Show More
@@ -1,2044 +1,2052 b'' | |||
|
1 | 1 | # manifest.py - manifest revision class for mercurial |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2005-2007 Matt Mackall <mpm@selenic.com> |
|
4 | 4 | # |
|
5 | 5 | # This software may be used and distributed according to the terms of the |
|
6 | 6 | # GNU General Public License version 2 or any later version. |
|
7 | 7 | |
|
8 | 8 | from __future__ import absolute_import |
|
9 | 9 | |
|
10 | 10 | import heapq |
|
11 | 11 | import itertools |
|
12 | 12 | import struct |
|
13 | 13 | import weakref |
|
14 | 14 | |
|
15 | 15 | from .i18n import _ |
|
16 | 16 | from .node import ( |
|
17 | 17 | bin, |
|
18 | 18 | hex, |
|
19 | 19 | nullid, |
|
20 | 20 | nullrev, |
|
21 | 21 | ) |
|
22 | 22 | from . import ( |
|
23 | 23 | error, |
|
24 | 24 | mdiff, |
|
25 | 25 | policy, |
|
26 | 26 | pycompat, |
|
27 | 27 | repository, |
|
28 | 28 | revlog, |
|
29 | 29 | util, |
|
30 | 30 | ) |
|
31 | 31 | from .utils import ( |
|
32 | 32 | interfaceutil, |
|
33 | 33 | ) |
|
34 | 34 | |
|
35 | 35 | parsers = policy.importmod(r'parsers') |
|
36 | 36 | propertycache = util.propertycache |
|
37 | 37 | |
|
38 | 38 | def _parse(data): |
|
39 | 39 | # This method does a little bit of excessive-looking |
|
40 | 40 | # precondition checking. This is so that the behavior of this |
|
41 | 41 | # class exactly matches its C counterpart to try and help |
|
42 | 42 | # prevent surprise breakage for anyone that develops against |
|
43 | 43 | # the pure version. |
|
44 | 44 | if data and data[-1:] != '\n': |
|
45 | 45 | raise ValueError('Manifest did not end in a newline.') |
|
46 | 46 | prev = None |
|
47 | 47 | for l in data.splitlines(): |
|
48 | 48 | if prev is not None and prev > l: |
|
49 | 49 | raise ValueError('Manifest lines not in sorted order.') |
|
50 | 50 | prev = l |
|
51 | 51 | f, n = l.split('\0') |
|
52 | 52 | if len(n) > 40: |
|
53 | 53 | yield f, bin(n[:40]), n[40:] |
|
54 | 54 | else: |
|
55 | 55 | yield f, bin(n), '' |
|
56 | 56 | |
|
57 | 57 | def _text(it): |
|
58 | 58 | files = [] |
|
59 | 59 | lines = [] |
|
60 | 60 | for f, n, fl in it: |
|
61 | 61 | files.append(f) |
|
62 | 62 | # if this is changed to support newlines in filenames, |
|
63 | 63 | # be sure to check the templates/ dir again (especially *-raw.tmpl) |
|
64 | 64 | lines.append("%s\0%s%s\n" % (f, hex(n), fl)) |
|
65 | 65 | |
|
66 | 66 | _checkforbidden(files) |
|
67 | 67 | return ''.join(lines) |
|
68 | 68 | |
|
69 | 69 | class lazymanifestiter(object): |
|
70 | 70 | def __init__(self, lm): |
|
71 | 71 | self.pos = 0 |
|
72 | 72 | self.lm = lm |
|
73 | 73 | |
|
74 | 74 | def __iter__(self): |
|
75 | 75 | return self |
|
76 | 76 | |
|
77 | 77 | def next(self): |
|
78 | 78 | try: |
|
79 | 79 | data, pos = self.lm._get(self.pos) |
|
80 | 80 | except IndexError: |
|
81 | 81 | raise StopIteration |
|
82 | 82 | if pos == -1: |
|
83 | 83 | self.pos += 1 |
|
84 | 84 | return data[0] |
|
85 | 85 | self.pos += 1 |
|
86 | 86 | zeropos = data.find('\x00', pos) |
|
87 | 87 | return data[pos:zeropos] |
|
88 | 88 | |
|
89 | 89 | __next__ = next |
|
90 | 90 | |
|
91 | 91 | class lazymanifestiterentries(object): |
|
92 | 92 | def __init__(self, lm): |
|
93 | 93 | self.lm = lm |
|
94 | 94 | self.pos = 0 |
|
95 | 95 | |
|
96 | 96 | def __iter__(self): |
|
97 | 97 | return self |
|
98 | 98 | |
|
99 | 99 | def next(self): |
|
100 | 100 | try: |
|
101 | 101 | data, pos = self.lm._get(self.pos) |
|
102 | 102 | except IndexError: |
|
103 | 103 | raise StopIteration |
|
104 | 104 | if pos == -1: |
|
105 | 105 | self.pos += 1 |
|
106 | 106 | return data |
|
107 | 107 | zeropos = data.find('\x00', pos) |
|
108 | 108 | hashval = unhexlify(data, self.lm.extrainfo[self.pos], |
|
109 | 109 | zeropos + 1, 40) |
|
110 | 110 | flags = self.lm._getflags(data, self.pos, zeropos) |
|
111 | 111 | self.pos += 1 |
|
112 | 112 | return (data[pos:zeropos], hashval, flags) |
|
113 | 113 | |
|
114 | 114 | __next__ = next |
|
115 | 115 | |
|
116 | 116 | def unhexlify(data, extra, pos, length): |
|
117 | 117 | s = bin(data[pos:pos + length]) |
|
118 | 118 | if extra: |
|
119 | 119 | s += chr(extra & 0xff) |
|
120 | 120 | return s |
|
121 | 121 | |
|
122 | 122 | def _cmp(a, b): |
|
123 | 123 | return (a > b) - (a < b) |
|
124 | 124 | |
|
125 | 125 | class _lazymanifest(object): |
|
126 | 126 | def __init__(self, data, positions=None, extrainfo=None, extradata=None): |
|
127 | 127 | if positions is None: |
|
128 | 128 | self.positions = self.findlines(data) |
|
129 | 129 | self.extrainfo = [0] * len(self.positions) |
|
130 | 130 | self.data = data |
|
131 | 131 | self.extradata = [] |
|
132 | 132 | else: |
|
133 | 133 | self.positions = positions[:] |
|
134 | 134 | self.extrainfo = extrainfo[:] |
|
135 | 135 | self.extradata = extradata[:] |
|
136 | 136 | self.data = data |
|
137 | 137 | |
|
138 | 138 | def findlines(self, data): |
|
139 | 139 | if not data: |
|
140 | 140 | return [] |
|
141 | 141 | pos = data.find("\n") |
|
142 | 142 | if pos == -1 or data[-1:] != '\n': |
|
143 | 143 | raise ValueError("Manifest did not end in a newline.") |
|
144 | 144 | positions = [0] |
|
145 | 145 | prev = data[:data.find('\x00')] |
|
146 | 146 | while pos < len(data) - 1 and pos != -1: |
|
147 | 147 | positions.append(pos + 1) |
|
148 | 148 | nexts = data[pos + 1:data.find('\x00', pos + 1)] |
|
149 | 149 | if nexts < prev: |
|
150 | 150 | raise ValueError("Manifest lines not in sorted order.") |
|
151 | 151 | prev = nexts |
|
152 | 152 | pos = data.find("\n", pos + 1) |
|
153 | 153 | return positions |
|
154 | 154 | |
|
155 | 155 | def _get(self, index): |
|
156 | 156 | # get the position encoded in pos: |
|
157 | 157 | # positive number is an index in 'data' |
|
158 | 158 | # negative number is in extrapieces |
|
159 | 159 | pos = self.positions[index] |
|
160 | 160 | if pos >= 0: |
|
161 | 161 | return self.data, pos |
|
162 | 162 | return self.extradata[-pos - 1], -1 |
|
163 | 163 | |
|
164 | 164 | def _getkey(self, pos): |
|
165 | 165 | if pos >= 0: |
|
166 | 166 | return self.data[pos:self.data.find('\x00', pos + 1)] |
|
167 | 167 | return self.extradata[-pos - 1][0] |
|
168 | 168 | |
|
169 | 169 | def bsearch(self, key): |
|
170 | 170 | first = 0 |
|
171 | 171 | last = len(self.positions) - 1 |
|
172 | 172 | |
|
173 | 173 | while first <= last: |
|
174 | 174 | midpoint = (first + last)//2 |
|
175 | 175 | nextpos = self.positions[midpoint] |
|
176 | 176 | candidate = self._getkey(nextpos) |
|
177 | 177 | r = _cmp(key, candidate) |
|
178 | 178 | if r == 0: |
|
179 | 179 | return midpoint |
|
180 | 180 | else: |
|
181 | 181 | if r < 0: |
|
182 | 182 | last = midpoint - 1 |
|
183 | 183 | else: |
|
184 | 184 | first = midpoint + 1 |
|
185 | 185 | return -1 |
|
186 | 186 | |
|
187 | 187 | def bsearch2(self, key): |
|
188 | 188 | # same as the above, but will always return the position |
|
189 | 189 | # done for performance reasons |
|
190 | 190 | first = 0 |
|
191 | 191 | last = len(self.positions) - 1 |
|
192 | 192 | |
|
193 | 193 | while first <= last: |
|
194 | 194 | midpoint = (first + last)//2 |
|
195 | 195 | nextpos = self.positions[midpoint] |
|
196 | 196 | candidate = self._getkey(nextpos) |
|
197 | 197 | r = _cmp(key, candidate) |
|
198 | 198 | if r == 0: |
|
199 | 199 | return (midpoint, True) |
|
200 | 200 | else: |
|
201 | 201 | if r < 0: |
|
202 | 202 | last = midpoint - 1 |
|
203 | 203 | else: |
|
204 | 204 | first = midpoint + 1 |
|
205 | 205 | return (first, False) |
|
206 | 206 | |
|
207 | 207 | def __contains__(self, key): |
|
208 | 208 | return self.bsearch(key) != -1 |
|
209 | 209 | |
|
210 | 210 | def _getflags(self, data, needle, pos): |
|
211 | 211 | start = pos + 41 |
|
212 | 212 | end = data.find("\n", start) |
|
213 | 213 | if end == -1: |
|
214 | 214 | end = len(data) - 1 |
|
215 | 215 | if start == end: |
|
216 | 216 | return '' |
|
217 | 217 | return self.data[start:end] |
|
218 | 218 | |
|
219 | 219 | def __getitem__(self, key): |
|
220 | 220 | if not isinstance(key, bytes): |
|
221 | 221 | raise TypeError("getitem: manifest keys must be a bytes.") |
|
222 | 222 | needle = self.bsearch(key) |
|
223 | 223 | if needle == -1: |
|
224 | 224 | raise KeyError |
|
225 | 225 | data, pos = self._get(needle) |
|
226 | 226 | if pos == -1: |
|
227 | 227 | return (data[1], data[2]) |
|
228 | 228 | zeropos = data.find('\x00', pos) |
|
229 | 229 | assert 0 <= needle <= len(self.positions) |
|
230 | 230 | assert len(self.extrainfo) == len(self.positions) |
|
231 | 231 | hashval = unhexlify(data, self.extrainfo[needle], zeropos + 1, 40) |
|
232 | 232 | flags = self._getflags(data, needle, zeropos) |
|
233 | 233 | return (hashval, flags) |
|
234 | 234 | |
|
235 | 235 | def __delitem__(self, key): |
|
236 | 236 | needle, found = self.bsearch2(key) |
|
237 | 237 | if not found: |
|
238 | 238 | raise KeyError |
|
239 | 239 | cur = self.positions[needle] |
|
240 | 240 | self.positions = self.positions[:needle] + self.positions[needle + 1:] |
|
241 | 241 | self.extrainfo = self.extrainfo[:needle] + self.extrainfo[needle + 1:] |
|
242 | 242 | if cur >= 0: |
|
243 | 243 | self.data = self.data[:cur] + '\x00' + self.data[cur + 1:] |
|
244 | 244 | |
|
245 | 245 | def __setitem__(self, key, value): |
|
246 | 246 | if not isinstance(key, bytes): |
|
247 | 247 | raise TypeError("setitem: manifest keys must be a byte string.") |
|
248 | 248 | if not isinstance(value, tuple) or len(value) != 2: |
|
249 | 249 | raise TypeError("Manifest values must be a tuple of (node, flags).") |
|
250 | 250 | hashval = value[0] |
|
251 | 251 | if not isinstance(hashval, bytes) or not 20 <= len(hashval) <= 22: |
|
252 | 252 | raise TypeError("node must be a 20-byte byte string") |
|
253 | 253 | flags = value[1] |
|
254 | 254 | if len(hashval) == 22: |
|
255 | 255 | hashval = hashval[:-1] |
|
256 | 256 | if not isinstance(flags, bytes) or len(flags) > 1: |
|
257 | 257 | raise TypeError("flags must a 0 or 1 byte string, got %r", flags) |
|
258 | 258 | needle, found = self.bsearch2(key) |
|
259 | 259 | if found: |
|
260 | 260 | # put the item |
|
261 | 261 | pos = self.positions[needle] |
|
262 | 262 | if pos < 0: |
|
263 | 263 | self.extradata[-pos - 1] = (key, hashval, value[1]) |
|
264 | 264 | else: |
|
265 | 265 | # just don't bother |
|
266 | 266 | self.extradata.append((key, hashval, value[1])) |
|
267 | 267 | self.positions[needle] = -len(self.extradata) |
|
268 | 268 | else: |
|
269 | 269 | # not found, put it in with extra positions |
|
270 | 270 | self.extradata.append((key, hashval, value[1])) |
|
271 | 271 | self.positions = (self.positions[:needle] + [-len(self.extradata)] |
|
272 | 272 | + self.positions[needle:]) |
|
273 | 273 | self.extrainfo = (self.extrainfo[:needle] + [0] + |
|
274 | 274 | self.extrainfo[needle:]) |
|
275 | 275 | |
|
276 | 276 | def copy(self): |
|
277 | 277 | # XXX call _compact like in C? |
|
278 | 278 | return _lazymanifest(self.data, self.positions, self.extrainfo, |
|
279 | 279 | self.extradata) |
|
280 | 280 | |
|
281 | 281 | def _compact(self): |
|
282 | 282 | # hopefully not called TOO often |
|
283 | 283 | if len(self.extradata) == 0: |
|
284 | 284 | return |
|
285 | 285 | l = [] |
|
286 | 286 | last_cut = 0 |
|
287 | 287 | i = 0 |
|
288 | 288 | offset = 0 |
|
289 | 289 | self.extrainfo = [0] * len(self.positions) |
|
290 | 290 | while i < len(self.positions): |
|
291 | 291 | if self.positions[i] >= 0: |
|
292 | 292 | cur = self.positions[i] |
|
293 | 293 | last_cut = cur |
|
294 | 294 | while True: |
|
295 | 295 | self.positions[i] = offset |
|
296 | 296 | i += 1 |
|
297 | 297 | if i == len(self.positions) or self.positions[i] < 0: |
|
298 | 298 | break |
|
299 | 299 | offset += self.positions[i] - cur |
|
300 | 300 | cur = self.positions[i] |
|
301 | 301 | end_cut = self.data.find('\n', cur) |
|
302 | 302 | if end_cut != -1: |
|
303 | 303 | end_cut += 1 |
|
304 | 304 | offset += end_cut - cur |
|
305 | 305 | l.append(self.data[last_cut:end_cut]) |
|
306 | 306 | else: |
|
307 | 307 | while i < len(self.positions) and self.positions[i] < 0: |
|
308 | 308 | cur = self.positions[i] |
|
309 | 309 | t = self.extradata[-cur - 1] |
|
310 | 310 | l.append(self._pack(t)) |
|
311 | 311 | self.positions[i] = offset |
|
312 | 312 | if len(t[1]) > 20: |
|
313 | 313 | self.extrainfo[i] = ord(t[1][21]) |
|
314 | 314 | offset += len(l[-1]) |
|
315 | 315 | i += 1 |
|
316 | 316 | self.data = ''.join(l) |
|
317 | 317 | self.extradata = [] |
|
318 | 318 | |
|
319 | 319 | def _pack(self, d): |
|
320 | 320 | return d[0] + '\x00' + hex(d[1][:20]) + d[2] + '\n' |
|
321 | 321 | |
|
322 | 322 | def text(self): |
|
323 | 323 | self._compact() |
|
324 | 324 | return self.data |
|
325 | 325 | |
|
326 | 326 | def diff(self, m2, clean=False): |
|
327 | 327 | '''Finds changes between the current manifest and m2.''' |
|
328 | 328 | # XXX think whether efficiency matters here |
|
329 | 329 | diff = {} |
|
330 | 330 | |
|
331 | 331 | for fn, e1, flags in self.iterentries(): |
|
332 | 332 | if fn not in m2: |
|
333 | 333 | diff[fn] = (e1, flags), (None, '') |
|
334 | 334 | else: |
|
335 | 335 | e2 = m2[fn] |
|
336 | 336 | if (e1, flags) != e2: |
|
337 | 337 | diff[fn] = (e1, flags), e2 |
|
338 | 338 | elif clean: |
|
339 | 339 | diff[fn] = None |
|
340 | 340 | |
|
341 | 341 | for fn, e2, flags in m2.iterentries(): |
|
342 | 342 | if fn not in self: |
|
343 | 343 | diff[fn] = (None, ''), (e2, flags) |
|
344 | 344 | |
|
345 | 345 | return diff |
|
346 | 346 | |
|
347 | 347 | def iterentries(self): |
|
348 | 348 | return lazymanifestiterentries(self) |
|
349 | 349 | |
|
350 | 350 | def iterkeys(self): |
|
351 | 351 | return lazymanifestiter(self) |
|
352 | 352 | |
|
353 | 353 | def __iter__(self): |
|
354 | 354 | return lazymanifestiter(self) |
|
355 | 355 | |
|
356 | 356 | def __len__(self): |
|
357 | 357 | return len(self.positions) |
|
358 | 358 | |
|
359 | 359 | def filtercopy(self, filterfn): |
|
360 | 360 | # XXX should be optimized |
|
361 | 361 | c = _lazymanifest('') |
|
362 | 362 | for f, n, fl in self.iterentries(): |
|
363 | 363 | if filterfn(f): |
|
364 | 364 | c[f] = n, fl |
|
365 | 365 | return c |
|
366 | 366 | |
|
367 | 367 | try: |
|
368 | 368 | _lazymanifest = parsers.lazymanifest |
|
369 | 369 | except AttributeError: |
|
370 | 370 | pass |
|
371 | 371 | |
|
372 | 372 | @interfaceutil.implementer(repository.imanifestdict) |
|
373 | 373 | class manifestdict(object): |
|
374 | 374 | def __init__(self, data=''): |
|
375 | 375 | self._lm = _lazymanifest(data) |
|
376 | 376 | |
|
377 | 377 | def __getitem__(self, key): |
|
378 | 378 | return self._lm[key][0] |
|
379 | 379 | |
|
380 | 380 | def find(self, key): |
|
381 | 381 | return self._lm[key] |
|
382 | 382 | |
|
383 | 383 | def __len__(self): |
|
384 | 384 | return len(self._lm) |
|
385 | 385 | |
|
386 | 386 | def __nonzero__(self): |
|
387 | 387 | # nonzero is covered by the __len__ function, but implementing it here |
|
388 | 388 | # makes it easier for extensions to override. |
|
389 | 389 | return len(self._lm) != 0 |
|
390 | 390 | |
|
391 | 391 | __bool__ = __nonzero__ |
|
392 | 392 | |
|
393 | 393 | def __setitem__(self, key, node): |
|
394 | 394 | self._lm[key] = node, self.flags(key, '') |
|
395 | 395 | |
|
396 | 396 | def __contains__(self, key): |
|
397 | 397 | if key is None: |
|
398 | 398 | return False |
|
399 | 399 | return key in self._lm |
|
400 | 400 | |
|
401 | 401 | def __delitem__(self, key): |
|
402 | 402 | del self._lm[key] |
|
403 | 403 | |
|
404 | 404 | def __iter__(self): |
|
405 | 405 | return self._lm.__iter__() |
|
406 | 406 | |
|
407 | 407 | def iterkeys(self): |
|
408 | 408 | return self._lm.iterkeys() |
|
409 | 409 | |
|
410 | 410 | def keys(self): |
|
411 | 411 | return list(self.iterkeys()) |
|
412 | 412 | |
|
413 | 413 | def filesnotin(self, m2, match=None): |
|
414 | 414 | '''Set of files in this manifest that are not in the other''' |
|
415 | 415 | if match: |
|
416 | 416 | m1 = self.matches(match) |
|
417 | 417 | m2 = m2.matches(match) |
|
418 | 418 | return m1.filesnotin(m2) |
|
419 | 419 | diff = self.diff(m2) |
|
420 | 420 | files = set(filepath |
|
421 | 421 | for filepath, hashflags in diff.iteritems() |
|
422 | 422 | if hashflags[1][0] is None) |
|
423 | 423 | return files |
|
424 | 424 | |
|
425 | 425 | @propertycache |
|
426 | 426 | def _dirs(self): |
|
427 | 427 | return util.dirs(self) |
|
428 | 428 | |
|
429 | 429 | def dirs(self): |
|
430 | 430 | return self._dirs |
|
431 | 431 | |
|
432 | 432 | def hasdir(self, dir): |
|
433 | 433 | return dir in self._dirs |
|
434 | 434 | |
|
435 | 435 | def _filesfastpath(self, match): |
|
436 | 436 | '''Checks whether we can correctly and quickly iterate over matcher |
|
437 | 437 | files instead of over manifest files.''' |
|
438 | 438 | files = match.files() |
|
439 | 439 | return (len(files) < 100 and (match.isexact() or |
|
440 | 440 | (match.prefix() and all(fn in self for fn in files)))) |
|
441 | 441 | |
|
442 | 442 | def walk(self, match): |
|
443 | 443 | '''Generates matching file names. |
|
444 | 444 | |
|
445 | 445 | Equivalent to manifest.matches(match).iterkeys(), but without creating |
|
446 | 446 | an entirely new manifest. |
|
447 | 447 | |
|
448 | 448 | It also reports nonexistent files by marking them bad with match.bad(). |
|
449 | 449 | ''' |
|
450 | 450 | if match.always(): |
|
451 | 451 | for f in iter(self): |
|
452 | 452 | yield f |
|
453 | 453 | return |
|
454 | 454 | |
|
455 | 455 | fset = set(match.files()) |
|
456 | 456 | |
|
457 | 457 | # avoid the entire walk if we're only looking for specific files |
|
458 | 458 | if self._filesfastpath(match): |
|
459 | 459 | for fn in sorted(fset): |
|
460 | 460 | yield fn |
|
461 | 461 | return |
|
462 | 462 | |
|
463 | 463 | for fn in self: |
|
464 | 464 | if fn in fset: |
|
465 | 465 | # specified pattern is the exact name |
|
466 | 466 | fset.remove(fn) |
|
467 | 467 | if match(fn): |
|
468 | 468 | yield fn |
|
469 | 469 | |
|
470 | 470 | # for dirstate.walk, files=['.'] means "walk the whole tree". |
|
471 | 471 | # follow that here, too |
|
472 | 472 | fset.discard('.') |
|
473 | 473 | |
|
474 | 474 | for fn in sorted(fset): |
|
475 | 475 | if not self.hasdir(fn): |
|
476 | 476 | match.bad(fn, None) |
|
477 | 477 | |
|
478 | 478 | def matches(self, match): |
|
479 | 479 | '''generate a new manifest filtered by the match argument''' |
|
480 | 480 | if match.always(): |
|
481 | 481 | return self.copy() |
|
482 | 482 | |
|
483 | 483 | if self._filesfastpath(match): |
|
484 | 484 | m = manifestdict() |
|
485 | 485 | lm = self._lm |
|
486 | 486 | for fn in match.files(): |
|
487 | 487 | if fn in lm: |
|
488 | 488 | m._lm[fn] = lm[fn] |
|
489 | 489 | return m |
|
490 | 490 | |
|
491 | 491 | m = manifestdict() |
|
492 | 492 | m._lm = self._lm.filtercopy(match) |
|
493 | 493 | return m |
|
494 | 494 | |
|
495 | 495 | def diff(self, m2, match=None, clean=False): |
|
496 | 496 | '''Finds changes between the current manifest and m2. |
|
497 | 497 | |
|
498 | 498 | Args: |
|
499 | 499 | m2: the manifest to which this manifest should be compared. |
|
500 | 500 | clean: if true, include files unchanged between these manifests |
|
501 | 501 | with a None value in the returned dictionary. |
|
502 | 502 | |
|
503 | 503 | The result is returned as a dict with filename as key and |
|
504 | 504 | values of the form ((n1,fl1),(n2,fl2)), where n1/n2 is the |
|
505 | 505 | nodeid in the current/other manifest and fl1/fl2 is the flag |
|
506 | 506 | in the current/other manifest. Where the file does not exist, |
|
507 | 507 | the nodeid will be None and the flags will be the empty |
|
508 | 508 | string. |
|
509 | 509 | ''' |
|
510 | 510 | if match: |
|
511 | 511 | m1 = self.matches(match) |
|
512 | 512 | m2 = m2.matches(match) |
|
513 | 513 | return m1.diff(m2, clean=clean) |
|
514 | 514 | return self._lm.diff(m2._lm, clean) |
|
515 | 515 | |
|
516 | 516 | def setflag(self, key, flag): |
|
517 | 517 | self._lm[key] = self[key], flag |
|
518 | 518 | |
|
519 | 519 | def get(self, key, default=None): |
|
520 | 520 | try: |
|
521 | 521 | return self._lm[key][0] |
|
522 | 522 | except KeyError: |
|
523 | 523 | return default |
|
524 | 524 | |
|
525 | 525 | def flags(self, key, default=''): |
|
526 | 526 | try: |
|
527 | 527 | return self._lm[key][1] |
|
528 | 528 | except KeyError: |
|
529 | 529 | return default |
|
530 | 530 | |
|
531 | 531 | def copy(self): |
|
532 | 532 | c = manifestdict() |
|
533 | 533 | c._lm = self._lm.copy() |
|
534 | 534 | return c |
|
535 | 535 | |
|
536 | 536 | def items(self): |
|
537 | 537 | return (x[:2] for x in self._lm.iterentries()) |
|
538 | 538 | |
|
539 | 539 | def iteritems(self): |
|
540 | 540 | return (x[:2] for x in self._lm.iterentries()) |
|
541 | 541 | |
|
542 | 542 | def iterentries(self): |
|
543 | 543 | return self._lm.iterentries() |
|
544 | 544 | |
|
545 | 545 | def text(self): |
|
546 | 546 | # most likely uses native version |
|
547 | 547 | return self._lm.text() |
|
548 | 548 | |
|
549 | 549 | def fastdelta(self, base, changes): |
|
550 | 550 | """Given a base manifest text as a bytearray and a list of changes |
|
551 | 551 | relative to that text, compute a delta that can be used by revlog. |
|
552 | 552 | """ |
|
553 | 553 | delta = [] |
|
554 | 554 | dstart = None |
|
555 | 555 | dend = None |
|
556 | 556 | dline = [""] |
|
557 | 557 | start = 0 |
|
558 | 558 | # zero copy representation of base as a buffer |
|
559 | 559 | addbuf = util.buffer(base) |
|
560 | 560 | |
|
561 | 561 | changes = list(changes) |
|
562 | 562 | if len(changes) < 1000: |
|
563 | 563 | # start with a readonly loop that finds the offset of |
|
564 | 564 | # each line and creates the deltas |
|
565 | 565 | for f, todelete in changes: |
|
566 | 566 | # bs will either be the index of the item or the insert point |
|
567 | 567 | start, end = _msearch(addbuf, f, start) |
|
568 | 568 | if not todelete: |
|
569 | 569 | h, fl = self._lm[f] |
|
570 | 570 | l = "%s\0%s%s\n" % (f, hex(h), fl) |
|
571 | 571 | else: |
|
572 | 572 | if start == end: |
|
573 | 573 | # item we want to delete was not found, error out |
|
574 | 574 | raise AssertionError( |
|
575 | 575 | _("failed to remove %s from manifest") % f) |
|
576 | 576 | l = "" |
|
577 | 577 | if dstart is not None and dstart <= start and dend >= start: |
|
578 | 578 | if dend < end: |
|
579 | 579 | dend = end |
|
580 | 580 | if l: |
|
581 | 581 | dline.append(l) |
|
582 | 582 | else: |
|
583 | 583 | if dstart is not None: |
|
584 | 584 | delta.append([dstart, dend, "".join(dline)]) |
|
585 | 585 | dstart = start |
|
586 | 586 | dend = end |
|
587 | 587 | dline = [l] |
|
588 | 588 | |
|
589 | 589 | if dstart is not None: |
|
590 | 590 | delta.append([dstart, dend, "".join(dline)]) |
|
591 | 591 | # apply the delta to the base, and get a delta for addrevision |
|
592 | 592 | deltatext, arraytext = _addlistdelta(base, delta) |
|
593 | 593 | else: |
|
594 | 594 | # For large changes, it's much cheaper to just build the text and |
|
595 | 595 | # diff it. |
|
596 | 596 | arraytext = bytearray(self.text()) |
|
597 | 597 | deltatext = mdiff.textdiff( |
|
598 | 598 | util.buffer(base), util.buffer(arraytext)) |
|
599 | 599 | |
|
600 | 600 | return arraytext, deltatext |
|
601 | 601 | |
|
602 | 602 | def _msearch(m, s, lo=0, hi=None): |
|
603 | 603 | '''return a tuple (start, end) that says where to find s within m. |
|
604 | 604 | |
|
605 | 605 | If the string is found m[start:end] are the line containing |
|
606 | 606 | that string. If start == end the string was not found and |
|
607 | 607 | they indicate the proper sorted insertion point. |
|
608 | 608 | |
|
609 | 609 | m should be a buffer, a memoryview or a byte string. |
|
610 | 610 | s is a byte string''' |
|
611 | 611 | def advance(i, c): |
|
612 | 612 | while i < lenm and m[i:i + 1] != c: |
|
613 | 613 | i += 1 |
|
614 | 614 | return i |
|
615 | 615 | if not s: |
|
616 | 616 | return (lo, lo) |
|
617 | 617 | lenm = len(m) |
|
618 | 618 | if not hi: |
|
619 | 619 | hi = lenm |
|
620 | 620 | while lo < hi: |
|
621 | 621 | mid = (lo + hi) // 2 |
|
622 | 622 | start = mid |
|
623 | 623 | while start > 0 and m[start - 1:start] != '\n': |
|
624 | 624 | start -= 1 |
|
625 | 625 | end = advance(start, '\0') |
|
626 | 626 | if bytes(m[start:end]) < s: |
|
627 | 627 | # we know that after the null there are 40 bytes of sha1 |
|
628 | 628 | # this translates to the bisect lo = mid + 1 |
|
629 | 629 | lo = advance(end + 40, '\n') + 1 |
|
630 | 630 | else: |
|
631 | 631 | # this translates to the bisect hi = mid |
|
632 | 632 | hi = start |
|
633 | 633 | end = advance(lo, '\0') |
|
634 | 634 | found = m[lo:end] |
|
635 | 635 | if s == found: |
|
636 | 636 | # we know that after the null there are 40 bytes of sha1 |
|
637 | 637 | end = advance(end + 40, '\n') |
|
638 | 638 | return (lo, end + 1) |
|
639 | 639 | else: |
|
640 | 640 | return (lo, lo) |
|
641 | 641 | |
|
642 | 642 | def _checkforbidden(l): |
|
643 | 643 | """Check filenames for illegal characters.""" |
|
644 | 644 | for f in l: |
|
645 | 645 | if '\n' in f or '\r' in f: |
|
646 | 646 | raise error.StorageError( |
|
647 | 647 | _("'\\n' and '\\r' disallowed in filenames: %r") |
|
648 | 648 | % pycompat.bytestr(f)) |
|
649 | 649 | |
|
650 | 650 | |
|
651 | 651 | # apply the changes collected during the bisect loop to our addlist |
|
652 | 652 | # return a delta suitable for addrevision |
|
653 | 653 | def _addlistdelta(addlist, x): |
|
654 | 654 | # for large addlist arrays, building a new array is cheaper |
|
655 | 655 | # than repeatedly modifying the existing one |
|
656 | 656 | currentposition = 0 |
|
657 | 657 | newaddlist = bytearray() |
|
658 | 658 | |
|
659 | 659 | for start, end, content in x: |
|
660 | 660 | newaddlist += addlist[currentposition:start] |
|
661 | 661 | if content: |
|
662 | 662 | newaddlist += bytearray(content) |
|
663 | 663 | |
|
664 | 664 | currentposition = end |
|
665 | 665 | |
|
666 | 666 | newaddlist += addlist[currentposition:] |
|
667 | 667 | |
|
668 | 668 | deltatext = "".join(struct.pack(">lll", start, end, len(content)) |
|
669 | 669 | + content for start, end, content in x) |
|
670 | 670 | return deltatext, newaddlist |
|
671 | 671 | |
|
672 | 672 | def _splittopdir(f): |
|
673 | 673 | if '/' in f: |
|
674 | 674 | dir, subpath = f.split('/', 1) |
|
675 | 675 | return dir + '/', subpath |
|
676 | 676 | else: |
|
677 | 677 | return '', f |
|
678 | 678 | |
|
679 | 679 | _noop = lambda s: None |
|
680 | 680 | |
|
681 | 681 | class treemanifest(object): |
|
682 | 682 | def __init__(self, dir='', text=''): |
|
683 | 683 | self._dir = dir |
|
684 | 684 | self._node = nullid |
|
685 | 685 | self._loadfunc = _noop |
|
686 | 686 | self._copyfunc = _noop |
|
687 | 687 | self._dirty = False |
|
688 | 688 | self._dirs = {} |
|
689 | 689 | self._lazydirs = {} |
|
690 | 690 | # Using _lazymanifest here is a little slower than plain old dicts |
|
691 | 691 | self._files = {} |
|
692 | 692 | self._flags = {} |
|
693 | 693 | if text: |
|
694 | 694 | def readsubtree(subdir, subm): |
|
695 | 695 | raise AssertionError('treemanifest constructor only accepts ' |
|
696 | 696 | 'flat manifests') |
|
697 | 697 | self.parse(text, readsubtree) |
|
698 | 698 | self._dirty = True # Mark flat manifest dirty after parsing |
|
699 | 699 | |
|
700 | 700 | def _subpath(self, path): |
|
701 | 701 | return self._dir + path |
|
702 | 702 | |
|
703 | 703 | def _loadalllazy(self): |
|
704 | 704 | selfdirs = self._dirs |
|
705 | 705 | for d, (path, node, readsubtree, docopy) in self._lazydirs.iteritems(): |
|
706 | 706 | if docopy: |
|
707 | 707 | selfdirs[d] = readsubtree(path, node).copy() |
|
708 | 708 | else: |
|
709 | 709 | selfdirs[d] = readsubtree(path, node) |
|
710 | 710 | self._lazydirs = {} |
|
711 | 711 | |
|
712 | 712 | def _loadlazy(self, d): |
|
713 | 713 | v = self._lazydirs.get(d) |
|
714 | 714 | if v: |
|
715 | 715 | path, node, readsubtree, docopy = v |
|
716 | 716 | if docopy: |
|
717 | 717 | self._dirs[d] = readsubtree(path, node).copy() |
|
718 | 718 | else: |
|
719 | 719 | self._dirs[d] = readsubtree(path, node) |
|
720 | 720 | del self._lazydirs[d] |
|
721 | 721 | |
|
722 | 722 | def _loadchildrensetlazy(self, visit): |
|
723 | 723 | if not visit: |
|
724 | 724 | return None |
|
725 | 725 | if visit == 'all' or visit == 'this': |
|
726 | 726 | self._loadalllazy() |
|
727 | 727 | return None |
|
728 | 728 | |
|
729 | 729 | loadlazy = self._loadlazy |
|
730 | 730 | for k in visit: |
|
731 | 731 | loadlazy(k + '/') |
|
732 | 732 | return visit |
|
733 | 733 | |
|
734 | 734 | def _loaddifflazy(self, t1, t2): |
|
735 | 735 | """load items in t1 and t2 if they're needed for diffing. |
|
736 | 736 | |
|
737 | 737 | The criteria currently is: |
|
738 | 738 | - if it's not present in _lazydirs in either t1 or t2, load it in the |
|
739 | 739 | other (it may already be loaded or it may not exist, doesn't matter) |
|
740 | 740 | - if it's present in _lazydirs in both, compare the nodeid; if it |
|
741 | 741 | differs, load it in both |
|
742 | 742 | """ |
|
743 | 743 | toloadlazy = [] |
|
744 | 744 | for d, v1 in t1._lazydirs.iteritems(): |
|
745 | 745 | v2 = t2._lazydirs.get(d) |
|
746 | 746 | if not v2 or v2[1] != v1[1]: |
|
747 | 747 | toloadlazy.append(d) |
|
748 | 748 | for d, v1 in t2._lazydirs.iteritems(): |
|
749 | 749 | if d not in t1._lazydirs: |
|
750 | 750 | toloadlazy.append(d) |
|
751 | 751 | |
|
752 | 752 | for d in toloadlazy: |
|
753 | 753 | t1._loadlazy(d) |
|
754 | 754 | t2._loadlazy(d) |
|
755 | 755 | |
|
756 | 756 | def __len__(self): |
|
757 | 757 | self._load() |
|
758 | 758 | size = len(self._files) |
|
759 | 759 | self._loadalllazy() |
|
760 | 760 | for m in self._dirs.values(): |
|
761 | 761 | size += m.__len__() |
|
762 | 762 | return size |
|
763 | 763 | |
|
764 | 764 | def __nonzero__(self): |
|
765 | 765 | # Faster than "__len() != 0" since it avoids loading sub-manifests |
|
766 | 766 | return not self._isempty() |
|
767 | 767 | |
|
768 | 768 | __bool__ = __nonzero__ |
|
769 | 769 | |
|
770 | 770 | def _isempty(self): |
|
771 | 771 | self._load() # for consistency; already loaded by all callers |
|
772 | 772 | # See if we can skip loading everything. |
|
773 | 773 | if self._files or (self._dirs and |
|
774 | 774 | any(not m._isempty() for m in self._dirs.values())): |
|
775 | 775 | return False |
|
776 | 776 | self._loadalllazy() |
|
777 | 777 | return (not self._dirs or |
|
778 | 778 | all(m._isempty() for m in self._dirs.values())) |
|
779 | 779 | |
|
780 | 780 | def __repr__(self): |
|
781 | 781 | return ('<treemanifest dir=%s, node=%s, loaded=%s, dirty=%s at 0x%x>' % |
|
782 | 782 | (self._dir, hex(self._node), |
|
783 | 783 | bool(self._loadfunc is _noop), |
|
784 | 784 | self._dirty, id(self))) |
|
785 | 785 | |
|
786 | 786 | def dir(self): |
|
787 | 787 | '''The directory that this tree manifest represents, including a |
|
788 | 788 | trailing '/'. Empty string for the repo root directory.''' |
|
789 | 789 | return self._dir |
|
790 | 790 | |
|
791 | 791 | def node(self): |
|
792 | 792 | '''This node of this instance. nullid for unsaved instances. Should |
|
793 | 793 | be updated when the instance is read or written from a revlog. |
|
794 | 794 | ''' |
|
795 | 795 | assert not self._dirty |
|
796 | 796 | return self._node |
|
797 | 797 | |
|
798 | 798 | def setnode(self, node): |
|
799 | 799 | self._node = node |
|
800 | 800 | self._dirty = False |
|
801 | 801 | |
|
802 | 802 | def iterentries(self): |
|
803 | 803 | self._load() |
|
804 | 804 | self._loadalllazy() |
|
805 | 805 | for p, n in sorted(itertools.chain(self._dirs.items(), |
|
806 | 806 | self._files.items())): |
|
807 | 807 | if p in self._files: |
|
808 | 808 | yield self._subpath(p), n, self._flags.get(p, '') |
|
809 | 809 | else: |
|
810 | 810 | for x in n.iterentries(): |
|
811 | 811 | yield x |
|
812 | 812 | |
|
813 | 813 | def items(self): |
|
814 | 814 | self._load() |
|
815 | 815 | self._loadalllazy() |
|
816 | 816 | for p, n in sorted(itertools.chain(self._dirs.items(), |
|
817 | 817 | self._files.items())): |
|
818 | 818 | if p in self._files: |
|
819 | 819 | yield self._subpath(p), n |
|
820 | 820 | else: |
|
821 | 821 | for f, sn in n.iteritems(): |
|
822 | 822 | yield f, sn |
|
823 | 823 | |
|
824 | 824 | iteritems = items |
|
825 | 825 | |
|
826 | 826 | def iterkeys(self): |
|
827 | 827 | self._load() |
|
828 | 828 | self._loadalllazy() |
|
829 | 829 | for p in sorted(itertools.chain(self._dirs, self._files)): |
|
830 | 830 | if p in self._files: |
|
831 | 831 | yield self._subpath(p) |
|
832 | 832 | else: |
|
833 | 833 | for f in self._dirs[p]: |
|
834 | 834 | yield f |
|
835 | 835 | |
|
836 | 836 | def keys(self): |
|
837 | 837 | return list(self.iterkeys()) |
|
838 | 838 | |
|
839 | 839 | def __iter__(self): |
|
840 | 840 | return self.iterkeys() |
|
841 | 841 | |
|
842 | 842 | def __contains__(self, f): |
|
843 | 843 | if f is None: |
|
844 | 844 | return False |
|
845 | 845 | self._load() |
|
846 | 846 | dir, subpath = _splittopdir(f) |
|
847 | 847 | if dir: |
|
848 | 848 | self._loadlazy(dir) |
|
849 | 849 | |
|
850 | 850 | if dir not in self._dirs: |
|
851 | 851 | return False |
|
852 | 852 | |
|
853 | 853 | return self._dirs[dir].__contains__(subpath) |
|
854 | 854 | else: |
|
855 | 855 | return f in self._files |
|
856 | 856 | |
|
857 | 857 | def get(self, f, default=None): |
|
858 | 858 | self._load() |
|
859 | 859 | dir, subpath = _splittopdir(f) |
|
860 | 860 | if dir: |
|
861 | 861 | self._loadlazy(dir) |
|
862 | 862 | |
|
863 | 863 | if dir not in self._dirs: |
|
864 | 864 | return default |
|
865 | 865 | return self._dirs[dir].get(subpath, default) |
|
866 | 866 | else: |
|
867 | 867 | return self._files.get(f, default) |
|
868 | 868 | |
|
869 | 869 | def __getitem__(self, f): |
|
870 | 870 | self._load() |
|
871 | 871 | dir, subpath = _splittopdir(f) |
|
872 | 872 | if dir: |
|
873 | 873 | self._loadlazy(dir) |
|
874 | 874 | |
|
875 | 875 | return self._dirs[dir].__getitem__(subpath) |
|
876 | 876 | else: |
|
877 | 877 | return self._files[f] |
|
878 | 878 | |
|
879 | 879 | def flags(self, f): |
|
880 | 880 | self._load() |
|
881 | 881 | dir, subpath = _splittopdir(f) |
|
882 | 882 | if dir: |
|
883 | 883 | self._loadlazy(dir) |
|
884 | 884 | |
|
885 | 885 | if dir not in self._dirs: |
|
886 | 886 | return '' |
|
887 | 887 | return self._dirs[dir].flags(subpath) |
|
888 | 888 | else: |
|
889 | 889 | if f in self._lazydirs or f in self._dirs: |
|
890 | 890 | return '' |
|
891 | 891 | return self._flags.get(f, '') |
|
892 | 892 | |
|
893 | 893 | def find(self, f): |
|
894 | 894 | self._load() |
|
895 | 895 | dir, subpath = _splittopdir(f) |
|
896 | 896 | if dir: |
|
897 | 897 | self._loadlazy(dir) |
|
898 | 898 | |
|
899 | 899 | return self._dirs[dir].find(subpath) |
|
900 | 900 | else: |
|
901 | 901 | return self._files[f], self._flags.get(f, '') |
|
902 | 902 | |
|
903 | 903 | def __delitem__(self, f): |
|
904 | 904 | self._load() |
|
905 | 905 | dir, subpath = _splittopdir(f) |
|
906 | 906 | if dir: |
|
907 | 907 | self._loadlazy(dir) |
|
908 | 908 | |
|
909 | 909 | self._dirs[dir].__delitem__(subpath) |
|
910 | 910 | # If the directory is now empty, remove it |
|
911 | 911 | if self._dirs[dir]._isempty(): |
|
912 | 912 | del self._dirs[dir] |
|
913 | 913 | else: |
|
914 | 914 | del self._files[f] |
|
915 | 915 | if f in self._flags: |
|
916 | 916 | del self._flags[f] |
|
917 | 917 | self._dirty = True |
|
918 | 918 | |
|
919 | 919 | def __setitem__(self, f, n): |
|
920 | 920 | assert n is not None |
|
921 | 921 | self._load() |
|
922 | 922 | dir, subpath = _splittopdir(f) |
|
923 | 923 | if dir: |
|
924 | 924 | self._loadlazy(dir) |
|
925 | 925 | if dir not in self._dirs: |
|
926 | 926 | self._dirs[dir] = treemanifest(self._subpath(dir)) |
|
927 | 927 | self._dirs[dir].__setitem__(subpath, n) |
|
928 | 928 | else: |
|
929 | 929 | self._files[f] = n[:21] # to match manifestdict's behavior |
|
930 | 930 | self._dirty = True |
|
931 | 931 | |
|
932 | 932 | def _load(self): |
|
933 | 933 | if self._loadfunc is not _noop: |
|
934 | 934 | lf, self._loadfunc = self._loadfunc, _noop |
|
935 | 935 | lf(self) |
|
936 | 936 | elif self._copyfunc is not _noop: |
|
937 | 937 | cf, self._copyfunc = self._copyfunc, _noop |
|
938 | 938 | cf(self) |
|
939 | 939 | |
|
940 | 940 | def setflag(self, f, flags): |
|
941 | 941 | """Set the flags (symlink, executable) for path f.""" |
|
942 | 942 | self._load() |
|
943 | 943 | dir, subpath = _splittopdir(f) |
|
944 | 944 | if dir: |
|
945 | 945 | self._loadlazy(dir) |
|
946 | 946 | if dir not in self._dirs: |
|
947 | 947 | self._dirs[dir] = treemanifest(self._subpath(dir)) |
|
948 | 948 | self._dirs[dir].setflag(subpath, flags) |
|
949 | 949 | else: |
|
950 | 950 | self._flags[f] = flags |
|
951 | 951 | self._dirty = True |
|
952 | 952 | |
|
953 | 953 | def copy(self): |
|
954 | 954 | copy = treemanifest(self._dir) |
|
955 | 955 | copy._node = self._node |
|
956 | 956 | copy._dirty = self._dirty |
|
957 | 957 | if self._copyfunc is _noop: |
|
958 | 958 | def _copyfunc(s): |
|
959 | 959 | self._load() |
|
960 | 960 | s._lazydirs = {d: (p, n, r, True) for |
|
961 | 961 | d, (p, n, r, c) in self._lazydirs.iteritems()} |
|
962 | 962 | sdirs = s._dirs |
|
963 | 963 | for d, v in self._dirs.iteritems(): |
|
964 | 964 | sdirs[d] = v.copy() |
|
965 | 965 | s._files = dict.copy(self._files) |
|
966 | 966 | s._flags = dict.copy(self._flags) |
|
967 | 967 | if self._loadfunc is _noop: |
|
968 | 968 | _copyfunc(copy) |
|
969 | 969 | else: |
|
970 | 970 | copy._copyfunc = _copyfunc |
|
971 | 971 | else: |
|
972 | 972 | copy._copyfunc = self._copyfunc |
|
973 | 973 | return copy |
|
974 | 974 | |
|
975 | 975 | def filesnotin(self, m2, match=None): |
|
976 | 976 | '''Set of files in this manifest that are not in the other''' |
|
977 | 977 | if match and not match.always(): |
|
978 | 978 | m1 = self.matches(match) |
|
979 | 979 | m2 = m2.matches(match) |
|
980 | 980 | return m1.filesnotin(m2) |
|
981 | 981 | |
|
982 | 982 | files = set() |
|
983 | 983 | def _filesnotin(t1, t2): |
|
984 | 984 | if t1._node == t2._node and not t1._dirty and not t2._dirty: |
|
985 | 985 | return |
|
986 | 986 | t1._load() |
|
987 | 987 | t2._load() |
|
988 | 988 | self._loaddifflazy(t1, t2) |
|
989 | 989 | for d, m1 in t1._dirs.iteritems(): |
|
990 | 990 | if d in t2._dirs: |
|
991 | 991 | m2 = t2._dirs[d] |
|
992 | 992 | _filesnotin(m1, m2) |
|
993 | 993 | else: |
|
994 | 994 | files.update(m1.iterkeys()) |
|
995 | 995 | |
|
996 | 996 | for fn in t1._files: |
|
997 | 997 | if fn not in t2._files: |
|
998 | 998 | files.add(t1._subpath(fn)) |
|
999 | 999 | |
|
1000 | 1000 | _filesnotin(self, m2) |
|
1001 | 1001 | return files |
|
1002 | 1002 | |
|
1003 | 1003 | @propertycache |
|
1004 | 1004 | def _alldirs(self): |
|
1005 | 1005 | return util.dirs(self) |
|
1006 | 1006 | |
|
1007 | 1007 | def dirs(self): |
|
1008 | 1008 | return self._alldirs |
|
1009 | 1009 | |
|
1010 | 1010 | def hasdir(self, dir): |
|
1011 | 1011 | self._load() |
|
1012 | 1012 | topdir, subdir = _splittopdir(dir) |
|
1013 | 1013 | if topdir: |
|
1014 | 1014 | self._loadlazy(topdir) |
|
1015 | 1015 | if topdir in self._dirs: |
|
1016 | 1016 | return self._dirs[topdir].hasdir(subdir) |
|
1017 | 1017 | return False |
|
1018 | 1018 | dirslash = dir + '/' |
|
1019 | 1019 | return dirslash in self._dirs or dirslash in self._lazydirs |
|
1020 | 1020 | |
|
1021 | 1021 | def walk(self, match): |
|
1022 | 1022 | '''Generates matching file names. |
|
1023 | 1023 | |
|
1024 | 1024 | Equivalent to manifest.matches(match).iterkeys(), but without creating |
|
1025 | 1025 | an entirely new manifest. |
|
1026 | 1026 | |
|
1027 | 1027 | It also reports nonexistent files by marking them bad with match.bad(). |
|
1028 | 1028 | ''' |
|
1029 | 1029 | if match.always(): |
|
1030 | 1030 | for f in iter(self): |
|
1031 | 1031 | yield f |
|
1032 | 1032 | return |
|
1033 | 1033 | |
|
1034 | 1034 | fset = set(match.files()) |
|
1035 | 1035 | |
|
1036 | 1036 | for fn in self._walk(match): |
|
1037 | 1037 | if fn in fset: |
|
1038 | 1038 | # specified pattern is the exact name |
|
1039 | 1039 | fset.remove(fn) |
|
1040 | 1040 | yield fn |
|
1041 | 1041 | |
|
1042 | 1042 | # for dirstate.walk, files=['.'] means "walk the whole tree". |
|
1043 | 1043 | # follow that here, too |
|
1044 | 1044 | fset.discard('.') |
|
1045 | 1045 | |
|
1046 | 1046 | for fn in sorted(fset): |
|
1047 | 1047 | if not self.hasdir(fn): |
|
1048 | 1048 | match.bad(fn, None) |
|
1049 | 1049 | |
|
1050 | 1050 | def _walk(self, match): |
|
1051 | 1051 | '''Recursively generates matching file names for walk().''' |
|
1052 | 1052 | visit = match.visitchildrenset(self._dir[:-1] or '.') |
|
1053 | 1053 | if not visit: |
|
1054 | 1054 | return |
|
1055 | 1055 | |
|
1056 | 1056 | # yield this dir's files and walk its submanifests |
|
1057 | 1057 | self._load() |
|
1058 | 1058 | visit = self._loadchildrensetlazy(visit) |
|
1059 | 1059 | for p in sorted(list(self._dirs) + list(self._files)): |
|
1060 | 1060 | if p in self._files: |
|
1061 | 1061 | fullp = self._subpath(p) |
|
1062 | 1062 | if match(fullp): |
|
1063 | 1063 | yield fullp |
|
1064 | 1064 | else: |
|
1065 | 1065 | if not visit or p[:-1] in visit: |
|
1066 | 1066 | for f in self._dirs[p]._walk(match): |
|
1067 | 1067 | yield f |
|
1068 | 1068 | |
|
1069 | 1069 | def matches(self, match): |
|
1070 | 1070 | '''generate a new manifest filtered by the match argument''' |
|
1071 | 1071 | if match.always(): |
|
1072 | 1072 | return self.copy() |
|
1073 | 1073 | |
|
1074 | 1074 | return self._matches(match) |
|
1075 | 1075 | |
|
1076 | 1076 | def _matches(self, match): |
|
1077 | 1077 | '''recursively generate a new manifest filtered by the match argument. |
|
1078 | 1078 | ''' |
|
1079 | 1079 | |
|
1080 | 1080 | visit = match.visitchildrenset(self._dir[:-1] or '.') |
|
1081 | 1081 | if visit == 'all': |
|
1082 | 1082 | return self.copy() |
|
1083 | 1083 | ret = treemanifest(self._dir) |
|
1084 | 1084 | if not visit: |
|
1085 | 1085 | return ret |
|
1086 | 1086 | |
|
1087 | 1087 | self._load() |
|
1088 | 1088 | for fn in self._files: |
|
1089 | 1089 | # While visitchildrenset *usually* lists only subdirs, this is |
|
1090 | 1090 | # actually up to the matcher and may have some files in the set(). |
|
1091 | 1091 | # If visit == 'this', we should obviously look at the files in this |
|
1092 | 1092 | # directory; if visit is a set, and fn is in it, we should inspect |
|
1093 | 1093 | # fn (but no need to inspect things not in the set). |
|
1094 | 1094 | if visit != 'this' and fn not in visit: |
|
1095 | 1095 | continue |
|
1096 | 1096 | fullp = self._subpath(fn) |
|
1097 | 1097 | # visitchildrenset isn't perfect, we still need to call the regular |
|
1098 | 1098 | # matcher code to further filter results. |
|
1099 | 1099 | if not match(fullp): |
|
1100 | 1100 | continue |
|
1101 | 1101 | ret._files[fn] = self._files[fn] |
|
1102 | 1102 | if fn in self._flags: |
|
1103 | 1103 | ret._flags[fn] = self._flags[fn] |
|
1104 | 1104 | |
|
1105 | 1105 | visit = self._loadchildrensetlazy(visit) |
|
1106 | 1106 | for dir, subm in self._dirs.iteritems(): |
|
1107 | 1107 | if visit and dir[:-1] not in visit: |
|
1108 | 1108 | continue |
|
1109 | 1109 | m = subm._matches(match) |
|
1110 | 1110 | if not m._isempty(): |
|
1111 | 1111 | ret._dirs[dir] = m |
|
1112 | 1112 | |
|
1113 | 1113 | if not ret._isempty(): |
|
1114 | 1114 | ret._dirty = True |
|
1115 | 1115 | return ret |
|
1116 | 1116 | |
|
1117 | 1117 | def diff(self, m2, match=None, clean=False): |
|
1118 | 1118 | '''Finds changes between the current manifest and m2. |
|
1119 | 1119 | |
|
1120 | 1120 | Args: |
|
1121 | 1121 | m2: the manifest to which this manifest should be compared. |
|
1122 | 1122 | clean: if true, include files unchanged between these manifests |
|
1123 | 1123 | with a None value in the returned dictionary. |
|
1124 | 1124 | |
|
1125 | 1125 | The result is returned as a dict with filename as key and |
|
1126 | 1126 | values of the form ((n1,fl1),(n2,fl2)), where n1/n2 is the |
|
1127 | 1127 | nodeid in the current/other manifest and fl1/fl2 is the flag |
|
1128 | 1128 | in the current/other manifest. Where the file does not exist, |
|
1129 | 1129 | the nodeid will be None and the flags will be the empty |
|
1130 | 1130 | string. |
|
1131 | 1131 | ''' |
|
1132 | 1132 | if match and not match.always(): |
|
1133 | 1133 | m1 = self.matches(match) |
|
1134 | 1134 | m2 = m2.matches(match) |
|
1135 | 1135 | return m1.diff(m2, clean=clean) |
|
1136 | 1136 | result = {} |
|
1137 | 1137 | emptytree = treemanifest() |
|
1138 | def _diff(t1, t2): | |
|
1138 | ||
|
1139 | def _iterativediff(t1, t2, stack): | |
|
1140 | """compares two tree manifests and append new tree-manifests which | |
|
1141 | needs to be compared to stack""" | |
|
1139 | 1142 | if t1._node == t2._node and not t1._dirty and not t2._dirty: |
|
1140 | 1143 | return |
|
1141 | 1144 | t1._load() |
|
1142 | 1145 | t2._load() |
|
1143 | 1146 | self._loaddifflazy(t1, t2) |
|
1144 | 1147 | |
|
1145 | 1148 | for d, m1 in t1._dirs.iteritems(): |
|
1146 | 1149 | m2 = t2._dirs.get(d, emptytree) |
|
1147 |
|
|
|
1150 | stack.append((m1, m2)) | |
|
1148 | 1151 | |
|
1149 | 1152 | for d, m2 in t2._dirs.iteritems(): |
|
1150 | 1153 | if d not in t1._dirs: |
|
1151 |
|
|
|
1154 | stack.append((emptytree, m2)) | |
|
1152 | 1155 | |
|
1153 | 1156 | for fn, n1 in t1._files.iteritems(): |
|
1154 | 1157 | fl1 = t1._flags.get(fn, '') |
|
1155 | 1158 | n2 = t2._files.get(fn, None) |
|
1156 | 1159 | fl2 = t2._flags.get(fn, '') |
|
1157 | 1160 | if n1 != n2 or fl1 != fl2: |
|
1158 | 1161 | result[t1._subpath(fn)] = ((n1, fl1), (n2, fl2)) |
|
1159 | 1162 | elif clean: |
|
1160 | 1163 | result[t1._subpath(fn)] = None |
|
1161 | 1164 | |
|
1162 | 1165 | for fn, n2 in t2._files.iteritems(): |
|
1163 | 1166 | if fn not in t1._files: |
|
1164 | 1167 | fl2 = t2._flags.get(fn, '') |
|
1165 | 1168 | result[t2._subpath(fn)] = ((None, ''), (n2, fl2)) |
|
1166 | 1169 | |
|
1167 | _diff(self, m2) | |
|
1170 | stackls = [] | |
|
1171 | _iterativediff(self, m2, stackls) | |
|
1172 | while stackls: | |
|
1173 | t1, t2 = stackls.pop() | |
|
1174 | # stackls is populated in the function call | |
|
1175 | _iterativediff(t1, t2, stackls) | |
|
1168 | 1176 | return result |
|
1169 | 1177 | |
|
1170 | 1178 | def unmodifiedsince(self, m2): |
|
1171 | 1179 | return not self._dirty and not m2._dirty and self._node == m2._node |
|
1172 | 1180 | |
|
1173 | 1181 | def parse(self, text, readsubtree): |
|
1174 | 1182 | selflazy = self._lazydirs |
|
1175 | 1183 | subpath = self._subpath |
|
1176 | 1184 | for f, n, fl in _parse(text): |
|
1177 | 1185 | if fl == 't': |
|
1178 | 1186 | f = f + '/' |
|
1179 | 1187 | # False below means "doesn't need to be copied" and can use the |
|
1180 | 1188 | # cached value from readsubtree directly. |
|
1181 | 1189 | selflazy[f] = (subpath(f), n, readsubtree, False) |
|
1182 | 1190 | elif '/' in f: |
|
1183 | 1191 | # This is a flat manifest, so use __setitem__ and setflag rather |
|
1184 | 1192 | # than assigning directly to _files and _flags, so we can |
|
1185 | 1193 | # assign a path in a subdirectory, and to mark dirty (compared |
|
1186 | 1194 | # to nullid). |
|
1187 | 1195 | self[f] = n |
|
1188 | 1196 | if fl: |
|
1189 | 1197 | self.setflag(f, fl) |
|
1190 | 1198 | else: |
|
1191 | 1199 | # Assigning to _files and _flags avoids marking as dirty, |
|
1192 | 1200 | # and should be a little faster. |
|
1193 | 1201 | self._files[f] = n |
|
1194 | 1202 | if fl: |
|
1195 | 1203 | self._flags[f] = fl |
|
1196 | 1204 | |
|
1197 | 1205 | def text(self): |
|
1198 | 1206 | """Get the full data of this manifest as a bytestring.""" |
|
1199 | 1207 | self._load() |
|
1200 | 1208 | return _text(self.iterentries()) |
|
1201 | 1209 | |
|
1202 | 1210 | def dirtext(self): |
|
1203 | 1211 | """Get the full data of this directory as a bytestring. Make sure that |
|
1204 | 1212 | any submanifests have been written first, so their nodeids are correct. |
|
1205 | 1213 | """ |
|
1206 | 1214 | self._load() |
|
1207 | 1215 | flags = self.flags |
|
1208 | 1216 | lazydirs = [(d[:-1], v[1], 't') for d, v in self._lazydirs.iteritems()] |
|
1209 | 1217 | dirs = [(d[:-1], self._dirs[d]._node, 't') for d in self._dirs] |
|
1210 | 1218 | files = [(f, self._files[f], flags(f)) for f in self._files] |
|
1211 | 1219 | return _text(sorted(dirs + files + lazydirs)) |
|
1212 | 1220 | |
|
1213 | 1221 | def read(self, gettext, readsubtree): |
|
1214 | 1222 | def _load_for_read(s): |
|
1215 | 1223 | s.parse(gettext(), readsubtree) |
|
1216 | 1224 | s._dirty = False |
|
1217 | 1225 | self._loadfunc = _load_for_read |
|
1218 | 1226 | |
|
1219 | 1227 | def writesubtrees(self, m1, m2, writesubtree, match): |
|
1220 | 1228 | self._load() # for consistency; should never have any effect here |
|
1221 | 1229 | m1._load() |
|
1222 | 1230 | m2._load() |
|
1223 | 1231 | emptytree = treemanifest() |
|
1224 | 1232 | def getnode(m, d): |
|
1225 | 1233 | ld = m._lazydirs.get(d) |
|
1226 | 1234 | if ld: |
|
1227 | 1235 | return ld[1] |
|
1228 | 1236 | return m._dirs.get(d, emptytree)._node |
|
1229 | 1237 | |
|
1230 | 1238 | # let's skip investigating things that `match` says we do not need. |
|
1231 | 1239 | visit = match.visitchildrenset(self._dir[:-1] or '.') |
|
1232 | 1240 | visit = self._loadchildrensetlazy(visit) |
|
1233 | 1241 | if visit == 'this' or visit == 'all': |
|
1234 | 1242 | visit = None |
|
1235 | 1243 | for d, subm in self._dirs.iteritems(): |
|
1236 | 1244 | if visit and d[:-1] not in visit: |
|
1237 | 1245 | continue |
|
1238 | 1246 | subp1 = getnode(m1, d) |
|
1239 | 1247 | subp2 = getnode(m2, d) |
|
1240 | 1248 | if subp1 == nullid: |
|
1241 | 1249 | subp1, subp2 = subp2, subp1 |
|
1242 | 1250 | writesubtree(subm, subp1, subp2, match) |
|
1243 | 1251 | |
|
1244 | 1252 | def walksubtrees(self, matcher=None): |
|
1245 | 1253 | """Returns an iterator of the subtrees of this manifest, including this |
|
1246 | 1254 | manifest itself. |
|
1247 | 1255 | |
|
1248 | 1256 | If `matcher` is provided, it only returns subtrees that match. |
|
1249 | 1257 | """ |
|
1250 | 1258 | if matcher and not matcher.visitdir(self._dir[:-1] or '.'): |
|
1251 | 1259 | return |
|
1252 | 1260 | if not matcher or matcher(self._dir[:-1]): |
|
1253 | 1261 | yield self |
|
1254 | 1262 | |
|
1255 | 1263 | self._load() |
|
1256 | 1264 | # OPT: use visitchildrenset to avoid loading everything. |
|
1257 | 1265 | self._loadalllazy() |
|
1258 | 1266 | for d, subm in self._dirs.iteritems(): |
|
1259 | 1267 | for subtree in subm.walksubtrees(matcher=matcher): |
|
1260 | 1268 | yield subtree |
|
1261 | 1269 | |
|
1262 | 1270 | class manifestfulltextcache(util.lrucachedict): |
|
1263 | 1271 | """File-backed LRU cache for the manifest cache |
|
1264 | 1272 | |
|
1265 | 1273 | File consists of entries, up to EOF: |
|
1266 | 1274 | |
|
1267 | 1275 | - 20 bytes node, 4 bytes length, <length> manifest data |
|
1268 | 1276 | |
|
1269 | 1277 | These are written in reverse cache order (oldest to newest). |
|
1270 | 1278 | |
|
1271 | 1279 | """ |
|
1272 | 1280 | def __init__(self, max): |
|
1273 | 1281 | super(manifestfulltextcache, self).__init__(max) |
|
1274 | 1282 | self._dirty = False |
|
1275 | 1283 | self._read = False |
|
1276 | 1284 | self._opener = None |
|
1277 | 1285 | |
|
1278 | 1286 | def read(self): |
|
1279 | 1287 | if self._read or self._opener is None: |
|
1280 | 1288 | return |
|
1281 | 1289 | |
|
1282 | 1290 | try: |
|
1283 | 1291 | with self._opener('manifestfulltextcache') as fp: |
|
1284 | 1292 | set = super(manifestfulltextcache, self).__setitem__ |
|
1285 | 1293 | # ignore trailing data, this is a cache, corruption is skipped |
|
1286 | 1294 | while True: |
|
1287 | 1295 | node = fp.read(20) |
|
1288 | 1296 | if len(node) < 20: |
|
1289 | 1297 | break |
|
1290 | 1298 | try: |
|
1291 | 1299 | size = struct.unpack('>L', fp.read(4))[0] |
|
1292 | 1300 | except struct.error: |
|
1293 | 1301 | break |
|
1294 | 1302 | value = bytearray(fp.read(size)) |
|
1295 | 1303 | if len(value) != size: |
|
1296 | 1304 | break |
|
1297 | 1305 | set(node, value) |
|
1298 | 1306 | except IOError: |
|
1299 | 1307 | # the file is allowed to be missing |
|
1300 | 1308 | pass |
|
1301 | 1309 | |
|
1302 | 1310 | self._read = True |
|
1303 | 1311 | self._dirty = False |
|
1304 | 1312 | |
|
1305 | 1313 | def write(self): |
|
1306 | 1314 | if not self._dirty or self._opener is None: |
|
1307 | 1315 | return |
|
1308 | 1316 | # rotate backwards to the first used node |
|
1309 | 1317 | with self._opener( |
|
1310 | 1318 | 'manifestfulltextcache', 'w', atomictemp=True, checkambig=True |
|
1311 | 1319 | ) as fp: |
|
1312 | 1320 | node = self._head.prev |
|
1313 | 1321 | while True: |
|
1314 | 1322 | if node.key in self._cache: |
|
1315 | 1323 | fp.write(node.key) |
|
1316 | 1324 | fp.write(struct.pack('>L', len(node.value))) |
|
1317 | 1325 | fp.write(node.value) |
|
1318 | 1326 | if node is self._head: |
|
1319 | 1327 | break |
|
1320 | 1328 | node = node.prev |
|
1321 | 1329 | |
|
1322 | 1330 | def __len__(self): |
|
1323 | 1331 | if not self._read: |
|
1324 | 1332 | self.read() |
|
1325 | 1333 | return super(manifestfulltextcache, self).__len__() |
|
1326 | 1334 | |
|
1327 | 1335 | def __contains__(self, k): |
|
1328 | 1336 | if not self._read: |
|
1329 | 1337 | self.read() |
|
1330 | 1338 | return super(manifestfulltextcache, self).__contains__(k) |
|
1331 | 1339 | |
|
1332 | 1340 | def __iter__(self): |
|
1333 | 1341 | if not self._read: |
|
1334 | 1342 | self.read() |
|
1335 | 1343 | return super(manifestfulltextcache, self).__iter__() |
|
1336 | 1344 | |
|
1337 | 1345 | def __getitem__(self, k): |
|
1338 | 1346 | if not self._read: |
|
1339 | 1347 | self.read() |
|
1340 | 1348 | # the cache lru order can change on read |
|
1341 | 1349 | setdirty = self._cache.get(k) is not self._head |
|
1342 | 1350 | value = super(manifestfulltextcache, self).__getitem__(k) |
|
1343 | 1351 | if setdirty: |
|
1344 | 1352 | self._dirty = True |
|
1345 | 1353 | return value |
|
1346 | 1354 | |
|
1347 | 1355 | def __setitem__(self, k, v): |
|
1348 | 1356 | if not self._read: |
|
1349 | 1357 | self.read() |
|
1350 | 1358 | super(manifestfulltextcache, self).__setitem__(k, v) |
|
1351 | 1359 | self._dirty = True |
|
1352 | 1360 | |
|
1353 | 1361 | def __delitem__(self, k): |
|
1354 | 1362 | if not self._read: |
|
1355 | 1363 | self.read() |
|
1356 | 1364 | super(manifestfulltextcache, self).__delitem__(k) |
|
1357 | 1365 | self._dirty = True |
|
1358 | 1366 | |
|
1359 | 1367 | def get(self, k, default=None): |
|
1360 | 1368 | if not self._read: |
|
1361 | 1369 | self.read() |
|
1362 | 1370 | return super(manifestfulltextcache, self).get(k, default=default) |
|
1363 | 1371 | |
|
1364 | 1372 | def clear(self, clear_persisted_data=False): |
|
1365 | 1373 | super(manifestfulltextcache, self).clear() |
|
1366 | 1374 | if clear_persisted_data: |
|
1367 | 1375 | self._dirty = True |
|
1368 | 1376 | self.write() |
|
1369 | 1377 | self._read = False |
|
1370 | 1378 | |
|
1371 | 1379 | @interfaceutil.implementer(repository.imanifeststorage) |
|
1372 | 1380 | class manifestrevlog(object): |
|
1373 | 1381 | '''A revlog that stores manifest texts. This is responsible for caching the |
|
1374 | 1382 | full-text manifest contents. |
|
1375 | 1383 | ''' |
|
1376 | 1384 | def __init__(self, opener, tree='', dirlogcache=None, indexfile=None, |
|
1377 | 1385 | treemanifest=False): |
|
1378 | 1386 | """Constructs a new manifest revlog |
|
1379 | 1387 | |
|
1380 | 1388 | `indexfile` - used by extensions to have two manifests at once, like |
|
1381 | 1389 | when transitioning between flatmanifeset and treemanifests. |
|
1382 | 1390 | |
|
1383 | 1391 | `treemanifest` - used to indicate this is a tree manifest revlog. Opener |
|
1384 | 1392 | options can also be used to make this a tree manifest revlog. The opener |
|
1385 | 1393 | option takes precedence, so if it is set to True, we ignore whatever |
|
1386 | 1394 | value is passed in to the constructor. |
|
1387 | 1395 | """ |
|
1388 | 1396 | # During normal operations, we expect to deal with not more than four |
|
1389 | 1397 | # revs at a time (such as during commit --amend). When rebasing large |
|
1390 | 1398 | # stacks of commits, the number can go up, hence the config knob below. |
|
1391 | 1399 | cachesize = 4 |
|
1392 | 1400 | optiontreemanifest = False |
|
1393 | 1401 | opts = getattr(opener, 'options', None) |
|
1394 | 1402 | if opts is not None: |
|
1395 | 1403 | cachesize = opts.get('manifestcachesize', cachesize) |
|
1396 | 1404 | optiontreemanifest = opts.get('treemanifest', False) |
|
1397 | 1405 | |
|
1398 | 1406 | self._treeondisk = optiontreemanifest or treemanifest |
|
1399 | 1407 | |
|
1400 | 1408 | self._fulltextcache = manifestfulltextcache(cachesize) |
|
1401 | 1409 | |
|
1402 | 1410 | if tree: |
|
1403 | 1411 | assert self._treeondisk, 'opts is %r' % opts |
|
1404 | 1412 | |
|
1405 | 1413 | if indexfile is None: |
|
1406 | 1414 | indexfile = '00manifest.i' |
|
1407 | 1415 | if tree: |
|
1408 | 1416 | indexfile = "meta/" + tree + indexfile |
|
1409 | 1417 | |
|
1410 | 1418 | self.tree = tree |
|
1411 | 1419 | |
|
1412 | 1420 | # The dirlogcache is kept on the root manifest log |
|
1413 | 1421 | if tree: |
|
1414 | 1422 | self._dirlogcache = dirlogcache |
|
1415 | 1423 | else: |
|
1416 | 1424 | self._dirlogcache = {'': self} |
|
1417 | 1425 | |
|
1418 | 1426 | self._revlog = revlog.revlog(opener, indexfile, |
|
1419 | 1427 | # only root indexfile is cached |
|
1420 | 1428 | checkambig=not bool(tree), |
|
1421 | 1429 | mmaplargeindex=True) |
|
1422 | 1430 | |
|
1423 | 1431 | self.index = self._revlog.index |
|
1424 | 1432 | self.version = self._revlog.version |
|
1425 | 1433 | self._generaldelta = self._revlog._generaldelta |
|
1426 | 1434 | |
|
1427 | 1435 | def _setupmanifestcachehooks(self, repo): |
|
1428 | 1436 | """Persist the manifestfulltextcache on lock release""" |
|
1429 | 1437 | if not util.safehasattr(repo, '_lockref'): |
|
1430 | 1438 | return |
|
1431 | 1439 | |
|
1432 | 1440 | self._fulltextcache._opener = repo.cachevfs |
|
1433 | 1441 | reporef = weakref.ref(repo) |
|
1434 | 1442 | manifestrevlogref = weakref.ref(self) |
|
1435 | 1443 | |
|
1436 | 1444 | def persistmanifestcache(): |
|
1437 | 1445 | repo = reporef() |
|
1438 | 1446 | self = manifestrevlogref() |
|
1439 | 1447 | if repo is None or self is None: |
|
1440 | 1448 | return |
|
1441 | 1449 | if repo.manifestlog.getstorage(b'') is not self: |
|
1442 | 1450 | # there's a different manifest in play now, abort |
|
1443 | 1451 | return |
|
1444 | 1452 | self._fulltextcache.write() |
|
1445 | 1453 | |
|
1446 | 1454 | if repo._currentlock(repo._lockref) is not None: |
|
1447 | 1455 | repo._afterlock(persistmanifestcache) |
|
1448 | 1456 | |
|
1449 | 1457 | @property |
|
1450 | 1458 | def fulltextcache(self): |
|
1451 | 1459 | return self._fulltextcache |
|
1452 | 1460 | |
|
1453 | 1461 | def clearcaches(self, clear_persisted_data=False): |
|
1454 | 1462 | self._revlog.clearcaches() |
|
1455 | 1463 | self._fulltextcache.clear(clear_persisted_data=clear_persisted_data) |
|
1456 | 1464 | self._dirlogcache = {self.tree: self} |
|
1457 | 1465 | |
|
1458 | 1466 | def dirlog(self, d): |
|
1459 | 1467 | if d: |
|
1460 | 1468 | assert self._treeondisk |
|
1461 | 1469 | if d not in self._dirlogcache: |
|
1462 | 1470 | mfrevlog = manifestrevlog(self.opener, d, |
|
1463 | 1471 | self._dirlogcache, |
|
1464 | 1472 | treemanifest=self._treeondisk) |
|
1465 | 1473 | self._dirlogcache[d] = mfrevlog |
|
1466 | 1474 | return self._dirlogcache[d] |
|
1467 | 1475 | |
|
1468 | 1476 | def add(self, m, transaction, link, p1, p2, added, removed, readtree=None, |
|
1469 | 1477 | match=None): |
|
1470 | 1478 | if p1 in self.fulltextcache and util.safehasattr(m, 'fastdelta'): |
|
1471 | 1479 | # If our first parent is in the manifest cache, we can |
|
1472 | 1480 | # compute a delta here using properties we know about the |
|
1473 | 1481 | # manifest up-front, which may save time later for the |
|
1474 | 1482 | # revlog layer. |
|
1475 | 1483 | |
|
1476 | 1484 | _checkforbidden(added) |
|
1477 | 1485 | # combine the changed lists into one sorted iterator |
|
1478 | 1486 | work = heapq.merge([(x, False) for x in added], |
|
1479 | 1487 | [(x, True) for x in removed]) |
|
1480 | 1488 | |
|
1481 | 1489 | arraytext, deltatext = m.fastdelta(self.fulltextcache[p1], work) |
|
1482 | 1490 | cachedelta = self._revlog.rev(p1), deltatext |
|
1483 | 1491 | text = util.buffer(arraytext) |
|
1484 | 1492 | n = self._revlog.addrevision(text, transaction, link, p1, p2, |
|
1485 | 1493 | cachedelta) |
|
1486 | 1494 | else: |
|
1487 | 1495 | # The first parent manifest isn't already loaded, so we'll |
|
1488 | 1496 | # just encode a fulltext of the manifest and pass that |
|
1489 | 1497 | # through to the revlog layer, and let it handle the delta |
|
1490 | 1498 | # process. |
|
1491 | 1499 | if self._treeondisk: |
|
1492 | 1500 | assert readtree, "readtree must be set for treemanifest writes" |
|
1493 | 1501 | assert match, "match must be specified for treemanifest writes" |
|
1494 | 1502 | m1 = readtree(self.tree, p1) |
|
1495 | 1503 | m2 = readtree(self.tree, p2) |
|
1496 | 1504 | n = self._addtree(m, transaction, link, m1, m2, readtree, |
|
1497 | 1505 | match=match) |
|
1498 | 1506 | arraytext = None |
|
1499 | 1507 | else: |
|
1500 | 1508 | text = m.text() |
|
1501 | 1509 | n = self._revlog.addrevision(text, transaction, link, p1, p2) |
|
1502 | 1510 | arraytext = bytearray(text) |
|
1503 | 1511 | |
|
1504 | 1512 | if arraytext is not None: |
|
1505 | 1513 | self.fulltextcache[n] = arraytext |
|
1506 | 1514 | |
|
1507 | 1515 | return n |
|
1508 | 1516 | |
|
1509 | 1517 | def _addtree(self, m, transaction, link, m1, m2, readtree, match): |
|
1510 | 1518 | # If the manifest is unchanged compared to one parent, |
|
1511 | 1519 | # don't write a new revision |
|
1512 | 1520 | if self.tree != '' and (m.unmodifiedsince(m1) or m.unmodifiedsince( |
|
1513 | 1521 | m2)): |
|
1514 | 1522 | return m.node() |
|
1515 | 1523 | def writesubtree(subm, subp1, subp2, match): |
|
1516 | 1524 | sublog = self.dirlog(subm.dir()) |
|
1517 | 1525 | sublog.add(subm, transaction, link, subp1, subp2, None, None, |
|
1518 | 1526 | readtree=readtree, match=match) |
|
1519 | 1527 | m.writesubtrees(m1, m2, writesubtree, match) |
|
1520 | 1528 | text = m.dirtext() |
|
1521 | 1529 | n = None |
|
1522 | 1530 | if self.tree != '': |
|
1523 | 1531 | # Double-check whether contents are unchanged to one parent |
|
1524 | 1532 | if text == m1.dirtext(): |
|
1525 | 1533 | n = m1.node() |
|
1526 | 1534 | elif text == m2.dirtext(): |
|
1527 | 1535 | n = m2.node() |
|
1528 | 1536 | |
|
1529 | 1537 | if not n: |
|
1530 | 1538 | n = self._revlog.addrevision(text, transaction, link, m1.node(), |
|
1531 | 1539 | m2.node()) |
|
1532 | 1540 | |
|
1533 | 1541 | # Save nodeid so parent manifest can calculate its nodeid |
|
1534 | 1542 | m.setnode(n) |
|
1535 | 1543 | return n |
|
1536 | 1544 | |
|
1537 | 1545 | def __len__(self): |
|
1538 | 1546 | return len(self._revlog) |
|
1539 | 1547 | |
|
1540 | 1548 | def __iter__(self): |
|
1541 | 1549 | return self._revlog.__iter__() |
|
1542 | 1550 | |
|
1543 | 1551 | def rev(self, node): |
|
1544 | 1552 | return self._revlog.rev(node) |
|
1545 | 1553 | |
|
1546 | 1554 | def node(self, rev): |
|
1547 | 1555 | return self._revlog.node(rev) |
|
1548 | 1556 | |
|
1549 | 1557 | def lookup(self, value): |
|
1550 | 1558 | return self._revlog.lookup(value) |
|
1551 | 1559 | |
|
1552 | 1560 | def parentrevs(self, rev): |
|
1553 | 1561 | return self._revlog.parentrevs(rev) |
|
1554 | 1562 | |
|
1555 | 1563 | def parents(self, node): |
|
1556 | 1564 | return self._revlog.parents(node) |
|
1557 | 1565 | |
|
1558 | 1566 | def linkrev(self, rev): |
|
1559 | 1567 | return self._revlog.linkrev(rev) |
|
1560 | 1568 | |
|
1561 | 1569 | def checksize(self): |
|
1562 | 1570 | return self._revlog.checksize() |
|
1563 | 1571 | |
|
1564 | 1572 | def revision(self, node, _df=None, raw=False): |
|
1565 | 1573 | return self._revlog.revision(node, _df=_df, raw=raw) |
|
1566 | 1574 | |
|
1567 | 1575 | def revdiff(self, rev1, rev2): |
|
1568 | 1576 | return self._revlog.revdiff(rev1, rev2) |
|
1569 | 1577 | |
|
1570 | 1578 | def cmp(self, node, text): |
|
1571 | 1579 | return self._revlog.cmp(node, text) |
|
1572 | 1580 | |
|
1573 | 1581 | def deltaparent(self, rev): |
|
1574 | 1582 | return self._revlog.deltaparent(rev) |
|
1575 | 1583 | |
|
1576 | 1584 | def emitrevisions(self, nodes, nodesorder=None, |
|
1577 | 1585 | revisiondata=False, assumehaveparentrevisions=False, |
|
1578 | 1586 | deltamode=repository.CG_DELTAMODE_STD): |
|
1579 | 1587 | return self._revlog.emitrevisions( |
|
1580 | 1588 | nodes, nodesorder=nodesorder, revisiondata=revisiondata, |
|
1581 | 1589 | assumehaveparentrevisions=assumehaveparentrevisions, |
|
1582 | 1590 | deltamode=deltamode) |
|
1583 | 1591 | |
|
1584 | 1592 | def addgroup(self, deltas, linkmapper, transaction, addrevisioncb=None): |
|
1585 | 1593 | return self._revlog.addgroup(deltas, linkmapper, transaction, |
|
1586 | 1594 | addrevisioncb=addrevisioncb) |
|
1587 | 1595 | |
|
1588 | 1596 | def rawsize(self, rev): |
|
1589 | 1597 | return self._revlog.rawsize(rev) |
|
1590 | 1598 | |
|
1591 | 1599 | def getstrippoint(self, minlink): |
|
1592 | 1600 | return self._revlog.getstrippoint(minlink) |
|
1593 | 1601 | |
|
1594 | 1602 | def strip(self, minlink, transaction): |
|
1595 | 1603 | return self._revlog.strip(minlink, transaction) |
|
1596 | 1604 | |
|
1597 | 1605 | def files(self): |
|
1598 | 1606 | return self._revlog.files() |
|
1599 | 1607 | |
|
1600 | 1608 | def clone(self, tr, destrevlog, **kwargs): |
|
1601 | 1609 | if not isinstance(destrevlog, manifestrevlog): |
|
1602 | 1610 | raise error.ProgrammingError('expected manifestrevlog to clone()') |
|
1603 | 1611 | |
|
1604 | 1612 | return self._revlog.clone(tr, destrevlog._revlog, **kwargs) |
|
1605 | 1613 | |
|
1606 | 1614 | def storageinfo(self, exclusivefiles=False, sharedfiles=False, |
|
1607 | 1615 | revisionscount=False, trackedsize=False, |
|
1608 | 1616 | storedsize=False): |
|
1609 | 1617 | return self._revlog.storageinfo( |
|
1610 | 1618 | exclusivefiles=exclusivefiles, sharedfiles=sharedfiles, |
|
1611 | 1619 | revisionscount=revisionscount, trackedsize=trackedsize, |
|
1612 | 1620 | storedsize=storedsize) |
|
1613 | 1621 | |
|
1614 | 1622 | @property |
|
1615 | 1623 | def indexfile(self): |
|
1616 | 1624 | return self._revlog.indexfile |
|
1617 | 1625 | |
|
1618 | 1626 | @indexfile.setter |
|
1619 | 1627 | def indexfile(self, value): |
|
1620 | 1628 | self._revlog.indexfile = value |
|
1621 | 1629 | |
|
1622 | 1630 | @property |
|
1623 | 1631 | def opener(self): |
|
1624 | 1632 | return self._revlog.opener |
|
1625 | 1633 | |
|
1626 | 1634 | @opener.setter |
|
1627 | 1635 | def opener(self, value): |
|
1628 | 1636 | self._revlog.opener = value |
|
1629 | 1637 | |
|
1630 | 1638 | @interfaceutil.implementer(repository.imanifestlog) |
|
1631 | 1639 | class manifestlog(object): |
|
1632 | 1640 | """A collection class representing the collection of manifest snapshots |
|
1633 | 1641 | referenced by commits in the repository. |
|
1634 | 1642 | |
|
1635 | 1643 | In this situation, 'manifest' refers to the abstract concept of a snapshot |
|
1636 | 1644 | of the list of files in the given commit. Consumers of the output of this |
|
1637 | 1645 | class do not care about the implementation details of the actual manifests |
|
1638 | 1646 | they receive (i.e. tree or flat or lazily loaded, etc).""" |
|
1639 | 1647 | def __init__(self, opener, repo, rootstore, narrowmatch): |
|
1640 | 1648 | usetreemanifest = False |
|
1641 | 1649 | cachesize = 4 |
|
1642 | 1650 | |
|
1643 | 1651 | opts = getattr(opener, 'options', None) |
|
1644 | 1652 | if opts is not None: |
|
1645 | 1653 | usetreemanifest = opts.get('treemanifest', usetreemanifest) |
|
1646 | 1654 | cachesize = opts.get('manifestcachesize', cachesize) |
|
1647 | 1655 | |
|
1648 | 1656 | self._treemanifests = usetreemanifest |
|
1649 | 1657 | |
|
1650 | 1658 | self._rootstore = rootstore |
|
1651 | 1659 | self._rootstore._setupmanifestcachehooks(repo) |
|
1652 | 1660 | self._narrowmatch = narrowmatch |
|
1653 | 1661 | |
|
1654 | 1662 | # A cache of the manifestctx or treemanifestctx for each directory |
|
1655 | 1663 | self._dirmancache = {} |
|
1656 | 1664 | self._dirmancache[''] = util.lrucachedict(cachesize) |
|
1657 | 1665 | |
|
1658 | 1666 | self._cachesize = cachesize |
|
1659 | 1667 | |
|
1660 | 1668 | def __getitem__(self, node): |
|
1661 | 1669 | """Retrieves the manifest instance for the given node. Throws a |
|
1662 | 1670 | LookupError if not found. |
|
1663 | 1671 | """ |
|
1664 | 1672 | return self.get('', node) |
|
1665 | 1673 | |
|
1666 | 1674 | def get(self, tree, node, verify=True): |
|
1667 | 1675 | """Retrieves the manifest instance for the given node. Throws a |
|
1668 | 1676 | LookupError if not found. |
|
1669 | 1677 | |
|
1670 | 1678 | `verify` - if True an exception will be thrown if the node is not in |
|
1671 | 1679 | the revlog |
|
1672 | 1680 | """ |
|
1673 | 1681 | if node in self._dirmancache.get(tree, ()): |
|
1674 | 1682 | return self._dirmancache[tree][node] |
|
1675 | 1683 | |
|
1676 | 1684 | if not self._narrowmatch.always(): |
|
1677 | 1685 | if not self._narrowmatch.visitdir(tree[:-1] or '.'): |
|
1678 | 1686 | return excludeddirmanifestctx(tree, node) |
|
1679 | 1687 | if tree: |
|
1680 | 1688 | if self._rootstore._treeondisk: |
|
1681 | 1689 | if verify: |
|
1682 | 1690 | # Side-effect is LookupError is raised if node doesn't |
|
1683 | 1691 | # exist. |
|
1684 | 1692 | self.getstorage(tree).rev(node) |
|
1685 | 1693 | |
|
1686 | 1694 | m = treemanifestctx(self, tree, node) |
|
1687 | 1695 | else: |
|
1688 | 1696 | raise error.Abort( |
|
1689 | 1697 | _("cannot ask for manifest directory '%s' in a flat " |
|
1690 | 1698 | "manifest") % tree) |
|
1691 | 1699 | else: |
|
1692 | 1700 | if verify: |
|
1693 | 1701 | # Side-effect is LookupError is raised if node doesn't exist. |
|
1694 | 1702 | self._rootstore.rev(node) |
|
1695 | 1703 | |
|
1696 | 1704 | if self._treemanifests: |
|
1697 | 1705 | m = treemanifestctx(self, '', node) |
|
1698 | 1706 | else: |
|
1699 | 1707 | m = manifestctx(self, node) |
|
1700 | 1708 | |
|
1701 | 1709 | if node != nullid: |
|
1702 | 1710 | mancache = self._dirmancache.get(tree) |
|
1703 | 1711 | if not mancache: |
|
1704 | 1712 | mancache = util.lrucachedict(self._cachesize) |
|
1705 | 1713 | self._dirmancache[tree] = mancache |
|
1706 | 1714 | mancache[node] = m |
|
1707 | 1715 | return m |
|
1708 | 1716 | |
|
1709 | 1717 | def getstorage(self, tree): |
|
1710 | 1718 | return self._rootstore.dirlog(tree) |
|
1711 | 1719 | |
|
1712 | 1720 | def clearcaches(self, clear_persisted_data=False): |
|
1713 | 1721 | self._dirmancache.clear() |
|
1714 | 1722 | self._rootstore.clearcaches(clear_persisted_data=clear_persisted_data) |
|
1715 | 1723 | |
|
1716 | 1724 | def rev(self, node): |
|
1717 | 1725 | return self._rootstore.rev(node) |
|
1718 | 1726 | |
|
1719 | 1727 | @interfaceutil.implementer(repository.imanifestrevisionwritable) |
|
1720 | 1728 | class memmanifestctx(object): |
|
1721 | 1729 | def __init__(self, manifestlog): |
|
1722 | 1730 | self._manifestlog = manifestlog |
|
1723 | 1731 | self._manifestdict = manifestdict() |
|
1724 | 1732 | |
|
1725 | 1733 | def _storage(self): |
|
1726 | 1734 | return self._manifestlog.getstorage(b'') |
|
1727 | 1735 | |
|
1728 | 1736 | def new(self): |
|
1729 | 1737 | return memmanifestctx(self._manifestlog) |
|
1730 | 1738 | |
|
1731 | 1739 | def copy(self): |
|
1732 | 1740 | memmf = memmanifestctx(self._manifestlog) |
|
1733 | 1741 | memmf._manifestdict = self.read().copy() |
|
1734 | 1742 | return memmf |
|
1735 | 1743 | |
|
1736 | 1744 | def read(self): |
|
1737 | 1745 | return self._manifestdict |
|
1738 | 1746 | |
|
1739 | 1747 | def write(self, transaction, link, p1, p2, added, removed, match=None): |
|
1740 | 1748 | return self._storage().add(self._manifestdict, transaction, link, |
|
1741 | 1749 | p1, p2, added, removed, match=match) |
|
1742 | 1750 | |
|
1743 | 1751 | @interfaceutil.implementer(repository.imanifestrevisionstored) |
|
1744 | 1752 | class manifestctx(object): |
|
1745 | 1753 | """A class representing a single revision of a manifest, including its |
|
1746 | 1754 | contents, its parent revs, and its linkrev. |
|
1747 | 1755 | """ |
|
1748 | 1756 | def __init__(self, manifestlog, node): |
|
1749 | 1757 | self._manifestlog = manifestlog |
|
1750 | 1758 | self._data = None |
|
1751 | 1759 | |
|
1752 | 1760 | self._node = node |
|
1753 | 1761 | |
|
1754 | 1762 | # TODO: We eventually want p1, p2, and linkrev exposed on this class, |
|
1755 | 1763 | # but let's add it later when something needs it and we can load it |
|
1756 | 1764 | # lazily. |
|
1757 | 1765 | #self.p1, self.p2 = store.parents(node) |
|
1758 | 1766 | #rev = store.rev(node) |
|
1759 | 1767 | #self.linkrev = store.linkrev(rev) |
|
1760 | 1768 | |
|
1761 | 1769 | def _storage(self): |
|
1762 | 1770 | return self._manifestlog.getstorage(b'') |
|
1763 | 1771 | |
|
1764 | 1772 | def node(self): |
|
1765 | 1773 | return self._node |
|
1766 | 1774 | |
|
1767 | 1775 | def new(self): |
|
1768 | 1776 | return memmanifestctx(self._manifestlog) |
|
1769 | 1777 | |
|
1770 | 1778 | def copy(self): |
|
1771 | 1779 | memmf = memmanifestctx(self._manifestlog) |
|
1772 | 1780 | memmf._manifestdict = self.read().copy() |
|
1773 | 1781 | return memmf |
|
1774 | 1782 | |
|
1775 | 1783 | @propertycache |
|
1776 | 1784 | def parents(self): |
|
1777 | 1785 | return self._storage().parents(self._node) |
|
1778 | 1786 | |
|
1779 | 1787 | def read(self): |
|
1780 | 1788 | if self._data is None: |
|
1781 | 1789 | if self._node == nullid: |
|
1782 | 1790 | self._data = manifestdict() |
|
1783 | 1791 | else: |
|
1784 | 1792 | store = self._storage() |
|
1785 | 1793 | if self._node in store.fulltextcache: |
|
1786 | 1794 | text = pycompat.bytestr(store.fulltextcache[self._node]) |
|
1787 | 1795 | else: |
|
1788 | 1796 | text = store.revision(self._node) |
|
1789 | 1797 | arraytext = bytearray(text) |
|
1790 | 1798 | store.fulltextcache[self._node] = arraytext |
|
1791 | 1799 | self._data = manifestdict(text) |
|
1792 | 1800 | return self._data |
|
1793 | 1801 | |
|
1794 | 1802 | def readfast(self, shallow=False): |
|
1795 | 1803 | '''Calls either readdelta or read, based on which would be less work. |
|
1796 | 1804 | readdelta is called if the delta is against the p1, and therefore can be |
|
1797 | 1805 | read quickly. |
|
1798 | 1806 | |
|
1799 | 1807 | If `shallow` is True, nothing changes since this is a flat manifest. |
|
1800 | 1808 | ''' |
|
1801 | 1809 | store = self._storage() |
|
1802 | 1810 | r = store.rev(self._node) |
|
1803 | 1811 | deltaparent = store.deltaparent(r) |
|
1804 | 1812 | if deltaparent != nullrev and deltaparent in store.parentrevs(r): |
|
1805 | 1813 | return self.readdelta() |
|
1806 | 1814 | return self.read() |
|
1807 | 1815 | |
|
1808 | 1816 | def readdelta(self, shallow=False): |
|
1809 | 1817 | '''Returns a manifest containing just the entries that are present |
|
1810 | 1818 | in this manifest, but not in its p1 manifest. This is efficient to read |
|
1811 | 1819 | if the revlog delta is already p1. |
|
1812 | 1820 | |
|
1813 | 1821 | Changing the value of `shallow` has no effect on flat manifests. |
|
1814 | 1822 | ''' |
|
1815 | 1823 | store = self._storage() |
|
1816 | 1824 | r = store.rev(self._node) |
|
1817 | 1825 | d = mdiff.patchtext(store.revdiff(store.deltaparent(r), r)) |
|
1818 | 1826 | return manifestdict(d) |
|
1819 | 1827 | |
|
1820 | 1828 | def find(self, key): |
|
1821 | 1829 | return self.read().find(key) |
|
1822 | 1830 | |
|
1823 | 1831 | @interfaceutil.implementer(repository.imanifestrevisionwritable) |
|
1824 | 1832 | class memtreemanifestctx(object): |
|
1825 | 1833 | def __init__(self, manifestlog, dir=''): |
|
1826 | 1834 | self._manifestlog = manifestlog |
|
1827 | 1835 | self._dir = dir |
|
1828 | 1836 | self._treemanifest = treemanifest() |
|
1829 | 1837 | |
|
1830 | 1838 | def _storage(self): |
|
1831 | 1839 | return self._manifestlog.getstorage(b'') |
|
1832 | 1840 | |
|
1833 | 1841 | def new(self, dir=''): |
|
1834 | 1842 | return memtreemanifestctx(self._manifestlog, dir=dir) |
|
1835 | 1843 | |
|
1836 | 1844 | def copy(self): |
|
1837 | 1845 | memmf = memtreemanifestctx(self._manifestlog, dir=self._dir) |
|
1838 | 1846 | memmf._treemanifest = self._treemanifest.copy() |
|
1839 | 1847 | return memmf |
|
1840 | 1848 | |
|
1841 | 1849 | def read(self): |
|
1842 | 1850 | return self._treemanifest |
|
1843 | 1851 | |
|
1844 | 1852 | def write(self, transaction, link, p1, p2, added, removed, match=None): |
|
1845 | 1853 | def readtree(dir, node): |
|
1846 | 1854 | return self._manifestlog.get(dir, node).read() |
|
1847 | 1855 | return self._storage().add(self._treemanifest, transaction, link, |
|
1848 | 1856 | p1, p2, added, removed, readtree=readtree, |
|
1849 | 1857 | match=match) |
|
1850 | 1858 | |
|
1851 | 1859 | @interfaceutil.implementer(repository.imanifestrevisionstored) |
|
1852 | 1860 | class treemanifestctx(object): |
|
1853 | 1861 | def __init__(self, manifestlog, dir, node): |
|
1854 | 1862 | self._manifestlog = manifestlog |
|
1855 | 1863 | self._dir = dir |
|
1856 | 1864 | self._data = None |
|
1857 | 1865 | |
|
1858 | 1866 | self._node = node |
|
1859 | 1867 | |
|
1860 | 1868 | # TODO: Load p1/p2/linkrev lazily. They need to be lazily loaded so that |
|
1861 | 1869 | # we can instantiate treemanifestctx objects for directories we don't |
|
1862 | 1870 | # have on disk. |
|
1863 | 1871 | #self.p1, self.p2 = store.parents(node) |
|
1864 | 1872 | #rev = store.rev(node) |
|
1865 | 1873 | #self.linkrev = store.linkrev(rev) |
|
1866 | 1874 | |
|
1867 | 1875 | def _storage(self): |
|
1868 | 1876 | narrowmatch = self._manifestlog._narrowmatch |
|
1869 | 1877 | if not narrowmatch.always(): |
|
1870 | 1878 | if not narrowmatch.visitdir(self._dir[:-1] or '.'): |
|
1871 | 1879 | return excludedmanifestrevlog(self._dir) |
|
1872 | 1880 | return self._manifestlog.getstorage(self._dir) |
|
1873 | 1881 | |
|
1874 | 1882 | def read(self): |
|
1875 | 1883 | if self._data is None: |
|
1876 | 1884 | store = self._storage() |
|
1877 | 1885 | if self._node == nullid: |
|
1878 | 1886 | self._data = treemanifest() |
|
1879 | 1887 | # TODO accessing non-public API |
|
1880 | 1888 | elif store._treeondisk: |
|
1881 | 1889 | m = treemanifest(dir=self._dir) |
|
1882 | 1890 | def gettext(): |
|
1883 | 1891 | return store.revision(self._node) |
|
1884 | 1892 | def readsubtree(dir, subm): |
|
1885 | 1893 | # Set verify to False since we need to be able to create |
|
1886 | 1894 | # subtrees for trees that don't exist on disk. |
|
1887 | 1895 | return self._manifestlog.get(dir, subm, verify=False).read() |
|
1888 | 1896 | m.read(gettext, readsubtree) |
|
1889 | 1897 | m.setnode(self._node) |
|
1890 | 1898 | self._data = m |
|
1891 | 1899 | else: |
|
1892 | 1900 | if self._node in store.fulltextcache: |
|
1893 | 1901 | text = pycompat.bytestr(store.fulltextcache[self._node]) |
|
1894 | 1902 | else: |
|
1895 | 1903 | text = store.revision(self._node) |
|
1896 | 1904 | arraytext = bytearray(text) |
|
1897 | 1905 | store.fulltextcache[self._node] = arraytext |
|
1898 | 1906 | self._data = treemanifest(dir=self._dir, text=text) |
|
1899 | 1907 | |
|
1900 | 1908 | return self._data |
|
1901 | 1909 | |
|
1902 | 1910 | def node(self): |
|
1903 | 1911 | return self._node |
|
1904 | 1912 | |
|
1905 | 1913 | def new(self, dir=''): |
|
1906 | 1914 | return memtreemanifestctx(self._manifestlog, dir=dir) |
|
1907 | 1915 | |
|
1908 | 1916 | def copy(self): |
|
1909 | 1917 | memmf = memtreemanifestctx(self._manifestlog, dir=self._dir) |
|
1910 | 1918 | memmf._treemanifest = self.read().copy() |
|
1911 | 1919 | return memmf |
|
1912 | 1920 | |
|
1913 | 1921 | @propertycache |
|
1914 | 1922 | def parents(self): |
|
1915 | 1923 | return self._storage().parents(self._node) |
|
1916 | 1924 | |
|
1917 | 1925 | def readdelta(self, shallow=False): |
|
1918 | 1926 | '''Returns a manifest containing just the entries that are present |
|
1919 | 1927 | in this manifest, but not in its p1 manifest. This is efficient to read |
|
1920 | 1928 | if the revlog delta is already p1. |
|
1921 | 1929 | |
|
1922 | 1930 | If `shallow` is True, this will read the delta for this directory, |
|
1923 | 1931 | without recursively reading subdirectory manifests. Instead, any |
|
1924 | 1932 | subdirectory entry will be reported as it appears in the manifest, i.e. |
|
1925 | 1933 | the subdirectory will be reported among files and distinguished only by |
|
1926 | 1934 | its 't' flag. |
|
1927 | 1935 | ''' |
|
1928 | 1936 | store = self._storage() |
|
1929 | 1937 | if shallow: |
|
1930 | 1938 | r = store.rev(self._node) |
|
1931 | 1939 | d = mdiff.patchtext(store.revdiff(store.deltaparent(r), r)) |
|
1932 | 1940 | return manifestdict(d) |
|
1933 | 1941 | else: |
|
1934 | 1942 | # Need to perform a slow delta |
|
1935 | 1943 | r0 = store.deltaparent(store.rev(self._node)) |
|
1936 | 1944 | m0 = self._manifestlog.get(self._dir, store.node(r0)).read() |
|
1937 | 1945 | m1 = self.read() |
|
1938 | 1946 | md = treemanifest(dir=self._dir) |
|
1939 | 1947 | for f, ((n0, fl0), (n1, fl1)) in m0.diff(m1).iteritems(): |
|
1940 | 1948 | if n1: |
|
1941 | 1949 | md[f] = n1 |
|
1942 | 1950 | if fl1: |
|
1943 | 1951 | md.setflag(f, fl1) |
|
1944 | 1952 | return md |
|
1945 | 1953 | |
|
1946 | 1954 | def readfast(self, shallow=False): |
|
1947 | 1955 | '''Calls either readdelta or read, based on which would be less work. |
|
1948 | 1956 | readdelta is called if the delta is against the p1, and therefore can be |
|
1949 | 1957 | read quickly. |
|
1950 | 1958 | |
|
1951 | 1959 | If `shallow` is True, it only returns the entries from this manifest, |
|
1952 | 1960 | and not any submanifests. |
|
1953 | 1961 | ''' |
|
1954 | 1962 | store = self._storage() |
|
1955 | 1963 | r = store.rev(self._node) |
|
1956 | 1964 | deltaparent = store.deltaparent(r) |
|
1957 | 1965 | if (deltaparent != nullrev and |
|
1958 | 1966 | deltaparent in store.parentrevs(r)): |
|
1959 | 1967 | return self.readdelta(shallow=shallow) |
|
1960 | 1968 | |
|
1961 | 1969 | if shallow: |
|
1962 | 1970 | return manifestdict(store.revision(self._node)) |
|
1963 | 1971 | else: |
|
1964 | 1972 | return self.read() |
|
1965 | 1973 | |
|
1966 | 1974 | def find(self, key): |
|
1967 | 1975 | return self.read().find(key) |
|
1968 | 1976 | |
|
1969 | 1977 | class excludeddir(treemanifest): |
|
1970 | 1978 | """Stand-in for a directory that is excluded from the repository. |
|
1971 | 1979 | |
|
1972 | 1980 | With narrowing active on a repository that uses treemanifests, |
|
1973 | 1981 | some of the directory revlogs will be excluded from the resulting |
|
1974 | 1982 | clone. This is a huge storage win for clients, but means we need |
|
1975 | 1983 | some sort of pseudo-manifest to surface to internals so we can |
|
1976 | 1984 | detect a merge conflict outside the narrowspec. That's what this |
|
1977 | 1985 | class is: it stands in for a directory whose node is known, but |
|
1978 | 1986 | whose contents are unknown. |
|
1979 | 1987 | """ |
|
1980 | 1988 | def __init__(self, dir, node): |
|
1981 | 1989 | super(excludeddir, self).__init__(dir) |
|
1982 | 1990 | self._node = node |
|
1983 | 1991 | # Add an empty file, which will be included by iterators and such, |
|
1984 | 1992 | # appearing as the directory itself (i.e. something like "dir/") |
|
1985 | 1993 | self._files[''] = node |
|
1986 | 1994 | self._flags[''] = 't' |
|
1987 | 1995 | |
|
1988 | 1996 | # Manifests outside the narrowspec should never be modified, so avoid |
|
1989 | 1997 | # copying. This makes a noticeable difference when there are very many |
|
1990 | 1998 | # directories outside the narrowspec. Also, it makes sense for the copy to |
|
1991 | 1999 | # be of the same type as the original, which would not happen with the |
|
1992 | 2000 | # super type's copy(). |
|
1993 | 2001 | def copy(self): |
|
1994 | 2002 | return self |
|
1995 | 2003 | |
|
1996 | 2004 | class excludeddirmanifestctx(treemanifestctx): |
|
1997 | 2005 | """context wrapper for excludeddir - see that docstring for rationale""" |
|
1998 | 2006 | def __init__(self, dir, node): |
|
1999 | 2007 | self._dir = dir |
|
2000 | 2008 | self._node = node |
|
2001 | 2009 | |
|
2002 | 2010 | def read(self): |
|
2003 | 2011 | return excludeddir(self._dir, self._node) |
|
2004 | 2012 | |
|
2005 | 2013 | def write(self, *args): |
|
2006 | 2014 | raise error.ProgrammingError( |
|
2007 | 2015 | 'attempt to write manifest from excluded dir %s' % self._dir) |
|
2008 | 2016 | |
|
2009 | 2017 | class excludedmanifestrevlog(manifestrevlog): |
|
2010 | 2018 | """Stand-in for excluded treemanifest revlogs. |
|
2011 | 2019 | |
|
2012 | 2020 | When narrowing is active on a treemanifest repository, we'll have |
|
2013 | 2021 | references to directories we can't see due to the revlog being |
|
2014 | 2022 | skipped. This class exists to conform to the manifestrevlog |
|
2015 | 2023 | interface for those directories and proactively prevent writes to |
|
2016 | 2024 | outside the narrowspec. |
|
2017 | 2025 | """ |
|
2018 | 2026 | |
|
2019 | 2027 | def __init__(self, dir): |
|
2020 | 2028 | self._dir = dir |
|
2021 | 2029 | |
|
2022 | 2030 | def __len__(self): |
|
2023 | 2031 | raise error.ProgrammingError( |
|
2024 | 2032 | 'attempt to get length of excluded dir %s' % self._dir) |
|
2025 | 2033 | |
|
2026 | 2034 | def rev(self, node): |
|
2027 | 2035 | raise error.ProgrammingError( |
|
2028 | 2036 | 'attempt to get rev from excluded dir %s' % self._dir) |
|
2029 | 2037 | |
|
2030 | 2038 | def linkrev(self, node): |
|
2031 | 2039 | raise error.ProgrammingError( |
|
2032 | 2040 | 'attempt to get linkrev from excluded dir %s' % self._dir) |
|
2033 | 2041 | |
|
2034 | 2042 | def node(self, rev): |
|
2035 | 2043 | raise error.ProgrammingError( |
|
2036 | 2044 | 'attempt to get node from excluded dir %s' % self._dir) |
|
2037 | 2045 | |
|
2038 | 2046 | def add(self, *args, **kwargs): |
|
2039 | 2047 | # We should never write entries in dirlogs outside the narrow clone. |
|
2040 | 2048 | # However, the method still gets called from writesubtree() in |
|
2041 | 2049 | # _addtree(), so we need to handle it. We should possibly make that |
|
2042 | 2050 | # avoid calling add() with a clean manifest (_dirty is always False |
|
2043 | 2051 | # in excludeddir instances). |
|
2044 | 2052 | pass |
General Comments 0
You need to be logged in to leave comments.
Login now