2835 | "return True if the node exist in the index"}, |
|
2835 | 2836 | {"rev", (PyCFunction)index_m_rev, METH_O, |
|
2836 | 2837 | "return `rev` associated with a node or raise RevlogError"}, |
|
2837 | 2838 | {"computephasesmapsets", (PyCFunction)compute_phases_map_sets, METH_VARARGS, |
|
2838 | 2839 | "compute phases"}, |
|
2839 | 2840 | {"reachableroots2", (PyCFunction)reachableroots2, METH_VARARGS, |
|
2840 | 2841 | "reachableroots"}, |
|
2841 | 2842 | {"replace_sidedata_info", (PyCFunction)index_replace_sidedata_info, |
|
2842 | 2843 | METH_VARARGS, "replace an existing index entry with a new value"}, |
|
2843 | 2844 | {"headrevs", (PyCFunction)index_headrevs, METH_VARARGS, |
|
2844 | 2845 | "get head revisions"}, /* Can do filtering since 3.2 */ |
|
2845 | 2846 | {"headrevsfiltered", (PyCFunction)index_headrevs, METH_VARARGS, |
|
2846 | 2847 | "get filtered head revisions"}, /* Can always do filtering */ |
|
2847 | 2848 | {"issnapshot", (PyCFunction)index_issnapshot, METH_O, |
|
2848 | 2849 | "True if the object is a snapshot"}, |
|
2849 | 2850 | {"findsnapshots", (PyCFunction)index_findsnapshots, METH_VARARGS, |
|
2850 | 2851 | "Gather snapshot data in a cache dict"}, |
|
2851 | 2852 | {"deltachain", (PyCFunction)index_deltachain, METH_VARARGS, |
|
2852 | 2853 | "determine revisions with deltas to reconstruct fulltext"}, |
|
2853 | 2854 | {"slicechunktodensity", (PyCFunction)index_slicechunktodensity, |
|
2854 | 2855 | METH_VARARGS, "determine revisions with deltas to reconstruct fulltext"}, |
|
2855 | 2856 | {"append", (PyCFunction)index_append, METH_O, "append an index entry"}, |
|
2856 | 2857 | {"partialmatch", (PyCFunction)index_partialmatch, METH_VARARGS, |
|
2857 | 2858 | "match a potentially ambiguous node ID"}, |
|
2858 | 2859 | {"shortest", (PyCFunction)index_shortest, METH_VARARGS, |
|
2859 | 2860 | "find length of shortest hex nodeid of a binary ID"}, |
|
2860 | 2861 | {"stats", (PyCFunction)index_stats, METH_NOARGS, "stats for the index"}, |
|
2861 | 2862 | {NULL} /* Sentinel */ |
|
2862 | 2863 | }; |
|
2863 | 2864 | |
|
2864 | 2865 | static PyGetSetDef index_getset[] = { |
|
2865 | 2866 | {"nodemap", (getter)index_nodemap, NULL, "nodemap", NULL}, |
|
2866 | 2867 | {NULL} /* Sentinel */ |
|
2867 | 2868 | }; |
|
2868 | 2869 | |
|
2870 | static PyMemberDef index_members[] = { | |
|
2871 | {"entry_size", T_LONG, offsetof(indexObject, hdrsize), 0, | |
|
2872 | "size of an index entry"}, | |
|
2873 | {NULL} /* Sentinel */ | |
|
2874 | }; | |
|
2875 | ||
|
2869 | 2876 | PyTypeObject HgRevlogIndex_Type = { |
|
2870 | 2877 | PyVarObject_HEAD_INIT(NULL, 0) /* header */ |
|
2871 | 2878 | "parsers.index", /* tp_name */ |
|
2872 | 2879 | sizeof(indexObject), /* tp_basicsize */ |
|
2873 | 2880 | 0, /* tp_itemsize */ |
|
2874 | 2881 | (destructor)index_dealloc, /* tp_dealloc */ |
|
2875 | 2882 | 0, /* tp_print */ |
|
2876 | 2883 | 0, /* tp_getattr */ |
|
2877 | 2884 | 0, /* tp_setattr */ |
|
2878 | 2885 | 0, /* tp_compare */ |
|
2879 | 2886 | 0, /* tp_repr */ |
|
2880 | 2887 | 0, /* tp_as_number */ |
|
2881 | 2888 | &index_sequence_methods, /* tp_as_sequence */ |
|
2882 | 2889 | &index_mapping_methods, /* tp_as_mapping */ |
|
2883 | 2890 | 0, /* tp_hash */ |
|
2884 | 2891 | 0, /* tp_call */ |
|
2885 | 2892 | 0, /* tp_str */ |
|
2886 | 2893 | 0, /* tp_getattro */ |
|
2887 | 2894 | 0, /* tp_setattro */ |
|
2888 | 2895 | 0, /* tp_as_buffer */ |
|
2889 | 2896 | Py_TPFLAGS_DEFAULT, /* tp_flags */ |
|
2890 | 2897 | "revlog index", /* tp_doc */ |
|
2891 | 2898 | 0, /* tp_traverse */ |
|
2892 | 2899 | 0, /* tp_clear */ |
|
2893 | 2900 | 0, /* tp_richcompare */ |
|
2894 | 2901 | 0, /* tp_weaklistoffset */ |
|
2895 | 2902 | 0, /* tp_iter */ |
|
2896 | 2903 | 0, /* tp_iternext */ |
|
2897 | 2904 | index_methods, /* tp_methods */ |
|
2898 |
|
|
|
2905 | index_members, /* tp_members */ | |
|
2899 | 2906 | index_getset, /* tp_getset */ |
|
2900 | 2907 | 0, /* tp_base */ |
|
2901 | 2908 | 0, /* tp_dict */ |
|
2902 | 2909 | 0, /* tp_descr_get */ |
|
2903 | 2910 | 0, /* tp_descr_set */ |
|
2904 | 2911 | 0, /* tp_dictoffset */ |
|
2905 | 2912 | (initproc)index_init, /* tp_init */ |
|
2906 | 2913 | 0, /* tp_alloc */ |
|
2907 | 2914 | }; |
|
2908 | 2915 | |
|
2909 | 2916 | /* |
|
2910 | 2917 | * returns a tuple of the form (index, cache) with elements as |
|
2911 | 2918 | * follows: |
|
2912 | 2919 | * |
|
2913 | 2920 | * index: an index object that lazily parses Revlog (v1 or v2) records |
|
2914 | 2921 | * cache: if data is inlined, a tuple (0, index_file_content), else None |
|
2915 | 2922 | * index_file_content could be a string, or a buffer |
|
2916 | 2923 | * |
|
2917 | 2924 | * added complications are for backwards compatibility |
|
2918 | 2925 | */ |
|
2919 | 2926 | PyObject *parse_index2(PyObject *self, PyObject *args, PyObject *kwargs) |
|
2920 | 2927 | { |
|
2921 | 2928 | PyObject *cache = NULL; |
|
2922 | 2929 | indexObject *idx; |
|
2923 | 2930 | int ret; |
|
2924 | 2931 | |
|
2925 | 2932 | idx = PyObject_New(indexObject, &HgRevlogIndex_Type); |
|
2926 | 2933 | if (idx == NULL) |
|
2927 | 2934 | goto bail; |
|
2928 | 2935 | |
|
2929 | 2936 | ret = index_init(idx, args, kwargs); |
|
2930 | 2937 | if (ret == -1) |
|
2931 | 2938 | goto bail; |
|
2932 | 2939 | |
|
2933 | 2940 | if (idx->inlined) { |
|
2934 | 2941 | cache = Py_BuildValue("iO", 0, idx->data); |
|
2935 | 2942 | if (cache == NULL) |
|
2936 | 2943 | goto bail; |
|
2937 | 2944 | } else { |
|
2938 | 2945 | cache = Py_None; |
|
2939 | 2946 | Py_INCREF(cache); |
|
2940 | 2947 | } |
|
2941 | 2948 | |
|
2942 | 2949 | return Py_BuildValue("NN", idx, cache); |
|
2943 | 2950 | |
|
2944 | 2951 | bail: |
|
2945 | 2952 | Py_XDECREF(idx); |
|
2946 | 2953 | Py_XDECREF(cache); |
|
2947 | 2954 | return NULL; |
|
2948 | 2955 | } |
|
2949 | 2956 | |
|
2950 | 2957 | static Revlog_CAPI CAPI = { |
|
2951 | 2958 | /* increment the abi_version field upon each change in the Revlog_CAPI |
|
2952 | 2959 | struct or in the ABI of the listed functions */ |
|
2953 | 2960 | 2, |
|
2954 | 2961 | index_length, |
|
2955 | 2962 | index_node, |
|
2956 | 2963 | HgRevlogIndex_GetParents, |
|
2957 | 2964 | }; |
|
2958 | 2965 | |
|
2959 | 2966 | void revlog_module_init(PyObject *mod) |
|
2960 | 2967 | { |
|
2961 | 2968 | PyObject *caps = NULL; |
|
2962 | 2969 | HgRevlogIndex_Type.tp_new = PyType_GenericNew; |
|
2963 | 2970 | if (PyType_Ready(&HgRevlogIndex_Type) < 0) |
|
2964 | 2971 | return; |
|
2965 | 2972 | Py_INCREF(&HgRevlogIndex_Type); |
|
2966 | 2973 | PyModule_AddObject(mod, "index", (PyObject *)&HgRevlogIndex_Type); |
|
2967 | 2974 | |
|
2968 | 2975 | nodetreeType.tp_new = PyType_GenericNew; |
|
2969 | 2976 | if (PyType_Ready(&nodetreeType) < 0) |
|
2970 | 2977 | return; |
|
2971 | 2978 | Py_INCREF(&nodetreeType); |
|
2972 | 2979 | PyModule_AddObject(mod, "nodetree", (PyObject *)&nodetreeType); |
|
2973 | 2980 | |
|
2974 | 2981 | caps = PyCapsule_New(&CAPI, "mercurial.cext.parsers.revlog_CAPI", NULL); |
|
2975 | 2982 | if (caps != NULL) |
|
2976 | 2983 | PyModule_AddObject(mod, "revlog_CAPI", caps); |
|
2977 | 2984 | } |
@@ -1,352 +1,353 b'' | |||
|
1 | 1 | # parsers.py - Python implementation of parsers.c |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2009 Olivia Mackall <olivia@selenic.com> and others |
|
4 | 4 | # |
|
5 | 5 | # This software may be used and distributed according to the terms of the |
|
6 | 6 | # GNU General Public License version 2 or any later version. |
|
7 | 7 | |
|
8 | 8 | from __future__ import absolute_import |
|
9 | 9 | |
|
10 | 10 | import struct |
|
11 | 11 | import zlib |
|
12 | 12 | |
|
13 | 13 | from ..node import nullid, nullrev |
|
14 | 14 | from .. import ( |
|
15 | 15 | pycompat, |
|
16 | 16 | util, |
|
17 | 17 | ) |
|
18 | 18 | |
|
19 | 19 | from ..revlogutils import nodemap as nodemaputil |
|
20 | 20 | from ..revlogutils import constants as revlog_constants |
|
21 | 21 | |
|
22 | 22 | stringio = pycompat.bytesio |
|
23 | 23 | |
|
24 | 24 | |
|
25 | 25 | _pack = struct.pack |
|
26 | 26 | _unpack = struct.unpack |
|
27 | 27 | _compress = zlib.compress |
|
28 | 28 | _decompress = zlib.decompress |
|
29 | 29 | |
|
30 | 30 | # Some code below makes tuples directly because it's more convenient. However, |
|
31 | 31 | # code outside this module should always use dirstatetuple. |
|
32 | 32 | def dirstatetuple(*x): |
|
33 | 33 | # x is a tuple |
|
34 | 34 | return x |
|
35 | 35 | |
|
36 | 36 | |
|
37 | 37 | def gettype(q): |
|
38 | 38 | return int(q & 0xFFFF) |
|
39 | 39 | |
|
40 | 40 | |
|
41 | 41 | def offset_type(offset, type): |
|
42 | 42 | return int(int(offset) << 16 | type) |
|
43 | 43 | |
|
44 | 44 | |
|
45 | 45 | class BaseIndexObject(object): |
|
46 | 46 | # Format of an index entry according to Python's `struct` language |
|
47 | 47 | index_format = revlog_constants.INDEX_ENTRY_V1 |
|
48 | 48 | # Size of a C unsigned long long int, platform independent |
|
49 | 49 | big_int_size = struct.calcsize(b'>Q') |
|
50 | 50 | # Size of a C long int, platform independent |
|
51 | 51 | int_size = struct.calcsize(b'>i') |
|
52 | # Size of the entire index format | |
|
53 | index_size = revlog_constants.INDEX_ENTRY_V1.size | |
|
54 | 52 | # An empty index entry, used as a default value to be overridden, or nullrev |
|
55 | 53 | null_item = (0, 0, 0, -1, -1, -1, -1, nullid) |
|
56 | 54 | |
|
55 | @util.propertycache | |
|
56 | def entry_size(self): | |
|
57 | return self.index_format.size | |
|
58 | ||
|
57 | 59 | @property |
|
58 | 60 | def nodemap(self): |
|
59 | 61 | msg = b"index.nodemap is deprecated, use index.[has_node|rev|get_rev]" |
|
60 | 62 | util.nouideprecwarn(msg, b'5.3', stacklevel=2) |
|
61 | 63 | return self._nodemap |
|
62 | 64 | |
|
63 | 65 | @util.propertycache |
|
64 | 66 | def _nodemap(self): |
|
65 | 67 | nodemap = nodemaputil.NodeMap({nullid: nullrev}) |
|
66 | 68 | for r in range(0, len(self)): |
|
67 | 69 | n = self[r][7] |
|
68 | 70 | nodemap[n] = r |
|
69 | 71 | return nodemap |
|
70 | 72 | |
|
71 | 73 | def has_node(self, node): |
|
72 | 74 | """return True if the node exist in the index""" |
|
73 | 75 | return node in self._nodemap |
|
74 | 76 | |
|
75 | 77 | def rev(self, node): |
|
76 | 78 | """return a revision for a node |
|
77 | 79 | |
|
78 | 80 | If the node is unknown, raise a RevlogError""" |
|
79 | 81 | return self._nodemap[node] |
|
80 | 82 | |
|
81 | 83 | def get_rev(self, node): |
|
82 | 84 | """return a revision for a node |
|
83 | 85 | |
|
84 | 86 | If the node is unknown, return None""" |
|
85 | 87 | return self._nodemap.get(node) |
|
86 | 88 | |
|
87 | 89 | def _stripnodes(self, start): |
|
88 | 90 | if '_nodemap' in vars(self): |
|
89 | 91 | for r in range(start, len(self)): |
|
90 | 92 | n = self[r][7] |
|
91 | 93 | del self._nodemap[n] |
|
92 | 94 | |
|
93 | 95 | def clearcaches(self): |
|
94 | 96 | self.__dict__.pop('_nodemap', None) |
|
95 | 97 | |
|
96 | 98 | def __len__(self): |
|
97 | 99 | return self._lgt + len(self._extra) |
|
98 | 100 | |
|
99 | 101 | def append(self, tup): |
|
100 | 102 | if '_nodemap' in vars(self): |
|
101 | 103 | self._nodemap[tup[7]] = len(self) |
|
102 | 104 | data = self.index_format.pack(*tup) |
|
103 | 105 | self._extra.append(data) |
|
104 | 106 | |
|
105 | 107 | def _check_index(self, i): |
|
106 | 108 | if not isinstance(i, int): |
|
107 | 109 | raise TypeError(b"expecting int indexes") |
|
108 | 110 | if i < 0 or i >= len(self): |
|
109 | 111 | raise IndexError |
|
110 | 112 | |
|
111 | 113 | def __getitem__(self, i): |
|
112 | 114 | if i == -1: |
|
113 | 115 | return self.null_item |
|
114 | 116 | self._check_index(i) |
|
115 | 117 | if i >= self._lgt: |
|
116 | 118 | data = self._extra[i - self._lgt] |
|
117 | 119 | else: |
|
118 | 120 | index = self._calculate_index(i) |
|
119 |
data = self._data[index : index + self. |
|
|
121 | data = self._data[index : index + self.entry_size] | |
|
120 | 122 | r = self.index_format.unpack(data) |
|
121 | 123 | if self._lgt and i == 0: |
|
122 | 124 | r = (offset_type(0, gettype(r[0])),) + r[1:] |
|
123 | 125 | return r |
|
124 | 126 | |
|
125 | 127 | |
|
126 | 128 | class IndexObject(BaseIndexObject): |
|
127 | 129 | def __init__(self, data): |
|
128 |
assert len(data) % self. |
|
|
130 | assert len(data) % self.entry_size == 0 | |
|
129 | 131 | self._data = data |
|
130 |
self._lgt = len(data) // self. |
|
|
132 | self._lgt = len(data) // self.entry_size | |
|
131 | 133 | self._extra = [] |
|
132 | 134 | |
|
133 | 135 | def _calculate_index(self, i): |
|
134 |
return i * self. |
|
|
136 | return i * self.entry_size | |
|
135 | 137 | |
|
136 | 138 | def __delitem__(self, i): |
|
137 | 139 | if not isinstance(i, slice) or not i.stop == -1 or i.step is not None: |
|
138 | 140 | raise ValueError(b"deleting slices only supports a:-1 with step 1") |
|
139 | 141 | i = i.start |
|
140 | 142 | self._check_index(i) |
|
141 | 143 | self._stripnodes(i) |
|
142 | 144 | if i < self._lgt: |
|
143 |
self._data = self._data[: i * self. |
|
|
145 | self._data = self._data[: i * self.entry_size] | |
|
144 | 146 | self._lgt = i |
|
145 | 147 | self._extra = [] |
|
146 | 148 | else: |
|
147 | 149 | self._extra = self._extra[: i - self._lgt] |
|
148 | 150 | |
|
149 | 151 | |
|
150 | 152 | class PersistentNodeMapIndexObject(IndexObject): |
|
151 | 153 | """a Debug oriented class to test persistent nodemap |
|
152 | 154 | |
|
153 | 155 | We need a simple python object to test API and higher level behavior. See |
|
154 | 156 | the Rust implementation for more serious usage. This should be used only |
|
155 | 157 | through the dedicated `devel.persistent-nodemap` config. |
|
156 | 158 | """ |
|
157 | 159 | |
|
158 | 160 | def nodemap_data_all(self): |
|
159 | 161 | """Return bytes containing a full serialization of a nodemap |
|
160 | 162 | |
|
161 | 163 | The nodemap should be valid for the full set of revisions in the |
|
162 | 164 | index.""" |
|
163 | 165 | return nodemaputil.persistent_data(self) |
|
164 | 166 | |
|
165 | 167 | def nodemap_data_incremental(self): |
|
166 | 168 | """Return bytes containing a incremental update to persistent nodemap |
|
167 | 169 | |
|
168 | 170 | This containst the data for an append-only update of the data provided |
|
169 | 171 | in the last call to `update_nodemap_data`. |
|
170 | 172 | """ |
|
171 | 173 | if self._nm_root is None: |
|
172 | 174 | return None |
|
173 | 175 | docket = self._nm_docket |
|
174 | 176 | changed, data = nodemaputil.update_persistent_data( |
|
175 | 177 | self, self._nm_root, self._nm_max_idx, self._nm_docket.tip_rev |
|
176 | 178 | ) |
|
177 | 179 | |
|
178 | 180 | self._nm_root = self._nm_max_idx = self._nm_docket = None |
|
179 | 181 | return docket, changed, data |
|
180 | 182 | |
|
181 | 183 | def update_nodemap_data(self, docket, nm_data): |
|
182 | 184 | """provide full block of persisted binary data for a nodemap |
|
183 | 185 | |
|
184 | 186 | The data are expected to come from disk. See `nodemap_data_all` for a |
|
185 | 187 | produceur of such data.""" |
|
186 | 188 | if nm_data is not None: |
|
187 | 189 | self._nm_root, self._nm_max_idx = nodemaputil.parse_data(nm_data) |
|
188 | 190 | if self._nm_root: |
|
189 | 191 | self._nm_docket = docket |
|
190 | 192 | else: |
|
191 | 193 | self._nm_root = self._nm_max_idx = self._nm_docket = None |
|
192 | 194 | |
|
193 | 195 | |
|
194 | 196 | class InlinedIndexObject(BaseIndexObject): |
|
195 | 197 | def __init__(self, data, inline=0): |
|
196 | 198 | self._data = data |
|
197 | 199 | self._lgt = self._inline_scan(None) |
|
198 | 200 | self._inline_scan(self._lgt) |
|
199 | 201 | self._extra = [] |
|
200 | 202 | |
|
201 | 203 | def _inline_scan(self, lgt): |
|
202 | 204 | off = 0 |
|
203 | 205 | if lgt is not None: |
|
204 | 206 | self._offsets = [0] * lgt |
|
205 | 207 | count = 0 |
|
206 |
while off <= len(self._data) - self. |
|
|
208 | while off <= len(self._data) - self.entry_size: | |
|
207 | 209 | start = off + self.big_int_size |
|
208 | 210 | (s,) = struct.unpack( |
|
209 | 211 | b'>i', |
|
210 | 212 | self._data[start : start + self.int_size], |
|
211 | 213 | ) |
|
212 | 214 | if lgt is not None: |
|
213 | 215 | self._offsets[count] = off |
|
214 | 216 | count += 1 |
|
215 |
off += self. |
|
|
217 | off += self.entry_size + s | |
|
216 | 218 | if off != len(self._data): |
|
217 | 219 | raise ValueError(b"corrupted data") |
|
218 | 220 | return count |
|
219 | 221 | |
|
220 | 222 | def __delitem__(self, i): |
|
221 | 223 | if not isinstance(i, slice) or not i.stop == -1 or i.step is not None: |
|
222 | 224 | raise ValueError(b"deleting slices only supports a:-1 with step 1") |
|
223 | 225 | i = i.start |
|
224 | 226 | self._check_index(i) |
|
225 | 227 | self._stripnodes(i) |
|
226 | 228 | if i < self._lgt: |
|
227 | 229 | self._offsets = self._offsets[:i] |
|
228 | 230 | self._lgt = i |
|
229 | 231 | self._extra = [] |
|
230 | 232 | else: |
|
231 | 233 | self._extra = self._extra[: i - self._lgt] |
|
232 | 234 | |
|
233 | 235 | def _calculate_index(self, i): |
|
234 | 236 | return self._offsets[i] |
|
235 | 237 | |
|
236 | 238 | |
|
237 | 239 | def parse_index2(data, inline, revlogv2=False): |
|
238 | 240 | if not inline: |
|
239 | 241 | cls = IndexObject2 if revlogv2 else IndexObject |
|
240 | 242 | return cls(data), None |
|
241 | 243 | cls = InlinedIndexObject2 if revlogv2 else InlinedIndexObject |
|
242 | 244 | return cls(data, inline), (0, data) |
|
243 | 245 | |
|
244 | 246 | |
|
245 | 247 | class Index2Mixin(object): |
|
246 | 248 | index_format = revlog_constants.INDEX_ENTRY_V2 |
|
247 | index_size = revlog_constants.INDEX_ENTRY_V2.size | |
|
248 | 249 | null_item = (0, 0, 0, -1, -1, -1, -1, nullid, 0, 0) |
|
249 | 250 | |
|
250 | 251 | def replace_sidedata_info(self, i, sidedata_offset, sidedata_length): |
|
251 | 252 | """ |
|
252 | 253 | Replace an existing index entry's sidedata offset and length with new |
|
253 | 254 | ones. |
|
254 | 255 | This cannot be used outside of the context of sidedata rewriting, |
|
255 | 256 | inside the transaction that creates the revision `i`. |
|
256 | 257 | """ |
|
257 | 258 | if i < 0: |
|
258 | 259 | raise KeyError |
|
259 | 260 | self._check_index(i) |
|
260 | 261 | sidedata_format = b">Qi" |
|
261 | 262 | packed_size = struct.calcsize(sidedata_format) |
|
262 | 263 | if i >= self._lgt: |
|
263 | 264 | packed = _pack(sidedata_format, sidedata_offset, sidedata_length) |
|
264 | 265 | old = self._extra[i - self._lgt] |
|
265 | 266 | new = old[:64] + packed + old[64 + packed_size :] |
|
266 | 267 | self._extra[i - self._lgt] = new |
|
267 | 268 | else: |
|
268 | 269 | msg = b"cannot rewrite entries outside of this transaction" |
|
269 | 270 | raise KeyError(msg) |
|
270 | 271 | |
|
271 | 272 | |
|
272 | 273 | class IndexObject2(Index2Mixin, IndexObject): |
|
273 | 274 | pass |
|
274 | 275 | |
|
275 | 276 | |
|
276 | 277 | class InlinedIndexObject2(Index2Mixin, InlinedIndexObject): |
|
277 | 278 | def _inline_scan(self, lgt): |
|
278 | 279 | sidedata_length_pos = 72 |
|
279 | 280 | off = 0 |
|
280 | 281 | if lgt is not None: |
|
281 | 282 | self._offsets = [0] * lgt |
|
282 | 283 | count = 0 |
|
283 |
while off <= len(self._data) - self. |
|
|
284 | while off <= len(self._data) - self.entry_size: | |
|
284 | 285 | start = off + self.big_int_size |
|
285 | 286 | (data_size,) = struct.unpack( |
|
286 | 287 | b'>i', |
|
287 | 288 | self._data[start : start + self.int_size], |
|
288 | 289 | ) |
|
289 | 290 | start = off + sidedata_length_pos |
|
290 | 291 | (side_data_size,) = struct.unpack( |
|
291 | 292 | b'>i', self._data[start : start + self.int_size] |
|
292 | 293 | ) |
|
293 | 294 | if lgt is not None: |
|
294 | 295 | self._offsets[count] = off |
|
295 | 296 | count += 1 |
|
296 |
off += self. |
|
|
297 | off += self.entry_size + data_size + side_data_size | |
|
297 | 298 | if off != len(self._data): |
|
298 | 299 | raise ValueError(b"corrupted data") |
|
299 | 300 | return count |
|
300 | 301 | |
|
301 | 302 | |
|
302 | 303 | def parse_index_devel_nodemap(data, inline): |
|
303 | 304 | """like parse_index2, but alway return a PersistentNodeMapIndexObject""" |
|
304 | 305 | return PersistentNodeMapIndexObject(data), None |
|
305 | 306 | |
|
306 | 307 | |
|
307 | 308 | def parse_dirstate(dmap, copymap, st): |
|
308 | 309 | parents = [st[:20], st[20:40]] |
|
309 | 310 | # dereference fields so they will be local in loop |
|
310 | 311 | format = b">cllll" |
|
311 | 312 | e_size = struct.calcsize(format) |
|
312 | 313 | pos1 = 40 |
|
313 | 314 | l = len(st) |
|
314 | 315 | |
|
315 | 316 | # the inner loop |
|
316 | 317 | while pos1 < l: |
|
317 | 318 | pos2 = pos1 + e_size |
|
318 | 319 | e = _unpack(b">cllll", st[pos1:pos2]) # a literal here is faster |
|
319 | 320 | pos1 = pos2 + e[4] |
|
320 | 321 | f = st[pos2:pos1] |
|
321 | 322 | if b'\0' in f: |
|
322 | 323 | f, c = f.split(b'\0') |
|
323 | 324 | copymap[f] = c |
|
324 | 325 | dmap[f] = e[:4] |
|
325 | 326 | return parents |
|
326 | 327 | |
|
327 | 328 | |
|
328 | 329 | def pack_dirstate(dmap, copymap, pl, now): |
|
329 | 330 | now = int(now) |
|
330 | 331 | cs = stringio() |
|
331 | 332 | write = cs.write |
|
332 | 333 | write(b"".join(pl)) |
|
333 | 334 | for f, e in pycompat.iteritems(dmap): |
|
334 | 335 | if e[0] == b'n' and e[3] == now: |
|
335 | 336 | # The file was last modified "simultaneously" with the current |
|
336 | 337 | # write to dirstate (i.e. within the same second for file- |
|
337 | 338 | # systems with a granularity of 1 sec). This commonly happens |
|
338 | 339 | # for at least a couple of files on 'update'. |
|
339 | 340 | # The user could change the file without changing its size |
|
340 | 341 | # within the same second. Invalidate the file's mtime in |
|
341 | 342 | # dirstate, forcing future 'status' calls to compare the |
|
342 | 343 | # contents of the file if the size is the same. This prevents |
|
343 | 344 | # mistakenly treating such files as clean. |
|
344 | 345 | e = dirstatetuple(e[0], e[1], e[2], -1) |
|
345 | 346 | dmap[f] = e |
|
346 | 347 | |
|
347 | 348 | if f in copymap: |
|
348 | 349 | f = b"%s\0%s" % (f, copymap[f]) |
|
349 | 350 | e = _pack(b">cllll", e[0], e[1], e[2], e[3], len(f)) |
|
350 | 351 | write(e) |
|
351 | 352 | write(f) |
|
352 | 353 | return cs.getvalue() |
@@ -1,3249 +1,3242 b'' | |||
|
1 | 1 | # revlog.py - storage back-end for mercurial |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2005-2007 Olivia Mackall <olivia@selenic.com> |
|
4 | 4 | # |
|
5 | 5 | # This software may be used and distributed according to the terms of the |
|
6 | 6 | # GNU General Public License version 2 or any later version. |
|
7 | 7 | |
|
8 | 8 | """Storage back-end for Mercurial. |
|
9 | 9 | |
|
10 | 10 | This provides efficient delta storage with O(1) retrieve and append |
|
11 | 11 | and O(changes) merge between branches. |
|
12 | 12 | """ |
|
13 | 13 | |
|
14 | 14 | from __future__ import absolute_import |
|
15 | 15 | |
|
16 | 16 | import binascii |
|
17 | 17 | import collections |
|
18 | 18 | import contextlib |
|
19 | 19 | import errno |
|
20 | 20 | import io |
|
21 | 21 | import os |
|
22 | 22 | import struct |
|
23 | 23 | import zlib |
|
24 | 24 | |
|
25 | 25 | # import stuff from node for others to import from revlog |
|
26 | 26 | from .node import ( |
|
27 | 27 | bin, |
|
28 | 28 | hex, |
|
29 | 29 | nullhex, |
|
30 | 30 | nullid, |
|
31 | 31 | nullrev, |
|
32 | 32 | sha1nodeconstants, |
|
33 | 33 | short, |
|
34 | 34 | wdirfilenodeids, |
|
35 | 35 | wdirhex, |
|
36 | 36 | wdirid, |
|
37 | 37 | wdirrev, |
|
38 | 38 | ) |
|
39 | 39 | from .i18n import _ |
|
40 | 40 | from .pycompat import getattr |
|
41 | 41 | from .revlogutils.constants import ( |
|
42 | 42 | FLAG_GENERALDELTA, |
|
43 | 43 | FLAG_INLINE_DATA, |
|
44 | 44 | INDEX_ENTRY_V0, |
|
45 | 45 | INDEX_ENTRY_V1, |
|
46 | 46 | INDEX_ENTRY_V2, |
|
47 | 47 | INDEX_HEADER, |
|
48 | 48 | REVLOGV0, |
|
49 | 49 | REVLOGV1, |
|
50 | 50 | REVLOGV1_FLAGS, |
|
51 | 51 | REVLOGV2, |
|
52 | 52 | REVLOGV2_FLAGS, |
|
53 | 53 | REVLOG_DEFAULT_FLAGS, |
|
54 | 54 | REVLOG_DEFAULT_FORMAT, |
|
55 | 55 | REVLOG_DEFAULT_VERSION, |
|
56 | 56 | ) |
|
57 | 57 | from .revlogutils.flagutil import ( |
|
58 | 58 | REVIDX_DEFAULT_FLAGS, |
|
59 | 59 | REVIDX_ELLIPSIS, |
|
60 | 60 | REVIDX_EXTSTORED, |
|
61 | 61 | REVIDX_FLAGS_ORDER, |
|
62 | 62 | REVIDX_HASCOPIESINFO, |
|
63 | 63 | REVIDX_ISCENSORED, |
|
64 | 64 | REVIDX_RAWTEXT_CHANGING_FLAGS, |
|
65 | 65 | REVIDX_SIDEDATA, |
|
66 | 66 | ) |
|
67 | 67 | from .thirdparty import attr |
|
68 | 68 | from . import ( |
|
69 | 69 | ancestor, |
|
70 | 70 | dagop, |
|
71 | 71 | error, |
|
72 | 72 | mdiff, |
|
73 | 73 | policy, |
|
74 | 74 | pycompat, |
|
75 | 75 | templatefilters, |
|
76 | 76 | util, |
|
77 | 77 | ) |
|
78 | 78 | from .interfaces import ( |
|
79 | 79 | repository, |
|
80 | 80 | util as interfaceutil, |
|
81 | 81 | ) |
|
82 | 82 | from .revlogutils import ( |
|
83 | 83 | deltas as deltautil, |
|
84 | 84 | flagutil, |
|
85 | 85 | nodemap as nodemaputil, |
|
86 | 86 | sidedata as sidedatautil, |
|
87 | 87 | ) |
|
88 | 88 | from .utils import ( |
|
89 | 89 | storageutil, |
|
90 | 90 | stringutil, |
|
91 | 91 | ) |
|
92 | 92 | |
|
93 | 93 | # blanked usage of all the name to prevent pyflakes constraints |
|
94 | 94 | # We need these name available in the module for extensions. |
|
95 | 95 | REVLOGV0 |
|
96 | 96 | REVLOGV1 |
|
97 | 97 | REVLOGV2 |
|
98 | 98 | FLAG_INLINE_DATA |
|
99 | 99 | FLAG_GENERALDELTA |
|
100 | 100 | REVLOG_DEFAULT_FLAGS |
|
101 | 101 | REVLOG_DEFAULT_FORMAT |
|
102 | 102 | REVLOG_DEFAULT_VERSION |
|
103 | 103 | REVLOGV1_FLAGS |
|
104 | 104 | REVLOGV2_FLAGS |
|
105 | 105 | REVIDX_ISCENSORED |
|
106 | 106 | REVIDX_ELLIPSIS |
|
107 | 107 | REVIDX_SIDEDATA |
|
108 | 108 | REVIDX_HASCOPIESINFO |
|
109 | 109 | REVIDX_EXTSTORED |
|
110 | 110 | REVIDX_DEFAULT_FLAGS |
|
111 | 111 | REVIDX_FLAGS_ORDER |
|
112 | 112 | REVIDX_RAWTEXT_CHANGING_FLAGS |
|
113 | 113 | |
|
114 | 114 | parsers = policy.importmod('parsers') |
|
115 | 115 | rustancestor = policy.importrust('ancestor') |
|
116 | 116 | rustdagop = policy.importrust('dagop') |
|
117 | 117 | rustrevlog = policy.importrust('revlog') |
|
118 | 118 | |
|
119 | 119 | # Aliased for performance. |
|
120 | 120 | _zlibdecompress = zlib.decompress |
|
121 | 121 | |
|
122 | 122 | # max size of revlog with inline data |
|
123 | 123 | _maxinline = 131072 |
|
124 | 124 | _chunksize = 1048576 |
|
125 | 125 | |
|
126 | 126 | # Flag processors for REVIDX_ELLIPSIS. |
|
127 | 127 | def ellipsisreadprocessor(rl, text): |
|
128 | 128 | return text, False |
|
129 | 129 | |
|
130 | 130 | |
|
131 | 131 | def ellipsiswriteprocessor(rl, text): |
|
132 | 132 | return text, False |
|
133 | 133 | |
|
134 | 134 | |
|
135 | 135 | def ellipsisrawprocessor(rl, text): |
|
136 | 136 | return False |
|
137 | 137 | |
|
138 | 138 | |
|
139 | 139 | ellipsisprocessor = ( |
|
140 | 140 | ellipsisreadprocessor, |
|
141 | 141 | ellipsiswriteprocessor, |
|
142 | 142 | ellipsisrawprocessor, |
|
143 | 143 | ) |
|
144 | 144 | |
|
145 | 145 | |
|
146 | 146 | def getoffset(q): |
|
147 | 147 | return int(q >> 16) |
|
148 | 148 | |
|
149 | 149 | |
|
150 | 150 | def gettype(q): |
|
151 | 151 | return int(q & 0xFFFF) |
|
152 | 152 | |
|
153 | 153 | |
|
154 | 154 | def offset_type(offset, type): |
|
155 | 155 | if (type & ~flagutil.REVIDX_KNOWN_FLAGS) != 0: |
|
156 | 156 | raise ValueError(b'unknown revlog index flags') |
|
157 | 157 | return int(int(offset) << 16 | type) |
|
158 | 158 | |
|
159 | 159 | |
|
160 | 160 | def _verify_revision(rl, skipflags, state, node): |
|
161 | 161 | """Verify the integrity of the given revlog ``node`` while providing a hook |
|
162 | 162 | point for extensions to influence the operation.""" |
|
163 | 163 | if skipflags: |
|
164 | 164 | state[b'skipread'].add(node) |
|
165 | 165 | else: |
|
166 | 166 | # Side-effect: read content and verify hash. |
|
167 | 167 | rl.revision(node) |
|
168 | 168 | |
|
169 | 169 | |
|
170 | 170 | # True if a fast implementation for persistent-nodemap is available |
|
171 | 171 | # |
|
172 | 172 | # We also consider we have a "fast" implementation in "pure" python because |
|
173 | 173 | # people using pure don't really have performance consideration (and a |
|
174 | 174 | # wheelbarrow of other slowness source) |
|
175 | 175 | HAS_FAST_PERSISTENT_NODEMAP = rustrevlog is not None or util.safehasattr( |
|
176 | 176 | parsers, 'BaseIndexObject' |
|
177 | 177 | ) |
|
178 | 178 | |
|
179 | 179 | |
|
180 | 180 | @attr.s(slots=True, frozen=True) |
|
181 | 181 | class _revisioninfo(object): |
|
182 | 182 | """Information about a revision that allows building its fulltext |
|
183 | 183 | node: expected hash of the revision |
|
184 | 184 | p1, p2: parent revs of the revision |
|
185 | 185 | btext: built text cache consisting of a one-element list |
|
186 | 186 | cachedelta: (baserev, uncompressed_delta) or None |
|
187 | 187 | flags: flags associated to the revision storage |
|
188 | 188 | |
|
189 | 189 | One of btext[0] or cachedelta must be set. |
|
190 | 190 | """ |
|
191 | 191 | |
|
192 | 192 | node = attr.ib() |
|
193 | 193 | p1 = attr.ib() |
|
194 | 194 | p2 = attr.ib() |
|
195 | 195 | btext = attr.ib() |
|
196 | 196 | textlen = attr.ib() |
|
197 | 197 | cachedelta = attr.ib() |
|
198 | 198 | flags = attr.ib() |
|
199 | 199 | |
|
200 | 200 | |
|
201 | 201 | @interfaceutil.implementer(repository.irevisiondelta) |
|
202 | 202 | @attr.s(slots=True) |
|
203 | 203 | class revlogrevisiondelta(object): |
|
204 | 204 | node = attr.ib() |
|
205 | 205 | p1node = attr.ib() |
|
206 | 206 | p2node = attr.ib() |
|
207 | 207 | basenode = attr.ib() |
|
208 | 208 | flags = attr.ib() |
|
209 | 209 | baserevisionsize = attr.ib() |
|
210 | 210 | revision = attr.ib() |
|
211 | 211 | delta = attr.ib() |
|
212 | 212 | sidedata = attr.ib() |
|
213 | 213 | linknode = attr.ib(default=None) |
|
214 | 214 | |
|
215 | 215 | |
|
216 | 216 | @interfaceutil.implementer(repository.iverifyproblem) |
|
217 | 217 | @attr.s(frozen=True) |
|
218 | 218 | class revlogproblem(object): |
|
219 | 219 | warning = attr.ib(default=None) |
|
220 | 220 | error = attr.ib(default=None) |
|
221 | 221 | node = attr.ib(default=None) |
|
222 | 222 | |
|
223 | 223 | |
|
224 | 224 | class revlogoldindex(list): |
|
225 | entry_size = INDEX_ENTRY_V0.size | |
|
226 | ||
|
225 | 227 | @property |
|
226 | 228 | def nodemap(self): |
|
227 | 229 | msg = b"index.nodemap is deprecated, use index.[has_node|rev|get_rev]" |
|
228 | 230 | util.nouideprecwarn(msg, b'5.3', stacklevel=2) |
|
229 | 231 | return self._nodemap |
|
230 | 232 | |
|
231 | 233 | @util.propertycache |
|
232 | 234 | def _nodemap(self): |
|
233 | 235 | nodemap = nodemaputil.NodeMap({nullid: nullrev}) |
|
234 | 236 | for r in range(0, len(self)): |
|
235 | 237 | n = self[r][7] |
|
236 | 238 | nodemap[n] = r |
|
237 | 239 | return nodemap |
|
238 | 240 | |
|
239 | 241 | def has_node(self, node): |
|
240 | 242 | """return True if the node exist in the index""" |
|
241 | 243 | return node in self._nodemap |
|
242 | 244 | |
|
243 | 245 | def rev(self, node): |
|
244 | 246 | """return a revision for a node |
|
245 | 247 | |
|
246 | 248 | If the node is unknown, raise a RevlogError""" |
|
247 | 249 | return self._nodemap[node] |
|
248 | 250 | |
|
249 | 251 | def get_rev(self, node): |
|
250 | 252 | """return a revision for a node |
|
251 | 253 | |
|
252 | 254 | If the node is unknown, return None""" |
|
253 | 255 | return self._nodemap.get(node) |
|
254 | 256 | |
|
255 | 257 | def append(self, tup): |
|
256 | 258 | self._nodemap[tup[7]] = len(self) |
|
257 | 259 | super(revlogoldindex, self).append(tup) |
|
258 | 260 | |
|
259 | 261 | def __delitem__(self, i): |
|
260 | 262 | if not isinstance(i, slice) or not i.stop == -1 or i.step is not None: |
|
261 | 263 | raise ValueError(b"deleting slices only supports a:-1 with step 1") |
|
262 | 264 | for r in pycompat.xrange(i.start, len(self)): |
|
263 | 265 | del self._nodemap[self[r][7]] |
|
264 | 266 | super(revlogoldindex, self).__delitem__(i) |
|
265 | 267 | |
|
266 | 268 | def clearcaches(self): |
|
267 | 269 | self.__dict__.pop('_nodemap', None) |
|
268 | 270 | |
|
269 | 271 | def __getitem__(self, i): |
|
270 | 272 | if i == -1: |
|
271 | 273 | return (0, 0, 0, -1, -1, -1, -1, nullid) |
|
272 | 274 | return list.__getitem__(self, i) |
|
273 | 275 | |
|
274 | 276 | |
|
275 | 277 | class revlogoldio(object): |
|
276 | def __init__(self): | |
|
277 | self.size = INDEX_ENTRY_V0.size | |
|
278 | ||
|
279 | 278 | def parseindex(self, data, inline): |
|
280 |
s = |
|
|
279 | s = INDEX_ENTRY_V0.size | |
|
281 | 280 | index = [] |
|
282 | 281 | nodemap = nodemaputil.NodeMap({nullid: nullrev}) |
|
283 | 282 | n = off = 0 |
|
284 | 283 | l = len(data) |
|
285 | 284 | while off + s <= l: |
|
286 | 285 | cur = data[off : off + s] |
|
287 | 286 | off += s |
|
288 | 287 | e = INDEX_ENTRY_V0.unpack(cur) |
|
289 | 288 | # transform to revlogv1 format |
|
290 | 289 | e2 = ( |
|
291 | 290 | offset_type(e[0], 0), |
|
292 | 291 | e[1], |
|
293 | 292 | -1, |
|
294 | 293 | e[2], |
|
295 | 294 | e[3], |
|
296 | 295 | nodemap.get(e[4], nullrev), |
|
297 | 296 | nodemap.get(e[5], nullrev), |
|
298 | 297 | e[6], |
|
299 | 298 | ) |
|
300 | 299 | index.append(e2) |
|
301 | 300 | nodemap[e[6]] = n |
|
302 | 301 | n += 1 |
|
303 | 302 | |
|
304 | 303 | index = revlogoldindex(index) |
|
305 | 304 | return index, None |
|
306 | 305 | |
|
307 | 306 | def packentry(self, entry, node, version, rev): |
|
308 | 307 | """return the binary representation of an entry |
|
309 | 308 | |
|
310 | 309 | entry: a tuple containing all the values (see index.__getitem__) |
|
311 | 310 | node: a callback to convert a revision to nodeid |
|
312 | 311 | version: the changelog version |
|
313 | 312 | rev: the revision number |
|
314 | 313 | """ |
|
315 | 314 | if gettype(entry[0]): |
|
316 | 315 | raise error.RevlogError( |
|
317 | 316 | _(b'index entry flags need revlog version 1') |
|
318 | 317 | ) |
|
319 | 318 | e2 = ( |
|
320 | 319 | getoffset(entry[0]), |
|
321 | 320 | entry[1], |
|
322 | 321 | entry[3], |
|
323 | 322 | entry[4], |
|
324 | 323 | node(entry[5]), |
|
325 | 324 | node(entry[6]), |
|
326 | 325 | entry[7], |
|
327 | 326 | ) |
|
328 | 327 | return INDEX_ENTRY_V0.pack(*e2) |
|
329 | 328 | |
|
330 | 329 | |
|
331 | 330 | # corresponds to uncompressed length of indexformatng (2 gigs, 4-byte |
|
332 | 331 | # signed integer) |
|
333 | 332 | _maxentrysize = 0x7FFFFFFF |
|
334 | 333 | |
|
335 | 334 | |
|
336 | 335 | class revlogio(object): |
|
337 | def __init__(self): | |
|
338 | self.size = INDEX_ENTRY_V1.size | |
|
339 | ||
|
340 | 336 | def parseindex(self, data, inline): |
|
341 | 337 | # call the C implementation to parse the index data |
|
342 | 338 | index, cache = parsers.parse_index2(data, inline) |
|
343 | 339 | return index, cache |
|
344 | 340 | |
|
345 | 341 | def packentry(self, entry, node, version, rev): |
|
346 | 342 | p = INDEX_ENTRY_V1.pack(*entry) |
|
347 | 343 | if rev == 0: |
|
348 | 344 | p = INDEX_HEADER.pack(version) + p[4:] |
|
349 | 345 | return p |
|
350 | 346 | |
|
351 | 347 | |
|
352 | 348 | class revlogv2io(object): |
|
353 | def __init__(self): | |
|
354 | self.size = INDEX_ENTRY_V2.size | |
|
355 | ||
|
356 | 349 | def parseindex(self, data, inline): |
|
357 | 350 | index, cache = parsers.parse_index2(data, inline, revlogv2=True) |
|
358 | 351 | return index, cache |
|
359 | 352 | |
|
360 | 353 | def packentry(self, entry, node, version, rev): |
|
361 | 354 | p = INDEX_ENTRY_V2.pack(*entry) |
|
362 | 355 | if rev == 0: |
|
363 | 356 | p = INDEX_HEADER.pack(version) + p[4:] |
|
364 | 357 | return p |
|
365 | 358 | |
|
366 | 359 | |
|
367 | 360 | NodemapRevlogIO = None |
|
368 | 361 | |
|
369 | 362 | if util.safehasattr(parsers, 'parse_index_devel_nodemap'): |
|
370 | 363 | |
|
371 | 364 | class NodemapRevlogIO(revlogio): |
|
372 | 365 | """A debug oriented IO class that return a PersistentNodeMapIndexObject |
|
373 | 366 | |
|
374 | 367 | The PersistentNodeMapIndexObject object is meant to test the persistent nodemap feature. |
|
375 | 368 | """ |
|
376 | 369 | |
|
377 | 370 | def parseindex(self, data, inline): |
|
378 | 371 | index, cache = parsers.parse_index_devel_nodemap(data, inline) |
|
379 | 372 | return index, cache |
|
380 | 373 | |
|
381 | 374 | |
|
382 | 375 | class rustrevlogio(revlogio): |
|
383 | 376 | def parseindex(self, data, inline): |
|
384 | 377 | index, cache = super(rustrevlogio, self).parseindex(data, inline) |
|
385 | 378 | return rustrevlog.MixedIndex(index), cache |
|
386 | 379 | |
|
387 | 380 | |
|
388 | 381 | class revlog(object): |
|
389 | 382 | """ |
|
390 | 383 | the underlying revision storage object |
|
391 | 384 | |
|
392 | 385 | A revlog consists of two parts, an index and the revision data. |
|
393 | 386 | |
|
394 | 387 | The index is a file with a fixed record size containing |
|
395 | 388 | information on each revision, including its nodeid (hash), the |
|
396 | 389 | nodeids of its parents, the position and offset of its data within |
|
397 | 390 | the data file, and the revision it's based on. Finally, each entry |
|
398 | 391 | contains a linkrev entry that can serve as a pointer to external |
|
399 | 392 | data. |
|
400 | 393 | |
|
401 | 394 | The revision data itself is a linear collection of data chunks. |
|
402 | 395 | Each chunk represents a revision and is usually represented as a |
|
403 | 396 | delta against the previous chunk. To bound lookup time, runs of |
|
404 | 397 | deltas are limited to about 2 times the length of the original |
|
405 | 398 | version data. This makes retrieval of a version proportional to |
|
406 | 399 | its size, or O(1) relative to the number of revisions. |
|
407 | 400 | |
|
408 | 401 | Both pieces of the revlog are written to in an append-only |
|
409 | 402 | fashion, which means we never need to rewrite a file to insert or |
|
410 | 403 | remove data, and can use some simple techniques to avoid the need |
|
411 | 404 | for locking while reading. |
|
412 | 405 | |
|
413 | 406 | If checkambig, indexfile is opened with checkambig=True at |
|
414 | 407 | writing, to avoid file stat ambiguity. |
|
415 | 408 | |
|
416 | 409 | If mmaplargeindex is True, and an mmapindexthreshold is set, the |
|
417 | 410 | index will be mmapped rather than read if it is larger than the |
|
418 | 411 | configured threshold. |
|
419 | 412 | |
|
420 | 413 | If censorable is True, the revlog can have censored revisions. |
|
421 | 414 | |
|
422 | 415 | If `upperboundcomp` is not None, this is the expected maximal gain from |
|
423 | 416 | compression for the data content. |
|
424 | 417 | |
|
425 | 418 | `concurrencychecker` is an optional function that receives 3 arguments: a |
|
426 | 419 | file handle, a filename, and an expected position. It should check whether |
|
427 | 420 | the current position in the file handle is valid, and log/warn/fail (by |
|
428 | 421 | raising). |
|
429 | 422 | """ |
|
430 | 423 | |
|
431 | 424 | _flagserrorclass = error.RevlogError |
|
432 | 425 | |
|
433 | 426 | def __init__( |
|
434 | 427 | self, |
|
435 | 428 | opener, |
|
436 | 429 | indexfile, |
|
437 | 430 | datafile=None, |
|
438 | 431 | checkambig=False, |
|
439 | 432 | mmaplargeindex=False, |
|
440 | 433 | censorable=False, |
|
441 | 434 | upperboundcomp=None, |
|
442 | 435 | persistentnodemap=False, |
|
443 | 436 | concurrencychecker=None, |
|
444 | 437 | ): |
|
445 | 438 | """ |
|
446 | 439 | create a revlog object |
|
447 | 440 | |
|
448 | 441 | opener is a function that abstracts the file opening operation |
|
449 | 442 | and can be used to implement COW semantics or the like. |
|
450 | 443 | |
|
451 | 444 | """ |
|
452 | 445 | self.upperboundcomp = upperboundcomp |
|
453 | 446 | self.indexfile = indexfile |
|
454 | 447 | self.datafile = datafile or (indexfile[:-2] + b".d") |
|
455 | 448 | self.nodemap_file = None |
|
456 | 449 | if persistentnodemap: |
|
457 | 450 | self.nodemap_file = nodemaputil.get_nodemap_file( |
|
458 | 451 | opener, self.indexfile |
|
459 | 452 | ) |
|
460 | 453 | |
|
461 | 454 | self.opener = opener |
|
462 | 455 | # When True, indexfile is opened with checkambig=True at writing, to |
|
463 | 456 | # avoid file stat ambiguity. |
|
464 | 457 | self._checkambig = checkambig |
|
465 | 458 | self._mmaplargeindex = mmaplargeindex |
|
466 | 459 | self._censorable = censorable |
|
467 | 460 | # 3-tuple of (node, rev, text) for a raw revision. |
|
468 | 461 | self._revisioncache = None |
|
469 | 462 | # Maps rev to chain base rev. |
|
470 | 463 | self._chainbasecache = util.lrucachedict(100) |
|
471 | 464 | # 2-tuple of (offset, data) of raw data from the revlog at an offset. |
|
472 | 465 | self._chunkcache = (0, b'') |
|
473 | 466 | # How much data to read and cache into the raw revlog data cache. |
|
474 | 467 | self._chunkcachesize = 65536 |
|
475 | 468 | self._maxchainlen = None |
|
476 | 469 | self._deltabothparents = True |
|
477 | 470 | self.index = None |
|
478 | 471 | self._nodemap_docket = None |
|
479 | 472 | # Mapping of partial identifiers to full nodes. |
|
480 | 473 | self._pcache = {} |
|
481 | 474 | # Mapping of revision integer to full node. |
|
482 | 475 | self._compengine = b'zlib' |
|
483 | 476 | self._compengineopts = {} |
|
484 | 477 | self._maxdeltachainspan = -1 |
|
485 | 478 | self._withsparseread = False |
|
486 | 479 | self._sparserevlog = False |
|
487 | 480 | self._srdensitythreshold = 0.50 |
|
488 | 481 | self._srmingapsize = 262144 |
|
489 | 482 | |
|
490 | 483 | # Make copy of flag processors so each revlog instance can support |
|
491 | 484 | # custom flags. |
|
492 | 485 | self._flagprocessors = dict(flagutil.flagprocessors) |
|
493 | 486 | |
|
494 | 487 | # 2-tuple of file handles being used for active writing. |
|
495 | 488 | self._writinghandles = None |
|
496 | 489 | |
|
497 | 490 | self._loadindex() |
|
498 | 491 | |
|
499 | 492 | self._concurrencychecker = concurrencychecker |
|
500 | 493 | |
|
501 | 494 | def _loadindex(self): |
|
502 | 495 | mmapindexthreshold = None |
|
503 | 496 | opts = self.opener.options |
|
504 | 497 | |
|
505 | 498 | if b'revlogv2' in opts: |
|
506 | 499 | newversionflags = REVLOGV2 | FLAG_INLINE_DATA |
|
507 | 500 | elif b'revlogv1' in opts: |
|
508 | 501 | newversionflags = REVLOGV1 | FLAG_INLINE_DATA |
|
509 | 502 | if b'generaldelta' in opts: |
|
510 | 503 | newversionflags |= FLAG_GENERALDELTA |
|
511 | 504 | elif b'revlogv0' in self.opener.options: |
|
512 | 505 | newversionflags = REVLOGV0 |
|
513 | 506 | else: |
|
514 | 507 | newversionflags = REVLOG_DEFAULT_VERSION |
|
515 | 508 | |
|
516 | 509 | if b'chunkcachesize' in opts: |
|
517 | 510 | self._chunkcachesize = opts[b'chunkcachesize'] |
|
518 | 511 | if b'maxchainlen' in opts: |
|
519 | 512 | self._maxchainlen = opts[b'maxchainlen'] |
|
520 | 513 | if b'deltabothparents' in opts: |
|
521 | 514 | self._deltabothparents = opts[b'deltabothparents'] |
|
522 | 515 | self._lazydelta = bool(opts.get(b'lazydelta', True)) |
|
523 | 516 | self._lazydeltabase = False |
|
524 | 517 | if self._lazydelta: |
|
525 | 518 | self._lazydeltabase = bool(opts.get(b'lazydeltabase', False)) |
|
526 | 519 | if b'compengine' in opts: |
|
527 | 520 | self._compengine = opts[b'compengine'] |
|
528 | 521 | if b'zlib.level' in opts: |
|
529 | 522 | self._compengineopts[b'zlib.level'] = opts[b'zlib.level'] |
|
530 | 523 | if b'zstd.level' in opts: |
|
531 | 524 | self._compengineopts[b'zstd.level'] = opts[b'zstd.level'] |
|
532 | 525 | if b'maxdeltachainspan' in opts: |
|
533 | 526 | self._maxdeltachainspan = opts[b'maxdeltachainspan'] |
|
534 | 527 | if self._mmaplargeindex and b'mmapindexthreshold' in opts: |
|
535 | 528 | mmapindexthreshold = opts[b'mmapindexthreshold'] |
|
536 | 529 | self.hassidedata = bool(opts.get(b'side-data', False)) |
|
537 | 530 | self._sparserevlog = bool(opts.get(b'sparse-revlog', False)) |
|
538 | 531 | withsparseread = bool(opts.get(b'with-sparse-read', False)) |
|
539 | 532 | # sparse-revlog forces sparse-read |
|
540 | 533 | self._withsparseread = self._sparserevlog or withsparseread |
|
541 | 534 | if b'sparse-read-density-threshold' in opts: |
|
542 | 535 | self._srdensitythreshold = opts[b'sparse-read-density-threshold'] |
|
543 | 536 | if b'sparse-read-min-gap-size' in opts: |
|
544 | 537 | self._srmingapsize = opts[b'sparse-read-min-gap-size'] |
|
545 | 538 | if opts.get(b'enableellipsis'): |
|
546 | 539 | self._flagprocessors[REVIDX_ELLIPSIS] = ellipsisprocessor |
|
547 | 540 | |
|
548 | 541 | # revlog v0 doesn't have flag processors |
|
549 | 542 | for flag, processor in pycompat.iteritems( |
|
550 | 543 | opts.get(b'flagprocessors', {}) |
|
551 | 544 | ): |
|
552 | 545 | flagutil.insertflagprocessor(flag, processor, self._flagprocessors) |
|
553 | 546 | |
|
554 | 547 | if self._chunkcachesize <= 0: |
|
555 | 548 | raise error.RevlogError( |
|
556 | 549 | _(b'revlog chunk cache size %r is not greater than 0') |
|
557 | 550 | % self._chunkcachesize |
|
558 | 551 | ) |
|
559 | 552 | elif self._chunkcachesize & (self._chunkcachesize - 1): |
|
560 | 553 | raise error.RevlogError( |
|
561 | 554 | _(b'revlog chunk cache size %r is not a power of 2') |
|
562 | 555 | % self._chunkcachesize |
|
563 | 556 | ) |
|
564 | 557 | |
|
565 | 558 | indexdata = b'' |
|
566 | 559 | self._initempty = True |
|
567 | 560 | try: |
|
568 | 561 | with self._indexfp() as f: |
|
569 | 562 | if ( |
|
570 | 563 | mmapindexthreshold is not None |
|
571 | 564 | and self.opener.fstat(f).st_size >= mmapindexthreshold |
|
572 | 565 | ): |
|
573 | 566 | # TODO: should .close() to release resources without |
|
574 | 567 | # relying on Python GC |
|
575 | 568 | indexdata = util.buffer(util.mmapread(f)) |
|
576 | 569 | else: |
|
577 | 570 | indexdata = f.read() |
|
578 | 571 | if len(indexdata) > 0: |
|
579 | 572 | versionflags = INDEX_HEADER.unpack(indexdata[:4])[0] |
|
580 | 573 | self._initempty = False |
|
581 | 574 | else: |
|
582 | 575 | versionflags = newversionflags |
|
583 | 576 | except IOError as inst: |
|
584 | 577 | if inst.errno != errno.ENOENT: |
|
585 | 578 | raise |
|
586 | 579 | |
|
587 | 580 | versionflags = newversionflags |
|
588 | 581 | |
|
589 | 582 | self.version = versionflags |
|
590 | 583 | |
|
591 | 584 | flags = versionflags & ~0xFFFF |
|
592 | 585 | fmt = versionflags & 0xFFFF |
|
593 | 586 | |
|
594 | 587 | if fmt == REVLOGV0: |
|
595 | 588 | if flags: |
|
596 | 589 | raise error.RevlogError( |
|
597 | 590 | _(b'unknown flags (%#04x) in version %d revlog %s') |
|
598 | 591 | % (flags >> 16, fmt, self.indexfile) |
|
599 | 592 | ) |
|
600 | 593 | |
|
601 | 594 | self._inline = False |
|
602 | 595 | self._generaldelta = False |
|
603 | 596 | |
|
604 | 597 | elif fmt == REVLOGV1: |
|
605 | 598 | if flags & ~REVLOGV1_FLAGS: |
|
606 | 599 | raise error.RevlogError( |
|
607 | 600 | _(b'unknown flags (%#04x) in version %d revlog %s') |
|
608 | 601 | % (flags >> 16, fmt, self.indexfile) |
|
609 | 602 | ) |
|
610 | 603 | |
|
611 | 604 | self._inline = versionflags & FLAG_INLINE_DATA |
|
612 | 605 | self._generaldelta = versionflags & FLAG_GENERALDELTA |
|
613 | 606 | |
|
614 | 607 | elif fmt == REVLOGV2: |
|
615 | 608 | if flags & ~REVLOGV2_FLAGS: |
|
616 | 609 | raise error.RevlogError( |
|
617 | 610 | _(b'unknown flags (%#04x) in version %d revlog %s') |
|
618 | 611 | % (flags >> 16, fmt, self.indexfile) |
|
619 | 612 | ) |
|
620 | 613 | |
|
621 | 614 | # There is a bug in the transaction handling when going from an |
|
622 | 615 | # inline revlog to a separate index and data file. Turn it off until |
|
623 | 616 | # it's fixed, since v2 revlogs sometimes get rewritten on exchange. |
|
624 | 617 | # See issue6485 |
|
625 | 618 | self._inline = False |
|
626 | 619 | # generaldelta implied by version 2 revlogs. |
|
627 | 620 | self._generaldelta = True |
|
628 | 621 | |
|
629 | 622 | else: |
|
630 | 623 | raise error.RevlogError( |
|
631 | 624 | _(b'unknown version (%d) in revlog %s') % (fmt, self.indexfile) |
|
632 | 625 | ) |
|
633 | 626 | |
|
634 | 627 | self.nodeconstants = sha1nodeconstants |
|
635 | 628 | self.nullid = self.nodeconstants.nullid |
|
636 | 629 | |
|
637 | 630 | # sparse-revlog can't be on without general-delta (issue6056) |
|
638 | 631 | if not self._generaldelta: |
|
639 | 632 | self._sparserevlog = False |
|
640 | 633 | |
|
641 | 634 | self._storedeltachains = True |
|
642 | 635 | |
|
643 | 636 | devel_nodemap = ( |
|
644 | 637 | self.nodemap_file |
|
645 | 638 | and opts.get(b'devel-force-nodemap', False) |
|
646 | 639 | and NodemapRevlogIO is not None |
|
647 | 640 | ) |
|
648 | 641 | |
|
649 | 642 | use_rust_index = False |
|
650 | 643 | if rustrevlog is not None: |
|
651 | 644 | if self.nodemap_file is not None: |
|
652 | 645 | use_rust_index = True |
|
653 | 646 | else: |
|
654 | 647 | use_rust_index = self.opener.options.get(b'rust.index') |
|
655 | 648 | |
|
656 | 649 | self._io = revlogio() |
|
657 | 650 | if self.version == REVLOGV0: |
|
658 | 651 | self._io = revlogoldio() |
|
659 | 652 | elif fmt == REVLOGV2: |
|
660 | 653 | self._io = revlogv2io() |
|
661 | 654 | elif devel_nodemap: |
|
662 | 655 | self._io = NodemapRevlogIO() |
|
663 | 656 | elif use_rust_index: |
|
664 | 657 | self._io = rustrevlogio() |
|
665 | 658 | try: |
|
666 | 659 | d = self._io.parseindex(indexdata, self._inline) |
|
667 | 660 | index, _chunkcache = d |
|
668 | 661 | use_nodemap = ( |
|
669 | 662 | not self._inline |
|
670 | 663 | and self.nodemap_file is not None |
|
671 | 664 | and util.safehasattr(index, 'update_nodemap_data') |
|
672 | 665 | ) |
|
673 | 666 | if use_nodemap: |
|
674 | 667 | nodemap_data = nodemaputil.persisted_data(self) |
|
675 | 668 | if nodemap_data is not None: |
|
676 | 669 | docket = nodemap_data[0] |
|
677 | 670 | if ( |
|
678 | 671 | len(d[0]) > docket.tip_rev |
|
679 | 672 | and d[0][docket.tip_rev][7] == docket.tip_node |
|
680 | 673 | ): |
|
681 | 674 | # no changelog tampering |
|
682 | 675 | self._nodemap_docket = docket |
|
683 | 676 | index.update_nodemap_data(*nodemap_data) |
|
684 | 677 | except (ValueError, IndexError): |
|
685 | 678 | raise error.RevlogError( |
|
686 | 679 | _(b"index %s is corrupted") % self.indexfile |
|
687 | 680 | ) |
|
688 | 681 | self.index, self._chunkcache = d |
|
689 | 682 | if not self._chunkcache: |
|
690 | 683 | self._chunkclear() |
|
691 | 684 | # revnum -> (chain-length, sum-delta-length) |
|
692 | 685 | self._chaininfocache = util.lrucachedict(500) |
|
693 | 686 | # revlog header -> revlog compressor |
|
694 | 687 | self._decompressors = {} |
|
695 | 688 | |
|
696 | 689 | @util.propertycache |
|
697 | 690 | def _compressor(self): |
|
698 | 691 | engine = util.compengines[self._compengine] |
|
699 | 692 | return engine.revlogcompressor(self._compengineopts) |
|
700 | 693 | |
|
701 | 694 | def _indexfp(self, mode=b'r'): |
|
702 | 695 | """file object for the revlog's index file""" |
|
703 | 696 | args = {'mode': mode} |
|
704 | 697 | if mode != b'r': |
|
705 | 698 | args['checkambig'] = self._checkambig |
|
706 | 699 | if mode == b'w': |
|
707 | 700 | args['atomictemp'] = True |
|
708 | 701 | return self.opener(self.indexfile, **args) |
|
709 | 702 | |
|
710 | 703 | def _datafp(self, mode=b'r'): |
|
711 | 704 | """file object for the revlog's data file""" |
|
712 | 705 | return self.opener(self.datafile, mode=mode) |
|
713 | 706 | |
|
714 | 707 | @contextlib.contextmanager |
|
715 | 708 | def _datareadfp(self, existingfp=None): |
|
716 | 709 | """file object suitable to read data""" |
|
717 | 710 | # Use explicit file handle, if given. |
|
718 | 711 | if existingfp is not None: |
|
719 | 712 | yield existingfp |
|
720 | 713 | |
|
721 | 714 | # Use a file handle being actively used for writes, if available. |
|
722 | 715 | # There is some danger to doing this because reads will seek the |
|
723 | 716 | # file. However, _writeentry() performs a SEEK_END before all writes, |
|
724 | 717 | # so we should be safe. |
|
725 | 718 | elif self._writinghandles: |
|
726 | 719 | if self._inline: |
|
727 | 720 | yield self._writinghandles[0] |
|
728 | 721 | else: |
|
729 | 722 | yield self._writinghandles[1] |
|
730 | 723 | |
|
731 | 724 | # Otherwise open a new file handle. |
|
732 | 725 | else: |
|
733 | 726 | if self._inline: |
|
734 | 727 | func = self._indexfp |
|
735 | 728 | else: |
|
736 | 729 | func = self._datafp |
|
737 | 730 | with func() as fp: |
|
738 | 731 | yield fp |
|
739 | 732 | |
|
740 | 733 | def tiprev(self): |
|
741 | 734 | return len(self.index) - 1 |
|
742 | 735 | |
|
743 | 736 | def tip(self): |
|
744 | 737 | return self.node(self.tiprev()) |
|
745 | 738 | |
|
746 | 739 | def __contains__(self, rev): |
|
747 | 740 | return 0 <= rev < len(self) |
|
748 | 741 | |
|
749 | 742 | def __len__(self): |
|
750 | 743 | return len(self.index) |
|
751 | 744 | |
|
752 | 745 | def __iter__(self): |
|
753 | 746 | return iter(pycompat.xrange(len(self))) |
|
754 | 747 | |
|
755 | 748 | def revs(self, start=0, stop=None): |
|
756 | 749 | """iterate over all rev in this revlog (from start to stop)""" |
|
757 | 750 | return storageutil.iterrevs(len(self), start=start, stop=stop) |
|
758 | 751 | |
|
759 | 752 | @property |
|
760 | 753 | def nodemap(self): |
|
761 | 754 | msg = ( |
|
762 | 755 | b"revlog.nodemap is deprecated, " |
|
763 | 756 | b"use revlog.index.[has_node|rev|get_rev]" |
|
764 | 757 | ) |
|
765 | 758 | util.nouideprecwarn(msg, b'5.3', stacklevel=2) |
|
766 | 759 | return self.index.nodemap |
|
767 | 760 | |
|
768 | 761 | @property |
|
769 | 762 | def _nodecache(self): |
|
770 | 763 | msg = b"revlog._nodecache is deprecated, use revlog.index.nodemap" |
|
771 | 764 | util.nouideprecwarn(msg, b'5.3', stacklevel=2) |
|
772 | 765 | return self.index.nodemap |
|
773 | 766 | |
|
774 | 767 | def hasnode(self, node): |
|
775 | 768 | try: |
|
776 | 769 | self.rev(node) |
|
777 | 770 | return True |
|
778 | 771 | except KeyError: |
|
779 | 772 | return False |
|
780 | 773 | |
|
781 | 774 | def candelta(self, baserev, rev): |
|
782 | 775 | """whether two revisions (baserev, rev) can be delta-ed or not""" |
|
783 | 776 | # Disable delta if either rev requires a content-changing flag |
|
784 | 777 | # processor (ex. LFS). This is because such flag processor can alter |
|
785 | 778 | # the rawtext content that the delta will be based on, and two clients |
|
786 | 779 | # could have a same revlog node with different flags (i.e. different |
|
787 | 780 | # rawtext contents) and the delta could be incompatible. |
|
788 | 781 | if (self.flags(baserev) & REVIDX_RAWTEXT_CHANGING_FLAGS) or ( |
|
789 | 782 | self.flags(rev) & REVIDX_RAWTEXT_CHANGING_FLAGS |
|
790 | 783 | ): |
|
791 | 784 | return False |
|
792 | 785 | return True |
|
793 | 786 | |
|
794 | 787 | def update_caches(self, transaction): |
|
795 | 788 | if self.nodemap_file is not None: |
|
796 | 789 | if transaction is None: |
|
797 | 790 | nodemaputil.update_persistent_nodemap(self) |
|
798 | 791 | else: |
|
799 | 792 | nodemaputil.setup_persistent_nodemap(transaction, self) |
|
800 | 793 | |
|
801 | 794 | def clearcaches(self): |
|
802 | 795 | self._revisioncache = None |
|
803 | 796 | self._chainbasecache.clear() |
|
804 | 797 | self._chunkcache = (0, b'') |
|
805 | 798 | self._pcache = {} |
|
806 | 799 | self._nodemap_docket = None |
|
807 | 800 | self.index.clearcaches() |
|
808 | 801 | # The python code is the one responsible for validating the docket, we |
|
809 | 802 | # end up having to refresh it here. |
|
810 | 803 | use_nodemap = ( |
|
811 | 804 | not self._inline |
|
812 | 805 | and self.nodemap_file is not None |
|
813 | 806 | and util.safehasattr(self.index, 'update_nodemap_data') |
|
814 | 807 | ) |
|
815 | 808 | if use_nodemap: |
|
816 | 809 | nodemap_data = nodemaputil.persisted_data(self) |
|
817 | 810 | if nodemap_data is not None: |
|
818 | 811 | self._nodemap_docket = nodemap_data[0] |
|
819 | 812 | self.index.update_nodemap_data(*nodemap_data) |
|
820 | 813 | |
|
821 | 814 | def rev(self, node): |
|
822 | 815 | try: |
|
823 | 816 | return self.index.rev(node) |
|
824 | 817 | except TypeError: |
|
825 | 818 | raise |
|
826 | 819 | except error.RevlogError: |
|
827 | 820 | # parsers.c radix tree lookup failed |
|
828 | 821 | if node == wdirid or node in wdirfilenodeids: |
|
829 | 822 | raise error.WdirUnsupported |
|
830 | 823 | raise error.LookupError(node, self.indexfile, _(b'no node')) |
|
831 | 824 | |
|
832 | 825 | # Accessors for index entries. |
|
833 | 826 | |
|
834 | 827 | # First tuple entry is 8 bytes. First 6 bytes are offset. Last 2 bytes |
|
835 | 828 | # are flags. |
|
836 | 829 | def start(self, rev): |
|
837 | 830 | return int(self.index[rev][0] >> 16) |
|
838 | 831 | |
|
839 | 832 | def flags(self, rev): |
|
840 | 833 | return self.index[rev][0] & 0xFFFF |
|
841 | 834 | |
|
842 | 835 | def length(self, rev): |
|
843 | 836 | return self.index[rev][1] |
|
844 | 837 | |
|
845 | 838 | def sidedata_length(self, rev): |
|
846 | 839 | if self.version & 0xFFFF != REVLOGV2: |
|
847 | 840 | return 0 |
|
848 | 841 | return self.index[rev][9] |
|
849 | 842 | |
|
850 | 843 | def rawsize(self, rev): |
|
851 | 844 | """return the length of the uncompressed text for a given revision""" |
|
852 | 845 | l = self.index[rev][2] |
|
853 | 846 | if l >= 0: |
|
854 | 847 | return l |
|
855 | 848 | |
|
856 | 849 | t = self.rawdata(rev) |
|
857 | 850 | return len(t) |
|
858 | 851 | |
|
859 | 852 | def size(self, rev): |
|
860 | 853 | """length of non-raw text (processed by a "read" flag processor)""" |
|
861 | 854 | # fast path: if no "read" flag processor could change the content, |
|
862 | 855 | # size is rawsize. note: ELLIPSIS is known to not change the content. |
|
863 | 856 | flags = self.flags(rev) |
|
864 | 857 | if flags & (flagutil.REVIDX_KNOWN_FLAGS ^ REVIDX_ELLIPSIS) == 0: |
|
865 | 858 | return self.rawsize(rev) |
|
866 | 859 | |
|
867 | 860 | return len(self.revision(rev, raw=False)) |
|
868 | 861 | |
|
869 | 862 | def chainbase(self, rev): |
|
870 | 863 | base = self._chainbasecache.get(rev) |
|
871 | 864 | if base is not None: |
|
872 | 865 | return base |
|
873 | 866 | |
|
874 | 867 | index = self.index |
|
875 | 868 | iterrev = rev |
|
876 | 869 | base = index[iterrev][3] |
|
877 | 870 | while base != iterrev: |
|
878 | 871 | iterrev = base |
|
879 | 872 | base = index[iterrev][3] |
|
880 | 873 | |
|
881 | 874 | self._chainbasecache[rev] = base |
|
882 | 875 | return base |
|
883 | 876 | |
|
884 | 877 | def linkrev(self, rev): |
|
885 | 878 | return self.index[rev][4] |
|
886 | 879 | |
|
887 | 880 | def parentrevs(self, rev): |
|
888 | 881 | try: |
|
889 | 882 | entry = self.index[rev] |
|
890 | 883 | except IndexError: |
|
891 | 884 | if rev == wdirrev: |
|
892 | 885 | raise error.WdirUnsupported |
|
893 | 886 | raise |
|
894 | 887 | if entry[5] == nullrev: |
|
895 | 888 | return entry[6], entry[5] |
|
896 | 889 | else: |
|
897 | 890 | return entry[5], entry[6] |
|
898 | 891 | |
|
899 | 892 | # fast parentrevs(rev) where rev isn't filtered |
|
900 | 893 | _uncheckedparentrevs = parentrevs |
|
901 | 894 | |
|
902 | 895 | def node(self, rev): |
|
903 | 896 | try: |
|
904 | 897 | return self.index[rev][7] |
|
905 | 898 | except IndexError: |
|
906 | 899 | if rev == wdirrev: |
|
907 | 900 | raise error.WdirUnsupported |
|
908 | 901 | raise |
|
909 | 902 | |
|
910 | 903 | # Derived from index values. |
|
911 | 904 | |
|
912 | 905 | def end(self, rev): |
|
913 | 906 | return self.start(rev) + self.length(rev) |
|
914 | 907 | |
|
915 | 908 | def parents(self, node): |
|
916 | 909 | i = self.index |
|
917 | 910 | d = i[self.rev(node)] |
|
918 | 911 | # inline node() to avoid function call overhead |
|
919 | 912 | if d[5] == nullid: |
|
920 | 913 | return i[d[6]][7], i[d[5]][7] |
|
921 | 914 | else: |
|
922 | 915 | return i[d[5]][7], i[d[6]][7] |
|
923 | 916 | |
|
924 | 917 | def chainlen(self, rev): |
|
925 | 918 | return self._chaininfo(rev)[0] |
|
926 | 919 | |
|
927 | 920 | def _chaininfo(self, rev): |
|
928 | 921 | chaininfocache = self._chaininfocache |
|
929 | 922 | if rev in chaininfocache: |
|
930 | 923 | return chaininfocache[rev] |
|
931 | 924 | index = self.index |
|
932 | 925 | generaldelta = self._generaldelta |
|
933 | 926 | iterrev = rev |
|
934 | 927 | e = index[iterrev] |
|
935 | 928 | clen = 0 |
|
936 | 929 | compresseddeltalen = 0 |
|
937 | 930 | while iterrev != e[3]: |
|
938 | 931 | clen += 1 |
|
939 | 932 | compresseddeltalen += e[1] |
|
940 | 933 | if generaldelta: |
|
941 | 934 | iterrev = e[3] |
|
942 | 935 | else: |
|
943 | 936 | iterrev -= 1 |
|
944 | 937 | if iterrev in chaininfocache: |
|
945 | 938 | t = chaininfocache[iterrev] |
|
946 | 939 | clen += t[0] |
|
947 | 940 | compresseddeltalen += t[1] |
|
948 | 941 | break |
|
949 | 942 | e = index[iterrev] |
|
950 | 943 | else: |
|
951 | 944 | # Add text length of base since decompressing that also takes |
|
952 | 945 | # work. For cache hits the length is already included. |
|
953 | 946 | compresseddeltalen += e[1] |
|
954 | 947 | r = (clen, compresseddeltalen) |
|
955 | 948 | chaininfocache[rev] = r |
|
956 | 949 | return r |
|
957 | 950 | |
|
958 | 951 | def _deltachain(self, rev, stoprev=None): |
|
959 | 952 | """Obtain the delta chain for a revision. |
|
960 | 953 | |
|
961 | 954 | ``stoprev`` specifies a revision to stop at. If not specified, we |
|
962 | 955 | stop at the base of the chain. |
|
963 | 956 | |
|
964 | 957 | Returns a 2-tuple of (chain, stopped) where ``chain`` is a list of |
|
965 | 958 | revs in ascending order and ``stopped`` is a bool indicating whether |
|
966 | 959 | ``stoprev`` was hit. |
|
967 | 960 | """ |
|
968 | 961 | # Try C implementation. |
|
969 | 962 | try: |
|
970 | 963 | return self.index.deltachain(rev, stoprev, self._generaldelta) |
|
971 | 964 | except AttributeError: |
|
972 | 965 | pass |
|
973 | 966 | |
|
974 | 967 | chain = [] |
|
975 | 968 | |
|
976 | 969 | # Alias to prevent attribute lookup in tight loop. |
|
977 | 970 | index = self.index |
|
978 | 971 | generaldelta = self._generaldelta |
|
979 | 972 | |
|
980 | 973 | iterrev = rev |
|
981 | 974 | e = index[iterrev] |
|
982 | 975 | while iterrev != e[3] and iterrev != stoprev: |
|
983 | 976 | chain.append(iterrev) |
|
984 | 977 | if generaldelta: |
|
985 | 978 | iterrev = e[3] |
|
986 | 979 | else: |
|
987 | 980 | iterrev -= 1 |
|
988 | 981 | e = index[iterrev] |
|
989 | 982 | |
|
990 | 983 | if iterrev == stoprev: |
|
991 | 984 | stopped = True |
|
992 | 985 | else: |
|
993 | 986 | chain.append(iterrev) |
|
994 | 987 | stopped = False |
|
995 | 988 | |
|
996 | 989 | chain.reverse() |
|
997 | 990 | return chain, stopped |
|
998 | 991 | |
|
999 | 992 | def ancestors(self, revs, stoprev=0, inclusive=False): |
|
1000 | 993 | """Generate the ancestors of 'revs' in reverse revision order. |
|
1001 | 994 | Does not generate revs lower than stoprev. |
|
1002 | 995 | |
|
1003 | 996 | See the documentation for ancestor.lazyancestors for more details.""" |
|
1004 | 997 | |
|
1005 | 998 | # first, make sure start revisions aren't filtered |
|
1006 | 999 | revs = list(revs) |
|
1007 | 1000 | checkrev = self.node |
|
1008 | 1001 | for r in revs: |
|
1009 | 1002 | checkrev(r) |
|
1010 | 1003 | # and we're sure ancestors aren't filtered as well |
|
1011 | 1004 | |
|
1012 | 1005 | if rustancestor is not None: |
|
1013 | 1006 | lazyancestors = rustancestor.LazyAncestors |
|
1014 | 1007 | arg = self.index |
|
1015 | 1008 | else: |
|
1016 | 1009 | lazyancestors = ancestor.lazyancestors |
|
1017 | 1010 | arg = self._uncheckedparentrevs |
|
1018 | 1011 | return lazyancestors(arg, revs, stoprev=stoprev, inclusive=inclusive) |
|
1019 | 1012 | |
|
1020 | 1013 | def descendants(self, revs): |
|
1021 | 1014 | return dagop.descendantrevs(revs, self.revs, self.parentrevs) |
|
1022 | 1015 | |
|
1023 | 1016 | def findcommonmissing(self, common=None, heads=None): |
|
1024 | 1017 | """Return a tuple of the ancestors of common and the ancestors of heads |
|
1025 | 1018 | that are not ancestors of common. In revset terminology, we return the |
|
1026 | 1019 | tuple: |
|
1027 | 1020 | |
|
1028 | 1021 | ::common, (::heads) - (::common) |
|
1029 | 1022 | |
|
1030 | 1023 | The list is sorted by revision number, meaning it is |
|
1031 | 1024 | topologically sorted. |
|
1032 | 1025 | |
|
1033 | 1026 | 'heads' and 'common' are both lists of node IDs. If heads is |
|
1034 | 1027 | not supplied, uses all of the revlog's heads. If common is not |
|
1035 | 1028 | supplied, uses nullid.""" |
|
1036 | 1029 | if common is None: |
|
1037 | 1030 | common = [nullid] |
|
1038 | 1031 | if heads is None: |
|
1039 | 1032 | heads = self.heads() |
|
1040 | 1033 | |
|
1041 | 1034 | common = [self.rev(n) for n in common] |
|
1042 | 1035 | heads = [self.rev(n) for n in heads] |
|
1043 | 1036 | |
|
1044 | 1037 | # we want the ancestors, but inclusive |
|
1045 | 1038 | class lazyset(object): |
|
1046 | 1039 | def __init__(self, lazyvalues): |
|
1047 | 1040 | self.addedvalues = set() |
|
1048 | 1041 | self.lazyvalues = lazyvalues |
|
1049 | 1042 | |
|
1050 | 1043 | def __contains__(self, value): |
|
1051 | 1044 | return value in self.addedvalues or value in self.lazyvalues |
|
1052 | 1045 | |
|
1053 | 1046 | def __iter__(self): |
|
1054 | 1047 | added = self.addedvalues |
|
1055 | 1048 | for r in added: |
|
1056 | 1049 | yield r |
|
1057 | 1050 | for r in self.lazyvalues: |
|
1058 | 1051 | if not r in added: |
|
1059 | 1052 | yield r |
|
1060 | 1053 | |
|
1061 | 1054 | def add(self, value): |
|
1062 | 1055 | self.addedvalues.add(value) |
|
1063 | 1056 | |
|
1064 | 1057 | def update(self, values): |
|
1065 | 1058 | self.addedvalues.update(values) |
|
1066 | 1059 | |
|
1067 | 1060 | has = lazyset(self.ancestors(common)) |
|
1068 | 1061 | has.add(nullrev) |
|
1069 | 1062 | has.update(common) |
|
1070 | 1063 | |
|
1071 | 1064 | # take all ancestors from heads that aren't in has |
|
1072 | 1065 | missing = set() |
|
1073 | 1066 | visit = collections.deque(r for r in heads if r not in has) |
|
1074 | 1067 | while visit: |
|
1075 | 1068 | r = visit.popleft() |
|
1076 | 1069 | if r in missing: |
|
1077 | 1070 | continue |
|
1078 | 1071 | else: |
|
1079 | 1072 | missing.add(r) |
|
1080 | 1073 | for p in self.parentrevs(r): |
|
1081 | 1074 | if p not in has: |
|
1082 | 1075 | visit.append(p) |
|
1083 | 1076 | missing = list(missing) |
|
1084 | 1077 | missing.sort() |
|
1085 | 1078 | return has, [self.node(miss) for miss in missing] |
|
1086 | 1079 | |
|
1087 | 1080 | def incrementalmissingrevs(self, common=None): |
|
1088 | 1081 | """Return an object that can be used to incrementally compute the |
|
1089 | 1082 | revision numbers of the ancestors of arbitrary sets that are not |
|
1090 | 1083 | ancestors of common. This is an ancestor.incrementalmissingancestors |
|
1091 | 1084 | object. |
|
1092 | 1085 | |
|
1093 | 1086 | 'common' is a list of revision numbers. If common is not supplied, uses |
|
1094 | 1087 | nullrev. |
|
1095 | 1088 | """ |
|
1096 | 1089 | if common is None: |
|
1097 | 1090 | common = [nullrev] |
|
1098 | 1091 | |
|
1099 | 1092 | if rustancestor is not None: |
|
1100 | 1093 | return rustancestor.MissingAncestors(self.index, common) |
|
1101 | 1094 | return ancestor.incrementalmissingancestors(self.parentrevs, common) |
|
1102 | 1095 | |
|
1103 | 1096 | def findmissingrevs(self, common=None, heads=None): |
|
1104 | 1097 | """Return the revision numbers of the ancestors of heads that |
|
1105 | 1098 | are not ancestors of common. |
|
1106 | 1099 | |
|
1107 | 1100 | More specifically, return a list of revision numbers corresponding to |
|
1108 | 1101 | nodes N such that every N satisfies the following constraints: |
|
1109 | 1102 | |
|
1110 | 1103 | 1. N is an ancestor of some node in 'heads' |
|
1111 | 1104 | 2. N is not an ancestor of any node in 'common' |
|
1112 | 1105 | |
|
1113 | 1106 | The list is sorted by revision number, meaning it is |
|
1114 | 1107 | topologically sorted. |
|
1115 | 1108 | |
|
1116 | 1109 | 'heads' and 'common' are both lists of revision numbers. If heads is |
|
1117 | 1110 | not supplied, uses all of the revlog's heads. If common is not |
|
1118 | 1111 | supplied, uses nullid.""" |
|
1119 | 1112 | if common is None: |
|
1120 | 1113 | common = [nullrev] |
|
1121 | 1114 | if heads is None: |
|
1122 | 1115 | heads = self.headrevs() |
|
1123 | 1116 | |
|
1124 | 1117 | inc = self.incrementalmissingrevs(common=common) |
|
1125 | 1118 | return inc.missingancestors(heads) |
|
1126 | 1119 | |
|
1127 | 1120 | def findmissing(self, common=None, heads=None): |
|
1128 | 1121 | """Return the ancestors of heads that are not ancestors of common. |
|
1129 | 1122 | |
|
1130 | 1123 | More specifically, return a list of nodes N such that every N |
|
1131 | 1124 | satisfies the following constraints: |
|
1132 | 1125 | |
|
1133 | 1126 | 1. N is an ancestor of some node in 'heads' |
|
1134 | 1127 | 2. N is not an ancestor of any node in 'common' |
|
1135 | 1128 | |
|
1136 | 1129 | The list is sorted by revision number, meaning it is |
|
1137 | 1130 | topologically sorted. |
|
1138 | 1131 | |
|
1139 | 1132 | 'heads' and 'common' are both lists of node IDs. If heads is |
|
1140 | 1133 | not supplied, uses all of the revlog's heads. If common is not |
|
1141 | 1134 | supplied, uses nullid.""" |
|
1142 | 1135 | if common is None: |
|
1143 | 1136 | common = [nullid] |
|
1144 | 1137 | if heads is None: |
|
1145 | 1138 | heads = self.heads() |
|
1146 | 1139 | |
|
1147 | 1140 | common = [self.rev(n) for n in common] |
|
1148 | 1141 | heads = [self.rev(n) for n in heads] |
|
1149 | 1142 | |
|
1150 | 1143 | inc = self.incrementalmissingrevs(common=common) |
|
1151 | 1144 | return [self.node(r) for r in inc.missingancestors(heads)] |
|
1152 | 1145 | |
|
1153 | 1146 | def nodesbetween(self, roots=None, heads=None): |
|
1154 | 1147 | """Return a topological path from 'roots' to 'heads'. |
|
1155 | 1148 | |
|
1156 | 1149 | Return a tuple (nodes, outroots, outheads) where 'nodes' is a |
|
1157 | 1150 | topologically sorted list of all nodes N that satisfy both of |
|
1158 | 1151 | these constraints: |
|
1159 | 1152 | |
|
1160 | 1153 | 1. N is a descendant of some node in 'roots' |
|
1161 | 1154 | 2. N is an ancestor of some node in 'heads' |
|
1162 | 1155 | |
|
1163 | 1156 | Every node is considered to be both a descendant and an ancestor |
|
1164 | 1157 | of itself, so every reachable node in 'roots' and 'heads' will be |
|
1165 | 1158 | included in 'nodes'. |
|
1166 | 1159 | |
|
1167 | 1160 | 'outroots' is the list of reachable nodes in 'roots', i.e., the |
|
1168 | 1161 | subset of 'roots' that is returned in 'nodes'. Likewise, |
|
1169 | 1162 | 'outheads' is the subset of 'heads' that is also in 'nodes'. |
|
1170 | 1163 | |
|
1171 | 1164 | 'roots' and 'heads' are both lists of node IDs. If 'roots' is |
|
1172 | 1165 | unspecified, uses nullid as the only root. If 'heads' is |
|
1173 | 1166 | unspecified, uses list of all of the revlog's heads.""" |
|
1174 | 1167 | nonodes = ([], [], []) |
|
1175 | 1168 | if roots is not None: |
|
1176 | 1169 | roots = list(roots) |
|
1177 | 1170 | if not roots: |
|
1178 | 1171 | return nonodes |
|
1179 | 1172 | lowestrev = min([self.rev(n) for n in roots]) |
|
1180 | 1173 | else: |
|
1181 | 1174 | roots = [nullid] # Everybody's a descendant of nullid |
|
1182 | 1175 | lowestrev = nullrev |
|
1183 | 1176 | if (lowestrev == nullrev) and (heads is None): |
|
1184 | 1177 | # We want _all_ the nodes! |
|
1185 | 1178 | return ([self.node(r) for r in self], [nullid], list(self.heads())) |
|
1186 | 1179 | if heads is None: |
|
1187 | 1180 | # All nodes are ancestors, so the latest ancestor is the last |
|
1188 | 1181 | # node. |
|
1189 | 1182 | highestrev = len(self) - 1 |
|
1190 | 1183 | # Set ancestors to None to signal that every node is an ancestor. |
|
1191 | 1184 | ancestors = None |
|
1192 | 1185 | # Set heads to an empty dictionary for later discovery of heads |
|
1193 | 1186 | heads = {} |
|
1194 | 1187 | else: |
|
1195 | 1188 | heads = list(heads) |
|
1196 | 1189 | if not heads: |
|
1197 | 1190 | return nonodes |
|
1198 | 1191 | ancestors = set() |
|
1199 | 1192 | # Turn heads into a dictionary so we can remove 'fake' heads. |
|
1200 | 1193 | # Also, later we will be using it to filter out the heads we can't |
|
1201 | 1194 | # find from roots. |
|
1202 | 1195 | heads = dict.fromkeys(heads, False) |
|
1203 | 1196 | # Start at the top and keep marking parents until we're done. |
|
1204 | 1197 | nodestotag = set(heads) |
|
1205 | 1198 | # Remember where the top was so we can use it as a limit later. |
|
1206 | 1199 | highestrev = max([self.rev(n) for n in nodestotag]) |
|
1207 | 1200 | while nodestotag: |
|
1208 | 1201 | # grab a node to tag |
|
1209 | 1202 | n = nodestotag.pop() |
|
1210 | 1203 | # Never tag nullid |
|
1211 | 1204 | if n == nullid: |
|
1212 | 1205 | continue |
|
1213 | 1206 | # A node's revision number represents its place in a |
|
1214 | 1207 | # topologically sorted list of nodes. |
|
1215 | 1208 | r = self.rev(n) |
|
1216 | 1209 | if r >= lowestrev: |
|
1217 | 1210 | if n not in ancestors: |
|
1218 | 1211 | # If we are possibly a descendant of one of the roots |
|
1219 | 1212 | # and we haven't already been marked as an ancestor |
|
1220 | 1213 | ancestors.add(n) # Mark as ancestor |
|
1221 | 1214 | # Add non-nullid parents to list of nodes to tag. |
|
1222 | 1215 | nodestotag.update( |
|
1223 | 1216 | [p for p in self.parents(n) if p != nullid] |
|
1224 | 1217 | ) |
|
1225 | 1218 | elif n in heads: # We've seen it before, is it a fake head? |
|
1226 | 1219 | # So it is, real heads should not be the ancestors of |
|
1227 | 1220 | # any other heads. |
|
1228 | 1221 | heads.pop(n) |
|
1229 | 1222 | if not ancestors: |
|
1230 | 1223 | return nonodes |
|
1231 | 1224 | # Now that we have our set of ancestors, we want to remove any |
|
1232 | 1225 | # roots that are not ancestors. |
|
1233 | 1226 | |
|
1234 | 1227 | # If one of the roots was nullid, everything is included anyway. |
|
1235 | 1228 | if lowestrev > nullrev: |
|
1236 | 1229 | # But, since we weren't, let's recompute the lowest rev to not |
|
1237 | 1230 | # include roots that aren't ancestors. |
|
1238 | 1231 | |
|
1239 | 1232 | # Filter out roots that aren't ancestors of heads |
|
1240 | 1233 | roots = [root for root in roots if root in ancestors] |
|
1241 | 1234 | # Recompute the lowest revision |
|
1242 | 1235 | if roots: |
|
1243 | 1236 | lowestrev = min([self.rev(root) for root in roots]) |
|
1244 | 1237 | else: |
|
1245 | 1238 | # No more roots? Return empty list |
|
1246 | 1239 | return nonodes |
|
1247 | 1240 | else: |
|
1248 | 1241 | # We are descending from nullid, and don't need to care about |
|
1249 | 1242 | # any other roots. |
|
1250 | 1243 | lowestrev = nullrev |
|
1251 | 1244 | roots = [nullid] |
|
1252 | 1245 | # Transform our roots list into a set. |
|
1253 | 1246 | descendants = set(roots) |
|
1254 | 1247 | # Also, keep the original roots so we can filter out roots that aren't |
|
1255 | 1248 | # 'real' roots (i.e. are descended from other roots). |
|
1256 | 1249 | roots = descendants.copy() |
|
1257 | 1250 | # Our topologically sorted list of output nodes. |
|
1258 | 1251 | orderedout = [] |
|
1259 | 1252 | # Don't start at nullid since we don't want nullid in our output list, |
|
1260 | 1253 | # and if nullid shows up in descendants, empty parents will look like |
|
1261 | 1254 | # they're descendants. |
|
1262 | 1255 | for r in self.revs(start=max(lowestrev, 0), stop=highestrev + 1): |
|
1263 | 1256 | n = self.node(r) |
|
1264 | 1257 | isdescendant = False |
|
1265 | 1258 | if lowestrev == nullrev: # Everybody is a descendant of nullid |
|
1266 | 1259 | isdescendant = True |
|
1267 | 1260 | elif n in descendants: |
|
1268 | 1261 | # n is already a descendant |
|
1269 | 1262 | isdescendant = True |
|
1270 | 1263 | # This check only needs to be done here because all the roots |
|
1271 | 1264 | # will start being marked is descendants before the loop. |
|
1272 | 1265 | if n in roots: |
|
1273 | 1266 | # If n was a root, check if it's a 'real' root. |
|
1274 | 1267 | p = tuple(self.parents(n)) |
|
1275 | 1268 | # If any of its parents are descendants, it's not a root. |
|
1276 | 1269 | if (p[0] in descendants) or (p[1] in descendants): |
|
1277 | 1270 | roots.remove(n) |
|
1278 | 1271 | else: |
|
1279 | 1272 | p = tuple(self.parents(n)) |
|
1280 | 1273 | # A node is a descendant if either of its parents are |
|
1281 | 1274 | # descendants. (We seeded the dependents list with the roots |
|
1282 | 1275 | # up there, remember?) |
|
1283 | 1276 | if (p[0] in descendants) or (p[1] in descendants): |
|
1284 | 1277 | descendants.add(n) |
|
1285 | 1278 | isdescendant = True |
|
1286 | 1279 | if isdescendant and ((ancestors is None) or (n in ancestors)): |
|
1287 | 1280 | # Only include nodes that are both descendants and ancestors. |
|
1288 | 1281 | orderedout.append(n) |
|
1289 | 1282 | if (ancestors is not None) and (n in heads): |
|
1290 | 1283 | # We're trying to figure out which heads are reachable |
|
1291 | 1284 | # from roots. |
|
1292 | 1285 | # Mark this head as having been reached |
|
1293 | 1286 | heads[n] = True |
|
1294 | 1287 | elif ancestors is None: |
|
1295 | 1288 | # Otherwise, we're trying to discover the heads. |
|
1296 | 1289 | # Assume this is a head because if it isn't, the next step |
|
1297 | 1290 | # will eventually remove it. |
|
1298 | 1291 | heads[n] = True |
|
1299 | 1292 | # But, obviously its parents aren't. |
|
1300 | 1293 | for p in self.parents(n): |
|
1301 | 1294 | heads.pop(p, None) |
|
1302 | 1295 | heads = [head for head, flag in pycompat.iteritems(heads) if flag] |
|
1303 | 1296 | roots = list(roots) |
|
1304 | 1297 | assert orderedout |
|
1305 | 1298 | assert roots |
|
1306 | 1299 | assert heads |
|
1307 | 1300 | return (orderedout, roots, heads) |
|
1308 | 1301 | |
|
1309 | 1302 | def headrevs(self, revs=None): |
|
1310 | 1303 | if revs is None: |
|
1311 | 1304 | try: |
|
1312 | 1305 | return self.index.headrevs() |
|
1313 | 1306 | except AttributeError: |
|
1314 | 1307 | return self._headrevs() |
|
1315 | 1308 | if rustdagop is not None: |
|
1316 | 1309 | return rustdagop.headrevs(self.index, revs) |
|
1317 | 1310 | return dagop.headrevs(revs, self._uncheckedparentrevs) |
|
1318 | 1311 | |
|
1319 | 1312 | def computephases(self, roots): |
|
1320 | 1313 | return self.index.computephasesmapsets(roots) |
|
1321 | 1314 | |
|
1322 | 1315 | def _headrevs(self): |
|
1323 | 1316 | count = len(self) |
|
1324 | 1317 | if not count: |
|
1325 | 1318 | return [nullrev] |
|
1326 | 1319 | # we won't iter over filtered rev so nobody is a head at start |
|
1327 | 1320 | ishead = [0] * (count + 1) |
|
1328 | 1321 | index = self.index |
|
1329 | 1322 | for r in self: |
|
1330 | 1323 | ishead[r] = 1 # I may be an head |
|
1331 | 1324 | e = index[r] |
|
1332 | 1325 | ishead[e[5]] = ishead[e[6]] = 0 # my parent are not |
|
1333 | 1326 | return [r for r, val in enumerate(ishead) if val] |
|
1334 | 1327 | |
|
1335 | 1328 | def heads(self, start=None, stop=None): |
|
1336 | 1329 | """return the list of all nodes that have no children |
|
1337 | 1330 | |
|
1338 | 1331 | if start is specified, only heads that are descendants of |
|
1339 | 1332 | start will be returned |
|
1340 | 1333 | if stop is specified, it will consider all the revs from stop |
|
1341 | 1334 | as if they had no children |
|
1342 | 1335 | """ |
|
1343 | 1336 | if start is None and stop is None: |
|
1344 | 1337 | if not len(self): |
|
1345 | 1338 | return [nullid] |
|
1346 | 1339 | return [self.node(r) for r in self.headrevs()] |
|
1347 | 1340 | |
|
1348 | 1341 | if start is None: |
|
1349 | 1342 | start = nullrev |
|
1350 | 1343 | else: |
|
1351 | 1344 | start = self.rev(start) |
|
1352 | 1345 | |
|
1353 | 1346 | stoprevs = {self.rev(n) for n in stop or []} |
|
1354 | 1347 | |
|
1355 | 1348 | revs = dagop.headrevssubset( |
|
1356 | 1349 | self.revs, self.parentrevs, startrev=start, stoprevs=stoprevs |
|
1357 | 1350 | ) |
|
1358 | 1351 | |
|
1359 | 1352 | return [self.node(rev) for rev in revs] |
|
1360 | 1353 | |
|
1361 | 1354 | def children(self, node): |
|
1362 | 1355 | """find the children of a given node""" |
|
1363 | 1356 | c = [] |
|
1364 | 1357 | p = self.rev(node) |
|
1365 | 1358 | for r in self.revs(start=p + 1): |
|
1366 | 1359 | prevs = [pr for pr in self.parentrevs(r) if pr != nullrev] |
|
1367 | 1360 | if prevs: |
|
1368 | 1361 | for pr in prevs: |
|
1369 | 1362 | if pr == p: |
|
1370 | 1363 | c.append(self.node(r)) |
|
1371 | 1364 | elif p == nullrev: |
|
1372 | 1365 | c.append(self.node(r)) |
|
1373 | 1366 | return c |
|
1374 | 1367 | |
|
1375 | 1368 | def commonancestorsheads(self, a, b): |
|
1376 | 1369 | """calculate all the heads of the common ancestors of nodes a and b""" |
|
1377 | 1370 | a, b = self.rev(a), self.rev(b) |
|
1378 | 1371 | ancs = self._commonancestorsheads(a, b) |
|
1379 | 1372 | return pycompat.maplist(self.node, ancs) |
|
1380 | 1373 | |
|
1381 | 1374 | def _commonancestorsheads(self, *revs): |
|
1382 | 1375 | """calculate all the heads of the common ancestors of revs""" |
|
1383 | 1376 | try: |
|
1384 | 1377 | ancs = self.index.commonancestorsheads(*revs) |
|
1385 | 1378 | except (AttributeError, OverflowError): # C implementation failed |
|
1386 | 1379 | ancs = ancestor.commonancestorsheads(self.parentrevs, *revs) |
|
1387 | 1380 | return ancs |
|
1388 | 1381 | |
|
1389 | 1382 | def isancestor(self, a, b): |
|
1390 | 1383 | """return True if node a is an ancestor of node b |
|
1391 | 1384 | |
|
1392 | 1385 | A revision is considered an ancestor of itself.""" |
|
1393 | 1386 | a, b = self.rev(a), self.rev(b) |
|
1394 | 1387 | return self.isancestorrev(a, b) |
|
1395 | 1388 | |
|
1396 | 1389 | def isancestorrev(self, a, b): |
|
1397 | 1390 | """return True if revision a is an ancestor of revision b |
|
1398 | 1391 | |
|
1399 | 1392 | A revision is considered an ancestor of itself. |
|
1400 | 1393 | |
|
1401 | 1394 | The implementation of this is trivial but the use of |
|
1402 | 1395 | reachableroots is not.""" |
|
1403 | 1396 | if a == nullrev: |
|
1404 | 1397 | return True |
|
1405 | 1398 | elif a == b: |
|
1406 | 1399 | return True |
|
1407 | 1400 | elif a > b: |
|
1408 | 1401 | return False |
|
1409 | 1402 | return bool(self.reachableroots(a, [b], [a], includepath=False)) |
|
1410 | 1403 | |
|
1411 | 1404 | def reachableroots(self, minroot, heads, roots, includepath=False): |
|
1412 | 1405 | """return (heads(::(<roots> and <roots>::<heads>))) |
|
1413 | 1406 | |
|
1414 | 1407 | If includepath is True, return (<roots>::<heads>).""" |
|
1415 | 1408 | try: |
|
1416 | 1409 | return self.index.reachableroots2( |
|
1417 | 1410 | minroot, heads, roots, includepath |
|
1418 | 1411 | ) |
|
1419 | 1412 | except AttributeError: |
|
1420 | 1413 | return dagop._reachablerootspure( |
|
1421 | 1414 | self.parentrevs, minroot, roots, heads, includepath |
|
1422 | 1415 | ) |
|
1423 | 1416 | |
|
1424 | 1417 | def ancestor(self, a, b): |
|
1425 | 1418 | """calculate the "best" common ancestor of nodes a and b""" |
|
1426 | 1419 | |
|
1427 | 1420 | a, b = self.rev(a), self.rev(b) |
|
1428 | 1421 | try: |
|
1429 | 1422 | ancs = self.index.ancestors(a, b) |
|
1430 | 1423 | except (AttributeError, OverflowError): |
|
1431 | 1424 | ancs = ancestor.ancestors(self.parentrevs, a, b) |
|
1432 | 1425 | if ancs: |
|
1433 | 1426 | # choose a consistent winner when there's a tie |
|
1434 | 1427 | return min(map(self.node, ancs)) |
|
1435 | 1428 | return nullid |
|
1436 | 1429 | |
|
1437 | 1430 | def _match(self, id): |
|
1438 | 1431 | if isinstance(id, int): |
|
1439 | 1432 | # rev |
|
1440 | 1433 | return self.node(id) |
|
1441 | 1434 | if len(id) == 20: |
|
1442 | 1435 | # possibly a binary node |
|
1443 | 1436 | # odds of a binary node being all hex in ASCII are 1 in 10**25 |
|
1444 | 1437 | try: |
|
1445 | 1438 | node = id |
|
1446 | 1439 | self.rev(node) # quick search the index |
|
1447 | 1440 | return node |
|
1448 | 1441 | except error.LookupError: |
|
1449 | 1442 | pass # may be partial hex id |
|
1450 | 1443 | try: |
|
1451 | 1444 | # str(rev) |
|
1452 | 1445 | rev = int(id) |
|
1453 | 1446 | if b"%d" % rev != id: |
|
1454 | 1447 | raise ValueError |
|
1455 | 1448 | if rev < 0: |
|
1456 | 1449 | rev = len(self) + rev |
|
1457 | 1450 | if rev < 0 or rev >= len(self): |
|
1458 | 1451 | raise ValueError |
|
1459 | 1452 | return self.node(rev) |
|
1460 | 1453 | except (ValueError, OverflowError): |
|
1461 | 1454 | pass |
|
1462 | 1455 | if len(id) == 40: |
|
1463 | 1456 | try: |
|
1464 | 1457 | # a full hex nodeid? |
|
1465 | 1458 | node = bin(id) |
|
1466 | 1459 | self.rev(node) |
|
1467 | 1460 | return node |
|
1468 | 1461 | except (TypeError, error.LookupError): |
|
1469 | 1462 | pass |
|
1470 | 1463 | |
|
1471 | 1464 | def _partialmatch(self, id): |
|
1472 | 1465 | # we don't care wdirfilenodeids as they should be always full hash |
|
1473 | 1466 | maybewdir = wdirhex.startswith(id) |
|
1474 | 1467 | try: |
|
1475 | 1468 | partial = self.index.partialmatch(id) |
|
1476 | 1469 | if partial and self.hasnode(partial): |
|
1477 | 1470 | if maybewdir: |
|
1478 | 1471 | # single 'ff...' match in radix tree, ambiguous with wdir |
|
1479 | 1472 | raise error.RevlogError |
|
1480 | 1473 | return partial |
|
1481 | 1474 | if maybewdir: |
|
1482 | 1475 | # no 'ff...' match in radix tree, wdir identified |
|
1483 | 1476 | raise error.WdirUnsupported |
|
1484 | 1477 | return None |
|
1485 | 1478 | except error.RevlogError: |
|
1486 | 1479 | # parsers.c radix tree lookup gave multiple matches |
|
1487 | 1480 | # fast path: for unfiltered changelog, radix tree is accurate |
|
1488 | 1481 | if not getattr(self, 'filteredrevs', None): |
|
1489 | 1482 | raise error.AmbiguousPrefixLookupError( |
|
1490 | 1483 | id, self.indexfile, _(b'ambiguous identifier') |
|
1491 | 1484 | ) |
|
1492 | 1485 | # fall through to slow path that filters hidden revisions |
|
1493 | 1486 | except (AttributeError, ValueError): |
|
1494 | 1487 | # we are pure python, or key was too short to search radix tree |
|
1495 | 1488 | pass |
|
1496 | 1489 | |
|
1497 | 1490 | if id in self._pcache: |
|
1498 | 1491 | return self._pcache[id] |
|
1499 | 1492 | |
|
1500 | 1493 | if len(id) <= 40: |
|
1501 | 1494 | try: |
|
1502 | 1495 | # hex(node)[:...] |
|
1503 | 1496 | l = len(id) // 2 # grab an even number of digits |
|
1504 | 1497 | prefix = bin(id[: l * 2]) |
|
1505 | 1498 | nl = [e[7] for e in self.index if e[7].startswith(prefix)] |
|
1506 | 1499 | nl = [ |
|
1507 | 1500 | n for n in nl if hex(n).startswith(id) and self.hasnode(n) |
|
1508 | 1501 | ] |
|
1509 | 1502 | if nullhex.startswith(id): |
|
1510 | 1503 | nl.append(nullid) |
|
1511 | 1504 | if len(nl) > 0: |
|
1512 | 1505 | if len(nl) == 1 and not maybewdir: |
|
1513 | 1506 | self._pcache[id] = nl[0] |
|
1514 | 1507 | return nl[0] |
|
1515 | 1508 | raise error.AmbiguousPrefixLookupError( |
|
1516 | 1509 | id, self.indexfile, _(b'ambiguous identifier') |
|
1517 | 1510 | ) |
|
1518 | 1511 | if maybewdir: |
|
1519 | 1512 | raise error.WdirUnsupported |
|
1520 | 1513 | return None |
|
1521 | 1514 | except TypeError: |
|
1522 | 1515 | pass |
|
1523 | 1516 | |
|
1524 | 1517 | def lookup(self, id): |
|
1525 | 1518 | """locate a node based on: |
|
1526 | 1519 | - revision number or str(revision number) |
|
1527 | 1520 | - nodeid or subset of hex nodeid |
|
1528 | 1521 | """ |
|
1529 | 1522 | n = self._match(id) |
|
1530 | 1523 | if n is not None: |
|
1531 | 1524 | return n |
|
1532 | 1525 | n = self._partialmatch(id) |
|
1533 | 1526 | if n: |
|
1534 | 1527 | return n |
|
1535 | 1528 | |
|
1536 | 1529 | raise error.LookupError(id, self.indexfile, _(b'no match found')) |
|
1537 | 1530 | |
|
1538 | 1531 | def shortest(self, node, minlength=1): |
|
1539 | 1532 | """Find the shortest unambiguous prefix that matches node.""" |
|
1540 | 1533 | |
|
1541 | 1534 | def isvalid(prefix): |
|
1542 | 1535 | try: |
|
1543 | 1536 | matchednode = self._partialmatch(prefix) |
|
1544 | 1537 | except error.AmbiguousPrefixLookupError: |
|
1545 | 1538 | return False |
|
1546 | 1539 | except error.WdirUnsupported: |
|
1547 | 1540 | # single 'ff...' match |
|
1548 | 1541 | return True |
|
1549 | 1542 | if matchednode is None: |
|
1550 | 1543 | raise error.LookupError(node, self.indexfile, _(b'no node')) |
|
1551 | 1544 | return True |
|
1552 | 1545 | |
|
1553 | 1546 | def maybewdir(prefix): |
|
1554 | 1547 | return all(c == b'f' for c in pycompat.iterbytestr(prefix)) |
|
1555 | 1548 | |
|
1556 | 1549 | hexnode = hex(node) |
|
1557 | 1550 | |
|
1558 | 1551 | def disambiguate(hexnode, minlength): |
|
1559 | 1552 | """Disambiguate against wdirid.""" |
|
1560 | 1553 | for length in range(minlength, len(hexnode) + 1): |
|
1561 | 1554 | prefix = hexnode[:length] |
|
1562 | 1555 | if not maybewdir(prefix): |
|
1563 | 1556 | return prefix |
|
1564 | 1557 | |
|
1565 | 1558 | if not getattr(self, 'filteredrevs', None): |
|
1566 | 1559 | try: |
|
1567 | 1560 | length = max(self.index.shortest(node), minlength) |
|
1568 | 1561 | return disambiguate(hexnode, length) |
|
1569 | 1562 | except error.RevlogError: |
|
1570 | 1563 | if node != wdirid: |
|
1571 | 1564 | raise error.LookupError(node, self.indexfile, _(b'no node')) |
|
1572 | 1565 | except AttributeError: |
|
1573 | 1566 | # Fall through to pure code |
|
1574 | 1567 | pass |
|
1575 | 1568 | |
|
1576 | 1569 | if node == wdirid: |
|
1577 | 1570 | for length in range(minlength, len(hexnode) + 1): |
|
1578 | 1571 | prefix = hexnode[:length] |
|
1579 | 1572 | if isvalid(prefix): |
|
1580 | 1573 | return prefix |
|
1581 | 1574 | |
|
1582 | 1575 | for length in range(minlength, len(hexnode) + 1): |
|
1583 | 1576 | prefix = hexnode[:length] |
|
1584 | 1577 | if isvalid(prefix): |
|
1585 | 1578 | return disambiguate(hexnode, length) |
|
1586 | 1579 | |
|
1587 | 1580 | def cmp(self, node, text): |
|
1588 | 1581 | """compare text with a given file revision |
|
1589 | 1582 | |
|
1590 | 1583 | returns True if text is different than what is stored. |
|
1591 | 1584 | """ |
|
1592 | 1585 | p1, p2 = self.parents(node) |
|
1593 | 1586 | return storageutil.hashrevisionsha1(text, p1, p2) != node |
|
1594 | 1587 | |
|
1595 | 1588 | def _cachesegment(self, offset, data): |
|
1596 | 1589 | """Add a segment to the revlog cache. |
|
1597 | 1590 | |
|
1598 | 1591 | Accepts an absolute offset and the data that is at that location. |
|
1599 | 1592 | """ |
|
1600 | 1593 | o, d = self._chunkcache |
|
1601 | 1594 | # try to add to existing cache |
|
1602 | 1595 | if o + len(d) == offset and len(d) + len(data) < _chunksize: |
|
1603 | 1596 | self._chunkcache = o, d + data |
|
1604 | 1597 | else: |
|
1605 | 1598 | self._chunkcache = offset, data |
|
1606 | 1599 | |
|
1607 | 1600 | def _readsegment(self, offset, length, df=None): |
|
1608 | 1601 | """Load a segment of raw data from the revlog. |
|
1609 | 1602 | |
|
1610 | 1603 | Accepts an absolute offset, length to read, and an optional existing |
|
1611 | 1604 | file handle to read from. |
|
1612 | 1605 | |
|
1613 | 1606 | If an existing file handle is passed, it will be seeked and the |
|
1614 | 1607 | original seek position will NOT be restored. |
|
1615 | 1608 | |
|
1616 | 1609 | Returns a str or buffer of raw byte data. |
|
1617 | 1610 | |
|
1618 | 1611 | Raises if the requested number of bytes could not be read. |
|
1619 | 1612 | """ |
|
1620 | 1613 | # Cache data both forward and backward around the requested |
|
1621 | 1614 | # data, in a fixed size window. This helps speed up operations |
|
1622 | 1615 | # involving reading the revlog backwards. |
|
1623 | 1616 | cachesize = self._chunkcachesize |
|
1624 | 1617 | realoffset = offset & ~(cachesize - 1) |
|
1625 | 1618 | reallength = ( |
|
1626 | 1619 | (offset + length + cachesize) & ~(cachesize - 1) |
|
1627 | 1620 | ) - realoffset |
|
1628 | 1621 | with self._datareadfp(df) as df: |
|
1629 | 1622 | df.seek(realoffset) |
|
1630 | 1623 | d = df.read(reallength) |
|
1631 | 1624 | |
|
1632 | 1625 | self._cachesegment(realoffset, d) |
|
1633 | 1626 | if offset != realoffset or reallength != length: |
|
1634 | 1627 | startoffset = offset - realoffset |
|
1635 | 1628 | if len(d) - startoffset < length: |
|
1636 | 1629 | raise error.RevlogError( |
|
1637 | 1630 | _( |
|
1638 | 1631 | b'partial read of revlog %s; expected %d bytes from ' |
|
1639 | 1632 | b'offset %d, got %d' |
|
1640 | 1633 | ) |
|
1641 | 1634 | % ( |
|
1642 | 1635 | self.indexfile if self._inline else self.datafile, |
|
1643 | 1636 | length, |
|
1644 | 1637 | realoffset, |
|
1645 | 1638 | len(d) - startoffset, |
|
1646 | 1639 | ) |
|
1647 | 1640 | ) |
|
1648 | 1641 | |
|
1649 | 1642 | return util.buffer(d, startoffset, length) |
|
1650 | 1643 | |
|
1651 | 1644 | if len(d) < length: |
|
1652 | 1645 | raise error.RevlogError( |
|
1653 | 1646 | _( |
|
1654 | 1647 | b'partial read of revlog %s; expected %d bytes from offset ' |
|
1655 | 1648 | b'%d, got %d' |
|
1656 | 1649 | ) |
|
1657 | 1650 | % ( |
|
1658 | 1651 | self.indexfile if self._inline else self.datafile, |
|
1659 | 1652 | length, |
|
1660 | 1653 | offset, |
|
1661 | 1654 | len(d), |
|
1662 | 1655 | ) |
|
1663 | 1656 | ) |
|
1664 | 1657 | |
|
1665 | 1658 | return d |
|
1666 | 1659 | |
|
1667 | 1660 | def _getsegment(self, offset, length, df=None): |
|
1668 | 1661 | """Obtain a segment of raw data from the revlog. |
|
1669 | 1662 | |
|
1670 | 1663 | Accepts an absolute offset, length of bytes to obtain, and an |
|
1671 | 1664 | optional file handle to the already-opened revlog. If the file |
|
1672 | 1665 | handle is used, it's original seek position will not be preserved. |
|
1673 | 1666 | |
|
1674 | 1667 | Requests for data may be returned from a cache. |
|
1675 | 1668 | |
|
1676 | 1669 | Returns a str or a buffer instance of raw byte data. |
|
1677 | 1670 | """ |
|
1678 | 1671 | o, d = self._chunkcache |
|
1679 | 1672 | l = len(d) |
|
1680 | 1673 | |
|
1681 | 1674 | # is it in the cache? |
|
1682 | 1675 | cachestart = offset - o |
|
1683 | 1676 | cacheend = cachestart + length |
|
1684 | 1677 | if cachestart >= 0 and cacheend <= l: |
|
1685 | 1678 | if cachestart == 0 and cacheend == l: |
|
1686 | 1679 | return d # avoid a copy |
|
1687 | 1680 | return util.buffer(d, cachestart, cacheend - cachestart) |
|
1688 | 1681 | |
|
1689 | 1682 | return self._readsegment(offset, length, df=df) |
|
1690 | 1683 | |
|
1691 | 1684 | def _getsegmentforrevs(self, startrev, endrev, df=None): |
|
1692 | 1685 | """Obtain a segment of raw data corresponding to a range of revisions. |
|
1693 | 1686 | |
|
1694 | 1687 | Accepts the start and end revisions and an optional already-open |
|
1695 | 1688 | file handle to be used for reading. If the file handle is read, its |
|
1696 | 1689 | seek position will not be preserved. |
|
1697 | 1690 | |
|
1698 | 1691 | Requests for data may be satisfied by a cache. |
|
1699 | 1692 | |
|
1700 | 1693 | Returns a 2-tuple of (offset, data) for the requested range of |
|
1701 | 1694 | revisions. Offset is the integer offset from the beginning of the |
|
1702 | 1695 | revlog and data is a str or buffer of the raw byte data. |
|
1703 | 1696 | |
|
1704 | 1697 | Callers will need to call ``self.start(rev)`` and ``self.length(rev)`` |
|
1705 | 1698 | to determine where each revision's data begins and ends. |
|
1706 | 1699 | """ |
|
1707 | 1700 | # Inlined self.start(startrev) & self.end(endrev) for perf reasons |
|
1708 | 1701 | # (functions are expensive). |
|
1709 | 1702 | index = self.index |
|
1710 | 1703 | istart = index[startrev] |
|
1711 | 1704 | start = int(istart[0] >> 16) |
|
1712 | 1705 | if startrev == endrev: |
|
1713 | 1706 | end = start + istart[1] |
|
1714 | 1707 | else: |
|
1715 | 1708 | iend = index[endrev] |
|
1716 | 1709 | end = int(iend[0] >> 16) + iend[1] |
|
1717 | 1710 | |
|
1718 | 1711 | if self._inline: |
|
1719 |
start += (startrev + 1) * self. |
|
|
1720 |
end += (endrev + 1) * self. |
|
|
1712 | start += (startrev + 1) * self.index.entry_size | |
|
1713 | end += (endrev + 1) * self.index.entry_size | |
|
1721 | 1714 | length = end - start |
|
1722 | 1715 | |
|
1723 | 1716 | return start, self._getsegment(start, length, df=df) |
|
1724 | 1717 | |
|
1725 | 1718 | def _chunk(self, rev, df=None): |
|
1726 | 1719 | """Obtain a single decompressed chunk for a revision. |
|
1727 | 1720 | |
|
1728 | 1721 | Accepts an integer revision and an optional already-open file handle |
|
1729 | 1722 | to be used for reading. If used, the seek position of the file will not |
|
1730 | 1723 | be preserved. |
|
1731 | 1724 | |
|
1732 | 1725 | Returns a str holding uncompressed data for the requested revision. |
|
1733 | 1726 | """ |
|
1734 | 1727 | return self.decompress(self._getsegmentforrevs(rev, rev, df=df)[1]) |
|
1735 | 1728 | |
|
1736 | 1729 | def _chunks(self, revs, df=None, targetsize=None): |
|
1737 | 1730 | """Obtain decompressed chunks for the specified revisions. |
|
1738 | 1731 | |
|
1739 | 1732 | Accepts an iterable of numeric revisions that are assumed to be in |
|
1740 | 1733 | ascending order. Also accepts an optional already-open file handle |
|
1741 | 1734 | to be used for reading. If used, the seek position of the file will |
|
1742 | 1735 | not be preserved. |
|
1743 | 1736 | |
|
1744 | 1737 | This function is similar to calling ``self._chunk()`` multiple times, |
|
1745 | 1738 | but is faster. |
|
1746 | 1739 | |
|
1747 | 1740 | Returns a list with decompressed data for each requested revision. |
|
1748 | 1741 | """ |
|
1749 | 1742 | if not revs: |
|
1750 | 1743 | return [] |
|
1751 | 1744 | start = self.start |
|
1752 | 1745 | length = self.length |
|
1753 | 1746 | inline = self._inline |
|
1754 |
iosize = self. |
|
|
1747 | iosize = self.index.entry_size | |
|
1755 | 1748 | buffer = util.buffer |
|
1756 | 1749 | |
|
1757 | 1750 | l = [] |
|
1758 | 1751 | ladd = l.append |
|
1759 | 1752 | |
|
1760 | 1753 | if not self._withsparseread: |
|
1761 | 1754 | slicedchunks = (revs,) |
|
1762 | 1755 | else: |
|
1763 | 1756 | slicedchunks = deltautil.slicechunk( |
|
1764 | 1757 | self, revs, targetsize=targetsize |
|
1765 | 1758 | ) |
|
1766 | 1759 | |
|
1767 | 1760 | for revschunk in slicedchunks: |
|
1768 | 1761 | firstrev = revschunk[0] |
|
1769 | 1762 | # Skip trailing revisions with empty diff |
|
1770 | 1763 | for lastrev in revschunk[::-1]: |
|
1771 | 1764 | if length(lastrev) != 0: |
|
1772 | 1765 | break |
|
1773 | 1766 | |
|
1774 | 1767 | try: |
|
1775 | 1768 | offset, data = self._getsegmentforrevs(firstrev, lastrev, df=df) |
|
1776 | 1769 | except OverflowError: |
|
1777 | 1770 | # issue4215 - we can't cache a run of chunks greater than |
|
1778 | 1771 | # 2G on Windows |
|
1779 | 1772 | return [self._chunk(rev, df=df) for rev in revschunk] |
|
1780 | 1773 | |
|
1781 | 1774 | decomp = self.decompress |
|
1782 | 1775 | for rev in revschunk: |
|
1783 | 1776 | chunkstart = start(rev) |
|
1784 | 1777 | if inline: |
|
1785 | 1778 | chunkstart += (rev + 1) * iosize |
|
1786 | 1779 | chunklength = length(rev) |
|
1787 | 1780 | ladd(decomp(buffer(data, chunkstart - offset, chunklength))) |
|
1788 | 1781 | |
|
1789 | 1782 | return l |
|
1790 | 1783 | |
|
1791 | 1784 | def _chunkclear(self): |
|
1792 | 1785 | """Clear the raw chunk cache.""" |
|
1793 | 1786 | self._chunkcache = (0, b'') |
|
1794 | 1787 | |
|
1795 | 1788 | def deltaparent(self, rev): |
|
1796 | 1789 | """return deltaparent of the given revision""" |
|
1797 | 1790 | base = self.index[rev][3] |
|
1798 | 1791 | if base == rev: |
|
1799 | 1792 | return nullrev |
|
1800 | 1793 | elif self._generaldelta: |
|
1801 | 1794 | return base |
|
1802 | 1795 | else: |
|
1803 | 1796 | return rev - 1 |
|
1804 | 1797 | |
|
1805 | 1798 | def issnapshot(self, rev): |
|
1806 | 1799 | """tells whether rev is a snapshot""" |
|
1807 | 1800 | if not self._sparserevlog: |
|
1808 | 1801 | return self.deltaparent(rev) == nullrev |
|
1809 | 1802 | elif util.safehasattr(self.index, b'issnapshot'): |
|
1810 | 1803 | # directly assign the method to cache the testing and access |
|
1811 | 1804 | self.issnapshot = self.index.issnapshot |
|
1812 | 1805 | return self.issnapshot(rev) |
|
1813 | 1806 | if rev == nullrev: |
|
1814 | 1807 | return True |
|
1815 | 1808 | entry = self.index[rev] |
|
1816 | 1809 | base = entry[3] |
|
1817 | 1810 | if base == rev: |
|
1818 | 1811 | return True |
|
1819 | 1812 | if base == nullrev: |
|
1820 | 1813 | return True |
|
1821 | 1814 | p1 = entry[5] |
|
1822 | 1815 | p2 = entry[6] |
|
1823 | 1816 | if base == p1 or base == p2: |
|
1824 | 1817 | return False |
|
1825 | 1818 | return self.issnapshot(base) |
|
1826 | 1819 | |
|
1827 | 1820 | def snapshotdepth(self, rev): |
|
1828 | 1821 | """number of snapshot in the chain before this one""" |
|
1829 | 1822 | if not self.issnapshot(rev): |
|
1830 | 1823 | raise error.ProgrammingError(b'revision %d not a snapshot') |
|
1831 | 1824 | return len(self._deltachain(rev)[0]) - 1 |
|
1832 | 1825 | |
|
1833 | 1826 | def revdiff(self, rev1, rev2): |
|
1834 | 1827 | """return or calculate a delta between two revisions |
|
1835 | 1828 | |
|
1836 | 1829 | The delta calculated is in binary form and is intended to be written to |
|
1837 | 1830 | revlog data directly. So this function needs raw revision data. |
|
1838 | 1831 | """ |
|
1839 | 1832 | if rev1 != nullrev and self.deltaparent(rev2) == rev1: |
|
1840 | 1833 | return bytes(self._chunk(rev2)) |
|
1841 | 1834 | |
|
1842 | 1835 | return mdiff.textdiff(self.rawdata(rev1), self.rawdata(rev2)) |
|
1843 | 1836 | |
|
1844 | 1837 | def _processflags(self, text, flags, operation, raw=False): |
|
1845 | 1838 | """deprecated entry point to access flag processors""" |
|
1846 | 1839 | msg = b'_processflag(...) use the specialized variant' |
|
1847 | 1840 | util.nouideprecwarn(msg, b'5.2', stacklevel=2) |
|
1848 | 1841 | if raw: |
|
1849 | 1842 | return text, flagutil.processflagsraw(self, text, flags) |
|
1850 | 1843 | elif operation == b'read': |
|
1851 | 1844 | return flagutil.processflagsread(self, text, flags) |
|
1852 | 1845 | else: # write operation |
|
1853 | 1846 | return flagutil.processflagswrite(self, text, flags) |
|
1854 | 1847 | |
|
1855 | 1848 | def revision(self, nodeorrev, _df=None, raw=False): |
|
1856 | 1849 | """return an uncompressed revision of a given node or revision |
|
1857 | 1850 | number. |
|
1858 | 1851 | |
|
1859 | 1852 | _df - an existing file handle to read from. (internal-only) |
|
1860 | 1853 | raw - an optional argument specifying if the revision data is to be |
|
1861 | 1854 | treated as raw data when applying flag transforms. 'raw' should be set |
|
1862 | 1855 | to True when generating changegroups or in debug commands. |
|
1863 | 1856 | """ |
|
1864 | 1857 | if raw: |
|
1865 | 1858 | msg = ( |
|
1866 | 1859 | b'revlog.revision(..., raw=True) is deprecated, ' |
|
1867 | 1860 | b'use revlog.rawdata(...)' |
|
1868 | 1861 | ) |
|
1869 | 1862 | util.nouideprecwarn(msg, b'5.2', stacklevel=2) |
|
1870 | 1863 | return self._revisiondata(nodeorrev, _df, raw=raw)[0] |
|
1871 | 1864 | |
|
1872 | 1865 | def sidedata(self, nodeorrev, _df=None): |
|
1873 | 1866 | """a map of extra data related to the changeset but not part of the hash |
|
1874 | 1867 | |
|
1875 | 1868 | This function currently return a dictionary. However, more advanced |
|
1876 | 1869 | mapping object will likely be used in the future for a more |
|
1877 | 1870 | efficient/lazy code. |
|
1878 | 1871 | """ |
|
1879 | 1872 | return self._revisiondata(nodeorrev, _df)[1] |
|
1880 | 1873 | |
|
1881 | 1874 | def _revisiondata(self, nodeorrev, _df=None, raw=False): |
|
1882 | 1875 | # deal with <nodeorrev> argument type |
|
1883 | 1876 | if isinstance(nodeorrev, int): |
|
1884 | 1877 | rev = nodeorrev |
|
1885 | 1878 | node = self.node(rev) |
|
1886 | 1879 | else: |
|
1887 | 1880 | node = nodeorrev |
|
1888 | 1881 | rev = None |
|
1889 | 1882 | |
|
1890 | 1883 | # fast path the special `nullid` rev |
|
1891 | 1884 | if node == nullid: |
|
1892 | 1885 | return b"", {} |
|
1893 | 1886 | |
|
1894 | 1887 | # ``rawtext`` is the text as stored inside the revlog. Might be the |
|
1895 | 1888 | # revision or might need to be processed to retrieve the revision. |
|
1896 | 1889 | rev, rawtext, validated = self._rawtext(node, rev, _df=_df) |
|
1897 | 1890 | |
|
1898 | 1891 | if self.version & 0xFFFF == REVLOGV2: |
|
1899 | 1892 | if rev is None: |
|
1900 | 1893 | rev = self.rev(node) |
|
1901 | 1894 | sidedata = self._sidedata(rev) |
|
1902 | 1895 | else: |
|
1903 | 1896 | sidedata = {} |
|
1904 | 1897 | |
|
1905 | 1898 | if raw and validated: |
|
1906 | 1899 | # if we don't want to process the raw text and that raw |
|
1907 | 1900 | # text is cached, we can exit early. |
|
1908 | 1901 | return rawtext, sidedata |
|
1909 | 1902 | if rev is None: |
|
1910 | 1903 | rev = self.rev(node) |
|
1911 | 1904 | # the revlog's flag for this revision |
|
1912 | 1905 | # (usually alter its state or content) |
|
1913 | 1906 | flags = self.flags(rev) |
|
1914 | 1907 | |
|
1915 | 1908 | if validated and flags == REVIDX_DEFAULT_FLAGS: |
|
1916 | 1909 | # no extra flags set, no flag processor runs, text = rawtext |
|
1917 | 1910 | return rawtext, sidedata |
|
1918 | 1911 | |
|
1919 | 1912 | if raw: |
|
1920 | 1913 | validatehash = flagutil.processflagsraw(self, rawtext, flags) |
|
1921 | 1914 | text = rawtext |
|
1922 | 1915 | else: |
|
1923 | 1916 | r = flagutil.processflagsread(self, rawtext, flags) |
|
1924 | 1917 | text, validatehash = r |
|
1925 | 1918 | if validatehash: |
|
1926 | 1919 | self.checkhash(text, node, rev=rev) |
|
1927 | 1920 | if not validated: |
|
1928 | 1921 | self._revisioncache = (node, rev, rawtext) |
|
1929 | 1922 | |
|
1930 | 1923 | return text, sidedata |
|
1931 | 1924 | |
|
1932 | 1925 | def _rawtext(self, node, rev, _df=None): |
|
1933 | 1926 | """return the possibly unvalidated rawtext for a revision |
|
1934 | 1927 | |
|
1935 | 1928 | returns (rev, rawtext, validated) |
|
1936 | 1929 | """ |
|
1937 | 1930 | |
|
1938 | 1931 | # revision in the cache (could be useful to apply delta) |
|
1939 | 1932 | cachedrev = None |
|
1940 | 1933 | # An intermediate text to apply deltas to |
|
1941 | 1934 | basetext = None |
|
1942 | 1935 | |
|
1943 | 1936 | # Check if we have the entry in cache |
|
1944 | 1937 | # The cache entry looks like (node, rev, rawtext) |
|
1945 | 1938 | if self._revisioncache: |
|
1946 | 1939 | if self._revisioncache[0] == node: |
|
1947 | 1940 | return (rev, self._revisioncache[2], True) |
|
1948 | 1941 | cachedrev = self._revisioncache[1] |
|
1949 | 1942 | |
|
1950 | 1943 | if rev is None: |
|
1951 | 1944 | rev = self.rev(node) |
|
1952 | 1945 | |
|
1953 | 1946 | chain, stopped = self._deltachain(rev, stoprev=cachedrev) |
|
1954 | 1947 | if stopped: |
|
1955 | 1948 | basetext = self._revisioncache[2] |
|
1956 | 1949 | |
|
1957 | 1950 | # drop cache to save memory, the caller is expected to |
|
1958 | 1951 | # update self._revisioncache after validating the text |
|
1959 | 1952 | self._revisioncache = None |
|
1960 | 1953 | |
|
1961 | 1954 | targetsize = None |
|
1962 | 1955 | rawsize = self.index[rev][2] |
|
1963 | 1956 | if 0 <= rawsize: |
|
1964 | 1957 | targetsize = 4 * rawsize |
|
1965 | 1958 | |
|
1966 | 1959 | bins = self._chunks(chain, df=_df, targetsize=targetsize) |
|
1967 | 1960 | if basetext is None: |
|
1968 | 1961 | basetext = bytes(bins[0]) |
|
1969 | 1962 | bins = bins[1:] |
|
1970 | 1963 | |
|
1971 | 1964 | rawtext = mdiff.patches(basetext, bins) |
|
1972 | 1965 | del basetext # let us have a chance to free memory early |
|
1973 | 1966 | return (rev, rawtext, False) |
|
1974 | 1967 | |
|
1975 | 1968 | def _sidedata(self, rev): |
|
1976 | 1969 | """Return the sidedata for a given revision number.""" |
|
1977 | 1970 | index_entry = self.index[rev] |
|
1978 | 1971 | sidedata_offset = index_entry[8] |
|
1979 | 1972 | sidedata_size = index_entry[9] |
|
1980 | 1973 | |
|
1981 | 1974 | if self._inline: |
|
1982 |
sidedata_offset += self. |
|
|
1975 | sidedata_offset += self.index.entry_size * (1 + rev) | |
|
1983 | 1976 | if sidedata_size == 0: |
|
1984 | 1977 | return {} |
|
1985 | 1978 | |
|
1986 | 1979 | segment = self._getsegment(sidedata_offset, sidedata_size) |
|
1987 | 1980 | sidedata = sidedatautil.deserialize_sidedata(segment) |
|
1988 | 1981 | return sidedata |
|
1989 | 1982 | |
|
1990 | 1983 | def rawdata(self, nodeorrev, _df=None): |
|
1991 | 1984 | """return an uncompressed raw data of a given node or revision number. |
|
1992 | 1985 | |
|
1993 | 1986 | _df - an existing file handle to read from. (internal-only) |
|
1994 | 1987 | """ |
|
1995 | 1988 | return self._revisiondata(nodeorrev, _df, raw=True)[0] |
|
1996 | 1989 | |
|
1997 | 1990 | def hash(self, text, p1, p2): |
|
1998 | 1991 | """Compute a node hash. |
|
1999 | 1992 | |
|
2000 | 1993 | Available as a function so that subclasses can replace the hash |
|
2001 | 1994 | as needed. |
|
2002 | 1995 | """ |
|
2003 | 1996 | return storageutil.hashrevisionsha1(text, p1, p2) |
|
2004 | 1997 | |
|
2005 | 1998 | def checkhash(self, text, node, p1=None, p2=None, rev=None): |
|
2006 | 1999 | """Check node hash integrity. |
|
2007 | 2000 | |
|
2008 | 2001 | Available as a function so that subclasses can extend hash mismatch |
|
2009 | 2002 | behaviors as needed. |
|
2010 | 2003 | """ |
|
2011 | 2004 | try: |
|
2012 | 2005 | if p1 is None and p2 is None: |
|
2013 | 2006 | p1, p2 = self.parents(node) |
|
2014 | 2007 | if node != self.hash(text, p1, p2): |
|
2015 | 2008 | # Clear the revision cache on hash failure. The revision cache |
|
2016 | 2009 | # only stores the raw revision and clearing the cache does have |
|
2017 | 2010 | # the side-effect that we won't have a cache hit when the raw |
|
2018 | 2011 | # revision data is accessed. But this case should be rare and |
|
2019 | 2012 | # it is extra work to teach the cache about the hash |
|
2020 | 2013 | # verification state. |
|
2021 | 2014 | if self._revisioncache and self._revisioncache[0] == node: |
|
2022 | 2015 | self._revisioncache = None |
|
2023 | 2016 | |
|
2024 | 2017 | revornode = rev |
|
2025 | 2018 | if revornode is None: |
|
2026 | 2019 | revornode = templatefilters.short(hex(node)) |
|
2027 | 2020 | raise error.RevlogError( |
|
2028 | 2021 | _(b"integrity check failed on %s:%s") |
|
2029 | 2022 | % (self.indexfile, pycompat.bytestr(revornode)) |
|
2030 | 2023 | ) |
|
2031 | 2024 | except error.RevlogError: |
|
2032 | 2025 | if self._censorable and storageutil.iscensoredtext(text): |
|
2033 | 2026 | raise error.CensoredNodeError(self.indexfile, node, text) |
|
2034 | 2027 | raise |
|
2035 | 2028 | |
|
2036 | 2029 | def _enforceinlinesize(self, tr, fp=None): |
|
2037 | 2030 | """Check if the revlog is too big for inline and convert if so. |
|
2038 | 2031 | |
|
2039 | 2032 | This should be called after revisions are added to the revlog. If the |
|
2040 | 2033 | revlog has grown too large to be an inline revlog, it will convert it |
|
2041 | 2034 | to use multiple index and data files. |
|
2042 | 2035 | """ |
|
2043 | 2036 | tiprev = len(self) - 1 |
|
2044 | 2037 | if ( |
|
2045 | 2038 | not self._inline |
|
2046 | 2039 | or (self.start(tiprev) + self.length(tiprev)) < _maxinline |
|
2047 | 2040 | ): |
|
2048 | 2041 | return |
|
2049 | 2042 | |
|
2050 | 2043 | troffset = tr.findoffset(self.indexfile) |
|
2051 | 2044 | if troffset is None: |
|
2052 | 2045 | raise error.RevlogError( |
|
2053 | 2046 | _(b"%s not found in the transaction") % self.indexfile |
|
2054 | 2047 | ) |
|
2055 | 2048 | trindex = 0 |
|
2056 | 2049 | tr.add(self.datafile, 0) |
|
2057 | 2050 | |
|
2058 | 2051 | if fp: |
|
2059 | 2052 | fp.flush() |
|
2060 | 2053 | fp.close() |
|
2061 | 2054 | # We can't use the cached file handle after close(). So prevent |
|
2062 | 2055 | # its usage. |
|
2063 | 2056 | self._writinghandles = None |
|
2064 | 2057 | |
|
2065 | 2058 | with self._indexfp(b'r') as ifh, self._datafp(b'w') as dfh: |
|
2066 | 2059 | for r in self: |
|
2067 | 2060 | dfh.write(self._getsegmentforrevs(r, r, df=ifh)[1]) |
|
2068 | 2061 | if troffset <= self.start(r): |
|
2069 | 2062 | trindex = r |
|
2070 | 2063 | |
|
2071 | 2064 | with self._indexfp(b'w') as fp: |
|
2072 | 2065 | self.version &= ~FLAG_INLINE_DATA |
|
2073 | 2066 | self._inline = False |
|
2074 | 2067 | io = self._io |
|
2075 | 2068 | for i in self: |
|
2076 | 2069 | e = io.packentry(self.index[i], self.node, self.version, i) |
|
2077 | 2070 | fp.write(e) |
|
2078 | 2071 | |
|
2079 | 2072 | # the temp file replace the real index when we exit the context |
|
2080 | 2073 | # manager |
|
2081 | 2074 | |
|
2082 |
tr.replace(self.indexfile, trindex * self. |
|
|
2075 | tr.replace(self.indexfile, trindex * self.index.entry_size) | |
|
2083 | 2076 | nodemaputil.setup_persistent_nodemap(tr, self) |
|
2084 | 2077 | self._chunkclear() |
|
2085 | 2078 | |
|
2086 | 2079 | def _nodeduplicatecallback(self, transaction, node): |
|
2087 | 2080 | """called when trying to add a node already stored.""" |
|
2088 | 2081 | |
|
2089 | 2082 | def addrevision( |
|
2090 | 2083 | self, |
|
2091 | 2084 | text, |
|
2092 | 2085 | transaction, |
|
2093 | 2086 | link, |
|
2094 | 2087 | p1, |
|
2095 | 2088 | p2, |
|
2096 | 2089 | cachedelta=None, |
|
2097 | 2090 | node=None, |
|
2098 | 2091 | flags=REVIDX_DEFAULT_FLAGS, |
|
2099 | 2092 | deltacomputer=None, |
|
2100 | 2093 | sidedata=None, |
|
2101 | 2094 | ): |
|
2102 | 2095 | """add a revision to the log |
|
2103 | 2096 | |
|
2104 | 2097 | text - the revision data to add |
|
2105 | 2098 | transaction - the transaction object used for rollback |
|
2106 | 2099 | link - the linkrev data to add |
|
2107 | 2100 | p1, p2 - the parent nodeids of the revision |
|
2108 | 2101 | cachedelta - an optional precomputed delta |
|
2109 | 2102 | node - nodeid of revision; typically node is not specified, and it is |
|
2110 | 2103 | computed by default as hash(text, p1, p2), however subclasses might |
|
2111 | 2104 | use different hashing method (and override checkhash() in such case) |
|
2112 | 2105 | flags - the known flags to set on the revision |
|
2113 | 2106 | deltacomputer - an optional deltacomputer instance shared between |
|
2114 | 2107 | multiple calls |
|
2115 | 2108 | """ |
|
2116 | 2109 | if link == nullrev: |
|
2117 | 2110 | raise error.RevlogError( |
|
2118 | 2111 | _(b"attempted to add linkrev -1 to %s") % self.indexfile |
|
2119 | 2112 | ) |
|
2120 | 2113 | |
|
2121 | 2114 | if sidedata is None: |
|
2122 | 2115 | sidedata = {} |
|
2123 | 2116 | elif not self.hassidedata: |
|
2124 | 2117 | raise error.ProgrammingError( |
|
2125 | 2118 | _(b"trying to add sidedata to a revlog who don't support them") |
|
2126 | 2119 | ) |
|
2127 | 2120 | |
|
2128 | 2121 | if flags: |
|
2129 | 2122 | node = node or self.hash(text, p1, p2) |
|
2130 | 2123 | |
|
2131 | 2124 | rawtext, validatehash = flagutil.processflagswrite(self, text, flags) |
|
2132 | 2125 | |
|
2133 | 2126 | # If the flag processor modifies the revision data, ignore any provided |
|
2134 | 2127 | # cachedelta. |
|
2135 | 2128 | if rawtext != text: |
|
2136 | 2129 | cachedelta = None |
|
2137 | 2130 | |
|
2138 | 2131 | if len(rawtext) > _maxentrysize: |
|
2139 | 2132 | raise error.RevlogError( |
|
2140 | 2133 | _( |
|
2141 | 2134 | b"%s: size of %d bytes exceeds maximum revlog storage of 2GiB" |
|
2142 | 2135 | ) |
|
2143 | 2136 | % (self.indexfile, len(rawtext)) |
|
2144 | 2137 | ) |
|
2145 | 2138 | |
|
2146 | 2139 | node = node or self.hash(rawtext, p1, p2) |
|
2147 | 2140 | rev = self.index.get_rev(node) |
|
2148 | 2141 | if rev is not None: |
|
2149 | 2142 | return rev |
|
2150 | 2143 | |
|
2151 | 2144 | if validatehash: |
|
2152 | 2145 | self.checkhash(rawtext, node, p1=p1, p2=p2) |
|
2153 | 2146 | |
|
2154 | 2147 | return self.addrawrevision( |
|
2155 | 2148 | rawtext, |
|
2156 | 2149 | transaction, |
|
2157 | 2150 | link, |
|
2158 | 2151 | p1, |
|
2159 | 2152 | p2, |
|
2160 | 2153 | node, |
|
2161 | 2154 | flags, |
|
2162 | 2155 | cachedelta=cachedelta, |
|
2163 | 2156 | deltacomputer=deltacomputer, |
|
2164 | 2157 | sidedata=sidedata, |
|
2165 | 2158 | ) |
|
2166 | 2159 | |
|
2167 | 2160 | def addrawrevision( |
|
2168 | 2161 | self, |
|
2169 | 2162 | rawtext, |
|
2170 | 2163 | transaction, |
|
2171 | 2164 | link, |
|
2172 | 2165 | p1, |
|
2173 | 2166 | p2, |
|
2174 | 2167 | node, |
|
2175 | 2168 | flags, |
|
2176 | 2169 | cachedelta=None, |
|
2177 | 2170 | deltacomputer=None, |
|
2178 | 2171 | sidedata=None, |
|
2179 | 2172 | ): |
|
2180 | 2173 | """add a raw revision with known flags, node and parents |
|
2181 | 2174 | useful when reusing a revision not stored in this revlog (ex: received |
|
2182 | 2175 | over wire, or read from an external bundle). |
|
2183 | 2176 | """ |
|
2184 | 2177 | dfh = None |
|
2185 | 2178 | if not self._inline: |
|
2186 | 2179 | dfh = self._datafp(b"a+") |
|
2187 | 2180 | ifh = self._indexfp(b"a+") |
|
2188 | 2181 | try: |
|
2189 | 2182 | return self._addrevision( |
|
2190 | 2183 | node, |
|
2191 | 2184 | rawtext, |
|
2192 | 2185 | transaction, |
|
2193 | 2186 | link, |
|
2194 | 2187 | p1, |
|
2195 | 2188 | p2, |
|
2196 | 2189 | flags, |
|
2197 | 2190 | cachedelta, |
|
2198 | 2191 | ifh, |
|
2199 | 2192 | dfh, |
|
2200 | 2193 | deltacomputer=deltacomputer, |
|
2201 | 2194 | sidedata=sidedata, |
|
2202 | 2195 | ) |
|
2203 | 2196 | finally: |
|
2204 | 2197 | if dfh: |
|
2205 | 2198 | dfh.close() |
|
2206 | 2199 | ifh.close() |
|
2207 | 2200 | |
|
2208 | 2201 | def compress(self, data): |
|
2209 | 2202 | """Generate a possibly-compressed representation of data.""" |
|
2210 | 2203 | if not data: |
|
2211 | 2204 | return b'', data |
|
2212 | 2205 | |
|
2213 | 2206 | compressed = self._compressor.compress(data) |
|
2214 | 2207 | |
|
2215 | 2208 | if compressed: |
|
2216 | 2209 | # The revlog compressor added the header in the returned data. |
|
2217 | 2210 | return b'', compressed |
|
2218 | 2211 | |
|
2219 | 2212 | if data[0:1] == b'\0': |
|
2220 | 2213 | return b'', data |
|
2221 | 2214 | return b'u', data |
|
2222 | 2215 | |
|
2223 | 2216 | def decompress(self, data): |
|
2224 | 2217 | """Decompress a revlog chunk. |
|
2225 | 2218 | |
|
2226 | 2219 | The chunk is expected to begin with a header identifying the |
|
2227 | 2220 | format type so it can be routed to an appropriate decompressor. |
|
2228 | 2221 | """ |
|
2229 | 2222 | if not data: |
|
2230 | 2223 | return data |
|
2231 | 2224 | |
|
2232 | 2225 | # Revlogs are read much more frequently than they are written and many |
|
2233 | 2226 | # chunks only take microseconds to decompress, so performance is |
|
2234 | 2227 | # important here. |
|
2235 | 2228 | # |
|
2236 | 2229 | # We can make a few assumptions about revlogs: |
|
2237 | 2230 | # |
|
2238 | 2231 | # 1) the majority of chunks will be compressed (as opposed to inline |
|
2239 | 2232 | # raw data). |
|
2240 | 2233 | # 2) decompressing *any* data will likely by at least 10x slower than |
|
2241 | 2234 | # returning raw inline data. |
|
2242 | 2235 | # 3) we want to prioritize common and officially supported compression |
|
2243 | 2236 | # engines |
|
2244 | 2237 | # |
|
2245 | 2238 | # It follows that we want to optimize for "decompress compressed data |
|
2246 | 2239 | # when encoded with common and officially supported compression engines" |
|
2247 | 2240 | # case over "raw data" and "data encoded by less common or non-official |
|
2248 | 2241 | # compression engines." That is why we have the inline lookup first |
|
2249 | 2242 | # followed by the compengines lookup. |
|
2250 | 2243 | # |
|
2251 | 2244 | # According to `hg perfrevlogchunks`, this is ~0.5% faster for zlib |
|
2252 | 2245 | # compressed chunks. And this matters for changelog and manifest reads. |
|
2253 | 2246 | t = data[0:1] |
|
2254 | 2247 | |
|
2255 | 2248 | if t == b'x': |
|
2256 | 2249 | try: |
|
2257 | 2250 | return _zlibdecompress(data) |
|
2258 | 2251 | except zlib.error as e: |
|
2259 | 2252 | raise error.RevlogError( |
|
2260 | 2253 | _(b'revlog decompress error: %s') |
|
2261 | 2254 | % stringutil.forcebytestr(e) |
|
2262 | 2255 | ) |
|
2263 | 2256 | # '\0' is more common than 'u' so it goes first. |
|
2264 | 2257 | elif t == b'\0': |
|
2265 | 2258 | return data |
|
2266 | 2259 | elif t == b'u': |
|
2267 | 2260 | return util.buffer(data, 1) |
|
2268 | 2261 | |
|
2269 | 2262 | try: |
|
2270 | 2263 | compressor = self._decompressors[t] |
|
2271 | 2264 | except KeyError: |
|
2272 | 2265 | try: |
|
2273 | 2266 | engine = util.compengines.forrevlogheader(t) |
|
2274 | 2267 | compressor = engine.revlogcompressor(self._compengineopts) |
|
2275 | 2268 | self._decompressors[t] = compressor |
|
2276 | 2269 | except KeyError: |
|
2277 | 2270 | raise error.RevlogError( |
|
2278 | 2271 | _(b'unknown compression type %s') % binascii.hexlify(t) |
|
2279 | 2272 | ) |
|
2280 | 2273 | |
|
2281 | 2274 | return compressor.decompress(data) |
|
2282 | 2275 | |
|
2283 | 2276 | def _addrevision( |
|
2284 | 2277 | self, |
|
2285 | 2278 | node, |
|
2286 | 2279 | rawtext, |
|
2287 | 2280 | transaction, |
|
2288 | 2281 | link, |
|
2289 | 2282 | p1, |
|
2290 | 2283 | p2, |
|
2291 | 2284 | flags, |
|
2292 | 2285 | cachedelta, |
|
2293 | 2286 | ifh, |
|
2294 | 2287 | dfh, |
|
2295 | 2288 | alwayscache=False, |
|
2296 | 2289 | deltacomputer=None, |
|
2297 | 2290 | sidedata=None, |
|
2298 | 2291 | ): |
|
2299 | 2292 | """internal function to add revisions to the log |
|
2300 | 2293 | |
|
2301 | 2294 | see addrevision for argument descriptions. |
|
2302 | 2295 | |
|
2303 | 2296 | note: "addrevision" takes non-raw text, "_addrevision" takes raw text. |
|
2304 | 2297 | |
|
2305 | 2298 | if "deltacomputer" is not provided or None, a defaultdeltacomputer will |
|
2306 | 2299 | be used. |
|
2307 | 2300 | |
|
2308 | 2301 | invariants: |
|
2309 | 2302 | - rawtext is optional (can be None); if not set, cachedelta must be set. |
|
2310 | 2303 | if both are set, they must correspond to each other. |
|
2311 | 2304 | """ |
|
2312 | 2305 | if node == nullid: |
|
2313 | 2306 | raise error.RevlogError( |
|
2314 | 2307 | _(b"%s: attempt to add null revision") % self.indexfile |
|
2315 | 2308 | ) |
|
2316 | 2309 | if node == wdirid or node in wdirfilenodeids: |
|
2317 | 2310 | raise error.RevlogError( |
|
2318 | 2311 | _(b"%s: attempt to add wdir revision") % self.indexfile |
|
2319 | 2312 | ) |
|
2320 | 2313 | |
|
2321 | 2314 | if self._inline: |
|
2322 | 2315 | fh = ifh |
|
2323 | 2316 | else: |
|
2324 | 2317 | fh = dfh |
|
2325 | 2318 | |
|
2326 | 2319 | btext = [rawtext] |
|
2327 | 2320 | |
|
2328 | 2321 | curr = len(self) |
|
2329 | 2322 | prev = curr - 1 |
|
2330 | 2323 | |
|
2331 | 2324 | offset = self._get_data_offset(prev) |
|
2332 | 2325 | |
|
2333 | 2326 | if self._concurrencychecker: |
|
2334 | 2327 | if self._inline: |
|
2335 | 2328 | # offset is "as if" it were in the .d file, so we need to add on |
|
2336 | 2329 | # the size of the entry metadata. |
|
2337 | 2330 | self._concurrencychecker( |
|
2338 |
ifh, self.indexfile, offset + curr * self. |
|
|
2331 | ifh, self.indexfile, offset + curr * self.index.entry_size | |
|
2339 | 2332 | ) |
|
2340 | 2333 | else: |
|
2341 | 2334 | # Entries in the .i are a consistent size. |
|
2342 | 2335 | self._concurrencychecker( |
|
2343 |
ifh, self.indexfile, curr * self. |
|
|
2336 | ifh, self.indexfile, curr * self.index.entry_size | |
|
2344 | 2337 | ) |
|
2345 | 2338 | self._concurrencychecker(dfh, self.datafile, offset) |
|
2346 | 2339 | |
|
2347 | 2340 | p1r, p2r = self.rev(p1), self.rev(p2) |
|
2348 | 2341 | |
|
2349 | 2342 | # full versions are inserted when the needed deltas |
|
2350 | 2343 | # become comparable to the uncompressed text |
|
2351 | 2344 | if rawtext is None: |
|
2352 | 2345 | # need rawtext size, before changed by flag processors, which is |
|
2353 | 2346 | # the non-raw size. use revlog explicitly to avoid filelog's extra |
|
2354 | 2347 | # logic that might remove metadata size. |
|
2355 | 2348 | textlen = mdiff.patchedsize( |
|
2356 | 2349 | revlog.size(self, cachedelta[0]), cachedelta[1] |
|
2357 | 2350 | ) |
|
2358 | 2351 | else: |
|
2359 | 2352 | textlen = len(rawtext) |
|
2360 | 2353 | |
|
2361 | 2354 | if deltacomputer is None: |
|
2362 | 2355 | deltacomputer = deltautil.deltacomputer(self) |
|
2363 | 2356 | |
|
2364 | 2357 | revinfo = _revisioninfo(node, p1, p2, btext, textlen, cachedelta, flags) |
|
2365 | 2358 | |
|
2366 | 2359 | deltainfo = deltacomputer.finddeltainfo(revinfo, fh) |
|
2367 | 2360 | |
|
2368 | 2361 | if sidedata: |
|
2369 | 2362 | serialized_sidedata = sidedatautil.serialize_sidedata(sidedata) |
|
2370 | 2363 | sidedata_offset = offset + deltainfo.deltalen |
|
2371 | 2364 | else: |
|
2372 | 2365 | serialized_sidedata = b"" |
|
2373 | 2366 | # Don't store the offset if the sidedata is empty, that way |
|
2374 | 2367 | # we can easily detect empty sidedata and they will be no different |
|
2375 | 2368 | # than ones we manually add. |
|
2376 | 2369 | sidedata_offset = 0 |
|
2377 | 2370 | |
|
2378 | 2371 | e = ( |
|
2379 | 2372 | offset_type(offset, flags), |
|
2380 | 2373 | deltainfo.deltalen, |
|
2381 | 2374 | textlen, |
|
2382 | 2375 | deltainfo.base, |
|
2383 | 2376 | link, |
|
2384 | 2377 | p1r, |
|
2385 | 2378 | p2r, |
|
2386 | 2379 | node, |
|
2387 | 2380 | sidedata_offset, |
|
2388 | 2381 | len(serialized_sidedata), |
|
2389 | 2382 | ) |
|
2390 | 2383 | |
|
2391 | 2384 | if self.version & 0xFFFF != REVLOGV2: |
|
2392 | 2385 | e = e[:8] |
|
2393 | 2386 | |
|
2394 | 2387 | self.index.append(e) |
|
2395 | 2388 | entry = self._io.packentry(e, self.node, self.version, curr) |
|
2396 | 2389 | self._writeentry( |
|
2397 | 2390 | transaction, |
|
2398 | 2391 | ifh, |
|
2399 | 2392 | dfh, |
|
2400 | 2393 | entry, |
|
2401 | 2394 | deltainfo.data, |
|
2402 | 2395 | link, |
|
2403 | 2396 | offset, |
|
2404 | 2397 | serialized_sidedata, |
|
2405 | 2398 | ) |
|
2406 | 2399 | |
|
2407 | 2400 | rawtext = btext[0] |
|
2408 | 2401 | |
|
2409 | 2402 | if alwayscache and rawtext is None: |
|
2410 | 2403 | rawtext = deltacomputer.buildtext(revinfo, fh) |
|
2411 | 2404 | |
|
2412 | 2405 | if type(rawtext) == bytes: # only accept immutable objects |
|
2413 | 2406 | self._revisioncache = (node, curr, rawtext) |
|
2414 | 2407 | self._chainbasecache[curr] = deltainfo.chainbase |
|
2415 | 2408 | return curr |
|
2416 | 2409 | |
|
2417 | 2410 | def _get_data_offset(self, prev): |
|
2418 | 2411 | """Returns the current offset in the (in-transaction) data file. |
|
2419 | 2412 | Versions < 2 of the revlog can get this 0(1), revlog v2 needs a docket |
|
2420 | 2413 | file to store that information: since sidedata can be rewritten to the |
|
2421 | 2414 | end of the data file within a transaction, you can have cases where, for |
|
2422 | 2415 | example, rev `n` does not have sidedata while rev `n - 1` does, leading |
|
2423 | 2416 | to `n - 1`'s sidedata being written after `n`'s data. |
|
2424 | 2417 | |
|
2425 | 2418 | TODO cache this in a docket file before getting out of experimental.""" |
|
2426 | 2419 | if self.version & 0xFFFF != REVLOGV2: |
|
2427 | 2420 | return self.end(prev) |
|
2428 | 2421 | |
|
2429 | 2422 | offset = 0 |
|
2430 | 2423 | for rev, entry in enumerate(self.index): |
|
2431 | 2424 | sidedata_end = entry[8] + entry[9] |
|
2432 | 2425 | # Sidedata for a previous rev has potentially been written after |
|
2433 | 2426 | # this rev's end, so take the max. |
|
2434 | 2427 | offset = max(self.end(rev), offset, sidedata_end) |
|
2435 | 2428 | return offset |
|
2436 | 2429 | |
|
2437 | 2430 | def _writeentry( |
|
2438 | 2431 | self, transaction, ifh, dfh, entry, data, link, offset, sidedata |
|
2439 | 2432 | ): |
|
2440 | 2433 | # Files opened in a+ mode have inconsistent behavior on various |
|
2441 | 2434 | # platforms. Windows requires that a file positioning call be made |
|
2442 | 2435 | # when the file handle transitions between reads and writes. See |
|
2443 | 2436 | # 3686fa2b8eee and the mixedfilemodewrapper in windows.py. On other |
|
2444 | 2437 | # platforms, Python or the platform itself can be buggy. Some versions |
|
2445 | 2438 | # of Solaris have been observed to not append at the end of the file |
|
2446 | 2439 | # if the file was seeked to before the end. See issue4943 for more. |
|
2447 | 2440 | # |
|
2448 | 2441 | # We work around this issue by inserting a seek() before writing. |
|
2449 | 2442 | # Note: This is likely not necessary on Python 3. However, because |
|
2450 | 2443 | # the file handle is reused for reads and may be seeked there, we need |
|
2451 | 2444 | # to be careful before changing this. |
|
2452 | 2445 | ifh.seek(0, os.SEEK_END) |
|
2453 | 2446 | if dfh: |
|
2454 | 2447 | dfh.seek(0, os.SEEK_END) |
|
2455 | 2448 | |
|
2456 | 2449 | curr = len(self) - 1 |
|
2457 | 2450 | if not self._inline: |
|
2458 | 2451 | transaction.add(self.datafile, offset) |
|
2459 | 2452 | transaction.add(self.indexfile, curr * len(entry)) |
|
2460 | 2453 | if data[0]: |
|
2461 | 2454 | dfh.write(data[0]) |
|
2462 | 2455 | dfh.write(data[1]) |
|
2463 | 2456 | if sidedata: |
|
2464 | 2457 | dfh.write(sidedata) |
|
2465 | 2458 | ifh.write(entry) |
|
2466 | 2459 | else: |
|
2467 |
offset += curr * self. |
|
|
2460 | offset += curr * self.index.entry_size | |
|
2468 | 2461 | transaction.add(self.indexfile, offset) |
|
2469 | 2462 | ifh.write(entry) |
|
2470 | 2463 | ifh.write(data[0]) |
|
2471 | 2464 | ifh.write(data[1]) |
|
2472 | 2465 | if sidedata: |
|
2473 | 2466 | ifh.write(sidedata) |
|
2474 | 2467 | self._enforceinlinesize(transaction, ifh) |
|
2475 | 2468 | nodemaputil.setup_persistent_nodemap(transaction, self) |
|
2476 | 2469 | |
|
2477 | 2470 | def addgroup( |
|
2478 | 2471 | self, |
|
2479 | 2472 | deltas, |
|
2480 | 2473 | linkmapper, |
|
2481 | 2474 | transaction, |
|
2482 | 2475 | alwayscache=False, |
|
2483 | 2476 | addrevisioncb=None, |
|
2484 | 2477 | duplicaterevisioncb=None, |
|
2485 | 2478 | ): |
|
2486 | 2479 | """ |
|
2487 | 2480 | add a delta group |
|
2488 | 2481 | |
|
2489 | 2482 | given a set of deltas, add them to the revision log. the |
|
2490 | 2483 | first delta is against its parent, which should be in our |
|
2491 | 2484 | log, the rest are against the previous delta. |
|
2492 | 2485 | |
|
2493 | 2486 | If ``addrevisioncb`` is defined, it will be called with arguments of |
|
2494 | 2487 | this revlog and the node that was added. |
|
2495 | 2488 | """ |
|
2496 | 2489 | |
|
2497 | 2490 | if self._writinghandles: |
|
2498 | 2491 | raise error.ProgrammingError(b'cannot nest addgroup() calls') |
|
2499 | 2492 | |
|
2500 | 2493 | r = len(self) |
|
2501 | 2494 | end = 0 |
|
2502 | 2495 | if r: |
|
2503 | 2496 | end = self.end(r - 1) |
|
2504 | 2497 | ifh = self._indexfp(b"a+") |
|
2505 |
isize = r * self. |
|
|
2498 | isize = r * self.index.entry_size | |
|
2506 | 2499 | if self._inline: |
|
2507 | 2500 | transaction.add(self.indexfile, end + isize) |
|
2508 | 2501 | dfh = None |
|
2509 | 2502 | else: |
|
2510 | 2503 | transaction.add(self.indexfile, isize) |
|
2511 | 2504 | transaction.add(self.datafile, end) |
|
2512 | 2505 | dfh = self._datafp(b"a+") |
|
2513 | 2506 | |
|
2514 | 2507 | def flush(): |
|
2515 | 2508 | if dfh: |
|
2516 | 2509 | dfh.flush() |
|
2517 | 2510 | ifh.flush() |
|
2518 | 2511 | |
|
2519 | 2512 | self._writinghandles = (ifh, dfh) |
|
2520 | 2513 | empty = True |
|
2521 | 2514 | |
|
2522 | 2515 | try: |
|
2523 | 2516 | deltacomputer = deltautil.deltacomputer(self) |
|
2524 | 2517 | # loop through our set of deltas |
|
2525 | 2518 | for data in deltas: |
|
2526 | 2519 | node, p1, p2, linknode, deltabase, delta, flags, sidedata = data |
|
2527 | 2520 | link = linkmapper(linknode) |
|
2528 | 2521 | flags = flags or REVIDX_DEFAULT_FLAGS |
|
2529 | 2522 | |
|
2530 | 2523 | rev = self.index.get_rev(node) |
|
2531 | 2524 | if rev is not None: |
|
2532 | 2525 | # this can happen if two branches make the same change |
|
2533 | 2526 | self._nodeduplicatecallback(transaction, rev) |
|
2534 | 2527 | if duplicaterevisioncb: |
|
2535 | 2528 | duplicaterevisioncb(self, rev) |
|
2536 | 2529 | empty = False |
|
2537 | 2530 | continue |
|
2538 | 2531 | |
|
2539 | 2532 | for p in (p1, p2): |
|
2540 | 2533 | if not self.index.has_node(p): |
|
2541 | 2534 | raise error.LookupError( |
|
2542 | 2535 | p, self.indexfile, _(b'unknown parent') |
|
2543 | 2536 | ) |
|
2544 | 2537 | |
|
2545 | 2538 | if not self.index.has_node(deltabase): |
|
2546 | 2539 | raise error.LookupError( |
|
2547 | 2540 | deltabase, self.indexfile, _(b'unknown delta base') |
|
2548 | 2541 | ) |
|
2549 | 2542 | |
|
2550 | 2543 | baserev = self.rev(deltabase) |
|
2551 | 2544 | |
|
2552 | 2545 | if baserev != nullrev and self.iscensored(baserev): |
|
2553 | 2546 | # if base is censored, delta must be full replacement in a |
|
2554 | 2547 | # single patch operation |
|
2555 | 2548 | hlen = struct.calcsize(b">lll") |
|
2556 | 2549 | oldlen = self.rawsize(baserev) |
|
2557 | 2550 | newlen = len(delta) - hlen |
|
2558 | 2551 | if delta[:hlen] != mdiff.replacediffheader(oldlen, newlen): |
|
2559 | 2552 | raise error.CensoredBaseError( |
|
2560 | 2553 | self.indexfile, self.node(baserev) |
|
2561 | 2554 | ) |
|
2562 | 2555 | |
|
2563 | 2556 | if not flags and self._peek_iscensored(baserev, delta, flush): |
|
2564 | 2557 | flags |= REVIDX_ISCENSORED |
|
2565 | 2558 | |
|
2566 | 2559 | # We assume consumers of addrevisioncb will want to retrieve |
|
2567 | 2560 | # the added revision, which will require a call to |
|
2568 | 2561 | # revision(). revision() will fast path if there is a cache |
|
2569 | 2562 | # hit. So, we tell _addrevision() to always cache in this case. |
|
2570 | 2563 | # We're only using addgroup() in the context of changegroup |
|
2571 | 2564 | # generation so the revision data can always be handled as raw |
|
2572 | 2565 | # by the flagprocessor. |
|
2573 | 2566 | rev = self._addrevision( |
|
2574 | 2567 | node, |
|
2575 | 2568 | None, |
|
2576 | 2569 | transaction, |
|
2577 | 2570 | link, |
|
2578 | 2571 | p1, |
|
2579 | 2572 | p2, |
|
2580 | 2573 | flags, |
|
2581 | 2574 | (baserev, delta), |
|
2582 | 2575 | ifh, |
|
2583 | 2576 | dfh, |
|
2584 | 2577 | alwayscache=alwayscache, |
|
2585 | 2578 | deltacomputer=deltacomputer, |
|
2586 | 2579 | sidedata=sidedata, |
|
2587 | 2580 | ) |
|
2588 | 2581 | |
|
2589 | 2582 | if addrevisioncb: |
|
2590 | 2583 | addrevisioncb(self, rev) |
|
2591 | 2584 | empty = False |
|
2592 | 2585 | |
|
2593 | 2586 | if not dfh and not self._inline: |
|
2594 | 2587 | # addrevision switched from inline to conventional |
|
2595 | 2588 | # reopen the index |
|
2596 | 2589 | ifh.close() |
|
2597 | 2590 | dfh = self._datafp(b"a+") |
|
2598 | 2591 | ifh = self._indexfp(b"a+") |
|
2599 | 2592 | self._writinghandles = (ifh, dfh) |
|
2600 | 2593 | finally: |
|
2601 | 2594 | self._writinghandles = None |
|
2602 | 2595 | |
|
2603 | 2596 | if dfh: |
|
2604 | 2597 | dfh.close() |
|
2605 | 2598 | ifh.close() |
|
2606 | 2599 | return not empty |
|
2607 | 2600 | |
|
2608 | 2601 | def iscensored(self, rev): |
|
2609 | 2602 | """Check if a file revision is censored.""" |
|
2610 | 2603 | if not self._censorable: |
|
2611 | 2604 | return False |
|
2612 | 2605 | |
|
2613 | 2606 | return self.flags(rev) & REVIDX_ISCENSORED |
|
2614 | 2607 | |
|
2615 | 2608 | def _peek_iscensored(self, baserev, delta, flush): |
|
2616 | 2609 | """Quickly check if a delta produces a censored revision.""" |
|
2617 | 2610 | if not self._censorable: |
|
2618 | 2611 | return False |
|
2619 | 2612 | |
|
2620 | 2613 | return storageutil.deltaiscensored(delta, baserev, self.rawsize) |
|
2621 | 2614 | |
|
2622 | 2615 | def getstrippoint(self, minlink): |
|
2623 | 2616 | """find the minimum rev that must be stripped to strip the linkrev |
|
2624 | 2617 | |
|
2625 | 2618 | Returns a tuple containing the minimum rev and a set of all revs that |
|
2626 | 2619 | have linkrevs that will be broken by this strip. |
|
2627 | 2620 | """ |
|
2628 | 2621 | return storageutil.resolvestripinfo( |
|
2629 | 2622 | minlink, |
|
2630 | 2623 | len(self) - 1, |
|
2631 | 2624 | self.headrevs(), |
|
2632 | 2625 | self.linkrev, |
|
2633 | 2626 | self.parentrevs, |
|
2634 | 2627 | ) |
|
2635 | 2628 | |
|
2636 | 2629 | def strip(self, minlink, transaction): |
|
2637 | 2630 | """truncate the revlog on the first revision with a linkrev >= minlink |
|
2638 | 2631 | |
|
2639 | 2632 | This function is called when we're stripping revision minlink and |
|
2640 | 2633 | its descendants from the repository. |
|
2641 | 2634 | |
|
2642 | 2635 | We have to remove all revisions with linkrev >= minlink, because |
|
2643 | 2636 | the equivalent changelog revisions will be renumbered after the |
|
2644 | 2637 | strip. |
|
2645 | 2638 | |
|
2646 | 2639 | So we truncate the revlog on the first of these revisions, and |
|
2647 | 2640 | trust that the caller has saved the revisions that shouldn't be |
|
2648 | 2641 | removed and that it'll re-add them after this truncation. |
|
2649 | 2642 | """ |
|
2650 | 2643 | if len(self) == 0: |
|
2651 | 2644 | return |
|
2652 | 2645 | |
|
2653 | 2646 | rev, _ = self.getstrippoint(minlink) |
|
2654 | 2647 | if rev == len(self): |
|
2655 | 2648 | return |
|
2656 | 2649 | |
|
2657 | 2650 | # first truncate the files on disk |
|
2658 | 2651 | end = self.start(rev) |
|
2659 | 2652 | if not self._inline: |
|
2660 | 2653 | transaction.add(self.datafile, end) |
|
2661 |
end = rev * self. |
|
|
2654 | end = rev * self.index.entry_size | |
|
2662 | 2655 | else: |
|
2663 |
end += rev * self. |
|
|
2656 | end += rev * self.index.entry_size | |
|
2664 | 2657 | |
|
2665 | 2658 | transaction.add(self.indexfile, end) |
|
2666 | 2659 | |
|
2667 | 2660 | # then reset internal state in memory to forget those revisions |
|
2668 | 2661 | self._revisioncache = None |
|
2669 | 2662 | self._chaininfocache = util.lrucachedict(500) |
|
2670 | 2663 | self._chunkclear() |
|
2671 | 2664 | |
|
2672 | 2665 | del self.index[rev:-1] |
|
2673 | 2666 | |
|
2674 | 2667 | def checksize(self): |
|
2675 | 2668 | """Check size of index and data files |
|
2676 | 2669 | |
|
2677 | 2670 | return a (dd, di) tuple. |
|
2678 | 2671 | - dd: extra bytes for the "data" file |
|
2679 | 2672 | - di: extra bytes for the "index" file |
|
2680 | 2673 | |
|
2681 | 2674 | A healthy revlog will return (0, 0). |
|
2682 | 2675 | """ |
|
2683 | 2676 | expected = 0 |
|
2684 | 2677 | if len(self): |
|
2685 | 2678 | expected = max(0, self.end(len(self) - 1)) |
|
2686 | 2679 | |
|
2687 | 2680 | try: |
|
2688 | 2681 | with self._datafp() as f: |
|
2689 | 2682 | f.seek(0, io.SEEK_END) |
|
2690 | 2683 | actual = f.tell() |
|
2691 | 2684 | dd = actual - expected |
|
2692 | 2685 | except IOError as inst: |
|
2693 | 2686 | if inst.errno != errno.ENOENT: |
|
2694 | 2687 | raise |
|
2695 | 2688 | dd = 0 |
|
2696 | 2689 | |
|
2697 | 2690 | try: |
|
2698 | 2691 | f = self.opener(self.indexfile) |
|
2699 | 2692 | f.seek(0, io.SEEK_END) |
|
2700 | 2693 | actual = f.tell() |
|
2701 | 2694 | f.close() |
|
2702 |
s = self. |
|
|
2695 | s = self.index.entry_size | |
|
2703 | 2696 | i = max(0, actual // s) |
|
2704 | 2697 | di = actual - (i * s) |
|
2705 | 2698 | if self._inline: |
|
2706 | 2699 | databytes = 0 |
|
2707 | 2700 | for r in self: |
|
2708 | 2701 | databytes += max(0, self.length(r)) |
|
2709 | 2702 | dd = 0 |
|
2710 | 2703 | di = actual - len(self) * s - databytes |
|
2711 | 2704 | except IOError as inst: |
|
2712 | 2705 | if inst.errno != errno.ENOENT: |
|
2713 | 2706 | raise |
|
2714 | 2707 | di = 0 |
|
2715 | 2708 | |
|
2716 | 2709 | return (dd, di) |
|
2717 | 2710 | |
|
2718 | 2711 | def files(self): |
|
2719 | 2712 | res = [self.indexfile] |
|
2720 | 2713 | if not self._inline: |
|
2721 | 2714 | res.append(self.datafile) |
|
2722 | 2715 | return res |
|
2723 | 2716 | |
|
2724 | 2717 | def emitrevisions( |
|
2725 | 2718 | self, |
|
2726 | 2719 | nodes, |
|
2727 | 2720 | nodesorder=None, |
|
2728 | 2721 | revisiondata=False, |
|
2729 | 2722 | assumehaveparentrevisions=False, |
|
2730 | 2723 | deltamode=repository.CG_DELTAMODE_STD, |
|
2731 | 2724 | sidedata_helpers=None, |
|
2732 | 2725 | ): |
|
2733 | 2726 | if nodesorder not in (b'nodes', b'storage', b'linear', None): |
|
2734 | 2727 | raise error.ProgrammingError( |
|
2735 | 2728 | b'unhandled value for nodesorder: %s' % nodesorder |
|
2736 | 2729 | ) |
|
2737 | 2730 | |
|
2738 | 2731 | if nodesorder is None and not self._generaldelta: |
|
2739 | 2732 | nodesorder = b'storage' |
|
2740 | 2733 | |
|
2741 | 2734 | if ( |
|
2742 | 2735 | not self._storedeltachains |
|
2743 | 2736 | and deltamode != repository.CG_DELTAMODE_PREV |
|
2744 | 2737 | ): |
|
2745 | 2738 | deltamode = repository.CG_DELTAMODE_FULL |
|
2746 | 2739 | |
|
2747 | 2740 | return storageutil.emitrevisions( |
|
2748 | 2741 | self, |
|
2749 | 2742 | nodes, |
|
2750 | 2743 | nodesorder, |
|
2751 | 2744 | revlogrevisiondelta, |
|
2752 | 2745 | deltaparentfn=self.deltaparent, |
|
2753 | 2746 | candeltafn=self.candelta, |
|
2754 | 2747 | rawsizefn=self.rawsize, |
|
2755 | 2748 | revdifffn=self.revdiff, |
|
2756 | 2749 | flagsfn=self.flags, |
|
2757 | 2750 | deltamode=deltamode, |
|
2758 | 2751 | revisiondata=revisiondata, |
|
2759 | 2752 | assumehaveparentrevisions=assumehaveparentrevisions, |
|
2760 | 2753 | sidedata_helpers=sidedata_helpers, |
|
2761 | 2754 | ) |
|
2762 | 2755 | |
|
2763 | 2756 | DELTAREUSEALWAYS = b'always' |
|
2764 | 2757 | DELTAREUSESAMEREVS = b'samerevs' |
|
2765 | 2758 | DELTAREUSENEVER = b'never' |
|
2766 | 2759 | |
|
2767 | 2760 | DELTAREUSEFULLADD = b'fulladd' |
|
2768 | 2761 | |
|
2769 | 2762 | DELTAREUSEALL = {b'always', b'samerevs', b'never', b'fulladd'} |
|
2770 | 2763 | |
|
2771 | 2764 | def clone( |
|
2772 | 2765 | self, |
|
2773 | 2766 | tr, |
|
2774 | 2767 | destrevlog, |
|
2775 | 2768 | addrevisioncb=None, |
|
2776 | 2769 | deltareuse=DELTAREUSESAMEREVS, |
|
2777 | 2770 | forcedeltabothparents=None, |
|
2778 | 2771 | sidedatacompanion=None, |
|
2779 | 2772 | ): |
|
2780 | 2773 | """Copy this revlog to another, possibly with format changes. |
|
2781 | 2774 | |
|
2782 | 2775 | The destination revlog will contain the same revisions and nodes. |
|
2783 | 2776 | However, it may not be bit-for-bit identical due to e.g. delta encoding |
|
2784 | 2777 | differences. |
|
2785 | 2778 | |
|
2786 | 2779 | The ``deltareuse`` argument control how deltas from the existing revlog |
|
2787 | 2780 | are preserved in the destination revlog. The argument can have the |
|
2788 | 2781 | following values: |
|
2789 | 2782 | |
|
2790 | 2783 | DELTAREUSEALWAYS |
|
2791 | 2784 | Deltas will always be reused (if possible), even if the destination |
|
2792 | 2785 | revlog would not select the same revisions for the delta. This is the |
|
2793 | 2786 | fastest mode of operation. |
|
2794 | 2787 | DELTAREUSESAMEREVS |
|
2795 | 2788 | Deltas will be reused if the destination revlog would pick the same |
|
2796 | 2789 | revisions for the delta. This mode strikes a balance between speed |
|
2797 | 2790 | and optimization. |
|
2798 | 2791 | DELTAREUSENEVER |
|
2799 | 2792 | Deltas will never be reused. This is the slowest mode of execution. |
|
2800 | 2793 | This mode can be used to recompute deltas (e.g. if the diff/delta |
|
2801 | 2794 | algorithm changes). |
|
2802 | 2795 | DELTAREUSEFULLADD |
|
2803 | 2796 | Revision will be re-added as if their were new content. This is |
|
2804 | 2797 | slower than DELTAREUSEALWAYS but allow more mechanism to kicks in. |
|
2805 | 2798 | eg: large file detection and handling. |
|
2806 | 2799 | |
|
2807 | 2800 | Delta computation can be slow, so the choice of delta reuse policy can |
|
2808 | 2801 | significantly affect run time. |
|
2809 | 2802 | |
|
2810 | 2803 | The default policy (``DELTAREUSESAMEREVS``) strikes a balance between |
|
2811 | 2804 | two extremes. Deltas will be reused if they are appropriate. But if the |
|
2812 | 2805 | delta could choose a better revision, it will do so. This means if you |
|
2813 | 2806 | are converting a non-generaldelta revlog to a generaldelta revlog, |
|
2814 | 2807 | deltas will be recomputed if the delta's parent isn't a parent of the |
|
2815 | 2808 | revision. |
|
2816 | 2809 | |
|
2817 | 2810 | In addition to the delta policy, the ``forcedeltabothparents`` |
|
2818 | 2811 | argument controls whether to force compute deltas against both parents |
|
2819 | 2812 | for merges. By default, the current default is used. |
|
2820 | 2813 | |
|
2821 | 2814 | If not None, the `sidedatacompanion` is callable that accept two |
|
2822 | 2815 | arguments: |
|
2823 | 2816 | |
|
2824 | 2817 | (srcrevlog, rev) |
|
2825 | 2818 | |
|
2826 | 2819 | and return a quintet that control changes to sidedata content from the |
|
2827 | 2820 | old revision to the new clone result: |
|
2828 | 2821 | |
|
2829 | 2822 | (dropall, filterout, update, new_flags, dropped_flags) |
|
2830 | 2823 | |
|
2831 | 2824 | * if `dropall` is True, all sidedata should be dropped |
|
2832 | 2825 | * `filterout` is a set of sidedata keys that should be dropped |
|
2833 | 2826 | * `update` is a mapping of additionnal/new key -> value |
|
2834 | 2827 | * new_flags is a bitfields of new flags that the revision should get |
|
2835 | 2828 | * dropped_flags is a bitfields of new flags that the revision shoudl not longer have |
|
2836 | 2829 | """ |
|
2837 | 2830 | if deltareuse not in self.DELTAREUSEALL: |
|
2838 | 2831 | raise ValueError( |
|
2839 | 2832 | _(b'value for deltareuse invalid: %s') % deltareuse |
|
2840 | 2833 | ) |
|
2841 | 2834 | |
|
2842 | 2835 | if len(destrevlog): |
|
2843 | 2836 | raise ValueError(_(b'destination revlog is not empty')) |
|
2844 | 2837 | |
|
2845 | 2838 | if getattr(self, 'filteredrevs', None): |
|
2846 | 2839 | raise ValueError(_(b'source revlog has filtered revisions')) |
|
2847 | 2840 | if getattr(destrevlog, 'filteredrevs', None): |
|
2848 | 2841 | raise ValueError(_(b'destination revlog has filtered revisions')) |
|
2849 | 2842 | |
|
2850 | 2843 | # lazydelta and lazydeltabase controls whether to reuse a cached delta, |
|
2851 | 2844 | # if possible. |
|
2852 | 2845 | oldlazydelta = destrevlog._lazydelta |
|
2853 | 2846 | oldlazydeltabase = destrevlog._lazydeltabase |
|
2854 | 2847 | oldamd = destrevlog._deltabothparents |
|
2855 | 2848 | |
|
2856 | 2849 | try: |
|
2857 | 2850 | if deltareuse == self.DELTAREUSEALWAYS: |
|
2858 | 2851 | destrevlog._lazydeltabase = True |
|
2859 | 2852 | destrevlog._lazydelta = True |
|
2860 | 2853 | elif deltareuse == self.DELTAREUSESAMEREVS: |
|
2861 | 2854 | destrevlog._lazydeltabase = False |
|
2862 | 2855 | destrevlog._lazydelta = True |
|
2863 | 2856 | elif deltareuse == self.DELTAREUSENEVER: |
|
2864 | 2857 | destrevlog._lazydeltabase = False |
|
2865 | 2858 | destrevlog._lazydelta = False |
|
2866 | 2859 | |
|
2867 | 2860 | destrevlog._deltabothparents = forcedeltabothparents or oldamd |
|
2868 | 2861 | |
|
2869 | 2862 | self._clone( |
|
2870 | 2863 | tr, |
|
2871 | 2864 | destrevlog, |
|
2872 | 2865 | addrevisioncb, |
|
2873 | 2866 | deltareuse, |
|
2874 | 2867 | forcedeltabothparents, |
|
2875 | 2868 | sidedatacompanion, |
|
2876 | 2869 | ) |
|
2877 | 2870 | |
|
2878 | 2871 | finally: |
|
2879 | 2872 | destrevlog._lazydelta = oldlazydelta |
|
2880 | 2873 | destrevlog._lazydeltabase = oldlazydeltabase |
|
2881 | 2874 | destrevlog._deltabothparents = oldamd |
|
2882 | 2875 | |
|
2883 | 2876 | def _clone( |
|
2884 | 2877 | self, |
|
2885 | 2878 | tr, |
|
2886 | 2879 | destrevlog, |
|
2887 | 2880 | addrevisioncb, |
|
2888 | 2881 | deltareuse, |
|
2889 | 2882 | forcedeltabothparents, |
|
2890 | 2883 | sidedatacompanion, |
|
2891 | 2884 | ): |
|
2892 | 2885 | """perform the core duty of `revlog.clone` after parameter processing""" |
|
2893 | 2886 | deltacomputer = deltautil.deltacomputer(destrevlog) |
|
2894 | 2887 | index = self.index |
|
2895 | 2888 | for rev in self: |
|
2896 | 2889 | entry = index[rev] |
|
2897 | 2890 | |
|
2898 | 2891 | # Some classes override linkrev to take filtered revs into |
|
2899 | 2892 | # account. Use raw entry from index. |
|
2900 | 2893 | flags = entry[0] & 0xFFFF |
|
2901 | 2894 | linkrev = entry[4] |
|
2902 | 2895 | p1 = index[entry[5]][7] |
|
2903 | 2896 | p2 = index[entry[6]][7] |
|
2904 | 2897 | node = entry[7] |
|
2905 | 2898 | |
|
2906 | 2899 | sidedataactions = (False, [], {}, 0, 0) |
|
2907 | 2900 | if sidedatacompanion is not None: |
|
2908 | 2901 | sidedataactions = sidedatacompanion(self, rev) |
|
2909 | 2902 | |
|
2910 | 2903 | # (Possibly) reuse the delta from the revlog if allowed and |
|
2911 | 2904 | # the revlog chunk is a delta. |
|
2912 | 2905 | cachedelta = None |
|
2913 | 2906 | rawtext = None |
|
2914 | 2907 | if any(sidedataactions) or deltareuse == self.DELTAREUSEFULLADD: |
|
2915 | 2908 | dropall = sidedataactions[0] |
|
2916 | 2909 | filterout = sidedataactions[1] |
|
2917 | 2910 | update = sidedataactions[2] |
|
2918 | 2911 | new_flags = sidedataactions[3] |
|
2919 | 2912 | dropped_flags = sidedataactions[4] |
|
2920 | 2913 | text, sidedata = self._revisiondata(rev) |
|
2921 | 2914 | if dropall: |
|
2922 | 2915 | sidedata = {} |
|
2923 | 2916 | for key in filterout: |
|
2924 | 2917 | sidedata.pop(key, None) |
|
2925 | 2918 | sidedata.update(update) |
|
2926 | 2919 | if not sidedata: |
|
2927 | 2920 | sidedata = None |
|
2928 | 2921 | |
|
2929 | 2922 | flags |= new_flags |
|
2930 | 2923 | flags &= ~dropped_flags |
|
2931 | 2924 | |
|
2932 | 2925 | destrevlog.addrevision( |
|
2933 | 2926 | text, |
|
2934 | 2927 | tr, |
|
2935 | 2928 | linkrev, |
|
2936 | 2929 | p1, |
|
2937 | 2930 | p2, |
|
2938 | 2931 | cachedelta=cachedelta, |
|
2939 | 2932 | node=node, |
|
2940 | 2933 | flags=flags, |
|
2941 | 2934 | deltacomputer=deltacomputer, |
|
2942 | 2935 | sidedata=sidedata, |
|
2943 | 2936 | ) |
|
2944 | 2937 | else: |
|
2945 | 2938 | if destrevlog._lazydelta: |
|
2946 | 2939 | dp = self.deltaparent(rev) |
|
2947 | 2940 | if dp != nullrev: |
|
2948 | 2941 | cachedelta = (dp, bytes(self._chunk(rev))) |
|
2949 | 2942 | |
|
2950 | 2943 | if not cachedelta: |
|
2951 | 2944 | rawtext = self.rawdata(rev) |
|
2952 | 2945 | |
|
2953 | 2946 | ifh = destrevlog.opener( |
|
2954 | 2947 | destrevlog.indexfile, b'a+', checkambig=False |
|
2955 | 2948 | ) |
|
2956 | 2949 | dfh = None |
|
2957 | 2950 | if not destrevlog._inline: |
|
2958 | 2951 | dfh = destrevlog.opener(destrevlog.datafile, b'a+') |
|
2959 | 2952 | try: |
|
2960 | 2953 | destrevlog._addrevision( |
|
2961 | 2954 | node, |
|
2962 | 2955 | rawtext, |
|
2963 | 2956 | tr, |
|
2964 | 2957 | linkrev, |
|
2965 | 2958 | p1, |
|
2966 | 2959 | p2, |
|
2967 | 2960 | flags, |
|
2968 | 2961 | cachedelta, |
|
2969 | 2962 | ifh, |
|
2970 | 2963 | dfh, |
|
2971 | 2964 | deltacomputer=deltacomputer, |
|
2972 | 2965 | ) |
|
2973 | 2966 | finally: |
|
2974 | 2967 | if dfh: |
|
2975 | 2968 | dfh.close() |
|
2976 | 2969 | ifh.close() |
|
2977 | 2970 | |
|
2978 | 2971 | if addrevisioncb: |
|
2979 | 2972 | addrevisioncb(self, rev, node) |
|
2980 | 2973 | |
|
2981 | 2974 | def censorrevision(self, tr, censornode, tombstone=b''): |
|
2982 | 2975 | if (self.version & 0xFFFF) == REVLOGV0: |
|
2983 | 2976 | raise error.RevlogError( |
|
2984 | 2977 | _(b'cannot censor with version %d revlogs') % self.version |
|
2985 | 2978 | ) |
|
2986 | 2979 | |
|
2987 | 2980 | censorrev = self.rev(censornode) |
|
2988 | 2981 | tombstone = storageutil.packmeta({b'censored': tombstone}, b'') |
|
2989 | 2982 | |
|
2990 | 2983 | if len(tombstone) > self.rawsize(censorrev): |
|
2991 | 2984 | raise error.Abort( |
|
2992 | 2985 | _(b'censor tombstone must be no longer than censored data') |
|
2993 | 2986 | ) |
|
2994 | 2987 | |
|
2995 | 2988 | # Rewriting the revlog in place is hard. Our strategy for censoring is |
|
2996 | 2989 | # to create a new revlog, copy all revisions to it, then replace the |
|
2997 | 2990 | # revlogs on transaction close. |
|
2998 | 2991 | |
|
2999 | 2992 | newindexfile = self.indexfile + b'.tmpcensored' |
|
3000 | 2993 | newdatafile = self.datafile + b'.tmpcensored' |
|
3001 | 2994 | |
|
3002 | 2995 | # This is a bit dangerous. We could easily have a mismatch of state. |
|
3003 | 2996 | newrl = revlog(self.opener, newindexfile, newdatafile, censorable=True) |
|
3004 | 2997 | newrl.version = self.version |
|
3005 | 2998 | newrl._generaldelta = self._generaldelta |
|
3006 | 2999 | newrl._io = self._io |
|
3007 | 3000 | |
|
3008 | 3001 | for rev in self.revs(): |
|
3009 | 3002 | node = self.node(rev) |
|
3010 | 3003 | p1, p2 = self.parents(node) |
|
3011 | 3004 | |
|
3012 | 3005 | if rev == censorrev: |
|
3013 | 3006 | newrl.addrawrevision( |
|
3014 | 3007 | tombstone, |
|
3015 | 3008 | tr, |
|
3016 | 3009 | self.linkrev(censorrev), |
|
3017 | 3010 | p1, |
|
3018 | 3011 | p2, |
|
3019 | 3012 | censornode, |
|
3020 | 3013 | REVIDX_ISCENSORED, |
|
3021 | 3014 | ) |
|
3022 | 3015 | |
|
3023 | 3016 | if newrl.deltaparent(rev) != nullrev: |
|
3024 | 3017 | raise error.Abort( |
|
3025 | 3018 | _( |
|
3026 | 3019 | b'censored revision stored as delta; ' |
|
3027 | 3020 | b'cannot censor' |
|
3028 | 3021 | ), |
|
3029 | 3022 | hint=_( |
|
3030 | 3023 | b'censoring of revlogs is not ' |
|
3031 | 3024 | b'fully implemented; please report ' |
|
3032 | 3025 | b'this bug' |
|
3033 | 3026 | ), |
|
3034 | 3027 | ) |
|
3035 | 3028 | continue |
|
3036 | 3029 | |
|
3037 | 3030 | if self.iscensored(rev): |
|
3038 | 3031 | if self.deltaparent(rev) != nullrev: |
|
3039 | 3032 | raise error.Abort( |
|
3040 | 3033 | _( |
|
3041 | 3034 | b'cannot censor due to censored ' |
|
3042 | 3035 | b'revision having delta stored' |
|
3043 | 3036 | ) |
|
3044 | 3037 | ) |
|
3045 | 3038 | rawtext = self._chunk(rev) |
|
3046 | 3039 | else: |
|
3047 | 3040 | rawtext = self.rawdata(rev) |
|
3048 | 3041 | |
|
3049 | 3042 | newrl.addrawrevision( |
|
3050 | 3043 | rawtext, tr, self.linkrev(rev), p1, p2, node, self.flags(rev) |
|
3051 | 3044 | ) |
|
3052 | 3045 | |
|
3053 | 3046 | tr.addbackup(self.indexfile, location=b'store') |
|
3054 | 3047 | if not self._inline: |
|
3055 | 3048 | tr.addbackup(self.datafile, location=b'store') |
|
3056 | 3049 | |
|
3057 | 3050 | self.opener.rename(newrl.indexfile, self.indexfile) |
|
3058 | 3051 | if not self._inline: |
|
3059 | 3052 | self.opener.rename(newrl.datafile, self.datafile) |
|
3060 | 3053 | |
|
3061 | 3054 | self.clearcaches() |
|
3062 | 3055 | self._loadindex() |
|
3063 | 3056 | |
|
3064 | 3057 | def verifyintegrity(self, state): |
|
3065 | 3058 | """Verifies the integrity of the revlog. |
|
3066 | 3059 | |
|
3067 | 3060 | Yields ``revlogproblem`` instances describing problems that are |
|
3068 | 3061 | found. |
|
3069 | 3062 | """ |
|
3070 | 3063 | dd, di = self.checksize() |
|
3071 | 3064 | if dd: |
|
3072 | 3065 | yield revlogproblem(error=_(b'data length off by %d bytes') % dd) |
|
3073 | 3066 | if di: |
|
3074 | 3067 | yield revlogproblem(error=_(b'index contains %d extra bytes') % di) |
|
3075 | 3068 | |
|
3076 | 3069 | version = self.version & 0xFFFF |
|
3077 | 3070 | |
|
3078 | 3071 | # The verifier tells us what version revlog we should be. |
|
3079 | 3072 | if version != state[b'expectedversion']: |
|
3080 | 3073 | yield revlogproblem( |
|
3081 | 3074 | warning=_(b"warning: '%s' uses revlog format %d; expected %d") |
|
3082 | 3075 | % (self.indexfile, version, state[b'expectedversion']) |
|
3083 | 3076 | ) |
|
3084 | 3077 | |
|
3085 | 3078 | state[b'skipread'] = set() |
|
3086 | 3079 | state[b'safe_renamed'] = set() |
|
3087 | 3080 | |
|
3088 | 3081 | for rev in self: |
|
3089 | 3082 | node = self.node(rev) |
|
3090 | 3083 | |
|
3091 | 3084 | # Verify contents. 4 cases to care about: |
|
3092 | 3085 | # |
|
3093 | 3086 | # common: the most common case |
|
3094 | 3087 | # rename: with a rename |
|
3095 | 3088 | # meta: file content starts with b'\1\n', the metadata |
|
3096 | 3089 | # header defined in filelog.py, but without a rename |
|
3097 | 3090 | # ext: content stored externally |
|
3098 | 3091 | # |
|
3099 | 3092 | # More formally, their differences are shown below: |
|
3100 | 3093 | # |
|
3101 | 3094 | # | common | rename | meta | ext |
|
3102 | 3095 | # ------------------------------------------------------- |
|
3103 | 3096 | # flags() | 0 | 0 | 0 | not 0 |
|
3104 | 3097 | # renamed() | False | True | False | ? |
|
3105 | 3098 | # rawtext[0:2]=='\1\n'| False | True | True | ? |
|
3106 | 3099 | # |
|
3107 | 3100 | # "rawtext" means the raw text stored in revlog data, which |
|
3108 | 3101 | # could be retrieved by "rawdata(rev)". "text" |
|
3109 | 3102 | # mentioned below is "revision(rev)". |
|
3110 | 3103 | # |
|
3111 | 3104 | # There are 3 different lengths stored physically: |
|
3112 | 3105 | # 1. L1: rawsize, stored in revlog index |
|
3113 | 3106 | # 2. L2: len(rawtext), stored in revlog data |
|
3114 | 3107 | # 3. L3: len(text), stored in revlog data if flags==0, or |
|
3115 | 3108 | # possibly somewhere else if flags!=0 |
|
3116 | 3109 | # |
|
3117 | 3110 | # L1 should be equal to L2. L3 could be different from them. |
|
3118 | 3111 | # "text" may or may not affect commit hash depending on flag |
|
3119 | 3112 | # processors (see flagutil.addflagprocessor). |
|
3120 | 3113 | # |
|
3121 | 3114 | # | common | rename | meta | ext |
|
3122 | 3115 | # ------------------------------------------------- |
|
3123 | 3116 | # rawsize() | L1 | L1 | L1 | L1 |
|
3124 | 3117 | # size() | L1 | L2-LM | L1(*) | L1 (?) |
|
3125 | 3118 | # len(rawtext) | L2 | L2 | L2 | L2 |
|
3126 | 3119 | # len(text) | L2 | L2 | L2 | L3 |
|
3127 | 3120 | # len(read()) | L2 | L2-LM | L2-LM | L3 (?) |
|
3128 | 3121 | # |
|
3129 | 3122 | # LM: length of metadata, depending on rawtext |
|
3130 | 3123 | # (*): not ideal, see comment in filelog.size |
|
3131 | 3124 | # (?): could be "- len(meta)" if the resolved content has |
|
3132 | 3125 | # rename metadata |
|
3133 | 3126 | # |
|
3134 | 3127 | # Checks needed to be done: |
|
3135 | 3128 | # 1. length check: L1 == L2, in all cases. |
|
3136 | 3129 | # 2. hash check: depending on flag processor, we may need to |
|
3137 | 3130 | # use either "text" (external), or "rawtext" (in revlog). |
|
3138 | 3131 | |
|
3139 | 3132 | try: |
|
3140 | 3133 | skipflags = state.get(b'skipflags', 0) |
|
3141 | 3134 | if skipflags: |
|
3142 | 3135 | skipflags &= self.flags(rev) |
|
3143 | 3136 | |
|
3144 | 3137 | _verify_revision(self, skipflags, state, node) |
|
3145 | 3138 | |
|
3146 | 3139 | l1 = self.rawsize(rev) |
|
3147 | 3140 | l2 = len(self.rawdata(node)) |
|
3148 | 3141 | |
|
3149 | 3142 | if l1 != l2: |
|
3150 | 3143 | yield revlogproblem( |
|
3151 | 3144 | error=_(b'unpacked size is %d, %d expected') % (l2, l1), |
|
3152 | 3145 | node=node, |
|
3153 | 3146 | ) |
|
3154 | 3147 | |
|
3155 | 3148 | except error.CensoredNodeError: |
|
3156 | 3149 | if state[b'erroroncensored']: |
|
3157 | 3150 | yield revlogproblem( |
|
3158 | 3151 | error=_(b'censored file data'), node=node |
|
3159 | 3152 | ) |
|
3160 | 3153 | state[b'skipread'].add(node) |
|
3161 | 3154 | except Exception as e: |
|
3162 | 3155 | yield revlogproblem( |
|
3163 | 3156 | error=_(b'unpacking %s: %s') |
|
3164 | 3157 | % (short(node), stringutil.forcebytestr(e)), |
|
3165 | 3158 | node=node, |
|
3166 | 3159 | ) |
|
3167 | 3160 | state[b'skipread'].add(node) |
|
3168 | 3161 | |
|
3169 | 3162 | def storageinfo( |
|
3170 | 3163 | self, |
|
3171 | 3164 | exclusivefiles=False, |
|
3172 | 3165 | sharedfiles=False, |
|
3173 | 3166 | revisionscount=False, |
|
3174 | 3167 | trackedsize=False, |
|
3175 | 3168 | storedsize=False, |
|
3176 | 3169 | ): |
|
3177 | 3170 | d = {} |
|
3178 | 3171 | |
|
3179 | 3172 | if exclusivefiles: |
|
3180 | 3173 | d[b'exclusivefiles'] = [(self.opener, self.indexfile)] |
|
3181 | 3174 | if not self._inline: |
|
3182 | 3175 | d[b'exclusivefiles'].append((self.opener, self.datafile)) |
|
3183 | 3176 | |
|
3184 | 3177 | if sharedfiles: |
|
3185 | 3178 | d[b'sharedfiles'] = [] |
|
3186 | 3179 | |
|
3187 | 3180 | if revisionscount: |
|
3188 | 3181 | d[b'revisionscount'] = len(self) |
|
3189 | 3182 | |
|
3190 | 3183 | if trackedsize: |
|
3191 | 3184 | d[b'trackedsize'] = sum(map(self.rawsize, iter(self))) |
|
3192 | 3185 | |
|
3193 | 3186 | if storedsize: |
|
3194 | 3187 | d[b'storedsize'] = sum( |
|
3195 | 3188 | self.opener.stat(path).st_size for path in self.files() |
|
3196 | 3189 | ) |
|
3197 | 3190 | |
|
3198 | 3191 | return d |
|
3199 | 3192 | |
|
3200 | 3193 | def rewrite_sidedata(self, helpers, startrev, endrev): |
|
3201 | 3194 | if self.version & 0xFFFF != REVLOGV2: |
|
3202 | 3195 | return |
|
3203 | 3196 | # inline are not yet supported because they suffer from an issue when |
|
3204 | 3197 | # rewriting them (since it's not an append-only operation). |
|
3205 | 3198 | # See issue6485. |
|
3206 | 3199 | assert not self._inline |
|
3207 | 3200 | if not helpers[1] and not helpers[2]: |
|
3208 | 3201 | # Nothing to generate or remove |
|
3209 | 3202 | return |
|
3210 | 3203 | |
|
3211 | 3204 | new_entries = [] |
|
3212 | 3205 | # append the new sidedata |
|
3213 | 3206 | with self._datafp(b'a+') as fp: |
|
3214 | 3207 | # Maybe this bug still exists, see revlog._writeentry |
|
3215 | 3208 | fp.seek(0, os.SEEK_END) |
|
3216 | 3209 | current_offset = fp.tell() |
|
3217 | 3210 | for rev in range(startrev, endrev + 1): |
|
3218 | 3211 | entry = self.index[rev] |
|
3219 | 3212 | new_sidedata = storageutil.run_sidedata_helpers( |
|
3220 | 3213 | store=self, |
|
3221 | 3214 | sidedata_helpers=helpers, |
|
3222 | 3215 | sidedata={}, |
|
3223 | 3216 | rev=rev, |
|
3224 | 3217 | ) |
|
3225 | 3218 | |
|
3226 | 3219 | serialized_sidedata = sidedatautil.serialize_sidedata( |
|
3227 | 3220 | new_sidedata |
|
3228 | 3221 | ) |
|
3229 | 3222 | if entry[8] != 0 or entry[9] != 0: |
|
3230 | 3223 | # rewriting entries that already have sidedata is not |
|
3231 | 3224 | # supported yet, because it introduces garbage data in the |
|
3232 | 3225 | # revlog. |
|
3233 | 3226 | msg = b"Rewriting existing sidedata is not supported yet" |
|
3234 | 3227 | raise error.Abort(msg) |
|
3235 | 3228 | entry = entry[:8] |
|
3236 | 3229 | entry += (current_offset, len(serialized_sidedata)) |
|
3237 | 3230 | |
|
3238 | 3231 | fp.write(serialized_sidedata) |
|
3239 | 3232 | new_entries.append(entry) |
|
3240 | 3233 | current_offset += len(serialized_sidedata) |
|
3241 | 3234 | |
|
3242 | 3235 | # rewrite the new index entries |
|
3243 | 3236 | with self._indexfp(b'w+') as fp: |
|
3244 |
fp.seek(startrev * self. |
|
|
3237 | fp.seek(startrev * self.index.entry_size) | |
|
3245 | 3238 | for i, entry in enumerate(new_entries): |
|
3246 | 3239 | rev = startrev + i |
|
3247 | 3240 | self.index.replace_sidedata_info(rev, entry[8], entry[9]) |
|
3248 | 3241 | packed = self._io.packentry(entry, self.node, self.version, rev) |
|
3249 | 3242 | fp.write(packed) |
@@ -1,490 +1,494 b'' | |||
|
1 | 1 | // revlog.rs |
|
2 | 2 | // |
|
3 | 3 | // Copyright 2019-2020 Georges Racinet <georges.racinet@octobus.net> |
|
4 | 4 | // |
|
5 | 5 | // This software may be used and distributed according to the terms of the |
|
6 | 6 | // GNU General Public License version 2 or any later version. |
|
7 | 7 | |
|
8 | 8 | use crate::{ |
|
9 | 9 | cindex, |
|
10 | 10 | utils::{node_from_py_bytes, node_from_py_object}, |
|
11 | 11 | }; |
|
12 | 12 | use cpython::{ |
|
13 | 13 | buffer::{Element, PyBuffer}, |
|
14 | 14 | exc::{IndexError, ValueError}, |
|
15 |
ObjectProtocol, PyBytes, PyClone, PyDict, PyErr, PyModule, |
|
|
16 | PyResult, PyString, PyTuple, Python, PythonObject, ToPyObject, | |
|
15 | ObjectProtocol, PyBytes, PyClone, PyDict, PyErr, PyInt, PyModule, | |
|
16 | PyObject, PyResult, PyString, PyTuple, Python, PythonObject, ToPyObject, | |
|
17 | 17 | }; |
|
18 | 18 | use hg::{ |
|
19 | 19 | nodemap::{Block, NodeMapError, NodeTree}, |
|
20 | 20 | revlog::{nodemap::NodeMap, NodePrefix, RevlogIndex}, |
|
21 | 21 | Revision, |
|
22 | 22 | }; |
|
23 | 23 | use std::cell::RefCell; |
|
24 | 24 | |
|
25 | 25 | /// Return a Struct implementing the Graph trait |
|
26 | 26 | pub(crate) fn pyindex_to_graph( |
|
27 | 27 | py: Python, |
|
28 | 28 | index: PyObject, |
|
29 | 29 | ) -> PyResult<cindex::Index> { |
|
30 | 30 | match index.extract::<MixedIndex>(py) { |
|
31 | 31 | Ok(midx) => Ok(midx.clone_cindex(py)), |
|
32 | 32 | Err(_) => cindex::Index::new(py, index), |
|
33 | 33 | } |
|
34 | 34 | } |
|
35 | 35 | |
|
36 | 36 | py_class!(pub class MixedIndex |py| { |
|
37 | 37 | data cindex: RefCell<cindex::Index>; |
|
38 | 38 | data nt: RefCell<Option<NodeTree>>; |
|
39 | 39 | data docket: RefCell<Option<PyObject>>; |
|
40 | 40 | // Holds a reference to the mmap'ed persistent nodemap data |
|
41 | 41 | data mmap: RefCell<Option<PyBuffer>>; |
|
42 | 42 | |
|
43 | 43 | def __new__(_cls, cindex: PyObject) -> PyResult<MixedIndex> { |
|
44 | 44 | Self::new(py, cindex) |
|
45 | 45 | } |
|
46 | 46 | |
|
47 | 47 | /// Compatibility layer used for Python consumers needing access to the C index |
|
48 | 48 | /// |
|
49 | 49 | /// Only use case so far is `scmutil.shortesthexnodeidprefix`, |
|
50 | 50 | /// that may need to build a custom `nodetree`, based on a specified revset. |
|
51 | 51 | /// With a Rust implementation of the nodemap, we will be able to get rid of |
|
52 | 52 | /// this, by exposing our own standalone nodemap class, |
|
53 | 53 | /// ready to accept `MixedIndex`. |
|
54 | 54 | def get_cindex(&self) -> PyResult<PyObject> { |
|
55 | 55 | Ok(self.cindex(py).borrow().inner().clone_ref(py)) |
|
56 | 56 | } |
|
57 | 57 | |
|
58 | 58 | // Index API involving nodemap, as defined in mercurial/pure/parsers.py |
|
59 | 59 | |
|
60 | 60 | /// Return Revision if found, raises a bare `error.RevlogError` |
|
61 | 61 | /// in case of ambiguity, same as C version does |
|
62 | 62 | def get_rev(&self, node: PyBytes) -> PyResult<Option<Revision>> { |
|
63 | 63 | let opt = self.get_nodetree(py)?.borrow(); |
|
64 | 64 | let nt = opt.as_ref().unwrap(); |
|
65 | 65 | let idx = &*self.cindex(py).borrow(); |
|
66 | 66 | let node = node_from_py_bytes(py, &node)?; |
|
67 | 67 | nt.find_bin(idx, node.into()).map_err(|e| nodemap_error(py, e)) |
|
68 | 68 | } |
|
69 | 69 | |
|
70 | 70 | /// same as `get_rev()` but raises a bare `error.RevlogError` if node |
|
71 | 71 | /// is not found. |
|
72 | 72 | /// |
|
73 | 73 | /// No need to repeat `node` in the exception, `mercurial/revlog.py` |
|
74 | 74 | /// will catch and rewrap with it |
|
75 | 75 | def rev(&self, node: PyBytes) -> PyResult<Revision> { |
|
76 | 76 | self.get_rev(py, node)?.ok_or_else(|| revlog_error(py)) |
|
77 | 77 | } |
|
78 | 78 | |
|
79 | 79 | /// return True if the node exist in the index |
|
80 | 80 | def has_node(&self, node: PyBytes) -> PyResult<bool> { |
|
81 | 81 | self.get_rev(py, node).map(|opt| opt.is_some()) |
|
82 | 82 | } |
|
83 | 83 | |
|
84 | 84 | /// find length of shortest hex nodeid of a binary ID |
|
85 | 85 | def shortest(&self, node: PyBytes) -> PyResult<usize> { |
|
86 | 86 | let opt = self.get_nodetree(py)?.borrow(); |
|
87 | 87 | let nt = opt.as_ref().unwrap(); |
|
88 | 88 | let idx = &*self.cindex(py).borrow(); |
|
89 | 89 | match nt.unique_prefix_len_node(idx, &node_from_py_bytes(py, &node)?) |
|
90 | 90 | { |
|
91 | 91 | Ok(Some(l)) => Ok(l), |
|
92 | 92 | Ok(None) => Err(revlog_error(py)), |
|
93 | 93 | Err(e) => Err(nodemap_error(py, e)), |
|
94 | 94 | } |
|
95 | 95 | } |
|
96 | 96 | |
|
97 | 97 | def partialmatch(&self, node: PyObject) -> PyResult<Option<PyBytes>> { |
|
98 | 98 | let opt = self.get_nodetree(py)?.borrow(); |
|
99 | 99 | let nt = opt.as_ref().unwrap(); |
|
100 | 100 | let idx = &*self.cindex(py).borrow(); |
|
101 | 101 | |
|
102 | 102 | let node_as_string = if cfg!(feature = "python3-sys") { |
|
103 | 103 | node.cast_as::<PyString>(py)?.to_string(py)?.to_string() |
|
104 | 104 | } |
|
105 | 105 | else { |
|
106 | 106 | let node = node.extract::<PyBytes>(py)?; |
|
107 | 107 | String::from_utf8_lossy(node.data(py)).to_string() |
|
108 | 108 | }; |
|
109 | 109 | |
|
110 | 110 | let prefix = NodePrefix::from_hex(&node_as_string).map_err(|_| PyErr::new::<ValueError, _>(py, "Invalid node or prefix"))?; |
|
111 | 111 | |
|
112 | 112 | nt.find_bin(idx, prefix) |
|
113 | 113 | // TODO make an inner API returning the node directly |
|
114 | 114 | .map(|opt| opt.map( |
|
115 | 115 | |rev| PyBytes::new(py, idx.node(rev).unwrap().as_bytes()))) |
|
116 | 116 | .map_err(|e| nodemap_error(py, e)) |
|
117 | 117 | |
|
118 | 118 | } |
|
119 | 119 | |
|
120 | 120 | /// append an index entry |
|
121 | 121 | def append(&self, tup: PyTuple) -> PyResult<PyObject> { |
|
122 | 122 | if tup.len(py) < 8 { |
|
123 | 123 | // this is better than the panic promised by tup.get_item() |
|
124 | 124 | return Err( |
|
125 | 125 | PyErr::new::<IndexError, _>(py, "tuple index out of range")) |
|
126 | 126 | } |
|
127 | 127 | let node_bytes = tup.get_item(py, 7).extract(py)?; |
|
128 | 128 | let node = node_from_py_object(py, &node_bytes)?; |
|
129 | 129 | |
|
130 | 130 | let mut idx = self.cindex(py).borrow_mut(); |
|
131 | 131 | let rev = idx.len() as Revision; |
|
132 | 132 | |
|
133 | 133 | idx.append(py, tup)?; |
|
134 | 134 | self.get_nodetree(py)?.borrow_mut().as_mut().unwrap() |
|
135 | 135 | .insert(&*idx, &node, rev) |
|
136 | 136 | .map_err(|e| nodemap_error(py, e))?; |
|
137 | 137 | Ok(py.None()) |
|
138 | 138 | } |
|
139 | 139 | |
|
140 | 140 | def __delitem__(&self, key: PyObject) -> PyResult<()> { |
|
141 | 141 | // __delitem__ is both for `del idx[r]` and `del idx[r1:r2]` |
|
142 | 142 | self.cindex(py).borrow().inner().del_item(py, key)?; |
|
143 | 143 | let mut opt = self.get_nodetree(py)?.borrow_mut(); |
|
144 | 144 | let mut nt = opt.as_mut().unwrap(); |
|
145 | 145 | nt.invalidate_all(); |
|
146 | 146 | self.fill_nodemap(py, &mut nt)?; |
|
147 | 147 | Ok(()) |
|
148 | 148 | } |
|
149 | 149 | |
|
150 | 150 | // |
|
151 | 151 | // Reforwarded C index API |
|
152 | 152 | // |
|
153 | 153 | |
|
154 | 154 | // index_methods (tp_methods). Same ordering as in revlog.c |
|
155 | 155 | |
|
156 | 156 | /// return the gca set of the given revs |
|
157 | 157 | def ancestors(&self, *args, **kw) -> PyResult<PyObject> { |
|
158 | 158 | self.call_cindex(py, "ancestors", args, kw) |
|
159 | 159 | } |
|
160 | 160 | |
|
161 | 161 | /// return the heads of the common ancestors of the given revs |
|
162 | 162 | def commonancestorsheads(&self, *args, **kw) -> PyResult<PyObject> { |
|
163 | 163 | self.call_cindex(py, "commonancestorsheads", args, kw) |
|
164 | 164 | } |
|
165 | 165 | |
|
166 | 166 | /// Clear the index caches and inner py_class data. |
|
167 | 167 | /// It is Python's responsibility to call `update_nodemap_data` again. |
|
168 | 168 | def clearcaches(&self, *args, **kw) -> PyResult<PyObject> { |
|
169 | 169 | self.nt(py).borrow_mut().take(); |
|
170 | 170 | self.docket(py).borrow_mut().take(); |
|
171 | 171 | self.mmap(py).borrow_mut().take(); |
|
172 | 172 | self.call_cindex(py, "clearcaches", args, kw) |
|
173 | 173 | } |
|
174 | 174 | |
|
175 | 175 | /// get an index entry |
|
176 | 176 | def get(&self, *args, **kw) -> PyResult<PyObject> { |
|
177 | 177 | self.call_cindex(py, "get", args, kw) |
|
178 | 178 | } |
|
179 | 179 | |
|
180 | 180 | /// compute phases |
|
181 | 181 | def computephasesmapsets(&self, *args, **kw) -> PyResult<PyObject> { |
|
182 | 182 | self.call_cindex(py, "computephasesmapsets", args, kw) |
|
183 | 183 | } |
|
184 | 184 | |
|
185 | 185 | /// reachableroots |
|
186 | 186 | def reachableroots2(&self, *args, **kw) -> PyResult<PyObject> { |
|
187 | 187 | self.call_cindex(py, "reachableroots2", args, kw) |
|
188 | 188 | } |
|
189 | 189 | |
|
190 | 190 | /// get head revisions |
|
191 | 191 | def headrevs(&self, *args, **kw) -> PyResult<PyObject> { |
|
192 | 192 | self.call_cindex(py, "headrevs", args, kw) |
|
193 | 193 | } |
|
194 | 194 | |
|
195 | 195 | /// get filtered head revisions |
|
196 | 196 | def headrevsfiltered(&self, *args, **kw) -> PyResult<PyObject> { |
|
197 | 197 | self.call_cindex(py, "headrevsfiltered", args, kw) |
|
198 | 198 | } |
|
199 | 199 | |
|
200 | 200 | /// True if the object is a snapshot |
|
201 | 201 | def issnapshot(&self, *args, **kw) -> PyResult<PyObject> { |
|
202 | 202 | self.call_cindex(py, "issnapshot", args, kw) |
|
203 | 203 | } |
|
204 | 204 | |
|
205 | 205 | /// Gather snapshot data in a cache dict |
|
206 | 206 | def findsnapshots(&self, *args, **kw) -> PyResult<PyObject> { |
|
207 | 207 | self.call_cindex(py, "findsnapshots", args, kw) |
|
208 | 208 | } |
|
209 | 209 | |
|
210 | 210 | /// determine revisions with deltas to reconstruct fulltext |
|
211 | 211 | def deltachain(&self, *args, **kw) -> PyResult<PyObject> { |
|
212 | 212 | self.call_cindex(py, "deltachain", args, kw) |
|
213 | 213 | } |
|
214 | 214 | |
|
215 | 215 | /// slice planned chunk read to reach a density threshold |
|
216 | 216 | def slicechunktodensity(&self, *args, **kw) -> PyResult<PyObject> { |
|
217 | 217 | self.call_cindex(py, "slicechunktodensity", args, kw) |
|
218 | 218 | } |
|
219 | 219 | |
|
220 | 220 | /// stats for the index |
|
221 | 221 | def stats(&self, *args, **kw) -> PyResult<PyObject> { |
|
222 | 222 | self.call_cindex(py, "stats", args, kw) |
|
223 | 223 | } |
|
224 | 224 | |
|
225 | 225 | // index_sequence_methods and index_mapping_methods. |
|
226 | 226 | // |
|
227 | 227 | // Since we call back through the high level Python API, |
|
228 | 228 | // there's no point making a distinction between index_get |
|
229 | 229 | // and index_getitem. |
|
230 | 230 | |
|
231 | 231 | def __len__(&self) -> PyResult<usize> { |
|
232 | 232 | self.cindex(py).borrow().inner().len(py) |
|
233 | 233 | } |
|
234 | 234 | |
|
235 | 235 | def __getitem__(&self, key: PyObject) -> PyResult<PyObject> { |
|
236 | 236 | // this conversion seems needless, but that's actually because |
|
237 | 237 | // `index_getitem` does not handle conversion from PyLong, |
|
238 | 238 | // which expressions such as [e for e in index] internally use. |
|
239 | 239 | // Note that we don't seem to have a direct way to call |
|
240 | 240 | // PySequence_GetItem (does the job), which would possibly be better |
|
241 | 241 | // for performance |
|
242 | 242 | let key = match key.extract::<Revision>(py) { |
|
243 | 243 | Ok(rev) => rev.to_py_object(py).into_object(), |
|
244 | 244 | Err(_) => key, |
|
245 | 245 | }; |
|
246 | 246 | self.cindex(py).borrow().inner().get_item(py, key) |
|
247 | 247 | } |
|
248 | 248 | |
|
249 | 249 | def __setitem__(&self, key: PyObject, value: PyObject) -> PyResult<()> { |
|
250 | 250 | self.cindex(py).borrow().inner().set_item(py, key, value) |
|
251 | 251 | } |
|
252 | 252 | |
|
253 | 253 | def __contains__(&self, item: PyObject) -> PyResult<bool> { |
|
254 | 254 | // ObjectProtocol does not seem to provide contains(), so |
|
255 | 255 | // this is an equivalent implementation of the index_contains() |
|
256 | 256 | // defined in revlog.c |
|
257 | 257 | let cindex = self.cindex(py).borrow(); |
|
258 | 258 | match item.extract::<Revision>(py) { |
|
259 | 259 | Ok(rev) => { |
|
260 | 260 | Ok(rev >= -1 && rev < cindex.inner().len(py)? as Revision) |
|
261 | 261 | } |
|
262 | 262 | Err(_) => { |
|
263 | 263 | cindex.inner().call_method( |
|
264 | 264 | py, |
|
265 | 265 | "has_node", |
|
266 | 266 | PyTuple::new(py, &[item]), |
|
267 | 267 | None)? |
|
268 | 268 | .extract(py) |
|
269 | 269 | } |
|
270 | 270 | } |
|
271 | 271 | } |
|
272 | 272 | |
|
273 | 273 | def nodemap_data_all(&self) -> PyResult<PyBytes> { |
|
274 | 274 | self.inner_nodemap_data_all(py) |
|
275 | 275 | } |
|
276 | 276 | |
|
277 | 277 | def nodemap_data_incremental(&self) -> PyResult<PyObject> { |
|
278 | 278 | self.inner_nodemap_data_incremental(py) |
|
279 | 279 | } |
|
280 | 280 | def update_nodemap_data( |
|
281 | 281 | &self, |
|
282 | 282 | docket: PyObject, |
|
283 | 283 | nm_data: PyObject |
|
284 | 284 | ) -> PyResult<PyObject> { |
|
285 | 285 | self.inner_update_nodemap_data(py, docket, nm_data) |
|
286 | 286 | } |
|
287 | 287 | |
|
288 | @property | |
|
289 | def entry_size(&self) -> PyResult<PyInt> { | |
|
290 | self.cindex(py).borrow().inner().getattr(py, "entry_size")?.extract::<PyInt>(py) | |
|
291 | } | |
|
288 | 292 | |
|
289 | 293 | }); |
|
290 | 294 | |
|
291 | 295 | impl MixedIndex { |
|
292 | 296 | fn new(py: Python, cindex: PyObject) -> PyResult<MixedIndex> { |
|
293 | 297 | Self::create_instance( |
|
294 | 298 | py, |
|
295 | 299 | RefCell::new(cindex::Index::new(py, cindex)?), |
|
296 | 300 | RefCell::new(None), |
|
297 | 301 | RefCell::new(None), |
|
298 | 302 | RefCell::new(None), |
|
299 | 303 | ) |
|
300 | 304 | } |
|
301 | 305 | |
|
302 | 306 | /// This is scaffolding at this point, but it could also become |
|
303 | 307 | /// a way to start a persistent nodemap or perform a |
|
304 | 308 | /// vacuum / repack operation |
|
305 | 309 | fn fill_nodemap( |
|
306 | 310 | &self, |
|
307 | 311 | py: Python, |
|
308 | 312 | nt: &mut NodeTree, |
|
309 | 313 | ) -> PyResult<PyObject> { |
|
310 | 314 | let index = self.cindex(py).borrow(); |
|
311 | 315 | for r in 0..index.len() { |
|
312 | 316 | let rev = r as Revision; |
|
313 | 317 | // in this case node() won't ever return None |
|
314 | 318 | nt.insert(&*index, index.node(rev).unwrap(), rev) |
|
315 | 319 | .map_err(|e| nodemap_error(py, e))? |
|
316 | 320 | } |
|
317 | 321 | Ok(py.None()) |
|
318 | 322 | } |
|
319 | 323 | |
|
320 | 324 | fn get_nodetree<'a>( |
|
321 | 325 | &'a self, |
|
322 | 326 | py: Python<'a>, |
|
323 | 327 | ) -> PyResult<&'a RefCell<Option<NodeTree>>> { |
|
324 | 328 | if self.nt(py).borrow().is_none() { |
|
325 | 329 | let readonly = Box::new(Vec::new()); |
|
326 | 330 | let mut nt = NodeTree::load_bytes(readonly, 0); |
|
327 | 331 | self.fill_nodemap(py, &mut nt)?; |
|
328 | 332 | self.nt(py).borrow_mut().replace(nt); |
|
329 | 333 | } |
|
330 | 334 | Ok(self.nt(py)) |
|
331 | 335 | } |
|
332 | 336 | |
|
333 | 337 | /// forward a method call to the underlying C index |
|
334 | 338 | fn call_cindex( |
|
335 | 339 | &self, |
|
336 | 340 | py: Python, |
|
337 | 341 | name: &str, |
|
338 | 342 | args: &PyTuple, |
|
339 | 343 | kwargs: Option<&PyDict>, |
|
340 | 344 | ) -> PyResult<PyObject> { |
|
341 | 345 | self.cindex(py) |
|
342 | 346 | .borrow() |
|
343 | 347 | .inner() |
|
344 | 348 | .call_method(py, name, args, kwargs) |
|
345 | 349 | } |
|
346 | 350 | |
|
347 | 351 | pub fn clone_cindex(&self, py: Python) -> cindex::Index { |
|
348 | 352 | self.cindex(py).borrow().clone_ref(py) |
|
349 | 353 | } |
|
350 | 354 | |
|
351 | 355 | /// Returns the full nodemap bytes to be written as-is to disk |
|
352 | 356 | fn inner_nodemap_data_all(&self, py: Python) -> PyResult<PyBytes> { |
|
353 | 357 | let nodemap = self.get_nodetree(py)?.borrow_mut().take().unwrap(); |
|
354 | 358 | let (readonly, bytes) = nodemap.into_readonly_and_added_bytes(); |
|
355 | 359 | |
|
356 | 360 | // If there's anything readonly, we need to build the data again from |
|
357 | 361 | // scratch |
|
358 | 362 | let bytes = if readonly.len() > 0 { |
|
359 | 363 | let mut nt = NodeTree::load_bytes(Box::new(vec![]), 0); |
|
360 | 364 | self.fill_nodemap(py, &mut nt)?; |
|
361 | 365 | |
|
362 | 366 | let (readonly, bytes) = nt.into_readonly_and_added_bytes(); |
|
363 | 367 | assert_eq!(readonly.len(), 0); |
|
364 | 368 | |
|
365 | 369 | bytes |
|
366 | 370 | } else { |
|
367 | 371 | bytes |
|
368 | 372 | }; |
|
369 | 373 | |
|
370 | 374 | let bytes = PyBytes::new(py, &bytes); |
|
371 | 375 | Ok(bytes) |
|
372 | 376 | } |
|
373 | 377 | |
|
374 | 378 | /// Returns the last saved docket along with the size of any changed data |
|
375 | 379 | /// (in number of blocks), and said data as bytes. |
|
376 | 380 | fn inner_nodemap_data_incremental( |
|
377 | 381 | &self, |
|
378 | 382 | py: Python, |
|
379 | 383 | ) -> PyResult<PyObject> { |
|
380 | 384 | let docket = self.docket(py).borrow(); |
|
381 | 385 | let docket = match docket.as_ref() { |
|
382 | 386 | Some(d) => d, |
|
383 | 387 | None => return Ok(py.None()), |
|
384 | 388 | }; |
|
385 | 389 | |
|
386 | 390 | let node_tree = self.get_nodetree(py)?.borrow_mut().take().unwrap(); |
|
387 | 391 | let masked_blocks = node_tree.masked_readonly_blocks(); |
|
388 | 392 | let (_, data) = node_tree.into_readonly_and_added_bytes(); |
|
389 | 393 | let changed = masked_blocks * std::mem::size_of::<Block>(); |
|
390 | 394 | |
|
391 | 395 | Ok((docket, changed, PyBytes::new(py, &data)) |
|
392 | 396 | .to_py_object(py) |
|
393 | 397 | .into_object()) |
|
394 | 398 | } |
|
395 | 399 | |
|
396 | 400 | /// Update the nodemap from the new (mmaped) data. |
|
397 | 401 | /// The docket is kept as a reference for later incremental calls. |
|
398 | 402 | fn inner_update_nodemap_data( |
|
399 | 403 | &self, |
|
400 | 404 | py: Python, |
|
401 | 405 | docket: PyObject, |
|
402 | 406 | nm_data: PyObject, |
|
403 | 407 | ) -> PyResult<PyObject> { |
|
404 | 408 | let buf = PyBuffer::get(py, &nm_data)?; |
|
405 | 409 | let len = buf.item_count(); |
|
406 | 410 | |
|
407 | 411 | // Build a slice from the mmap'ed buffer data |
|
408 | 412 | let cbuf = buf.buf_ptr(); |
|
409 | 413 | let bytes = if std::mem::size_of::<u8>() == buf.item_size() |
|
410 | 414 | && buf.is_c_contiguous() |
|
411 | 415 | && u8::is_compatible_format(buf.format()) |
|
412 | 416 | { |
|
413 | 417 | unsafe { std::slice::from_raw_parts(cbuf as *const u8, len) } |
|
414 | 418 | } else { |
|
415 | 419 | return Err(PyErr::new::<ValueError, _>( |
|
416 | 420 | py, |
|
417 | 421 | "Nodemap data buffer has an invalid memory representation" |
|
418 | 422 | .to_string(), |
|
419 | 423 | )); |
|
420 | 424 | }; |
|
421 | 425 | |
|
422 | 426 | // Keep a reference to the mmap'ed buffer, otherwise we get a dangling |
|
423 | 427 | // pointer. |
|
424 | 428 | self.mmap(py).borrow_mut().replace(buf); |
|
425 | 429 | |
|
426 | 430 | let mut nt = NodeTree::load_bytes(Box::new(bytes), len); |
|
427 | 431 | |
|
428 | 432 | let data_tip = |
|
429 | 433 | docket.getattr(py, "tip_rev")?.extract::<Revision>(py)?; |
|
430 | 434 | self.docket(py).borrow_mut().replace(docket.clone_ref(py)); |
|
431 | 435 | let idx = self.cindex(py).borrow(); |
|
432 | 436 | let current_tip = idx.len(); |
|
433 | 437 | |
|
434 | 438 | for r in (data_tip + 1)..current_tip as Revision { |
|
435 | 439 | let rev = r as Revision; |
|
436 | 440 | // in this case node() won't ever return None |
|
437 | 441 | nt.insert(&*idx, idx.node(rev).unwrap(), rev) |
|
438 | 442 | .map_err(|e| nodemap_error(py, e))? |
|
439 | 443 | } |
|
440 | 444 | |
|
441 | 445 | *self.nt(py).borrow_mut() = Some(nt); |
|
442 | 446 | |
|
443 | 447 | Ok(py.None()) |
|
444 | 448 | } |
|
445 | 449 | } |
|
446 | 450 | |
|
447 | 451 | fn revlog_error(py: Python) -> PyErr { |
|
448 | 452 | match py |
|
449 | 453 | .import("mercurial.error") |
|
450 | 454 | .and_then(|m| m.get(py, "RevlogError")) |
|
451 | 455 | { |
|
452 | 456 | Err(e) => e, |
|
453 | 457 | Ok(cls) => PyErr::from_instance(py, cls), |
|
454 | 458 | } |
|
455 | 459 | } |
|
456 | 460 | |
|
457 | 461 | fn rev_not_in_index(py: Python, rev: Revision) -> PyErr { |
|
458 | 462 | PyErr::new::<ValueError, _>( |
|
459 | 463 | py, |
|
460 | 464 | format!( |
|
461 | 465 | "Inconsistency: Revision {} found in nodemap \ |
|
462 | 466 | is not in revlog index", |
|
463 | 467 | rev |
|
464 | 468 | ), |
|
465 | 469 | ) |
|
466 | 470 | } |
|
467 | 471 | |
|
468 | 472 | /// Standard treatment of NodeMapError |
|
469 | 473 | fn nodemap_error(py: Python, err: NodeMapError) -> PyErr { |
|
470 | 474 | match err { |
|
471 | 475 | NodeMapError::MultipleResults => revlog_error(py), |
|
472 | 476 | NodeMapError::RevisionNotInIndex(r) => rev_not_in_index(py, r), |
|
473 | 477 | } |
|
474 | 478 | } |
|
475 | 479 | |
|
476 | 480 | /// Create the module, with __package__ given from parent |
|
477 | 481 | pub fn init_module(py: Python, package: &str) -> PyResult<PyModule> { |
|
478 | 482 | let dotted_name = &format!("{}.revlog", package); |
|
479 | 483 | let m = PyModule::new(py, dotted_name)?; |
|
480 | 484 | m.add(py, "__package__", package)?; |
|
481 | 485 | m.add(py, "__doc__", "RevLog - Rust implementations")?; |
|
482 | 486 | |
|
483 | 487 | m.add_class::<MixedIndex>(py)?; |
|
484 | 488 | |
|
485 | 489 | let sys = PyModule::import(py, "sys")?; |
|
486 | 490 | let sys_modules: PyDict = sys.get(py, "modules")?.extract(py)?; |
|
487 | 491 | sys_modules.set_item(py, dotted_name, &m)?; |
|
488 | 492 | |
|
489 | 493 | Ok(m) |
|
490 | 494 | } |
General Comments 0
You need to be logged in to leave comments.
Login now