##// END OF EJS Templates
revlog: avoid string slice when decompressing u* chunks...
Gregory Szorc -
r27475:a2e2a8fa default
parent child Browse files
Show More
@@ -1,1786 +1,1786 b''
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 errno
18 18 import os
19 19 import struct
20 20 import zlib
21 21
22 22 # import stuff from node for others to import from revlog
23 23 from .node import (
24 24 bin,
25 25 hex,
26 26 nullid,
27 27 nullrev,
28 28 )
29 29 from .i18n import _
30 30 from . import (
31 31 ancestor,
32 32 error,
33 33 mdiff,
34 34 parsers,
35 35 templatefilters,
36 36 util,
37 37 )
38 38
39 39 _pack = struct.pack
40 40 _unpack = struct.unpack
41 41 _compress = zlib.compress
42 42 _decompress = zlib.decompress
43 43 _sha = util.sha1
44 44
45 45 # revlog header flags
46 46 REVLOGV0 = 0
47 47 REVLOGNG = 1
48 48 REVLOGNGINLINEDATA = (1 << 16)
49 49 REVLOGGENERALDELTA = (1 << 17)
50 50 REVLOG_DEFAULT_FLAGS = REVLOGNGINLINEDATA
51 51 REVLOG_DEFAULT_FORMAT = REVLOGNG
52 52 REVLOG_DEFAULT_VERSION = REVLOG_DEFAULT_FORMAT | REVLOG_DEFAULT_FLAGS
53 53 REVLOGNG_FLAGS = REVLOGNGINLINEDATA | REVLOGGENERALDELTA
54 54
55 55 # revlog index flags
56 56 REVIDX_ISCENSORED = (1 << 15) # revision has censor metadata, must be verified
57 57 REVIDX_DEFAULT_FLAGS = 0
58 58 REVIDX_KNOWN_FLAGS = REVIDX_ISCENSORED
59 59
60 60 # max size of revlog with inline data
61 61 _maxinline = 131072
62 62 _chunksize = 1048576
63 63
64 64 RevlogError = error.RevlogError
65 65 LookupError = error.LookupError
66 66 CensoredNodeError = error.CensoredNodeError
67 67
68 68 def getoffset(q):
69 69 return int(q >> 16)
70 70
71 71 def gettype(q):
72 72 return int(q & 0xFFFF)
73 73
74 74 def offset_type(offset, type):
75 75 return long(long(offset) << 16 | type)
76 76
77 77 _nullhash = _sha(nullid)
78 78
79 79 def hash(text, p1, p2):
80 80 """generate a hash from the given text and its parent hashes
81 81
82 82 This hash combines both the current file contents and its history
83 83 in a manner that makes it easy to distinguish nodes with the same
84 84 content in the revision graph.
85 85 """
86 86 # As of now, if one of the parent node is null, p2 is null
87 87 if p2 == nullid:
88 88 # deep copy of a hash is faster than creating one
89 89 s = _nullhash.copy()
90 90 s.update(p1)
91 91 else:
92 92 # none of the parent nodes are nullid
93 93 l = [p1, p2]
94 94 l.sort()
95 95 s = _sha(l[0])
96 96 s.update(l[1])
97 97 s.update(text)
98 98 return s.digest()
99 99
100 100 def decompress(bin):
101 101 """ decompress the given input """
102 102 if not bin:
103 103 return bin
104 104 t = bin[0]
105 105 if t == '\0':
106 106 return bin
107 107 if t == 'x':
108 108 try:
109 109 return _decompress(bin)
110 110 except zlib.error as e:
111 111 raise RevlogError(_("revlog decompress error: %s") % str(e))
112 112 if t == 'u':
113 return bin[1:]
113 return util.buffer(bin, 1)
114 114 raise RevlogError(_("unknown compression type %r") % t)
115 115
116 116 # index v0:
117 117 # 4 bytes: offset
118 118 # 4 bytes: compressed length
119 119 # 4 bytes: base rev
120 120 # 4 bytes: link rev
121 121 # 20 bytes: parent 1 nodeid
122 122 # 20 bytes: parent 2 nodeid
123 123 # 20 bytes: nodeid
124 124 indexformatv0 = ">4l20s20s20s"
125 125
126 126 class revlogoldio(object):
127 127 def __init__(self):
128 128 self.size = struct.calcsize(indexformatv0)
129 129
130 130 def parseindex(self, data, inline):
131 131 s = self.size
132 132 index = []
133 133 nodemap = {nullid: nullrev}
134 134 n = off = 0
135 135 l = len(data)
136 136 while off + s <= l:
137 137 cur = data[off:off + s]
138 138 off += s
139 139 e = _unpack(indexformatv0, cur)
140 140 # transform to revlogv1 format
141 141 e2 = (offset_type(e[0], 0), e[1], -1, e[2], e[3],
142 142 nodemap.get(e[4], nullrev), nodemap.get(e[5], nullrev), e[6])
143 143 index.append(e2)
144 144 nodemap[e[6]] = n
145 145 n += 1
146 146
147 147 # add the magic null revision at -1
148 148 index.append((0, 0, 0, -1, -1, -1, -1, nullid))
149 149
150 150 return index, nodemap, None
151 151
152 152 def packentry(self, entry, node, version, rev):
153 153 if gettype(entry[0]):
154 154 raise RevlogError(_("index entry flags need RevlogNG"))
155 155 e2 = (getoffset(entry[0]), entry[1], entry[3], entry[4],
156 156 node(entry[5]), node(entry[6]), entry[7])
157 157 return _pack(indexformatv0, *e2)
158 158
159 159 # index ng:
160 160 # 6 bytes: offset
161 161 # 2 bytes: flags
162 162 # 4 bytes: compressed length
163 163 # 4 bytes: uncompressed length
164 164 # 4 bytes: base rev
165 165 # 4 bytes: link rev
166 166 # 4 bytes: parent 1 rev
167 167 # 4 bytes: parent 2 rev
168 168 # 32 bytes: nodeid
169 169 indexformatng = ">Qiiiiii20s12x"
170 170 versionformat = ">I"
171 171
172 172 # corresponds to uncompressed length of indexformatng (2 gigs, 4-byte
173 173 # signed integer)
174 174 _maxentrysize = 0x7fffffff
175 175
176 176 class revlogio(object):
177 177 def __init__(self):
178 178 self.size = struct.calcsize(indexformatng)
179 179
180 180 def parseindex(self, data, inline):
181 181 # call the C implementation to parse the index data
182 182 index, cache = parsers.parse_index2(data, inline)
183 183 return index, getattr(index, 'nodemap', None), cache
184 184
185 185 def packentry(self, entry, node, version, rev):
186 186 p = _pack(indexformatng, *entry)
187 187 if rev == 0:
188 188 p = _pack(versionformat, version) + p[4:]
189 189 return p
190 190
191 191 class revlog(object):
192 192 """
193 193 the underlying revision storage object
194 194
195 195 A revlog consists of two parts, an index and the revision data.
196 196
197 197 The index is a file with a fixed record size containing
198 198 information on each revision, including its nodeid (hash), the
199 199 nodeids of its parents, the position and offset of its data within
200 200 the data file, and the revision it's based on. Finally, each entry
201 201 contains a linkrev entry that can serve as a pointer to external
202 202 data.
203 203
204 204 The revision data itself is a linear collection of data chunks.
205 205 Each chunk represents a revision and is usually represented as a
206 206 delta against the previous chunk. To bound lookup time, runs of
207 207 deltas are limited to about 2 times the length of the original
208 208 version data. This makes retrieval of a version proportional to
209 209 its size, or O(1) relative to the number of revisions.
210 210
211 211 Both pieces of the revlog are written to in an append-only
212 212 fashion, which means we never need to rewrite a file to insert or
213 213 remove data, and can use some simple techniques to avoid the need
214 214 for locking while reading.
215 215 """
216 216 def __init__(self, opener, indexfile):
217 217 """
218 218 create a revlog object
219 219
220 220 opener is a function that abstracts the file opening operation
221 221 and can be used to implement COW semantics or the like.
222 222 """
223 223 self.indexfile = indexfile
224 224 self.datafile = indexfile[:-2] + ".d"
225 225 self.opener = opener
226 226 # 3-tuple of (node, rev, text) for a raw revision.
227 227 self._cache = None
228 228 # 2-tuple of (rev, baserev) defining the base revision the delta chain
229 229 # begins at for a revision.
230 230 self._basecache = None
231 231 # 2-tuple of (offset, data) of raw data from the revlog at an offset.
232 232 self._chunkcache = (0, '')
233 233 # How much data to read and cache into the raw revlog data cache.
234 234 self._chunkcachesize = 65536
235 235 self._maxchainlen = None
236 236 self._aggressivemergedeltas = False
237 237 self.index = []
238 238 # Mapping of partial identifiers to full nodes.
239 239 self._pcache = {}
240 240 # Mapping of revision integer to full node.
241 241 self._nodecache = {nullid: nullrev}
242 242 self._nodepos = None
243 243
244 244 v = REVLOG_DEFAULT_VERSION
245 245 opts = getattr(opener, 'options', None)
246 246 if opts is not None:
247 247 if 'revlogv1' in opts:
248 248 if 'generaldelta' in opts:
249 249 v |= REVLOGGENERALDELTA
250 250 else:
251 251 v = 0
252 252 if 'chunkcachesize' in opts:
253 253 self._chunkcachesize = opts['chunkcachesize']
254 254 if 'maxchainlen' in opts:
255 255 self._maxchainlen = opts['maxchainlen']
256 256 if 'aggressivemergedeltas' in opts:
257 257 self._aggressivemergedeltas = opts['aggressivemergedeltas']
258 258 self._lazydeltabase = bool(opts.get('lazydeltabase', False))
259 259
260 260 if self._chunkcachesize <= 0:
261 261 raise RevlogError(_('revlog chunk cache size %r is not greater '
262 262 'than 0') % self._chunkcachesize)
263 263 elif self._chunkcachesize & (self._chunkcachesize - 1):
264 264 raise RevlogError(_('revlog chunk cache size %r is not a power '
265 265 'of 2') % self._chunkcachesize)
266 266
267 267 indexdata = ''
268 268 self._initempty = True
269 269 try:
270 270 f = self.opener(self.indexfile)
271 271 indexdata = f.read()
272 272 f.close()
273 273 if len(indexdata) > 0:
274 274 v = struct.unpack(versionformat, indexdata[:4])[0]
275 275 self._initempty = False
276 276 except IOError as inst:
277 277 if inst.errno != errno.ENOENT:
278 278 raise
279 279
280 280 self.version = v
281 281 self._inline = v & REVLOGNGINLINEDATA
282 282 self._generaldelta = v & REVLOGGENERALDELTA
283 283 flags = v & ~0xFFFF
284 284 fmt = v & 0xFFFF
285 285 if fmt == REVLOGV0 and flags:
286 286 raise RevlogError(_("index %s unknown flags %#04x for format v0")
287 287 % (self.indexfile, flags >> 16))
288 288 elif fmt == REVLOGNG and flags & ~REVLOGNG_FLAGS:
289 289 raise RevlogError(_("index %s unknown flags %#04x for revlogng")
290 290 % (self.indexfile, flags >> 16))
291 291 elif fmt > REVLOGNG:
292 292 raise RevlogError(_("index %s unknown format %d")
293 293 % (self.indexfile, fmt))
294 294
295 295 self._io = revlogio()
296 296 if self.version == REVLOGV0:
297 297 self._io = revlogoldio()
298 298 try:
299 299 d = self._io.parseindex(indexdata, self._inline)
300 300 except (ValueError, IndexError):
301 301 raise RevlogError(_("index %s is corrupted") % (self.indexfile))
302 302 self.index, nodemap, self._chunkcache = d
303 303 if nodemap is not None:
304 304 self.nodemap = self._nodecache = nodemap
305 305 if not self._chunkcache:
306 306 self._chunkclear()
307 307 # revnum -> (chain-length, sum-delta-length)
308 308 self._chaininfocache = {}
309 309
310 310 def tip(self):
311 311 return self.node(len(self.index) - 2)
312 312 def __contains__(self, rev):
313 313 return 0 <= rev < len(self)
314 314 def __len__(self):
315 315 return len(self.index) - 1
316 316 def __iter__(self):
317 317 return iter(xrange(len(self)))
318 318 def revs(self, start=0, stop=None):
319 319 """iterate over all rev in this revlog (from start to stop)"""
320 320 step = 1
321 321 if stop is not None:
322 322 if start > stop:
323 323 step = -1
324 324 stop += step
325 325 else:
326 326 stop = len(self)
327 327 return xrange(start, stop, step)
328 328
329 329 @util.propertycache
330 330 def nodemap(self):
331 331 self.rev(self.node(0))
332 332 return self._nodecache
333 333
334 334 def hasnode(self, node):
335 335 try:
336 336 self.rev(node)
337 337 return True
338 338 except KeyError:
339 339 return False
340 340
341 341 def clearcaches(self):
342 342 self._cache = None
343 343 self._basecache = None
344 344 self._chunkcache = (0, '')
345 345 self._pcache = {}
346 346
347 347 try:
348 348 self._nodecache.clearcaches()
349 349 except AttributeError:
350 350 self._nodecache = {nullid: nullrev}
351 351 self._nodepos = None
352 352
353 353 def rev(self, node):
354 354 try:
355 355 return self._nodecache[node]
356 356 except TypeError:
357 357 raise
358 358 except RevlogError:
359 359 # parsers.c radix tree lookup failed
360 360 raise LookupError(node, self.indexfile, _('no node'))
361 361 except KeyError:
362 362 # pure python cache lookup failed
363 363 n = self._nodecache
364 364 i = self.index
365 365 p = self._nodepos
366 366 if p is None:
367 367 p = len(i) - 2
368 368 for r in xrange(p, -1, -1):
369 369 v = i[r][7]
370 370 n[v] = r
371 371 if v == node:
372 372 self._nodepos = r - 1
373 373 return r
374 374 raise LookupError(node, self.indexfile, _('no node'))
375 375
376 376 def node(self, rev):
377 377 return self.index[rev][7]
378 378 def linkrev(self, rev):
379 379 return self.index[rev][4]
380 380 def parents(self, node):
381 381 i = self.index
382 382 d = i[self.rev(node)]
383 383 return i[d[5]][7], i[d[6]][7] # map revisions to nodes inline
384 384 def parentrevs(self, rev):
385 385 return self.index[rev][5:7]
386 386 def start(self, rev):
387 387 return int(self.index[rev][0] >> 16)
388 388 def end(self, rev):
389 389 return self.start(rev) + self.length(rev)
390 390 def length(self, rev):
391 391 return self.index[rev][1]
392 392 def chainbase(self, rev):
393 393 index = self.index
394 394 base = index[rev][3]
395 395 while base != rev:
396 396 rev = base
397 397 base = index[rev][3]
398 398 return base
399 399 def chainlen(self, rev):
400 400 return self._chaininfo(rev)[0]
401 401
402 402 def _chaininfo(self, rev):
403 403 chaininfocache = self._chaininfocache
404 404 if rev in chaininfocache:
405 405 return chaininfocache[rev]
406 406 index = self.index
407 407 generaldelta = self._generaldelta
408 408 iterrev = rev
409 409 e = index[iterrev]
410 410 clen = 0
411 411 compresseddeltalen = 0
412 412 while iterrev != e[3]:
413 413 clen += 1
414 414 compresseddeltalen += e[1]
415 415 if generaldelta:
416 416 iterrev = e[3]
417 417 else:
418 418 iterrev -= 1
419 419 if iterrev in chaininfocache:
420 420 t = chaininfocache[iterrev]
421 421 clen += t[0]
422 422 compresseddeltalen += t[1]
423 423 break
424 424 e = index[iterrev]
425 425 else:
426 426 # Add text length of base since decompressing that also takes
427 427 # work. For cache hits the length is already included.
428 428 compresseddeltalen += e[1]
429 429 r = (clen, compresseddeltalen)
430 430 chaininfocache[rev] = r
431 431 return r
432 432
433 433 def _deltachain(self, rev, stoprev=None):
434 434 """Obtain the delta chain for a revision.
435 435
436 436 ``stoprev`` specifies a revision to stop at. If not specified, we
437 437 stop at the base of the chain.
438 438
439 439 Returns a 2-tuple of (chain, stopped) where ``chain`` is a list of
440 440 revs in ascending order and ``stopped`` is a bool indicating whether
441 441 ``stoprev`` was hit.
442 442 """
443 443 chain = []
444 444
445 445 # Alias to prevent attribute lookup in tight loop.
446 446 index = self.index
447 447 generaldelta = self._generaldelta
448 448
449 449 iterrev = rev
450 450 e = index[iterrev]
451 451 while iterrev != e[3] and iterrev != stoprev:
452 452 chain.append(iterrev)
453 453 if generaldelta:
454 454 iterrev = e[3]
455 455 else:
456 456 iterrev -= 1
457 457 e = index[iterrev]
458 458
459 459 if iterrev == stoprev:
460 460 stopped = True
461 461 else:
462 462 chain.append(iterrev)
463 463 stopped = False
464 464
465 465 chain.reverse()
466 466 return chain, stopped
467 467
468 468 def flags(self, rev):
469 469 return self.index[rev][0] & 0xFFFF
470 470 def rawsize(self, rev):
471 471 """return the length of the uncompressed text for a given revision"""
472 472 l = self.index[rev][2]
473 473 if l >= 0:
474 474 return l
475 475
476 476 t = self.revision(self.node(rev))
477 477 return len(t)
478 478 size = rawsize
479 479
480 480 def ancestors(self, revs, stoprev=0, inclusive=False):
481 481 """Generate the ancestors of 'revs' in reverse topological order.
482 482 Does not generate revs lower than stoprev.
483 483
484 484 See the documentation for ancestor.lazyancestors for more details."""
485 485
486 486 return ancestor.lazyancestors(self.parentrevs, revs, stoprev=stoprev,
487 487 inclusive=inclusive)
488 488
489 489 def descendants(self, revs):
490 490 """Generate the descendants of 'revs' in revision order.
491 491
492 492 Yield a sequence of revision numbers starting with a child of
493 493 some rev in revs, i.e., each revision is *not* considered a
494 494 descendant of itself. Results are ordered by revision number (a
495 495 topological sort)."""
496 496 first = min(revs)
497 497 if first == nullrev:
498 498 for i in self:
499 499 yield i
500 500 return
501 501
502 502 seen = set(revs)
503 503 for i in self.revs(start=first + 1):
504 504 for x in self.parentrevs(i):
505 505 if x != nullrev and x in seen:
506 506 seen.add(i)
507 507 yield i
508 508 break
509 509
510 510 def findcommonmissing(self, common=None, heads=None):
511 511 """Return a tuple of the ancestors of common and the ancestors of heads
512 512 that are not ancestors of common. In revset terminology, we return the
513 513 tuple:
514 514
515 515 ::common, (::heads) - (::common)
516 516
517 517 The list is sorted by revision number, meaning it is
518 518 topologically sorted.
519 519
520 520 'heads' and 'common' are both lists of node IDs. If heads is
521 521 not supplied, uses all of the revlog's heads. If common is not
522 522 supplied, uses nullid."""
523 523 if common is None:
524 524 common = [nullid]
525 525 if heads is None:
526 526 heads = self.heads()
527 527
528 528 common = [self.rev(n) for n in common]
529 529 heads = [self.rev(n) for n in heads]
530 530
531 531 # we want the ancestors, but inclusive
532 532 class lazyset(object):
533 533 def __init__(self, lazyvalues):
534 534 self.addedvalues = set()
535 535 self.lazyvalues = lazyvalues
536 536
537 537 def __contains__(self, value):
538 538 return value in self.addedvalues or value in self.lazyvalues
539 539
540 540 def __iter__(self):
541 541 added = self.addedvalues
542 542 for r in added:
543 543 yield r
544 544 for r in self.lazyvalues:
545 545 if not r in added:
546 546 yield r
547 547
548 548 def add(self, value):
549 549 self.addedvalues.add(value)
550 550
551 551 def update(self, values):
552 552 self.addedvalues.update(values)
553 553
554 554 has = lazyset(self.ancestors(common))
555 555 has.add(nullrev)
556 556 has.update(common)
557 557
558 558 # take all ancestors from heads that aren't in has
559 559 missing = set()
560 560 visit = collections.deque(r for r in heads if r not in has)
561 561 while visit:
562 562 r = visit.popleft()
563 563 if r in missing:
564 564 continue
565 565 else:
566 566 missing.add(r)
567 567 for p in self.parentrevs(r):
568 568 if p not in has:
569 569 visit.append(p)
570 570 missing = list(missing)
571 571 missing.sort()
572 572 return has, [self.node(r) for r in missing]
573 573
574 574 def incrementalmissingrevs(self, common=None):
575 575 """Return an object that can be used to incrementally compute the
576 576 revision numbers of the ancestors of arbitrary sets that are not
577 577 ancestors of common. This is an ancestor.incrementalmissingancestors
578 578 object.
579 579
580 580 'common' is a list of revision numbers. If common is not supplied, uses
581 581 nullrev.
582 582 """
583 583 if common is None:
584 584 common = [nullrev]
585 585
586 586 return ancestor.incrementalmissingancestors(self.parentrevs, common)
587 587
588 588 def findmissingrevs(self, common=None, heads=None):
589 589 """Return the revision numbers of the ancestors of heads that
590 590 are not ancestors of common.
591 591
592 592 More specifically, return a list of revision numbers corresponding to
593 593 nodes N such that every N satisfies the following constraints:
594 594
595 595 1. N is an ancestor of some node in 'heads'
596 596 2. N is not an ancestor of any node in 'common'
597 597
598 598 The list is sorted by revision number, meaning it is
599 599 topologically sorted.
600 600
601 601 'heads' and 'common' are both lists of revision numbers. If heads is
602 602 not supplied, uses all of the revlog's heads. If common is not
603 603 supplied, uses nullid."""
604 604 if common is None:
605 605 common = [nullrev]
606 606 if heads is None:
607 607 heads = self.headrevs()
608 608
609 609 inc = self.incrementalmissingrevs(common=common)
610 610 return inc.missingancestors(heads)
611 611
612 612 def findmissing(self, common=None, heads=None):
613 613 """Return the ancestors of heads that are not ancestors of common.
614 614
615 615 More specifically, return a list of nodes N such that every N
616 616 satisfies the following constraints:
617 617
618 618 1. N is an ancestor of some node in 'heads'
619 619 2. N is not an ancestor of any node in 'common'
620 620
621 621 The list is sorted by revision number, meaning it is
622 622 topologically sorted.
623 623
624 624 'heads' and 'common' are both lists of node IDs. If heads is
625 625 not supplied, uses all of the revlog's heads. If common is not
626 626 supplied, uses nullid."""
627 627 if common is None:
628 628 common = [nullid]
629 629 if heads is None:
630 630 heads = self.heads()
631 631
632 632 common = [self.rev(n) for n in common]
633 633 heads = [self.rev(n) for n in heads]
634 634
635 635 inc = self.incrementalmissingrevs(common=common)
636 636 return [self.node(r) for r in inc.missingancestors(heads)]
637 637
638 638 def nodesbetween(self, roots=None, heads=None):
639 639 """Return a topological path from 'roots' to 'heads'.
640 640
641 641 Return a tuple (nodes, outroots, outheads) where 'nodes' is a
642 642 topologically sorted list of all nodes N that satisfy both of
643 643 these constraints:
644 644
645 645 1. N is a descendant of some node in 'roots'
646 646 2. N is an ancestor of some node in 'heads'
647 647
648 648 Every node is considered to be both a descendant and an ancestor
649 649 of itself, so every reachable node in 'roots' and 'heads' will be
650 650 included in 'nodes'.
651 651
652 652 'outroots' is the list of reachable nodes in 'roots', i.e., the
653 653 subset of 'roots' that is returned in 'nodes'. Likewise,
654 654 'outheads' is the subset of 'heads' that is also in 'nodes'.
655 655
656 656 'roots' and 'heads' are both lists of node IDs. If 'roots' is
657 657 unspecified, uses nullid as the only root. If 'heads' is
658 658 unspecified, uses list of all of the revlog's heads."""
659 659 nonodes = ([], [], [])
660 660 if roots is not None:
661 661 roots = list(roots)
662 662 if not roots:
663 663 return nonodes
664 664 lowestrev = min([self.rev(n) for n in roots])
665 665 else:
666 666 roots = [nullid] # Everybody's a descendant of nullid
667 667 lowestrev = nullrev
668 668 if (lowestrev == nullrev) and (heads is None):
669 669 # We want _all_ the nodes!
670 670 return ([self.node(r) for r in self], [nullid], list(self.heads()))
671 671 if heads is None:
672 672 # All nodes are ancestors, so the latest ancestor is the last
673 673 # node.
674 674 highestrev = len(self) - 1
675 675 # Set ancestors to None to signal that every node is an ancestor.
676 676 ancestors = None
677 677 # Set heads to an empty dictionary for later discovery of heads
678 678 heads = {}
679 679 else:
680 680 heads = list(heads)
681 681 if not heads:
682 682 return nonodes
683 683 ancestors = set()
684 684 # Turn heads into a dictionary so we can remove 'fake' heads.
685 685 # Also, later we will be using it to filter out the heads we can't
686 686 # find from roots.
687 687 heads = dict.fromkeys(heads, False)
688 688 # Start at the top and keep marking parents until we're done.
689 689 nodestotag = set(heads)
690 690 # Remember where the top was so we can use it as a limit later.
691 691 highestrev = max([self.rev(n) for n in nodestotag])
692 692 while nodestotag:
693 693 # grab a node to tag
694 694 n = nodestotag.pop()
695 695 # Never tag nullid
696 696 if n == nullid:
697 697 continue
698 698 # A node's revision number represents its place in a
699 699 # topologically sorted list of nodes.
700 700 r = self.rev(n)
701 701 if r >= lowestrev:
702 702 if n not in ancestors:
703 703 # If we are possibly a descendant of one of the roots
704 704 # and we haven't already been marked as an ancestor
705 705 ancestors.add(n) # Mark as ancestor
706 706 # Add non-nullid parents to list of nodes to tag.
707 707 nodestotag.update([p for p in self.parents(n) if
708 708 p != nullid])
709 709 elif n in heads: # We've seen it before, is it a fake head?
710 710 # So it is, real heads should not be the ancestors of
711 711 # any other heads.
712 712 heads.pop(n)
713 713 if not ancestors:
714 714 return nonodes
715 715 # Now that we have our set of ancestors, we want to remove any
716 716 # roots that are not ancestors.
717 717
718 718 # If one of the roots was nullid, everything is included anyway.
719 719 if lowestrev > nullrev:
720 720 # But, since we weren't, let's recompute the lowest rev to not
721 721 # include roots that aren't ancestors.
722 722
723 723 # Filter out roots that aren't ancestors of heads
724 724 roots = [n for n in roots if n in ancestors]
725 725 # Recompute the lowest revision
726 726 if roots:
727 727 lowestrev = min([self.rev(n) for n in roots])
728 728 else:
729 729 # No more roots? Return empty list
730 730 return nonodes
731 731 else:
732 732 # We are descending from nullid, and don't need to care about
733 733 # any other roots.
734 734 lowestrev = nullrev
735 735 roots = [nullid]
736 736 # Transform our roots list into a set.
737 737 descendants = set(roots)
738 738 # Also, keep the original roots so we can filter out roots that aren't
739 739 # 'real' roots (i.e. are descended from other roots).
740 740 roots = descendants.copy()
741 741 # Our topologically sorted list of output nodes.
742 742 orderedout = []
743 743 # Don't start at nullid since we don't want nullid in our output list,
744 744 # and if nullid shows up in descendants, empty parents will look like
745 745 # they're descendants.
746 746 for r in self.revs(start=max(lowestrev, 0), stop=highestrev + 1):
747 747 n = self.node(r)
748 748 isdescendant = False
749 749 if lowestrev == nullrev: # Everybody is a descendant of nullid
750 750 isdescendant = True
751 751 elif n in descendants:
752 752 # n is already a descendant
753 753 isdescendant = True
754 754 # This check only needs to be done here because all the roots
755 755 # will start being marked is descendants before the loop.
756 756 if n in roots:
757 757 # If n was a root, check if it's a 'real' root.
758 758 p = tuple(self.parents(n))
759 759 # If any of its parents are descendants, it's not a root.
760 760 if (p[0] in descendants) or (p[1] in descendants):
761 761 roots.remove(n)
762 762 else:
763 763 p = tuple(self.parents(n))
764 764 # A node is a descendant if either of its parents are
765 765 # descendants. (We seeded the dependents list with the roots
766 766 # up there, remember?)
767 767 if (p[0] in descendants) or (p[1] in descendants):
768 768 descendants.add(n)
769 769 isdescendant = True
770 770 if isdescendant and ((ancestors is None) or (n in ancestors)):
771 771 # Only include nodes that are both descendants and ancestors.
772 772 orderedout.append(n)
773 773 if (ancestors is not None) and (n in heads):
774 774 # We're trying to figure out which heads are reachable
775 775 # from roots.
776 776 # Mark this head as having been reached
777 777 heads[n] = True
778 778 elif ancestors is None:
779 779 # Otherwise, we're trying to discover the heads.
780 780 # Assume this is a head because if it isn't, the next step
781 781 # will eventually remove it.
782 782 heads[n] = True
783 783 # But, obviously its parents aren't.
784 784 for p in self.parents(n):
785 785 heads.pop(p, None)
786 786 heads = [n for n, flag in heads.iteritems() if flag]
787 787 roots = list(roots)
788 788 assert orderedout
789 789 assert roots
790 790 assert heads
791 791 return (orderedout, roots, heads)
792 792
793 793 def headrevs(self):
794 794 try:
795 795 return self.index.headrevs()
796 796 except AttributeError:
797 797 return self._headrevs()
798 798
799 799 def computephases(self, roots):
800 800 return self.index.computephasesmapsets(roots)
801 801
802 802 def _headrevs(self):
803 803 count = len(self)
804 804 if not count:
805 805 return [nullrev]
806 806 # we won't iter over filtered rev so nobody is a head at start
807 807 ishead = [0] * (count + 1)
808 808 index = self.index
809 809 for r in self:
810 810 ishead[r] = 1 # I may be an head
811 811 e = index[r]
812 812 ishead[e[5]] = ishead[e[6]] = 0 # my parent are not
813 813 return [r for r, val in enumerate(ishead) if val]
814 814
815 815 def heads(self, start=None, stop=None):
816 816 """return the list of all nodes that have no children
817 817
818 818 if start is specified, only heads that are descendants of
819 819 start will be returned
820 820 if stop is specified, it will consider all the revs from stop
821 821 as if they had no children
822 822 """
823 823 if start is None and stop is None:
824 824 if not len(self):
825 825 return [nullid]
826 826 return [self.node(r) for r in self.headrevs()]
827 827
828 828 if start is None:
829 829 start = nullid
830 830 if stop is None:
831 831 stop = []
832 832 stoprevs = set([self.rev(n) for n in stop])
833 833 startrev = self.rev(start)
834 834 reachable = set((startrev,))
835 835 heads = set((startrev,))
836 836
837 837 parentrevs = self.parentrevs
838 838 for r in self.revs(start=startrev + 1):
839 839 for p in parentrevs(r):
840 840 if p in reachable:
841 841 if r not in stoprevs:
842 842 reachable.add(r)
843 843 heads.add(r)
844 844 if p in heads and p not in stoprevs:
845 845 heads.remove(p)
846 846
847 847 return [self.node(r) for r in heads]
848 848
849 849 def children(self, node):
850 850 """find the children of a given node"""
851 851 c = []
852 852 p = self.rev(node)
853 853 for r in self.revs(start=p + 1):
854 854 prevs = [pr for pr in self.parentrevs(r) if pr != nullrev]
855 855 if prevs:
856 856 for pr in prevs:
857 857 if pr == p:
858 858 c.append(self.node(r))
859 859 elif p == nullrev:
860 860 c.append(self.node(r))
861 861 return c
862 862
863 863 def descendant(self, start, end):
864 864 if start == nullrev:
865 865 return True
866 866 for i in self.descendants([start]):
867 867 if i == end:
868 868 return True
869 869 elif i > end:
870 870 break
871 871 return False
872 872
873 873 def commonancestorsheads(self, a, b):
874 874 """calculate all the heads of the common ancestors of nodes a and b"""
875 875 a, b = self.rev(a), self.rev(b)
876 876 try:
877 877 ancs = self.index.commonancestorsheads(a, b)
878 878 except (AttributeError, OverflowError): # C implementation failed
879 879 ancs = ancestor.commonancestorsheads(self.parentrevs, a, b)
880 880 return map(self.node, ancs)
881 881
882 882 def isancestor(self, a, b):
883 883 """return True if node a is an ancestor of node b
884 884
885 885 The implementation of this is trivial but the use of
886 886 commonancestorsheads is not."""
887 887 return a in self.commonancestorsheads(a, b)
888 888
889 889 def ancestor(self, a, b):
890 890 """calculate the "best" common ancestor of nodes a and b"""
891 891
892 892 a, b = self.rev(a), self.rev(b)
893 893 try:
894 894 ancs = self.index.ancestors(a, b)
895 895 except (AttributeError, OverflowError):
896 896 ancs = ancestor.ancestors(self.parentrevs, a, b)
897 897 if ancs:
898 898 # choose a consistent winner when there's a tie
899 899 return min(map(self.node, ancs))
900 900 return nullid
901 901
902 902 def _match(self, id):
903 903 if isinstance(id, int):
904 904 # rev
905 905 return self.node(id)
906 906 if len(id) == 20:
907 907 # possibly a binary node
908 908 # odds of a binary node being all hex in ASCII are 1 in 10**25
909 909 try:
910 910 node = id
911 911 self.rev(node) # quick search the index
912 912 return node
913 913 except LookupError:
914 914 pass # may be partial hex id
915 915 try:
916 916 # str(rev)
917 917 rev = int(id)
918 918 if str(rev) != id:
919 919 raise ValueError
920 920 if rev < 0:
921 921 rev = len(self) + rev
922 922 if rev < 0 or rev >= len(self):
923 923 raise ValueError
924 924 return self.node(rev)
925 925 except (ValueError, OverflowError):
926 926 pass
927 927 if len(id) == 40:
928 928 try:
929 929 # a full hex nodeid?
930 930 node = bin(id)
931 931 self.rev(node)
932 932 return node
933 933 except (TypeError, LookupError):
934 934 pass
935 935
936 936 def _partialmatch(self, id):
937 937 try:
938 938 n = self.index.partialmatch(id)
939 939 if n and self.hasnode(n):
940 940 return n
941 941 return None
942 942 except RevlogError:
943 943 # parsers.c radix tree lookup gave multiple matches
944 944 # fall through to slow path that filters hidden revisions
945 945 pass
946 946 except (AttributeError, ValueError):
947 947 # we are pure python, or key was too short to search radix tree
948 948 pass
949 949
950 950 if id in self._pcache:
951 951 return self._pcache[id]
952 952
953 953 if len(id) < 40:
954 954 try:
955 955 # hex(node)[:...]
956 956 l = len(id) // 2 # grab an even number of digits
957 957 prefix = bin(id[:l * 2])
958 958 nl = [e[7] for e in self.index if e[7].startswith(prefix)]
959 959 nl = [n for n in nl if hex(n).startswith(id) and
960 960 self.hasnode(n)]
961 961 if len(nl) > 0:
962 962 if len(nl) == 1:
963 963 self._pcache[id] = nl[0]
964 964 return nl[0]
965 965 raise LookupError(id, self.indexfile,
966 966 _('ambiguous identifier'))
967 967 return None
968 968 except TypeError:
969 969 pass
970 970
971 971 def lookup(self, id):
972 972 """locate a node based on:
973 973 - revision number or str(revision number)
974 974 - nodeid or subset of hex nodeid
975 975 """
976 976 n = self._match(id)
977 977 if n is not None:
978 978 return n
979 979 n = self._partialmatch(id)
980 980 if n:
981 981 return n
982 982
983 983 raise LookupError(id, self.indexfile, _('no match found'))
984 984
985 985 def cmp(self, node, text):
986 986 """compare text with a given file revision
987 987
988 988 returns True if text is different than what is stored.
989 989 """
990 990 p1, p2 = self.parents(node)
991 991 return hash(text, p1, p2) != node
992 992
993 993 def _addchunk(self, offset, data):
994 994 """Add a segment to the revlog cache.
995 995
996 996 Accepts an absolute offset and the data that is at that location.
997 997 """
998 998 o, d = self._chunkcache
999 999 # try to add to existing cache
1000 1000 if o + len(d) == offset and len(d) + len(data) < _chunksize:
1001 1001 self._chunkcache = o, d + data
1002 1002 else:
1003 1003 self._chunkcache = offset, data
1004 1004
1005 1005 def _loadchunk(self, offset, length, df=None):
1006 1006 """Load a segment of raw data from the revlog.
1007 1007
1008 1008 Accepts an absolute offset, length to read, and an optional existing
1009 1009 file handle to read from.
1010 1010
1011 1011 If an existing file handle is passed, it will be seeked and the
1012 1012 original seek position will NOT be restored.
1013 1013
1014 1014 Returns a str or buffer of raw byte data.
1015 1015 """
1016 1016 if df is not None:
1017 1017 closehandle = False
1018 1018 else:
1019 1019 if self._inline:
1020 1020 df = self.opener(self.indexfile)
1021 1021 else:
1022 1022 df = self.opener(self.datafile)
1023 1023 closehandle = True
1024 1024
1025 1025 # Cache data both forward and backward around the requested
1026 1026 # data, in a fixed size window. This helps speed up operations
1027 1027 # involving reading the revlog backwards.
1028 1028 cachesize = self._chunkcachesize
1029 1029 realoffset = offset & ~(cachesize - 1)
1030 1030 reallength = (((offset + length + cachesize) & ~(cachesize - 1))
1031 1031 - realoffset)
1032 1032 df.seek(realoffset)
1033 1033 d = df.read(reallength)
1034 1034 if closehandle:
1035 1035 df.close()
1036 1036 self._addchunk(realoffset, d)
1037 1037 if offset != realoffset or reallength != length:
1038 1038 return util.buffer(d, offset - realoffset, length)
1039 1039 return d
1040 1040
1041 1041 def _getchunk(self, offset, length, df=None):
1042 1042 """Obtain a segment of raw data from the revlog.
1043 1043
1044 1044 Accepts an absolute offset, length of bytes to obtain, and an
1045 1045 optional file handle to the already-opened revlog. If the file
1046 1046 handle is used, it's original seek position will not be preserved.
1047 1047
1048 1048 Requests for data may be returned from a cache.
1049 1049
1050 1050 Returns a str or a buffer instance of raw byte data.
1051 1051 """
1052 1052 o, d = self._chunkcache
1053 1053 l = len(d)
1054 1054
1055 1055 # is it in the cache?
1056 1056 cachestart = offset - o
1057 1057 cacheend = cachestart + length
1058 1058 if cachestart >= 0 and cacheend <= l:
1059 1059 if cachestart == 0 and cacheend == l:
1060 1060 return d # avoid a copy
1061 1061 return util.buffer(d, cachestart, cacheend - cachestart)
1062 1062
1063 1063 return self._loadchunk(offset, length, df=df)
1064 1064
1065 1065 def _chunkraw(self, startrev, endrev, df=None):
1066 1066 """Obtain a segment of raw data corresponding to a range of revisions.
1067 1067
1068 1068 Accepts the start and end revisions and an optional already-open
1069 1069 file handle to be used for reading. If the file handle is read, its
1070 1070 seek position will not be preserved.
1071 1071
1072 1072 Requests for data may be satisfied by a cache.
1073 1073
1074 1074 Returns a str or a buffer instance of raw byte data. Callers will
1075 1075 need to call ``self.start(rev)`` and ``self.length()`` to determine
1076 1076 where each revision's data begins and ends.
1077 1077 """
1078 1078 start = self.start(startrev)
1079 1079 end = self.end(endrev)
1080 1080 if self._inline:
1081 1081 start += (startrev + 1) * self._io.size
1082 1082 end += (endrev + 1) * self._io.size
1083 1083 length = end - start
1084 1084 return self._getchunk(start, length, df=df)
1085 1085
1086 1086 def _chunk(self, rev, df=None):
1087 1087 """Obtain a single decompressed chunk for a revision.
1088 1088
1089 1089 Accepts an integer revision and an optional already-open file handle
1090 1090 to be used for reading. If used, the seek position of the file will not
1091 1091 be preserved.
1092 1092
1093 1093 Returns a str holding uncompressed data for the requested revision.
1094 1094 """
1095 1095 return decompress(self._chunkraw(rev, rev, df=df))
1096 1096
1097 1097 def _chunks(self, revs, df=None):
1098 1098 """Obtain decompressed chunks for the specified revisions.
1099 1099
1100 1100 Accepts an iterable of numeric revisions that are assumed to be in
1101 1101 ascending order. Also accepts an optional already-open file handle
1102 1102 to be used for reading. If used, the seek position of the file will
1103 1103 not be preserved.
1104 1104
1105 1105 This function is similar to calling ``self._chunk()`` multiple times,
1106 1106 but is faster.
1107 1107
1108 1108 Returns a list with decompressed data for each requested revision.
1109 1109 """
1110 1110 if not revs:
1111 1111 return []
1112 1112 start = self.start
1113 1113 length = self.length
1114 1114 inline = self._inline
1115 1115 iosize = self._io.size
1116 1116 buffer = util.buffer
1117 1117
1118 1118 l = []
1119 1119 ladd = l.append
1120 1120
1121 1121 # preload the cache
1122 1122 try:
1123 1123 while True:
1124 1124 # ensure that the cache doesn't change out from under us
1125 1125 _cache = self._chunkcache
1126 1126 self._chunkraw(revs[0], revs[-1], df=df)
1127 1127 if _cache == self._chunkcache:
1128 1128 break
1129 1129 offset, data = _cache
1130 1130 except OverflowError:
1131 1131 # issue4215 - we can't cache a run of chunks greater than
1132 1132 # 2G on Windows
1133 1133 return [self._chunk(rev, df=df) for rev in revs]
1134 1134
1135 1135 for rev in revs:
1136 1136 chunkstart = start(rev)
1137 1137 if inline:
1138 1138 chunkstart += (rev + 1) * iosize
1139 1139 chunklength = length(rev)
1140 1140 ladd(decompress(buffer(data, chunkstart - offset, chunklength)))
1141 1141
1142 1142 return l
1143 1143
1144 1144 def _chunkclear(self):
1145 1145 """Clear the raw chunk cache."""
1146 1146 self._chunkcache = (0, '')
1147 1147
1148 1148 def deltaparent(self, rev):
1149 1149 """return deltaparent of the given revision"""
1150 1150 base = self.index[rev][3]
1151 1151 if base == rev:
1152 1152 return nullrev
1153 1153 elif self._generaldelta:
1154 1154 return base
1155 1155 else:
1156 1156 return rev - 1
1157 1157
1158 1158 def revdiff(self, rev1, rev2):
1159 1159 """return or calculate a delta between two revisions"""
1160 1160 if rev1 != nullrev and self.deltaparent(rev2) == rev1:
1161 1161 return str(self._chunk(rev2))
1162 1162
1163 1163 return mdiff.textdiff(self.revision(rev1),
1164 1164 self.revision(rev2))
1165 1165
1166 1166 def revision(self, nodeorrev, _df=None):
1167 1167 """return an uncompressed revision of a given node or revision
1168 1168 number.
1169 1169
1170 1170 _df is an existing file handle to read from. It is meant to only be
1171 1171 used internally.
1172 1172 """
1173 1173 if isinstance(nodeorrev, int):
1174 1174 rev = nodeorrev
1175 1175 node = self.node(rev)
1176 1176 else:
1177 1177 node = nodeorrev
1178 1178 rev = None
1179 1179
1180 1180 cachedrev = None
1181 1181 if node == nullid:
1182 1182 return ""
1183 1183 if self._cache:
1184 1184 if self._cache[0] == node:
1185 1185 return self._cache[2]
1186 1186 cachedrev = self._cache[1]
1187 1187
1188 1188 # look up what we need to read
1189 1189 text = None
1190 1190 if rev is None:
1191 1191 rev = self.rev(node)
1192 1192
1193 1193 # check rev flags
1194 1194 if self.flags(rev) & ~REVIDX_KNOWN_FLAGS:
1195 1195 raise RevlogError(_('incompatible revision flag %x') %
1196 1196 (self.flags(rev) & ~REVIDX_KNOWN_FLAGS))
1197 1197
1198 1198 chain, stopped = self._deltachain(rev, stoprev=cachedrev)
1199 1199 if stopped:
1200 1200 text = self._cache[2]
1201 1201
1202 1202 # drop cache to save memory
1203 1203 self._cache = None
1204 1204
1205 1205 bins = self._chunks(chain, df=_df)
1206 1206 if text is None:
1207 1207 text = str(bins[0])
1208 1208 bins = bins[1:]
1209 1209
1210 1210 text = mdiff.patches(text, bins)
1211 1211
1212 1212 text = self._checkhash(text, node, rev)
1213 1213
1214 1214 self._cache = (node, rev, text)
1215 1215 return text
1216 1216
1217 1217 def hash(self, text, p1, p2):
1218 1218 """Compute a node hash.
1219 1219
1220 1220 Available as a function so that subclasses can replace the hash
1221 1221 as needed.
1222 1222 """
1223 1223 return hash(text, p1, p2)
1224 1224
1225 1225 def _checkhash(self, text, node, rev):
1226 1226 p1, p2 = self.parents(node)
1227 1227 self.checkhash(text, p1, p2, node, rev)
1228 1228 return text
1229 1229
1230 1230 def checkhash(self, text, p1, p2, node, rev=None):
1231 1231 if node != self.hash(text, p1, p2):
1232 1232 revornode = rev
1233 1233 if revornode is None:
1234 1234 revornode = templatefilters.short(hex(node))
1235 1235 raise RevlogError(_("integrity check failed on %s:%s")
1236 1236 % (self.indexfile, revornode))
1237 1237
1238 1238 def checkinlinesize(self, tr, fp=None):
1239 1239 """Check if the revlog is too big for inline and convert if so.
1240 1240
1241 1241 This should be called after revisions are added to the revlog. If the
1242 1242 revlog has grown too large to be an inline revlog, it will convert it
1243 1243 to use multiple index and data files.
1244 1244 """
1245 1245 if not self._inline or (self.start(-2) + self.length(-2)) < _maxinline:
1246 1246 return
1247 1247
1248 1248 trinfo = tr.find(self.indexfile)
1249 1249 if trinfo is None:
1250 1250 raise RevlogError(_("%s not found in the transaction")
1251 1251 % self.indexfile)
1252 1252
1253 1253 trindex = trinfo[2]
1254 1254 if trindex is not None:
1255 1255 dataoff = self.start(trindex)
1256 1256 else:
1257 1257 # revlog was stripped at start of transaction, use all leftover data
1258 1258 trindex = len(self) - 1
1259 1259 dataoff = self.end(-2)
1260 1260
1261 1261 tr.add(self.datafile, dataoff)
1262 1262
1263 1263 if fp:
1264 1264 fp.flush()
1265 1265 fp.close()
1266 1266
1267 1267 df = self.opener(self.datafile, 'w')
1268 1268 try:
1269 1269 for r in self:
1270 1270 df.write(self._chunkraw(r, r))
1271 1271 finally:
1272 1272 df.close()
1273 1273
1274 1274 fp = self.opener(self.indexfile, 'w', atomictemp=True)
1275 1275 self.version &= ~(REVLOGNGINLINEDATA)
1276 1276 self._inline = False
1277 1277 for i in self:
1278 1278 e = self._io.packentry(self.index[i], self.node, self.version, i)
1279 1279 fp.write(e)
1280 1280
1281 1281 # if we don't call close, the temp file will never replace the
1282 1282 # real index
1283 1283 fp.close()
1284 1284
1285 1285 tr.replace(self.indexfile, trindex * self._io.size)
1286 1286 self._chunkclear()
1287 1287
1288 1288 def addrevision(self, text, transaction, link, p1, p2, cachedelta=None,
1289 1289 node=None):
1290 1290 """add a revision to the log
1291 1291
1292 1292 text - the revision data to add
1293 1293 transaction - the transaction object used for rollback
1294 1294 link - the linkrev data to add
1295 1295 p1, p2 - the parent nodeids of the revision
1296 1296 cachedelta - an optional precomputed delta
1297 1297 node - nodeid of revision; typically node is not specified, and it is
1298 1298 computed by default as hash(text, p1, p2), however subclasses might
1299 1299 use different hashing method (and override checkhash() in such case)
1300 1300 """
1301 1301 if link == nullrev:
1302 1302 raise RevlogError(_("attempted to add linkrev -1 to %s")
1303 1303 % self.indexfile)
1304 1304
1305 1305 if len(text) > _maxentrysize:
1306 1306 raise RevlogError(
1307 1307 _("%s: size of %d bytes exceeds maximum revlog storage of 2GiB")
1308 1308 % (self.indexfile, len(text)))
1309 1309
1310 1310 node = node or self.hash(text, p1, p2)
1311 1311 if node in self.nodemap:
1312 1312 return node
1313 1313
1314 1314 dfh = None
1315 1315 if not self._inline:
1316 1316 dfh = self.opener(self.datafile, "a+")
1317 1317 ifh = self.opener(self.indexfile, "a+")
1318 1318 try:
1319 1319 return self._addrevision(node, text, transaction, link, p1, p2,
1320 1320 REVIDX_DEFAULT_FLAGS, cachedelta, ifh, dfh)
1321 1321 finally:
1322 1322 if dfh:
1323 1323 dfh.close()
1324 1324 ifh.close()
1325 1325
1326 1326 def compress(self, text):
1327 1327 """ generate a possibly-compressed representation of text """
1328 1328 if not text:
1329 1329 return ("", text)
1330 1330 l = len(text)
1331 1331 bin = None
1332 1332 if l < 44:
1333 1333 pass
1334 1334 elif l > 1000000:
1335 1335 # zlib makes an internal copy, thus doubling memory usage for
1336 1336 # large files, so lets do this in pieces
1337 1337 z = zlib.compressobj()
1338 1338 p = []
1339 1339 pos = 0
1340 1340 while pos < l:
1341 1341 pos2 = pos + 2**20
1342 1342 p.append(z.compress(text[pos:pos2]))
1343 1343 pos = pos2
1344 1344 p.append(z.flush())
1345 1345 if sum(map(len, p)) < l:
1346 1346 bin = "".join(p)
1347 1347 else:
1348 1348 bin = _compress(text)
1349 1349 if bin is None or len(bin) > l:
1350 1350 if text[0] == '\0':
1351 1351 return ("", text)
1352 1352 return ('u', text)
1353 1353 return ("", bin)
1354 1354
1355 1355 def _isgooddelta(self, d, textlen):
1356 1356 """Returns True if the given delta is good. Good means that it is within
1357 1357 the disk span, disk size, and chain length bounds that we know to be
1358 1358 performant."""
1359 1359 if d is None:
1360 1360 return False
1361 1361
1362 1362 # - 'dist' is the distance from the base revision -- bounding it limits
1363 1363 # the amount of I/O we need to do.
1364 1364 # - 'compresseddeltalen' is the sum of the total size of deltas we need
1365 1365 # to apply -- bounding it limits the amount of CPU we consume.
1366 1366 dist, l, data, base, chainbase, chainlen, compresseddeltalen = d
1367 1367 if (dist > textlen * 4 or l > textlen or
1368 1368 compresseddeltalen > textlen * 2 or
1369 1369 (self._maxchainlen and chainlen > self._maxchainlen)):
1370 1370 return False
1371 1371
1372 1372 return True
1373 1373
1374 1374 def _addrevision(self, node, text, transaction, link, p1, p2, flags,
1375 1375 cachedelta, ifh, dfh, alwayscache=False):
1376 1376 """internal function to add revisions to the log
1377 1377
1378 1378 see addrevision for argument descriptions.
1379 1379 invariants:
1380 1380 - text is optional (can be None); if not set, cachedelta must be set.
1381 1381 if both are set, they must correspond to each other.
1382 1382 """
1383 1383 btext = [text]
1384 1384 def buildtext():
1385 1385 if btext[0] is not None:
1386 1386 return btext[0]
1387 1387 baserev = cachedelta[0]
1388 1388 delta = cachedelta[1]
1389 1389 # special case deltas which replace entire base; no need to decode
1390 1390 # base revision. this neatly avoids censored bases, which throw when
1391 1391 # they're decoded.
1392 1392 hlen = struct.calcsize(">lll")
1393 1393 if delta[:hlen] == mdiff.replacediffheader(self.rawsize(baserev),
1394 1394 len(delta) - hlen):
1395 1395 btext[0] = delta[hlen:]
1396 1396 else:
1397 1397 if self._inline:
1398 1398 fh = ifh
1399 1399 else:
1400 1400 fh = dfh
1401 1401 basetext = self.revision(self.node(baserev), _df=fh)
1402 1402 btext[0] = mdiff.patch(basetext, delta)
1403 1403 try:
1404 1404 self.checkhash(btext[0], p1, p2, node)
1405 1405 if flags & REVIDX_ISCENSORED:
1406 1406 raise RevlogError(_('node %s is not censored') % node)
1407 1407 except CensoredNodeError:
1408 1408 # must pass the censored index flag to add censored revisions
1409 1409 if not flags & REVIDX_ISCENSORED:
1410 1410 raise
1411 1411 return btext[0]
1412 1412
1413 1413 def builddelta(rev):
1414 1414 # can we use the cached delta?
1415 1415 if cachedelta and cachedelta[0] == rev:
1416 1416 delta = cachedelta[1]
1417 1417 else:
1418 1418 t = buildtext()
1419 1419 if self.iscensored(rev):
1420 1420 # deltas based on a censored revision must replace the
1421 1421 # full content in one patch, so delta works everywhere
1422 1422 header = mdiff.replacediffheader(self.rawsize(rev), len(t))
1423 1423 delta = header + t
1424 1424 else:
1425 1425 if self._inline:
1426 1426 fh = ifh
1427 1427 else:
1428 1428 fh = dfh
1429 1429 ptext = self.revision(self.node(rev), _df=fh)
1430 1430 delta = mdiff.textdiff(ptext, t)
1431 1431 data = self.compress(delta)
1432 1432 l = len(data[1]) + len(data[0])
1433 1433 if basecache[0] == rev:
1434 1434 chainbase = basecache[1]
1435 1435 else:
1436 1436 chainbase = self.chainbase(rev)
1437 1437 dist = l + offset - self.start(chainbase)
1438 1438 if self._generaldelta:
1439 1439 base = rev
1440 1440 else:
1441 1441 base = chainbase
1442 1442 chainlen, compresseddeltalen = self._chaininfo(rev)
1443 1443 chainlen += 1
1444 1444 compresseddeltalen += l
1445 1445 return dist, l, data, base, chainbase, chainlen, compresseddeltalen
1446 1446
1447 1447 curr = len(self)
1448 1448 prev = curr - 1
1449 1449 base = chainbase = curr
1450 1450 offset = self.end(prev)
1451 1451 delta = None
1452 1452 if self._basecache is None:
1453 1453 self._basecache = (prev, self.chainbase(prev))
1454 1454 basecache = self._basecache
1455 1455 p1r, p2r = self.rev(p1), self.rev(p2)
1456 1456
1457 1457 # full versions are inserted when the needed deltas
1458 1458 # become comparable to the uncompressed text
1459 1459 if text is None:
1460 1460 textlen = mdiff.patchedsize(self.rawsize(cachedelta[0]),
1461 1461 cachedelta[1])
1462 1462 else:
1463 1463 textlen = len(text)
1464 1464
1465 1465 # should we try to build a delta?
1466 1466 if prev != nullrev:
1467 1467 tested = set()
1468 1468 if cachedelta and self._generaldelta and self._lazydeltabase:
1469 1469 # Assume what we received from the server is a good choice
1470 1470 # build delta will reuse the cache
1471 1471 candidatedelta = builddelta(cachedelta[0])
1472 1472 tested.add(cachedelta[0])
1473 1473 if self._isgooddelta(candidatedelta, textlen):
1474 1474 delta = candidatedelta
1475 1475 if delta is None and self._generaldelta:
1476 1476 # exclude already lazy tested base if any
1477 1477 parents = [p for p in (p1r, p2r)
1478 1478 if p != nullrev and p not in tested]
1479 1479 if parents and not self._aggressivemergedeltas:
1480 1480 # Pick whichever parent is closer to us (to minimize the
1481 1481 # chance of having to build a fulltext).
1482 1482 parents = [max(parents)]
1483 1483 tested.update(parents)
1484 1484 pdeltas = []
1485 1485 for p in parents:
1486 1486 pd = builddelta(p)
1487 1487 if self._isgooddelta(pd, textlen):
1488 1488 pdeltas.append(pd)
1489 1489 if pdeltas:
1490 1490 delta = min(pdeltas, key=lambda x: x[1])
1491 1491 if delta is None and prev not in tested:
1492 1492 # other approach failed try against prev to hopefully save us a
1493 1493 # fulltext.
1494 1494 candidatedelta = builddelta(prev)
1495 1495 if self._isgooddelta(candidatedelta, textlen):
1496 1496 delta = candidatedelta
1497 1497 if delta is not None:
1498 1498 dist, l, data, base, chainbase, chainlen, compresseddeltalen = delta
1499 1499 else:
1500 1500 text = buildtext()
1501 1501 data = self.compress(text)
1502 1502 l = len(data[1]) + len(data[0])
1503 1503 base = chainbase = curr
1504 1504
1505 1505 e = (offset_type(offset, flags), l, textlen,
1506 1506 base, link, p1r, p2r, node)
1507 1507 self.index.insert(-1, e)
1508 1508 self.nodemap[node] = curr
1509 1509
1510 1510 entry = self._io.packentry(e, self.node, self.version, curr)
1511 1511 self._writeentry(transaction, ifh, dfh, entry, data, link, offset)
1512 1512
1513 1513 if alwayscache and text is None:
1514 1514 text = buildtext()
1515 1515
1516 1516 if type(text) == str: # only accept immutable objects
1517 1517 self._cache = (node, curr, text)
1518 1518 self._basecache = (curr, chainbase)
1519 1519 return node
1520 1520
1521 1521 def _writeentry(self, transaction, ifh, dfh, entry, data, link, offset):
1522 1522 # Files opened in a+ mode have inconsistent behavior on various
1523 1523 # platforms. Windows requires that a file positioning call be made
1524 1524 # when the file handle transitions between reads and writes. See
1525 1525 # 3686fa2b8eee and the mixedfilemodewrapper in windows.py. On other
1526 1526 # platforms, Python or the platform itself can be buggy. Some versions
1527 1527 # of Solaris have been observed to not append at the end of the file
1528 1528 # if the file was seeked to before the end. See issue4943 for more.
1529 1529 #
1530 1530 # We work around this issue by inserting a seek() before writing.
1531 1531 # Note: This is likely not necessary on Python 3.
1532 1532 ifh.seek(0, os.SEEK_END)
1533 1533 if dfh:
1534 1534 dfh.seek(0, os.SEEK_END)
1535 1535
1536 1536 curr = len(self) - 1
1537 1537 if not self._inline:
1538 1538 transaction.add(self.datafile, offset)
1539 1539 transaction.add(self.indexfile, curr * len(entry))
1540 1540 if data[0]:
1541 1541 dfh.write(data[0])
1542 1542 dfh.write(data[1])
1543 1543 ifh.write(entry)
1544 1544 else:
1545 1545 offset += curr * self._io.size
1546 1546 transaction.add(self.indexfile, offset, curr)
1547 1547 ifh.write(entry)
1548 1548 ifh.write(data[0])
1549 1549 ifh.write(data[1])
1550 1550 self.checkinlinesize(transaction, ifh)
1551 1551
1552 1552 def addgroup(self, cg, linkmapper, transaction, addrevisioncb=None):
1553 1553 """
1554 1554 add a delta group
1555 1555
1556 1556 given a set of deltas, add them to the revision log. the
1557 1557 first delta is against its parent, which should be in our
1558 1558 log, the rest are against the previous delta.
1559 1559
1560 1560 If ``addrevisioncb`` is defined, it will be called with arguments of
1561 1561 this revlog and the node that was added.
1562 1562 """
1563 1563
1564 1564 # track the base of the current delta log
1565 1565 content = []
1566 1566 node = None
1567 1567
1568 1568 r = len(self)
1569 1569 end = 0
1570 1570 if r:
1571 1571 end = self.end(r - 1)
1572 1572 ifh = self.opener(self.indexfile, "a+")
1573 1573 isize = r * self._io.size
1574 1574 if self._inline:
1575 1575 transaction.add(self.indexfile, end + isize, r)
1576 1576 dfh = None
1577 1577 else:
1578 1578 transaction.add(self.indexfile, isize, r)
1579 1579 transaction.add(self.datafile, end)
1580 1580 dfh = self.opener(self.datafile, "a+")
1581 1581 def flush():
1582 1582 if dfh:
1583 1583 dfh.flush()
1584 1584 ifh.flush()
1585 1585 try:
1586 1586 # loop through our set of deltas
1587 1587 chain = None
1588 1588 while True:
1589 1589 chunkdata = cg.deltachunk(chain)
1590 1590 if not chunkdata:
1591 1591 break
1592 1592 node = chunkdata['node']
1593 1593 p1 = chunkdata['p1']
1594 1594 p2 = chunkdata['p2']
1595 1595 cs = chunkdata['cs']
1596 1596 deltabase = chunkdata['deltabase']
1597 1597 delta = chunkdata['delta']
1598 1598 flags = chunkdata['flags'] or REVIDX_DEFAULT_FLAGS
1599 1599
1600 1600 content.append(node)
1601 1601
1602 1602 link = linkmapper(cs)
1603 1603 if node in self.nodemap:
1604 1604 # this can happen if two branches make the same change
1605 1605 chain = node
1606 1606 continue
1607 1607
1608 1608 for p in (p1, p2):
1609 1609 if p not in self.nodemap:
1610 1610 raise LookupError(p, self.indexfile,
1611 1611 _('unknown parent'))
1612 1612
1613 1613 if deltabase not in self.nodemap:
1614 1614 raise LookupError(deltabase, self.indexfile,
1615 1615 _('unknown delta base'))
1616 1616
1617 1617 baserev = self.rev(deltabase)
1618 1618
1619 1619 if baserev != nullrev and self.iscensored(baserev):
1620 1620 # if base is censored, delta must be full replacement in a
1621 1621 # single patch operation
1622 1622 hlen = struct.calcsize(">lll")
1623 1623 oldlen = self.rawsize(baserev)
1624 1624 newlen = len(delta) - hlen
1625 1625 if delta[:hlen] != mdiff.replacediffheader(oldlen, newlen):
1626 1626 raise error.CensoredBaseError(self.indexfile,
1627 1627 self.node(baserev))
1628 1628
1629 1629 if not flags and self._peek_iscensored(baserev, delta, flush):
1630 1630 flags |= REVIDX_ISCENSORED
1631 1631
1632 1632 # We assume consumers of addrevisioncb will want to retrieve
1633 1633 # the added revision, which will require a call to
1634 1634 # revision(). revision() will fast path if there is a cache
1635 1635 # hit. So, we tell _addrevision() to always cache in this case.
1636 1636 chain = self._addrevision(node, None, transaction, link,
1637 1637 p1, p2, flags, (baserev, delta),
1638 1638 ifh, dfh,
1639 1639 alwayscache=bool(addrevisioncb))
1640 1640
1641 1641 if addrevisioncb:
1642 1642 addrevisioncb(self, chain)
1643 1643
1644 1644 if not dfh and not self._inline:
1645 1645 # addrevision switched from inline to conventional
1646 1646 # reopen the index
1647 1647 ifh.close()
1648 1648 dfh = self.opener(self.datafile, "a+")
1649 1649 ifh = self.opener(self.indexfile, "a+")
1650 1650 finally:
1651 1651 if dfh:
1652 1652 dfh.close()
1653 1653 ifh.close()
1654 1654
1655 1655 return content
1656 1656
1657 1657 def iscensored(self, rev):
1658 1658 """Check if a file revision is censored."""
1659 1659 return False
1660 1660
1661 1661 def _peek_iscensored(self, baserev, delta, flush):
1662 1662 """Quickly check if a delta produces a censored revision."""
1663 1663 return False
1664 1664
1665 1665 def getstrippoint(self, minlink):
1666 1666 """find the minimum rev that must be stripped to strip the linkrev
1667 1667
1668 1668 Returns a tuple containing the minimum rev and a set of all revs that
1669 1669 have linkrevs that will be broken by this strip.
1670 1670 """
1671 1671 brokenrevs = set()
1672 1672 strippoint = len(self)
1673 1673
1674 1674 heads = {}
1675 1675 futurelargelinkrevs = set()
1676 1676 for head in self.headrevs():
1677 1677 headlinkrev = self.linkrev(head)
1678 1678 heads[head] = headlinkrev
1679 1679 if headlinkrev >= minlink:
1680 1680 futurelargelinkrevs.add(headlinkrev)
1681 1681
1682 1682 # This algorithm involves walking down the rev graph, starting at the
1683 1683 # heads. Since the revs are topologically sorted according to linkrev,
1684 1684 # once all head linkrevs are below the minlink, we know there are
1685 1685 # no more revs that could have a linkrev greater than minlink.
1686 1686 # So we can stop walking.
1687 1687 while futurelargelinkrevs:
1688 1688 strippoint -= 1
1689 1689 linkrev = heads.pop(strippoint)
1690 1690
1691 1691 if linkrev < minlink:
1692 1692 brokenrevs.add(strippoint)
1693 1693 else:
1694 1694 futurelargelinkrevs.remove(linkrev)
1695 1695
1696 1696 for p in self.parentrevs(strippoint):
1697 1697 if p != nullrev:
1698 1698 plinkrev = self.linkrev(p)
1699 1699 heads[p] = plinkrev
1700 1700 if plinkrev >= minlink:
1701 1701 futurelargelinkrevs.add(plinkrev)
1702 1702
1703 1703 return strippoint, brokenrevs
1704 1704
1705 1705 def strip(self, minlink, transaction):
1706 1706 """truncate the revlog on the first revision with a linkrev >= minlink
1707 1707
1708 1708 This function is called when we're stripping revision minlink and
1709 1709 its descendants from the repository.
1710 1710
1711 1711 We have to remove all revisions with linkrev >= minlink, because
1712 1712 the equivalent changelog revisions will be renumbered after the
1713 1713 strip.
1714 1714
1715 1715 So we truncate the revlog on the first of these revisions, and
1716 1716 trust that the caller has saved the revisions that shouldn't be
1717 1717 removed and that it'll re-add them after this truncation.
1718 1718 """
1719 1719 if len(self) == 0:
1720 1720 return
1721 1721
1722 1722 rev, _ = self.getstrippoint(minlink)
1723 1723 if rev == len(self):
1724 1724 return
1725 1725
1726 1726 # first truncate the files on disk
1727 1727 end = self.start(rev)
1728 1728 if not self._inline:
1729 1729 transaction.add(self.datafile, end)
1730 1730 end = rev * self._io.size
1731 1731 else:
1732 1732 end += rev * self._io.size
1733 1733
1734 1734 transaction.add(self.indexfile, end)
1735 1735
1736 1736 # then reset internal state in memory to forget those revisions
1737 1737 self._cache = None
1738 1738 self._chaininfocache = {}
1739 1739 self._chunkclear()
1740 1740 for x in xrange(rev, len(self)):
1741 1741 del self.nodemap[self.node(x)]
1742 1742
1743 1743 del self.index[rev:-1]
1744 1744
1745 1745 def checksize(self):
1746 1746 expected = 0
1747 1747 if len(self):
1748 1748 expected = max(0, self.end(len(self) - 1))
1749 1749
1750 1750 try:
1751 1751 f = self.opener(self.datafile)
1752 1752 f.seek(0, 2)
1753 1753 actual = f.tell()
1754 1754 f.close()
1755 1755 dd = actual - expected
1756 1756 except IOError as inst:
1757 1757 if inst.errno != errno.ENOENT:
1758 1758 raise
1759 1759 dd = 0
1760 1760
1761 1761 try:
1762 1762 f = self.opener(self.indexfile)
1763 1763 f.seek(0, 2)
1764 1764 actual = f.tell()
1765 1765 f.close()
1766 1766 s = self._io.size
1767 1767 i = max(0, actual // s)
1768 1768 di = actual - (i * s)
1769 1769 if self._inline:
1770 1770 databytes = 0
1771 1771 for r in self:
1772 1772 databytes += max(0, self.length(r))
1773 1773 dd = 0
1774 1774 di = actual - len(self) * s - databytes
1775 1775 except IOError as inst:
1776 1776 if inst.errno != errno.ENOENT:
1777 1777 raise
1778 1778 di = 0
1779 1779
1780 1780 return (dd, di)
1781 1781
1782 1782 def files(self):
1783 1783 res = [self.indexfile]
1784 1784 if not self._inline:
1785 1785 res.append(self.datafile)
1786 1786 return res
General Comments 0
You need to be logged in to leave comments. Login now