Show More
@@ -1,4706 +1,4725 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 | (by default, the first iteration is benchmarked) |
|
24 | 24 | |
|
25 | 25 | ``profiled-runs`` |
|
26 | 26 | list of iteration to profile (starting from 0) |
|
27 | 27 | |
|
28 | 28 | ``run-limits`` |
|
29 | 29 | Control the number of runs each benchmark will perform. The option value |
|
30 | 30 | should be a list of `<time>-<numberofrun>` pairs. After each run the |
|
31 | 31 | conditions are considered in order with the following logic: |
|
32 | 32 | |
|
33 | 33 | If benchmark has been running for <time> seconds, and we have performed |
|
34 | 34 | <numberofrun> iterations, stop the benchmark, |
|
35 | 35 | |
|
36 | 36 | The default value is: `3.0-100, 10.0-3` |
|
37 | 37 | |
|
38 | 38 | ``stub`` |
|
39 | 39 | When set, benchmarks will only be run once, useful for testing |
|
40 | 40 | (default: off) |
|
41 | 41 | ''' |
|
42 | 42 | |
|
43 | 43 | # "historical portability" policy of perf.py: |
|
44 | 44 | # |
|
45 | 45 | # We have to do: |
|
46 | 46 | # - make perf.py "loadable" with as wide Mercurial version as possible |
|
47 | 47 | # This doesn't mean that perf commands work correctly with that Mercurial. |
|
48 | 48 | # BTW, perf.py itself has been available since 1.1 (or eb240755386d). |
|
49 | 49 | # - make historical perf command work correctly with as wide Mercurial |
|
50 | 50 | # version as possible |
|
51 | 51 | # |
|
52 | 52 | # We have to do, if possible with reasonable cost: |
|
53 | 53 | # - make recent perf command for historical feature work correctly |
|
54 | 54 | # with early Mercurial |
|
55 | 55 | # |
|
56 | 56 | # We don't have to do: |
|
57 | 57 | # - make perf command for recent feature work correctly with early |
|
58 | 58 | # Mercurial |
|
59 | 59 | |
|
60 | 60 | import contextlib |
|
61 | 61 | import functools |
|
62 | 62 | import gc |
|
63 | 63 | import os |
|
64 | 64 | import random |
|
65 | 65 | import shutil |
|
66 | 66 | import struct |
|
67 | 67 | import sys |
|
68 | 68 | import tempfile |
|
69 | 69 | import threading |
|
70 | 70 | import time |
|
71 | 71 | |
|
72 | 72 | import mercurial.revlog |
|
73 | 73 | from mercurial import ( |
|
74 | 74 | changegroup, |
|
75 | 75 | cmdutil, |
|
76 | 76 | commands, |
|
77 | 77 | copies, |
|
78 | 78 | error, |
|
79 | 79 | extensions, |
|
80 | 80 | hg, |
|
81 | 81 | mdiff, |
|
82 | 82 | merge, |
|
83 | 83 | util, |
|
84 | 84 | ) |
|
85 | 85 | |
|
86 | 86 | # for "historical portability": |
|
87 | 87 | # try to import modules separately (in dict order), and ignore |
|
88 | 88 | # failure, because these aren't available with early Mercurial |
|
89 | 89 | try: |
|
90 | 90 | from mercurial import branchmap # since 2.5 (or bcee63733aad) |
|
91 | 91 | except ImportError: |
|
92 | 92 | pass |
|
93 | 93 | try: |
|
94 | 94 | from mercurial import obsolete # since 2.3 (or ad0d6c2b3279) |
|
95 | 95 | except ImportError: |
|
96 | 96 | pass |
|
97 | 97 | try: |
|
98 | 98 | from mercurial import registrar # since 3.7 (or 37d50250b696) |
|
99 | 99 | |
|
100 | 100 | dir(registrar) # forcibly load it |
|
101 | 101 | except ImportError: |
|
102 | 102 | registrar = None |
|
103 | 103 | try: |
|
104 | 104 | from mercurial import repoview # since 2.5 (or 3a6ddacb7198) |
|
105 | 105 | except ImportError: |
|
106 | 106 | pass |
|
107 | 107 | try: |
|
108 | 108 | from mercurial.utils import repoviewutil # since 5.0 |
|
109 | 109 | except ImportError: |
|
110 | 110 | repoviewutil = None |
|
111 | 111 | try: |
|
112 | 112 | from mercurial import scmutil # since 1.9 (or 8b252e826c68) |
|
113 | 113 | except ImportError: |
|
114 | 114 | pass |
|
115 | 115 | try: |
|
116 | 116 | from mercurial import setdiscovery # since 1.9 (or cb98fed52495) |
|
117 | 117 | except ImportError: |
|
118 | 118 | pass |
|
119 | 119 | |
|
120 | 120 | try: |
|
121 | 121 | from mercurial import profiling |
|
122 | 122 | except ImportError: |
|
123 | 123 | profiling = None |
|
124 | 124 | |
|
125 | 125 | try: |
|
126 | 126 | from mercurial.revlogutils import constants as revlog_constants |
|
127 | 127 | |
|
128 | 128 | perf_rl_kind = (revlog_constants.KIND_OTHER, b'created-by-perf') |
|
129 | 129 | |
|
130 | 130 | def revlog(opener, *args, **kwargs): |
|
131 | 131 | return mercurial.revlog.revlog(opener, perf_rl_kind, *args, **kwargs) |
|
132 | 132 | |
|
133 | 133 | |
|
134 | 134 | except (ImportError, AttributeError): |
|
135 | 135 | perf_rl_kind = None |
|
136 | 136 | |
|
137 | 137 | def revlog(opener, *args, **kwargs): |
|
138 | 138 | return mercurial.revlog.revlog(opener, *args, **kwargs) |
|
139 | 139 | |
|
140 | 140 | |
|
141 | 141 | def identity(a): |
|
142 | 142 | return a |
|
143 | 143 | |
|
144 | 144 | |
|
145 | 145 | try: |
|
146 | 146 | from mercurial import pycompat |
|
147 | 147 | |
|
148 | 148 | getargspec = pycompat.getargspec # added to module after 4.5 |
|
149 | 149 | _byteskwargs = pycompat.byteskwargs # since 4.1 (or fbc3f73dc802) |
|
150 | 150 | _sysstr = pycompat.sysstr # since 4.0 (or 2219f4f82ede) |
|
151 | 151 | _bytestr = pycompat.bytestr # since 4.2 (or b70407bd84d5) |
|
152 | 152 | _xrange = pycompat.xrange # since 4.8 (or 7eba8f83129b) |
|
153 | 153 | fsencode = pycompat.fsencode # since 3.9 (or f4a5e0e86a7e) |
|
154 | 154 | if pycompat.ispy3: |
|
155 | 155 | _maxint = sys.maxsize # per py3 docs for replacing maxint |
|
156 | 156 | else: |
|
157 | 157 | _maxint = sys.maxint |
|
158 | 158 | except (NameError, ImportError, AttributeError): |
|
159 | 159 | import inspect |
|
160 | 160 | |
|
161 | 161 | getargspec = inspect.getargspec |
|
162 | 162 | _byteskwargs = identity |
|
163 | 163 | _bytestr = str |
|
164 | 164 | fsencode = identity # no py3 support |
|
165 | 165 | _maxint = sys.maxint # no py3 support |
|
166 | 166 | _sysstr = lambda x: x # no py3 support |
|
167 | 167 | _xrange = xrange |
|
168 | 168 | |
|
169 | 169 | try: |
|
170 | 170 | # 4.7+ |
|
171 | 171 | queue = pycompat.queue.Queue |
|
172 | 172 | except (NameError, AttributeError, ImportError): |
|
173 | 173 | # <4.7. |
|
174 | 174 | try: |
|
175 | 175 | queue = pycompat.queue |
|
176 | 176 | except (NameError, AttributeError, ImportError): |
|
177 | 177 | import Queue as queue |
|
178 | 178 | |
|
179 | 179 | try: |
|
180 | 180 | from mercurial import logcmdutil |
|
181 | 181 | |
|
182 | 182 | makelogtemplater = logcmdutil.maketemplater |
|
183 | 183 | except (AttributeError, ImportError): |
|
184 | 184 | try: |
|
185 | 185 | makelogtemplater = cmdutil.makelogtemplater |
|
186 | 186 | except (AttributeError, ImportError): |
|
187 | 187 | makelogtemplater = None |
|
188 | 188 | |
|
189 | 189 | # for "historical portability": |
|
190 | 190 | # define util.safehasattr forcibly, because util.safehasattr has been |
|
191 | 191 | # available since 1.9.3 (or 94b200a11cf7) |
|
192 | 192 | _undefined = object() |
|
193 | 193 | |
|
194 | 194 | |
|
195 | 195 | def safehasattr(thing, attr): |
|
196 | 196 | return getattr(thing, _sysstr(attr), _undefined) is not _undefined |
|
197 | 197 | |
|
198 | 198 | |
|
199 | 199 | setattr(util, 'safehasattr', safehasattr) |
|
200 | 200 | |
|
201 | 201 | # for "historical portability": |
|
202 | 202 | # define util.timer forcibly, because util.timer has been available |
|
203 | 203 | # since ae5d60bb70c9 |
|
204 | 204 | if safehasattr(time, 'perf_counter'): |
|
205 | 205 | util.timer = time.perf_counter |
|
206 | 206 | elif os.name == b'nt': |
|
207 | 207 | util.timer = time.clock |
|
208 | 208 | else: |
|
209 | 209 | util.timer = time.time |
|
210 | 210 | |
|
211 | 211 | # for "historical portability": |
|
212 | 212 | # use locally defined empty option list, if formatteropts isn't |
|
213 | 213 | # available, because commands.formatteropts has been available since |
|
214 | 214 | # 3.2 (or 7a7eed5176a4), even though formatting itself has been |
|
215 | 215 | # available since 2.2 (or ae5f92e154d3) |
|
216 | 216 | formatteropts = getattr( |
|
217 | 217 | cmdutil, "formatteropts", getattr(commands, "formatteropts", []) |
|
218 | 218 | ) |
|
219 | 219 | |
|
220 | 220 | # for "historical portability": |
|
221 | 221 | # use locally defined option list, if debugrevlogopts isn't available, |
|
222 | 222 | # because commands.debugrevlogopts has been available since 3.7 (or |
|
223 | 223 | # 5606f7d0d063), even though cmdutil.openrevlog() has been available |
|
224 | 224 | # since 1.9 (or a79fea6b3e77). |
|
225 | 225 | revlogopts = getattr( |
|
226 | 226 | cmdutil, |
|
227 | 227 | "debugrevlogopts", |
|
228 | 228 | getattr( |
|
229 | 229 | commands, |
|
230 | 230 | "debugrevlogopts", |
|
231 | 231 | [ |
|
232 | 232 | (b'c', b'changelog', False, b'open changelog'), |
|
233 | 233 | (b'm', b'manifest', False, b'open manifest'), |
|
234 | 234 | (b'', b'dir', False, b'open directory manifest'), |
|
235 | 235 | ], |
|
236 | 236 | ), |
|
237 | 237 | ) |
|
238 | 238 | |
|
239 | 239 | cmdtable = {} |
|
240 | 240 | |
|
241 | 241 | |
|
242 | 242 | # for "historical portability": |
|
243 | 243 | # define parsealiases locally, because cmdutil.parsealiases has been |
|
244 | 244 | # available since 1.5 (or 6252852b4332) |
|
245 | 245 | def parsealiases(cmd): |
|
246 | 246 | return cmd.split(b"|") |
|
247 | 247 | |
|
248 | 248 | |
|
249 | 249 | if safehasattr(registrar, 'command'): |
|
250 | 250 | command = registrar.command(cmdtable) |
|
251 | 251 | elif safehasattr(cmdutil, 'command'): |
|
252 | 252 | command = cmdutil.command(cmdtable) |
|
253 | 253 | if 'norepo' not in getargspec(command).args: |
|
254 | 254 | # for "historical portability": |
|
255 | 255 | # wrap original cmdutil.command, because "norepo" option has |
|
256 | 256 | # been available since 3.1 (or 75a96326cecb) |
|
257 | 257 | _command = command |
|
258 | 258 | |
|
259 | 259 | def command(name, options=(), synopsis=None, norepo=False): |
|
260 | 260 | if norepo: |
|
261 | 261 | commands.norepo += b' %s' % b' '.join(parsealiases(name)) |
|
262 | 262 | return _command(name, list(options), synopsis) |
|
263 | 263 | |
|
264 | 264 | |
|
265 | 265 | else: |
|
266 | 266 | # for "historical portability": |
|
267 | 267 | # define "@command" annotation locally, because cmdutil.command |
|
268 | 268 | # has been available since 1.9 (or 2daa5179e73f) |
|
269 | 269 | def command(name, options=(), synopsis=None, norepo=False): |
|
270 | 270 | def decorator(func): |
|
271 | 271 | if synopsis: |
|
272 | 272 | cmdtable[name] = func, list(options), synopsis |
|
273 | 273 | else: |
|
274 | 274 | cmdtable[name] = func, list(options) |
|
275 | 275 | if norepo: |
|
276 | 276 | commands.norepo += b' %s' % b' '.join(parsealiases(name)) |
|
277 | 277 | return func |
|
278 | 278 | |
|
279 | 279 | return decorator |
|
280 | 280 | |
|
281 | 281 | |
|
282 | 282 | try: |
|
283 | 283 | import mercurial.registrar |
|
284 | 284 | import mercurial.configitems |
|
285 | 285 | |
|
286 | 286 | configtable = {} |
|
287 | 287 | configitem = mercurial.registrar.configitem(configtable) |
|
288 | 288 | configitem( |
|
289 | 289 | b'perf', |
|
290 | 290 | b'presleep', |
|
291 | 291 | default=mercurial.configitems.dynamicdefault, |
|
292 | 292 | experimental=True, |
|
293 | 293 | ) |
|
294 | 294 | configitem( |
|
295 | 295 | b'perf', |
|
296 | 296 | b'stub', |
|
297 | 297 | default=mercurial.configitems.dynamicdefault, |
|
298 | 298 | experimental=True, |
|
299 | 299 | ) |
|
300 | 300 | configitem( |
|
301 | 301 | b'perf', |
|
302 | 302 | b'parentscount', |
|
303 | 303 | default=mercurial.configitems.dynamicdefault, |
|
304 | 304 | experimental=True, |
|
305 | 305 | ) |
|
306 | 306 | configitem( |
|
307 | 307 | b'perf', |
|
308 | 308 | b'all-timing', |
|
309 | 309 | default=mercurial.configitems.dynamicdefault, |
|
310 | 310 | experimental=True, |
|
311 | 311 | ) |
|
312 | 312 | configitem( |
|
313 | 313 | b'perf', |
|
314 | 314 | b'pre-run', |
|
315 | 315 | default=mercurial.configitems.dynamicdefault, |
|
316 | 316 | ) |
|
317 | 317 | configitem( |
|
318 | 318 | b'perf', |
|
319 | 319 | b'profile-benchmark', |
|
320 | 320 | default=mercurial.configitems.dynamicdefault, |
|
321 | 321 | ) |
|
322 | 322 | configitem( |
|
323 | 323 | b'perf', |
|
324 | 324 | b'profiled-runs', |
|
325 | 325 | default=mercurial.configitems.dynamicdefault, |
|
326 | 326 | ) |
|
327 | 327 | configitem( |
|
328 | 328 | b'perf', |
|
329 | 329 | b'run-limits', |
|
330 | 330 | default=mercurial.configitems.dynamicdefault, |
|
331 | 331 | experimental=True, |
|
332 | 332 | ) |
|
333 | 333 | except (ImportError, AttributeError): |
|
334 | 334 | pass |
|
335 | 335 | except TypeError: |
|
336 | 336 | # compatibility fix for a11fd395e83f |
|
337 | 337 | # hg version: 5.2 |
|
338 | 338 | configitem( |
|
339 | 339 | b'perf', |
|
340 | 340 | b'presleep', |
|
341 | 341 | default=mercurial.configitems.dynamicdefault, |
|
342 | 342 | ) |
|
343 | 343 | configitem( |
|
344 | 344 | b'perf', |
|
345 | 345 | b'stub', |
|
346 | 346 | default=mercurial.configitems.dynamicdefault, |
|
347 | 347 | ) |
|
348 | 348 | configitem( |
|
349 | 349 | b'perf', |
|
350 | 350 | b'parentscount', |
|
351 | 351 | default=mercurial.configitems.dynamicdefault, |
|
352 | 352 | ) |
|
353 | 353 | configitem( |
|
354 | 354 | b'perf', |
|
355 | 355 | b'all-timing', |
|
356 | 356 | default=mercurial.configitems.dynamicdefault, |
|
357 | 357 | ) |
|
358 | 358 | configitem( |
|
359 | 359 | b'perf', |
|
360 | 360 | b'pre-run', |
|
361 | 361 | default=mercurial.configitems.dynamicdefault, |
|
362 | 362 | ) |
|
363 | 363 | configitem( |
|
364 | 364 | b'perf', |
|
365 | 365 | b'profiled-runs', |
|
366 | 366 | default=mercurial.configitems.dynamicdefault, |
|
367 | 367 | ) |
|
368 | 368 | configitem( |
|
369 | 369 | b'perf', |
|
370 | 370 | b'run-limits', |
|
371 | 371 | default=mercurial.configitems.dynamicdefault, |
|
372 | 372 | ) |
|
373 | 373 | |
|
374 | 374 | |
|
375 | 375 | def getlen(ui): |
|
376 | 376 | if ui.configbool(b"perf", b"stub", False): |
|
377 | 377 | return lambda x: 1 |
|
378 | 378 | return len |
|
379 | 379 | |
|
380 | 380 | |
|
381 | 381 | class noop: |
|
382 | 382 | """dummy context manager""" |
|
383 | 383 | |
|
384 | 384 | def __enter__(self): |
|
385 | 385 | pass |
|
386 | 386 | |
|
387 | 387 | def __exit__(self, *args): |
|
388 | 388 | pass |
|
389 | 389 | |
|
390 | 390 | |
|
391 | 391 | NOOPCTX = noop() |
|
392 | 392 | |
|
393 | 393 | |
|
394 | 394 | def gettimer(ui, opts=None): |
|
395 | 395 | """return a timer function and formatter: (timer, formatter) |
|
396 | 396 | |
|
397 | 397 | This function exists to gather the creation of formatter in a single |
|
398 | 398 | place instead of duplicating it in all performance commands.""" |
|
399 | 399 | |
|
400 | 400 | # enforce an idle period before execution to counteract power management |
|
401 | 401 | # experimental config: perf.presleep |
|
402 | 402 | time.sleep(getint(ui, b"perf", b"presleep", 1)) |
|
403 | 403 | |
|
404 | 404 | if opts is None: |
|
405 | 405 | opts = {} |
|
406 | 406 | # redirect all to stderr unless buffer api is in use |
|
407 | 407 | if not ui._buffers: |
|
408 | 408 | ui = ui.copy() |
|
409 | 409 | uifout = safeattrsetter(ui, b'fout', ignoremissing=True) |
|
410 | 410 | if uifout: |
|
411 | 411 | # for "historical portability": |
|
412 | 412 | # ui.fout/ferr have been available since 1.9 (or 4e1ccd4c2b6d) |
|
413 | 413 | uifout.set(ui.ferr) |
|
414 | 414 | |
|
415 | 415 | # get a formatter |
|
416 | 416 | uiformatter = getattr(ui, 'formatter', None) |
|
417 | 417 | if uiformatter: |
|
418 | 418 | fm = uiformatter(b'perf', opts) |
|
419 | 419 | else: |
|
420 | 420 | # for "historical portability": |
|
421 | 421 | # define formatter locally, because ui.formatter has been |
|
422 | 422 | # available since 2.2 (or ae5f92e154d3) |
|
423 | 423 | from mercurial import node |
|
424 | 424 | |
|
425 | 425 | class defaultformatter: |
|
426 | 426 | """Minimized composition of baseformatter and plainformatter""" |
|
427 | 427 | |
|
428 | 428 | def __init__(self, ui, topic, opts): |
|
429 | 429 | self._ui = ui |
|
430 | 430 | if ui.debugflag: |
|
431 | 431 | self.hexfunc = node.hex |
|
432 | 432 | else: |
|
433 | 433 | self.hexfunc = node.short |
|
434 | 434 | |
|
435 | 435 | def __nonzero__(self): |
|
436 | 436 | return False |
|
437 | 437 | |
|
438 | 438 | __bool__ = __nonzero__ |
|
439 | 439 | |
|
440 | 440 | def startitem(self): |
|
441 | 441 | pass |
|
442 | 442 | |
|
443 | 443 | def data(self, **data): |
|
444 | 444 | pass |
|
445 | 445 | |
|
446 | 446 | def write(self, fields, deftext, *fielddata, **opts): |
|
447 | 447 | self._ui.write(deftext % fielddata, **opts) |
|
448 | 448 | |
|
449 | 449 | def condwrite(self, cond, fields, deftext, *fielddata, **opts): |
|
450 | 450 | if cond: |
|
451 | 451 | self._ui.write(deftext % fielddata, **opts) |
|
452 | 452 | |
|
453 | 453 | def plain(self, text, **opts): |
|
454 | 454 | self._ui.write(text, **opts) |
|
455 | 455 | |
|
456 | 456 | def end(self): |
|
457 | 457 | pass |
|
458 | 458 | |
|
459 | 459 | fm = defaultformatter(ui, b'perf', opts) |
|
460 | 460 | |
|
461 | 461 | # stub function, runs code only once instead of in a loop |
|
462 | 462 | # experimental config: perf.stub |
|
463 | 463 | if ui.configbool(b"perf", b"stub", False): |
|
464 | 464 | return functools.partial(stub_timer, fm), fm |
|
465 | 465 | |
|
466 | 466 | # experimental config: perf.all-timing |
|
467 | 467 | displayall = ui.configbool(b"perf", b"all-timing", True) |
|
468 | 468 | |
|
469 | 469 | # experimental config: perf.run-limits |
|
470 | 470 | limitspec = ui.configlist(b"perf", b"run-limits", []) |
|
471 | 471 | limits = [] |
|
472 | 472 | for item in limitspec: |
|
473 | 473 | parts = item.split(b'-', 1) |
|
474 | 474 | if len(parts) < 2: |
|
475 | 475 | ui.warn((b'malformatted run limit entry, missing "-": %s\n' % item)) |
|
476 | 476 | continue |
|
477 | 477 | try: |
|
478 | 478 | time_limit = float(_sysstr(parts[0])) |
|
479 | 479 | except ValueError as e: |
|
480 | 480 | ui.warn( |
|
481 | 481 | ( |
|
482 | 482 | b'malformatted run limit entry, %s: %s\n' |
|
483 | 483 | % (_bytestr(e), item) |
|
484 | 484 | ) |
|
485 | 485 | ) |
|
486 | 486 | continue |
|
487 | 487 | try: |
|
488 | 488 | run_limit = int(_sysstr(parts[1])) |
|
489 | 489 | except ValueError as e: |
|
490 | 490 | ui.warn( |
|
491 | 491 | ( |
|
492 | 492 | b'malformatted run limit entry, %s: %s\n' |
|
493 | 493 | % (_bytestr(e), item) |
|
494 | 494 | ) |
|
495 | 495 | ) |
|
496 | 496 | continue |
|
497 | 497 | limits.append((time_limit, run_limit)) |
|
498 | 498 | if not limits: |
|
499 | 499 | limits = DEFAULTLIMITS |
|
500 | 500 | |
|
501 | 501 | profiler = None |
|
502 | 502 | profiled_runs = set() |
|
503 | 503 | if profiling is not None: |
|
504 | 504 | if ui.configbool(b"perf", b"profile-benchmark", False): |
|
505 | 505 | profiler = lambda: profiling.profile(ui) |
|
506 | 506 | for run in ui.configlist(b"perf", b"profiled-runs", [0]): |
|
507 | 507 | profiled_runs.add(int(run)) |
|
508 | 508 | |
|
509 | 509 | prerun = getint(ui, b"perf", b"pre-run", 0) |
|
510 | 510 | t = functools.partial( |
|
511 | 511 | _timer, |
|
512 | 512 | fm, |
|
513 | 513 | displayall=displayall, |
|
514 | 514 | limits=limits, |
|
515 | 515 | prerun=prerun, |
|
516 | 516 | profiler=profiler, |
|
517 | 517 | profiled_runs=profiled_runs, |
|
518 | 518 | ) |
|
519 | 519 | return t, fm |
|
520 | 520 | |
|
521 | 521 | |
|
522 | 522 | def stub_timer(fm, func, setup=None, title=None): |
|
523 | 523 | if setup is not None: |
|
524 | 524 | setup() |
|
525 | 525 | func() |
|
526 | 526 | |
|
527 | 527 | |
|
528 | 528 | @contextlib.contextmanager |
|
529 | 529 | def timeone(): |
|
530 | 530 | r = [] |
|
531 | 531 | ostart = os.times() |
|
532 | 532 | cstart = util.timer() |
|
533 | 533 | yield r |
|
534 | 534 | cstop = util.timer() |
|
535 | 535 | ostop = os.times() |
|
536 | 536 | a, b = ostart, ostop |
|
537 | 537 | r.append((cstop - cstart, b[0] - a[0], b[1] - a[1])) |
|
538 | 538 | |
|
539 | 539 | |
|
540 | 540 | # list of stop condition (elapsed time, minimal run count) |
|
541 | 541 | DEFAULTLIMITS = ( |
|
542 | 542 | (3.0, 100), |
|
543 | 543 | (10.0, 3), |
|
544 | 544 | ) |
|
545 | 545 | |
|
546 | 546 | |
|
547 | 547 | @contextlib.contextmanager |
|
548 | 548 | def noop_context(): |
|
549 | 549 | yield |
|
550 | 550 | |
|
551 | 551 | |
|
552 | 552 | def _timer( |
|
553 | 553 | fm, |
|
554 | 554 | func, |
|
555 | 555 | setup=None, |
|
556 | 556 | context=noop_context, |
|
557 | 557 | title=None, |
|
558 | 558 | displayall=False, |
|
559 | 559 | limits=DEFAULTLIMITS, |
|
560 | 560 | prerun=0, |
|
561 | 561 | profiler=None, |
|
562 | 562 | profiled_runs=(0,), |
|
563 | 563 | ): |
|
564 | 564 | gc.collect() |
|
565 | 565 | results = [] |
|
566 | 566 | count = 0 |
|
567 | 567 | if profiler is None: |
|
568 | 568 | profiler = lambda: NOOPCTX |
|
569 | 569 | for i in range(prerun): |
|
570 | 570 | if setup is not None: |
|
571 | 571 | setup() |
|
572 | 572 | with context(): |
|
573 | 573 | func() |
|
574 | 574 | begin = util.timer() |
|
575 | 575 | keepgoing = True |
|
576 | 576 | while keepgoing: |
|
577 | 577 | if count in profiled_runs: |
|
578 | 578 | prof = profiler() |
|
579 | 579 | else: |
|
580 | 580 | prof = NOOPCTX |
|
581 | 581 | if setup is not None: |
|
582 | 582 | setup() |
|
583 | 583 | with context(): |
|
584 | 584 | gc.collect() |
|
585 | 585 | with prof: |
|
586 | 586 | with timeone() as item: |
|
587 | 587 | r = func() |
|
588 | 588 | count += 1 |
|
589 | 589 | results.append(item[0]) |
|
590 | 590 | cstop = util.timer() |
|
591 | 591 | # Look for a stop condition. |
|
592 | 592 | elapsed = cstop - begin |
|
593 | 593 | for t, mincount in limits: |
|
594 | 594 | if elapsed >= t and count >= mincount: |
|
595 | 595 | keepgoing = False |
|
596 | 596 | break |
|
597 | 597 | |
|
598 | 598 | formatone(fm, results, title=title, result=r, displayall=displayall) |
|
599 | 599 | |
|
600 | 600 | |
|
601 | 601 | def formatone(fm, timings, title=None, result=None, displayall=False): |
|
602 | 602 | count = len(timings) |
|
603 | 603 | |
|
604 | 604 | fm.startitem() |
|
605 | 605 | |
|
606 | 606 | if title: |
|
607 | 607 | fm.write(b'title', b'! %s\n', title) |
|
608 | 608 | if result: |
|
609 | 609 | fm.write(b'result', b'! result: %s\n', result) |
|
610 | 610 | |
|
611 | 611 | def display(role, entry): |
|
612 | 612 | prefix = b'' |
|
613 | 613 | if role != b'best': |
|
614 | 614 | prefix = b'%s.' % role |
|
615 | 615 | fm.plain(b'!') |
|
616 | 616 | fm.write(prefix + b'wall', b' wall %f', entry[0]) |
|
617 | 617 | fm.write(prefix + b'comb', b' comb %f', entry[1] + entry[2]) |
|
618 | 618 | fm.write(prefix + b'user', b' user %f', entry[1]) |
|
619 | 619 | fm.write(prefix + b'sys', b' sys %f', entry[2]) |
|
620 | 620 | fm.write(prefix + b'count', b' (%s of %%d)' % role, count) |
|
621 | 621 | fm.plain(b'\n') |
|
622 | 622 | |
|
623 | 623 | timings.sort() |
|
624 | 624 | min_val = timings[0] |
|
625 | 625 | display(b'best', min_val) |
|
626 | 626 | if displayall: |
|
627 | 627 | max_val = timings[-1] |
|
628 | 628 | display(b'max', max_val) |
|
629 | 629 | avg = tuple([sum(x) / count for x in zip(*timings)]) |
|
630 | 630 | display(b'avg', avg) |
|
631 | 631 | median = timings[len(timings) // 2] |
|
632 | 632 | display(b'median', median) |
|
633 | 633 | |
|
634 | 634 | |
|
635 | 635 | # utilities for historical portability |
|
636 | 636 | |
|
637 | 637 | |
|
638 | 638 | def getint(ui, section, name, default): |
|
639 | 639 | # for "historical portability": |
|
640 | 640 | # ui.configint has been available since 1.9 (or fa2b596db182) |
|
641 | 641 | v = ui.config(section, name, None) |
|
642 | 642 | if v is None: |
|
643 | 643 | return default |
|
644 | 644 | try: |
|
645 | 645 | return int(v) |
|
646 | 646 | except ValueError: |
|
647 | 647 | raise error.ConfigError( |
|
648 | 648 | b"%s.%s is not an integer ('%s')" % (section, name, v) |
|
649 | 649 | ) |
|
650 | 650 | |
|
651 | 651 | |
|
652 | 652 | def safeattrsetter(obj, name, ignoremissing=False): |
|
653 | 653 | """Ensure that 'obj' has 'name' attribute before subsequent setattr |
|
654 | 654 | |
|
655 | 655 | This function is aborted, if 'obj' doesn't have 'name' attribute |
|
656 | 656 | at runtime. This avoids overlooking removal of an attribute, which |
|
657 | 657 | breaks assumption of performance measurement, in the future. |
|
658 | 658 | |
|
659 | 659 | This function returns the object to (1) assign a new value, and |
|
660 | 660 | (2) restore an original value to the attribute. |
|
661 | 661 | |
|
662 | 662 | If 'ignoremissing' is true, missing 'name' attribute doesn't cause |
|
663 | 663 | abortion, and this function returns None. This is useful to |
|
664 | 664 | examine an attribute, which isn't ensured in all Mercurial |
|
665 | 665 | versions. |
|
666 | 666 | """ |
|
667 | 667 | if not util.safehasattr(obj, name): |
|
668 | 668 | if ignoremissing: |
|
669 | 669 | return None |
|
670 | 670 | raise error.Abort( |
|
671 | 671 | ( |
|
672 | 672 | b"missing attribute %s of %s might break assumption" |
|
673 | 673 | b" of performance measurement" |
|
674 | 674 | ) |
|
675 | 675 | % (name, obj) |
|
676 | 676 | ) |
|
677 | 677 | |
|
678 | 678 | origvalue = getattr(obj, _sysstr(name)) |
|
679 | 679 | |
|
680 | 680 | class attrutil: |
|
681 | 681 | def set(self, newvalue): |
|
682 | 682 | setattr(obj, _sysstr(name), newvalue) |
|
683 | 683 | |
|
684 | 684 | def restore(self): |
|
685 | 685 | setattr(obj, _sysstr(name), origvalue) |
|
686 | 686 | |
|
687 | 687 | return attrutil() |
|
688 | 688 | |
|
689 | 689 | |
|
690 | 690 | # utilities to examine each internal API changes |
|
691 | 691 | |
|
692 | 692 | |
|
693 | 693 | def getbranchmapsubsettable(): |
|
694 | 694 | # for "historical portability": |
|
695 | 695 | # subsettable is defined in: |
|
696 | 696 | # - branchmap since 2.9 (or 175c6fd8cacc) |
|
697 | 697 | # - repoview since 2.5 (or 59a9f18d4587) |
|
698 | 698 | # - repoviewutil since 5.0 |
|
699 | 699 | for mod in (branchmap, repoview, repoviewutil): |
|
700 | 700 | subsettable = getattr(mod, 'subsettable', None) |
|
701 | 701 | if subsettable: |
|
702 | 702 | return subsettable |
|
703 | 703 | |
|
704 | 704 | # bisecting in bcee63733aad::59a9f18d4587 can reach here (both |
|
705 | 705 | # branchmap and repoview modules exist, but subsettable attribute |
|
706 | 706 | # doesn't) |
|
707 | 707 | raise error.Abort( |
|
708 | 708 | b"perfbranchmap not available with this Mercurial", |
|
709 | 709 | hint=b"use 2.5 or later", |
|
710 | 710 | ) |
|
711 | 711 | |
|
712 | 712 | |
|
713 | 713 | def getsvfs(repo): |
|
714 | 714 | """Return appropriate object to access files under .hg/store""" |
|
715 | 715 | # for "historical portability": |
|
716 | 716 | # repo.svfs has been available since 2.3 (or 7034365089bf) |
|
717 | 717 | svfs = getattr(repo, 'svfs', None) |
|
718 | 718 | if svfs: |
|
719 | 719 | return svfs |
|
720 | 720 | else: |
|
721 | 721 | return getattr(repo, 'sopener') |
|
722 | 722 | |
|
723 | 723 | |
|
724 | 724 | def getvfs(repo): |
|
725 | 725 | """Return appropriate object to access files under .hg""" |
|
726 | 726 | # for "historical portability": |
|
727 | 727 | # repo.vfs has been available since 2.3 (or 7034365089bf) |
|
728 | 728 | vfs = getattr(repo, 'vfs', None) |
|
729 | 729 | if vfs: |
|
730 | 730 | return vfs |
|
731 | 731 | else: |
|
732 | 732 | return getattr(repo, 'opener') |
|
733 | 733 | |
|
734 | 734 | |
|
735 | 735 | def repocleartagscachefunc(repo): |
|
736 | 736 | """Return the function to clear tags cache according to repo internal API""" |
|
737 | 737 | if util.safehasattr(repo, b'_tagscache'): # since 2.0 (or 9dca7653b525) |
|
738 | 738 | # in this case, setattr(repo, '_tagscache', None) or so isn't |
|
739 | 739 | # correct way to clear tags cache, because existing code paths |
|
740 | 740 | # expect _tagscache to be a structured object. |
|
741 | 741 | def clearcache(): |
|
742 | 742 | # _tagscache has been filteredpropertycache since 2.5 (or |
|
743 | 743 | # 98c867ac1330), and delattr() can't work in such case |
|
744 | 744 | if '_tagscache' in vars(repo): |
|
745 | 745 | del repo.__dict__['_tagscache'] |
|
746 | 746 | |
|
747 | 747 | return clearcache |
|
748 | 748 | |
|
749 | 749 | repotags = safeattrsetter(repo, b'_tags', ignoremissing=True) |
|
750 | 750 | if repotags: # since 1.4 (or 5614a628d173) |
|
751 | 751 | return lambda: repotags.set(None) |
|
752 | 752 | |
|
753 | 753 | repotagscache = safeattrsetter(repo, b'tagscache', ignoremissing=True) |
|
754 | 754 | if repotagscache: # since 0.6 (or d7df759d0e97) |
|
755 | 755 | return lambda: repotagscache.set(None) |
|
756 | 756 | |
|
757 | 757 | # Mercurial earlier than 0.6 (or d7df759d0e97) logically reaches |
|
758 | 758 | # this point, but it isn't so problematic, because: |
|
759 | 759 | # - repo.tags of such Mercurial isn't "callable", and repo.tags() |
|
760 | 760 | # in perftags() causes failure soon |
|
761 | 761 | # - perf.py itself has been available since 1.1 (or eb240755386d) |
|
762 | 762 | raise error.Abort(b"tags API of this hg command is unknown") |
|
763 | 763 | |
|
764 | 764 | |
|
765 | 765 | # utilities to clear cache |
|
766 | 766 | |
|
767 | 767 | |
|
768 | 768 | def clearfilecache(obj, attrname): |
|
769 | 769 | unfiltered = getattr(obj, 'unfiltered', None) |
|
770 | 770 | if unfiltered is not None: |
|
771 | 771 | obj = obj.unfiltered() |
|
772 | 772 | if attrname in vars(obj): |
|
773 | 773 | delattr(obj, attrname) |
|
774 | 774 | obj._filecache.pop(attrname, None) |
|
775 | 775 | |
|
776 | 776 | |
|
777 | 777 | def clearchangelog(repo): |
|
778 | 778 | if repo is not repo.unfiltered(): |
|
779 | 779 | object.__setattr__(repo, '_clcachekey', None) |
|
780 | 780 | object.__setattr__(repo, '_clcache', None) |
|
781 | 781 | clearfilecache(repo.unfiltered(), 'changelog') |
|
782 | 782 | |
|
783 | 783 | |
|
784 | 784 | # perf commands |
|
785 | 785 | |
|
786 | 786 | |
|
787 | 787 | @command(b'perf::walk|perfwalk', formatteropts) |
|
788 | 788 | def perfwalk(ui, repo, *pats, **opts): |
|
789 | 789 | opts = _byteskwargs(opts) |
|
790 | 790 | timer, fm = gettimer(ui, opts) |
|
791 | 791 | m = scmutil.match(repo[None], pats, {}) |
|
792 | 792 | timer( |
|
793 | 793 | lambda: len( |
|
794 | 794 | list( |
|
795 | 795 | repo.dirstate.walk(m, subrepos=[], unknown=True, ignored=False) |
|
796 | 796 | ) |
|
797 | 797 | ) |
|
798 | 798 | ) |
|
799 | 799 | fm.end() |
|
800 | 800 | |
|
801 | 801 | |
|
802 | 802 | @command(b'perf::annotate|perfannotate', formatteropts) |
|
803 | 803 | def perfannotate(ui, repo, f, **opts): |
|
804 | 804 | opts = _byteskwargs(opts) |
|
805 | 805 | timer, fm = gettimer(ui, opts) |
|
806 | 806 | fc = repo[b'.'][f] |
|
807 | 807 | timer(lambda: len(fc.annotate(True))) |
|
808 | 808 | fm.end() |
|
809 | 809 | |
|
810 | 810 | |
|
811 | 811 | @command( |
|
812 | 812 | b'perf::status|perfstatus', |
|
813 | 813 | [ |
|
814 | 814 | (b'u', b'unknown', False, b'ask status to look for unknown files'), |
|
815 | 815 | (b'', b'dirstate', False, b'benchmark the internal dirstate call'), |
|
816 | 816 | ] |
|
817 | 817 | + formatteropts, |
|
818 | 818 | ) |
|
819 | 819 | def perfstatus(ui, repo, **opts): |
|
820 | 820 | """benchmark the performance of a single status call |
|
821 | 821 | |
|
822 | 822 | The repository data are preserved between each call. |
|
823 | 823 | |
|
824 | 824 | By default, only the status of the tracked file are requested. If |
|
825 | 825 | `--unknown` is passed, the "unknown" files are also tracked. |
|
826 | 826 | """ |
|
827 | 827 | opts = _byteskwargs(opts) |
|
828 | 828 | # m = match.always(repo.root, repo.getcwd()) |
|
829 | 829 | # timer(lambda: sum(map(len, repo.dirstate.status(m, [], False, False, |
|
830 | 830 | # False)))) |
|
831 | 831 | timer, fm = gettimer(ui, opts) |
|
832 | 832 | if opts[b'dirstate']: |
|
833 | 833 | dirstate = repo.dirstate |
|
834 | 834 | m = scmutil.matchall(repo) |
|
835 | 835 | unknown = opts[b'unknown'] |
|
836 | 836 | |
|
837 | 837 | def status_dirstate(): |
|
838 | 838 | s = dirstate.status( |
|
839 | 839 | m, subrepos=[], ignored=False, clean=False, unknown=unknown |
|
840 | 840 | ) |
|
841 | 841 | sum(map(bool, s)) |
|
842 | 842 | |
|
843 | 843 | if util.safehasattr(dirstate, 'running_status'): |
|
844 | 844 | with dirstate.running_status(repo): |
|
845 | 845 | timer(status_dirstate) |
|
846 | 846 | dirstate.invalidate() |
|
847 | 847 | else: |
|
848 | 848 | timer(status_dirstate) |
|
849 | 849 | else: |
|
850 | 850 | timer(lambda: sum(map(len, repo.status(unknown=opts[b'unknown'])))) |
|
851 | 851 | fm.end() |
|
852 | 852 | |
|
853 | 853 | |
|
854 | 854 | @command(b'perf::addremove|perfaddremove', formatteropts) |
|
855 | 855 | def perfaddremove(ui, repo, **opts): |
|
856 | 856 | opts = _byteskwargs(opts) |
|
857 | 857 | timer, fm = gettimer(ui, opts) |
|
858 | 858 | try: |
|
859 | 859 | oldquiet = repo.ui.quiet |
|
860 | 860 | repo.ui.quiet = True |
|
861 | 861 | matcher = scmutil.match(repo[None]) |
|
862 | 862 | opts[b'dry_run'] = True |
|
863 | 863 | if 'uipathfn' in getargspec(scmutil.addremove).args: |
|
864 | 864 | uipathfn = scmutil.getuipathfn(repo) |
|
865 | 865 | timer(lambda: scmutil.addremove(repo, matcher, b"", uipathfn, opts)) |
|
866 | 866 | else: |
|
867 | 867 | timer(lambda: scmutil.addremove(repo, matcher, b"", opts)) |
|
868 | 868 | finally: |
|
869 | 869 | repo.ui.quiet = oldquiet |
|
870 | 870 | fm.end() |
|
871 | 871 | |
|
872 | 872 | |
|
873 | 873 | def clearcaches(cl): |
|
874 | 874 | # behave somewhat consistently across internal API changes |
|
875 | 875 | if util.safehasattr(cl, b'clearcaches'): |
|
876 | 876 | cl.clearcaches() |
|
877 | 877 | elif util.safehasattr(cl, b'_nodecache'): |
|
878 | 878 | # <= hg-5.2 |
|
879 | 879 | from mercurial.node import nullid, nullrev |
|
880 | 880 | |
|
881 | 881 | cl._nodecache = {nullid: nullrev} |
|
882 | 882 | cl._nodepos = None |
|
883 | 883 | |
|
884 | 884 | |
|
885 | 885 | @command(b'perf::heads|perfheads', formatteropts) |
|
886 | 886 | def perfheads(ui, repo, **opts): |
|
887 | 887 | """benchmark the computation of a changelog heads""" |
|
888 | 888 | opts = _byteskwargs(opts) |
|
889 | 889 | timer, fm = gettimer(ui, opts) |
|
890 | 890 | cl = repo.changelog |
|
891 | 891 | |
|
892 | 892 | def s(): |
|
893 | 893 | clearcaches(cl) |
|
894 | 894 | |
|
895 | 895 | def d(): |
|
896 | 896 | len(cl.headrevs()) |
|
897 | 897 | |
|
898 | 898 | timer(d, setup=s) |
|
899 | 899 | fm.end() |
|
900 | 900 | |
|
901 | 901 | |
|
902 | 902 | def _default_clear_on_disk_tags_cache(repo): |
|
903 | 903 | from mercurial import tags |
|
904 | 904 | |
|
905 | 905 | repo.cachevfs.tryunlink(tags._filename(repo)) |
|
906 | 906 | |
|
907 | 907 | |
|
908 | 908 | def _default_clear_on_disk_tags_fnodes_cache(repo): |
|
909 | 909 | from mercurial import tags |
|
910 | 910 | |
|
911 | 911 | repo.cachevfs.tryunlink(tags._fnodescachefile) |
|
912 | 912 | |
|
913 | 913 | |
|
914 | 914 | def _default_forget_fnodes(repo, revs): |
|
915 | 915 | """function used by the perf extension to prune some entries from the |
|
916 | 916 | fnodes cache""" |
|
917 | 917 | from mercurial import tags |
|
918 | 918 | |
|
919 | 919 | missing_1 = b'\xff' * 4 |
|
920 | 920 | missing_2 = b'\xff' * 20 |
|
921 | 921 | cache = tags.hgtagsfnodescache(repo.unfiltered()) |
|
922 | 922 | for r in revs: |
|
923 | 923 | cache._writeentry(r * tags._fnodesrecsize, missing_1, missing_2) |
|
924 | 924 | cache.write() |
|
925 | 925 | |
|
926 | 926 | |
|
927 | 927 | @command( |
|
928 | 928 | b'perf::tags|perftags', |
|
929 | 929 | formatteropts |
|
930 | 930 | + [ |
|
931 | 931 | (b'', b'clear-revlogs', False, b'refresh changelog and manifest'), |
|
932 | 932 | ( |
|
933 | 933 | b'', |
|
934 | 934 | b'clear-on-disk-cache', |
|
935 | 935 | False, |
|
936 | 936 | b'clear on disk tags cache (DESTRUCTIVE)', |
|
937 | 937 | ), |
|
938 | 938 | ( |
|
939 | 939 | b'', |
|
940 | 940 | b'clear-fnode-cache-all', |
|
941 | 941 | False, |
|
942 | 942 | b'clear on disk file node cache (DESTRUCTIVE),', |
|
943 | 943 | ), |
|
944 | 944 | ( |
|
945 | 945 | b'', |
|
946 | 946 | b'clear-fnode-cache-rev', |
|
947 | 947 | [], |
|
948 | 948 | b'clear on disk file node cache (DESTRUCTIVE),', |
|
949 | 949 | b'REVS', |
|
950 | 950 | ), |
|
951 | 951 | ( |
|
952 | 952 | b'', |
|
953 | 953 | b'update-last', |
|
954 | 954 | b'', |
|
955 | 955 | b'simulate an update over the last N revisions (DESTRUCTIVE),', |
|
956 | 956 | b'N', |
|
957 | 957 | ), |
|
958 | 958 | ], |
|
959 | 959 | ) |
|
960 | 960 | def perftags(ui, repo, **opts): |
|
961 | 961 | """Benchmark tags retrieval in various situation |
|
962 | 962 | |
|
963 | 963 | The option marked as (DESTRUCTIVE) will alter the on-disk cache, possibly |
|
964 | 964 | altering performance after the command was run. However, it does not |
|
965 | 965 | destroy any stored data. |
|
966 | 966 | """ |
|
967 | 967 | from mercurial import tags |
|
968 | 968 | |
|
969 | 969 | opts = _byteskwargs(opts) |
|
970 | 970 | timer, fm = gettimer(ui, opts) |
|
971 | 971 | repocleartagscache = repocleartagscachefunc(repo) |
|
972 | 972 | clearrevlogs = opts[b'clear_revlogs'] |
|
973 | 973 | clear_disk = opts[b'clear_on_disk_cache'] |
|
974 | 974 | clear_fnode = opts[b'clear_fnode_cache_all'] |
|
975 | 975 | |
|
976 | 976 | clear_fnode_revs = opts[b'clear_fnode_cache_rev'] |
|
977 | 977 | update_last_str = opts[b'update_last'] |
|
978 | 978 | update_last = None |
|
979 | 979 | if update_last_str: |
|
980 | 980 | try: |
|
981 | 981 | update_last = int(update_last_str) |
|
982 | 982 | except ValueError: |
|
983 | 983 | msg = b'could not parse value for update-last: "%s"' |
|
984 | 984 | msg %= update_last_str |
|
985 | 985 | hint = b'value should be an integer' |
|
986 | 986 | raise error.Abort(msg, hint=hint) |
|
987 | 987 | |
|
988 | 988 | clear_disk_fn = getattr( |
|
989 | 989 | tags, |
|
990 | 990 | "clear_cache_on_disk", |
|
991 | 991 | _default_clear_on_disk_tags_cache, |
|
992 | 992 | ) |
|
993 | 993 | if getattr(tags, 'clear_cache_fnodes_is_working', False): |
|
994 | 994 | clear_fnodes_fn = tags.clear_cache_fnodes |
|
995 | 995 | else: |
|
996 | 996 | clear_fnodes_fn = _default_clear_on_disk_tags_fnodes_cache |
|
997 | 997 | clear_fnodes_rev_fn = getattr( |
|
998 | 998 | tags, |
|
999 | 999 | "forget_fnodes", |
|
1000 | 1000 | _default_forget_fnodes, |
|
1001 | 1001 | ) |
|
1002 | 1002 | |
|
1003 | 1003 | clear_revs = [] |
|
1004 | 1004 | if clear_fnode_revs: |
|
1005 | 1005 | clear_revs.extend(scmutil.revrange(repo, clear_fnode_revs)) |
|
1006 | 1006 | |
|
1007 | 1007 | if update_last: |
|
1008 | 1008 | revset = b'last(all(), %d)' % update_last |
|
1009 | 1009 | last_revs = repo.unfiltered().revs(revset) |
|
1010 | 1010 | clear_revs.extend(last_revs) |
|
1011 | 1011 | |
|
1012 | 1012 | from mercurial import repoview |
|
1013 | 1013 | |
|
1014 | 1014 | rev_filter = {(b'experimental', b'extra-filter-revs'): revset} |
|
1015 | 1015 | with repo.ui.configoverride(rev_filter, source=b"perf"): |
|
1016 | 1016 | filter_id = repoview.extrafilter(repo.ui) |
|
1017 | 1017 | |
|
1018 | 1018 | filter_name = b'%s%%%s' % (repo.filtername, filter_id) |
|
1019 | 1019 | pre_repo = repo.filtered(filter_name) |
|
1020 | 1020 | pre_repo.tags() # warm the cache |
|
1021 | 1021 | old_tags_path = repo.cachevfs.join(tags._filename(pre_repo)) |
|
1022 | 1022 | new_tags_path = repo.cachevfs.join(tags._filename(repo)) |
|
1023 | 1023 | |
|
1024 | 1024 | clear_revs = sorted(set(clear_revs)) |
|
1025 | 1025 | |
|
1026 | 1026 | def s(): |
|
1027 | 1027 | if update_last: |
|
1028 | 1028 | util.copyfile(old_tags_path, new_tags_path) |
|
1029 | 1029 | if clearrevlogs: |
|
1030 | 1030 | clearchangelog(repo) |
|
1031 | 1031 | clearfilecache(repo.unfiltered(), 'manifest') |
|
1032 | 1032 | if clear_disk: |
|
1033 | 1033 | clear_disk_fn(repo) |
|
1034 | 1034 | if clear_fnode: |
|
1035 | 1035 | clear_fnodes_fn(repo) |
|
1036 | 1036 | elif clear_revs: |
|
1037 | 1037 | clear_fnodes_rev_fn(repo, clear_revs) |
|
1038 | 1038 | repocleartagscache() |
|
1039 | 1039 | |
|
1040 | 1040 | def t(): |
|
1041 | 1041 | len(repo.tags()) |
|
1042 | 1042 | |
|
1043 | 1043 | timer(t, setup=s) |
|
1044 | 1044 | fm.end() |
|
1045 | 1045 | |
|
1046 | 1046 | |
|
1047 | 1047 | @command(b'perf::ancestors|perfancestors', formatteropts) |
|
1048 | 1048 | def perfancestors(ui, repo, **opts): |
|
1049 | 1049 | opts = _byteskwargs(opts) |
|
1050 | 1050 | timer, fm = gettimer(ui, opts) |
|
1051 | 1051 | heads = repo.changelog.headrevs() |
|
1052 | 1052 | |
|
1053 | 1053 | def d(): |
|
1054 | 1054 | for a in repo.changelog.ancestors(heads): |
|
1055 | 1055 | pass |
|
1056 | 1056 | |
|
1057 | 1057 | timer(d) |
|
1058 | 1058 | fm.end() |
|
1059 | 1059 | |
|
1060 | 1060 | |
|
1061 | 1061 | @command(b'perf::ancestorset|perfancestorset', formatteropts) |
|
1062 | 1062 | def perfancestorset(ui, repo, revset, **opts): |
|
1063 | 1063 | opts = _byteskwargs(opts) |
|
1064 | 1064 | timer, fm = gettimer(ui, opts) |
|
1065 | 1065 | revs = repo.revs(revset) |
|
1066 | 1066 | heads = repo.changelog.headrevs() |
|
1067 | 1067 | |
|
1068 | 1068 | def d(): |
|
1069 | 1069 | s = repo.changelog.ancestors(heads) |
|
1070 | 1070 | for rev in revs: |
|
1071 | 1071 | rev in s |
|
1072 | 1072 | |
|
1073 | 1073 | timer(d) |
|
1074 | 1074 | fm.end() |
|
1075 | 1075 | |
|
1076 | 1076 | |
|
1077 | 1077 | @command( |
|
1078 | 1078 | b'perf::delta-find', |
|
1079 | 1079 | revlogopts + formatteropts, |
|
1080 | 1080 | b'-c|-m|FILE REV', |
|
1081 | 1081 | ) |
|
1082 | 1082 | def perf_delta_find(ui, repo, arg_1, arg_2=None, **opts): |
|
1083 | 1083 | """benchmark the process of finding a valid delta for a revlog revision |
|
1084 | 1084 | |
|
1085 | 1085 | When a revlog receives a new revision (e.g. from a commit, or from an |
|
1086 | 1086 | incoming bundle), it searches for a suitable delta-base to produce a delta. |
|
1087 | 1087 | This perf command measures how much time we spend in this process. It |
|
1088 | 1088 | operates on an already stored revision. |
|
1089 | 1089 | |
|
1090 | 1090 | See `hg help debug-delta-find` for another related command. |
|
1091 | 1091 | """ |
|
1092 | 1092 | from mercurial import revlogutils |
|
1093 | 1093 | import mercurial.revlogutils.deltas as deltautil |
|
1094 | 1094 | |
|
1095 | 1095 | opts = _byteskwargs(opts) |
|
1096 | 1096 | if arg_2 is None: |
|
1097 | 1097 | file_ = None |
|
1098 | 1098 | rev = arg_1 |
|
1099 | 1099 | else: |
|
1100 | 1100 | file_ = arg_1 |
|
1101 | 1101 | rev = arg_2 |
|
1102 | 1102 | |
|
1103 | 1103 | repo = repo.unfiltered() |
|
1104 | 1104 | |
|
1105 | 1105 | timer, fm = gettimer(ui, opts) |
|
1106 | 1106 | |
|
1107 | 1107 | rev = int(rev) |
|
1108 | 1108 | |
|
1109 | 1109 | revlog = cmdutil.openrevlog(repo, b'perf::delta-find', file_, opts) |
|
1110 | 1110 | |
|
1111 | 1111 | deltacomputer = deltautil.deltacomputer(revlog) |
|
1112 | 1112 | |
|
1113 | 1113 | node = revlog.node(rev) |
|
1114 | 1114 | p1r, p2r = revlog.parentrevs(rev) |
|
1115 | 1115 | p1 = revlog.node(p1r) |
|
1116 | 1116 | p2 = revlog.node(p2r) |
|
1117 | 1117 | full_text = revlog.revision(rev) |
|
1118 | 1118 | textlen = len(full_text) |
|
1119 | 1119 | cachedelta = None |
|
1120 | 1120 | flags = revlog.flags(rev) |
|
1121 | 1121 | |
|
1122 | 1122 | revinfo = revlogutils.revisioninfo( |
|
1123 | 1123 | node, |
|
1124 | 1124 | p1, |
|
1125 | 1125 | p2, |
|
1126 | 1126 | [full_text], # btext |
|
1127 | 1127 | textlen, |
|
1128 | 1128 | cachedelta, |
|
1129 | 1129 | flags, |
|
1130 | 1130 | ) |
|
1131 | 1131 | |
|
1132 | 1132 | # Note: we should probably purge the potential caches (like the full |
|
1133 | 1133 | # manifest cache) between runs. |
|
1134 | 1134 | def find_one(): |
|
1135 | 1135 | with revlog._datafp() as fh: |
|
1136 | 1136 | deltacomputer.finddeltainfo(revinfo, fh, target_rev=rev) |
|
1137 | 1137 | |
|
1138 | 1138 | timer(find_one) |
|
1139 | 1139 | fm.end() |
|
1140 | 1140 | |
|
1141 | 1141 | |
|
1142 | 1142 | @command(b'perf::discovery|perfdiscovery', formatteropts, b'PATH') |
|
1143 | 1143 | def perfdiscovery(ui, repo, path, **opts): |
|
1144 | 1144 | """benchmark discovery between local repo and the peer at given path""" |
|
1145 | 1145 | repos = [repo, None] |
|
1146 | 1146 | timer, fm = gettimer(ui, opts) |
|
1147 | 1147 | |
|
1148 | 1148 | try: |
|
1149 | 1149 | from mercurial.utils.urlutil import get_unique_pull_path_obj |
|
1150 | 1150 | |
|
1151 | 1151 | path = get_unique_pull_path_obj(b'perfdiscovery', ui, path) |
|
1152 | 1152 | except ImportError: |
|
1153 | 1153 | try: |
|
1154 | 1154 | from mercurial.utils.urlutil import get_unique_pull_path |
|
1155 | 1155 | |
|
1156 | 1156 | path = get_unique_pull_path(b'perfdiscovery', repo, ui, path)[0] |
|
1157 | 1157 | except ImportError: |
|
1158 | 1158 | path = ui.expandpath(path) |
|
1159 | 1159 | |
|
1160 | 1160 | def s(): |
|
1161 | 1161 | repos[1] = hg.peer(ui, opts, path) |
|
1162 | 1162 | |
|
1163 | 1163 | def d(): |
|
1164 | 1164 | setdiscovery.findcommonheads(ui, *repos) |
|
1165 | 1165 | |
|
1166 | 1166 | timer(d, setup=s) |
|
1167 | 1167 | fm.end() |
|
1168 | 1168 | |
|
1169 | 1169 | |
|
1170 | 1170 | @command( |
|
1171 | 1171 | b'perf::bookmarks|perfbookmarks', |
|
1172 | 1172 | formatteropts |
|
1173 | 1173 | + [ |
|
1174 | 1174 | (b'', b'clear-revlogs', False, b'refresh changelog and manifest'), |
|
1175 | 1175 | ], |
|
1176 | 1176 | ) |
|
1177 | 1177 | def perfbookmarks(ui, repo, **opts): |
|
1178 | 1178 | """benchmark parsing bookmarks from disk to memory""" |
|
1179 | 1179 | opts = _byteskwargs(opts) |
|
1180 | 1180 | timer, fm = gettimer(ui, opts) |
|
1181 | 1181 | |
|
1182 | 1182 | clearrevlogs = opts[b'clear_revlogs'] |
|
1183 | 1183 | |
|
1184 | 1184 | def s(): |
|
1185 | 1185 | if clearrevlogs: |
|
1186 | 1186 | clearchangelog(repo) |
|
1187 | 1187 | clearfilecache(repo, b'_bookmarks') |
|
1188 | 1188 | |
|
1189 | 1189 | def d(): |
|
1190 | 1190 | repo._bookmarks |
|
1191 | 1191 | |
|
1192 | 1192 | timer(d, setup=s) |
|
1193 | 1193 | fm.end() |
|
1194 | 1194 | |
|
1195 | 1195 | |
|
1196 | 1196 | @command( |
|
1197 | 1197 | b'perf::bundle', |
|
1198 | 1198 | [ |
|
1199 | 1199 | ( |
|
1200 | 1200 | b'r', |
|
1201 | 1201 | b'rev', |
|
1202 | 1202 | [], |
|
1203 | 1203 | b'changesets to bundle', |
|
1204 | 1204 | b'REV', |
|
1205 | 1205 | ), |
|
1206 | 1206 | ( |
|
1207 | 1207 | b't', |
|
1208 | 1208 | b'type', |
|
1209 | 1209 | b'none', |
|
1210 | 1210 | b'bundlespec to use (see `hg help bundlespec`)', |
|
1211 | 1211 | b'TYPE', |
|
1212 | 1212 | ), |
|
1213 | 1213 | ] |
|
1214 | 1214 | + formatteropts, |
|
1215 | 1215 | b'REVS', |
|
1216 | 1216 | ) |
|
1217 | 1217 | def perfbundle(ui, repo, *revs, **opts): |
|
1218 | 1218 | """benchmark the creation of a bundle from a repository |
|
1219 | 1219 | |
|
1220 | 1220 | For now, this only supports "none" compression. |
|
1221 | 1221 | """ |
|
1222 | 1222 | try: |
|
1223 | 1223 | from mercurial import bundlecaches |
|
1224 | 1224 | |
|
1225 | 1225 | parsebundlespec = bundlecaches.parsebundlespec |
|
1226 | 1226 | except ImportError: |
|
1227 | 1227 | from mercurial import exchange |
|
1228 | 1228 | |
|
1229 | 1229 | parsebundlespec = exchange.parsebundlespec |
|
1230 | 1230 | |
|
1231 | 1231 | from mercurial import discovery |
|
1232 | 1232 | from mercurial import bundle2 |
|
1233 | 1233 | |
|
1234 | 1234 | opts = _byteskwargs(opts) |
|
1235 | 1235 | timer, fm = gettimer(ui, opts) |
|
1236 | 1236 | |
|
1237 | 1237 | cl = repo.changelog |
|
1238 | 1238 | revs = list(revs) |
|
1239 | 1239 | revs.extend(opts.get(b'rev', ())) |
|
1240 | 1240 | revs = scmutil.revrange(repo, revs) |
|
1241 | 1241 | if not revs: |
|
1242 | 1242 | raise error.Abort(b"not revision specified") |
|
1243 | 1243 | # make it a consistent set (ie: without topological gaps) |
|
1244 | 1244 | old_len = len(revs) |
|
1245 | 1245 | revs = list(repo.revs(b"%ld::%ld", revs, revs)) |
|
1246 | 1246 | if old_len != len(revs): |
|
1247 | 1247 | new_count = len(revs) - old_len |
|
1248 | 1248 | msg = b"add %d new revisions to make it a consistent set\n" |
|
1249 | 1249 | ui.write_err(msg % new_count) |
|
1250 | 1250 | |
|
1251 | 1251 | targets = [cl.node(r) for r in repo.revs(b"heads(::%ld)", revs)] |
|
1252 | 1252 | bases = [cl.node(r) for r in repo.revs(b"heads(::%ld - %ld)", revs, revs)] |
|
1253 | 1253 | outgoing = discovery.outgoing(repo, bases, targets) |
|
1254 | 1254 | |
|
1255 | 1255 | bundle_spec = opts.get(b'type') |
|
1256 | 1256 | |
|
1257 | 1257 | bundle_spec = parsebundlespec(repo, bundle_spec, strict=False) |
|
1258 | 1258 | |
|
1259 | 1259 | cgversion = bundle_spec.params.get(b"cg.version") |
|
1260 | 1260 | if cgversion is None: |
|
1261 | 1261 | if bundle_spec.version == b'v1': |
|
1262 | 1262 | cgversion = b'01' |
|
1263 | 1263 | if bundle_spec.version == b'v2': |
|
1264 | 1264 | cgversion = b'02' |
|
1265 | 1265 | if cgversion not in changegroup.supportedoutgoingversions(repo): |
|
1266 | 1266 | err = b"repository does not support bundle version %s" |
|
1267 | 1267 | raise error.Abort(err % cgversion) |
|
1268 | 1268 | |
|
1269 | 1269 | if cgversion == b'01': # bundle1 |
|
1270 | 1270 | bversion = b'HG10' + bundle_spec.wirecompression |
|
1271 | 1271 | bcompression = None |
|
1272 | 1272 | elif cgversion in (b'02', b'03'): |
|
1273 | 1273 | bversion = b'HG20' |
|
1274 | 1274 | bcompression = bundle_spec.wirecompression |
|
1275 | 1275 | else: |
|
1276 | 1276 | err = b'perf::bundle: unexpected changegroup version %s' |
|
1277 | 1277 | raise error.ProgrammingError(err % cgversion) |
|
1278 | 1278 | |
|
1279 | 1279 | if bcompression is None: |
|
1280 | 1280 | bcompression = b'UN' |
|
1281 | 1281 | |
|
1282 | 1282 | if bcompression != b'UN': |
|
1283 | 1283 | err = b'perf::bundle: compression currently unsupported: %s' |
|
1284 | 1284 | raise error.ProgrammingError(err % bcompression) |
|
1285 | 1285 | |
|
1286 | 1286 | def do_bundle(): |
|
1287 | 1287 | bundle2.writenewbundle( |
|
1288 | 1288 | ui, |
|
1289 | 1289 | repo, |
|
1290 | 1290 | b'perf::bundle', |
|
1291 | 1291 | os.devnull, |
|
1292 | 1292 | bversion, |
|
1293 | 1293 | outgoing, |
|
1294 | 1294 | bundle_spec.params, |
|
1295 | 1295 | ) |
|
1296 | 1296 | |
|
1297 | 1297 | timer(do_bundle) |
|
1298 | 1298 | fm.end() |
|
1299 | 1299 | |
|
1300 | 1300 | |
|
1301 | 1301 | @command(b'perf::bundleread|perfbundleread', formatteropts, b'BUNDLE') |
|
1302 | 1302 | def perfbundleread(ui, repo, bundlepath, **opts): |
|
1303 | 1303 | """Benchmark reading of bundle files. |
|
1304 | 1304 | |
|
1305 | 1305 | This command is meant to isolate the I/O part of bundle reading as |
|
1306 | 1306 | much as possible. |
|
1307 | 1307 | """ |
|
1308 | 1308 | from mercurial import ( |
|
1309 | 1309 | bundle2, |
|
1310 | 1310 | exchange, |
|
1311 | 1311 | streamclone, |
|
1312 | 1312 | ) |
|
1313 | 1313 | |
|
1314 | 1314 | opts = _byteskwargs(opts) |
|
1315 | 1315 | |
|
1316 | 1316 | def makebench(fn): |
|
1317 | 1317 | def run(): |
|
1318 | 1318 | with open(bundlepath, b'rb') as fh: |
|
1319 | 1319 | bundle = exchange.readbundle(ui, fh, bundlepath) |
|
1320 | 1320 | fn(bundle) |
|
1321 | 1321 | |
|
1322 | 1322 | return run |
|
1323 | 1323 | |
|
1324 | 1324 | def makereadnbytes(size): |
|
1325 | 1325 | def run(): |
|
1326 | 1326 | with open(bundlepath, b'rb') as fh: |
|
1327 | 1327 | bundle = exchange.readbundle(ui, fh, bundlepath) |
|
1328 | 1328 | while bundle.read(size): |
|
1329 | 1329 | pass |
|
1330 | 1330 | |
|
1331 | 1331 | return run |
|
1332 | 1332 | |
|
1333 | 1333 | def makestdioread(size): |
|
1334 | 1334 | def run(): |
|
1335 | 1335 | with open(bundlepath, b'rb') as fh: |
|
1336 | 1336 | while fh.read(size): |
|
1337 | 1337 | pass |
|
1338 | 1338 | |
|
1339 | 1339 | return run |
|
1340 | 1340 | |
|
1341 | 1341 | # bundle1 |
|
1342 | 1342 | |
|
1343 | 1343 | def deltaiter(bundle): |
|
1344 | 1344 | for delta in bundle.deltaiter(): |
|
1345 | 1345 | pass |
|
1346 | 1346 | |
|
1347 | 1347 | def iterchunks(bundle): |
|
1348 | 1348 | for chunk in bundle.getchunks(): |
|
1349 | 1349 | pass |
|
1350 | 1350 | |
|
1351 | 1351 | # bundle2 |
|
1352 | 1352 | |
|
1353 | 1353 | def forwardchunks(bundle): |
|
1354 | 1354 | for chunk in bundle._forwardchunks(): |
|
1355 | 1355 | pass |
|
1356 | 1356 | |
|
1357 | 1357 | def iterparts(bundle): |
|
1358 | 1358 | for part in bundle.iterparts(): |
|
1359 | 1359 | pass |
|
1360 | 1360 | |
|
1361 | 1361 | def iterpartsseekable(bundle): |
|
1362 | 1362 | for part in bundle.iterparts(seekable=True): |
|
1363 | 1363 | pass |
|
1364 | 1364 | |
|
1365 | 1365 | def seek(bundle): |
|
1366 | 1366 | for part in bundle.iterparts(seekable=True): |
|
1367 | 1367 | part.seek(0, os.SEEK_END) |
|
1368 | 1368 | |
|
1369 | 1369 | def makepartreadnbytes(size): |
|
1370 | 1370 | def run(): |
|
1371 | 1371 | with open(bundlepath, b'rb') as fh: |
|
1372 | 1372 | bundle = exchange.readbundle(ui, fh, bundlepath) |
|
1373 | 1373 | for part in bundle.iterparts(): |
|
1374 | 1374 | while part.read(size): |
|
1375 | 1375 | pass |
|
1376 | 1376 | |
|
1377 | 1377 | return run |
|
1378 | 1378 | |
|
1379 | 1379 | benches = [ |
|
1380 | 1380 | (makestdioread(8192), b'read(8k)'), |
|
1381 | 1381 | (makestdioread(16384), b'read(16k)'), |
|
1382 | 1382 | (makestdioread(32768), b'read(32k)'), |
|
1383 | 1383 | (makestdioread(131072), b'read(128k)'), |
|
1384 | 1384 | ] |
|
1385 | 1385 | |
|
1386 | 1386 | with open(bundlepath, b'rb') as fh: |
|
1387 | 1387 | bundle = exchange.readbundle(ui, fh, bundlepath) |
|
1388 | 1388 | |
|
1389 | 1389 | if isinstance(bundle, changegroup.cg1unpacker): |
|
1390 | 1390 | benches.extend( |
|
1391 | 1391 | [ |
|
1392 | 1392 | (makebench(deltaiter), b'cg1 deltaiter()'), |
|
1393 | 1393 | (makebench(iterchunks), b'cg1 getchunks()'), |
|
1394 | 1394 | (makereadnbytes(8192), b'cg1 read(8k)'), |
|
1395 | 1395 | (makereadnbytes(16384), b'cg1 read(16k)'), |
|
1396 | 1396 | (makereadnbytes(32768), b'cg1 read(32k)'), |
|
1397 | 1397 | (makereadnbytes(131072), b'cg1 read(128k)'), |
|
1398 | 1398 | ] |
|
1399 | 1399 | ) |
|
1400 | 1400 | elif isinstance(bundle, bundle2.unbundle20): |
|
1401 | 1401 | benches.extend( |
|
1402 | 1402 | [ |
|
1403 | 1403 | (makebench(forwardchunks), b'bundle2 forwardchunks()'), |
|
1404 | 1404 | (makebench(iterparts), b'bundle2 iterparts()'), |
|
1405 | 1405 | ( |
|
1406 | 1406 | makebench(iterpartsseekable), |
|
1407 | 1407 | b'bundle2 iterparts() seekable', |
|
1408 | 1408 | ), |
|
1409 | 1409 | (makebench(seek), b'bundle2 part seek()'), |
|
1410 | 1410 | (makepartreadnbytes(8192), b'bundle2 part read(8k)'), |
|
1411 | 1411 | (makepartreadnbytes(16384), b'bundle2 part read(16k)'), |
|
1412 | 1412 | (makepartreadnbytes(32768), b'bundle2 part read(32k)'), |
|
1413 | 1413 | (makepartreadnbytes(131072), b'bundle2 part read(128k)'), |
|
1414 | 1414 | ] |
|
1415 | 1415 | ) |
|
1416 | 1416 | elif isinstance(bundle, streamclone.streamcloneapplier): |
|
1417 | 1417 | raise error.Abort(b'stream clone bundles not supported') |
|
1418 | 1418 | else: |
|
1419 | 1419 | raise error.Abort(b'unhandled bundle type: %s' % type(bundle)) |
|
1420 | 1420 | |
|
1421 | 1421 | for fn, title in benches: |
|
1422 | 1422 | timer, fm = gettimer(ui, opts) |
|
1423 | 1423 | timer(fn, title=title) |
|
1424 | 1424 | fm.end() |
|
1425 | 1425 | |
|
1426 | 1426 | |
|
1427 | 1427 | @command( |
|
1428 | 1428 | b'perf::changegroupchangelog|perfchangegroupchangelog', |
|
1429 | 1429 | formatteropts |
|
1430 | 1430 | + [ |
|
1431 | 1431 | (b'', b'cgversion', b'02', b'changegroup version'), |
|
1432 | 1432 | (b'r', b'rev', b'', b'revisions to add to changegroup'), |
|
1433 | 1433 | ], |
|
1434 | 1434 | ) |
|
1435 | 1435 | def perfchangegroupchangelog(ui, repo, cgversion=b'02', rev=None, **opts): |
|
1436 | 1436 | """Benchmark producing a changelog group for a changegroup. |
|
1437 | 1437 | |
|
1438 | 1438 | This measures the time spent processing the changelog during a |
|
1439 | 1439 | bundle operation. This occurs during `hg bundle` and on a server |
|
1440 | 1440 | processing a `getbundle` wire protocol request (handles clones |
|
1441 | 1441 | and pull requests). |
|
1442 | 1442 | |
|
1443 | 1443 | By default, all revisions are added to the changegroup. |
|
1444 | 1444 | """ |
|
1445 | 1445 | opts = _byteskwargs(opts) |
|
1446 | 1446 | cl = repo.changelog |
|
1447 | 1447 | nodes = [cl.lookup(r) for r in repo.revs(rev or b'all()')] |
|
1448 | 1448 | bundler = changegroup.getbundler(cgversion, repo) |
|
1449 | 1449 | |
|
1450 | 1450 | def d(): |
|
1451 | 1451 | state, chunks = bundler._generatechangelog(cl, nodes) |
|
1452 | 1452 | for chunk in chunks: |
|
1453 | 1453 | pass |
|
1454 | 1454 | |
|
1455 | 1455 | timer, fm = gettimer(ui, opts) |
|
1456 | 1456 | |
|
1457 | 1457 | # Terminal printing can interfere with timing. So disable it. |
|
1458 | 1458 | with ui.configoverride({(b'progress', b'disable'): True}): |
|
1459 | 1459 | timer(d) |
|
1460 | 1460 | |
|
1461 | 1461 | fm.end() |
|
1462 | 1462 | |
|
1463 | 1463 | |
|
1464 | 1464 | @command(b'perf::dirs|perfdirs', formatteropts) |
|
1465 | 1465 | def perfdirs(ui, repo, **opts): |
|
1466 | 1466 | opts = _byteskwargs(opts) |
|
1467 | 1467 | timer, fm = gettimer(ui, opts) |
|
1468 | 1468 | dirstate = repo.dirstate |
|
1469 | 1469 | b'a' in dirstate |
|
1470 | 1470 | |
|
1471 | 1471 | def d(): |
|
1472 | 1472 | dirstate.hasdir(b'a') |
|
1473 | 1473 | try: |
|
1474 | 1474 | del dirstate._map._dirs |
|
1475 | 1475 | except AttributeError: |
|
1476 | 1476 | pass |
|
1477 | 1477 | |
|
1478 | 1478 | timer(d) |
|
1479 | 1479 | fm.end() |
|
1480 | 1480 | |
|
1481 | 1481 | |
|
1482 | 1482 | @command( |
|
1483 | 1483 | b'perf::dirstate|perfdirstate', |
|
1484 | 1484 | [ |
|
1485 | 1485 | ( |
|
1486 | 1486 | b'', |
|
1487 | 1487 | b'iteration', |
|
1488 | 1488 | None, |
|
1489 | 1489 | b'benchmark a full iteration for the dirstate', |
|
1490 | 1490 | ), |
|
1491 | 1491 | ( |
|
1492 | 1492 | b'', |
|
1493 | 1493 | b'contains', |
|
1494 | 1494 | None, |
|
1495 | 1495 | b'benchmark a large amount of `nf in dirstate` calls', |
|
1496 | 1496 | ), |
|
1497 | 1497 | ] |
|
1498 | 1498 | + formatteropts, |
|
1499 | 1499 | ) |
|
1500 | 1500 | def perfdirstate(ui, repo, **opts): |
|
1501 | 1501 | """benchmap the time of various distate operations |
|
1502 | 1502 | |
|
1503 | 1503 | By default benchmark the time necessary to load a dirstate from scratch. |
|
1504 | 1504 | The dirstate is loaded to the point were a "contains" request can be |
|
1505 | 1505 | answered. |
|
1506 | 1506 | """ |
|
1507 | 1507 | opts = _byteskwargs(opts) |
|
1508 | 1508 | timer, fm = gettimer(ui, opts) |
|
1509 | 1509 | b"a" in repo.dirstate |
|
1510 | 1510 | |
|
1511 | 1511 | if opts[b'iteration'] and opts[b'contains']: |
|
1512 | 1512 | msg = b'only specify one of --iteration or --contains' |
|
1513 | 1513 | raise error.Abort(msg) |
|
1514 | 1514 | |
|
1515 | 1515 | if opts[b'iteration']: |
|
1516 | 1516 | setup = None |
|
1517 | 1517 | dirstate = repo.dirstate |
|
1518 | 1518 | |
|
1519 | 1519 | def d(): |
|
1520 | 1520 | for f in dirstate: |
|
1521 | 1521 | pass |
|
1522 | 1522 | |
|
1523 | 1523 | elif opts[b'contains']: |
|
1524 | 1524 | setup = None |
|
1525 | 1525 | dirstate = repo.dirstate |
|
1526 | 1526 | allfiles = list(dirstate) |
|
1527 | 1527 | # also add file path that will be "missing" from the dirstate |
|
1528 | 1528 | allfiles.extend([f[::-1] for f in allfiles]) |
|
1529 | 1529 | |
|
1530 | 1530 | def d(): |
|
1531 | 1531 | for f in allfiles: |
|
1532 | 1532 | f in dirstate |
|
1533 | 1533 | |
|
1534 | 1534 | else: |
|
1535 | 1535 | |
|
1536 | 1536 | def setup(): |
|
1537 | 1537 | repo.dirstate.invalidate() |
|
1538 | 1538 | |
|
1539 | 1539 | def d(): |
|
1540 | 1540 | b"a" in repo.dirstate |
|
1541 | 1541 | |
|
1542 | 1542 | timer(d, setup=setup) |
|
1543 | 1543 | fm.end() |
|
1544 | 1544 | |
|
1545 | 1545 | |
|
1546 | 1546 | @command(b'perf::dirstatedirs|perfdirstatedirs', formatteropts) |
|
1547 | 1547 | def perfdirstatedirs(ui, repo, **opts): |
|
1548 | 1548 | """benchmap a 'dirstate.hasdir' call from an empty `dirs` cache""" |
|
1549 | 1549 | opts = _byteskwargs(opts) |
|
1550 | 1550 | timer, fm = gettimer(ui, opts) |
|
1551 | 1551 | repo.dirstate.hasdir(b"a") |
|
1552 | 1552 | |
|
1553 | 1553 | def setup(): |
|
1554 | 1554 | try: |
|
1555 | 1555 | del repo.dirstate._map._dirs |
|
1556 | 1556 | except AttributeError: |
|
1557 | 1557 | pass |
|
1558 | 1558 | |
|
1559 | 1559 | def d(): |
|
1560 | 1560 | repo.dirstate.hasdir(b"a") |
|
1561 | 1561 | |
|
1562 | 1562 | timer(d, setup=setup) |
|
1563 | 1563 | fm.end() |
|
1564 | 1564 | |
|
1565 | 1565 | |
|
1566 | 1566 | @command(b'perf::dirstatefoldmap|perfdirstatefoldmap', formatteropts) |
|
1567 | 1567 | def perfdirstatefoldmap(ui, repo, **opts): |
|
1568 | 1568 | """benchmap a `dirstate._map.filefoldmap.get()` request |
|
1569 | 1569 | |
|
1570 | 1570 | The dirstate filefoldmap cache is dropped between every request. |
|
1571 | 1571 | """ |
|
1572 | 1572 | opts = _byteskwargs(opts) |
|
1573 | 1573 | timer, fm = gettimer(ui, opts) |
|
1574 | 1574 | dirstate = repo.dirstate |
|
1575 | 1575 | dirstate._map.filefoldmap.get(b'a') |
|
1576 | 1576 | |
|
1577 | 1577 | def setup(): |
|
1578 | 1578 | del dirstate._map.filefoldmap |
|
1579 | 1579 | |
|
1580 | 1580 | def d(): |
|
1581 | 1581 | dirstate._map.filefoldmap.get(b'a') |
|
1582 | 1582 | |
|
1583 | 1583 | timer(d, setup=setup) |
|
1584 | 1584 | fm.end() |
|
1585 | 1585 | |
|
1586 | 1586 | |
|
1587 | 1587 | @command(b'perf::dirfoldmap|perfdirfoldmap', formatteropts) |
|
1588 | 1588 | def perfdirfoldmap(ui, repo, **opts): |
|
1589 | 1589 | """benchmap a `dirstate._map.dirfoldmap.get()` request |
|
1590 | 1590 | |
|
1591 | 1591 | The dirstate dirfoldmap cache is dropped between every request. |
|
1592 | 1592 | """ |
|
1593 | 1593 | opts = _byteskwargs(opts) |
|
1594 | 1594 | timer, fm = gettimer(ui, opts) |
|
1595 | 1595 | dirstate = repo.dirstate |
|
1596 | 1596 | dirstate._map.dirfoldmap.get(b'a') |
|
1597 | 1597 | |
|
1598 | 1598 | def setup(): |
|
1599 | 1599 | del dirstate._map.dirfoldmap |
|
1600 | 1600 | try: |
|
1601 | 1601 | del dirstate._map._dirs |
|
1602 | 1602 | except AttributeError: |
|
1603 | 1603 | pass |
|
1604 | 1604 | |
|
1605 | 1605 | def d(): |
|
1606 | 1606 | dirstate._map.dirfoldmap.get(b'a') |
|
1607 | 1607 | |
|
1608 | 1608 | timer(d, setup=setup) |
|
1609 | 1609 | fm.end() |
|
1610 | 1610 | |
|
1611 | 1611 | |
|
1612 | 1612 | @command(b'perf::dirstatewrite|perfdirstatewrite', formatteropts) |
|
1613 | 1613 | def perfdirstatewrite(ui, repo, **opts): |
|
1614 | 1614 | """benchmap the time it take to write a dirstate on disk""" |
|
1615 | 1615 | opts = _byteskwargs(opts) |
|
1616 | 1616 | timer, fm = gettimer(ui, opts) |
|
1617 | 1617 | ds = repo.dirstate |
|
1618 | 1618 | b"a" in ds |
|
1619 | 1619 | |
|
1620 | 1620 | def setup(): |
|
1621 | 1621 | ds._dirty = True |
|
1622 | 1622 | |
|
1623 | 1623 | def d(): |
|
1624 | 1624 | ds.write(repo.currenttransaction()) |
|
1625 | 1625 | |
|
1626 | 1626 | with repo.wlock(): |
|
1627 | 1627 | timer(d, setup=setup) |
|
1628 | 1628 | fm.end() |
|
1629 | 1629 | |
|
1630 | 1630 | |
|
1631 | 1631 | def _getmergerevs(repo, opts): |
|
1632 | 1632 | """parse command argument to return rev involved in merge |
|
1633 | 1633 | |
|
1634 | 1634 | input: options dictionnary with `rev`, `from` and `bse` |
|
1635 | 1635 | output: (localctx, otherctx, basectx) |
|
1636 | 1636 | """ |
|
1637 | 1637 | if opts[b'from']: |
|
1638 | 1638 | fromrev = scmutil.revsingle(repo, opts[b'from']) |
|
1639 | 1639 | wctx = repo[fromrev] |
|
1640 | 1640 | else: |
|
1641 | 1641 | wctx = repo[None] |
|
1642 | 1642 | # we don't want working dir files to be stat'd in the benchmark, so |
|
1643 | 1643 | # prime that cache |
|
1644 | 1644 | wctx.dirty() |
|
1645 | 1645 | rctx = scmutil.revsingle(repo, opts[b'rev'], opts[b'rev']) |
|
1646 | 1646 | if opts[b'base']: |
|
1647 | 1647 | fromrev = scmutil.revsingle(repo, opts[b'base']) |
|
1648 | 1648 | ancestor = repo[fromrev] |
|
1649 | 1649 | else: |
|
1650 | 1650 | ancestor = wctx.ancestor(rctx) |
|
1651 | 1651 | return (wctx, rctx, ancestor) |
|
1652 | 1652 | |
|
1653 | 1653 | |
|
1654 | 1654 | @command( |
|
1655 | 1655 | b'perf::mergecalculate|perfmergecalculate', |
|
1656 | 1656 | [ |
|
1657 | 1657 | (b'r', b'rev', b'.', b'rev to merge against'), |
|
1658 | 1658 | (b'', b'from', b'', b'rev to merge from'), |
|
1659 | 1659 | (b'', b'base', b'', b'the revision to use as base'), |
|
1660 | 1660 | ] |
|
1661 | 1661 | + formatteropts, |
|
1662 | 1662 | ) |
|
1663 | 1663 | def perfmergecalculate(ui, repo, **opts): |
|
1664 | 1664 | opts = _byteskwargs(opts) |
|
1665 | 1665 | timer, fm = gettimer(ui, opts) |
|
1666 | 1666 | |
|
1667 | 1667 | wctx, rctx, ancestor = _getmergerevs(repo, opts) |
|
1668 | 1668 | |
|
1669 | 1669 | def d(): |
|
1670 | 1670 | # acceptremote is True because we don't want prompts in the middle of |
|
1671 | 1671 | # our benchmark |
|
1672 | 1672 | merge.calculateupdates( |
|
1673 | 1673 | repo, |
|
1674 | 1674 | wctx, |
|
1675 | 1675 | rctx, |
|
1676 | 1676 | [ancestor], |
|
1677 | 1677 | branchmerge=False, |
|
1678 | 1678 | force=False, |
|
1679 | 1679 | acceptremote=True, |
|
1680 | 1680 | followcopies=True, |
|
1681 | 1681 | ) |
|
1682 | 1682 | |
|
1683 | 1683 | timer(d) |
|
1684 | 1684 | fm.end() |
|
1685 | 1685 | |
|
1686 | 1686 | |
|
1687 | 1687 | @command( |
|
1688 | 1688 | b'perf::mergecopies|perfmergecopies', |
|
1689 | 1689 | [ |
|
1690 | 1690 | (b'r', b'rev', b'.', b'rev to merge against'), |
|
1691 | 1691 | (b'', b'from', b'', b'rev to merge from'), |
|
1692 | 1692 | (b'', b'base', b'', b'the revision to use as base'), |
|
1693 | 1693 | ] |
|
1694 | 1694 | + formatteropts, |
|
1695 | 1695 | ) |
|
1696 | 1696 | def perfmergecopies(ui, repo, **opts): |
|
1697 | 1697 | """measure runtime of `copies.mergecopies`""" |
|
1698 | 1698 | opts = _byteskwargs(opts) |
|
1699 | 1699 | timer, fm = gettimer(ui, opts) |
|
1700 | 1700 | wctx, rctx, ancestor = _getmergerevs(repo, opts) |
|
1701 | 1701 | |
|
1702 | 1702 | def d(): |
|
1703 | 1703 | # acceptremote is True because we don't want prompts in the middle of |
|
1704 | 1704 | # our benchmark |
|
1705 | 1705 | copies.mergecopies(repo, wctx, rctx, ancestor) |
|
1706 | 1706 | |
|
1707 | 1707 | timer(d) |
|
1708 | 1708 | fm.end() |
|
1709 | 1709 | |
|
1710 | 1710 | |
|
1711 | 1711 | @command(b'perf::pathcopies|perfpathcopies', [], b"REV REV") |
|
1712 | 1712 | def perfpathcopies(ui, repo, rev1, rev2, **opts): |
|
1713 | 1713 | """benchmark the copy tracing logic""" |
|
1714 | 1714 | opts = _byteskwargs(opts) |
|
1715 | 1715 | timer, fm = gettimer(ui, opts) |
|
1716 | 1716 | ctx1 = scmutil.revsingle(repo, rev1, rev1) |
|
1717 | 1717 | ctx2 = scmutil.revsingle(repo, rev2, rev2) |
|
1718 | 1718 | |
|
1719 | 1719 | def d(): |
|
1720 | 1720 | copies.pathcopies(ctx1, ctx2) |
|
1721 | 1721 | |
|
1722 | 1722 | timer(d) |
|
1723 | 1723 | fm.end() |
|
1724 | 1724 | |
|
1725 | 1725 | |
|
1726 | 1726 | @command( |
|
1727 | 1727 | b'perf::phases|perfphases', |
|
1728 | 1728 | [ |
|
1729 | 1729 | (b'', b'full', False, b'include file reading time too'), |
|
1730 | 1730 | ] |
|
1731 | 1731 | + formatteropts, |
|
1732 | 1732 | b"", |
|
1733 | 1733 | ) |
|
1734 | 1734 | def perfphases(ui, repo, **opts): |
|
1735 | 1735 | """benchmark phasesets computation""" |
|
1736 | 1736 | opts = _byteskwargs(opts) |
|
1737 | 1737 | timer, fm = gettimer(ui, opts) |
|
1738 | 1738 | _phases = repo._phasecache |
|
1739 | 1739 | full = opts.get(b'full') |
|
1740 | 1740 | tip_rev = repo.changelog.tiprev() |
|
1741 | 1741 | |
|
1742 | 1742 | def d(): |
|
1743 | 1743 | phases = _phases |
|
1744 | 1744 | if full: |
|
1745 | 1745 | clearfilecache(repo, b'_phasecache') |
|
1746 | 1746 | phases = repo._phasecache |
|
1747 | 1747 | phases.invalidate() |
|
1748 | 1748 | phases.phase(repo, tip_rev) |
|
1749 | 1749 | |
|
1750 | 1750 | timer(d) |
|
1751 | 1751 | fm.end() |
|
1752 | 1752 | |
|
1753 | 1753 | |
|
1754 | 1754 | @command(b'perf::phasesremote|perfphasesremote', [], b"[DEST]") |
|
1755 | 1755 | def perfphasesremote(ui, repo, dest=None, **opts): |
|
1756 | 1756 | """benchmark time needed to analyse phases of the remote server""" |
|
1757 | 1757 | from mercurial.node import bin |
|
1758 | 1758 | from mercurial import ( |
|
1759 | 1759 | exchange, |
|
1760 | 1760 | hg, |
|
1761 | 1761 | phases, |
|
1762 | 1762 | ) |
|
1763 | 1763 | |
|
1764 | 1764 | opts = _byteskwargs(opts) |
|
1765 | 1765 | timer, fm = gettimer(ui, opts) |
|
1766 | 1766 | |
|
1767 | 1767 | path = ui.getpath(dest, default=(b'default-push', b'default')) |
|
1768 | 1768 | if not path: |
|
1769 | 1769 | raise error.Abort( |
|
1770 | 1770 | b'default repository not configured!', |
|
1771 | 1771 | hint=b"see 'hg help config.paths'", |
|
1772 | 1772 | ) |
|
1773 | 1773 | if util.safehasattr(path, 'main_path'): |
|
1774 | 1774 | path = path.get_push_variant() |
|
1775 | 1775 | dest = path.loc |
|
1776 | 1776 | else: |
|
1777 | 1777 | dest = path.pushloc or path.loc |
|
1778 | 1778 | ui.statusnoi18n(b'analysing phase of %s\n' % util.hidepassword(dest)) |
|
1779 | 1779 | other = hg.peer(repo, opts, dest) |
|
1780 | 1780 | |
|
1781 | 1781 | # easier to perform discovery through the operation |
|
1782 | 1782 | op = exchange.pushoperation(repo, other) |
|
1783 | 1783 | exchange._pushdiscoverychangeset(op) |
|
1784 | 1784 | |
|
1785 | 1785 | remotesubset = op.fallbackheads |
|
1786 | 1786 | |
|
1787 | 1787 | with other.commandexecutor() as e: |
|
1788 | 1788 | remotephases = e.callcommand( |
|
1789 | 1789 | b'listkeys', {b'namespace': b'phases'} |
|
1790 | 1790 | ).result() |
|
1791 | 1791 | del other |
|
1792 | 1792 | publishing = remotephases.get(b'publishing', False) |
|
1793 | 1793 | if publishing: |
|
1794 | 1794 | ui.statusnoi18n(b'publishing: yes\n') |
|
1795 | 1795 | else: |
|
1796 | 1796 | ui.statusnoi18n(b'publishing: no\n') |
|
1797 | 1797 | |
|
1798 | 1798 | has_node = getattr(repo.changelog.index, 'has_node', None) |
|
1799 | 1799 | if has_node is None: |
|
1800 | 1800 | has_node = repo.changelog.nodemap.__contains__ |
|
1801 | 1801 | nonpublishroots = 0 |
|
1802 | 1802 | for nhex, phase in remotephases.iteritems(): |
|
1803 | 1803 | if nhex == b'publishing': # ignore data related to publish option |
|
1804 | 1804 | continue |
|
1805 | 1805 | node = bin(nhex) |
|
1806 | 1806 | if has_node(node) and int(phase): |
|
1807 | 1807 | nonpublishroots += 1 |
|
1808 | 1808 | ui.statusnoi18n(b'number of roots: %d\n' % len(remotephases)) |
|
1809 | 1809 | ui.statusnoi18n(b'number of known non public roots: %d\n' % nonpublishroots) |
|
1810 | 1810 | |
|
1811 | 1811 | def d(): |
|
1812 | 1812 | phases.remotephasessummary(repo, remotesubset, remotephases) |
|
1813 | 1813 | |
|
1814 | 1814 | timer(d) |
|
1815 | 1815 | fm.end() |
|
1816 | 1816 | |
|
1817 | 1817 | |
|
1818 | 1818 | @command( |
|
1819 | 1819 | b'perf::manifest|perfmanifest', |
|
1820 | 1820 | [ |
|
1821 | 1821 | (b'm', b'manifest-rev', False, b'Look up a manifest node revision'), |
|
1822 | 1822 | (b'', b'clear-disk', False, b'clear on-disk caches too'), |
|
1823 | 1823 | ] |
|
1824 | 1824 | + formatteropts, |
|
1825 | 1825 | b'REV|NODE', |
|
1826 | 1826 | ) |
|
1827 | 1827 | def perfmanifest(ui, repo, rev, manifest_rev=False, clear_disk=False, **opts): |
|
1828 | 1828 | """benchmark the time to read a manifest from disk and return a usable |
|
1829 | 1829 | dict-like object |
|
1830 | 1830 | |
|
1831 | 1831 | Manifest caches are cleared before retrieval.""" |
|
1832 | 1832 | opts = _byteskwargs(opts) |
|
1833 | 1833 | timer, fm = gettimer(ui, opts) |
|
1834 | 1834 | if not manifest_rev: |
|
1835 | 1835 | ctx = scmutil.revsingle(repo, rev, rev) |
|
1836 | 1836 | t = ctx.manifestnode() |
|
1837 | 1837 | else: |
|
1838 | 1838 | from mercurial.node import bin |
|
1839 | 1839 | |
|
1840 | 1840 | if len(rev) == 40: |
|
1841 | 1841 | t = bin(rev) |
|
1842 | 1842 | else: |
|
1843 | 1843 | try: |
|
1844 | 1844 | rev = int(rev) |
|
1845 | 1845 | |
|
1846 | 1846 | if util.safehasattr(repo.manifestlog, b'getstorage'): |
|
1847 | 1847 | t = repo.manifestlog.getstorage(b'').node(rev) |
|
1848 | 1848 | else: |
|
1849 | 1849 | t = repo.manifestlog._revlog.lookup(rev) |
|
1850 | 1850 | except ValueError: |
|
1851 | 1851 | raise error.Abort( |
|
1852 | 1852 | b'manifest revision must be integer or full node' |
|
1853 | 1853 | ) |
|
1854 | 1854 | |
|
1855 | 1855 | def d(): |
|
1856 | 1856 | repo.manifestlog.clearcaches(clear_persisted_data=clear_disk) |
|
1857 | 1857 | repo.manifestlog[t].read() |
|
1858 | 1858 | |
|
1859 | 1859 | timer(d) |
|
1860 | 1860 | fm.end() |
|
1861 | 1861 | |
|
1862 | 1862 | |
|
1863 | 1863 | @command(b'perf::changeset|perfchangeset', formatteropts) |
|
1864 | 1864 | def perfchangeset(ui, repo, rev, **opts): |
|
1865 | 1865 | opts = _byteskwargs(opts) |
|
1866 | 1866 | timer, fm = gettimer(ui, opts) |
|
1867 | 1867 | n = scmutil.revsingle(repo, rev).node() |
|
1868 | 1868 | |
|
1869 | 1869 | def d(): |
|
1870 | 1870 | repo.changelog.read(n) |
|
1871 | 1871 | # repo.changelog._cache = None |
|
1872 | 1872 | |
|
1873 | 1873 | timer(d) |
|
1874 | 1874 | fm.end() |
|
1875 | 1875 | |
|
1876 | 1876 | |
|
1877 | 1877 | @command(b'perf::ignore|perfignore', formatteropts) |
|
1878 | 1878 | def perfignore(ui, repo, **opts): |
|
1879 | 1879 | """benchmark operation related to computing ignore""" |
|
1880 | 1880 | opts = _byteskwargs(opts) |
|
1881 | 1881 | timer, fm = gettimer(ui, opts) |
|
1882 | 1882 | dirstate = repo.dirstate |
|
1883 | 1883 | |
|
1884 | 1884 | def setupone(): |
|
1885 | 1885 | dirstate.invalidate() |
|
1886 | 1886 | clearfilecache(dirstate, b'_ignore') |
|
1887 | 1887 | |
|
1888 | 1888 | def runone(): |
|
1889 | 1889 | dirstate._ignore |
|
1890 | 1890 | |
|
1891 | 1891 | timer(runone, setup=setupone, title=b"load") |
|
1892 | 1892 | fm.end() |
|
1893 | 1893 | |
|
1894 | 1894 | |
|
1895 | 1895 | @command( |
|
1896 | 1896 | b'perf::index|perfindex', |
|
1897 | 1897 | [ |
|
1898 | 1898 | (b'', b'rev', [], b'revision to be looked up (default tip)'), |
|
1899 | 1899 | (b'', b'no-lookup', None, b'do not revision lookup post creation'), |
|
1900 | 1900 | ] |
|
1901 | 1901 | + formatteropts, |
|
1902 | 1902 | ) |
|
1903 | 1903 | def perfindex(ui, repo, **opts): |
|
1904 | 1904 | """benchmark index creation time followed by a lookup |
|
1905 | 1905 | |
|
1906 | 1906 | The default is to look `tip` up. Depending on the index implementation, |
|
1907 | 1907 | the revision looked up can matters. For example, an implementation |
|
1908 | 1908 | scanning the index will have a faster lookup time for `--rev tip` than for |
|
1909 | 1909 | `--rev 0`. The number of looked up revisions and their order can also |
|
1910 | 1910 | matters. |
|
1911 | 1911 | |
|
1912 | 1912 | Example of useful set to test: |
|
1913 | 1913 | |
|
1914 | 1914 | * tip |
|
1915 | 1915 | * 0 |
|
1916 | 1916 | * -10: |
|
1917 | 1917 | * :10 |
|
1918 | 1918 | * -10: + :10 |
|
1919 | 1919 | * :10: + -10: |
|
1920 | 1920 | * -10000: |
|
1921 | 1921 | * -10000: + 0 |
|
1922 | 1922 | |
|
1923 | 1923 | It is not currently possible to check for lookup of a missing node. For |
|
1924 | 1924 | deeper lookup benchmarking, checkout the `perfnodemap` command.""" |
|
1925 | 1925 | import mercurial.revlog |
|
1926 | 1926 | |
|
1927 | 1927 | opts = _byteskwargs(opts) |
|
1928 | 1928 | timer, fm = gettimer(ui, opts) |
|
1929 | 1929 | mercurial.revlog._prereadsize = 2 ** 24 # disable lazy parser in old hg |
|
1930 | 1930 | if opts[b'no_lookup']: |
|
1931 | 1931 | if opts['rev']: |
|
1932 | 1932 | raise error.Abort('--no-lookup and --rev are mutually exclusive') |
|
1933 | 1933 | nodes = [] |
|
1934 | 1934 | elif not opts[b'rev']: |
|
1935 | 1935 | nodes = [repo[b"tip"].node()] |
|
1936 | 1936 | else: |
|
1937 | 1937 | revs = scmutil.revrange(repo, opts[b'rev']) |
|
1938 | 1938 | cl = repo.changelog |
|
1939 | 1939 | nodes = [cl.node(r) for r in revs] |
|
1940 | 1940 | |
|
1941 | 1941 | unfi = repo.unfiltered() |
|
1942 | 1942 | # find the filecache func directly |
|
1943 | 1943 | # This avoid polluting the benchmark with the filecache logic |
|
1944 | 1944 | makecl = unfi.__class__.changelog.func |
|
1945 | 1945 | |
|
1946 | 1946 | def setup(): |
|
1947 | 1947 | # probably not necessary, but for good measure |
|
1948 | 1948 | clearchangelog(unfi) |
|
1949 | 1949 | |
|
1950 | 1950 | def d(): |
|
1951 | 1951 | cl = makecl(unfi) |
|
1952 | 1952 | for n in nodes: |
|
1953 | 1953 | cl.rev(n) |
|
1954 | 1954 | |
|
1955 | 1955 | timer(d, setup=setup) |
|
1956 | 1956 | fm.end() |
|
1957 | 1957 | |
|
1958 | 1958 | |
|
1959 | 1959 | @command( |
|
1960 | 1960 | b'perf::nodemap|perfnodemap', |
|
1961 | 1961 | [ |
|
1962 | 1962 | (b'', b'rev', [], b'revision to be looked up (default tip)'), |
|
1963 | 1963 | (b'', b'clear-caches', True, b'clear revlog cache between calls'), |
|
1964 | 1964 | ] |
|
1965 | 1965 | + formatteropts, |
|
1966 | 1966 | ) |
|
1967 | 1967 | def perfnodemap(ui, repo, **opts): |
|
1968 | 1968 | """benchmark the time necessary to look up revision from a cold nodemap |
|
1969 | 1969 | |
|
1970 | 1970 | Depending on the implementation, the amount and order of revision we look |
|
1971 | 1971 | up can varies. Example of useful set to test: |
|
1972 | 1972 | * tip |
|
1973 | 1973 | * 0 |
|
1974 | 1974 | * -10: |
|
1975 | 1975 | * :10 |
|
1976 | 1976 | * -10: + :10 |
|
1977 | 1977 | * :10: + -10: |
|
1978 | 1978 | * -10000: |
|
1979 | 1979 | * -10000: + 0 |
|
1980 | 1980 | |
|
1981 | 1981 | The command currently focus on valid binary lookup. Benchmarking for |
|
1982 | 1982 | hexlookup, prefix lookup and missing lookup would also be valuable. |
|
1983 | 1983 | """ |
|
1984 | 1984 | import mercurial.revlog |
|
1985 | 1985 | |
|
1986 | 1986 | opts = _byteskwargs(opts) |
|
1987 | 1987 | timer, fm = gettimer(ui, opts) |
|
1988 | 1988 | mercurial.revlog._prereadsize = 2 ** 24 # disable lazy parser in old hg |
|
1989 | 1989 | |
|
1990 | 1990 | unfi = repo.unfiltered() |
|
1991 | 1991 | clearcaches = opts[b'clear_caches'] |
|
1992 | 1992 | # find the filecache func directly |
|
1993 | 1993 | # This avoid polluting the benchmark with the filecache logic |
|
1994 | 1994 | makecl = unfi.__class__.changelog.func |
|
1995 | 1995 | if not opts[b'rev']: |
|
1996 | 1996 | raise error.Abort(b'use --rev to specify revisions to look up') |
|
1997 | 1997 | revs = scmutil.revrange(repo, opts[b'rev']) |
|
1998 | 1998 | cl = repo.changelog |
|
1999 | 1999 | nodes = [cl.node(r) for r in revs] |
|
2000 | 2000 | |
|
2001 | 2001 | # use a list to pass reference to a nodemap from one closure to the next |
|
2002 | 2002 | nodeget = [None] |
|
2003 | 2003 | |
|
2004 | 2004 | def setnodeget(): |
|
2005 | 2005 | # probably not necessary, but for good measure |
|
2006 | 2006 | clearchangelog(unfi) |
|
2007 | 2007 | cl = makecl(unfi) |
|
2008 | 2008 | if util.safehasattr(cl.index, 'get_rev'): |
|
2009 | 2009 | nodeget[0] = cl.index.get_rev |
|
2010 | 2010 | else: |
|
2011 | 2011 | nodeget[0] = cl.nodemap.get |
|
2012 | 2012 | |
|
2013 | 2013 | def d(): |
|
2014 | 2014 | get = nodeget[0] |
|
2015 | 2015 | for n in nodes: |
|
2016 | 2016 | get(n) |
|
2017 | 2017 | |
|
2018 | 2018 | setup = None |
|
2019 | 2019 | if clearcaches: |
|
2020 | 2020 | |
|
2021 | 2021 | def setup(): |
|
2022 | 2022 | setnodeget() |
|
2023 | 2023 | |
|
2024 | 2024 | else: |
|
2025 | 2025 | setnodeget() |
|
2026 | 2026 | d() # prewarm the data structure |
|
2027 | 2027 | timer(d, setup=setup) |
|
2028 | 2028 | fm.end() |
|
2029 | 2029 | |
|
2030 | 2030 | |
|
2031 | 2031 | @command(b'perf::startup|perfstartup', formatteropts) |
|
2032 | 2032 | def perfstartup(ui, repo, **opts): |
|
2033 | 2033 | opts = _byteskwargs(opts) |
|
2034 | 2034 | timer, fm = gettimer(ui, opts) |
|
2035 | 2035 | |
|
2036 | 2036 | def d(): |
|
2037 | 2037 | if os.name != 'nt': |
|
2038 | 2038 | os.system( |
|
2039 | 2039 | b"HGRCPATH= %s version -q > /dev/null" % fsencode(sys.argv[0]) |
|
2040 | 2040 | ) |
|
2041 | 2041 | else: |
|
2042 | 2042 | os.environ['HGRCPATH'] = r' ' |
|
2043 | 2043 | os.system("%s version -q > NUL" % sys.argv[0]) |
|
2044 | 2044 | |
|
2045 | 2045 | timer(d) |
|
2046 | 2046 | fm.end() |
|
2047 | 2047 | |
|
2048 | 2048 | |
|
2049 | def _clear_store_audit_cache(repo): | |
|
2050 | vfs = getsvfs(repo) | |
|
2051 | # unwrap the fncache proxy | |
|
2052 | if not hasattr(vfs, "audit"): | |
|
2053 | vfs = getattr(vfs, "vfs", vfs) | |
|
2054 | auditor = vfs.audit | |
|
2055 | if hasattr(auditor, "clear_audit_cache"): | |
|
2056 | auditor.clear_audit_cache() | |
|
2057 | elif hasattr(auditor, "audited"): | |
|
2058 | auditor.audited.clear() | |
|
2059 | auditor.auditeddir.clear() | |
|
2060 | ||
|
2061 | ||
|
2049 | 2062 | def _find_stream_generator(version): |
|
2050 | 2063 | """find the proper generator function for this stream version""" |
|
2051 | 2064 | import mercurial.streamclone |
|
2052 | 2065 | |
|
2053 | 2066 | available = {} |
|
2054 | 2067 | |
|
2055 | 2068 | # try to fetch a v1 generator |
|
2056 | 2069 | generatev1 = getattr(mercurial.streamclone, "generatev1", None) |
|
2057 | 2070 | if generatev1 is not None: |
|
2058 | 2071 | |
|
2059 | 2072 | def generate(repo): |
|
2060 | 2073 | entries, bytes, data = generatev1(repo, None, None, True) |
|
2061 | 2074 | return data |
|
2062 | 2075 | |
|
2063 | 2076 | available[b'v1'] = generatev1 |
|
2064 | 2077 | # try to fetch a v2 generator |
|
2065 | 2078 | generatev2 = getattr(mercurial.streamclone, "generatev2", None) |
|
2066 | 2079 | if generatev2 is not None: |
|
2067 | 2080 | |
|
2068 | 2081 | def generate(repo): |
|
2069 | 2082 | entries, bytes, data = generatev2(repo, None, None, True) |
|
2070 | 2083 | return data |
|
2071 | 2084 | |
|
2072 | 2085 | available[b'v2'] = generate |
|
2073 | 2086 | # try to fetch a v3 generator |
|
2074 | 2087 | generatev3 = getattr(mercurial.streamclone, "generatev3", None) |
|
2075 | 2088 | if generatev3 is not None: |
|
2076 | 2089 | |
|
2077 | 2090 | def generate(repo): |
|
2078 | 2091 | return generatev3(repo, None, None, True) |
|
2079 | 2092 | |
|
2080 | 2093 | available[b'v3-exp'] = generate |
|
2081 | 2094 | |
|
2082 | 2095 | # resolve the request |
|
2083 | 2096 | if version == b"latest": |
|
2084 | 2097 | # latest is the highest non experimental version |
|
2085 | 2098 | latest_key = max(v for v in available if b'-exp' not in v) |
|
2086 | 2099 | return available[latest_key] |
|
2087 | 2100 | elif version in available: |
|
2088 | 2101 | return available[version] |
|
2089 | 2102 | else: |
|
2090 | 2103 | msg = b"unkown or unavailable version: %s" |
|
2091 | 2104 | msg %= version |
|
2092 | 2105 | hint = b"available versions: %s" |
|
2093 | 2106 | hint %= b', '.join(sorted(available)) |
|
2094 | 2107 | raise error.Abort(msg, hint=hint) |
|
2095 | 2108 | |
|
2096 | 2109 | |
|
2097 | 2110 | @command( |
|
2098 | 2111 | b'perf::stream-locked-section', |
|
2099 | 2112 | [ |
|
2100 | 2113 | ( |
|
2101 | 2114 | b'', |
|
2102 | 2115 | b'stream-version', |
|
2103 | 2116 | b'latest', |
|
2104 | 2117 | b'stream version to use ("v1", "v2", "v3-exp" ' |
|
2105 | 2118 | b'or "latest", (the default))', |
|
2106 | 2119 | ), |
|
2107 | 2120 | ] |
|
2108 | 2121 | + formatteropts, |
|
2109 | 2122 | ) |
|
2110 | 2123 | def perf_stream_clone_scan(ui, repo, stream_version, **opts): |
|
2111 | 2124 | """benchmark the initial, repo-locked, section of a stream-clone""" |
|
2112 | 2125 | |
|
2113 | 2126 | opts = _byteskwargs(opts) |
|
2114 | 2127 | timer, fm = gettimer(ui, opts) |
|
2115 | 2128 | |
|
2116 | 2129 | # deletion of the generator may trigger some cleanup that we do not want to |
|
2117 | 2130 | # measure |
|
2118 | 2131 | result_holder = [None] |
|
2119 | 2132 | |
|
2120 | 2133 | def setupone(): |
|
2121 | 2134 | result_holder[0] = None |
|
2135 | # This is important for the full generation, even if it does not | |
|
2136 | # currently matters, it seems safer to also real it here. | |
|
2137 | _clear_store_audit_cache(repo) | |
|
2122 | 2138 | |
|
2123 | 2139 | generate = _find_stream_generator(stream_version) |
|
2124 | 2140 | |
|
2125 | 2141 | def runone(): |
|
2126 | 2142 | # the lock is held for the duration the initialisation |
|
2127 | 2143 | result_holder[0] = generate(repo) |
|
2128 | 2144 | |
|
2129 | 2145 | timer(runone, setup=setupone, title=b"load") |
|
2130 | 2146 | fm.end() |
|
2131 | 2147 | |
|
2132 | 2148 | |
|
2133 | 2149 | @command( |
|
2134 | 2150 | b'perf::stream-generate', |
|
2135 | 2151 | [ |
|
2136 | 2152 | ( |
|
2137 | 2153 | b'', |
|
2138 | 2154 | b'stream-version', |
|
2139 | 2155 | b'latest', |
|
2140 | 2156 | b'stream version to us ("v1", "v2", "v3-exp" ' |
|
2141 | 2157 | b'or "latest", (the default))', |
|
2142 | 2158 | ), |
|
2143 | 2159 | ] |
|
2144 | 2160 | + formatteropts, |
|
2145 | 2161 | ) |
|
2146 | 2162 | def perf_stream_clone_generate(ui, repo, stream_version, **opts): |
|
2147 | 2163 | """benchmark the full generation of a stream clone""" |
|
2148 | 2164 | |
|
2149 | 2165 | opts = _byteskwargs(opts) |
|
2150 | 2166 | timer, fm = gettimer(ui, opts) |
|
2151 | 2167 | |
|
2152 | 2168 | # deletion of the generator may trigger some cleanup that we do not want to |
|
2153 | 2169 | # measure |
|
2154 | 2170 | |
|
2155 | 2171 | generate = _find_stream_generator(stream_version) |
|
2156 | 2172 | |
|
2173 | def setup(): | |
|
2174 | _clear_store_audit_cache(repo) | |
|
2175 | ||
|
2157 | 2176 | def runone(): |
|
2158 | 2177 | # the lock is held for the duration the initialisation |
|
2159 | 2178 | for chunk in generate(repo): |
|
2160 | 2179 | pass |
|
2161 | 2180 | |
|
2162 | timer(runone, title=b"generate") | |
|
2181 | timer(runone, setup=setup, title=b"generate") | |
|
2163 | 2182 | fm.end() |
|
2164 | 2183 | |
|
2165 | 2184 | |
|
2166 | 2185 | @command( |
|
2167 | 2186 | b'perf::stream-consume', |
|
2168 | 2187 | formatteropts, |
|
2169 | 2188 | ) |
|
2170 | 2189 | def perf_stream_clone_consume(ui, repo, filename, **opts): |
|
2171 | 2190 | """benchmark the full application of a stream clone |
|
2172 | 2191 | |
|
2173 | 2192 | This include the creation of the repository |
|
2174 | 2193 | """ |
|
2175 | 2194 | # try except to appease check code |
|
2176 | 2195 | msg = b"mercurial too old, missing necessary module: %s" |
|
2177 | 2196 | try: |
|
2178 | 2197 | from mercurial import bundle2 |
|
2179 | 2198 | except ImportError as exc: |
|
2180 | 2199 | msg %= _bytestr(exc) |
|
2181 | 2200 | raise error.Abort(msg) |
|
2182 | 2201 | try: |
|
2183 | 2202 | from mercurial import exchange |
|
2184 | 2203 | except ImportError as exc: |
|
2185 | 2204 | msg %= _bytestr(exc) |
|
2186 | 2205 | raise error.Abort(msg) |
|
2187 | 2206 | try: |
|
2188 | 2207 | from mercurial import hg |
|
2189 | 2208 | except ImportError as exc: |
|
2190 | 2209 | msg %= _bytestr(exc) |
|
2191 | 2210 | raise error.Abort(msg) |
|
2192 | 2211 | try: |
|
2193 | 2212 | from mercurial import localrepo |
|
2194 | 2213 | except ImportError as exc: |
|
2195 | 2214 | msg %= _bytestr(exc) |
|
2196 | 2215 | raise error.Abort(msg) |
|
2197 | 2216 | |
|
2198 | 2217 | opts = _byteskwargs(opts) |
|
2199 | 2218 | timer, fm = gettimer(ui, opts) |
|
2200 | 2219 | |
|
2201 | 2220 | # deletion of the generator may trigger some cleanup that we do not want to |
|
2202 | 2221 | # measure |
|
2203 | 2222 | if not (os.path.isfile(filename) and os.access(filename, os.R_OK)): |
|
2204 | 2223 | raise error.Abort("not a readable file: %s" % filename) |
|
2205 | 2224 | |
|
2206 | 2225 | run_variables = [None, None] |
|
2207 | 2226 | |
|
2208 | 2227 | # we create the new repository next to the other one for two reasons: |
|
2209 | 2228 | # - this way we use the same file system, which are relevant for benchmark |
|
2210 | 2229 | # - if /tmp/ is small, the operation could overfills it. |
|
2211 | 2230 | source_repo_dir = os.path.dirname(repo.root) |
|
2212 | 2231 | |
|
2213 | 2232 | @contextlib.contextmanager |
|
2214 | 2233 | def context(): |
|
2215 | 2234 | with open(filename, mode='rb') as bundle: |
|
2216 | 2235 | with tempfile.TemporaryDirectory( |
|
2217 | 2236 | prefix=b'hg-perf-stream-consume-', |
|
2218 | 2237 | dir=source_repo_dir, |
|
2219 | 2238 | ) as tmp_dir: |
|
2220 | 2239 | tmp_dir = fsencode(tmp_dir) |
|
2221 | 2240 | run_variables[0] = bundle |
|
2222 | 2241 | run_variables[1] = tmp_dir |
|
2223 | 2242 | yield |
|
2224 | 2243 | run_variables[0] = None |
|
2225 | 2244 | run_variables[1] = None |
|
2226 | 2245 | |
|
2227 | 2246 | def runone(): |
|
2228 | 2247 | bundle = run_variables[0] |
|
2229 | 2248 | tmp_dir = run_variables[1] |
|
2230 | 2249 | |
|
2231 | 2250 | # we actually wants to copy all config to ensure the repo config is |
|
2232 | 2251 | # taken in account during the benchmark |
|
2233 | 2252 | new_ui = repo.ui.__class__(repo.ui) |
|
2234 | 2253 | # only pass ui when no srcrepo |
|
2235 | 2254 | localrepo.createrepository( |
|
2236 | 2255 | new_ui, tmp_dir, requirements=repo.requirements |
|
2237 | 2256 | ) |
|
2238 | 2257 | target = hg.repository(new_ui, tmp_dir) |
|
2239 | 2258 | gen = exchange.readbundle(target.ui, bundle, bundle.name) |
|
2240 | 2259 | # stream v1 |
|
2241 | 2260 | if util.safehasattr(gen, 'apply'): |
|
2242 | 2261 | gen.apply(target) |
|
2243 | 2262 | else: |
|
2244 | 2263 | with target.transaction(b"perf::stream-consume") as tr: |
|
2245 | 2264 | bundle2.applybundle( |
|
2246 | 2265 | target, |
|
2247 | 2266 | gen, |
|
2248 | 2267 | tr, |
|
2249 | 2268 | source=b'unbundle', |
|
2250 | 2269 | url=filename, |
|
2251 | 2270 | ) |
|
2252 | 2271 | |
|
2253 | 2272 | timer(runone, context=context, title=b"consume") |
|
2254 | 2273 | fm.end() |
|
2255 | 2274 | |
|
2256 | 2275 | |
|
2257 | 2276 | @command(b'perf::parents|perfparents', formatteropts) |
|
2258 | 2277 | def perfparents(ui, repo, **opts): |
|
2259 | 2278 | """benchmark the time necessary to fetch one changeset's parents. |
|
2260 | 2279 | |
|
2261 | 2280 | The fetch is done using the `node identifier`, traversing all object layers |
|
2262 | 2281 | from the repository object. The first N revisions will be used for this |
|
2263 | 2282 | benchmark. N is controlled by the ``perf.parentscount`` config option |
|
2264 | 2283 | (default: 1000). |
|
2265 | 2284 | """ |
|
2266 | 2285 | opts = _byteskwargs(opts) |
|
2267 | 2286 | timer, fm = gettimer(ui, opts) |
|
2268 | 2287 | # control the number of commits perfparents iterates over |
|
2269 | 2288 | # experimental config: perf.parentscount |
|
2270 | 2289 | count = getint(ui, b"perf", b"parentscount", 1000) |
|
2271 | 2290 | if len(repo.changelog) < count: |
|
2272 | 2291 | raise error.Abort(b"repo needs %d commits for this test" % count) |
|
2273 | 2292 | repo = repo.unfiltered() |
|
2274 | 2293 | nl = [repo.changelog.node(i) for i in _xrange(count)] |
|
2275 | 2294 | |
|
2276 | 2295 | def d(): |
|
2277 | 2296 | for n in nl: |
|
2278 | 2297 | repo.changelog.parents(n) |
|
2279 | 2298 | |
|
2280 | 2299 | timer(d) |
|
2281 | 2300 | fm.end() |
|
2282 | 2301 | |
|
2283 | 2302 | |
|
2284 | 2303 | @command(b'perf::ctxfiles|perfctxfiles', formatteropts) |
|
2285 | 2304 | def perfctxfiles(ui, repo, x, **opts): |
|
2286 | 2305 | opts = _byteskwargs(opts) |
|
2287 | 2306 | x = int(x) |
|
2288 | 2307 | timer, fm = gettimer(ui, opts) |
|
2289 | 2308 | |
|
2290 | 2309 | def d(): |
|
2291 | 2310 | len(repo[x].files()) |
|
2292 | 2311 | |
|
2293 | 2312 | timer(d) |
|
2294 | 2313 | fm.end() |
|
2295 | 2314 | |
|
2296 | 2315 | |
|
2297 | 2316 | @command(b'perf::rawfiles|perfrawfiles', formatteropts) |
|
2298 | 2317 | def perfrawfiles(ui, repo, x, **opts): |
|
2299 | 2318 | opts = _byteskwargs(opts) |
|
2300 | 2319 | x = int(x) |
|
2301 | 2320 | timer, fm = gettimer(ui, opts) |
|
2302 | 2321 | cl = repo.changelog |
|
2303 | 2322 | |
|
2304 | 2323 | def d(): |
|
2305 | 2324 | len(cl.read(x)[3]) |
|
2306 | 2325 | |
|
2307 | 2326 | timer(d) |
|
2308 | 2327 | fm.end() |
|
2309 | 2328 | |
|
2310 | 2329 | |
|
2311 | 2330 | @command(b'perf::lookup|perflookup', formatteropts) |
|
2312 | 2331 | def perflookup(ui, repo, rev, **opts): |
|
2313 | 2332 | opts = _byteskwargs(opts) |
|
2314 | 2333 | timer, fm = gettimer(ui, opts) |
|
2315 | 2334 | timer(lambda: len(repo.lookup(rev))) |
|
2316 | 2335 | fm.end() |
|
2317 | 2336 | |
|
2318 | 2337 | |
|
2319 | 2338 | @command( |
|
2320 | 2339 | b'perf::linelogedits|perflinelogedits', |
|
2321 | 2340 | [ |
|
2322 | 2341 | (b'n', b'edits', 10000, b'number of edits'), |
|
2323 | 2342 | (b'', b'max-hunk-lines', 10, b'max lines in a hunk'), |
|
2324 | 2343 | ], |
|
2325 | 2344 | norepo=True, |
|
2326 | 2345 | ) |
|
2327 | 2346 | def perflinelogedits(ui, **opts): |
|
2328 | 2347 | from mercurial import linelog |
|
2329 | 2348 | |
|
2330 | 2349 | opts = _byteskwargs(opts) |
|
2331 | 2350 | |
|
2332 | 2351 | edits = opts[b'edits'] |
|
2333 | 2352 | maxhunklines = opts[b'max_hunk_lines'] |
|
2334 | 2353 | |
|
2335 | 2354 | maxb1 = 100000 |
|
2336 | 2355 | random.seed(0) |
|
2337 | 2356 | randint = random.randint |
|
2338 | 2357 | currentlines = 0 |
|
2339 | 2358 | arglist = [] |
|
2340 | 2359 | for rev in _xrange(edits): |
|
2341 | 2360 | a1 = randint(0, currentlines) |
|
2342 | 2361 | a2 = randint(a1, min(currentlines, a1 + maxhunklines)) |
|
2343 | 2362 | b1 = randint(0, maxb1) |
|
2344 | 2363 | b2 = randint(b1, b1 + maxhunklines) |
|
2345 | 2364 | currentlines += (b2 - b1) - (a2 - a1) |
|
2346 | 2365 | arglist.append((rev, a1, a2, b1, b2)) |
|
2347 | 2366 | |
|
2348 | 2367 | def d(): |
|
2349 | 2368 | ll = linelog.linelog() |
|
2350 | 2369 | for args in arglist: |
|
2351 | 2370 | ll.replacelines(*args) |
|
2352 | 2371 | |
|
2353 | 2372 | timer, fm = gettimer(ui, opts) |
|
2354 | 2373 | timer(d) |
|
2355 | 2374 | fm.end() |
|
2356 | 2375 | |
|
2357 | 2376 | |
|
2358 | 2377 | @command(b'perf::revrange|perfrevrange', formatteropts) |
|
2359 | 2378 | def perfrevrange(ui, repo, *specs, **opts): |
|
2360 | 2379 | opts = _byteskwargs(opts) |
|
2361 | 2380 | timer, fm = gettimer(ui, opts) |
|
2362 | 2381 | revrange = scmutil.revrange |
|
2363 | 2382 | timer(lambda: len(revrange(repo, specs))) |
|
2364 | 2383 | fm.end() |
|
2365 | 2384 | |
|
2366 | 2385 | |
|
2367 | 2386 | @command(b'perf::nodelookup|perfnodelookup', formatteropts) |
|
2368 | 2387 | def perfnodelookup(ui, repo, rev, **opts): |
|
2369 | 2388 | opts = _byteskwargs(opts) |
|
2370 | 2389 | timer, fm = gettimer(ui, opts) |
|
2371 | 2390 | import mercurial.revlog |
|
2372 | 2391 | |
|
2373 | 2392 | mercurial.revlog._prereadsize = 2 ** 24 # disable lazy parser in old hg |
|
2374 | 2393 | n = scmutil.revsingle(repo, rev).node() |
|
2375 | 2394 | |
|
2376 | 2395 | try: |
|
2377 | 2396 | cl = revlog(getsvfs(repo), radix=b"00changelog") |
|
2378 | 2397 | except TypeError: |
|
2379 | 2398 | cl = revlog(getsvfs(repo), indexfile=b"00changelog.i") |
|
2380 | 2399 | |
|
2381 | 2400 | def d(): |
|
2382 | 2401 | cl.rev(n) |
|
2383 | 2402 | clearcaches(cl) |
|
2384 | 2403 | |
|
2385 | 2404 | timer(d) |
|
2386 | 2405 | fm.end() |
|
2387 | 2406 | |
|
2388 | 2407 | |
|
2389 | 2408 | @command( |
|
2390 | 2409 | b'perf::log|perflog', |
|
2391 | 2410 | [(b'', b'rename', False, b'ask log to follow renames')] + formatteropts, |
|
2392 | 2411 | ) |
|
2393 | 2412 | def perflog(ui, repo, rev=None, **opts): |
|
2394 | 2413 | opts = _byteskwargs(opts) |
|
2395 | 2414 | if rev is None: |
|
2396 | 2415 | rev = [] |
|
2397 | 2416 | timer, fm = gettimer(ui, opts) |
|
2398 | 2417 | ui.pushbuffer() |
|
2399 | 2418 | timer( |
|
2400 | 2419 | lambda: commands.log( |
|
2401 | 2420 | ui, repo, rev=rev, date=b'', user=b'', copies=opts.get(b'rename') |
|
2402 | 2421 | ) |
|
2403 | 2422 | ) |
|
2404 | 2423 | ui.popbuffer() |
|
2405 | 2424 | fm.end() |
|
2406 | 2425 | |
|
2407 | 2426 | |
|
2408 | 2427 | @command(b'perf::moonwalk|perfmoonwalk', formatteropts) |
|
2409 | 2428 | def perfmoonwalk(ui, repo, **opts): |
|
2410 | 2429 | """benchmark walking the changelog backwards |
|
2411 | 2430 | |
|
2412 | 2431 | This also loads the changelog data for each revision in the changelog. |
|
2413 | 2432 | """ |
|
2414 | 2433 | opts = _byteskwargs(opts) |
|
2415 | 2434 | timer, fm = gettimer(ui, opts) |
|
2416 | 2435 | |
|
2417 | 2436 | def moonwalk(): |
|
2418 | 2437 | for i in repo.changelog.revs(start=(len(repo) - 1), stop=-1): |
|
2419 | 2438 | ctx = repo[i] |
|
2420 | 2439 | ctx.branch() # read changelog data (in addition to the index) |
|
2421 | 2440 | |
|
2422 | 2441 | timer(moonwalk) |
|
2423 | 2442 | fm.end() |
|
2424 | 2443 | |
|
2425 | 2444 | |
|
2426 | 2445 | @command( |
|
2427 | 2446 | b'perf::templating|perftemplating', |
|
2428 | 2447 | [ |
|
2429 | 2448 | (b'r', b'rev', [], b'revisions to run the template on'), |
|
2430 | 2449 | ] |
|
2431 | 2450 | + formatteropts, |
|
2432 | 2451 | ) |
|
2433 | 2452 | def perftemplating(ui, repo, testedtemplate=None, **opts): |
|
2434 | 2453 | """test the rendering time of a given template""" |
|
2435 | 2454 | if makelogtemplater is None: |
|
2436 | 2455 | raise error.Abort( |
|
2437 | 2456 | b"perftemplating not available with this Mercurial", |
|
2438 | 2457 | hint=b"use 4.3 or later", |
|
2439 | 2458 | ) |
|
2440 | 2459 | |
|
2441 | 2460 | opts = _byteskwargs(opts) |
|
2442 | 2461 | |
|
2443 | 2462 | nullui = ui.copy() |
|
2444 | 2463 | nullui.fout = open(os.devnull, 'wb') |
|
2445 | 2464 | nullui.disablepager() |
|
2446 | 2465 | revs = opts.get(b'rev') |
|
2447 | 2466 | if not revs: |
|
2448 | 2467 | revs = [b'all()'] |
|
2449 | 2468 | revs = list(scmutil.revrange(repo, revs)) |
|
2450 | 2469 | |
|
2451 | 2470 | defaulttemplate = ( |
|
2452 | 2471 | b'{date|shortdate} [{rev}:{node|short}]' |
|
2453 | 2472 | b' {author|person}: {desc|firstline}\n' |
|
2454 | 2473 | ) |
|
2455 | 2474 | if testedtemplate is None: |
|
2456 | 2475 | testedtemplate = defaulttemplate |
|
2457 | 2476 | displayer = makelogtemplater(nullui, repo, testedtemplate) |
|
2458 | 2477 | |
|
2459 | 2478 | def format(): |
|
2460 | 2479 | for r in revs: |
|
2461 | 2480 | ctx = repo[r] |
|
2462 | 2481 | displayer.show(ctx) |
|
2463 | 2482 | displayer.flush(ctx) |
|
2464 | 2483 | |
|
2465 | 2484 | timer, fm = gettimer(ui, opts) |
|
2466 | 2485 | timer(format) |
|
2467 | 2486 | fm.end() |
|
2468 | 2487 | |
|
2469 | 2488 | |
|
2470 | 2489 | def _displaystats(ui, opts, entries, data): |
|
2471 | 2490 | # use a second formatter because the data are quite different, not sure |
|
2472 | 2491 | # how it flies with the templater. |
|
2473 | 2492 | fm = ui.formatter(b'perf-stats', opts) |
|
2474 | 2493 | for key, title in entries: |
|
2475 | 2494 | values = data[key] |
|
2476 | 2495 | nbvalues = len(data) |
|
2477 | 2496 | values.sort() |
|
2478 | 2497 | stats = { |
|
2479 | 2498 | 'key': key, |
|
2480 | 2499 | 'title': title, |
|
2481 | 2500 | 'nbitems': len(values), |
|
2482 | 2501 | 'min': values[0][0], |
|
2483 | 2502 | '10%': values[(nbvalues * 10) // 100][0], |
|
2484 | 2503 | '25%': values[(nbvalues * 25) // 100][0], |
|
2485 | 2504 | '50%': values[(nbvalues * 50) // 100][0], |
|
2486 | 2505 | '75%': values[(nbvalues * 75) // 100][0], |
|
2487 | 2506 | '80%': values[(nbvalues * 80) // 100][0], |
|
2488 | 2507 | '85%': values[(nbvalues * 85) // 100][0], |
|
2489 | 2508 | '90%': values[(nbvalues * 90) // 100][0], |
|
2490 | 2509 | '95%': values[(nbvalues * 95) // 100][0], |
|
2491 | 2510 | '99%': values[(nbvalues * 99) // 100][0], |
|
2492 | 2511 | 'max': values[-1][0], |
|
2493 | 2512 | } |
|
2494 | 2513 | fm.startitem() |
|
2495 | 2514 | fm.data(**stats) |
|
2496 | 2515 | # make node pretty for the human output |
|
2497 | 2516 | fm.plain('### %s (%d items)\n' % (title, len(values))) |
|
2498 | 2517 | lines = [ |
|
2499 | 2518 | 'min', |
|
2500 | 2519 | '10%', |
|
2501 | 2520 | '25%', |
|
2502 | 2521 | '50%', |
|
2503 | 2522 | '75%', |
|
2504 | 2523 | '80%', |
|
2505 | 2524 | '85%', |
|
2506 | 2525 | '90%', |
|
2507 | 2526 | '95%', |
|
2508 | 2527 | '99%', |
|
2509 | 2528 | 'max', |
|
2510 | 2529 | ] |
|
2511 | 2530 | for l in lines: |
|
2512 | 2531 | fm.plain('%s: %s\n' % (l, stats[l])) |
|
2513 | 2532 | fm.end() |
|
2514 | 2533 | |
|
2515 | 2534 | |
|
2516 | 2535 | @command( |
|
2517 | 2536 | b'perf::helper-mergecopies|perfhelper-mergecopies', |
|
2518 | 2537 | formatteropts |
|
2519 | 2538 | + [ |
|
2520 | 2539 | (b'r', b'revs', [], b'restrict search to these revisions'), |
|
2521 | 2540 | (b'', b'timing', False, b'provides extra data (costly)'), |
|
2522 | 2541 | (b'', b'stats', False, b'provides statistic about the measured data'), |
|
2523 | 2542 | ], |
|
2524 | 2543 | ) |
|
2525 | 2544 | def perfhelpermergecopies(ui, repo, revs=[], **opts): |
|
2526 | 2545 | """find statistics about potential parameters for `perfmergecopies` |
|
2527 | 2546 | |
|
2528 | 2547 | This command find (base, p1, p2) triplet relevant for copytracing |
|
2529 | 2548 | benchmarking in the context of a merge. It reports values for some of the |
|
2530 | 2549 | parameters that impact merge copy tracing time during merge. |
|
2531 | 2550 | |
|
2532 | 2551 | If `--timing` is set, rename detection is run and the associated timing |
|
2533 | 2552 | will be reported. The extra details come at the cost of slower command |
|
2534 | 2553 | execution. |
|
2535 | 2554 | |
|
2536 | 2555 | Since rename detection is only run once, other factors might easily |
|
2537 | 2556 | affect the precision of the timing. However it should give a good |
|
2538 | 2557 | approximation of which revision triplets are very costly. |
|
2539 | 2558 | """ |
|
2540 | 2559 | opts = _byteskwargs(opts) |
|
2541 | 2560 | fm = ui.formatter(b'perf', opts) |
|
2542 | 2561 | dotiming = opts[b'timing'] |
|
2543 | 2562 | dostats = opts[b'stats'] |
|
2544 | 2563 | |
|
2545 | 2564 | output_template = [ |
|
2546 | 2565 | ("base", "%(base)12s"), |
|
2547 | 2566 | ("p1", "%(p1.node)12s"), |
|
2548 | 2567 | ("p2", "%(p2.node)12s"), |
|
2549 | 2568 | ("p1.nb-revs", "%(p1.nbrevs)12d"), |
|
2550 | 2569 | ("p1.nb-files", "%(p1.nbmissingfiles)12d"), |
|
2551 | 2570 | ("p1.renames", "%(p1.renamedfiles)12d"), |
|
2552 | 2571 | ("p1.time", "%(p1.time)12.3f"), |
|
2553 | 2572 | ("p2.nb-revs", "%(p2.nbrevs)12d"), |
|
2554 | 2573 | ("p2.nb-files", "%(p2.nbmissingfiles)12d"), |
|
2555 | 2574 | ("p2.renames", "%(p2.renamedfiles)12d"), |
|
2556 | 2575 | ("p2.time", "%(p2.time)12.3f"), |
|
2557 | 2576 | ("renames", "%(nbrenamedfiles)12d"), |
|
2558 | 2577 | ("total.time", "%(time)12.3f"), |
|
2559 | 2578 | ] |
|
2560 | 2579 | if not dotiming: |
|
2561 | 2580 | output_template = [ |
|
2562 | 2581 | i |
|
2563 | 2582 | for i in output_template |
|
2564 | 2583 | if not ('time' in i[0] or 'renames' in i[0]) |
|
2565 | 2584 | ] |
|
2566 | 2585 | header_names = [h for (h, v) in output_template] |
|
2567 | 2586 | output = ' '.join([v for (h, v) in output_template]) + '\n' |
|
2568 | 2587 | header = ' '.join(['%12s'] * len(header_names)) + '\n' |
|
2569 | 2588 | fm.plain(header % tuple(header_names)) |
|
2570 | 2589 | |
|
2571 | 2590 | if not revs: |
|
2572 | 2591 | revs = ['all()'] |
|
2573 | 2592 | revs = scmutil.revrange(repo, revs) |
|
2574 | 2593 | |
|
2575 | 2594 | if dostats: |
|
2576 | 2595 | alldata = { |
|
2577 | 2596 | 'nbrevs': [], |
|
2578 | 2597 | 'nbmissingfiles': [], |
|
2579 | 2598 | } |
|
2580 | 2599 | if dotiming: |
|
2581 | 2600 | alldata['parentnbrenames'] = [] |
|
2582 | 2601 | alldata['totalnbrenames'] = [] |
|
2583 | 2602 | alldata['parenttime'] = [] |
|
2584 | 2603 | alldata['totaltime'] = [] |
|
2585 | 2604 | |
|
2586 | 2605 | roi = repo.revs('merge() and %ld', revs) |
|
2587 | 2606 | for r in roi: |
|
2588 | 2607 | ctx = repo[r] |
|
2589 | 2608 | p1 = ctx.p1() |
|
2590 | 2609 | p2 = ctx.p2() |
|
2591 | 2610 | bases = repo.changelog._commonancestorsheads(p1.rev(), p2.rev()) |
|
2592 | 2611 | for b in bases: |
|
2593 | 2612 | b = repo[b] |
|
2594 | 2613 | p1missing = copies._computeforwardmissing(b, p1) |
|
2595 | 2614 | p2missing = copies._computeforwardmissing(b, p2) |
|
2596 | 2615 | data = { |
|
2597 | 2616 | b'base': b.hex(), |
|
2598 | 2617 | b'p1.node': p1.hex(), |
|
2599 | 2618 | b'p1.nbrevs': len(repo.revs('only(%d, %d)', p1.rev(), b.rev())), |
|
2600 | 2619 | b'p1.nbmissingfiles': len(p1missing), |
|
2601 | 2620 | b'p2.node': p2.hex(), |
|
2602 | 2621 | b'p2.nbrevs': len(repo.revs('only(%d, %d)', p2.rev(), b.rev())), |
|
2603 | 2622 | b'p2.nbmissingfiles': len(p2missing), |
|
2604 | 2623 | } |
|
2605 | 2624 | if dostats: |
|
2606 | 2625 | if p1missing: |
|
2607 | 2626 | alldata['nbrevs'].append( |
|
2608 | 2627 | (data['p1.nbrevs'], b.hex(), p1.hex()) |
|
2609 | 2628 | ) |
|
2610 | 2629 | alldata['nbmissingfiles'].append( |
|
2611 | 2630 | (data['p1.nbmissingfiles'], b.hex(), p1.hex()) |
|
2612 | 2631 | ) |
|
2613 | 2632 | if p2missing: |
|
2614 | 2633 | alldata['nbrevs'].append( |
|
2615 | 2634 | (data['p2.nbrevs'], b.hex(), p2.hex()) |
|
2616 | 2635 | ) |
|
2617 | 2636 | alldata['nbmissingfiles'].append( |
|
2618 | 2637 | (data['p2.nbmissingfiles'], b.hex(), p2.hex()) |
|
2619 | 2638 | ) |
|
2620 | 2639 | if dotiming: |
|
2621 | 2640 | begin = util.timer() |
|
2622 | 2641 | mergedata = copies.mergecopies(repo, p1, p2, b) |
|
2623 | 2642 | end = util.timer() |
|
2624 | 2643 | # not very stable timing since we did only one run |
|
2625 | 2644 | data['time'] = end - begin |
|
2626 | 2645 | # mergedata contains five dicts: "copy", "movewithdir", |
|
2627 | 2646 | # "diverge", "renamedelete" and "dirmove". |
|
2628 | 2647 | # The first 4 are about renamed file so lets count that. |
|
2629 | 2648 | renames = len(mergedata[0]) |
|
2630 | 2649 | renames += len(mergedata[1]) |
|
2631 | 2650 | renames += len(mergedata[2]) |
|
2632 | 2651 | renames += len(mergedata[3]) |
|
2633 | 2652 | data['nbrenamedfiles'] = renames |
|
2634 | 2653 | begin = util.timer() |
|
2635 | 2654 | p1renames = copies.pathcopies(b, p1) |
|
2636 | 2655 | end = util.timer() |
|
2637 | 2656 | data['p1.time'] = end - begin |
|
2638 | 2657 | begin = util.timer() |
|
2639 | 2658 | p2renames = copies.pathcopies(b, p2) |
|
2640 | 2659 | end = util.timer() |
|
2641 | 2660 | data['p2.time'] = end - begin |
|
2642 | 2661 | data['p1.renamedfiles'] = len(p1renames) |
|
2643 | 2662 | data['p2.renamedfiles'] = len(p2renames) |
|
2644 | 2663 | |
|
2645 | 2664 | if dostats: |
|
2646 | 2665 | if p1missing: |
|
2647 | 2666 | alldata['parentnbrenames'].append( |
|
2648 | 2667 | (data['p1.renamedfiles'], b.hex(), p1.hex()) |
|
2649 | 2668 | ) |
|
2650 | 2669 | alldata['parenttime'].append( |
|
2651 | 2670 | (data['p1.time'], b.hex(), p1.hex()) |
|
2652 | 2671 | ) |
|
2653 | 2672 | if p2missing: |
|
2654 | 2673 | alldata['parentnbrenames'].append( |
|
2655 | 2674 | (data['p2.renamedfiles'], b.hex(), p2.hex()) |
|
2656 | 2675 | ) |
|
2657 | 2676 | alldata['parenttime'].append( |
|
2658 | 2677 | (data['p2.time'], b.hex(), p2.hex()) |
|
2659 | 2678 | ) |
|
2660 | 2679 | if p1missing or p2missing: |
|
2661 | 2680 | alldata['totalnbrenames'].append( |
|
2662 | 2681 | ( |
|
2663 | 2682 | data['nbrenamedfiles'], |
|
2664 | 2683 | b.hex(), |
|
2665 | 2684 | p1.hex(), |
|
2666 | 2685 | p2.hex(), |
|
2667 | 2686 | ) |
|
2668 | 2687 | ) |
|
2669 | 2688 | alldata['totaltime'].append( |
|
2670 | 2689 | (data['time'], b.hex(), p1.hex(), p2.hex()) |
|
2671 | 2690 | ) |
|
2672 | 2691 | fm.startitem() |
|
2673 | 2692 | fm.data(**data) |
|
2674 | 2693 | # make node pretty for the human output |
|
2675 | 2694 | out = data.copy() |
|
2676 | 2695 | out['base'] = fm.hexfunc(b.node()) |
|
2677 | 2696 | out['p1.node'] = fm.hexfunc(p1.node()) |
|
2678 | 2697 | out['p2.node'] = fm.hexfunc(p2.node()) |
|
2679 | 2698 | fm.plain(output % out) |
|
2680 | 2699 | |
|
2681 | 2700 | fm.end() |
|
2682 | 2701 | if dostats: |
|
2683 | 2702 | # use a second formatter because the data are quite different, not sure |
|
2684 | 2703 | # how it flies with the templater. |
|
2685 | 2704 | entries = [ |
|
2686 | 2705 | ('nbrevs', 'number of revision covered'), |
|
2687 | 2706 | ('nbmissingfiles', 'number of missing files at head'), |
|
2688 | 2707 | ] |
|
2689 | 2708 | if dotiming: |
|
2690 | 2709 | entries.append( |
|
2691 | 2710 | ('parentnbrenames', 'rename from one parent to base') |
|
2692 | 2711 | ) |
|
2693 | 2712 | entries.append(('totalnbrenames', 'total number of renames')) |
|
2694 | 2713 | entries.append(('parenttime', 'time for one parent')) |
|
2695 | 2714 | entries.append(('totaltime', 'time for both parents')) |
|
2696 | 2715 | _displaystats(ui, opts, entries, alldata) |
|
2697 | 2716 | |
|
2698 | 2717 | |
|
2699 | 2718 | @command( |
|
2700 | 2719 | b'perf::helper-pathcopies|perfhelper-pathcopies', |
|
2701 | 2720 | formatteropts |
|
2702 | 2721 | + [ |
|
2703 | 2722 | (b'r', b'revs', [], b'restrict search to these revisions'), |
|
2704 | 2723 | (b'', b'timing', False, b'provides extra data (costly)'), |
|
2705 | 2724 | (b'', b'stats', False, b'provides statistic about the measured data'), |
|
2706 | 2725 | ], |
|
2707 | 2726 | ) |
|
2708 | 2727 | def perfhelperpathcopies(ui, repo, revs=[], **opts): |
|
2709 | 2728 | """find statistic about potential parameters for the `perftracecopies` |
|
2710 | 2729 | |
|
2711 | 2730 | This command find source-destination pair relevant for copytracing testing. |
|
2712 | 2731 | It report value for some of the parameters that impact copy tracing time. |
|
2713 | 2732 | |
|
2714 | 2733 | If `--timing` is set, rename detection is run and the associated timing |
|
2715 | 2734 | will be reported. The extra details comes at the cost of a slower command |
|
2716 | 2735 | execution. |
|
2717 | 2736 | |
|
2718 | 2737 | Since the rename detection is only run once, other factors might easily |
|
2719 | 2738 | affect the precision of the timing. However it should give a good |
|
2720 | 2739 | approximation of which revision pairs are very costly. |
|
2721 | 2740 | """ |
|
2722 | 2741 | opts = _byteskwargs(opts) |
|
2723 | 2742 | fm = ui.formatter(b'perf', opts) |
|
2724 | 2743 | dotiming = opts[b'timing'] |
|
2725 | 2744 | dostats = opts[b'stats'] |
|
2726 | 2745 | |
|
2727 | 2746 | if dotiming: |
|
2728 | 2747 | header = '%12s %12s %12s %12s %12s %12s\n' |
|
2729 | 2748 | output = ( |
|
2730 | 2749 | "%(source)12s %(destination)12s " |
|
2731 | 2750 | "%(nbrevs)12d %(nbmissingfiles)12d " |
|
2732 | 2751 | "%(nbrenamedfiles)12d %(time)18.5f\n" |
|
2733 | 2752 | ) |
|
2734 | 2753 | header_names = ( |
|
2735 | 2754 | "source", |
|
2736 | 2755 | "destination", |
|
2737 | 2756 | "nb-revs", |
|
2738 | 2757 | "nb-files", |
|
2739 | 2758 | "nb-renames", |
|
2740 | 2759 | "time", |
|
2741 | 2760 | ) |
|
2742 | 2761 | fm.plain(header % header_names) |
|
2743 | 2762 | else: |
|
2744 | 2763 | header = '%12s %12s %12s %12s\n' |
|
2745 | 2764 | output = ( |
|
2746 | 2765 | "%(source)12s %(destination)12s " |
|
2747 | 2766 | "%(nbrevs)12d %(nbmissingfiles)12d\n" |
|
2748 | 2767 | ) |
|
2749 | 2768 | fm.plain(header % ("source", "destination", "nb-revs", "nb-files")) |
|
2750 | 2769 | |
|
2751 | 2770 | if not revs: |
|
2752 | 2771 | revs = ['all()'] |
|
2753 | 2772 | revs = scmutil.revrange(repo, revs) |
|
2754 | 2773 | |
|
2755 | 2774 | if dostats: |
|
2756 | 2775 | alldata = { |
|
2757 | 2776 | 'nbrevs': [], |
|
2758 | 2777 | 'nbmissingfiles': [], |
|
2759 | 2778 | } |
|
2760 | 2779 | if dotiming: |
|
2761 | 2780 | alldata['nbrenames'] = [] |
|
2762 | 2781 | alldata['time'] = [] |
|
2763 | 2782 | |
|
2764 | 2783 | roi = repo.revs('merge() and %ld', revs) |
|
2765 | 2784 | for r in roi: |
|
2766 | 2785 | ctx = repo[r] |
|
2767 | 2786 | p1 = ctx.p1().rev() |
|
2768 | 2787 | p2 = ctx.p2().rev() |
|
2769 | 2788 | bases = repo.changelog._commonancestorsheads(p1, p2) |
|
2770 | 2789 | for p in (p1, p2): |
|
2771 | 2790 | for b in bases: |
|
2772 | 2791 | base = repo[b] |
|
2773 | 2792 | parent = repo[p] |
|
2774 | 2793 | missing = copies._computeforwardmissing(base, parent) |
|
2775 | 2794 | if not missing: |
|
2776 | 2795 | continue |
|
2777 | 2796 | data = { |
|
2778 | 2797 | b'source': base.hex(), |
|
2779 | 2798 | b'destination': parent.hex(), |
|
2780 | 2799 | b'nbrevs': len(repo.revs('only(%d, %d)', p, b)), |
|
2781 | 2800 | b'nbmissingfiles': len(missing), |
|
2782 | 2801 | } |
|
2783 | 2802 | if dostats: |
|
2784 | 2803 | alldata['nbrevs'].append( |
|
2785 | 2804 | ( |
|
2786 | 2805 | data['nbrevs'], |
|
2787 | 2806 | base.hex(), |
|
2788 | 2807 | parent.hex(), |
|
2789 | 2808 | ) |
|
2790 | 2809 | ) |
|
2791 | 2810 | alldata['nbmissingfiles'].append( |
|
2792 | 2811 | ( |
|
2793 | 2812 | data['nbmissingfiles'], |
|
2794 | 2813 | base.hex(), |
|
2795 | 2814 | parent.hex(), |
|
2796 | 2815 | ) |
|
2797 | 2816 | ) |
|
2798 | 2817 | if dotiming: |
|
2799 | 2818 | begin = util.timer() |
|
2800 | 2819 | renames = copies.pathcopies(base, parent) |
|
2801 | 2820 | end = util.timer() |
|
2802 | 2821 | # not very stable timing since we did only one run |
|
2803 | 2822 | data['time'] = end - begin |
|
2804 | 2823 | data['nbrenamedfiles'] = len(renames) |
|
2805 | 2824 | if dostats: |
|
2806 | 2825 | alldata['time'].append( |
|
2807 | 2826 | ( |
|
2808 | 2827 | data['time'], |
|
2809 | 2828 | base.hex(), |
|
2810 | 2829 | parent.hex(), |
|
2811 | 2830 | ) |
|
2812 | 2831 | ) |
|
2813 | 2832 | alldata['nbrenames'].append( |
|
2814 | 2833 | ( |
|
2815 | 2834 | data['nbrenamedfiles'], |
|
2816 | 2835 | base.hex(), |
|
2817 | 2836 | parent.hex(), |
|
2818 | 2837 | ) |
|
2819 | 2838 | ) |
|
2820 | 2839 | fm.startitem() |
|
2821 | 2840 | fm.data(**data) |
|
2822 | 2841 | out = data.copy() |
|
2823 | 2842 | out['source'] = fm.hexfunc(base.node()) |
|
2824 | 2843 | out['destination'] = fm.hexfunc(parent.node()) |
|
2825 | 2844 | fm.plain(output % out) |
|
2826 | 2845 | |
|
2827 | 2846 | fm.end() |
|
2828 | 2847 | if dostats: |
|
2829 | 2848 | entries = [ |
|
2830 | 2849 | ('nbrevs', 'number of revision covered'), |
|
2831 | 2850 | ('nbmissingfiles', 'number of missing files at head'), |
|
2832 | 2851 | ] |
|
2833 | 2852 | if dotiming: |
|
2834 | 2853 | entries.append(('nbrenames', 'renamed files')) |
|
2835 | 2854 | entries.append(('time', 'time')) |
|
2836 | 2855 | _displaystats(ui, opts, entries, alldata) |
|
2837 | 2856 | |
|
2838 | 2857 | |
|
2839 | 2858 | @command(b'perf::cca|perfcca', formatteropts) |
|
2840 | 2859 | def perfcca(ui, repo, **opts): |
|
2841 | 2860 | opts = _byteskwargs(opts) |
|
2842 | 2861 | timer, fm = gettimer(ui, opts) |
|
2843 | 2862 | timer(lambda: scmutil.casecollisionauditor(ui, False, repo.dirstate)) |
|
2844 | 2863 | fm.end() |
|
2845 | 2864 | |
|
2846 | 2865 | |
|
2847 | 2866 | @command(b'perf::fncacheload|perffncacheload', formatteropts) |
|
2848 | 2867 | def perffncacheload(ui, repo, **opts): |
|
2849 | 2868 | opts = _byteskwargs(opts) |
|
2850 | 2869 | timer, fm = gettimer(ui, opts) |
|
2851 | 2870 | s = repo.store |
|
2852 | 2871 | |
|
2853 | 2872 | def d(): |
|
2854 | 2873 | s.fncache._load() |
|
2855 | 2874 | |
|
2856 | 2875 | timer(d) |
|
2857 | 2876 | fm.end() |
|
2858 | 2877 | |
|
2859 | 2878 | |
|
2860 | 2879 | @command(b'perf::fncachewrite|perffncachewrite', formatteropts) |
|
2861 | 2880 | def perffncachewrite(ui, repo, **opts): |
|
2862 | 2881 | opts = _byteskwargs(opts) |
|
2863 | 2882 | timer, fm = gettimer(ui, opts) |
|
2864 | 2883 | s = repo.store |
|
2865 | 2884 | lock = repo.lock() |
|
2866 | 2885 | s.fncache._load() |
|
2867 | 2886 | tr = repo.transaction(b'perffncachewrite') |
|
2868 | 2887 | tr.addbackup(b'fncache') |
|
2869 | 2888 | |
|
2870 | 2889 | def d(): |
|
2871 | 2890 | s.fncache._dirty = True |
|
2872 | 2891 | s.fncache.write(tr) |
|
2873 | 2892 | |
|
2874 | 2893 | timer(d) |
|
2875 | 2894 | tr.close() |
|
2876 | 2895 | lock.release() |
|
2877 | 2896 | fm.end() |
|
2878 | 2897 | |
|
2879 | 2898 | |
|
2880 | 2899 | @command(b'perf::fncacheencode|perffncacheencode', formatteropts) |
|
2881 | 2900 | def perffncacheencode(ui, repo, **opts): |
|
2882 | 2901 | opts = _byteskwargs(opts) |
|
2883 | 2902 | timer, fm = gettimer(ui, opts) |
|
2884 | 2903 | s = repo.store |
|
2885 | 2904 | s.fncache._load() |
|
2886 | 2905 | |
|
2887 | 2906 | def d(): |
|
2888 | 2907 | for p in s.fncache.entries: |
|
2889 | 2908 | s.encode(p) |
|
2890 | 2909 | |
|
2891 | 2910 | timer(d) |
|
2892 | 2911 | fm.end() |
|
2893 | 2912 | |
|
2894 | 2913 | |
|
2895 | 2914 | def _bdiffworker(q, blocks, xdiff, ready, done): |
|
2896 | 2915 | while not done.is_set(): |
|
2897 | 2916 | pair = q.get() |
|
2898 | 2917 | while pair is not None: |
|
2899 | 2918 | if xdiff: |
|
2900 | 2919 | mdiff.bdiff.xdiffblocks(*pair) |
|
2901 | 2920 | elif blocks: |
|
2902 | 2921 | mdiff.bdiff.blocks(*pair) |
|
2903 | 2922 | else: |
|
2904 | 2923 | mdiff.textdiff(*pair) |
|
2905 | 2924 | q.task_done() |
|
2906 | 2925 | pair = q.get() |
|
2907 | 2926 | q.task_done() # for the None one |
|
2908 | 2927 | with ready: |
|
2909 | 2928 | ready.wait() |
|
2910 | 2929 | |
|
2911 | 2930 | |
|
2912 | 2931 | def _manifestrevision(repo, mnode): |
|
2913 | 2932 | ml = repo.manifestlog |
|
2914 | 2933 | |
|
2915 | 2934 | if util.safehasattr(ml, b'getstorage'): |
|
2916 | 2935 | store = ml.getstorage(b'') |
|
2917 | 2936 | else: |
|
2918 | 2937 | store = ml._revlog |
|
2919 | 2938 | |
|
2920 | 2939 | return store.revision(mnode) |
|
2921 | 2940 | |
|
2922 | 2941 | |
|
2923 | 2942 | @command( |
|
2924 | 2943 | b'perf::bdiff|perfbdiff', |
|
2925 | 2944 | revlogopts |
|
2926 | 2945 | + formatteropts |
|
2927 | 2946 | + [ |
|
2928 | 2947 | ( |
|
2929 | 2948 | b'', |
|
2930 | 2949 | b'count', |
|
2931 | 2950 | 1, |
|
2932 | 2951 | b'number of revisions to test (when using --startrev)', |
|
2933 | 2952 | ), |
|
2934 | 2953 | (b'', b'alldata', False, b'test bdiffs for all associated revisions'), |
|
2935 | 2954 | (b'', b'threads', 0, b'number of thread to use (disable with 0)'), |
|
2936 | 2955 | (b'', b'blocks', False, b'test computing diffs into blocks'), |
|
2937 | 2956 | (b'', b'xdiff', False, b'use xdiff algorithm'), |
|
2938 | 2957 | ], |
|
2939 | 2958 | b'-c|-m|FILE REV', |
|
2940 | 2959 | ) |
|
2941 | 2960 | def perfbdiff(ui, repo, file_, rev=None, count=None, threads=0, **opts): |
|
2942 | 2961 | """benchmark a bdiff between revisions |
|
2943 | 2962 | |
|
2944 | 2963 | By default, benchmark a bdiff between its delta parent and itself. |
|
2945 | 2964 | |
|
2946 | 2965 | With ``--count``, benchmark bdiffs between delta parents and self for N |
|
2947 | 2966 | revisions starting at the specified revision. |
|
2948 | 2967 | |
|
2949 | 2968 | With ``--alldata``, assume the requested revision is a changeset and |
|
2950 | 2969 | measure bdiffs for all changes related to that changeset (manifest |
|
2951 | 2970 | and filelogs). |
|
2952 | 2971 | """ |
|
2953 | 2972 | opts = _byteskwargs(opts) |
|
2954 | 2973 | |
|
2955 | 2974 | if opts[b'xdiff'] and not opts[b'blocks']: |
|
2956 | 2975 | raise error.CommandError(b'perfbdiff', b'--xdiff requires --blocks') |
|
2957 | 2976 | |
|
2958 | 2977 | if opts[b'alldata']: |
|
2959 | 2978 | opts[b'changelog'] = True |
|
2960 | 2979 | |
|
2961 | 2980 | if opts.get(b'changelog') or opts.get(b'manifest'): |
|
2962 | 2981 | file_, rev = None, file_ |
|
2963 | 2982 | elif rev is None: |
|
2964 | 2983 | raise error.CommandError(b'perfbdiff', b'invalid arguments') |
|
2965 | 2984 | |
|
2966 | 2985 | blocks = opts[b'blocks'] |
|
2967 | 2986 | xdiff = opts[b'xdiff'] |
|
2968 | 2987 | textpairs = [] |
|
2969 | 2988 | |
|
2970 | 2989 | r = cmdutil.openrevlog(repo, b'perfbdiff', file_, opts) |
|
2971 | 2990 | |
|
2972 | 2991 | startrev = r.rev(r.lookup(rev)) |
|
2973 | 2992 | for rev in range(startrev, min(startrev + count, len(r) - 1)): |
|
2974 | 2993 | if opts[b'alldata']: |
|
2975 | 2994 | # Load revisions associated with changeset. |
|
2976 | 2995 | ctx = repo[rev] |
|
2977 | 2996 | mtext = _manifestrevision(repo, ctx.manifestnode()) |
|
2978 | 2997 | for pctx in ctx.parents(): |
|
2979 | 2998 | pman = _manifestrevision(repo, pctx.manifestnode()) |
|
2980 | 2999 | textpairs.append((pman, mtext)) |
|
2981 | 3000 | |
|
2982 | 3001 | # Load filelog revisions by iterating manifest delta. |
|
2983 | 3002 | man = ctx.manifest() |
|
2984 | 3003 | pman = ctx.p1().manifest() |
|
2985 | 3004 | for filename, change in pman.diff(man).items(): |
|
2986 | 3005 | fctx = repo.file(filename) |
|
2987 | 3006 | f1 = fctx.revision(change[0][0] or -1) |
|
2988 | 3007 | f2 = fctx.revision(change[1][0] or -1) |
|
2989 | 3008 | textpairs.append((f1, f2)) |
|
2990 | 3009 | else: |
|
2991 | 3010 | dp = r.deltaparent(rev) |
|
2992 | 3011 | textpairs.append((r.revision(dp), r.revision(rev))) |
|
2993 | 3012 | |
|
2994 | 3013 | withthreads = threads > 0 |
|
2995 | 3014 | if not withthreads: |
|
2996 | 3015 | |
|
2997 | 3016 | def d(): |
|
2998 | 3017 | for pair in textpairs: |
|
2999 | 3018 | if xdiff: |
|
3000 | 3019 | mdiff.bdiff.xdiffblocks(*pair) |
|
3001 | 3020 | elif blocks: |
|
3002 | 3021 | mdiff.bdiff.blocks(*pair) |
|
3003 | 3022 | else: |
|
3004 | 3023 | mdiff.textdiff(*pair) |
|
3005 | 3024 | |
|
3006 | 3025 | else: |
|
3007 | 3026 | q = queue() |
|
3008 | 3027 | for i in _xrange(threads): |
|
3009 | 3028 | q.put(None) |
|
3010 | 3029 | ready = threading.Condition() |
|
3011 | 3030 | done = threading.Event() |
|
3012 | 3031 | for i in _xrange(threads): |
|
3013 | 3032 | threading.Thread( |
|
3014 | 3033 | target=_bdiffworker, args=(q, blocks, xdiff, ready, done) |
|
3015 | 3034 | ).start() |
|
3016 | 3035 | q.join() |
|
3017 | 3036 | |
|
3018 | 3037 | def d(): |
|
3019 | 3038 | for pair in textpairs: |
|
3020 | 3039 | q.put(pair) |
|
3021 | 3040 | for i in _xrange(threads): |
|
3022 | 3041 | q.put(None) |
|
3023 | 3042 | with ready: |
|
3024 | 3043 | ready.notify_all() |
|
3025 | 3044 | q.join() |
|
3026 | 3045 | |
|
3027 | 3046 | timer, fm = gettimer(ui, opts) |
|
3028 | 3047 | timer(d) |
|
3029 | 3048 | fm.end() |
|
3030 | 3049 | |
|
3031 | 3050 | if withthreads: |
|
3032 | 3051 | done.set() |
|
3033 | 3052 | for i in _xrange(threads): |
|
3034 | 3053 | q.put(None) |
|
3035 | 3054 | with ready: |
|
3036 | 3055 | ready.notify_all() |
|
3037 | 3056 | |
|
3038 | 3057 | |
|
3039 | 3058 | @command( |
|
3040 | 3059 | b'perf::unbundle', |
|
3041 | 3060 | [ |
|
3042 | 3061 | (b'', b'as-push', None, b'pretend the bundle comes from a push'), |
|
3043 | 3062 | ] |
|
3044 | 3063 | + formatteropts, |
|
3045 | 3064 | b'BUNDLE_FILE', |
|
3046 | 3065 | ) |
|
3047 | 3066 | def perf_unbundle(ui, repo, fname, **opts): |
|
3048 | 3067 | """benchmark application of a bundle in a repository. |
|
3049 | 3068 | |
|
3050 | 3069 | This does not include the final transaction processing |
|
3051 | 3070 | |
|
3052 | 3071 | The --as-push option make the unbundle operation appears like it comes from |
|
3053 | 3072 | a client push. It change some aspect of the processing and associated |
|
3054 | 3073 | performance profile. |
|
3055 | 3074 | """ |
|
3056 | 3075 | |
|
3057 | 3076 | from mercurial import exchange |
|
3058 | 3077 | from mercurial import bundle2 |
|
3059 | 3078 | from mercurial import transaction |
|
3060 | 3079 | |
|
3061 | 3080 | opts = _byteskwargs(opts) |
|
3062 | 3081 | |
|
3063 | 3082 | ### some compatibility hotfix |
|
3064 | 3083 | # |
|
3065 | 3084 | # the data attribute is dropped in 63edc384d3b7 a changeset introducing a |
|
3066 | 3085 | # critical regression that break transaction rollback for files that are |
|
3067 | 3086 | # de-inlined. |
|
3068 | 3087 | method = transaction.transaction._addentry |
|
3069 | 3088 | pre_63edc384d3b7 = "data" in getargspec(method).args |
|
3070 | 3089 | # the `detailed_exit_code` attribute is introduced in 33c0c25d0b0f |
|
3071 | 3090 | # a changeset that is a close descendant of 18415fc918a1, the changeset |
|
3072 | 3091 | # that conclude the fix run for the bug introduced in 63edc384d3b7. |
|
3073 | 3092 | args = getargspec(error.Abort.__init__).args |
|
3074 | 3093 | post_18415fc918a1 = "detailed_exit_code" in args |
|
3075 | 3094 | |
|
3076 | 3095 | unbundle_source = b'perf::unbundle' |
|
3077 | 3096 | if opts[b'as_push']: |
|
3078 | 3097 | unbundle_source = b'push' |
|
3079 | 3098 | |
|
3080 | 3099 | old_max_inline = None |
|
3081 | 3100 | try: |
|
3082 | 3101 | if not (pre_63edc384d3b7 or post_18415fc918a1): |
|
3083 | 3102 | # disable inlining |
|
3084 | 3103 | old_max_inline = mercurial.revlog._maxinline |
|
3085 | 3104 | # large enough to never happen |
|
3086 | 3105 | mercurial.revlog._maxinline = 2 ** 50 |
|
3087 | 3106 | |
|
3088 | 3107 | with repo.lock(): |
|
3089 | 3108 | bundle = [None, None] |
|
3090 | 3109 | orig_quiet = repo.ui.quiet |
|
3091 | 3110 | try: |
|
3092 | 3111 | repo.ui.quiet = True |
|
3093 | 3112 | with open(fname, mode="rb") as f: |
|
3094 | 3113 | |
|
3095 | 3114 | def noop_report(*args, **kwargs): |
|
3096 | 3115 | pass |
|
3097 | 3116 | |
|
3098 | 3117 | def setup(): |
|
3099 | 3118 | gen, tr = bundle |
|
3100 | 3119 | if tr is not None: |
|
3101 | 3120 | tr.abort() |
|
3102 | 3121 | bundle[:] = [None, None] |
|
3103 | 3122 | f.seek(0) |
|
3104 | 3123 | bundle[0] = exchange.readbundle(ui, f, fname) |
|
3105 | 3124 | bundle[1] = repo.transaction(b'perf::unbundle') |
|
3106 | 3125 | # silence the transaction |
|
3107 | 3126 | bundle[1]._report = noop_report |
|
3108 | 3127 | |
|
3109 | 3128 | def apply(): |
|
3110 | 3129 | gen, tr = bundle |
|
3111 | 3130 | bundle2.applybundle( |
|
3112 | 3131 | repo, |
|
3113 | 3132 | gen, |
|
3114 | 3133 | tr, |
|
3115 | 3134 | source=unbundle_source, |
|
3116 | 3135 | url=fname, |
|
3117 | 3136 | ) |
|
3118 | 3137 | |
|
3119 | 3138 | timer, fm = gettimer(ui, opts) |
|
3120 | 3139 | timer(apply, setup=setup) |
|
3121 | 3140 | fm.end() |
|
3122 | 3141 | finally: |
|
3123 | 3142 | repo.ui.quiet == orig_quiet |
|
3124 | 3143 | gen, tr = bundle |
|
3125 | 3144 | if tr is not None: |
|
3126 | 3145 | tr.abort() |
|
3127 | 3146 | finally: |
|
3128 | 3147 | if old_max_inline is not None: |
|
3129 | 3148 | mercurial.revlog._maxinline = old_max_inline |
|
3130 | 3149 | |
|
3131 | 3150 | |
|
3132 | 3151 | @command( |
|
3133 | 3152 | b'perf::unidiff|perfunidiff', |
|
3134 | 3153 | revlogopts |
|
3135 | 3154 | + formatteropts |
|
3136 | 3155 | + [ |
|
3137 | 3156 | ( |
|
3138 | 3157 | b'', |
|
3139 | 3158 | b'count', |
|
3140 | 3159 | 1, |
|
3141 | 3160 | b'number of revisions to test (when using --startrev)', |
|
3142 | 3161 | ), |
|
3143 | 3162 | (b'', b'alldata', False, b'test unidiffs for all associated revisions'), |
|
3144 | 3163 | ], |
|
3145 | 3164 | b'-c|-m|FILE REV', |
|
3146 | 3165 | ) |
|
3147 | 3166 | def perfunidiff(ui, repo, file_, rev=None, count=None, **opts): |
|
3148 | 3167 | """benchmark a unified diff between revisions |
|
3149 | 3168 | |
|
3150 | 3169 | This doesn't include any copy tracing - it's just a unified diff |
|
3151 | 3170 | of the texts. |
|
3152 | 3171 | |
|
3153 | 3172 | By default, benchmark a diff between its delta parent and itself. |
|
3154 | 3173 | |
|
3155 | 3174 | With ``--count``, benchmark diffs between delta parents and self for N |
|
3156 | 3175 | revisions starting at the specified revision. |
|
3157 | 3176 | |
|
3158 | 3177 | With ``--alldata``, assume the requested revision is a changeset and |
|
3159 | 3178 | measure diffs for all changes related to that changeset (manifest |
|
3160 | 3179 | and filelogs). |
|
3161 | 3180 | """ |
|
3162 | 3181 | opts = _byteskwargs(opts) |
|
3163 | 3182 | if opts[b'alldata']: |
|
3164 | 3183 | opts[b'changelog'] = True |
|
3165 | 3184 | |
|
3166 | 3185 | if opts.get(b'changelog') or opts.get(b'manifest'): |
|
3167 | 3186 | file_, rev = None, file_ |
|
3168 | 3187 | elif rev is None: |
|
3169 | 3188 | raise error.CommandError(b'perfunidiff', b'invalid arguments') |
|
3170 | 3189 | |
|
3171 | 3190 | textpairs = [] |
|
3172 | 3191 | |
|
3173 | 3192 | r = cmdutil.openrevlog(repo, b'perfunidiff', file_, opts) |
|
3174 | 3193 | |
|
3175 | 3194 | startrev = r.rev(r.lookup(rev)) |
|
3176 | 3195 | for rev in range(startrev, min(startrev + count, len(r) - 1)): |
|
3177 | 3196 | if opts[b'alldata']: |
|
3178 | 3197 | # Load revisions associated with changeset. |
|
3179 | 3198 | ctx = repo[rev] |
|
3180 | 3199 | mtext = _manifestrevision(repo, ctx.manifestnode()) |
|
3181 | 3200 | for pctx in ctx.parents(): |
|
3182 | 3201 | pman = _manifestrevision(repo, pctx.manifestnode()) |
|
3183 | 3202 | textpairs.append((pman, mtext)) |
|
3184 | 3203 | |
|
3185 | 3204 | # Load filelog revisions by iterating manifest delta. |
|
3186 | 3205 | man = ctx.manifest() |
|
3187 | 3206 | pman = ctx.p1().manifest() |
|
3188 | 3207 | for filename, change in pman.diff(man).items(): |
|
3189 | 3208 | fctx = repo.file(filename) |
|
3190 | 3209 | f1 = fctx.revision(change[0][0] or -1) |
|
3191 | 3210 | f2 = fctx.revision(change[1][0] or -1) |
|
3192 | 3211 | textpairs.append((f1, f2)) |
|
3193 | 3212 | else: |
|
3194 | 3213 | dp = r.deltaparent(rev) |
|
3195 | 3214 | textpairs.append((r.revision(dp), r.revision(rev))) |
|
3196 | 3215 | |
|
3197 | 3216 | def d(): |
|
3198 | 3217 | for left, right in textpairs: |
|
3199 | 3218 | # The date strings don't matter, so we pass empty strings. |
|
3200 | 3219 | headerlines, hunks = mdiff.unidiff( |
|
3201 | 3220 | left, b'', right, b'', b'left', b'right', binary=False |
|
3202 | 3221 | ) |
|
3203 | 3222 | # consume iterators in roughly the way patch.py does |
|
3204 | 3223 | b'\n'.join(headerlines) |
|
3205 | 3224 | b''.join(sum((list(hlines) for hrange, hlines in hunks), [])) |
|
3206 | 3225 | |
|
3207 | 3226 | timer, fm = gettimer(ui, opts) |
|
3208 | 3227 | timer(d) |
|
3209 | 3228 | fm.end() |
|
3210 | 3229 | |
|
3211 | 3230 | |
|
3212 | 3231 | @command(b'perf::diffwd|perfdiffwd', formatteropts) |
|
3213 | 3232 | def perfdiffwd(ui, repo, **opts): |
|
3214 | 3233 | """Profile diff of working directory changes""" |
|
3215 | 3234 | opts = _byteskwargs(opts) |
|
3216 | 3235 | timer, fm = gettimer(ui, opts) |
|
3217 | 3236 | options = { |
|
3218 | 3237 | 'w': 'ignore_all_space', |
|
3219 | 3238 | 'b': 'ignore_space_change', |
|
3220 | 3239 | 'B': 'ignore_blank_lines', |
|
3221 | 3240 | } |
|
3222 | 3241 | |
|
3223 | 3242 | for diffopt in ('', 'w', 'b', 'B', 'wB'): |
|
3224 | 3243 | opts = {options[c]: b'1' for c in diffopt} |
|
3225 | 3244 | |
|
3226 | 3245 | def d(): |
|
3227 | 3246 | ui.pushbuffer() |
|
3228 | 3247 | commands.diff(ui, repo, **opts) |
|
3229 | 3248 | ui.popbuffer() |
|
3230 | 3249 | |
|
3231 | 3250 | diffopt = diffopt.encode('ascii') |
|
3232 | 3251 | title = b'diffopts: %s' % (diffopt and (b'-' + diffopt) or b'none') |
|
3233 | 3252 | timer(d, title=title) |
|
3234 | 3253 | fm.end() |
|
3235 | 3254 | |
|
3236 | 3255 | |
|
3237 | 3256 | @command( |
|
3238 | 3257 | b'perf::revlogindex|perfrevlogindex', |
|
3239 | 3258 | revlogopts + formatteropts, |
|
3240 | 3259 | b'-c|-m|FILE', |
|
3241 | 3260 | ) |
|
3242 | 3261 | def perfrevlogindex(ui, repo, file_=None, **opts): |
|
3243 | 3262 | """Benchmark operations against a revlog index. |
|
3244 | 3263 | |
|
3245 | 3264 | This tests constructing a revlog instance, reading index data, |
|
3246 | 3265 | parsing index data, and performing various operations related to |
|
3247 | 3266 | index data. |
|
3248 | 3267 | """ |
|
3249 | 3268 | |
|
3250 | 3269 | opts = _byteskwargs(opts) |
|
3251 | 3270 | |
|
3252 | 3271 | rl = cmdutil.openrevlog(repo, b'perfrevlogindex', file_, opts) |
|
3253 | 3272 | |
|
3254 | 3273 | opener = getattr(rl, 'opener') # trick linter |
|
3255 | 3274 | # compat with hg <= 5.8 |
|
3256 | 3275 | radix = getattr(rl, 'radix', None) |
|
3257 | 3276 | indexfile = getattr(rl, '_indexfile', None) |
|
3258 | 3277 | if indexfile is None: |
|
3259 | 3278 | # compatibility with <= hg-5.8 |
|
3260 | 3279 | indexfile = getattr(rl, 'indexfile') |
|
3261 | 3280 | data = opener.read(indexfile) |
|
3262 | 3281 | |
|
3263 | 3282 | header = struct.unpack(b'>I', data[0:4])[0] |
|
3264 | 3283 | version = header & 0xFFFF |
|
3265 | 3284 | if version == 1: |
|
3266 | 3285 | inline = header & (1 << 16) |
|
3267 | 3286 | else: |
|
3268 | 3287 | raise error.Abort(b'unsupported revlog version: %d' % version) |
|
3269 | 3288 | |
|
3270 | 3289 | parse_index_v1 = getattr(mercurial.revlog, 'parse_index_v1', None) |
|
3271 | 3290 | if parse_index_v1 is None: |
|
3272 | 3291 | parse_index_v1 = mercurial.revlog.revlogio().parseindex |
|
3273 | 3292 | |
|
3274 | 3293 | rllen = len(rl) |
|
3275 | 3294 | |
|
3276 | 3295 | node0 = rl.node(0) |
|
3277 | 3296 | node25 = rl.node(rllen // 4) |
|
3278 | 3297 | node50 = rl.node(rllen // 2) |
|
3279 | 3298 | node75 = rl.node(rllen // 4 * 3) |
|
3280 | 3299 | node100 = rl.node(rllen - 1) |
|
3281 | 3300 | |
|
3282 | 3301 | allrevs = range(rllen) |
|
3283 | 3302 | allrevsrev = list(reversed(allrevs)) |
|
3284 | 3303 | allnodes = [rl.node(rev) for rev in range(rllen)] |
|
3285 | 3304 | allnodesrev = list(reversed(allnodes)) |
|
3286 | 3305 | |
|
3287 | 3306 | def constructor(): |
|
3288 | 3307 | if radix is not None: |
|
3289 | 3308 | revlog(opener, radix=radix) |
|
3290 | 3309 | else: |
|
3291 | 3310 | # hg <= 5.8 |
|
3292 | 3311 | revlog(opener, indexfile=indexfile) |
|
3293 | 3312 | |
|
3294 | 3313 | def read(): |
|
3295 | 3314 | with opener(indexfile) as fh: |
|
3296 | 3315 | fh.read() |
|
3297 | 3316 | |
|
3298 | 3317 | def parseindex(): |
|
3299 | 3318 | parse_index_v1(data, inline) |
|
3300 | 3319 | |
|
3301 | 3320 | def getentry(revornode): |
|
3302 | 3321 | index = parse_index_v1(data, inline)[0] |
|
3303 | 3322 | index[revornode] |
|
3304 | 3323 | |
|
3305 | 3324 | def getentries(revs, count=1): |
|
3306 | 3325 | index = parse_index_v1(data, inline)[0] |
|
3307 | 3326 | |
|
3308 | 3327 | for i in range(count): |
|
3309 | 3328 | for rev in revs: |
|
3310 | 3329 | index[rev] |
|
3311 | 3330 | |
|
3312 | 3331 | def resolvenode(node): |
|
3313 | 3332 | index = parse_index_v1(data, inline)[0] |
|
3314 | 3333 | rev = getattr(index, 'rev', None) |
|
3315 | 3334 | if rev is None: |
|
3316 | 3335 | nodemap = getattr(parse_index_v1(data, inline)[0], 'nodemap', None) |
|
3317 | 3336 | # This only works for the C code. |
|
3318 | 3337 | if nodemap is None: |
|
3319 | 3338 | return |
|
3320 | 3339 | rev = nodemap.__getitem__ |
|
3321 | 3340 | |
|
3322 | 3341 | try: |
|
3323 | 3342 | rev(node) |
|
3324 | 3343 | except error.RevlogError: |
|
3325 | 3344 | pass |
|
3326 | 3345 | |
|
3327 | 3346 | def resolvenodes(nodes, count=1): |
|
3328 | 3347 | index = parse_index_v1(data, inline)[0] |
|
3329 | 3348 | rev = getattr(index, 'rev', None) |
|
3330 | 3349 | if rev is None: |
|
3331 | 3350 | nodemap = getattr(parse_index_v1(data, inline)[0], 'nodemap', None) |
|
3332 | 3351 | # This only works for the C code. |
|
3333 | 3352 | if nodemap is None: |
|
3334 | 3353 | return |
|
3335 | 3354 | rev = nodemap.__getitem__ |
|
3336 | 3355 | |
|
3337 | 3356 | for i in range(count): |
|
3338 | 3357 | for node in nodes: |
|
3339 | 3358 | try: |
|
3340 | 3359 | rev(node) |
|
3341 | 3360 | except error.RevlogError: |
|
3342 | 3361 | pass |
|
3343 | 3362 | |
|
3344 | 3363 | benches = [ |
|
3345 | 3364 | (constructor, b'revlog constructor'), |
|
3346 | 3365 | (read, b'read'), |
|
3347 | 3366 | (parseindex, b'create index object'), |
|
3348 | 3367 | (lambda: getentry(0), b'retrieve index entry for rev 0'), |
|
3349 | 3368 | (lambda: resolvenode(b'a' * 20), b'look up missing node'), |
|
3350 | 3369 | (lambda: resolvenode(node0), b'look up node at rev 0'), |
|
3351 | 3370 | (lambda: resolvenode(node25), b'look up node at 1/4 len'), |
|
3352 | 3371 | (lambda: resolvenode(node50), b'look up node at 1/2 len'), |
|
3353 | 3372 | (lambda: resolvenode(node75), b'look up node at 3/4 len'), |
|
3354 | 3373 | (lambda: resolvenode(node100), b'look up node at tip'), |
|
3355 | 3374 | # 2x variation is to measure caching impact. |
|
3356 | 3375 | (lambda: resolvenodes(allnodes), b'look up all nodes (forward)'), |
|
3357 | 3376 | (lambda: resolvenodes(allnodes, 2), b'look up all nodes 2x (forward)'), |
|
3358 | 3377 | (lambda: resolvenodes(allnodesrev), b'look up all nodes (reverse)'), |
|
3359 | 3378 | ( |
|
3360 | 3379 | lambda: resolvenodes(allnodesrev, 2), |
|
3361 | 3380 | b'look up all nodes 2x (reverse)', |
|
3362 | 3381 | ), |
|
3363 | 3382 | (lambda: getentries(allrevs), b'retrieve all index entries (forward)'), |
|
3364 | 3383 | ( |
|
3365 | 3384 | lambda: getentries(allrevs, 2), |
|
3366 | 3385 | b'retrieve all index entries 2x (forward)', |
|
3367 | 3386 | ), |
|
3368 | 3387 | ( |
|
3369 | 3388 | lambda: getentries(allrevsrev), |
|
3370 | 3389 | b'retrieve all index entries (reverse)', |
|
3371 | 3390 | ), |
|
3372 | 3391 | ( |
|
3373 | 3392 | lambda: getentries(allrevsrev, 2), |
|
3374 | 3393 | b'retrieve all index entries 2x (reverse)', |
|
3375 | 3394 | ), |
|
3376 | 3395 | ] |
|
3377 | 3396 | |
|
3378 | 3397 | for fn, title in benches: |
|
3379 | 3398 | timer, fm = gettimer(ui, opts) |
|
3380 | 3399 | timer(fn, title=title) |
|
3381 | 3400 | fm.end() |
|
3382 | 3401 | |
|
3383 | 3402 | |
|
3384 | 3403 | @command( |
|
3385 | 3404 | b'perf::revlogrevisions|perfrevlogrevisions', |
|
3386 | 3405 | revlogopts |
|
3387 | 3406 | + formatteropts |
|
3388 | 3407 | + [ |
|
3389 | 3408 | (b'd', b'dist', 100, b'distance between the revisions'), |
|
3390 | 3409 | (b's', b'startrev', 0, b'revision to start reading at'), |
|
3391 | 3410 | (b'', b'reverse', False, b'read in reverse'), |
|
3392 | 3411 | ], |
|
3393 | 3412 | b'-c|-m|FILE', |
|
3394 | 3413 | ) |
|
3395 | 3414 | def perfrevlogrevisions( |
|
3396 | 3415 | ui, repo, file_=None, startrev=0, reverse=False, **opts |
|
3397 | 3416 | ): |
|
3398 | 3417 | """Benchmark reading a series of revisions from a revlog. |
|
3399 | 3418 | |
|
3400 | 3419 | By default, we read every ``-d/--dist`` revision from 0 to tip of |
|
3401 | 3420 | the specified revlog. |
|
3402 | 3421 | |
|
3403 | 3422 | The start revision can be defined via ``-s/--startrev``. |
|
3404 | 3423 | """ |
|
3405 | 3424 | opts = _byteskwargs(opts) |
|
3406 | 3425 | |
|
3407 | 3426 | rl = cmdutil.openrevlog(repo, b'perfrevlogrevisions', file_, opts) |
|
3408 | 3427 | rllen = getlen(ui)(rl) |
|
3409 | 3428 | |
|
3410 | 3429 | if startrev < 0: |
|
3411 | 3430 | startrev = rllen + startrev |
|
3412 | 3431 | |
|
3413 | 3432 | def d(): |
|
3414 | 3433 | rl.clearcaches() |
|
3415 | 3434 | |
|
3416 | 3435 | beginrev = startrev |
|
3417 | 3436 | endrev = rllen |
|
3418 | 3437 | dist = opts[b'dist'] |
|
3419 | 3438 | |
|
3420 | 3439 | if reverse: |
|
3421 | 3440 | beginrev, endrev = endrev - 1, beginrev - 1 |
|
3422 | 3441 | dist = -1 * dist |
|
3423 | 3442 | |
|
3424 | 3443 | for x in _xrange(beginrev, endrev, dist): |
|
3425 | 3444 | # Old revisions don't support passing int. |
|
3426 | 3445 | n = rl.node(x) |
|
3427 | 3446 | rl.revision(n) |
|
3428 | 3447 | |
|
3429 | 3448 | timer, fm = gettimer(ui, opts) |
|
3430 | 3449 | timer(d) |
|
3431 | 3450 | fm.end() |
|
3432 | 3451 | |
|
3433 | 3452 | |
|
3434 | 3453 | @command( |
|
3435 | 3454 | b'perf::revlogwrite|perfrevlogwrite', |
|
3436 | 3455 | revlogopts |
|
3437 | 3456 | + formatteropts |
|
3438 | 3457 | + [ |
|
3439 | 3458 | (b's', b'startrev', 1000, b'revision to start writing at'), |
|
3440 | 3459 | (b'', b'stoprev', -1, b'last revision to write'), |
|
3441 | 3460 | (b'', b'count', 3, b'number of passes to perform'), |
|
3442 | 3461 | (b'', b'details', False, b'print timing for every revisions tested'), |
|
3443 | 3462 | (b'', b'source', b'full', b'the kind of data feed in the revlog'), |
|
3444 | 3463 | (b'', b'lazydeltabase', True, b'try the provided delta first'), |
|
3445 | 3464 | (b'', b'clear-caches', True, b'clear revlog cache between calls'), |
|
3446 | 3465 | ], |
|
3447 | 3466 | b'-c|-m|FILE', |
|
3448 | 3467 | ) |
|
3449 | 3468 | def perfrevlogwrite(ui, repo, file_=None, startrev=1000, stoprev=-1, **opts): |
|
3450 | 3469 | """Benchmark writing a series of revisions to a revlog. |
|
3451 | 3470 | |
|
3452 | 3471 | Possible source values are: |
|
3453 | 3472 | * `full`: add from a full text (default). |
|
3454 | 3473 | * `parent-1`: add from a delta to the first parent |
|
3455 | 3474 | * `parent-2`: add from a delta to the second parent if it exists |
|
3456 | 3475 | (use a delta from the first parent otherwise) |
|
3457 | 3476 | * `parent-smallest`: add from the smallest delta (either p1 or p2) |
|
3458 | 3477 | * `storage`: add from the existing precomputed deltas |
|
3459 | 3478 | |
|
3460 | 3479 | Note: This performance command measures performance in a custom way. As a |
|
3461 | 3480 | result some of the global configuration of the 'perf' command does not |
|
3462 | 3481 | apply to it: |
|
3463 | 3482 | |
|
3464 | 3483 | * ``pre-run``: disabled |
|
3465 | 3484 | |
|
3466 | 3485 | * ``profile-benchmark``: disabled |
|
3467 | 3486 | |
|
3468 | 3487 | * ``run-limits``: disabled use --count instead |
|
3469 | 3488 | """ |
|
3470 | 3489 | opts = _byteskwargs(opts) |
|
3471 | 3490 | |
|
3472 | 3491 | rl = cmdutil.openrevlog(repo, b'perfrevlogwrite', file_, opts) |
|
3473 | 3492 | rllen = getlen(ui)(rl) |
|
3474 | 3493 | if startrev < 0: |
|
3475 | 3494 | startrev = rllen + startrev |
|
3476 | 3495 | if stoprev < 0: |
|
3477 | 3496 | stoprev = rllen + stoprev |
|
3478 | 3497 | |
|
3479 | 3498 | lazydeltabase = opts['lazydeltabase'] |
|
3480 | 3499 | source = opts['source'] |
|
3481 | 3500 | clearcaches = opts['clear_caches'] |
|
3482 | 3501 | validsource = ( |
|
3483 | 3502 | b'full', |
|
3484 | 3503 | b'parent-1', |
|
3485 | 3504 | b'parent-2', |
|
3486 | 3505 | b'parent-smallest', |
|
3487 | 3506 | b'storage', |
|
3488 | 3507 | ) |
|
3489 | 3508 | if source not in validsource: |
|
3490 | 3509 | raise error.Abort('invalid source type: %s' % source) |
|
3491 | 3510 | |
|
3492 | 3511 | ### actually gather results |
|
3493 | 3512 | count = opts['count'] |
|
3494 | 3513 | if count <= 0: |
|
3495 | 3514 | raise error.Abort('invalide run count: %d' % count) |
|
3496 | 3515 | allresults = [] |
|
3497 | 3516 | for c in range(count): |
|
3498 | 3517 | timing = _timeonewrite( |
|
3499 | 3518 | ui, |
|
3500 | 3519 | rl, |
|
3501 | 3520 | source, |
|
3502 | 3521 | startrev, |
|
3503 | 3522 | stoprev, |
|
3504 | 3523 | c + 1, |
|
3505 | 3524 | lazydeltabase=lazydeltabase, |
|
3506 | 3525 | clearcaches=clearcaches, |
|
3507 | 3526 | ) |
|
3508 | 3527 | allresults.append(timing) |
|
3509 | 3528 | |
|
3510 | 3529 | ### consolidate the results in a single list |
|
3511 | 3530 | results = [] |
|
3512 | 3531 | for idx, (rev, t) in enumerate(allresults[0]): |
|
3513 | 3532 | ts = [t] |
|
3514 | 3533 | for other in allresults[1:]: |
|
3515 | 3534 | orev, ot = other[idx] |
|
3516 | 3535 | assert orev == rev |
|
3517 | 3536 | ts.append(ot) |
|
3518 | 3537 | results.append((rev, ts)) |
|
3519 | 3538 | resultcount = len(results) |
|
3520 | 3539 | |
|
3521 | 3540 | ### Compute and display relevant statistics |
|
3522 | 3541 | |
|
3523 | 3542 | # get a formatter |
|
3524 | 3543 | fm = ui.formatter(b'perf', opts) |
|
3525 | 3544 | displayall = ui.configbool(b"perf", b"all-timing", True) |
|
3526 | 3545 | |
|
3527 | 3546 | # print individual details if requested |
|
3528 | 3547 | if opts['details']: |
|
3529 | 3548 | for idx, item in enumerate(results, 1): |
|
3530 | 3549 | rev, data = item |
|
3531 | 3550 | title = 'revisions #%d of %d, rev %d' % (idx, resultcount, rev) |
|
3532 | 3551 | formatone(fm, data, title=title, displayall=displayall) |
|
3533 | 3552 | |
|
3534 | 3553 | # sorts results by median time |
|
3535 | 3554 | results.sort(key=lambda x: sorted(x[1])[len(x[1]) // 2]) |
|
3536 | 3555 | # list of (name, index) to display) |
|
3537 | 3556 | relevants = [ |
|
3538 | 3557 | ("min", 0), |
|
3539 | 3558 | ("10%", resultcount * 10 // 100), |
|
3540 | 3559 | ("25%", resultcount * 25 // 100), |
|
3541 | 3560 | ("50%", resultcount * 70 // 100), |
|
3542 | 3561 | ("75%", resultcount * 75 // 100), |
|
3543 | 3562 | ("90%", resultcount * 90 // 100), |
|
3544 | 3563 | ("95%", resultcount * 95 // 100), |
|
3545 | 3564 | ("99%", resultcount * 99 // 100), |
|
3546 | 3565 | ("99.9%", resultcount * 999 // 1000), |
|
3547 | 3566 | ("99.99%", resultcount * 9999 // 10000), |
|
3548 | 3567 | ("99.999%", resultcount * 99999 // 100000), |
|
3549 | 3568 | ("max", -1), |
|
3550 | 3569 | ] |
|
3551 | 3570 | if not ui.quiet: |
|
3552 | 3571 | for name, idx in relevants: |
|
3553 | 3572 | data = results[idx] |
|
3554 | 3573 | title = '%s of %d, rev %d' % (name, resultcount, data[0]) |
|
3555 | 3574 | formatone(fm, data[1], title=title, displayall=displayall) |
|
3556 | 3575 | |
|
3557 | 3576 | # XXX summing that many float will not be very precise, we ignore this fact |
|
3558 | 3577 | # for now |
|
3559 | 3578 | totaltime = [] |
|
3560 | 3579 | for item in allresults: |
|
3561 | 3580 | totaltime.append( |
|
3562 | 3581 | ( |
|
3563 | 3582 | sum(x[1][0] for x in item), |
|
3564 | 3583 | sum(x[1][1] for x in item), |
|
3565 | 3584 | sum(x[1][2] for x in item), |
|
3566 | 3585 | ) |
|
3567 | 3586 | ) |
|
3568 | 3587 | formatone( |
|
3569 | 3588 | fm, |
|
3570 | 3589 | totaltime, |
|
3571 | 3590 | title="total time (%d revs)" % resultcount, |
|
3572 | 3591 | displayall=displayall, |
|
3573 | 3592 | ) |
|
3574 | 3593 | fm.end() |
|
3575 | 3594 | |
|
3576 | 3595 | |
|
3577 | 3596 | class _faketr: |
|
3578 | 3597 | def add(s, x, y, z=None): |
|
3579 | 3598 | return None |
|
3580 | 3599 | |
|
3581 | 3600 | |
|
3582 | 3601 | def _timeonewrite( |
|
3583 | 3602 | ui, |
|
3584 | 3603 | orig, |
|
3585 | 3604 | source, |
|
3586 | 3605 | startrev, |
|
3587 | 3606 | stoprev, |
|
3588 | 3607 | runidx=None, |
|
3589 | 3608 | lazydeltabase=True, |
|
3590 | 3609 | clearcaches=True, |
|
3591 | 3610 | ): |
|
3592 | 3611 | timings = [] |
|
3593 | 3612 | tr = _faketr() |
|
3594 | 3613 | with _temprevlog(ui, orig, startrev) as dest: |
|
3595 | 3614 | if hasattr(dest, "delta_config"): |
|
3596 | 3615 | dest.delta_config.lazy_delta_base = lazydeltabase |
|
3597 | 3616 | else: |
|
3598 | 3617 | dest._lazydeltabase = lazydeltabase |
|
3599 | 3618 | revs = list(orig.revs(startrev, stoprev)) |
|
3600 | 3619 | total = len(revs) |
|
3601 | 3620 | topic = 'adding' |
|
3602 | 3621 | if runidx is not None: |
|
3603 | 3622 | topic += ' (run #%d)' % runidx |
|
3604 | 3623 | # Support both old and new progress API |
|
3605 | 3624 | if util.safehasattr(ui, 'makeprogress'): |
|
3606 | 3625 | progress = ui.makeprogress(topic, unit='revs', total=total) |
|
3607 | 3626 | |
|
3608 | 3627 | def updateprogress(pos): |
|
3609 | 3628 | progress.update(pos) |
|
3610 | 3629 | |
|
3611 | 3630 | def completeprogress(): |
|
3612 | 3631 | progress.complete() |
|
3613 | 3632 | |
|
3614 | 3633 | else: |
|
3615 | 3634 | |
|
3616 | 3635 | def updateprogress(pos): |
|
3617 | 3636 | ui.progress(topic, pos, unit='revs', total=total) |
|
3618 | 3637 | |
|
3619 | 3638 | def completeprogress(): |
|
3620 | 3639 | ui.progress(topic, None, unit='revs', total=total) |
|
3621 | 3640 | |
|
3622 | 3641 | for idx, rev in enumerate(revs): |
|
3623 | 3642 | updateprogress(idx) |
|
3624 | 3643 | addargs, addkwargs = _getrevisionseed(orig, rev, tr, source) |
|
3625 | 3644 | if clearcaches: |
|
3626 | 3645 | dest.index.clearcaches() |
|
3627 | 3646 | dest.clearcaches() |
|
3628 | 3647 | with timeone() as r: |
|
3629 | 3648 | dest.addrawrevision(*addargs, **addkwargs) |
|
3630 | 3649 | timings.append((rev, r[0])) |
|
3631 | 3650 | updateprogress(total) |
|
3632 | 3651 | completeprogress() |
|
3633 | 3652 | return timings |
|
3634 | 3653 | |
|
3635 | 3654 | |
|
3636 | 3655 | def _getrevisionseed(orig, rev, tr, source): |
|
3637 | 3656 | from mercurial.node import nullid |
|
3638 | 3657 | |
|
3639 | 3658 | linkrev = orig.linkrev(rev) |
|
3640 | 3659 | node = orig.node(rev) |
|
3641 | 3660 | p1, p2 = orig.parents(node) |
|
3642 | 3661 | flags = orig.flags(rev) |
|
3643 | 3662 | cachedelta = None |
|
3644 | 3663 | text = None |
|
3645 | 3664 | |
|
3646 | 3665 | if source == b'full': |
|
3647 | 3666 | text = orig.revision(rev) |
|
3648 | 3667 | elif source == b'parent-1': |
|
3649 | 3668 | baserev = orig.rev(p1) |
|
3650 | 3669 | cachedelta = (baserev, orig.revdiff(p1, rev)) |
|
3651 | 3670 | elif source == b'parent-2': |
|
3652 | 3671 | parent = p2 |
|
3653 | 3672 | if p2 == nullid: |
|
3654 | 3673 | parent = p1 |
|
3655 | 3674 | baserev = orig.rev(parent) |
|
3656 | 3675 | cachedelta = (baserev, orig.revdiff(parent, rev)) |
|
3657 | 3676 | elif source == b'parent-smallest': |
|
3658 | 3677 | p1diff = orig.revdiff(p1, rev) |
|
3659 | 3678 | parent = p1 |
|
3660 | 3679 | diff = p1diff |
|
3661 | 3680 | if p2 != nullid: |
|
3662 | 3681 | p2diff = orig.revdiff(p2, rev) |
|
3663 | 3682 | if len(p1diff) > len(p2diff): |
|
3664 | 3683 | parent = p2 |
|
3665 | 3684 | diff = p2diff |
|
3666 | 3685 | baserev = orig.rev(parent) |
|
3667 | 3686 | cachedelta = (baserev, diff) |
|
3668 | 3687 | elif source == b'storage': |
|
3669 | 3688 | baserev = orig.deltaparent(rev) |
|
3670 | 3689 | cachedelta = (baserev, orig.revdiff(orig.node(baserev), rev)) |
|
3671 | 3690 | |
|
3672 | 3691 | return ( |
|
3673 | 3692 | (text, tr, linkrev, p1, p2), |
|
3674 | 3693 | {'node': node, 'flags': flags, 'cachedelta': cachedelta}, |
|
3675 | 3694 | ) |
|
3676 | 3695 | |
|
3677 | 3696 | |
|
3678 | 3697 | @contextlib.contextmanager |
|
3679 | 3698 | def _temprevlog(ui, orig, truncaterev): |
|
3680 | 3699 | from mercurial import vfs as vfsmod |
|
3681 | 3700 | |
|
3682 | 3701 | if orig._inline: |
|
3683 | 3702 | raise error.Abort('not supporting inline revlog (yet)') |
|
3684 | 3703 | revlogkwargs = {} |
|
3685 | 3704 | k = 'upperboundcomp' |
|
3686 | 3705 | if util.safehasattr(orig, k): |
|
3687 | 3706 | revlogkwargs[k] = getattr(orig, k) |
|
3688 | 3707 | |
|
3689 | 3708 | indexfile = getattr(orig, '_indexfile', None) |
|
3690 | 3709 | if indexfile is None: |
|
3691 | 3710 | # compatibility with <= hg-5.8 |
|
3692 | 3711 | indexfile = getattr(orig, 'indexfile') |
|
3693 | 3712 | origindexpath = orig.opener.join(indexfile) |
|
3694 | 3713 | |
|
3695 | 3714 | datafile = getattr(orig, '_datafile', getattr(orig, 'datafile')) |
|
3696 | 3715 | origdatapath = orig.opener.join(datafile) |
|
3697 | 3716 | radix = b'revlog' |
|
3698 | 3717 | indexname = b'revlog.i' |
|
3699 | 3718 | dataname = b'revlog.d' |
|
3700 | 3719 | |
|
3701 | 3720 | tmpdir = tempfile.mkdtemp(prefix='tmp-hgperf-') |
|
3702 | 3721 | try: |
|
3703 | 3722 | # copy the data file in a temporary directory |
|
3704 | 3723 | ui.debug('copying data in %s\n' % tmpdir) |
|
3705 | 3724 | destindexpath = os.path.join(tmpdir, 'revlog.i') |
|
3706 | 3725 | destdatapath = os.path.join(tmpdir, 'revlog.d') |
|
3707 | 3726 | shutil.copyfile(origindexpath, destindexpath) |
|
3708 | 3727 | shutil.copyfile(origdatapath, destdatapath) |
|
3709 | 3728 | |
|
3710 | 3729 | # remove the data we want to add again |
|
3711 | 3730 | ui.debug('truncating data to be rewritten\n') |
|
3712 | 3731 | with open(destindexpath, 'ab') as index: |
|
3713 | 3732 | index.seek(0) |
|
3714 | 3733 | index.truncate(truncaterev * orig._io.size) |
|
3715 | 3734 | with open(destdatapath, 'ab') as data: |
|
3716 | 3735 | data.seek(0) |
|
3717 | 3736 | data.truncate(orig.start(truncaterev)) |
|
3718 | 3737 | |
|
3719 | 3738 | # instantiate a new revlog from the temporary copy |
|
3720 | 3739 | ui.debug('truncating adding to be rewritten\n') |
|
3721 | 3740 | vfs = vfsmod.vfs(tmpdir) |
|
3722 | 3741 | vfs.options = getattr(orig.opener, 'options', None) |
|
3723 | 3742 | |
|
3724 | 3743 | try: |
|
3725 | 3744 | dest = revlog(vfs, radix=radix, **revlogkwargs) |
|
3726 | 3745 | except TypeError: |
|
3727 | 3746 | dest = revlog( |
|
3728 | 3747 | vfs, indexfile=indexname, datafile=dataname, **revlogkwargs |
|
3729 | 3748 | ) |
|
3730 | 3749 | if dest._inline: |
|
3731 | 3750 | raise error.Abort('not supporting inline revlog (yet)') |
|
3732 | 3751 | # make sure internals are initialized |
|
3733 | 3752 | dest.revision(len(dest) - 1) |
|
3734 | 3753 | yield dest |
|
3735 | 3754 | del dest, vfs |
|
3736 | 3755 | finally: |
|
3737 | 3756 | shutil.rmtree(tmpdir, True) |
|
3738 | 3757 | |
|
3739 | 3758 | |
|
3740 | 3759 | @command( |
|
3741 | 3760 | b'perf::revlogchunks|perfrevlogchunks', |
|
3742 | 3761 | revlogopts |
|
3743 | 3762 | + formatteropts |
|
3744 | 3763 | + [ |
|
3745 | 3764 | (b'e', b'engines', b'', b'compression engines to use'), |
|
3746 | 3765 | (b's', b'startrev', 0, b'revision to start at'), |
|
3747 | 3766 | ], |
|
3748 | 3767 | b'-c|-m|FILE', |
|
3749 | 3768 | ) |
|
3750 | 3769 | def perfrevlogchunks(ui, repo, file_=None, engines=None, startrev=0, **opts): |
|
3751 | 3770 | """Benchmark operations on revlog chunks. |
|
3752 | 3771 | |
|
3753 | 3772 | Logically, each revlog is a collection of fulltext revisions. However, |
|
3754 | 3773 | stored within each revlog are "chunks" of possibly compressed data. This |
|
3755 | 3774 | data needs to be read and decompressed or compressed and written. |
|
3756 | 3775 | |
|
3757 | 3776 | This command measures the time it takes to read+decompress and recompress |
|
3758 | 3777 | chunks in a revlog. It effectively isolates I/O and compression performance. |
|
3759 | 3778 | For measurements of higher-level operations like resolving revisions, |
|
3760 | 3779 | see ``perfrevlogrevisions`` and ``perfrevlogrevision``. |
|
3761 | 3780 | """ |
|
3762 | 3781 | opts = _byteskwargs(opts) |
|
3763 | 3782 | |
|
3764 | 3783 | rl = cmdutil.openrevlog(repo, b'perfrevlogchunks', file_, opts) |
|
3765 | 3784 | |
|
3766 | 3785 | # - _chunkraw was renamed to _getsegmentforrevs |
|
3767 | 3786 | # - _getsegmentforrevs was moved on the inner object |
|
3768 | 3787 | try: |
|
3769 | 3788 | segmentforrevs = rl._inner.get_segment_for_revs |
|
3770 | 3789 | except AttributeError: |
|
3771 | 3790 | try: |
|
3772 | 3791 | segmentforrevs = rl._getsegmentforrevs |
|
3773 | 3792 | except AttributeError: |
|
3774 | 3793 | segmentforrevs = rl._chunkraw |
|
3775 | 3794 | |
|
3776 | 3795 | # Verify engines argument. |
|
3777 | 3796 | if engines: |
|
3778 | 3797 | engines = {e.strip() for e in engines.split(b',')} |
|
3779 | 3798 | for engine in engines: |
|
3780 | 3799 | try: |
|
3781 | 3800 | util.compressionengines[engine] |
|
3782 | 3801 | except KeyError: |
|
3783 | 3802 | raise error.Abort(b'unknown compression engine: %s' % engine) |
|
3784 | 3803 | else: |
|
3785 | 3804 | engines = [] |
|
3786 | 3805 | for e in util.compengines: |
|
3787 | 3806 | engine = util.compengines[e] |
|
3788 | 3807 | try: |
|
3789 | 3808 | if engine.available(): |
|
3790 | 3809 | engine.revlogcompressor().compress(b'dummy') |
|
3791 | 3810 | engines.append(e) |
|
3792 | 3811 | except NotImplementedError: |
|
3793 | 3812 | pass |
|
3794 | 3813 | |
|
3795 | 3814 | revs = list(rl.revs(startrev, len(rl) - 1)) |
|
3796 | 3815 | |
|
3797 | 3816 | @contextlib.contextmanager |
|
3798 | 3817 | def reading(rl): |
|
3799 | 3818 | if getattr(rl, 'reading', None) is not None: |
|
3800 | 3819 | with rl.reading(): |
|
3801 | 3820 | yield None |
|
3802 | 3821 | elif rl._inline: |
|
3803 | 3822 | indexfile = getattr(rl, '_indexfile', None) |
|
3804 | 3823 | if indexfile is None: |
|
3805 | 3824 | # compatibility with <= hg-5.8 |
|
3806 | 3825 | indexfile = getattr(rl, 'indexfile') |
|
3807 | 3826 | yield getsvfs(repo)(indexfile) |
|
3808 | 3827 | else: |
|
3809 | 3828 | datafile = getattr(rl, 'datafile', getattr(rl, 'datafile')) |
|
3810 | 3829 | yield getsvfs(repo)(datafile) |
|
3811 | 3830 | |
|
3812 | 3831 | if getattr(rl, 'reading', None) is not None: |
|
3813 | 3832 | |
|
3814 | 3833 | @contextlib.contextmanager |
|
3815 | 3834 | def lazy_reading(rl): |
|
3816 | 3835 | with rl.reading(): |
|
3817 | 3836 | yield |
|
3818 | 3837 | |
|
3819 | 3838 | else: |
|
3820 | 3839 | |
|
3821 | 3840 | @contextlib.contextmanager |
|
3822 | 3841 | def lazy_reading(rl): |
|
3823 | 3842 | yield |
|
3824 | 3843 | |
|
3825 | 3844 | def doread(): |
|
3826 | 3845 | rl.clearcaches() |
|
3827 | 3846 | for rev in revs: |
|
3828 | 3847 | with lazy_reading(rl): |
|
3829 | 3848 | segmentforrevs(rev, rev) |
|
3830 | 3849 | |
|
3831 | 3850 | def doreadcachedfh(): |
|
3832 | 3851 | rl.clearcaches() |
|
3833 | 3852 | with reading(rl) as fh: |
|
3834 | 3853 | if fh is not None: |
|
3835 | 3854 | for rev in revs: |
|
3836 | 3855 | segmentforrevs(rev, rev, df=fh) |
|
3837 | 3856 | else: |
|
3838 | 3857 | for rev in revs: |
|
3839 | 3858 | segmentforrevs(rev, rev) |
|
3840 | 3859 | |
|
3841 | 3860 | def doreadbatch(): |
|
3842 | 3861 | rl.clearcaches() |
|
3843 | 3862 | with lazy_reading(rl): |
|
3844 | 3863 | segmentforrevs(revs[0], revs[-1]) |
|
3845 | 3864 | |
|
3846 | 3865 | def doreadbatchcachedfh(): |
|
3847 | 3866 | rl.clearcaches() |
|
3848 | 3867 | with reading(rl) as fh: |
|
3849 | 3868 | if fh is not None: |
|
3850 | 3869 | segmentforrevs(revs[0], revs[-1], df=fh) |
|
3851 | 3870 | else: |
|
3852 | 3871 | segmentforrevs(revs[0], revs[-1]) |
|
3853 | 3872 | |
|
3854 | 3873 | def dochunk(): |
|
3855 | 3874 | rl.clearcaches() |
|
3856 | 3875 | # chunk used to be available directly on the revlog |
|
3857 | 3876 | _chunk = getattr(rl, '_inner', rl)._chunk |
|
3858 | 3877 | with reading(rl) as fh: |
|
3859 | 3878 | if fh is not None: |
|
3860 | 3879 | for rev in revs: |
|
3861 | 3880 | _chunk(rev, df=fh) |
|
3862 | 3881 | else: |
|
3863 | 3882 | for rev in revs: |
|
3864 | 3883 | _chunk(rev) |
|
3865 | 3884 | |
|
3866 | 3885 | chunks = [None] |
|
3867 | 3886 | |
|
3868 | 3887 | def dochunkbatch(): |
|
3869 | 3888 | rl.clearcaches() |
|
3870 | 3889 | _chunks = getattr(rl, '_inner', rl)._chunks |
|
3871 | 3890 | with reading(rl) as fh: |
|
3872 | 3891 | if fh is not None: |
|
3873 | 3892 | # Save chunks as a side-effect. |
|
3874 | 3893 | chunks[0] = _chunks(revs, df=fh) |
|
3875 | 3894 | else: |
|
3876 | 3895 | # Save chunks as a side-effect. |
|
3877 | 3896 | chunks[0] = _chunks(revs) |
|
3878 | 3897 | |
|
3879 | 3898 | def docompress(compressor): |
|
3880 | 3899 | rl.clearcaches() |
|
3881 | 3900 | |
|
3882 | 3901 | compressor_holder = getattr(rl, '_inner', rl) |
|
3883 | 3902 | |
|
3884 | 3903 | try: |
|
3885 | 3904 | # Swap in the requested compression engine. |
|
3886 | 3905 | oldcompressor = compressor_holder._compressor |
|
3887 | 3906 | compressor_holder._compressor = compressor |
|
3888 | 3907 | for chunk in chunks[0]: |
|
3889 | 3908 | rl.compress(chunk) |
|
3890 | 3909 | finally: |
|
3891 | 3910 | compressor_holder._compressor = oldcompressor |
|
3892 | 3911 | |
|
3893 | 3912 | benches = [ |
|
3894 | 3913 | (lambda: doread(), b'read'), |
|
3895 | 3914 | (lambda: doreadcachedfh(), b'read w/ reused fd'), |
|
3896 | 3915 | (lambda: doreadbatch(), b'read batch'), |
|
3897 | 3916 | (lambda: doreadbatchcachedfh(), b'read batch w/ reused fd'), |
|
3898 | 3917 | (lambda: dochunk(), b'chunk'), |
|
3899 | 3918 | (lambda: dochunkbatch(), b'chunk batch'), |
|
3900 | 3919 | ] |
|
3901 | 3920 | |
|
3902 | 3921 | for engine in sorted(engines): |
|
3903 | 3922 | compressor = util.compengines[engine].revlogcompressor() |
|
3904 | 3923 | benches.append( |
|
3905 | 3924 | ( |
|
3906 | 3925 | functools.partial(docompress, compressor), |
|
3907 | 3926 | b'compress w/ %s' % engine, |
|
3908 | 3927 | ) |
|
3909 | 3928 | ) |
|
3910 | 3929 | |
|
3911 | 3930 | for fn, title in benches: |
|
3912 | 3931 | timer, fm = gettimer(ui, opts) |
|
3913 | 3932 | timer(fn, title=title) |
|
3914 | 3933 | fm.end() |
|
3915 | 3934 | |
|
3916 | 3935 | |
|
3917 | 3936 | @command( |
|
3918 | 3937 | b'perf::revlogrevision|perfrevlogrevision', |
|
3919 | 3938 | revlogopts |
|
3920 | 3939 | + formatteropts |
|
3921 | 3940 | + [(b'', b'cache', False, b'use caches instead of clearing')], |
|
3922 | 3941 | b'-c|-m|FILE REV', |
|
3923 | 3942 | ) |
|
3924 | 3943 | def perfrevlogrevision(ui, repo, file_, rev=None, cache=None, **opts): |
|
3925 | 3944 | """Benchmark obtaining a revlog revision. |
|
3926 | 3945 | |
|
3927 | 3946 | Obtaining a revlog revision consists of roughly the following steps: |
|
3928 | 3947 | |
|
3929 | 3948 | 1. Compute the delta chain |
|
3930 | 3949 | 2. Slice the delta chain if applicable |
|
3931 | 3950 | 3. Obtain the raw chunks for that delta chain |
|
3932 | 3951 | 4. Decompress each raw chunk |
|
3933 | 3952 | 5. Apply binary patches to obtain fulltext |
|
3934 | 3953 | 6. Verify hash of fulltext |
|
3935 | 3954 | |
|
3936 | 3955 | This command measures the time spent in each of these phases. |
|
3937 | 3956 | """ |
|
3938 | 3957 | opts = _byteskwargs(opts) |
|
3939 | 3958 | |
|
3940 | 3959 | if opts.get(b'changelog') or opts.get(b'manifest'): |
|
3941 | 3960 | file_, rev = None, file_ |
|
3942 | 3961 | elif rev is None: |
|
3943 | 3962 | raise error.CommandError(b'perfrevlogrevision', b'invalid arguments') |
|
3944 | 3963 | |
|
3945 | 3964 | r = cmdutil.openrevlog(repo, b'perfrevlogrevision', file_, opts) |
|
3946 | 3965 | |
|
3947 | 3966 | # _chunkraw was renamed to _getsegmentforrevs. |
|
3948 | 3967 | try: |
|
3949 | 3968 | segmentforrevs = r._inner.get_segment_for_revs |
|
3950 | 3969 | except AttributeError: |
|
3951 | 3970 | try: |
|
3952 | 3971 | segmentforrevs = r._getsegmentforrevs |
|
3953 | 3972 | except AttributeError: |
|
3954 | 3973 | segmentforrevs = r._chunkraw |
|
3955 | 3974 | |
|
3956 | 3975 | node = r.lookup(rev) |
|
3957 | 3976 | rev = r.rev(node) |
|
3958 | 3977 | |
|
3959 | 3978 | if getattr(r, 'reading', None) is not None: |
|
3960 | 3979 | |
|
3961 | 3980 | @contextlib.contextmanager |
|
3962 | 3981 | def lazy_reading(r): |
|
3963 | 3982 | with r.reading(): |
|
3964 | 3983 | yield |
|
3965 | 3984 | |
|
3966 | 3985 | else: |
|
3967 | 3986 | |
|
3968 | 3987 | @contextlib.contextmanager |
|
3969 | 3988 | def lazy_reading(r): |
|
3970 | 3989 | yield |
|
3971 | 3990 | |
|
3972 | 3991 | def getrawchunks(data, chain): |
|
3973 | 3992 | start = r.start |
|
3974 | 3993 | length = r.length |
|
3975 | 3994 | inline = r._inline |
|
3976 | 3995 | try: |
|
3977 | 3996 | iosize = r.index.entry_size |
|
3978 | 3997 | except AttributeError: |
|
3979 | 3998 | iosize = r._io.size |
|
3980 | 3999 | buffer = util.buffer |
|
3981 | 4000 | |
|
3982 | 4001 | chunks = [] |
|
3983 | 4002 | ladd = chunks.append |
|
3984 | 4003 | for idx, item in enumerate(chain): |
|
3985 | 4004 | offset = start(item[0]) |
|
3986 | 4005 | bits = data[idx] |
|
3987 | 4006 | for rev in item: |
|
3988 | 4007 | chunkstart = start(rev) |
|
3989 | 4008 | if inline: |
|
3990 | 4009 | chunkstart += (rev + 1) * iosize |
|
3991 | 4010 | chunklength = length(rev) |
|
3992 | 4011 | ladd(buffer(bits, chunkstart - offset, chunklength)) |
|
3993 | 4012 | |
|
3994 | 4013 | return chunks |
|
3995 | 4014 | |
|
3996 | 4015 | def dodeltachain(rev): |
|
3997 | 4016 | if not cache: |
|
3998 | 4017 | r.clearcaches() |
|
3999 | 4018 | r._deltachain(rev) |
|
4000 | 4019 | |
|
4001 | 4020 | def doread(chain): |
|
4002 | 4021 | if not cache: |
|
4003 | 4022 | r.clearcaches() |
|
4004 | 4023 | for item in slicedchain: |
|
4005 | 4024 | with lazy_reading(r): |
|
4006 | 4025 | segmentforrevs(item[0], item[-1]) |
|
4007 | 4026 | |
|
4008 | 4027 | def doslice(r, chain, size): |
|
4009 | 4028 | for s in slicechunk(r, chain, targetsize=size): |
|
4010 | 4029 | pass |
|
4011 | 4030 | |
|
4012 | 4031 | def dorawchunks(data, chain): |
|
4013 | 4032 | if not cache: |
|
4014 | 4033 | r.clearcaches() |
|
4015 | 4034 | getrawchunks(data, chain) |
|
4016 | 4035 | |
|
4017 | 4036 | def dodecompress(chunks): |
|
4018 | 4037 | decomp = r.decompress |
|
4019 | 4038 | for chunk in chunks: |
|
4020 | 4039 | decomp(chunk) |
|
4021 | 4040 | |
|
4022 | 4041 | def dopatch(text, bins): |
|
4023 | 4042 | if not cache: |
|
4024 | 4043 | r.clearcaches() |
|
4025 | 4044 | mdiff.patches(text, bins) |
|
4026 | 4045 | |
|
4027 | 4046 | def dohash(text): |
|
4028 | 4047 | if not cache: |
|
4029 | 4048 | r.clearcaches() |
|
4030 | 4049 | r.checkhash(text, node, rev=rev) |
|
4031 | 4050 | |
|
4032 | 4051 | def dorevision(): |
|
4033 | 4052 | if not cache: |
|
4034 | 4053 | r.clearcaches() |
|
4035 | 4054 | r.revision(node) |
|
4036 | 4055 | |
|
4037 | 4056 | try: |
|
4038 | 4057 | from mercurial.revlogutils.deltas import slicechunk |
|
4039 | 4058 | except ImportError: |
|
4040 | 4059 | slicechunk = getattr(revlog, '_slicechunk', None) |
|
4041 | 4060 | |
|
4042 | 4061 | size = r.length(rev) |
|
4043 | 4062 | chain = r._deltachain(rev)[0] |
|
4044 | 4063 | |
|
4045 | 4064 | with_sparse_read = False |
|
4046 | 4065 | if hasattr(r, 'data_config'): |
|
4047 | 4066 | with_sparse_read = r.data_config.with_sparse_read |
|
4048 | 4067 | elif hasattr(r, '_withsparseread'): |
|
4049 | 4068 | with_sparse_read = r._withsparseread |
|
4050 | 4069 | if with_sparse_read: |
|
4051 | 4070 | slicedchain = (chain,) |
|
4052 | 4071 | else: |
|
4053 | 4072 | slicedchain = tuple(slicechunk(r, chain, targetsize=size)) |
|
4054 | 4073 | data = [segmentforrevs(seg[0], seg[-1])[1] for seg in slicedchain] |
|
4055 | 4074 | rawchunks = getrawchunks(data, slicedchain) |
|
4056 | 4075 | bins = r._inner._chunks(chain) |
|
4057 | 4076 | text = bytes(bins[0]) |
|
4058 | 4077 | bins = bins[1:] |
|
4059 | 4078 | text = mdiff.patches(text, bins) |
|
4060 | 4079 | |
|
4061 | 4080 | benches = [ |
|
4062 | 4081 | (lambda: dorevision(), b'full'), |
|
4063 | 4082 | (lambda: dodeltachain(rev), b'deltachain'), |
|
4064 | 4083 | (lambda: doread(chain), b'read'), |
|
4065 | 4084 | ] |
|
4066 | 4085 | |
|
4067 | 4086 | if with_sparse_read: |
|
4068 | 4087 | slicing = (lambda: doslice(r, chain, size), b'slice-sparse-chain') |
|
4069 | 4088 | benches.append(slicing) |
|
4070 | 4089 | |
|
4071 | 4090 | benches.extend( |
|
4072 | 4091 | [ |
|
4073 | 4092 | (lambda: dorawchunks(data, slicedchain), b'rawchunks'), |
|
4074 | 4093 | (lambda: dodecompress(rawchunks), b'decompress'), |
|
4075 | 4094 | (lambda: dopatch(text, bins), b'patch'), |
|
4076 | 4095 | (lambda: dohash(text), b'hash'), |
|
4077 | 4096 | ] |
|
4078 | 4097 | ) |
|
4079 | 4098 | |
|
4080 | 4099 | timer, fm = gettimer(ui, opts) |
|
4081 | 4100 | for fn, title in benches: |
|
4082 | 4101 | timer(fn, title=title) |
|
4083 | 4102 | fm.end() |
|
4084 | 4103 | |
|
4085 | 4104 | |
|
4086 | 4105 | @command( |
|
4087 | 4106 | b'perf::revset|perfrevset', |
|
4088 | 4107 | [ |
|
4089 | 4108 | (b'C', b'clear', False, b'clear volatile cache between each call.'), |
|
4090 | 4109 | (b'', b'contexts', False, b'obtain changectx for each revision'), |
|
4091 | 4110 | ] |
|
4092 | 4111 | + formatteropts, |
|
4093 | 4112 | b"REVSET", |
|
4094 | 4113 | ) |
|
4095 | 4114 | def perfrevset(ui, repo, expr, clear=False, contexts=False, **opts): |
|
4096 | 4115 | """benchmark the execution time of a revset |
|
4097 | 4116 | |
|
4098 | 4117 | Use the --clean option if need to evaluate the impact of build volatile |
|
4099 | 4118 | revisions set cache on the revset execution. Volatile cache hold filtered |
|
4100 | 4119 | and obsolete related cache.""" |
|
4101 | 4120 | opts = _byteskwargs(opts) |
|
4102 | 4121 | |
|
4103 | 4122 | timer, fm = gettimer(ui, opts) |
|
4104 | 4123 | |
|
4105 | 4124 | def d(): |
|
4106 | 4125 | if clear: |
|
4107 | 4126 | repo.invalidatevolatilesets() |
|
4108 | 4127 | if contexts: |
|
4109 | 4128 | for ctx in repo.set(expr): |
|
4110 | 4129 | pass |
|
4111 | 4130 | else: |
|
4112 | 4131 | for r in repo.revs(expr): |
|
4113 | 4132 | pass |
|
4114 | 4133 | |
|
4115 | 4134 | timer(d) |
|
4116 | 4135 | fm.end() |
|
4117 | 4136 | |
|
4118 | 4137 | |
|
4119 | 4138 | @command( |
|
4120 | 4139 | b'perf::volatilesets|perfvolatilesets', |
|
4121 | 4140 | [ |
|
4122 | 4141 | (b'', b'clear-obsstore', False, b'drop obsstore between each call.'), |
|
4123 | 4142 | ] |
|
4124 | 4143 | + formatteropts, |
|
4125 | 4144 | ) |
|
4126 | 4145 | def perfvolatilesets(ui, repo, *names, **opts): |
|
4127 | 4146 | """benchmark the computation of various volatile set |
|
4128 | 4147 | |
|
4129 | 4148 | Volatile set computes element related to filtering and obsolescence.""" |
|
4130 | 4149 | opts = _byteskwargs(opts) |
|
4131 | 4150 | timer, fm = gettimer(ui, opts) |
|
4132 | 4151 | repo = repo.unfiltered() |
|
4133 | 4152 | |
|
4134 | 4153 | def getobs(name): |
|
4135 | 4154 | def d(): |
|
4136 | 4155 | repo.invalidatevolatilesets() |
|
4137 | 4156 | if opts[b'clear_obsstore']: |
|
4138 | 4157 | clearfilecache(repo, b'obsstore') |
|
4139 | 4158 | obsolete.getrevs(repo, name) |
|
4140 | 4159 | |
|
4141 | 4160 | return d |
|
4142 | 4161 | |
|
4143 | 4162 | allobs = sorted(obsolete.cachefuncs) |
|
4144 | 4163 | if names: |
|
4145 | 4164 | allobs = [n for n in allobs if n in names] |
|
4146 | 4165 | |
|
4147 | 4166 | for name in allobs: |
|
4148 | 4167 | timer(getobs(name), title=name) |
|
4149 | 4168 | |
|
4150 | 4169 | def getfiltered(name): |
|
4151 | 4170 | def d(): |
|
4152 | 4171 | repo.invalidatevolatilesets() |
|
4153 | 4172 | if opts[b'clear_obsstore']: |
|
4154 | 4173 | clearfilecache(repo, b'obsstore') |
|
4155 | 4174 | repoview.filterrevs(repo, name) |
|
4156 | 4175 | |
|
4157 | 4176 | return d |
|
4158 | 4177 | |
|
4159 | 4178 | allfilter = sorted(repoview.filtertable) |
|
4160 | 4179 | if names: |
|
4161 | 4180 | allfilter = [n for n in allfilter if n in names] |
|
4162 | 4181 | |
|
4163 | 4182 | for name in allfilter: |
|
4164 | 4183 | timer(getfiltered(name), title=name) |
|
4165 | 4184 | fm.end() |
|
4166 | 4185 | |
|
4167 | 4186 | |
|
4168 | 4187 | @command( |
|
4169 | 4188 | b'perf::branchmap|perfbranchmap', |
|
4170 | 4189 | [ |
|
4171 | 4190 | (b'f', b'full', False, b'Includes build time of subset'), |
|
4172 | 4191 | ( |
|
4173 | 4192 | b'', |
|
4174 | 4193 | b'clear-revbranch', |
|
4175 | 4194 | False, |
|
4176 | 4195 | b'purge the revbranch cache between computation', |
|
4177 | 4196 | ), |
|
4178 | 4197 | ] |
|
4179 | 4198 | + formatteropts, |
|
4180 | 4199 | ) |
|
4181 | 4200 | def perfbranchmap(ui, repo, *filternames, **opts): |
|
4182 | 4201 | """benchmark the update of a branchmap |
|
4183 | 4202 | |
|
4184 | 4203 | This benchmarks the full repo.branchmap() call with read and write disabled |
|
4185 | 4204 | """ |
|
4186 | 4205 | opts = _byteskwargs(opts) |
|
4187 | 4206 | full = opts.get(b"full", False) |
|
4188 | 4207 | clear_revbranch = opts.get(b"clear_revbranch", False) |
|
4189 | 4208 | timer, fm = gettimer(ui, opts) |
|
4190 | 4209 | |
|
4191 | 4210 | def getbranchmap(filtername): |
|
4192 | 4211 | """generate a benchmark function for the filtername""" |
|
4193 | 4212 | if filtername is None: |
|
4194 | 4213 | view = repo |
|
4195 | 4214 | else: |
|
4196 | 4215 | view = repo.filtered(filtername) |
|
4197 | 4216 | if util.safehasattr(view._branchcaches, '_per_filter'): |
|
4198 | 4217 | filtered = view._branchcaches._per_filter |
|
4199 | 4218 | else: |
|
4200 | 4219 | # older versions |
|
4201 | 4220 | filtered = view._branchcaches |
|
4202 | 4221 | |
|
4203 | 4222 | def d(): |
|
4204 | 4223 | if clear_revbranch: |
|
4205 | 4224 | repo.revbranchcache()._clear() |
|
4206 | 4225 | if full: |
|
4207 | 4226 | view._branchcaches.clear() |
|
4208 | 4227 | else: |
|
4209 | 4228 | filtered.pop(filtername, None) |
|
4210 | 4229 | view.branchmap() |
|
4211 | 4230 | |
|
4212 | 4231 | return d |
|
4213 | 4232 | |
|
4214 | 4233 | # add filter in smaller subset to bigger subset |
|
4215 | 4234 | possiblefilters = set(repoview.filtertable) |
|
4216 | 4235 | if filternames: |
|
4217 | 4236 | possiblefilters &= set(filternames) |
|
4218 | 4237 | subsettable = getbranchmapsubsettable() |
|
4219 | 4238 | allfilters = [] |
|
4220 | 4239 | while possiblefilters: |
|
4221 | 4240 | for name in possiblefilters: |
|
4222 | 4241 | subset = subsettable.get(name) |
|
4223 | 4242 | if subset not in possiblefilters: |
|
4224 | 4243 | break |
|
4225 | 4244 | else: |
|
4226 | 4245 | assert False, b'subset cycle %s!' % possiblefilters |
|
4227 | 4246 | allfilters.append(name) |
|
4228 | 4247 | possiblefilters.remove(name) |
|
4229 | 4248 | |
|
4230 | 4249 | # warm the cache |
|
4231 | 4250 | if not full: |
|
4232 | 4251 | for name in allfilters: |
|
4233 | 4252 | repo.filtered(name).branchmap() |
|
4234 | 4253 | if not filternames or b'unfiltered' in filternames: |
|
4235 | 4254 | # add unfiltered |
|
4236 | 4255 | allfilters.append(None) |
|
4237 | 4256 | |
|
4238 | 4257 | old_branch_cache_from_file = None |
|
4239 | 4258 | branchcacheread = None |
|
4240 | 4259 | if util.safehasattr(branchmap, 'branch_cache_from_file'): |
|
4241 | 4260 | old_branch_cache_from_file = branchmap.branch_cache_from_file |
|
4242 | 4261 | branchmap.branch_cache_from_file = lambda *args: None |
|
4243 | 4262 | elif util.safehasattr(branchmap.branchcache, 'fromfile'): |
|
4244 | 4263 | branchcacheread = safeattrsetter(branchmap.branchcache, b'fromfile') |
|
4245 | 4264 | branchcacheread.set(classmethod(lambda *args: None)) |
|
4246 | 4265 | else: |
|
4247 | 4266 | # older versions |
|
4248 | 4267 | branchcacheread = safeattrsetter(branchmap, b'read') |
|
4249 | 4268 | branchcacheread.set(lambda *args: None) |
|
4250 | 4269 | if util.safehasattr(branchmap, '_LocalBranchCache'): |
|
4251 | 4270 | branchcachewrite = safeattrsetter(branchmap._LocalBranchCache, b'write') |
|
4252 | 4271 | branchcachewrite.set(lambda *args: None) |
|
4253 | 4272 | else: |
|
4254 | 4273 | branchcachewrite = safeattrsetter(branchmap.branchcache, b'write') |
|
4255 | 4274 | branchcachewrite.set(lambda *args: None) |
|
4256 | 4275 | try: |
|
4257 | 4276 | for name in allfilters: |
|
4258 | 4277 | printname = name |
|
4259 | 4278 | if name is None: |
|
4260 | 4279 | printname = b'unfiltered' |
|
4261 | 4280 | timer(getbranchmap(name), title=printname) |
|
4262 | 4281 | finally: |
|
4263 | 4282 | if old_branch_cache_from_file is not None: |
|
4264 | 4283 | branchmap.branch_cache_from_file = old_branch_cache_from_file |
|
4265 | 4284 | if branchcacheread is not None: |
|
4266 | 4285 | branchcacheread.restore() |
|
4267 | 4286 | branchcachewrite.restore() |
|
4268 | 4287 | fm.end() |
|
4269 | 4288 | |
|
4270 | 4289 | |
|
4271 | 4290 | @command( |
|
4272 | 4291 | b'perf::branchmapupdate|perfbranchmapupdate', |
|
4273 | 4292 | [ |
|
4274 | 4293 | (b'', b'base', [], b'subset of revision to start from'), |
|
4275 | 4294 | (b'', b'target', [], b'subset of revision to end with'), |
|
4276 | 4295 | (b'', b'clear-caches', False, b'clear cache between each runs'), |
|
4277 | 4296 | ] |
|
4278 | 4297 | + formatteropts, |
|
4279 | 4298 | ) |
|
4280 | 4299 | def perfbranchmapupdate(ui, repo, base=(), target=(), **opts): |
|
4281 | 4300 | """benchmark branchmap update from for <base> revs to <target> revs |
|
4282 | 4301 | |
|
4283 | 4302 | If `--clear-caches` is passed, the following items will be reset before |
|
4284 | 4303 | each update: |
|
4285 | 4304 | * the changelog instance and associated indexes |
|
4286 | 4305 | * the rev-branch-cache instance |
|
4287 | 4306 | |
|
4288 | 4307 | Examples: |
|
4289 | 4308 | |
|
4290 | 4309 | # update for the one last revision |
|
4291 | 4310 | $ hg perfbranchmapupdate --base 'not tip' --target 'tip' |
|
4292 | 4311 | |
|
4293 | 4312 | $ update for change coming with a new branch |
|
4294 | 4313 | $ hg perfbranchmapupdate --base 'stable' --target 'default' |
|
4295 | 4314 | """ |
|
4296 | 4315 | from mercurial import branchmap |
|
4297 | 4316 | from mercurial import repoview |
|
4298 | 4317 | |
|
4299 | 4318 | opts = _byteskwargs(opts) |
|
4300 | 4319 | timer, fm = gettimer(ui, opts) |
|
4301 | 4320 | clearcaches = opts[b'clear_caches'] |
|
4302 | 4321 | unfi = repo.unfiltered() |
|
4303 | 4322 | x = [None] # used to pass data between closure |
|
4304 | 4323 | |
|
4305 | 4324 | # we use a `list` here to avoid possible side effect from smartset |
|
4306 | 4325 | baserevs = list(scmutil.revrange(repo, base)) |
|
4307 | 4326 | targetrevs = list(scmutil.revrange(repo, target)) |
|
4308 | 4327 | if not baserevs: |
|
4309 | 4328 | raise error.Abort(b'no revisions selected for --base') |
|
4310 | 4329 | if not targetrevs: |
|
4311 | 4330 | raise error.Abort(b'no revisions selected for --target') |
|
4312 | 4331 | |
|
4313 | 4332 | # make sure the target branchmap also contains the one in the base |
|
4314 | 4333 | targetrevs = list(set(baserevs) | set(targetrevs)) |
|
4315 | 4334 | targetrevs.sort() |
|
4316 | 4335 | |
|
4317 | 4336 | cl = repo.changelog |
|
4318 | 4337 | allbaserevs = list(cl.ancestors(baserevs, inclusive=True)) |
|
4319 | 4338 | allbaserevs.sort() |
|
4320 | 4339 | alltargetrevs = frozenset(cl.ancestors(targetrevs, inclusive=True)) |
|
4321 | 4340 | |
|
4322 | 4341 | newrevs = list(alltargetrevs.difference(allbaserevs)) |
|
4323 | 4342 | newrevs.sort() |
|
4324 | 4343 | |
|
4325 | 4344 | allrevs = frozenset(unfi.changelog.revs()) |
|
4326 | 4345 | basefilterrevs = frozenset(allrevs.difference(allbaserevs)) |
|
4327 | 4346 | targetfilterrevs = frozenset(allrevs.difference(alltargetrevs)) |
|
4328 | 4347 | |
|
4329 | 4348 | def basefilter(repo, visibilityexceptions=None): |
|
4330 | 4349 | return basefilterrevs |
|
4331 | 4350 | |
|
4332 | 4351 | def targetfilter(repo, visibilityexceptions=None): |
|
4333 | 4352 | return targetfilterrevs |
|
4334 | 4353 | |
|
4335 | 4354 | msg = b'benchmark of branchmap with %d revisions with %d new ones\n' |
|
4336 | 4355 | ui.status(msg % (len(allbaserevs), len(newrevs))) |
|
4337 | 4356 | if targetfilterrevs: |
|
4338 | 4357 | msg = b'(%d revisions still filtered)\n' |
|
4339 | 4358 | ui.status(msg % len(targetfilterrevs)) |
|
4340 | 4359 | |
|
4341 | 4360 | try: |
|
4342 | 4361 | repoview.filtertable[b'__perf_branchmap_update_base'] = basefilter |
|
4343 | 4362 | repoview.filtertable[b'__perf_branchmap_update_target'] = targetfilter |
|
4344 | 4363 | |
|
4345 | 4364 | baserepo = repo.filtered(b'__perf_branchmap_update_base') |
|
4346 | 4365 | targetrepo = repo.filtered(b'__perf_branchmap_update_target') |
|
4347 | 4366 | |
|
4348 | 4367 | bcache = repo.branchmap() |
|
4349 | 4368 | copy_method = 'copy' |
|
4350 | 4369 | |
|
4351 | 4370 | copy_base_kwargs = copy_base_kwargs = {} |
|
4352 | 4371 | if hasattr(bcache, 'copy'): |
|
4353 | 4372 | if 'repo' in getargspec(bcache.copy).args: |
|
4354 | 4373 | copy_base_kwargs = {"repo": baserepo} |
|
4355 | 4374 | copy_target_kwargs = {"repo": targetrepo} |
|
4356 | 4375 | else: |
|
4357 | 4376 | copy_method = 'inherit_for' |
|
4358 | 4377 | copy_base_kwargs = {"repo": baserepo} |
|
4359 | 4378 | copy_target_kwargs = {"repo": targetrepo} |
|
4360 | 4379 | |
|
4361 | 4380 | # try to find an existing branchmap to reuse |
|
4362 | 4381 | subsettable = getbranchmapsubsettable() |
|
4363 | 4382 | candidatefilter = subsettable.get(None) |
|
4364 | 4383 | while candidatefilter is not None: |
|
4365 | 4384 | candidatebm = repo.filtered(candidatefilter).branchmap() |
|
4366 | 4385 | if candidatebm.validfor(baserepo): |
|
4367 | 4386 | filtered = repoview.filterrevs(repo, candidatefilter) |
|
4368 | 4387 | missing = [r for r in allbaserevs if r in filtered] |
|
4369 | 4388 | base = getattr(candidatebm, copy_method)(**copy_base_kwargs) |
|
4370 | 4389 | base.update(baserepo, missing) |
|
4371 | 4390 | break |
|
4372 | 4391 | candidatefilter = subsettable.get(candidatefilter) |
|
4373 | 4392 | else: |
|
4374 | 4393 | # no suitable subset where found |
|
4375 | 4394 | base = branchmap.branchcache() |
|
4376 | 4395 | base.update(baserepo, allbaserevs) |
|
4377 | 4396 | |
|
4378 | 4397 | def setup(): |
|
4379 | 4398 | x[0] = getattr(base, copy_method)(**copy_target_kwargs) |
|
4380 | 4399 | if clearcaches: |
|
4381 | 4400 | unfi._revbranchcache = None |
|
4382 | 4401 | clearchangelog(repo) |
|
4383 | 4402 | |
|
4384 | 4403 | def bench(): |
|
4385 | 4404 | x[0].update(targetrepo, newrevs) |
|
4386 | 4405 | |
|
4387 | 4406 | timer(bench, setup=setup) |
|
4388 | 4407 | fm.end() |
|
4389 | 4408 | finally: |
|
4390 | 4409 | repoview.filtertable.pop(b'__perf_branchmap_update_base', None) |
|
4391 | 4410 | repoview.filtertable.pop(b'__perf_branchmap_update_target', None) |
|
4392 | 4411 | |
|
4393 | 4412 | |
|
4394 | 4413 | @command( |
|
4395 | 4414 | b'perf::branchmapload|perfbranchmapload', |
|
4396 | 4415 | [ |
|
4397 | 4416 | (b'f', b'filter', b'', b'Specify repoview filter'), |
|
4398 | 4417 | (b'', b'list', False, b'List brachmap filter caches'), |
|
4399 | 4418 | (b'', b'clear-revlogs', False, b'refresh changelog and manifest'), |
|
4400 | 4419 | ] |
|
4401 | 4420 | + formatteropts, |
|
4402 | 4421 | ) |
|
4403 | 4422 | def perfbranchmapload(ui, repo, filter=b'', list=False, **opts): |
|
4404 | 4423 | """benchmark reading the branchmap""" |
|
4405 | 4424 | opts = _byteskwargs(opts) |
|
4406 | 4425 | clearrevlogs = opts[b'clear_revlogs'] |
|
4407 | 4426 | |
|
4408 | 4427 | if list: |
|
4409 | 4428 | for name, kind, st in repo.cachevfs.readdir(stat=True): |
|
4410 | 4429 | if name.startswith(b'branch2'): |
|
4411 | 4430 | filtername = name.partition(b'-')[2] or b'unfiltered' |
|
4412 | 4431 | ui.status( |
|
4413 | 4432 | b'%s - %s\n' % (filtername, util.bytecount(st.st_size)) |
|
4414 | 4433 | ) |
|
4415 | 4434 | return |
|
4416 | 4435 | if not filter: |
|
4417 | 4436 | filter = None |
|
4418 | 4437 | subsettable = getbranchmapsubsettable() |
|
4419 | 4438 | if filter is None: |
|
4420 | 4439 | repo = repo.unfiltered() |
|
4421 | 4440 | else: |
|
4422 | 4441 | repo = repoview.repoview(repo, filter) |
|
4423 | 4442 | |
|
4424 | 4443 | repo.branchmap() # make sure we have a relevant, up to date branchmap |
|
4425 | 4444 | |
|
4426 | 4445 | fromfile = getattr(branchmap, 'branch_cache_from_file', None) |
|
4427 | 4446 | if fromfile is None: |
|
4428 | 4447 | fromfile = getattr(branchmap.branchcache, 'fromfile', None) |
|
4429 | 4448 | if fromfile is None: |
|
4430 | 4449 | fromfile = branchmap.read |
|
4431 | 4450 | |
|
4432 | 4451 | currentfilter = filter |
|
4433 | 4452 | # try once without timer, the filter may not be cached |
|
4434 | 4453 | while fromfile(repo) is None: |
|
4435 | 4454 | currentfilter = subsettable.get(currentfilter) |
|
4436 | 4455 | if currentfilter is None: |
|
4437 | 4456 | raise error.Abort( |
|
4438 | 4457 | b'No branchmap cached for %s repo' % (filter or b'unfiltered') |
|
4439 | 4458 | ) |
|
4440 | 4459 | repo = repo.filtered(currentfilter) |
|
4441 | 4460 | timer, fm = gettimer(ui, opts) |
|
4442 | 4461 | |
|
4443 | 4462 | def setup(): |
|
4444 | 4463 | if clearrevlogs: |
|
4445 | 4464 | clearchangelog(repo) |
|
4446 | 4465 | |
|
4447 | 4466 | def bench(): |
|
4448 | 4467 | fromfile(repo) |
|
4449 | 4468 | |
|
4450 | 4469 | timer(bench, setup=setup) |
|
4451 | 4470 | fm.end() |
|
4452 | 4471 | |
|
4453 | 4472 | |
|
4454 | 4473 | @command(b'perf::loadmarkers|perfloadmarkers') |
|
4455 | 4474 | def perfloadmarkers(ui, repo): |
|
4456 | 4475 | """benchmark the time to parse the on-disk markers for a repo |
|
4457 | 4476 | |
|
4458 | 4477 | Result is the number of markers in the repo.""" |
|
4459 | 4478 | timer, fm = gettimer(ui) |
|
4460 | 4479 | svfs = getsvfs(repo) |
|
4461 | 4480 | timer(lambda: len(obsolete.obsstore(repo, svfs))) |
|
4462 | 4481 | fm.end() |
|
4463 | 4482 | |
|
4464 | 4483 | |
|
4465 | 4484 | @command( |
|
4466 | 4485 | b'perf::lrucachedict|perflrucachedict', |
|
4467 | 4486 | formatteropts |
|
4468 | 4487 | + [ |
|
4469 | 4488 | (b'', b'costlimit', 0, b'maximum total cost of items in cache'), |
|
4470 | 4489 | (b'', b'mincost', 0, b'smallest cost of items in cache'), |
|
4471 | 4490 | (b'', b'maxcost', 100, b'maximum cost of items in cache'), |
|
4472 | 4491 | (b'', b'size', 4, b'size of cache'), |
|
4473 | 4492 | (b'', b'gets', 10000, b'number of key lookups'), |
|
4474 | 4493 | (b'', b'sets', 10000, b'number of key sets'), |
|
4475 | 4494 | (b'', b'mixed', 10000, b'number of mixed mode operations'), |
|
4476 | 4495 | ( |
|
4477 | 4496 | b'', |
|
4478 | 4497 | b'mixedgetfreq', |
|
4479 | 4498 | 50, |
|
4480 | 4499 | b'frequency of get vs set ops in mixed mode', |
|
4481 | 4500 | ), |
|
4482 | 4501 | ], |
|
4483 | 4502 | norepo=True, |
|
4484 | 4503 | ) |
|
4485 | 4504 | def perflrucache( |
|
4486 | 4505 | ui, |
|
4487 | 4506 | mincost=0, |
|
4488 | 4507 | maxcost=100, |
|
4489 | 4508 | costlimit=0, |
|
4490 | 4509 | size=4, |
|
4491 | 4510 | gets=10000, |
|
4492 | 4511 | sets=10000, |
|
4493 | 4512 | mixed=10000, |
|
4494 | 4513 | mixedgetfreq=50, |
|
4495 | 4514 | **opts |
|
4496 | 4515 | ): |
|
4497 | 4516 | opts = _byteskwargs(opts) |
|
4498 | 4517 | |
|
4499 | 4518 | def doinit(): |
|
4500 | 4519 | for i in _xrange(10000): |
|
4501 | 4520 | util.lrucachedict(size) |
|
4502 | 4521 | |
|
4503 | 4522 | costrange = list(range(mincost, maxcost + 1)) |
|
4504 | 4523 | |
|
4505 | 4524 | values = [] |
|
4506 | 4525 | for i in _xrange(size): |
|
4507 | 4526 | values.append(random.randint(0, _maxint)) |
|
4508 | 4527 | |
|
4509 | 4528 | # Get mode fills the cache and tests raw lookup performance with no |
|
4510 | 4529 | # eviction. |
|
4511 | 4530 | getseq = [] |
|
4512 | 4531 | for i in _xrange(gets): |
|
4513 | 4532 | getseq.append(random.choice(values)) |
|
4514 | 4533 | |
|
4515 | 4534 | def dogets(): |
|
4516 | 4535 | d = util.lrucachedict(size) |
|
4517 | 4536 | for v in values: |
|
4518 | 4537 | d[v] = v |
|
4519 | 4538 | for key in getseq: |
|
4520 | 4539 | value = d[key] |
|
4521 | 4540 | value # silence pyflakes warning |
|
4522 | 4541 | |
|
4523 | 4542 | def dogetscost(): |
|
4524 | 4543 | d = util.lrucachedict(size, maxcost=costlimit) |
|
4525 | 4544 | for i, v in enumerate(values): |
|
4526 | 4545 | d.insert(v, v, cost=costs[i]) |
|
4527 | 4546 | for key in getseq: |
|
4528 | 4547 | try: |
|
4529 | 4548 | value = d[key] |
|
4530 | 4549 | value # silence pyflakes warning |
|
4531 | 4550 | except KeyError: |
|
4532 | 4551 | pass |
|
4533 | 4552 | |
|
4534 | 4553 | # Set mode tests insertion speed with cache eviction. |
|
4535 | 4554 | setseq = [] |
|
4536 | 4555 | costs = [] |
|
4537 | 4556 | for i in _xrange(sets): |
|
4538 | 4557 | setseq.append(random.randint(0, _maxint)) |
|
4539 | 4558 | costs.append(random.choice(costrange)) |
|
4540 | 4559 | |
|
4541 | 4560 | def doinserts(): |
|
4542 | 4561 | d = util.lrucachedict(size) |
|
4543 | 4562 | for v in setseq: |
|
4544 | 4563 | d.insert(v, v) |
|
4545 | 4564 | |
|
4546 | 4565 | def doinsertscost(): |
|
4547 | 4566 | d = util.lrucachedict(size, maxcost=costlimit) |
|
4548 | 4567 | for i, v in enumerate(setseq): |
|
4549 | 4568 | d.insert(v, v, cost=costs[i]) |
|
4550 | 4569 | |
|
4551 | 4570 | def dosets(): |
|
4552 | 4571 | d = util.lrucachedict(size) |
|
4553 | 4572 | for v in setseq: |
|
4554 | 4573 | d[v] = v |
|
4555 | 4574 | |
|
4556 | 4575 | # Mixed mode randomly performs gets and sets with eviction. |
|
4557 | 4576 | mixedops = [] |
|
4558 | 4577 | for i in _xrange(mixed): |
|
4559 | 4578 | r = random.randint(0, 100) |
|
4560 | 4579 | if r < mixedgetfreq: |
|
4561 | 4580 | op = 0 |
|
4562 | 4581 | else: |
|
4563 | 4582 | op = 1 |
|
4564 | 4583 | |
|
4565 | 4584 | mixedops.append( |
|
4566 | 4585 | (op, random.randint(0, size * 2), random.choice(costrange)) |
|
4567 | 4586 | ) |
|
4568 | 4587 | |
|
4569 | 4588 | def domixed(): |
|
4570 | 4589 | d = util.lrucachedict(size) |
|
4571 | 4590 | |
|
4572 | 4591 | for op, v, cost in mixedops: |
|
4573 | 4592 | if op == 0: |
|
4574 | 4593 | try: |
|
4575 | 4594 | d[v] |
|
4576 | 4595 | except KeyError: |
|
4577 | 4596 | pass |
|
4578 | 4597 | else: |
|
4579 | 4598 | d[v] = v |
|
4580 | 4599 | |
|
4581 | 4600 | def domixedcost(): |
|
4582 | 4601 | d = util.lrucachedict(size, maxcost=costlimit) |
|
4583 | 4602 | |
|
4584 | 4603 | for op, v, cost in mixedops: |
|
4585 | 4604 | if op == 0: |
|
4586 | 4605 | try: |
|
4587 | 4606 | d[v] |
|
4588 | 4607 | except KeyError: |
|
4589 | 4608 | pass |
|
4590 | 4609 | else: |
|
4591 | 4610 | d.insert(v, v, cost=cost) |
|
4592 | 4611 | |
|
4593 | 4612 | benches = [ |
|
4594 | 4613 | (doinit, b'init'), |
|
4595 | 4614 | ] |
|
4596 | 4615 | |
|
4597 | 4616 | if costlimit: |
|
4598 | 4617 | benches.extend( |
|
4599 | 4618 | [ |
|
4600 | 4619 | (dogetscost, b'gets w/ cost limit'), |
|
4601 | 4620 | (doinsertscost, b'inserts w/ cost limit'), |
|
4602 | 4621 | (domixedcost, b'mixed w/ cost limit'), |
|
4603 | 4622 | ] |
|
4604 | 4623 | ) |
|
4605 | 4624 | else: |
|
4606 | 4625 | benches.extend( |
|
4607 | 4626 | [ |
|
4608 | 4627 | (dogets, b'gets'), |
|
4609 | 4628 | (doinserts, b'inserts'), |
|
4610 | 4629 | (dosets, b'sets'), |
|
4611 | 4630 | (domixed, b'mixed'), |
|
4612 | 4631 | ] |
|
4613 | 4632 | ) |
|
4614 | 4633 | |
|
4615 | 4634 | for fn, title in benches: |
|
4616 | 4635 | timer, fm = gettimer(ui, opts) |
|
4617 | 4636 | timer(fn, title=title) |
|
4618 | 4637 | fm.end() |
|
4619 | 4638 | |
|
4620 | 4639 | |
|
4621 | 4640 | @command( |
|
4622 | 4641 | b'perf::write|perfwrite', |
|
4623 | 4642 | formatteropts |
|
4624 | 4643 | + [ |
|
4625 | 4644 | (b'', b'write-method', b'write', b'ui write method'), |
|
4626 | 4645 | (b'', b'nlines', 100, b'number of lines'), |
|
4627 | 4646 | (b'', b'nitems', 100, b'number of items (per line)'), |
|
4628 | 4647 | (b'', b'item', b'x', b'item that is written'), |
|
4629 | 4648 | (b'', b'batch-line', None, b'pass whole line to write method at once'), |
|
4630 | 4649 | (b'', b'flush-line', None, b'flush after each line'), |
|
4631 | 4650 | ], |
|
4632 | 4651 | ) |
|
4633 | 4652 | def perfwrite(ui, repo, **opts): |
|
4634 | 4653 | """microbenchmark ui.write (and others)""" |
|
4635 | 4654 | opts = _byteskwargs(opts) |
|
4636 | 4655 | |
|
4637 | 4656 | write = getattr(ui, _sysstr(opts[b'write_method'])) |
|
4638 | 4657 | nlines = int(opts[b'nlines']) |
|
4639 | 4658 | nitems = int(opts[b'nitems']) |
|
4640 | 4659 | item = opts[b'item'] |
|
4641 | 4660 | batch_line = opts.get(b'batch_line') |
|
4642 | 4661 | flush_line = opts.get(b'flush_line') |
|
4643 | 4662 | |
|
4644 | 4663 | if batch_line: |
|
4645 | 4664 | line = item * nitems + b'\n' |
|
4646 | 4665 | |
|
4647 | 4666 | def benchmark(): |
|
4648 | 4667 | for i in pycompat.xrange(nlines): |
|
4649 | 4668 | if batch_line: |
|
4650 | 4669 | write(line) |
|
4651 | 4670 | else: |
|
4652 | 4671 | for i in pycompat.xrange(nitems): |
|
4653 | 4672 | write(item) |
|
4654 | 4673 | write(b'\n') |
|
4655 | 4674 | if flush_line: |
|
4656 | 4675 | ui.flush() |
|
4657 | 4676 | ui.flush() |
|
4658 | 4677 | |
|
4659 | 4678 | timer, fm = gettimer(ui, opts) |
|
4660 | 4679 | timer(benchmark) |
|
4661 | 4680 | fm.end() |
|
4662 | 4681 | |
|
4663 | 4682 | |
|
4664 | 4683 | def uisetup(ui): |
|
4665 | 4684 | if util.safehasattr(cmdutil, b'openrevlog') and not util.safehasattr( |
|
4666 | 4685 | commands, b'debugrevlogopts' |
|
4667 | 4686 | ): |
|
4668 | 4687 | # for "historical portability": |
|
4669 | 4688 | # In this case, Mercurial should be 1.9 (or a79fea6b3e77) - |
|
4670 | 4689 | # 3.7 (or 5606f7d0d063). Therefore, '--dir' option for |
|
4671 | 4690 | # openrevlog() should cause failure, because it has been |
|
4672 | 4691 | # available since 3.5 (or 49c583ca48c4). |
|
4673 | 4692 | def openrevlog(orig, repo, cmd, file_, opts): |
|
4674 | 4693 | if opts.get(b'dir') and not util.safehasattr(repo, b'dirlog'): |
|
4675 | 4694 | raise error.Abort( |
|
4676 | 4695 | b"This version doesn't support --dir option", |
|
4677 | 4696 | hint=b"use 3.5 or later", |
|
4678 | 4697 | ) |
|
4679 | 4698 | return orig(repo, cmd, file_, opts) |
|
4680 | 4699 | |
|
4681 | 4700 | name = _sysstr(b'openrevlog') |
|
4682 | 4701 | extensions.wrapfunction(cmdutil, name, openrevlog) |
|
4683 | 4702 | |
|
4684 | 4703 | |
|
4685 | 4704 | @command( |
|
4686 | 4705 | b'perf::progress|perfprogress', |
|
4687 | 4706 | formatteropts |
|
4688 | 4707 | + [ |
|
4689 | 4708 | (b'', b'topic', b'topic', b'topic for progress messages'), |
|
4690 | 4709 | (b'c', b'total', 1000000, b'total value we are progressing to'), |
|
4691 | 4710 | ], |
|
4692 | 4711 | norepo=True, |
|
4693 | 4712 | ) |
|
4694 | 4713 | def perfprogress(ui, topic=None, total=None, **opts): |
|
4695 | 4714 | """printing of progress bars""" |
|
4696 | 4715 | opts = _byteskwargs(opts) |
|
4697 | 4716 | |
|
4698 | 4717 | timer, fm = gettimer(ui, opts) |
|
4699 | 4718 | |
|
4700 | 4719 | def doprogress(): |
|
4701 | 4720 | with ui.makeprogress(topic, total=total) as progress: |
|
4702 | 4721 | for i in _xrange(total): |
|
4703 | 4722 | progress.increment() |
|
4704 | 4723 | |
|
4705 | 4724 | timer(doprogress) |
|
4706 | 4725 | fm.end() |
@@ -1,387 +1,394 b'' | |||
|
1 | 1 | import contextlib |
|
2 | 2 | import errno |
|
3 | 3 | import os |
|
4 | 4 | import posixpath |
|
5 | 5 | import stat |
|
6 | 6 | |
|
7 | 7 | from typing import ( |
|
8 | 8 | Any, |
|
9 | 9 | Callable, |
|
10 | 10 | Iterator, |
|
11 | 11 | Optional, |
|
12 | 12 | ) |
|
13 | 13 | |
|
14 | 14 | from .i18n import _ |
|
15 | 15 | from . import ( |
|
16 | 16 | encoding, |
|
17 | 17 | error, |
|
18 | 18 | policy, |
|
19 | 19 | pycompat, |
|
20 | 20 | util, |
|
21 | 21 | ) |
|
22 | 22 | |
|
23 | 23 | rustdirs = policy.importrust('dirstate', 'Dirs') |
|
24 | 24 | parsers = policy.importmod('parsers') |
|
25 | 25 | |
|
26 | 26 | |
|
27 | 27 | def _lowerclean(s: bytes) -> bytes: |
|
28 | 28 | return encoding.hfsignoreclean(s.lower()) |
|
29 | 29 | |
|
30 | 30 | |
|
31 | 31 | class pathauditor: |
|
32 | 32 | """ensure that a filesystem path contains no banned components. |
|
33 | 33 | the following properties of a path are checked: |
|
34 | 34 | |
|
35 | 35 | - ends with a directory separator |
|
36 | 36 | - under top-level .hg |
|
37 | 37 | - starts at the root of a windows drive |
|
38 | 38 | - contains ".." |
|
39 | 39 | |
|
40 | 40 | More check are also done about the file system states: |
|
41 | 41 | - traverses a symlink (e.g. a/symlink_here/b) |
|
42 | 42 | - inside a nested repository (a callback can be used to approve |
|
43 | 43 | some nested repositories, e.g., subrepositories) |
|
44 | 44 | |
|
45 | 45 | The file system checks are only done when 'realfs' is set to True (the |
|
46 | 46 | default). They should be disable then we are auditing path for operation on |
|
47 | 47 | stored history. |
|
48 | 48 | |
|
49 | 49 | If 'cached' is set to True, audited paths and sub-directories are cached. |
|
50 | 50 | Be careful to not keep the cache of unmanaged directories for long because |
|
51 | 51 | audited paths may be replaced with symlinks. |
|
52 | 52 | """ |
|
53 | 53 | |
|
54 | 54 | def __init__(self, root, callback=None, realfs=True, cached=False): |
|
55 | 55 | self.audited = set() |
|
56 | 56 | self.auditeddir = dict() |
|
57 | 57 | self.root = root |
|
58 | 58 | self._realfs = realfs |
|
59 | 59 | self._cached = cached |
|
60 | 60 | self.callback = callback |
|
61 | 61 | if os.path.lexists(root) and not util.fscasesensitive(root): |
|
62 | 62 | self.normcase = util.normcase |
|
63 | 63 | else: |
|
64 | 64 | self.normcase = lambda x: x |
|
65 | 65 | |
|
66 | 66 | def __call__(self, path: bytes, mode: Optional[Any] = None) -> None: |
|
67 | 67 | """Check the relative path. |
|
68 | 68 | path may contain a pattern (e.g. foodir/**.txt)""" |
|
69 | 69 | |
|
70 | 70 | path = util.localpath(path) |
|
71 | 71 | if path in self.audited: |
|
72 | 72 | return |
|
73 | 73 | # AIX ignores "/" at end of path, others raise EISDIR. |
|
74 | 74 | if util.endswithsep(path): |
|
75 | 75 | raise error.InputError( |
|
76 | 76 | _(b"path ends in directory separator: %s") % path |
|
77 | 77 | ) |
|
78 | 78 | parts = util.splitpath(path) |
|
79 | 79 | if ( |
|
80 | 80 | os.path.splitdrive(path)[0] |
|
81 | 81 | or _lowerclean(parts[0]) in (b'.hg', b'.hg.', b'') |
|
82 | 82 | or pycompat.ospardir in parts |
|
83 | 83 | ): |
|
84 | 84 | raise error.InputError( |
|
85 | 85 | _(b"path contains illegal component: %s") % path |
|
86 | 86 | ) |
|
87 | 87 | # Windows shortname aliases |
|
88 | 88 | if b"~" in path: |
|
89 | 89 | for p in parts: |
|
90 | 90 | if b"~" in p: |
|
91 | 91 | first, last = p.split(b"~", 1) |
|
92 | 92 | if last.isdigit() and first.upper() in [b"HG", b"HG8B6C"]: |
|
93 | 93 | raise error.InputError( |
|
94 | 94 | _(b"path contains illegal component: %s") % path |
|
95 | 95 | ) |
|
96 | 96 | if b'.hg' in _lowerclean(path): |
|
97 | 97 | lparts = [_lowerclean(p) for p in parts] |
|
98 | 98 | for p in b'.hg', b'.hg.': |
|
99 | 99 | if p in lparts[1:]: |
|
100 | 100 | pos = lparts.index(p) |
|
101 | 101 | base = os.path.join(*parts[:pos]) |
|
102 | 102 | raise error.InputError( |
|
103 | 103 | _(b"path '%s' is inside nested repo %r") |
|
104 | 104 | % (path, pycompat.bytestr(base)) |
|
105 | 105 | ) |
|
106 | 106 | |
|
107 | 107 | if self._realfs: |
|
108 | 108 | # It's important that we check the path parts starting from the root. |
|
109 | 109 | # We don't want to add "foo/bar/baz" to auditeddir before checking if |
|
110 | 110 | # there's a "foo/.hg" directory. This also means we won't accidentally |
|
111 | 111 | # traverse a symlink into some other filesystem (which is potentially |
|
112 | 112 | # expensive to access). |
|
113 | 113 | for prefix in finddirs_rev_noroot(path): |
|
114 | 114 | if prefix in self.auditeddir: |
|
115 | 115 | res = self.auditeddir[prefix] |
|
116 | 116 | else: |
|
117 | 117 | res = pathauditor._checkfs_exists( |
|
118 | 118 | self.root, prefix, path, self.callback |
|
119 | 119 | ) |
|
120 | 120 | if self._cached: |
|
121 | 121 | self.auditeddir[prefix] = res |
|
122 | 122 | if not res: |
|
123 | 123 | break |
|
124 | 124 | |
|
125 | 125 | if self._cached: |
|
126 | 126 | self.audited.add(path) |
|
127 | 127 | |
|
128 | 128 | @staticmethod |
|
129 | 129 | def _checkfs_exists( |
|
130 | 130 | root, |
|
131 | 131 | prefix: bytes, |
|
132 | 132 | path: bytes, |
|
133 | 133 | callback: Optional[Callable[[bytes], bool]] = None, |
|
134 | 134 | ): |
|
135 | 135 | """raise exception if a file system backed check fails. |
|
136 | 136 | |
|
137 | 137 | Return a bool that indicates that the directory (or file) exists.""" |
|
138 | 138 | curpath = os.path.join(root, prefix) |
|
139 | 139 | try: |
|
140 | 140 | st = os.lstat(curpath) |
|
141 | 141 | except OSError as err: |
|
142 | 142 | if err.errno == errno.ENOENT: |
|
143 | 143 | return False |
|
144 | 144 | # EINVAL can be raised as invalid path syntax under win32. |
|
145 | 145 | # They must be ignored for patterns can be checked too. |
|
146 | 146 | if err.errno not in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL): |
|
147 | 147 | raise |
|
148 | 148 | else: |
|
149 | 149 | if stat.S_ISLNK(st.st_mode): |
|
150 | 150 | msg = _(b'path %r traverses symbolic link %r') % ( |
|
151 | 151 | pycompat.bytestr(path), |
|
152 | 152 | pycompat.bytestr(prefix), |
|
153 | 153 | ) |
|
154 | 154 | raise error.Abort(msg) |
|
155 | 155 | elif stat.S_ISDIR(st.st_mode) and os.path.isdir( |
|
156 | 156 | os.path.join(curpath, b'.hg') |
|
157 | 157 | ): |
|
158 | 158 | if not callback or not callback(curpath): |
|
159 | 159 | msg = _(b"path '%s' is inside nested repo %r") |
|
160 | 160 | raise error.Abort(msg % (path, pycompat.bytestr(prefix))) |
|
161 | 161 | return True |
|
162 | 162 | |
|
163 | 163 | def check(self, path: bytes) -> bool: |
|
164 | 164 | try: |
|
165 | 165 | self(path) |
|
166 | 166 | return True |
|
167 | 167 | except (OSError, error.Abort): |
|
168 | 168 | return False |
|
169 | 169 | |
|
170 | 170 | @contextlib.contextmanager |
|
171 | 171 | def cached(self): |
|
172 | 172 | if self._cached: |
|
173 | 173 | yield |
|
174 | 174 | else: |
|
175 | 175 | try: |
|
176 | 176 | self._cached = True |
|
177 | 177 | yield |
|
178 | 178 | finally: |
|
179 | 179 | self.audited.clear() |
|
180 | 180 | self.auditeddir.clear() |
|
181 | 181 | self._cached = False |
|
182 | 182 | |
|
183 | def clear_audit_cache(self): | |
|
184 | """reset all audit cache | |
|
185 | ||
|
186 | intended for debug and performance benchmark purposes""" | |
|
187 | self.audited.clear() | |
|
188 | self.auditeddir.clear() | |
|
189 | ||
|
183 | 190 | |
|
184 | 191 | def canonpath( |
|
185 | 192 | root: bytes, |
|
186 | 193 | cwd: bytes, |
|
187 | 194 | myname: bytes, |
|
188 | 195 | auditor: Optional[pathauditor] = None, |
|
189 | 196 | ) -> bytes: |
|
190 | 197 | """return the canonical path of myname, given cwd and root |
|
191 | 198 | |
|
192 | 199 | >>> def check(root, cwd, myname): |
|
193 | 200 | ... a = pathauditor(root, realfs=False) |
|
194 | 201 | ... try: |
|
195 | 202 | ... return canonpath(root, cwd, myname, a) |
|
196 | 203 | ... except error.Abort: |
|
197 | 204 | ... return 'aborted' |
|
198 | 205 | >>> def unixonly(root, cwd, myname, expected='aborted'): |
|
199 | 206 | ... if pycompat.iswindows: |
|
200 | 207 | ... return expected |
|
201 | 208 | ... return check(root, cwd, myname) |
|
202 | 209 | >>> def winonly(root, cwd, myname, expected='aborted'): |
|
203 | 210 | ... if not pycompat.iswindows: |
|
204 | 211 | ... return expected |
|
205 | 212 | ... return check(root, cwd, myname) |
|
206 | 213 | >>> winonly(b'd:\\\\repo', b'c:\\\\dir', b'filename') |
|
207 | 214 | 'aborted' |
|
208 | 215 | >>> winonly(b'c:\\\\repo', b'c:\\\\dir', b'filename') |
|
209 | 216 | 'aborted' |
|
210 | 217 | >>> winonly(b'c:\\\\repo', b'c:\\\\', b'filename') |
|
211 | 218 | 'aborted' |
|
212 | 219 | >>> winonly(b'c:\\\\repo', b'c:\\\\', b'repo\\\\filename', |
|
213 | 220 | ... b'filename') |
|
214 | 221 | 'filename' |
|
215 | 222 | >>> winonly(b'c:\\\\repo', b'c:\\\\repo', b'filename', b'filename') |
|
216 | 223 | 'filename' |
|
217 | 224 | >>> winonly(b'c:\\\\repo', b'c:\\\\repo\\\\subdir', b'filename', |
|
218 | 225 | ... b'subdir/filename') |
|
219 | 226 | 'subdir/filename' |
|
220 | 227 | >>> unixonly(b'/repo', b'/dir', b'filename') |
|
221 | 228 | 'aborted' |
|
222 | 229 | >>> unixonly(b'/repo', b'/', b'filename') |
|
223 | 230 | 'aborted' |
|
224 | 231 | >>> unixonly(b'/repo', b'/', b'repo/filename', b'filename') |
|
225 | 232 | 'filename' |
|
226 | 233 | >>> unixonly(b'/repo', b'/repo', b'filename', b'filename') |
|
227 | 234 | 'filename' |
|
228 | 235 | >>> unixonly(b'/repo', b'/repo/subdir', b'filename', b'subdir/filename') |
|
229 | 236 | 'subdir/filename' |
|
230 | 237 | """ |
|
231 | 238 | if util.endswithsep(root): |
|
232 | 239 | rootsep = root |
|
233 | 240 | else: |
|
234 | 241 | rootsep = root + pycompat.ossep |
|
235 | 242 | name = myname |
|
236 | 243 | if not os.path.isabs(name): |
|
237 | 244 | name = os.path.join(root, cwd, name) |
|
238 | 245 | name = os.path.normpath(name) |
|
239 | 246 | if auditor is None: |
|
240 | 247 | auditor = pathauditor(root) |
|
241 | 248 | if name != rootsep and name.startswith(rootsep): |
|
242 | 249 | name = name[len(rootsep) :] |
|
243 | 250 | auditor(name) |
|
244 | 251 | return util.pconvert(name) |
|
245 | 252 | elif name == root: |
|
246 | 253 | return b'' |
|
247 | 254 | else: |
|
248 | 255 | # Determine whether `name' is in the hierarchy at or beneath `root', |
|
249 | 256 | # by iterating name=dirname(name) until that causes no change (can't |
|
250 | 257 | # check name == '/', because that doesn't work on windows). The list |
|
251 | 258 | # `rel' holds the reversed list of components making up the relative |
|
252 | 259 | # file name we want. |
|
253 | 260 | rel = [] |
|
254 | 261 | while True: |
|
255 | 262 | try: |
|
256 | 263 | s = util.samefile(name, root) |
|
257 | 264 | except OSError: |
|
258 | 265 | s = False |
|
259 | 266 | if s: |
|
260 | 267 | if not rel: |
|
261 | 268 | # name was actually the same as root (maybe a symlink) |
|
262 | 269 | return b'' |
|
263 | 270 | rel.reverse() |
|
264 | 271 | name = os.path.join(*rel) |
|
265 | 272 | auditor(name) |
|
266 | 273 | return util.pconvert(name) |
|
267 | 274 | dirname, basename = util.split(name) |
|
268 | 275 | rel.append(basename) |
|
269 | 276 | if dirname == name: |
|
270 | 277 | break |
|
271 | 278 | name = dirname |
|
272 | 279 | |
|
273 | 280 | # A common mistake is to use -R, but specify a file relative to the repo |
|
274 | 281 | # instead of cwd. Detect that case, and provide a hint to the user. |
|
275 | 282 | hint = None |
|
276 | 283 | try: |
|
277 | 284 | if cwd != root: |
|
278 | 285 | canonpath(root, root, myname, auditor) |
|
279 | 286 | relpath = util.pathto(root, cwd, b'') |
|
280 | 287 | if relpath.endswith(pycompat.ossep): |
|
281 | 288 | relpath = relpath[:-1] |
|
282 | 289 | hint = _(b"consider using '--cwd %s'") % relpath |
|
283 | 290 | except error.Abort: |
|
284 | 291 | pass |
|
285 | 292 | |
|
286 | 293 | raise error.Abort( |
|
287 | 294 | _(b"%s not under root '%s'") % (myname, root), hint=hint |
|
288 | 295 | ) |
|
289 | 296 | |
|
290 | 297 | |
|
291 | 298 | def normasprefix(path: bytes) -> bytes: |
|
292 | 299 | """normalize the specified path as path prefix |
|
293 | 300 | |
|
294 | 301 | Returned value can be used safely for "p.startswith(prefix)", |
|
295 | 302 | "p[len(prefix):]", and so on. |
|
296 | 303 | |
|
297 | 304 | For efficiency, this expects "path" argument to be already |
|
298 | 305 | normalized by "os.path.normpath", "os.path.realpath", and so on. |
|
299 | 306 | |
|
300 | 307 | See also issue3033 for detail about need of this function. |
|
301 | 308 | |
|
302 | 309 | >>> normasprefix(b'/foo/bar').replace(pycompat.ossep, b'/') |
|
303 | 310 | '/foo/bar/' |
|
304 | 311 | >>> normasprefix(b'/').replace(pycompat.ossep, b'/') |
|
305 | 312 | '/' |
|
306 | 313 | """ |
|
307 | 314 | d, p = os.path.splitdrive(path) |
|
308 | 315 | if len(p) != len(pycompat.ossep): |
|
309 | 316 | return path + pycompat.ossep |
|
310 | 317 | else: |
|
311 | 318 | return path |
|
312 | 319 | |
|
313 | 320 | |
|
314 | 321 | def finddirs(path: bytes) -> Iterator[bytes]: |
|
315 | 322 | pos = path.rfind(b'/') |
|
316 | 323 | while pos != -1: |
|
317 | 324 | yield path[:pos] |
|
318 | 325 | pos = path.rfind(b'/', 0, pos) |
|
319 | 326 | yield b'' |
|
320 | 327 | |
|
321 | 328 | |
|
322 | 329 | def finddirs_rev_noroot(path: bytes) -> Iterator[bytes]: |
|
323 | 330 | pos = path.find(pycompat.ossep) |
|
324 | 331 | while pos != -1: |
|
325 | 332 | yield path[:pos] |
|
326 | 333 | pos = path.find(pycompat.ossep, pos + 1) |
|
327 | 334 | |
|
328 | 335 | |
|
329 | 336 | class dirs: |
|
330 | 337 | '''a multiset of directory names from a set of file paths''' |
|
331 | 338 | |
|
332 | 339 | def __init__(self, map, only_tracked=False): |
|
333 | 340 | """ |
|
334 | 341 | a dict map indicates a dirstate while a list indicates a manifest |
|
335 | 342 | """ |
|
336 | 343 | self._dirs = {} |
|
337 | 344 | addpath = self.addpath |
|
338 | 345 | if isinstance(map, dict) and only_tracked: |
|
339 | 346 | for f, s in map.items(): |
|
340 | 347 | if s.state != b'r': |
|
341 | 348 | addpath(f) |
|
342 | 349 | elif only_tracked: |
|
343 | 350 | msg = b"`only_tracked` is only supported with a dict source" |
|
344 | 351 | raise error.ProgrammingError(msg) |
|
345 | 352 | else: |
|
346 | 353 | for f in map: |
|
347 | 354 | addpath(f) |
|
348 | 355 | |
|
349 | 356 | def addpath(self, path: bytes) -> None: |
|
350 | 357 | dirs = self._dirs |
|
351 | 358 | for base in finddirs(path): |
|
352 | 359 | if base.endswith(b'/'): |
|
353 | 360 | raise ValueError( |
|
354 | 361 | "found invalid consecutive slashes in path: %r" % base |
|
355 | 362 | ) |
|
356 | 363 | if base in dirs: |
|
357 | 364 | dirs[base] += 1 |
|
358 | 365 | return |
|
359 | 366 | dirs[base] = 1 |
|
360 | 367 | |
|
361 | 368 | def delpath(self, path: bytes) -> None: |
|
362 | 369 | dirs = self._dirs |
|
363 | 370 | for base in finddirs(path): |
|
364 | 371 | if dirs[base] > 1: |
|
365 | 372 | dirs[base] -= 1 |
|
366 | 373 | return |
|
367 | 374 | del dirs[base] |
|
368 | 375 | |
|
369 | 376 | def __iter__(self): |
|
370 | 377 | return iter(self._dirs) |
|
371 | 378 | |
|
372 | 379 | def __contains__(self, d: bytes) -> bool: |
|
373 | 380 | return d in self._dirs |
|
374 | 381 | |
|
375 | 382 | |
|
376 | 383 | if hasattr(parsers, 'dirs'): |
|
377 | 384 | dirs = parsers.dirs |
|
378 | 385 | |
|
379 | 386 | if rustdirs is not None: |
|
380 | 387 | dirs = rustdirs |
|
381 | 388 | |
|
382 | 389 | |
|
383 | 390 | # forward two methods from posixpath that do what we need, but we'd |
|
384 | 391 | # rather not let our internals know that we're thinking in posix terms |
|
385 | 392 | # - instead we'll let them be oblivious. |
|
386 | 393 | join = posixpath.join |
|
387 | 394 | dirname: Callable[[bytes], bytes] = posixpath.dirname |
General Comments 0
You need to be logged in to leave comments.
Login now