##// END OF EJS Templates
revlog: make pure version of _partialmatch() support 40-byte hex nodeids...
Martin von Zweigbergk -
r37837:ee3d58b4 stable
parent child Browse files
Show More
@@ -1,2594 +1,2594
1 1 # revlog.py - storage back-end 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 """Storage back-end for Mercurial.
9 9
10 10 This provides efficient delta storage with O(1) retrieve and append
11 11 and O(changes) merge between branches.
12 12 """
13 13
14 14 from __future__ import absolute_import
15 15
16 16 import collections
17 17 import contextlib
18 18 import errno
19 19 import hashlib
20 20 import heapq
21 21 import os
22 22 import re
23 23 import struct
24 24 import zlib
25 25
26 26 # import stuff from node for others to import from revlog
27 27 from .node import (
28 28 bin,
29 29 hex,
30 30 nullid,
31 31 nullrev,
32 32 wdirfilenodeids,
33 33 wdirhex,
34 34 wdirid,
35 35 wdirrev,
36 36 )
37 37 from .i18n import _
38 38 from .thirdparty import (
39 39 attr,
40 40 )
41 41 from . import (
42 42 ancestor,
43 43 error,
44 44 mdiff,
45 45 policy,
46 46 pycompat,
47 47 templatefilters,
48 48 util,
49 49 )
50 50 from .utils import (
51 51 stringutil,
52 52 )
53 53
54 54 parsers = policy.importmod(r'parsers')
55 55
56 56 # Aliased for performance.
57 57 _zlibdecompress = zlib.decompress
58 58
59 59 # revlog header flags
60 60 REVLOGV0 = 0
61 61 REVLOGV1 = 1
62 62 # Dummy value until file format is finalized.
63 63 # Reminder: change the bounds check in revlog.__init__ when this is changed.
64 64 REVLOGV2 = 0xDEAD
65 65 FLAG_INLINE_DATA = (1 << 16)
66 66 FLAG_GENERALDELTA = (1 << 17)
67 67 REVLOG_DEFAULT_FLAGS = FLAG_INLINE_DATA
68 68 REVLOG_DEFAULT_FORMAT = REVLOGV1
69 69 REVLOG_DEFAULT_VERSION = REVLOG_DEFAULT_FORMAT | REVLOG_DEFAULT_FLAGS
70 70 REVLOGV1_FLAGS = FLAG_INLINE_DATA | FLAG_GENERALDELTA
71 71 REVLOGV2_FLAGS = REVLOGV1_FLAGS
72 72
73 73 # revlog index flags
74 74 REVIDX_ISCENSORED = (1 << 15) # revision has censor metadata, must be verified
75 75 REVIDX_ELLIPSIS = (1 << 14) # revision hash does not match data (narrowhg)
76 76 REVIDX_EXTSTORED = (1 << 13) # revision data is stored externally
77 77 REVIDX_DEFAULT_FLAGS = 0
78 78 # stable order in which flags need to be processed and their processors applied
79 79 REVIDX_FLAGS_ORDER = [
80 80 REVIDX_ISCENSORED,
81 81 REVIDX_ELLIPSIS,
82 82 REVIDX_EXTSTORED,
83 83 ]
84 84 REVIDX_KNOWN_FLAGS = util.bitsfrom(REVIDX_FLAGS_ORDER)
85 85 # bitmark for flags that could cause rawdata content change
86 86 REVIDX_RAWTEXT_CHANGING_FLAGS = REVIDX_ISCENSORED | REVIDX_EXTSTORED
87 87
88 88 # max size of revlog with inline data
89 89 _maxinline = 131072
90 90 _chunksize = 1048576
91 91
92 92 RevlogError = error.RevlogError
93 93 LookupError = error.LookupError
94 94 CensoredNodeError = error.CensoredNodeError
95 95 ProgrammingError = error.ProgrammingError
96 96
97 97 # Store flag processors (cf. 'addflagprocessor()' to register)
98 98 _flagprocessors = {
99 99 REVIDX_ISCENSORED: None,
100 100 }
101 101
102 102 _mdre = re.compile('\1\n')
103 103 def parsemeta(text):
104 104 """return (metadatadict, metadatasize)"""
105 105 # text can be buffer, so we can't use .startswith or .index
106 106 if text[:2] != '\1\n':
107 107 return None, None
108 108 s = _mdre.search(text, 2).start()
109 109 mtext = text[2:s]
110 110 meta = {}
111 111 for l in mtext.splitlines():
112 112 k, v = l.split(": ", 1)
113 113 meta[k] = v
114 114 return meta, (s + 2)
115 115
116 116 def packmeta(meta, text):
117 117 keys = sorted(meta)
118 118 metatext = "".join("%s: %s\n" % (k, meta[k]) for k in keys)
119 119 return "\1\n%s\1\n%s" % (metatext, text)
120 120
121 121 def _censoredtext(text):
122 122 m, offs = parsemeta(text)
123 123 return m and "censored" in m
124 124
125 125 def addflagprocessor(flag, processor):
126 126 """Register a flag processor on a revision data flag.
127 127
128 128 Invariant:
129 129 - Flags need to be defined in REVIDX_KNOWN_FLAGS and REVIDX_FLAGS_ORDER,
130 130 and REVIDX_RAWTEXT_CHANGING_FLAGS if they can alter rawtext.
131 131 - Only one flag processor can be registered on a specific flag.
132 132 - flagprocessors must be 3-tuples of functions (read, write, raw) with the
133 133 following signatures:
134 134 - (read) f(self, rawtext) -> text, bool
135 135 - (write) f(self, text) -> rawtext, bool
136 136 - (raw) f(self, rawtext) -> bool
137 137 "text" is presented to the user. "rawtext" is stored in revlog data, not
138 138 directly visible to the user.
139 139 The boolean returned by these transforms is used to determine whether
140 140 the returned text can be used for hash integrity checking. For example,
141 141 if "write" returns False, then "text" is used to generate hash. If
142 142 "write" returns True, that basically means "rawtext" returned by "write"
143 143 should be used to generate hash. Usually, "write" and "read" return
144 144 different booleans. And "raw" returns a same boolean as "write".
145 145
146 146 Note: The 'raw' transform is used for changegroup generation and in some
147 147 debug commands. In this case the transform only indicates whether the
148 148 contents can be used for hash integrity checks.
149 149 """
150 150 if not flag & REVIDX_KNOWN_FLAGS:
151 151 msg = _("cannot register processor on unknown flag '%#x'.") % (flag)
152 152 raise ProgrammingError(msg)
153 153 if flag not in REVIDX_FLAGS_ORDER:
154 154 msg = _("flag '%#x' undefined in REVIDX_FLAGS_ORDER.") % (flag)
155 155 raise ProgrammingError(msg)
156 156 if flag in _flagprocessors:
157 157 msg = _("cannot register multiple processors on flag '%#x'.") % (flag)
158 158 raise error.Abort(msg)
159 159 _flagprocessors[flag] = processor
160 160
161 161 def getoffset(q):
162 162 return int(q >> 16)
163 163
164 164 def gettype(q):
165 165 return int(q & 0xFFFF)
166 166
167 167 def offset_type(offset, type):
168 168 if (type & ~REVIDX_KNOWN_FLAGS) != 0:
169 169 raise ValueError('unknown revlog index flags')
170 170 return int(int(offset) << 16 | type)
171 171
172 172 _nullhash = hashlib.sha1(nullid)
173 173
174 174 def hash(text, p1, p2):
175 175 """generate a hash from the given text and its parent hashes
176 176
177 177 This hash combines both the current file contents and its history
178 178 in a manner that makes it easy to distinguish nodes with the same
179 179 content in the revision graph.
180 180 """
181 181 # As of now, if one of the parent node is null, p2 is null
182 182 if p2 == nullid:
183 183 # deep copy of a hash is faster than creating one
184 184 s = _nullhash.copy()
185 185 s.update(p1)
186 186 else:
187 187 # none of the parent nodes are nullid
188 188 if p1 < p2:
189 189 a = p1
190 190 b = p2
191 191 else:
192 192 a = p2
193 193 b = p1
194 194 s = hashlib.sha1(a)
195 195 s.update(b)
196 196 s.update(text)
197 197 return s.digest()
198 198
199 199 def _trimchunk(revlog, revs, startidx, endidx=None):
200 200 """returns revs[startidx:endidx] without empty trailing revs
201 201 """
202 202 length = revlog.length
203 203
204 204 if endidx is None:
205 205 endidx = len(revs)
206 206
207 207 # Trim empty revs at the end, but never the very first revision of a chain
208 208 while endidx > 1 and endidx > startidx and length(revs[endidx - 1]) == 0:
209 209 endidx -= 1
210 210
211 211 return revs[startidx:endidx]
212 212
213 213 def _slicechunk(revlog, revs):
214 214 """slice revs to reduce the amount of unrelated data to be read from disk.
215 215
216 216 ``revs`` is sliced into groups that should be read in one time.
217 217 Assume that revs are sorted.
218 218 """
219 219 start = revlog.start
220 220 length = revlog.length
221 221
222 222 if len(revs) <= 1:
223 223 yield revs
224 224 return
225 225
226 226 startbyte = start(revs[0])
227 227 endbyte = start(revs[-1]) + length(revs[-1])
228 228 readdata = deltachainspan = endbyte - startbyte
229 229
230 230 chainpayload = sum(length(r) for r in revs)
231 231
232 232 if deltachainspan:
233 233 density = chainpayload / float(deltachainspan)
234 234 else:
235 235 density = 1.0
236 236
237 237 # Store the gaps in a heap to have them sorted by decreasing size
238 238 gapsheap = []
239 239 heapq.heapify(gapsheap)
240 240 prevend = None
241 241 for i, rev in enumerate(revs):
242 242 revstart = start(rev)
243 243 revlen = length(rev)
244 244
245 245 # Skip empty revisions to form larger holes
246 246 if revlen == 0:
247 247 continue
248 248
249 249 if prevend is not None:
250 250 gapsize = revstart - prevend
251 251 # only consider holes that are large enough
252 252 if gapsize > revlog._srmingapsize:
253 253 heapq.heappush(gapsheap, (-gapsize, i))
254 254
255 255 prevend = revstart + revlen
256 256
257 257 # Collect the indices of the largest holes until the density is acceptable
258 258 indicesheap = []
259 259 heapq.heapify(indicesheap)
260 260 while gapsheap and density < revlog._srdensitythreshold:
261 261 oppgapsize, gapidx = heapq.heappop(gapsheap)
262 262
263 263 heapq.heappush(indicesheap, gapidx)
264 264
265 265 # the gap sizes are stored as negatives to be sorted decreasingly
266 266 # by the heap
267 267 readdata -= (-oppgapsize)
268 268 if readdata > 0:
269 269 density = chainpayload / float(readdata)
270 270 else:
271 271 density = 1.0
272 272
273 273 # Cut the revs at collected indices
274 274 previdx = 0
275 275 while indicesheap:
276 276 idx = heapq.heappop(indicesheap)
277 277
278 278 chunk = _trimchunk(revlog, revs, previdx, idx)
279 279 if chunk:
280 280 yield chunk
281 281
282 282 previdx = idx
283 283
284 284 chunk = _trimchunk(revlog, revs, previdx)
285 285 if chunk:
286 286 yield chunk
287 287
288 288 @attr.s(slots=True, frozen=True)
289 289 class _deltainfo(object):
290 290 distance = attr.ib()
291 291 deltalen = attr.ib()
292 292 data = attr.ib()
293 293 base = attr.ib()
294 294 chainbase = attr.ib()
295 295 chainlen = attr.ib()
296 296 compresseddeltalen = attr.ib()
297 297
298 298 class _deltacomputer(object):
299 299 def __init__(self, revlog):
300 300 self.revlog = revlog
301 301
302 302 def _getcandidaterevs(self, p1, p2, cachedelta):
303 303 """
304 304 Provides revisions that present an interest to be diffed against,
305 305 grouped by level of easiness.
306 306 """
307 307 revlog = self.revlog
308 308 curr = len(revlog)
309 309 prev = curr - 1
310 310 p1r, p2r = revlog.rev(p1), revlog.rev(p2)
311 311
312 312 # should we try to build a delta?
313 313 if prev != nullrev and revlog.storedeltachains:
314 314 tested = set()
315 315 # This condition is true most of the time when processing
316 316 # changegroup data into a generaldelta repo. The only time it
317 317 # isn't true is if this is the first revision in a delta chain
318 318 # or if ``format.generaldelta=true`` disabled ``lazydeltabase``.
319 319 if cachedelta and revlog._generaldelta and revlog._lazydeltabase:
320 320 # Assume what we received from the server is a good choice
321 321 # build delta will reuse the cache
322 322 yield (cachedelta[0],)
323 323 tested.add(cachedelta[0])
324 324
325 325 if revlog._generaldelta:
326 326 # exclude already lazy tested base if any
327 327 parents = [p for p in (p1r, p2r)
328 328 if p != nullrev and p not in tested]
329 329 if parents and not revlog._aggressivemergedeltas:
330 330 # Pick whichever parent is closer to us (to minimize the
331 331 # chance of having to build a fulltext).
332 332 parents = [max(parents)]
333 333 tested.update(parents)
334 334 yield parents
335 335
336 336 if prev not in tested:
337 337 # other approach failed try against prev to hopefully save us a
338 338 # fulltext.
339 339 yield (prev,)
340 340
341 341 def buildtext(self, revinfo, fh):
342 342 """Builds a fulltext version of a revision
343 343
344 344 revinfo: _revisioninfo instance that contains all needed info
345 345 fh: file handle to either the .i or the .d revlog file,
346 346 depending on whether it is inlined or not
347 347 """
348 348 btext = revinfo.btext
349 349 if btext[0] is not None:
350 350 return btext[0]
351 351
352 352 revlog = self.revlog
353 353 cachedelta = revinfo.cachedelta
354 354 flags = revinfo.flags
355 355 node = revinfo.node
356 356
357 357 baserev = cachedelta[0]
358 358 delta = cachedelta[1]
359 359 # special case deltas which replace entire base; no need to decode
360 360 # base revision. this neatly avoids censored bases, which throw when
361 361 # they're decoded.
362 362 hlen = struct.calcsize(">lll")
363 363 if delta[:hlen] == mdiff.replacediffheader(revlog.rawsize(baserev),
364 364 len(delta) - hlen):
365 365 btext[0] = delta[hlen:]
366 366 else:
367 367 # deltabase is rawtext before changed by flag processors, which is
368 368 # equivalent to non-raw text
369 369 basetext = revlog.revision(baserev, _df=fh, raw=False)
370 370 btext[0] = mdiff.patch(basetext, delta)
371 371
372 372 try:
373 373 res = revlog._processflags(btext[0], flags, 'read', raw=True)
374 374 btext[0], validatehash = res
375 375 if validatehash:
376 376 revlog.checkhash(btext[0], node, p1=revinfo.p1, p2=revinfo.p2)
377 377 if flags & REVIDX_ISCENSORED:
378 378 raise RevlogError(_('node %s is not censored') % node)
379 379 except CensoredNodeError:
380 380 # must pass the censored index flag to add censored revisions
381 381 if not flags & REVIDX_ISCENSORED:
382 382 raise
383 383 return btext[0]
384 384
385 385 def _builddeltadiff(self, base, revinfo, fh):
386 386 revlog = self.revlog
387 387 t = self.buildtext(revinfo, fh)
388 388 if revlog.iscensored(base):
389 389 # deltas based on a censored revision must replace the
390 390 # full content in one patch, so delta works everywhere
391 391 header = mdiff.replacediffheader(revlog.rawsize(base), len(t))
392 392 delta = header + t
393 393 else:
394 394 ptext = revlog.revision(base, _df=fh, raw=True)
395 395 delta = mdiff.textdiff(ptext, t)
396 396
397 397 return delta
398 398
399 399 def _builddeltainfo(self, revinfo, base, fh):
400 400 # can we use the cached delta?
401 401 if revinfo.cachedelta and revinfo.cachedelta[0] == base:
402 402 delta = revinfo.cachedelta[1]
403 403 else:
404 404 delta = self._builddeltadiff(base, revinfo, fh)
405 405 revlog = self.revlog
406 406 header, data = revlog.compress(delta)
407 407 deltalen = len(header) + len(data)
408 408 chainbase = revlog.chainbase(base)
409 409 offset = revlog.end(len(revlog) - 1)
410 410 dist = deltalen + offset - revlog.start(chainbase)
411 411 if revlog._generaldelta:
412 412 deltabase = base
413 413 else:
414 414 deltabase = chainbase
415 415 chainlen, compresseddeltalen = revlog._chaininfo(base)
416 416 chainlen += 1
417 417 compresseddeltalen += deltalen
418 418 return _deltainfo(dist, deltalen, (header, data), deltabase,
419 419 chainbase, chainlen, compresseddeltalen)
420 420
421 421 def finddeltainfo(self, revinfo, fh):
422 422 """Find an acceptable delta against a candidate revision
423 423
424 424 revinfo: information about the revision (instance of _revisioninfo)
425 425 fh: file handle to either the .i or the .d revlog file,
426 426 depending on whether it is inlined or not
427 427
428 428 Returns the first acceptable candidate revision, as ordered by
429 429 _getcandidaterevs
430 430 """
431 431 cachedelta = revinfo.cachedelta
432 432 p1 = revinfo.p1
433 433 p2 = revinfo.p2
434 434 revlog = self.revlog
435 435
436 436 deltainfo = None
437 437 for candidaterevs in self._getcandidaterevs(p1, p2, cachedelta):
438 438 nominateddeltas = []
439 439 for candidaterev in candidaterevs:
440 440 # no delta for rawtext-changing revs (see "candelta" for why)
441 441 if revlog.flags(candidaterev) & REVIDX_RAWTEXT_CHANGING_FLAGS:
442 442 continue
443 443 candidatedelta = self._builddeltainfo(revinfo, candidaterev, fh)
444 444 if revlog._isgooddeltainfo(candidatedelta, revinfo.textlen):
445 445 nominateddeltas.append(candidatedelta)
446 446 if nominateddeltas:
447 447 deltainfo = min(nominateddeltas, key=lambda x: x.deltalen)
448 448 break
449 449
450 450 return deltainfo
451 451
452 452 @attr.s(slots=True, frozen=True)
453 453 class _revisioninfo(object):
454 454 """Information about a revision that allows building its fulltext
455 455 node: expected hash of the revision
456 456 p1, p2: parent revs of the revision
457 457 btext: built text cache consisting of a one-element list
458 458 cachedelta: (baserev, uncompressed_delta) or None
459 459 flags: flags associated to the revision storage
460 460
461 461 One of btext[0] or cachedelta must be set.
462 462 """
463 463 node = attr.ib()
464 464 p1 = attr.ib()
465 465 p2 = attr.ib()
466 466 btext = attr.ib()
467 467 textlen = attr.ib()
468 468 cachedelta = attr.ib()
469 469 flags = attr.ib()
470 470
471 471 # index v0:
472 472 # 4 bytes: offset
473 473 # 4 bytes: compressed length
474 474 # 4 bytes: base rev
475 475 # 4 bytes: link rev
476 476 # 20 bytes: parent 1 nodeid
477 477 # 20 bytes: parent 2 nodeid
478 478 # 20 bytes: nodeid
479 479 indexformatv0 = struct.Struct(">4l20s20s20s")
480 480 indexformatv0_pack = indexformatv0.pack
481 481 indexformatv0_unpack = indexformatv0.unpack
482 482
483 483 class revlogoldio(object):
484 484 def __init__(self):
485 485 self.size = indexformatv0.size
486 486
487 487 def parseindex(self, data, inline):
488 488 s = self.size
489 489 index = []
490 490 nodemap = {nullid: nullrev}
491 491 n = off = 0
492 492 l = len(data)
493 493 while off + s <= l:
494 494 cur = data[off:off + s]
495 495 off += s
496 496 e = indexformatv0_unpack(cur)
497 497 # transform to revlogv1 format
498 498 e2 = (offset_type(e[0], 0), e[1], -1, e[2], e[3],
499 499 nodemap.get(e[4], nullrev), nodemap.get(e[5], nullrev), e[6])
500 500 index.append(e2)
501 501 nodemap[e[6]] = n
502 502 n += 1
503 503
504 504 # add the magic null revision at -1
505 505 index.append((0, 0, 0, -1, -1, -1, -1, nullid))
506 506
507 507 return index, nodemap, None
508 508
509 509 def packentry(self, entry, node, version, rev):
510 510 if gettype(entry[0]):
511 511 raise RevlogError(_('index entry flags need revlog version 1'))
512 512 e2 = (getoffset(entry[0]), entry[1], entry[3], entry[4],
513 513 node(entry[5]), node(entry[6]), entry[7])
514 514 return indexformatv0_pack(*e2)
515 515
516 516 # index ng:
517 517 # 6 bytes: offset
518 518 # 2 bytes: flags
519 519 # 4 bytes: compressed length
520 520 # 4 bytes: uncompressed length
521 521 # 4 bytes: base rev
522 522 # 4 bytes: link rev
523 523 # 4 bytes: parent 1 rev
524 524 # 4 bytes: parent 2 rev
525 525 # 32 bytes: nodeid
526 526 indexformatng = struct.Struct(">Qiiiiii20s12x")
527 527 indexformatng_pack = indexformatng.pack
528 528 versionformat = struct.Struct(">I")
529 529 versionformat_pack = versionformat.pack
530 530 versionformat_unpack = versionformat.unpack
531 531
532 532 # corresponds to uncompressed length of indexformatng (2 gigs, 4-byte
533 533 # signed integer)
534 534 _maxentrysize = 0x7fffffff
535 535
536 536 class revlogio(object):
537 537 def __init__(self):
538 538 self.size = indexformatng.size
539 539
540 540 def parseindex(self, data, inline):
541 541 # call the C implementation to parse the index data
542 542 index, cache = parsers.parse_index2(data, inline)
543 543 return index, getattr(index, 'nodemap', None), cache
544 544
545 545 def packentry(self, entry, node, version, rev):
546 546 p = indexformatng_pack(*entry)
547 547 if rev == 0:
548 548 p = versionformat_pack(version) + p[4:]
549 549 return p
550 550
551 551 class revlog(object):
552 552 """
553 553 the underlying revision storage object
554 554
555 555 A revlog consists of two parts, an index and the revision data.
556 556
557 557 The index is a file with a fixed record size containing
558 558 information on each revision, including its nodeid (hash), the
559 559 nodeids of its parents, the position and offset of its data within
560 560 the data file, and the revision it's based on. Finally, each entry
561 561 contains a linkrev entry that can serve as a pointer to external
562 562 data.
563 563
564 564 The revision data itself is a linear collection of data chunks.
565 565 Each chunk represents a revision and is usually represented as a
566 566 delta against the previous chunk. To bound lookup time, runs of
567 567 deltas are limited to about 2 times the length of the original
568 568 version data. This makes retrieval of a version proportional to
569 569 its size, or O(1) relative to the number of revisions.
570 570
571 571 Both pieces of the revlog are written to in an append-only
572 572 fashion, which means we never need to rewrite a file to insert or
573 573 remove data, and can use some simple techniques to avoid the need
574 574 for locking while reading.
575 575
576 576 If checkambig, indexfile is opened with checkambig=True at
577 577 writing, to avoid file stat ambiguity.
578 578
579 579 If mmaplargeindex is True, and an mmapindexthreshold is set, the
580 580 index will be mmapped rather than read if it is larger than the
581 581 configured threshold.
582 582
583 583 If censorable is True, the revlog can have censored revisions.
584 584 """
585 585 def __init__(self, opener, indexfile, datafile=None, checkambig=False,
586 586 mmaplargeindex=False, censorable=False):
587 587 """
588 588 create a revlog object
589 589
590 590 opener is a function that abstracts the file opening operation
591 591 and can be used to implement COW semantics or the like.
592 592 """
593 593 self.indexfile = indexfile
594 594 self.datafile = datafile or (indexfile[:-2] + ".d")
595 595 self.opener = opener
596 596 # When True, indexfile is opened with checkambig=True at writing, to
597 597 # avoid file stat ambiguity.
598 598 self._checkambig = checkambig
599 599 self._censorable = censorable
600 600 # 3-tuple of (node, rev, text) for a raw revision.
601 601 self._cache = None
602 602 # Maps rev to chain base rev.
603 603 self._chainbasecache = util.lrucachedict(100)
604 604 # 2-tuple of (offset, data) of raw data from the revlog at an offset.
605 605 self._chunkcache = (0, '')
606 606 # How much data to read and cache into the raw revlog data cache.
607 607 self._chunkcachesize = 65536
608 608 self._maxchainlen = None
609 609 self._aggressivemergedeltas = False
610 610 self.index = []
611 611 # Mapping of partial identifiers to full nodes.
612 612 self._pcache = {}
613 613 # Mapping of revision integer to full node.
614 614 self._nodecache = {nullid: nullrev}
615 615 self._nodepos = None
616 616 self._compengine = 'zlib'
617 617 self._maxdeltachainspan = -1
618 618 self._withsparseread = False
619 619 self._srdensitythreshold = 0.25
620 620 self._srmingapsize = 262144
621 621
622 622 mmapindexthreshold = None
623 623 v = REVLOG_DEFAULT_VERSION
624 624 opts = getattr(opener, 'options', None)
625 625 if opts is not None:
626 626 if 'revlogv2' in opts:
627 627 # version 2 revlogs always use generaldelta.
628 628 v = REVLOGV2 | FLAG_GENERALDELTA | FLAG_INLINE_DATA
629 629 elif 'revlogv1' in opts:
630 630 if 'generaldelta' in opts:
631 631 v |= FLAG_GENERALDELTA
632 632 else:
633 633 v = 0
634 634 if 'chunkcachesize' in opts:
635 635 self._chunkcachesize = opts['chunkcachesize']
636 636 if 'maxchainlen' in opts:
637 637 self._maxchainlen = opts['maxchainlen']
638 638 if 'aggressivemergedeltas' in opts:
639 639 self._aggressivemergedeltas = opts['aggressivemergedeltas']
640 640 self._lazydeltabase = bool(opts.get('lazydeltabase', False))
641 641 if 'compengine' in opts:
642 642 self._compengine = opts['compengine']
643 643 if 'maxdeltachainspan' in opts:
644 644 self._maxdeltachainspan = opts['maxdeltachainspan']
645 645 if mmaplargeindex and 'mmapindexthreshold' in opts:
646 646 mmapindexthreshold = opts['mmapindexthreshold']
647 647 self._withsparseread = bool(opts.get('with-sparse-read', False))
648 648 if 'sparse-read-density-threshold' in opts:
649 649 self._srdensitythreshold = opts['sparse-read-density-threshold']
650 650 if 'sparse-read-min-gap-size' in opts:
651 651 self._srmingapsize = opts['sparse-read-min-gap-size']
652 652
653 653 if self._chunkcachesize <= 0:
654 654 raise RevlogError(_('revlog chunk cache size %r is not greater '
655 655 'than 0') % self._chunkcachesize)
656 656 elif self._chunkcachesize & (self._chunkcachesize - 1):
657 657 raise RevlogError(_('revlog chunk cache size %r is not a power '
658 658 'of 2') % self._chunkcachesize)
659 659
660 660 indexdata = ''
661 661 self._initempty = True
662 662 try:
663 663 with self._indexfp() as f:
664 664 if (mmapindexthreshold is not None and
665 665 self.opener.fstat(f).st_size >= mmapindexthreshold):
666 666 indexdata = util.buffer(util.mmapread(f))
667 667 else:
668 668 indexdata = f.read()
669 669 if len(indexdata) > 0:
670 670 v = versionformat_unpack(indexdata[:4])[0]
671 671 self._initempty = False
672 672 except IOError as inst:
673 673 if inst.errno != errno.ENOENT:
674 674 raise
675 675
676 676 self.version = v
677 677 self._inline = v & FLAG_INLINE_DATA
678 678 self._generaldelta = v & FLAG_GENERALDELTA
679 679 flags = v & ~0xFFFF
680 680 fmt = v & 0xFFFF
681 681 if fmt == REVLOGV0:
682 682 if flags:
683 683 raise RevlogError(_('unknown flags (%#04x) in version %d '
684 684 'revlog %s') %
685 685 (flags >> 16, fmt, self.indexfile))
686 686 elif fmt == REVLOGV1:
687 687 if flags & ~REVLOGV1_FLAGS:
688 688 raise RevlogError(_('unknown flags (%#04x) in version %d '
689 689 'revlog %s') %
690 690 (flags >> 16, fmt, self.indexfile))
691 691 elif fmt == REVLOGV2:
692 692 if flags & ~REVLOGV2_FLAGS:
693 693 raise RevlogError(_('unknown flags (%#04x) in version %d '
694 694 'revlog %s') %
695 695 (flags >> 16, fmt, self.indexfile))
696 696 else:
697 697 raise RevlogError(_('unknown version (%d) in revlog %s') %
698 698 (fmt, self.indexfile))
699 699
700 700 self.storedeltachains = True
701 701
702 702 self._io = revlogio()
703 703 if self.version == REVLOGV0:
704 704 self._io = revlogoldio()
705 705 try:
706 706 d = self._io.parseindex(indexdata, self._inline)
707 707 except (ValueError, IndexError):
708 708 raise RevlogError(_("index %s is corrupted") % (self.indexfile))
709 709 self.index, nodemap, self._chunkcache = d
710 710 if nodemap is not None:
711 711 self.nodemap = self._nodecache = nodemap
712 712 if not self._chunkcache:
713 713 self._chunkclear()
714 714 # revnum -> (chain-length, sum-delta-length)
715 715 self._chaininfocache = {}
716 716 # revlog header -> revlog compressor
717 717 self._decompressors = {}
718 718
719 719 @util.propertycache
720 720 def _compressor(self):
721 721 return util.compengines[self._compengine].revlogcompressor()
722 722
723 723 def _indexfp(self, mode='r'):
724 724 """file object for the revlog's index file"""
725 725 args = {r'mode': mode}
726 726 if mode != 'r':
727 727 args[r'checkambig'] = self._checkambig
728 728 if mode == 'w':
729 729 args[r'atomictemp'] = True
730 730 return self.opener(self.indexfile, **args)
731 731
732 732 def _datafp(self, mode='r'):
733 733 """file object for the revlog's data file"""
734 734 return self.opener(self.datafile, mode=mode)
735 735
736 736 @contextlib.contextmanager
737 737 def _datareadfp(self, existingfp=None):
738 738 """file object suitable to read data"""
739 739 if existingfp is not None:
740 740 yield existingfp
741 741 else:
742 742 if self._inline:
743 743 func = self._indexfp
744 744 else:
745 745 func = self._datafp
746 746 with func() as fp:
747 747 yield fp
748 748
749 749 def tip(self):
750 750 return self.node(len(self.index) - 2)
751 751 def __contains__(self, rev):
752 752 return 0 <= rev < len(self)
753 753 def __len__(self):
754 754 return len(self.index) - 1
755 755 def __iter__(self):
756 756 return iter(xrange(len(self)))
757 757 def revs(self, start=0, stop=None):
758 758 """iterate over all rev in this revlog (from start to stop)"""
759 759 step = 1
760 760 if stop is not None:
761 761 if start > stop:
762 762 step = -1
763 763 stop += step
764 764 else:
765 765 stop = len(self)
766 766 return xrange(start, stop, step)
767 767
768 768 @util.propertycache
769 769 def nodemap(self):
770 770 self.rev(self.node(0))
771 771 return self._nodecache
772 772
773 773 def hasnode(self, node):
774 774 try:
775 775 self.rev(node)
776 776 return True
777 777 except KeyError:
778 778 return False
779 779
780 780 def candelta(self, baserev, rev):
781 781 """whether two revisions (baserev, rev) can be delta-ed or not"""
782 782 # Disable delta if either rev requires a content-changing flag
783 783 # processor (ex. LFS). This is because such flag processor can alter
784 784 # the rawtext content that the delta will be based on, and two clients
785 785 # could have a same revlog node with different flags (i.e. different
786 786 # rawtext contents) and the delta could be incompatible.
787 787 if ((self.flags(baserev) & REVIDX_RAWTEXT_CHANGING_FLAGS)
788 788 or (self.flags(rev) & REVIDX_RAWTEXT_CHANGING_FLAGS)):
789 789 return False
790 790 return True
791 791
792 792 def clearcaches(self):
793 793 self._cache = None
794 794 self._chainbasecache.clear()
795 795 self._chunkcache = (0, '')
796 796 self._pcache = {}
797 797
798 798 try:
799 799 self._nodecache.clearcaches()
800 800 except AttributeError:
801 801 self._nodecache = {nullid: nullrev}
802 802 self._nodepos = None
803 803
804 804 def rev(self, node):
805 805 try:
806 806 return self._nodecache[node]
807 807 except TypeError:
808 808 raise
809 809 except RevlogError:
810 810 # parsers.c radix tree lookup failed
811 811 if node == wdirid or node in wdirfilenodeids:
812 812 raise error.WdirUnsupported
813 813 raise LookupError(node, self.indexfile, _('no node'))
814 814 except KeyError:
815 815 # pure python cache lookup failed
816 816 n = self._nodecache
817 817 i = self.index
818 818 p = self._nodepos
819 819 if p is None:
820 820 p = len(i) - 2
821 821 else:
822 822 assert p < len(i)
823 823 for r in xrange(p, -1, -1):
824 824 v = i[r][7]
825 825 n[v] = r
826 826 if v == node:
827 827 self._nodepos = r - 1
828 828 return r
829 829 if node == wdirid or node in wdirfilenodeids:
830 830 raise error.WdirUnsupported
831 831 raise LookupError(node, self.indexfile, _('no node'))
832 832
833 833 # Accessors for index entries.
834 834
835 835 # First tuple entry is 8 bytes. First 6 bytes are offset. Last 2 bytes
836 836 # are flags.
837 837 def start(self, rev):
838 838 return int(self.index[rev][0] >> 16)
839 839
840 840 def flags(self, rev):
841 841 return self.index[rev][0] & 0xFFFF
842 842
843 843 def length(self, rev):
844 844 return self.index[rev][1]
845 845
846 846 def rawsize(self, rev):
847 847 """return the length of the uncompressed text for a given revision"""
848 848 l = self.index[rev][2]
849 849 if l >= 0:
850 850 return l
851 851
852 852 t = self.revision(rev, raw=True)
853 853 return len(t)
854 854
855 855 def size(self, rev):
856 856 """length of non-raw text (processed by a "read" flag processor)"""
857 857 # fast path: if no "read" flag processor could change the content,
858 858 # size is rawsize. note: ELLIPSIS is known to not change the content.
859 859 flags = self.flags(rev)
860 860 if flags & (REVIDX_KNOWN_FLAGS ^ REVIDX_ELLIPSIS) == 0:
861 861 return self.rawsize(rev)
862 862
863 863 return len(self.revision(rev, raw=False))
864 864
865 865 def chainbase(self, rev):
866 866 base = self._chainbasecache.get(rev)
867 867 if base is not None:
868 868 return base
869 869
870 870 index = self.index
871 871 base = index[rev][3]
872 872 while base != rev:
873 873 rev = base
874 874 base = index[rev][3]
875 875
876 876 self._chainbasecache[rev] = base
877 877 return base
878 878
879 879 def linkrev(self, rev):
880 880 return self.index[rev][4]
881 881
882 882 def parentrevs(self, rev):
883 883 try:
884 884 entry = self.index[rev]
885 885 except IndexError:
886 886 if rev == wdirrev:
887 887 raise error.WdirUnsupported
888 888 raise
889 889
890 890 return entry[5], entry[6]
891 891
892 892 def node(self, rev):
893 893 try:
894 894 return self.index[rev][7]
895 895 except IndexError:
896 896 if rev == wdirrev:
897 897 raise error.WdirUnsupported
898 898 raise
899 899
900 900 # Derived from index values.
901 901
902 902 def end(self, rev):
903 903 return self.start(rev) + self.length(rev)
904 904
905 905 def parents(self, node):
906 906 i = self.index
907 907 d = i[self.rev(node)]
908 908 return i[d[5]][7], i[d[6]][7] # map revisions to nodes inline
909 909
910 910 def chainlen(self, rev):
911 911 return self._chaininfo(rev)[0]
912 912
913 913 def _chaininfo(self, rev):
914 914 chaininfocache = self._chaininfocache
915 915 if rev in chaininfocache:
916 916 return chaininfocache[rev]
917 917 index = self.index
918 918 generaldelta = self._generaldelta
919 919 iterrev = rev
920 920 e = index[iterrev]
921 921 clen = 0
922 922 compresseddeltalen = 0
923 923 while iterrev != e[3]:
924 924 clen += 1
925 925 compresseddeltalen += e[1]
926 926 if generaldelta:
927 927 iterrev = e[3]
928 928 else:
929 929 iterrev -= 1
930 930 if iterrev in chaininfocache:
931 931 t = chaininfocache[iterrev]
932 932 clen += t[0]
933 933 compresseddeltalen += t[1]
934 934 break
935 935 e = index[iterrev]
936 936 else:
937 937 # Add text length of base since decompressing that also takes
938 938 # work. For cache hits the length is already included.
939 939 compresseddeltalen += e[1]
940 940 r = (clen, compresseddeltalen)
941 941 chaininfocache[rev] = r
942 942 return r
943 943
944 944 def _deltachain(self, rev, stoprev=None):
945 945 """Obtain the delta chain for a revision.
946 946
947 947 ``stoprev`` specifies a revision to stop at. If not specified, we
948 948 stop at the base of the chain.
949 949
950 950 Returns a 2-tuple of (chain, stopped) where ``chain`` is a list of
951 951 revs in ascending order and ``stopped`` is a bool indicating whether
952 952 ``stoprev`` was hit.
953 953 """
954 954 # Try C implementation.
955 955 try:
956 956 return self.index.deltachain(rev, stoprev, self._generaldelta)
957 957 except AttributeError:
958 958 pass
959 959
960 960 chain = []
961 961
962 962 # Alias to prevent attribute lookup in tight loop.
963 963 index = self.index
964 964 generaldelta = self._generaldelta
965 965
966 966 iterrev = rev
967 967 e = index[iterrev]
968 968 while iterrev != e[3] and iterrev != stoprev:
969 969 chain.append(iterrev)
970 970 if generaldelta:
971 971 iterrev = e[3]
972 972 else:
973 973 iterrev -= 1
974 974 e = index[iterrev]
975 975
976 976 if iterrev == stoprev:
977 977 stopped = True
978 978 else:
979 979 chain.append(iterrev)
980 980 stopped = False
981 981
982 982 chain.reverse()
983 983 return chain, stopped
984 984
985 985 def ancestors(self, revs, stoprev=0, inclusive=False):
986 986 """Generate the ancestors of 'revs' in reverse topological order.
987 987 Does not generate revs lower than stoprev.
988 988
989 989 See the documentation for ancestor.lazyancestors for more details."""
990 990
991 991 return ancestor.lazyancestors(self.parentrevs, revs, stoprev=stoprev,
992 992 inclusive=inclusive)
993 993
994 994 def descendants(self, revs):
995 995 """Generate the descendants of 'revs' in revision order.
996 996
997 997 Yield a sequence of revision numbers starting with a child of
998 998 some rev in revs, i.e., each revision is *not* considered a
999 999 descendant of itself. Results are ordered by revision number (a
1000 1000 topological sort)."""
1001 1001 first = min(revs)
1002 1002 if first == nullrev:
1003 1003 for i in self:
1004 1004 yield i
1005 1005 return
1006 1006
1007 1007 seen = set(revs)
1008 1008 for i in self.revs(start=first + 1):
1009 1009 for x in self.parentrevs(i):
1010 1010 if x != nullrev and x in seen:
1011 1011 seen.add(i)
1012 1012 yield i
1013 1013 break
1014 1014
1015 1015 def findcommonmissing(self, common=None, heads=None):
1016 1016 """Return a tuple of the ancestors of common and the ancestors of heads
1017 1017 that are not ancestors of common. In revset terminology, we return the
1018 1018 tuple:
1019 1019
1020 1020 ::common, (::heads) - (::common)
1021 1021
1022 1022 The list is sorted by revision number, meaning it is
1023 1023 topologically sorted.
1024 1024
1025 1025 'heads' and 'common' are both lists of node IDs. If heads is
1026 1026 not supplied, uses all of the revlog's heads. If common is not
1027 1027 supplied, uses nullid."""
1028 1028 if common is None:
1029 1029 common = [nullid]
1030 1030 if heads is None:
1031 1031 heads = self.heads()
1032 1032
1033 1033 common = [self.rev(n) for n in common]
1034 1034 heads = [self.rev(n) for n in heads]
1035 1035
1036 1036 # we want the ancestors, but inclusive
1037 1037 class lazyset(object):
1038 1038 def __init__(self, lazyvalues):
1039 1039 self.addedvalues = set()
1040 1040 self.lazyvalues = lazyvalues
1041 1041
1042 1042 def __contains__(self, value):
1043 1043 return value in self.addedvalues or value in self.lazyvalues
1044 1044
1045 1045 def __iter__(self):
1046 1046 added = self.addedvalues
1047 1047 for r in added:
1048 1048 yield r
1049 1049 for r in self.lazyvalues:
1050 1050 if not r in added:
1051 1051 yield r
1052 1052
1053 1053 def add(self, value):
1054 1054 self.addedvalues.add(value)
1055 1055
1056 1056 def update(self, values):
1057 1057 self.addedvalues.update(values)
1058 1058
1059 1059 has = lazyset(self.ancestors(common))
1060 1060 has.add(nullrev)
1061 1061 has.update(common)
1062 1062
1063 1063 # take all ancestors from heads that aren't in has
1064 1064 missing = set()
1065 1065 visit = collections.deque(r for r in heads if r not in has)
1066 1066 while visit:
1067 1067 r = visit.popleft()
1068 1068 if r in missing:
1069 1069 continue
1070 1070 else:
1071 1071 missing.add(r)
1072 1072 for p in self.parentrevs(r):
1073 1073 if p not in has:
1074 1074 visit.append(p)
1075 1075 missing = list(missing)
1076 1076 missing.sort()
1077 1077 return has, [self.node(miss) for miss in missing]
1078 1078
1079 1079 def incrementalmissingrevs(self, common=None):
1080 1080 """Return an object that can be used to incrementally compute the
1081 1081 revision numbers of the ancestors of arbitrary sets that are not
1082 1082 ancestors of common. This is an ancestor.incrementalmissingancestors
1083 1083 object.
1084 1084
1085 1085 'common' is a list of revision numbers. If common is not supplied, uses
1086 1086 nullrev.
1087 1087 """
1088 1088 if common is None:
1089 1089 common = [nullrev]
1090 1090
1091 1091 return ancestor.incrementalmissingancestors(self.parentrevs, common)
1092 1092
1093 1093 def findmissingrevs(self, common=None, heads=None):
1094 1094 """Return the revision numbers of the ancestors of heads that
1095 1095 are not ancestors of common.
1096 1096
1097 1097 More specifically, return a list of revision numbers corresponding to
1098 1098 nodes N such that every N satisfies the following constraints:
1099 1099
1100 1100 1. N is an ancestor of some node in 'heads'
1101 1101 2. N is not an ancestor of any node in 'common'
1102 1102
1103 1103 The list is sorted by revision number, meaning it is
1104 1104 topologically sorted.
1105 1105
1106 1106 'heads' and 'common' are both lists of revision numbers. If heads is
1107 1107 not supplied, uses all of the revlog's heads. If common is not
1108 1108 supplied, uses nullid."""
1109 1109 if common is None:
1110 1110 common = [nullrev]
1111 1111 if heads is None:
1112 1112 heads = self.headrevs()
1113 1113
1114 1114 inc = self.incrementalmissingrevs(common=common)
1115 1115 return inc.missingancestors(heads)
1116 1116
1117 1117 def findmissing(self, common=None, heads=None):
1118 1118 """Return the ancestors of heads that are not ancestors of common.
1119 1119
1120 1120 More specifically, return a list of nodes N such that every N
1121 1121 satisfies the following constraints:
1122 1122
1123 1123 1. N is an ancestor of some node in 'heads'
1124 1124 2. N is not an ancestor of any node in 'common'
1125 1125
1126 1126 The list is sorted by revision number, meaning it is
1127 1127 topologically sorted.
1128 1128
1129 1129 'heads' and 'common' are both lists of node IDs. If heads is
1130 1130 not supplied, uses all of the revlog's heads. If common is not
1131 1131 supplied, uses nullid."""
1132 1132 if common is None:
1133 1133 common = [nullid]
1134 1134 if heads is None:
1135 1135 heads = self.heads()
1136 1136
1137 1137 common = [self.rev(n) for n in common]
1138 1138 heads = [self.rev(n) for n in heads]
1139 1139
1140 1140 inc = self.incrementalmissingrevs(common=common)
1141 1141 return [self.node(r) for r in inc.missingancestors(heads)]
1142 1142
1143 1143 def nodesbetween(self, roots=None, heads=None):
1144 1144 """Return a topological path from 'roots' to 'heads'.
1145 1145
1146 1146 Return a tuple (nodes, outroots, outheads) where 'nodes' is a
1147 1147 topologically sorted list of all nodes N that satisfy both of
1148 1148 these constraints:
1149 1149
1150 1150 1. N is a descendant of some node in 'roots'
1151 1151 2. N is an ancestor of some node in 'heads'
1152 1152
1153 1153 Every node is considered to be both a descendant and an ancestor
1154 1154 of itself, so every reachable node in 'roots' and 'heads' will be
1155 1155 included in 'nodes'.
1156 1156
1157 1157 'outroots' is the list of reachable nodes in 'roots', i.e., the
1158 1158 subset of 'roots' that is returned in 'nodes'. Likewise,
1159 1159 'outheads' is the subset of 'heads' that is also in 'nodes'.
1160 1160
1161 1161 'roots' and 'heads' are both lists of node IDs. If 'roots' is
1162 1162 unspecified, uses nullid as the only root. If 'heads' is
1163 1163 unspecified, uses list of all of the revlog's heads."""
1164 1164 nonodes = ([], [], [])
1165 1165 if roots is not None:
1166 1166 roots = list(roots)
1167 1167 if not roots:
1168 1168 return nonodes
1169 1169 lowestrev = min([self.rev(n) for n in roots])
1170 1170 else:
1171 1171 roots = [nullid] # Everybody's a descendant of nullid
1172 1172 lowestrev = nullrev
1173 1173 if (lowestrev == nullrev) and (heads is None):
1174 1174 # We want _all_ the nodes!
1175 1175 return ([self.node(r) for r in self], [nullid], list(self.heads()))
1176 1176 if heads is None:
1177 1177 # All nodes are ancestors, so the latest ancestor is the last
1178 1178 # node.
1179 1179 highestrev = len(self) - 1
1180 1180 # Set ancestors to None to signal that every node is an ancestor.
1181 1181 ancestors = None
1182 1182 # Set heads to an empty dictionary for later discovery of heads
1183 1183 heads = {}
1184 1184 else:
1185 1185 heads = list(heads)
1186 1186 if not heads:
1187 1187 return nonodes
1188 1188 ancestors = set()
1189 1189 # Turn heads into a dictionary so we can remove 'fake' heads.
1190 1190 # Also, later we will be using it to filter out the heads we can't
1191 1191 # find from roots.
1192 1192 heads = dict.fromkeys(heads, False)
1193 1193 # Start at the top and keep marking parents until we're done.
1194 1194 nodestotag = set(heads)
1195 1195 # Remember where the top was so we can use it as a limit later.
1196 1196 highestrev = max([self.rev(n) for n in nodestotag])
1197 1197 while nodestotag:
1198 1198 # grab a node to tag
1199 1199 n = nodestotag.pop()
1200 1200 # Never tag nullid
1201 1201 if n == nullid:
1202 1202 continue
1203 1203 # A node's revision number represents its place in a
1204 1204 # topologically sorted list of nodes.
1205 1205 r = self.rev(n)
1206 1206 if r >= lowestrev:
1207 1207 if n not in ancestors:
1208 1208 # If we are possibly a descendant of one of the roots
1209 1209 # and we haven't already been marked as an ancestor
1210 1210 ancestors.add(n) # Mark as ancestor
1211 1211 # Add non-nullid parents to list of nodes to tag.
1212 1212 nodestotag.update([p for p in self.parents(n) if
1213 1213 p != nullid])
1214 1214 elif n in heads: # We've seen it before, is it a fake head?
1215 1215 # So it is, real heads should not be the ancestors of
1216 1216 # any other heads.
1217 1217 heads.pop(n)
1218 1218 if not ancestors:
1219 1219 return nonodes
1220 1220 # Now that we have our set of ancestors, we want to remove any
1221 1221 # roots that are not ancestors.
1222 1222
1223 1223 # If one of the roots was nullid, everything is included anyway.
1224 1224 if lowestrev > nullrev:
1225 1225 # But, since we weren't, let's recompute the lowest rev to not
1226 1226 # include roots that aren't ancestors.
1227 1227
1228 1228 # Filter out roots that aren't ancestors of heads
1229 1229 roots = [root for root in roots if root in ancestors]
1230 1230 # Recompute the lowest revision
1231 1231 if roots:
1232 1232 lowestrev = min([self.rev(root) for root in roots])
1233 1233 else:
1234 1234 # No more roots? Return empty list
1235 1235 return nonodes
1236 1236 else:
1237 1237 # We are descending from nullid, and don't need to care about
1238 1238 # any other roots.
1239 1239 lowestrev = nullrev
1240 1240 roots = [nullid]
1241 1241 # Transform our roots list into a set.
1242 1242 descendants = set(roots)
1243 1243 # Also, keep the original roots so we can filter out roots that aren't
1244 1244 # 'real' roots (i.e. are descended from other roots).
1245 1245 roots = descendants.copy()
1246 1246 # Our topologically sorted list of output nodes.
1247 1247 orderedout = []
1248 1248 # Don't start at nullid since we don't want nullid in our output list,
1249 1249 # and if nullid shows up in descendants, empty parents will look like
1250 1250 # they're descendants.
1251 1251 for r in self.revs(start=max(lowestrev, 0), stop=highestrev + 1):
1252 1252 n = self.node(r)
1253 1253 isdescendant = False
1254 1254 if lowestrev == nullrev: # Everybody is a descendant of nullid
1255 1255 isdescendant = True
1256 1256 elif n in descendants:
1257 1257 # n is already a descendant
1258 1258 isdescendant = True
1259 1259 # This check only needs to be done here because all the roots
1260 1260 # will start being marked is descendants before the loop.
1261 1261 if n in roots:
1262 1262 # If n was a root, check if it's a 'real' root.
1263 1263 p = tuple(self.parents(n))
1264 1264 # If any of its parents are descendants, it's not a root.
1265 1265 if (p[0] in descendants) or (p[1] in descendants):
1266 1266 roots.remove(n)
1267 1267 else:
1268 1268 p = tuple(self.parents(n))
1269 1269 # A node is a descendant if either of its parents are
1270 1270 # descendants. (We seeded the dependents list with the roots
1271 1271 # up there, remember?)
1272 1272 if (p[0] in descendants) or (p[1] in descendants):
1273 1273 descendants.add(n)
1274 1274 isdescendant = True
1275 1275 if isdescendant and ((ancestors is None) or (n in ancestors)):
1276 1276 # Only include nodes that are both descendants and ancestors.
1277 1277 orderedout.append(n)
1278 1278 if (ancestors is not None) and (n in heads):
1279 1279 # We're trying to figure out which heads are reachable
1280 1280 # from roots.
1281 1281 # Mark this head as having been reached
1282 1282 heads[n] = True
1283 1283 elif ancestors is None:
1284 1284 # Otherwise, we're trying to discover the heads.
1285 1285 # Assume this is a head because if it isn't, the next step
1286 1286 # will eventually remove it.
1287 1287 heads[n] = True
1288 1288 # But, obviously its parents aren't.
1289 1289 for p in self.parents(n):
1290 1290 heads.pop(p, None)
1291 1291 heads = [head for head, flag in heads.iteritems() if flag]
1292 1292 roots = list(roots)
1293 1293 assert orderedout
1294 1294 assert roots
1295 1295 assert heads
1296 1296 return (orderedout, roots, heads)
1297 1297
1298 1298 def headrevs(self):
1299 1299 try:
1300 1300 return self.index.headrevs()
1301 1301 except AttributeError:
1302 1302 return self._headrevs()
1303 1303
1304 1304 def computephases(self, roots):
1305 1305 return self.index.computephasesmapsets(roots)
1306 1306
1307 1307 def _headrevs(self):
1308 1308 count = len(self)
1309 1309 if not count:
1310 1310 return [nullrev]
1311 1311 # we won't iter over filtered rev so nobody is a head at start
1312 1312 ishead = [0] * (count + 1)
1313 1313 index = self.index
1314 1314 for r in self:
1315 1315 ishead[r] = 1 # I may be an head
1316 1316 e = index[r]
1317 1317 ishead[e[5]] = ishead[e[6]] = 0 # my parent are not
1318 1318 return [r for r, val in enumerate(ishead) if val]
1319 1319
1320 1320 def heads(self, start=None, stop=None):
1321 1321 """return the list of all nodes that have no children
1322 1322
1323 1323 if start is specified, only heads that are descendants of
1324 1324 start will be returned
1325 1325 if stop is specified, it will consider all the revs from stop
1326 1326 as if they had no children
1327 1327 """
1328 1328 if start is None and stop is None:
1329 1329 if not len(self):
1330 1330 return [nullid]
1331 1331 return [self.node(r) for r in self.headrevs()]
1332 1332
1333 1333 if start is None:
1334 1334 start = nullid
1335 1335 if stop is None:
1336 1336 stop = []
1337 1337 stoprevs = set([self.rev(n) for n in stop])
1338 1338 startrev = self.rev(start)
1339 1339 reachable = {startrev}
1340 1340 heads = {startrev}
1341 1341
1342 1342 parentrevs = self.parentrevs
1343 1343 for r in self.revs(start=startrev + 1):
1344 1344 for p in parentrevs(r):
1345 1345 if p in reachable:
1346 1346 if r not in stoprevs:
1347 1347 reachable.add(r)
1348 1348 heads.add(r)
1349 1349 if p in heads and p not in stoprevs:
1350 1350 heads.remove(p)
1351 1351
1352 1352 return [self.node(r) for r in heads]
1353 1353
1354 1354 def children(self, node):
1355 1355 """find the children of a given node"""
1356 1356 c = []
1357 1357 p = self.rev(node)
1358 1358 for r in self.revs(start=p + 1):
1359 1359 prevs = [pr for pr in self.parentrevs(r) if pr != nullrev]
1360 1360 if prevs:
1361 1361 for pr in prevs:
1362 1362 if pr == p:
1363 1363 c.append(self.node(r))
1364 1364 elif p == nullrev:
1365 1365 c.append(self.node(r))
1366 1366 return c
1367 1367
1368 1368 def descendant(self, start, end):
1369 1369 if start == nullrev:
1370 1370 return True
1371 1371 for i in self.descendants([start]):
1372 1372 if i == end:
1373 1373 return True
1374 1374 elif i > end:
1375 1375 break
1376 1376 return False
1377 1377
1378 1378 def commonancestorsheads(self, a, b):
1379 1379 """calculate all the heads of the common ancestors of nodes a and b"""
1380 1380 a, b = self.rev(a), self.rev(b)
1381 1381 try:
1382 1382 ancs = self.index.commonancestorsheads(a, b)
1383 1383 except (AttributeError, OverflowError): # C implementation failed
1384 1384 ancs = ancestor.commonancestorsheads(self.parentrevs, a, b)
1385 1385 return pycompat.maplist(self.node, ancs)
1386 1386
1387 1387 def isancestor(self, a, b):
1388 1388 """return True if node a is an ancestor of node b
1389 1389
1390 1390 The implementation of this is trivial but the use of
1391 1391 commonancestorsheads is not."""
1392 1392 return a in self.commonancestorsheads(a, b)
1393 1393
1394 1394 def ancestor(self, a, b):
1395 1395 """calculate the "best" common ancestor of nodes a and b"""
1396 1396
1397 1397 a, b = self.rev(a), self.rev(b)
1398 1398 try:
1399 1399 ancs = self.index.ancestors(a, b)
1400 1400 except (AttributeError, OverflowError):
1401 1401 ancs = ancestor.ancestors(self.parentrevs, a, b)
1402 1402 if ancs:
1403 1403 # choose a consistent winner when there's a tie
1404 1404 return min(map(self.node, ancs))
1405 1405 return nullid
1406 1406
1407 1407 def _match(self, id):
1408 1408 if isinstance(id, int):
1409 1409 # rev
1410 1410 return self.node(id)
1411 1411 if len(id) == 20:
1412 1412 # possibly a binary node
1413 1413 # odds of a binary node being all hex in ASCII are 1 in 10**25
1414 1414 try:
1415 1415 node = id
1416 1416 self.rev(node) # quick search the index
1417 1417 return node
1418 1418 except LookupError:
1419 1419 pass # may be partial hex id
1420 1420 try:
1421 1421 # str(rev)
1422 1422 rev = int(id)
1423 1423 if "%d" % rev != id:
1424 1424 raise ValueError
1425 1425 if rev < 0:
1426 1426 rev = len(self) + rev
1427 1427 if rev < 0 or rev >= len(self):
1428 1428 raise ValueError
1429 1429 return self.node(rev)
1430 1430 except (ValueError, OverflowError):
1431 1431 pass
1432 1432 if len(id) == 40:
1433 1433 try:
1434 1434 # a full hex nodeid?
1435 1435 node = bin(id)
1436 1436 self.rev(node)
1437 1437 return node
1438 1438 except (TypeError, LookupError):
1439 1439 pass
1440 1440
1441 1441 def _partialmatch(self, id):
1442 1442 # we don't care wdirfilenodeids as they should be always full hash
1443 1443 maybewdir = wdirhex.startswith(id)
1444 1444 try:
1445 1445 partial = self.index.partialmatch(id)
1446 1446 if partial and self.hasnode(partial):
1447 1447 if maybewdir:
1448 1448 # single 'ff...' match in radix tree, ambiguous with wdir
1449 1449 raise RevlogError
1450 1450 return partial
1451 1451 if maybewdir:
1452 1452 # no 'ff...' match in radix tree, wdir identified
1453 1453 raise error.WdirUnsupported
1454 1454 return None
1455 1455 except RevlogError:
1456 1456 # parsers.c radix tree lookup gave multiple matches
1457 1457 # fast path: for unfiltered changelog, radix tree is accurate
1458 1458 if not getattr(self, 'filteredrevs', None):
1459 1459 raise LookupError(id, self.indexfile,
1460 1460 _('ambiguous identifier'))
1461 1461 # fall through to slow path that filters hidden revisions
1462 1462 except (AttributeError, ValueError):
1463 1463 # we are pure python, or key was too short to search radix tree
1464 1464 pass
1465 1465
1466 1466 if id in self._pcache:
1467 1467 return self._pcache[id]
1468 1468
1469 if len(id) < 40:
1469 if len(id) <= 40:
1470 1470 try:
1471 1471 # hex(node)[:...]
1472 1472 l = len(id) // 2 # grab an even number of digits
1473 1473 prefix = bin(id[:l * 2])
1474 1474 nl = [e[7] for e in self.index if e[7].startswith(prefix)]
1475 1475 nl = [n for n in nl if hex(n).startswith(id) and
1476 1476 self.hasnode(n)]
1477 1477 if len(nl) > 0:
1478 1478 if len(nl) == 1 and not maybewdir:
1479 1479 self._pcache[id] = nl[0]
1480 1480 return nl[0]
1481 1481 raise LookupError(id, self.indexfile,
1482 1482 _('ambiguous identifier'))
1483 1483 if maybewdir:
1484 1484 raise error.WdirUnsupported
1485 1485 return None
1486 1486 except TypeError:
1487 1487 pass
1488 1488
1489 1489 def lookup(self, id):
1490 1490 """locate a node based on:
1491 1491 - revision number or str(revision number)
1492 1492 - nodeid or subset of hex nodeid
1493 1493 """
1494 1494 n = self._match(id)
1495 1495 if n is not None:
1496 1496 return n
1497 1497 n = self._partialmatch(id)
1498 1498 if n:
1499 1499 return n
1500 1500
1501 1501 raise LookupError(id, self.indexfile, _('no match found'))
1502 1502
1503 1503 def shortest(self, node, minlength=1):
1504 1504 """Find the shortest unambiguous prefix that matches node."""
1505 1505 def isvalid(test):
1506 1506 try:
1507 1507 if self._partialmatch(test) is None:
1508 1508 return False
1509 1509
1510 1510 try:
1511 1511 i = int(test)
1512 1512 # if we are a pure int, then starting with zero will not be
1513 1513 # confused as a rev; or, obviously, if the int is larger
1514 1514 # than the value of the tip rev
1515 1515 if test[0] == '0' or i > len(self):
1516 1516 return True
1517 1517 return False
1518 1518 except ValueError:
1519 1519 return True
1520 1520 except error.RevlogError:
1521 1521 return False
1522 1522 except error.WdirUnsupported:
1523 1523 # single 'ff...' match
1524 1524 return True
1525 1525
1526 1526 hexnode = hex(node)
1527 1527 shortest = hexnode
1528 1528 startlength = max(6, minlength)
1529 1529 length = startlength
1530 1530 while True:
1531 1531 test = hexnode[:length]
1532 1532 if isvalid(test):
1533 1533 shortest = test
1534 1534 if length == minlength or length > startlength:
1535 1535 return shortest
1536 1536 length -= 1
1537 1537 else:
1538 1538 length += 1
1539 1539 if len(shortest) <= length:
1540 1540 return shortest
1541 1541
1542 1542 def cmp(self, node, text):
1543 1543 """compare text with a given file revision
1544 1544
1545 1545 returns True if text is different than what is stored.
1546 1546 """
1547 1547 p1, p2 = self.parents(node)
1548 1548 return hash(text, p1, p2) != node
1549 1549
1550 1550 def _cachesegment(self, offset, data):
1551 1551 """Add a segment to the revlog cache.
1552 1552
1553 1553 Accepts an absolute offset and the data that is at that location.
1554 1554 """
1555 1555 o, d = self._chunkcache
1556 1556 # try to add to existing cache
1557 1557 if o + len(d) == offset and len(d) + len(data) < _chunksize:
1558 1558 self._chunkcache = o, d + data
1559 1559 else:
1560 1560 self._chunkcache = offset, data
1561 1561
1562 1562 def _readsegment(self, offset, length, df=None):
1563 1563 """Load a segment of raw data from the revlog.
1564 1564
1565 1565 Accepts an absolute offset, length to read, and an optional existing
1566 1566 file handle to read from.
1567 1567
1568 1568 If an existing file handle is passed, it will be seeked and the
1569 1569 original seek position will NOT be restored.
1570 1570
1571 1571 Returns a str or buffer of raw byte data.
1572 1572 """
1573 1573 # Cache data both forward and backward around the requested
1574 1574 # data, in a fixed size window. This helps speed up operations
1575 1575 # involving reading the revlog backwards.
1576 1576 cachesize = self._chunkcachesize
1577 1577 realoffset = offset & ~(cachesize - 1)
1578 1578 reallength = (((offset + length + cachesize) & ~(cachesize - 1))
1579 1579 - realoffset)
1580 1580 with self._datareadfp(df) as df:
1581 1581 df.seek(realoffset)
1582 1582 d = df.read(reallength)
1583 1583 self._cachesegment(realoffset, d)
1584 1584 if offset != realoffset or reallength != length:
1585 1585 return util.buffer(d, offset - realoffset, length)
1586 1586 return d
1587 1587
1588 1588 def _getsegment(self, offset, length, df=None):
1589 1589 """Obtain a segment of raw data from the revlog.
1590 1590
1591 1591 Accepts an absolute offset, length of bytes to obtain, and an
1592 1592 optional file handle to the already-opened revlog. If the file
1593 1593 handle is used, it's original seek position will not be preserved.
1594 1594
1595 1595 Requests for data may be returned from a cache.
1596 1596
1597 1597 Returns a str or a buffer instance of raw byte data.
1598 1598 """
1599 1599 o, d = self._chunkcache
1600 1600 l = len(d)
1601 1601
1602 1602 # is it in the cache?
1603 1603 cachestart = offset - o
1604 1604 cacheend = cachestart + length
1605 1605 if cachestart >= 0 and cacheend <= l:
1606 1606 if cachestart == 0 and cacheend == l:
1607 1607 return d # avoid a copy
1608 1608 return util.buffer(d, cachestart, cacheend - cachestart)
1609 1609
1610 1610 return self._readsegment(offset, length, df=df)
1611 1611
1612 1612 def _getsegmentforrevs(self, startrev, endrev, df=None):
1613 1613 """Obtain a segment of raw data corresponding to a range of revisions.
1614 1614
1615 1615 Accepts the start and end revisions and an optional already-open
1616 1616 file handle to be used for reading. If the file handle is read, its
1617 1617 seek position will not be preserved.
1618 1618
1619 1619 Requests for data may be satisfied by a cache.
1620 1620
1621 1621 Returns a 2-tuple of (offset, data) for the requested range of
1622 1622 revisions. Offset is the integer offset from the beginning of the
1623 1623 revlog and data is a str or buffer of the raw byte data.
1624 1624
1625 1625 Callers will need to call ``self.start(rev)`` and ``self.length(rev)``
1626 1626 to determine where each revision's data begins and ends.
1627 1627 """
1628 1628 # Inlined self.start(startrev) & self.end(endrev) for perf reasons
1629 1629 # (functions are expensive).
1630 1630 index = self.index
1631 1631 istart = index[startrev]
1632 1632 start = int(istart[0] >> 16)
1633 1633 if startrev == endrev:
1634 1634 end = start + istart[1]
1635 1635 else:
1636 1636 iend = index[endrev]
1637 1637 end = int(iend[0] >> 16) + iend[1]
1638 1638
1639 1639 if self._inline:
1640 1640 start += (startrev + 1) * self._io.size
1641 1641 end += (endrev + 1) * self._io.size
1642 1642 length = end - start
1643 1643
1644 1644 return start, self._getsegment(start, length, df=df)
1645 1645
1646 1646 def _chunk(self, rev, df=None):
1647 1647 """Obtain a single decompressed chunk for a revision.
1648 1648
1649 1649 Accepts an integer revision and an optional already-open file handle
1650 1650 to be used for reading. If used, the seek position of the file will not
1651 1651 be preserved.
1652 1652
1653 1653 Returns a str holding uncompressed data for the requested revision.
1654 1654 """
1655 1655 return self.decompress(self._getsegmentforrevs(rev, rev, df=df)[1])
1656 1656
1657 1657 def _chunks(self, revs, df=None):
1658 1658 """Obtain decompressed chunks for the specified revisions.
1659 1659
1660 1660 Accepts an iterable of numeric revisions that are assumed to be in
1661 1661 ascending order. Also accepts an optional already-open file handle
1662 1662 to be used for reading. If used, the seek position of the file will
1663 1663 not be preserved.
1664 1664
1665 1665 This function is similar to calling ``self._chunk()`` multiple times,
1666 1666 but is faster.
1667 1667
1668 1668 Returns a list with decompressed data for each requested revision.
1669 1669 """
1670 1670 if not revs:
1671 1671 return []
1672 1672 start = self.start
1673 1673 length = self.length
1674 1674 inline = self._inline
1675 1675 iosize = self._io.size
1676 1676 buffer = util.buffer
1677 1677
1678 1678 l = []
1679 1679 ladd = l.append
1680 1680
1681 1681 if not self._withsparseread:
1682 1682 slicedchunks = (revs,)
1683 1683 else:
1684 1684 slicedchunks = _slicechunk(self, revs)
1685 1685
1686 1686 for revschunk in slicedchunks:
1687 1687 firstrev = revschunk[0]
1688 1688 # Skip trailing revisions with empty diff
1689 1689 for lastrev in revschunk[::-1]:
1690 1690 if length(lastrev) != 0:
1691 1691 break
1692 1692
1693 1693 try:
1694 1694 offset, data = self._getsegmentforrevs(firstrev, lastrev, df=df)
1695 1695 except OverflowError:
1696 1696 # issue4215 - we can't cache a run of chunks greater than
1697 1697 # 2G on Windows
1698 1698 return [self._chunk(rev, df=df) for rev in revschunk]
1699 1699
1700 1700 decomp = self.decompress
1701 1701 for rev in revschunk:
1702 1702 chunkstart = start(rev)
1703 1703 if inline:
1704 1704 chunkstart += (rev + 1) * iosize
1705 1705 chunklength = length(rev)
1706 1706 ladd(decomp(buffer(data, chunkstart - offset, chunklength)))
1707 1707
1708 1708 return l
1709 1709
1710 1710 def _chunkclear(self):
1711 1711 """Clear the raw chunk cache."""
1712 1712 self._chunkcache = (0, '')
1713 1713
1714 1714 def deltaparent(self, rev):
1715 1715 """return deltaparent of the given revision"""
1716 1716 base = self.index[rev][3]
1717 1717 if base == rev:
1718 1718 return nullrev
1719 1719 elif self._generaldelta:
1720 1720 return base
1721 1721 else:
1722 1722 return rev - 1
1723 1723
1724 1724 def revdiff(self, rev1, rev2):
1725 1725 """return or calculate a delta between two revisions
1726 1726
1727 1727 The delta calculated is in binary form and is intended to be written to
1728 1728 revlog data directly. So this function needs raw revision data.
1729 1729 """
1730 1730 if rev1 != nullrev and self.deltaparent(rev2) == rev1:
1731 1731 return bytes(self._chunk(rev2))
1732 1732
1733 1733 return mdiff.textdiff(self.revision(rev1, raw=True),
1734 1734 self.revision(rev2, raw=True))
1735 1735
1736 1736 def revision(self, nodeorrev, _df=None, raw=False):
1737 1737 """return an uncompressed revision of a given node or revision
1738 1738 number.
1739 1739
1740 1740 _df - an existing file handle to read from. (internal-only)
1741 1741 raw - an optional argument specifying if the revision data is to be
1742 1742 treated as raw data when applying flag transforms. 'raw' should be set
1743 1743 to True when generating changegroups or in debug commands.
1744 1744 """
1745 1745 if isinstance(nodeorrev, int):
1746 1746 rev = nodeorrev
1747 1747 node = self.node(rev)
1748 1748 else:
1749 1749 node = nodeorrev
1750 1750 rev = None
1751 1751
1752 1752 cachedrev = None
1753 1753 flags = None
1754 1754 rawtext = None
1755 1755 if node == nullid:
1756 1756 return ""
1757 1757 if self._cache:
1758 1758 if self._cache[0] == node:
1759 1759 # _cache only stores rawtext
1760 1760 if raw:
1761 1761 return self._cache[2]
1762 1762 # duplicated, but good for perf
1763 1763 if rev is None:
1764 1764 rev = self.rev(node)
1765 1765 if flags is None:
1766 1766 flags = self.flags(rev)
1767 1767 # no extra flags set, no flag processor runs, text = rawtext
1768 1768 if flags == REVIDX_DEFAULT_FLAGS:
1769 1769 return self._cache[2]
1770 1770 # rawtext is reusable. need to run flag processor
1771 1771 rawtext = self._cache[2]
1772 1772
1773 1773 cachedrev = self._cache[1]
1774 1774
1775 1775 # look up what we need to read
1776 1776 if rawtext is None:
1777 1777 if rev is None:
1778 1778 rev = self.rev(node)
1779 1779
1780 1780 chain, stopped = self._deltachain(rev, stoprev=cachedrev)
1781 1781 if stopped:
1782 1782 rawtext = self._cache[2]
1783 1783
1784 1784 # drop cache to save memory
1785 1785 self._cache = None
1786 1786
1787 1787 bins = self._chunks(chain, df=_df)
1788 1788 if rawtext is None:
1789 1789 rawtext = bytes(bins[0])
1790 1790 bins = bins[1:]
1791 1791
1792 1792 rawtext = mdiff.patches(rawtext, bins)
1793 1793 self._cache = (node, rev, rawtext)
1794 1794
1795 1795 if flags is None:
1796 1796 if rev is None:
1797 1797 rev = self.rev(node)
1798 1798 flags = self.flags(rev)
1799 1799
1800 1800 text, validatehash = self._processflags(rawtext, flags, 'read', raw=raw)
1801 1801 if validatehash:
1802 1802 self.checkhash(text, node, rev=rev)
1803 1803
1804 1804 return text
1805 1805
1806 1806 def hash(self, text, p1, p2):
1807 1807 """Compute a node hash.
1808 1808
1809 1809 Available as a function so that subclasses can replace the hash
1810 1810 as needed.
1811 1811 """
1812 1812 return hash(text, p1, p2)
1813 1813
1814 1814 def _processflags(self, text, flags, operation, raw=False):
1815 1815 """Inspect revision data flags and applies transforms defined by
1816 1816 registered flag processors.
1817 1817
1818 1818 ``text`` - the revision data to process
1819 1819 ``flags`` - the revision flags
1820 1820 ``operation`` - the operation being performed (read or write)
1821 1821 ``raw`` - an optional argument describing if the raw transform should be
1822 1822 applied.
1823 1823
1824 1824 This method processes the flags in the order (or reverse order if
1825 1825 ``operation`` is 'write') defined by REVIDX_FLAGS_ORDER, applying the
1826 1826 flag processors registered for present flags. The order of flags defined
1827 1827 in REVIDX_FLAGS_ORDER needs to be stable to allow non-commutativity.
1828 1828
1829 1829 Returns a 2-tuple of ``(text, validatehash)`` where ``text`` is the
1830 1830 processed text and ``validatehash`` is a bool indicating whether the
1831 1831 returned text should be checked for hash integrity.
1832 1832
1833 1833 Note: If the ``raw`` argument is set, it has precedence over the
1834 1834 operation and will only update the value of ``validatehash``.
1835 1835 """
1836 1836 # fast path: no flag processors will run
1837 1837 if flags == 0:
1838 1838 return text, True
1839 1839 if not operation in ('read', 'write'):
1840 1840 raise ProgrammingError(_("invalid '%s' operation ") % (operation))
1841 1841 # Check all flags are known.
1842 1842 if flags & ~REVIDX_KNOWN_FLAGS:
1843 1843 raise RevlogError(_("incompatible revision flag '%#x'") %
1844 1844 (flags & ~REVIDX_KNOWN_FLAGS))
1845 1845 validatehash = True
1846 1846 # Depending on the operation (read or write), the order might be
1847 1847 # reversed due to non-commutative transforms.
1848 1848 orderedflags = REVIDX_FLAGS_ORDER
1849 1849 if operation == 'write':
1850 1850 orderedflags = reversed(orderedflags)
1851 1851
1852 1852 for flag in orderedflags:
1853 1853 # If a flagprocessor has been registered for a known flag, apply the
1854 1854 # related operation transform and update result tuple.
1855 1855 if flag & flags:
1856 1856 vhash = True
1857 1857
1858 1858 if flag not in _flagprocessors:
1859 1859 message = _("missing processor for flag '%#x'") % (flag)
1860 1860 raise RevlogError(message)
1861 1861
1862 1862 processor = _flagprocessors[flag]
1863 1863 if processor is not None:
1864 1864 readtransform, writetransform, rawtransform = processor
1865 1865
1866 1866 if raw:
1867 1867 vhash = rawtransform(self, text)
1868 1868 elif operation == 'read':
1869 1869 text, vhash = readtransform(self, text)
1870 1870 else: # write operation
1871 1871 text, vhash = writetransform(self, text)
1872 1872 validatehash = validatehash and vhash
1873 1873
1874 1874 return text, validatehash
1875 1875
1876 1876 def checkhash(self, text, node, p1=None, p2=None, rev=None):
1877 1877 """Check node hash integrity.
1878 1878
1879 1879 Available as a function so that subclasses can extend hash mismatch
1880 1880 behaviors as needed.
1881 1881 """
1882 1882 try:
1883 1883 if p1 is None and p2 is None:
1884 1884 p1, p2 = self.parents(node)
1885 1885 if node != self.hash(text, p1, p2):
1886 1886 revornode = rev
1887 1887 if revornode is None:
1888 1888 revornode = templatefilters.short(hex(node))
1889 1889 raise RevlogError(_("integrity check failed on %s:%s")
1890 1890 % (self.indexfile, pycompat.bytestr(revornode)))
1891 1891 except RevlogError:
1892 1892 if self._censorable and _censoredtext(text):
1893 1893 raise error.CensoredNodeError(self.indexfile, node, text)
1894 1894 raise
1895 1895
1896 1896 def _enforceinlinesize(self, tr, fp=None):
1897 1897 """Check if the revlog is too big for inline and convert if so.
1898 1898
1899 1899 This should be called after revisions are added to the revlog. If the
1900 1900 revlog has grown too large to be an inline revlog, it will convert it
1901 1901 to use multiple index and data files.
1902 1902 """
1903 1903 if not self._inline or (self.start(-2) + self.length(-2)) < _maxinline:
1904 1904 return
1905 1905
1906 1906 trinfo = tr.find(self.indexfile)
1907 1907 if trinfo is None:
1908 1908 raise RevlogError(_("%s not found in the transaction")
1909 1909 % self.indexfile)
1910 1910
1911 1911 trindex = trinfo[2]
1912 1912 if trindex is not None:
1913 1913 dataoff = self.start(trindex)
1914 1914 else:
1915 1915 # revlog was stripped at start of transaction, use all leftover data
1916 1916 trindex = len(self) - 1
1917 1917 dataoff = self.end(-2)
1918 1918
1919 1919 tr.add(self.datafile, dataoff)
1920 1920
1921 1921 if fp:
1922 1922 fp.flush()
1923 1923 fp.close()
1924 1924
1925 1925 with self._datafp('w') as df:
1926 1926 for r in self:
1927 1927 df.write(self._getsegmentforrevs(r, r)[1])
1928 1928
1929 1929 with self._indexfp('w') as fp:
1930 1930 self.version &= ~FLAG_INLINE_DATA
1931 1931 self._inline = False
1932 1932 io = self._io
1933 1933 for i in self:
1934 1934 e = io.packentry(self.index[i], self.node, self.version, i)
1935 1935 fp.write(e)
1936 1936
1937 1937 # the temp file replace the real index when we exit the context
1938 1938 # manager
1939 1939
1940 1940 tr.replace(self.indexfile, trindex * self._io.size)
1941 1941 self._chunkclear()
1942 1942
1943 1943 def addrevision(self, text, transaction, link, p1, p2, cachedelta=None,
1944 1944 node=None, flags=REVIDX_DEFAULT_FLAGS, deltacomputer=None):
1945 1945 """add a revision to the log
1946 1946
1947 1947 text - the revision data to add
1948 1948 transaction - the transaction object used for rollback
1949 1949 link - the linkrev data to add
1950 1950 p1, p2 - the parent nodeids of the revision
1951 1951 cachedelta - an optional precomputed delta
1952 1952 node - nodeid of revision; typically node is not specified, and it is
1953 1953 computed by default as hash(text, p1, p2), however subclasses might
1954 1954 use different hashing method (and override checkhash() in such case)
1955 1955 flags - the known flags to set on the revision
1956 1956 deltacomputer - an optional _deltacomputer instance shared between
1957 1957 multiple calls
1958 1958 """
1959 1959 if link == nullrev:
1960 1960 raise RevlogError(_("attempted to add linkrev -1 to %s")
1961 1961 % self.indexfile)
1962 1962
1963 1963 if flags:
1964 1964 node = node or self.hash(text, p1, p2)
1965 1965
1966 1966 rawtext, validatehash = self._processflags(text, flags, 'write')
1967 1967
1968 1968 # If the flag processor modifies the revision data, ignore any provided
1969 1969 # cachedelta.
1970 1970 if rawtext != text:
1971 1971 cachedelta = None
1972 1972
1973 1973 if len(rawtext) > _maxentrysize:
1974 1974 raise RevlogError(
1975 1975 _("%s: size of %d bytes exceeds maximum revlog storage of 2GiB")
1976 1976 % (self.indexfile, len(rawtext)))
1977 1977
1978 1978 node = node or self.hash(rawtext, p1, p2)
1979 1979 if node in self.nodemap:
1980 1980 return node
1981 1981
1982 1982 if validatehash:
1983 1983 self.checkhash(rawtext, node, p1=p1, p2=p2)
1984 1984
1985 1985 return self.addrawrevision(rawtext, transaction, link, p1, p2, node,
1986 1986 flags, cachedelta=cachedelta,
1987 1987 deltacomputer=deltacomputer)
1988 1988
1989 1989 def addrawrevision(self, rawtext, transaction, link, p1, p2, node, flags,
1990 1990 cachedelta=None, deltacomputer=None):
1991 1991 """add a raw revision with known flags, node and parents
1992 1992 useful when reusing a revision not stored in this revlog (ex: received
1993 1993 over wire, or read from an external bundle).
1994 1994 """
1995 1995 dfh = None
1996 1996 if not self._inline:
1997 1997 dfh = self._datafp("a+")
1998 1998 ifh = self._indexfp("a+")
1999 1999 try:
2000 2000 return self._addrevision(node, rawtext, transaction, link, p1, p2,
2001 2001 flags, cachedelta, ifh, dfh,
2002 2002 deltacomputer=deltacomputer)
2003 2003 finally:
2004 2004 if dfh:
2005 2005 dfh.close()
2006 2006 ifh.close()
2007 2007
2008 2008 def compress(self, data):
2009 2009 """Generate a possibly-compressed representation of data."""
2010 2010 if not data:
2011 2011 return '', data
2012 2012
2013 2013 compressed = self._compressor.compress(data)
2014 2014
2015 2015 if compressed:
2016 2016 # The revlog compressor added the header in the returned data.
2017 2017 return '', compressed
2018 2018
2019 2019 if data[0:1] == '\0':
2020 2020 return '', data
2021 2021 return 'u', data
2022 2022
2023 2023 def decompress(self, data):
2024 2024 """Decompress a revlog chunk.
2025 2025
2026 2026 The chunk is expected to begin with a header identifying the
2027 2027 format type so it can be routed to an appropriate decompressor.
2028 2028 """
2029 2029 if not data:
2030 2030 return data
2031 2031
2032 2032 # Revlogs are read much more frequently than they are written and many
2033 2033 # chunks only take microseconds to decompress, so performance is
2034 2034 # important here.
2035 2035 #
2036 2036 # We can make a few assumptions about revlogs:
2037 2037 #
2038 2038 # 1) the majority of chunks will be compressed (as opposed to inline
2039 2039 # raw data).
2040 2040 # 2) decompressing *any* data will likely by at least 10x slower than
2041 2041 # returning raw inline data.
2042 2042 # 3) we want to prioritize common and officially supported compression
2043 2043 # engines
2044 2044 #
2045 2045 # It follows that we want to optimize for "decompress compressed data
2046 2046 # when encoded with common and officially supported compression engines"
2047 2047 # case over "raw data" and "data encoded by less common or non-official
2048 2048 # compression engines." That is why we have the inline lookup first
2049 2049 # followed by the compengines lookup.
2050 2050 #
2051 2051 # According to `hg perfrevlogchunks`, this is ~0.5% faster for zlib
2052 2052 # compressed chunks. And this matters for changelog and manifest reads.
2053 2053 t = data[0:1]
2054 2054
2055 2055 if t == 'x':
2056 2056 try:
2057 2057 return _zlibdecompress(data)
2058 2058 except zlib.error as e:
2059 2059 raise RevlogError(_('revlog decompress error: %s') %
2060 2060 stringutil.forcebytestr(e))
2061 2061 # '\0' is more common than 'u' so it goes first.
2062 2062 elif t == '\0':
2063 2063 return data
2064 2064 elif t == 'u':
2065 2065 return util.buffer(data, 1)
2066 2066
2067 2067 try:
2068 2068 compressor = self._decompressors[t]
2069 2069 except KeyError:
2070 2070 try:
2071 2071 engine = util.compengines.forrevlogheader(t)
2072 2072 compressor = engine.revlogcompressor()
2073 2073 self._decompressors[t] = compressor
2074 2074 except KeyError:
2075 2075 raise RevlogError(_('unknown compression type %r') % t)
2076 2076
2077 2077 return compressor.decompress(data)
2078 2078
2079 2079 def _isgooddeltainfo(self, d, textlen):
2080 2080 """Returns True if the given delta is good. Good means that it is within
2081 2081 the disk span, disk size, and chain length bounds that we know to be
2082 2082 performant."""
2083 2083 if d is None:
2084 2084 return False
2085 2085
2086 2086 # - 'd.distance' is the distance from the base revision -- bounding it
2087 2087 # limits the amount of I/O we need to do.
2088 2088 # - 'd.compresseddeltalen' is the sum of the total size of deltas we
2089 2089 # need to apply -- bounding it limits the amount of CPU we consume.
2090 2090
2091 2091 defaultmax = textlen * 4
2092 2092 maxdist = self._maxdeltachainspan
2093 2093 if not maxdist:
2094 2094 maxdist = d.distance # ensure the conditional pass
2095 2095 maxdist = max(maxdist, defaultmax)
2096 2096 if (d.distance > maxdist or d.deltalen > textlen or
2097 2097 d.compresseddeltalen > textlen * 2 or
2098 2098 (self._maxchainlen and d.chainlen > self._maxchainlen)):
2099 2099 return False
2100 2100
2101 2101 return True
2102 2102
2103 2103 def _addrevision(self, node, rawtext, transaction, link, p1, p2, flags,
2104 2104 cachedelta, ifh, dfh, alwayscache=False,
2105 2105 deltacomputer=None):
2106 2106 """internal function to add revisions to the log
2107 2107
2108 2108 see addrevision for argument descriptions.
2109 2109
2110 2110 note: "addrevision" takes non-raw text, "_addrevision" takes raw text.
2111 2111
2112 2112 if "deltacomputer" is not provided or None, a defaultdeltacomputer will
2113 2113 be used.
2114 2114
2115 2115 invariants:
2116 2116 - rawtext is optional (can be None); if not set, cachedelta must be set.
2117 2117 if both are set, they must correspond to each other.
2118 2118 """
2119 2119 if node == nullid:
2120 2120 raise RevlogError(_("%s: attempt to add null revision") %
2121 2121 (self.indexfile))
2122 2122 if node == wdirid or node in wdirfilenodeids:
2123 2123 raise RevlogError(_("%s: attempt to add wdir revision") %
2124 2124 (self.indexfile))
2125 2125
2126 2126 if self._inline:
2127 2127 fh = ifh
2128 2128 else:
2129 2129 fh = dfh
2130 2130
2131 2131 btext = [rawtext]
2132 2132
2133 2133 curr = len(self)
2134 2134 prev = curr - 1
2135 2135 offset = self.end(prev)
2136 2136 p1r, p2r = self.rev(p1), self.rev(p2)
2137 2137
2138 2138 # full versions are inserted when the needed deltas
2139 2139 # become comparable to the uncompressed text
2140 2140 if rawtext is None:
2141 2141 # need rawtext size, before changed by flag processors, which is
2142 2142 # the non-raw size. use revlog explicitly to avoid filelog's extra
2143 2143 # logic that might remove metadata size.
2144 2144 textlen = mdiff.patchedsize(revlog.size(self, cachedelta[0]),
2145 2145 cachedelta[1])
2146 2146 else:
2147 2147 textlen = len(rawtext)
2148 2148
2149 2149 if deltacomputer is None:
2150 2150 deltacomputer = _deltacomputer(self)
2151 2151
2152 2152 revinfo = _revisioninfo(node, p1, p2, btext, textlen, cachedelta, flags)
2153 2153
2154 2154 # no delta for flag processor revision (see "candelta" for why)
2155 2155 # not calling candelta since only one revision needs test, also to
2156 2156 # avoid overhead fetching flags again.
2157 2157 if flags & REVIDX_RAWTEXT_CHANGING_FLAGS:
2158 2158 deltainfo = None
2159 2159 else:
2160 2160 deltainfo = deltacomputer.finddeltainfo(revinfo, fh)
2161 2161
2162 2162 if deltainfo is not None:
2163 2163 base = deltainfo.base
2164 2164 chainbase = deltainfo.chainbase
2165 2165 data = deltainfo.data
2166 2166 l = deltainfo.deltalen
2167 2167 else:
2168 2168 rawtext = deltacomputer.buildtext(revinfo, fh)
2169 2169 data = self.compress(rawtext)
2170 2170 l = len(data[1]) + len(data[0])
2171 2171 base = chainbase = curr
2172 2172
2173 2173 e = (offset_type(offset, flags), l, textlen,
2174 2174 base, link, p1r, p2r, node)
2175 2175 self.index.insert(-1, e)
2176 2176 self.nodemap[node] = curr
2177 2177
2178 2178 entry = self._io.packentry(e, self.node, self.version, curr)
2179 2179 self._writeentry(transaction, ifh, dfh, entry, data, link, offset)
2180 2180
2181 2181 if alwayscache and rawtext is None:
2182 2182 rawtext = deltacomputer._buildtext(revinfo, fh)
2183 2183
2184 2184 if type(rawtext) == bytes: # only accept immutable objects
2185 2185 self._cache = (node, curr, rawtext)
2186 2186 self._chainbasecache[curr] = chainbase
2187 2187 return node
2188 2188
2189 2189 def _writeentry(self, transaction, ifh, dfh, entry, data, link, offset):
2190 2190 # Files opened in a+ mode have inconsistent behavior on various
2191 2191 # platforms. Windows requires that a file positioning call be made
2192 2192 # when the file handle transitions between reads and writes. See
2193 2193 # 3686fa2b8eee and the mixedfilemodewrapper in windows.py. On other
2194 2194 # platforms, Python or the platform itself can be buggy. Some versions
2195 2195 # of Solaris have been observed to not append at the end of the file
2196 2196 # if the file was seeked to before the end. See issue4943 for more.
2197 2197 #
2198 2198 # We work around this issue by inserting a seek() before writing.
2199 2199 # Note: This is likely not necessary on Python 3.
2200 2200 ifh.seek(0, os.SEEK_END)
2201 2201 if dfh:
2202 2202 dfh.seek(0, os.SEEK_END)
2203 2203
2204 2204 curr = len(self) - 1
2205 2205 if not self._inline:
2206 2206 transaction.add(self.datafile, offset)
2207 2207 transaction.add(self.indexfile, curr * len(entry))
2208 2208 if data[0]:
2209 2209 dfh.write(data[0])
2210 2210 dfh.write(data[1])
2211 2211 ifh.write(entry)
2212 2212 else:
2213 2213 offset += curr * self._io.size
2214 2214 transaction.add(self.indexfile, offset, curr)
2215 2215 ifh.write(entry)
2216 2216 ifh.write(data[0])
2217 2217 ifh.write(data[1])
2218 2218 self._enforceinlinesize(transaction, ifh)
2219 2219
2220 2220 def addgroup(self, deltas, linkmapper, transaction, addrevisioncb=None):
2221 2221 """
2222 2222 add a delta group
2223 2223
2224 2224 given a set of deltas, add them to the revision log. the
2225 2225 first delta is against its parent, which should be in our
2226 2226 log, the rest are against the previous delta.
2227 2227
2228 2228 If ``addrevisioncb`` is defined, it will be called with arguments of
2229 2229 this revlog and the node that was added.
2230 2230 """
2231 2231
2232 2232 nodes = []
2233 2233
2234 2234 r = len(self)
2235 2235 end = 0
2236 2236 if r:
2237 2237 end = self.end(r - 1)
2238 2238 ifh = self._indexfp("a+")
2239 2239 isize = r * self._io.size
2240 2240 if self._inline:
2241 2241 transaction.add(self.indexfile, end + isize, r)
2242 2242 dfh = None
2243 2243 else:
2244 2244 transaction.add(self.indexfile, isize, r)
2245 2245 transaction.add(self.datafile, end)
2246 2246 dfh = self._datafp("a+")
2247 2247 def flush():
2248 2248 if dfh:
2249 2249 dfh.flush()
2250 2250 ifh.flush()
2251 2251 try:
2252 2252 deltacomputer = _deltacomputer(self)
2253 2253 # loop through our set of deltas
2254 2254 for data in deltas:
2255 2255 node, p1, p2, linknode, deltabase, delta, flags = data
2256 2256 link = linkmapper(linknode)
2257 2257 flags = flags or REVIDX_DEFAULT_FLAGS
2258 2258
2259 2259 nodes.append(node)
2260 2260
2261 2261 if node in self.nodemap:
2262 2262 # this can happen if two branches make the same change
2263 2263 continue
2264 2264
2265 2265 for p in (p1, p2):
2266 2266 if p not in self.nodemap:
2267 2267 raise LookupError(p, self.indexfile,
2268 2268 _('unknown parent'))
2269 2269
2270 2270 if deltabase not in self.nodemap:
2271 2271 raise LookupError(deltabase, self.indexfile,
2272 2272 _('unknown delta base'))
2273 2273
2274 2274 baserev = self.rev(deltabase)
2275 2275
2276 2276 if baserev != nullrev and self.iscensored(baserev):
2277 2277 # if base is censored, delta must be full replacement in a
2278 2278 # single patch operation
2279 2279 hlen = struct.calcsize(">lll")
2280 2280 oldlen = self.rawsize(baserev)
2281 2281 newlen = len(delta) - hlen
2282 2282 if delta[:hlen] != mdiff.replacediffheader(oldlen, newlen):
2283 2283 raise error.CensoredBaseError(self.indexfile,
2284 2284 self.node(baserev))
2285 2285
2286 2286 if not flags and self._peek_iscensored(baserev, delta, flush):
2287 2287 flags |= REVIDX_ISCENSORED
2288 2288
2289 2289 # We assume consumers of addrevisioncb will want to retrieve
2290 2290 # the added revision, which will require a call to
2291 2291 # revision(). revision() will fast path if there is a cache
2292 2292 # hit. So, we tell _addrevision() to always cache in this case.
2293 2293 # We're only using addgroup() in the context of changegroup
2294 2294 # generation so the revision data can always be handled as raw
2295 2295 # by the flagprocessor.
2296 2296 self._addrevision(node, None, transaction, link,
2297 2297 p1, p2, flags, (baserev, delta),
2298 2298 ifh, dfh,
2299 2299 alwayscache=bool(addrevisioncb),
2300 2300 deltacomputer=deltacomputer)
2301 2301
2302 2302 if addrevisioncb:
2303 2303 addrevisioncb(self, node)
2304 2304
2305 2305 if not dfh and not self._inline:
2306 2306 # addrevision switched from inline to conventional
2307 2307 # reopen the index
2308 2308 ifh.close()
2309 2309 dfh = self._datafp("a+")
2310 2310 ifh = self._indexfp("a+")
2311 2311 finally:
2312 2312 if dfh:
2313 2313 dfh.close()
2314 2314 ifh.close()
2315 2315
2316 2316 return nodes
2317 2317
2318 2318 def iscensored(self, rev):
2319 2319 """Check if a file revision is censored."""
2320 2320 if not self._censorable:
2321 2321 return False
2322 2322
2323 2323 return self.flags(rev) & REVIDX_ISCENSORED
2324 2324
2325 2325 def _peek_iscensored(self, baserev, delta, flush):
2326 2326 """Quickly check if a delta produces a censored revision."""
2327 2327 if not self._censorable:
2328 2328 return False
2329 2329
2330 2330 # Fragile heuristic: unless new file meta keys are added alphabetically
2331 2331 # preceding "censored", all censored revisions are prefixed by
2332 2332 # "\1\ncensored:". A delta producing such a censored revision must be a
2333 2333 # full-replacement delta, so we inspect the first and only patch in the
2334 2334 # delta for this prefix.
2335 2335 hlen = struct.calcsize(">lll")
2336 2336 if len(delta) <= hlen:
2337 2337 return False
2338 2338
2339 2339 oldlen = self.rawsize(baserev)
2340 2340 newlen = len(delta) - hlen
2341 2341 if delta[:hlen] != mdiff.replacediffheader(oldlen, newlen):
2342 2342 return False
2343 2343
2344 2344 add = "\1\ncensored:"
2345 2345 addlen = len(add)
2346 2346 return newlen >= addlen and delta[hlen:hlen + addlen] == add
2347 2347
2348 2348 def getstrippoint(self, minlink):
2349 2349 """find the minimum rev that must be stripped to strip the linkrev
2350 2350
2351 2351 Returns a tuple containing the minimum rev and a set of all revs that
2352 2352 have linkrevs that will be broken by this strip.
2353 2353 """
2354 2354 brokenrevs = set()
2355 2355 strippoint = len(self)
2356 2356
2357 2357 heads = {}
2358 2358 futurelargelinkrevs = set()
2359 2359 for head in self.headrevs():
2360 2360 headlinkrev = self.linkrev(head)
2361 2361 heads[head] = headlinkrev
2362 2362 if headlinkrev >= minlink:
2363 2363 futurelargelinkrevs.add(headlinkrev)
2364 2364
2365 2365 # This algorithm involves walking down the rev graph, starting at the
2366 2366 # heads. Since the revs are topologically sorted according to linkrev,
2367 2367 # once all head linkrevs are below the minlink, we know there are
2368 2368 # no more revs that could have a linkrev greater than minlink.
2369 2369 # So we can stop walking.
2370 2370 while futurelargelinkrevs:
2371 2371 strippoint -= 1
2372 2372 linkrev = heads.pop(strippoint)
2373 2373
2374 2374 if linkrev < minlink:
2375 2375 brokenrevs.add(strippoint)
2376 2376 else:
2377 2377 futurelargelinkrevs.remove(linkrev)
2378 2378
2379 2379 for p in self.parentrevs(strippoint):
2380 2380 if p != nullrev:
2381 2381 plinkrev = self.linkrev(p)
2382 2382 heads[p] = plinkrev
2383 2383 if plinkrev >= minlink:
2384 2384 futurelargelinkrevs.add(plinkrev)
2385 2385
2386 2386 return strippoint, brokenrevs
2387 2387
2388 2388 def strip(self, minlink, transaction):
2389 2389 """truncate the revlog on the first revision with a linkrev >= minlink
2390 2390
2391 2391 This function is called when we're stripping revision minlink and
2392 2392 its descendants from the repository.
2393 2393
2394 2394 We have to remove all revisions with linkrev >= minlink, because
2395 2395 the equivalent changelog revisions will be renumbered after the
2396 2396 strip.
2397 2397
2398 2398 So we truncate the revlog on the first of these revisions, and
2399 2399 trust that the caller has saved the revisions that shouldn't be
2400 2400 removed and that it'll re-add them after this truncation.
2401 2401 """
2402 2402 if len(self) == 0:
2403 2403 return
2404 2404
2405 2405 rev, _ = self.getstrippoint(minlink)
2406 2406 if rev == len(self):
2407 2407 return
2408 2408
2409 2409 # first truncate the files on disk
2410 2410 end = self.start(rev)
2411 2411 if not self._inline:
2412 2412 transaction.add(self.datafile, end)
2413 2413 end = rev * self._io.size
2414 2414 else:
2415 2415 end += rev * self._io.size
2416 2416
2417 2417 transaction.add(self.indexfile, end)
2418 2418
2419 2419 # then reset internal state in memory to forget those revisions
2420 2420 self._cache = None
2421 2421 self._chaininfocache = {}
2422 2422 self._chunkclear()
2423 2423 for x in xrange(rev, len(self)):
2424 2424 del self.nodemap[self.node(x)]
2425 2425
2426 2426 del self.index[rev:-1]
2427 2427 self._nodepos = None
2428 2428
2429 2429 def checksize(self):
2430 2430 expected = 0
2431 2431 if len(self):
2432 2432 expected = max(0, self.end(len(self) - 1))
2433 2433
2434 2434 try:
2435 2435 with self._datafp() as f:
2436 2436 f.seek(0, 2)
2437 2437 actual = f.tell()
2438 2438 dd = actual - expected
2439 2439 except IOError as inst:
2440 2440 if inst.errno != errno.ENOENT:
2441 2441 raise
2442 2442 dd = 0
2443 2443
2444 2444 try:
2445 2445 f = self.opener(self.indexfile)
2446 2446 f.seek(0, 2)
2447 2447 actual = f.tell()
2448 2448 f.close()
2449 2449 s = self._io.size
2450 2450 i = max(0, actual // s)
2451 2451 di = actual - (i * s)
2452 2452 if self._inline:
2453 2453 databytes = 0
2454 2454 for r in self:
2455 2455 databytes += max(0, self.length(r))
2456 2456 dd = 0
2457 2457 di = actual - len(self) * s - databytes
2458 2458 except IOError as inst:
2459 2459 if inst.errno != errno.ENOENT:
2460 2460 raise
2461 2461 di = 0
2462 2462
2463 2463 return (dd, di)
2464 2464
2465 2465 def files(self):
2466 2466 res = [self.indexfile]
2467 2467 if not self._inline:
2468 2468 res.append(self.datafile)
2469 2469 return res
2470 2470
2471 2471 DELTAREUSEALWAYS = 'always'
2472 2472 DELTAREUSESAMEREVS = 'samerevs'
2473 2473 DELTAREUSENEVER = 'never'
2474 2474
2475 2475 DELTAREUSEFULLADD = 'fulladd'
2476 2476
2477 2477 DELTAREUSEALL = {'always', 'samerevs', 'never', 'fulladd'}
2478 2478
2479 2479 def clone(self, tr, destrevlog, addrevisioncb=None,
2480 2480 deltareuse=DELTAREUSESAMEREVS, aggressivemergedeltas=None):
2481 2481 """Copy this revlog to another, possibly with format changes.
2482 2482
2483 2483 The destination revlog will contain the same revisions and nodes.
2484 2484 However, it may not be bit-for-bit identical due to e.g. delta encoding
2485 2485 differences.
2486 2486
2487 2487 The ``deltareuse`` argument control how deltas from the existing revlog
2488 2488 are preserved in the destination revlog. The argument can have the
2489 2489 following values:
2490 2490
2491 2491 DELTAREUSEALWAYS
2492 2492 Deltas will always be reused (if possible), even if the destination
2493 2493 revlog would not select the same revisions for the delta. This is the
2494 2494 fastest mode of operation.
2495 2495 DELTAREUSESAMEREVS
2496 2496 Deltas will be reused if the destination revlog would pick the same
2497 2497 revisions for the delta. This mode strikes a balance between speed
2498 2498 and optimization.
2499 2499 DELTAREUSENEVER
2500 2500 Deltas will never be reused. This is the slowest mode of execution.
2501 2501 This mode can be used to recompute deltas (e.g. if the diff/delta
2502 2502 algorithm changes).
2503 2503
2504 2504 Delta computation can be slow, so the choice of delta reuse policy can
2505 2505 significantly affect run time.
2506 2506
2507 2507 The default policy (``DELTAREUSESAMEREVS``) strikes a balance between
2508 2508 two extremes. Deltas will be reused if they are appropriate. But if the
2509 2509 delta could choose a better revision, it will do so. This means if you
2510 2510 are converting a non-generaldelta revlog to a generaldelta revlog,
2511 2511 deltas will be recomputed if the delta's parent isn't a parent of the
2512 2512 revision.
2513 2513
2514 2514 In addition to the delta policy, the ``aggressivemergedeltas`` argument
2515 2515 controls whether to compute deltas against both parents for merges.
2516 2516 By default, the current default is used.
2517 2517 """
2518 2518 if deltareuse not in self.DELTAREUSEALL:
2519 2519 raise ValueError(_('value for deltareuse invalid: %s') % deltareuse)
2520 2520
2521 2521 if len(destrevlog):
2522 2522 raise ValueError(_('destination revlog is not empty'))
2523 2523
2524 2524 if getattr(self, 'filteredrevs', None):
2525 2525 raise ValueError(_('source revlog has filtered revisions'))
2526 2526 if getattr(destrevlog, 'filteredrevs', None):
2527 2527 raise ValueError(_('destination revlog has filtered revisions'))
2528 2528
2529 2529 # lazydeltabase controls whether to reuse a cached delta, if possible.
2530 2530 oldlazydeltabase = destrevlog._lazydeltabase
2531 2531 oldamd = destrevlog._aggressivemergedeltas
2532 2532
2533 2533 try:
2534 2534 if deltareuse == self.DELTAREUSEALWAYS:
2535 2535 destrevlog._lazydeltabase = True
2536 2536 elif deltareuse == self.DELTAREUSESAMEREVS:
2537 2537 destrevlog._lazydeltabase = False
2538 2538
2539 2539 destrevlog._aggressivemergedeltas = aggressivemergedeltas or oldamd
2540 2540
2541 2541 populatecachedelta = deltareuse in (self.DELTAREUSEALWAYS,
2542 2542 self.DELTAREUSESAMEREVS)
2543 2543
2544 2544 deltacomputer = _deltacomputer(destrevlog)
2545 2545 index = self.index
2546 2546 for rev in self:
2547 2547 entry = index[rev]
2548 2548
2549 2549 # Some classes override linkrev to take filtered revs into
2550 2550 # account. Use raw entry from index.
2551 2551 flags = entry[0] & 0xffff
2552 2552 linkrev = entry[4]
2553 2553 p1 = index[entry[5]][7]
2554 2554 p2 = index[entry[6]][7]
2555 2555 node = entry[7]
2556 2556
2557 2557 # (Possibly) reuse the delta from the revlog if allowed and
2558 2558 # the revlog chunk is a delta.
2559 2559 cachedelta = None
2560 2560 rawtext = None
2561 2561 if populatecachedelta:
2562 2562 dp = self.deltaparent(rev)
2563 2563 if dp != nullrev:
2564 2564 cachedelta = (dp, bytes(self._chunk(rev)))
2565 2565
2566 2566 if not cachedelta:
2567 2567 rawtext = self.revision(rev, raw=True)
2568 2568
2569 2569
2570 2570 if deltareuse == self.DELTAREUSEFULLADD:
2571 2571 destrevlog.addrevision(rawtext, tr, linkrev, p1, p2,
2572 2572 cachedelta=cachedelta,
2573 2573 node=node, flags=flags,
2574 2574 deltacomputer=deltacomputer)
2575 2575 else:
2576 2576 ifh = destrevlog.opener(destrevlog.indexfile, 'a+',
2577 2577 checkambig=False)
2578 2578 dfh = None
2579 2579 if not destrevlog._inline:
2580 2580 dfh = destrevlog.opener(destrevlog.datafile, 'a+')
2581 2581 try:
2582 2582 destrevlog._addrevision(node, rawtext, tr, linkrev, p1,
2583 2583 p2, flags, cachedelta, ifh, dfh,
2584 2584 deltacomputer=deltacomputer)
2585 2585 finally:
2586 2586 if dfh:
2587 2587 dfh.close()
2588 2588 ifh.close()
2589 2589
2590 2590 if addrevisioncb:
2591 2591 addrevisioncb(self, rev, node)
2592 2592 finally:
2593 2593 destrevlog._lazydeltabase = oldlazydeltabase
2594 2594 destrevlog._aggressivemergedeltas = oldamd
General Comments 0
You need to be logged in to leave comments. Login now