52
52
is that sequence numbers are stable references. But not every
53
53
repository in the world will assign the same sequence numbers,
54
54
therefore the SHA-1 is the only universally unique reference.
56
This is meant to scale to hold 100,000 revisions of a single file, by
57
which time the index file will be ~4.8MB and a bit big to read
60
Some of the reserved fields could be used to implement a (semi?)
61
balanced tree indexed by SHA1 so we can much more efficiently find the
62
index associated with a particular hash. For 100,000 revs we would be
63
able to find it in about 17 random reads, which is not too bad.
65
This performs pretty well except when trying to calculate deltas of
66
really large files. For that the main thing would be to plug in
67
something faster than difflib, which is after all pure Python.
68
Another approach is to just store the gzipped full text of big files,
69
though perhaps that's too perverse?
55
71
The iter method here will generally read through the whole index file
56
72
in one go. With readahead in the kernel and python/libc (typically
57
73
128kB) this means that there should be no seeks and often only one
73
89
# if there are gaps and that can happen if we're interrupted while
74
90
# writing to the datafile. Overlapping would be very bad though.
92
# TODO: Shouldn't need to lock if we always write in append mode and
93
# then ftell after writing to see where it went. In any case we
94
# assume the whole branch is protected by a lock.
78
96
import sys, zlib, struct, mdiff, stat, os, sha
79
97
from binascii import hexlify, unhexlify
85
101
_HEADER = "bzr revfile v1\n"
227
242
return self._add_compressed(text_sha, text, _NO_RECORD, compress)
246
def _choose_base(self, seed, base):
248
if base == _NO_RECORD:
251
if idxrec[I_BASE] == _NO_RECORD:
254
base = idxrec[I_BASE]
257
return base # relative to this full text
230
261
def _add_delta(self, text, text_sha, base, compress):
231
262
"""Add a text stored relative to a previous text."""
232
263
self._check_index(base)
235
base_text = self.get(base, recursion_limit=CHAIN_LIMIT)
266
base_text = self.get(base, CHAIN_LIMIT)
236
267
except LimitHitException:
237
268
return self._add_full_text(text, text_sha, compress)
239
270
data = mdiff.bdiff(base_text, text)
273
if True: # paranoid early check for bad diff
274
result = mdiff.bpatch(base_text, data)
275
assert result == text
241
278
# If the delta is larger than the text, we might as well just
242
279
# store the text. (OK, the delta might be more compressible,
248
285
return self._add_compressed(text_sha, data, base, compress)
251
def add(self, text, base=_NO_RECORD, compress=True):
288
def add(self, text, base=None, compress=True):
252
289
"""Add a new text to the revfile.
254
291
If the text is already present them its existing id is
272
312
# it's the same, in case someone ever breaks SHA-1.
273
313
return idx # already present
315
# base = self._choose_base(ord(text_sha[0]), base)
275
317
if base == _NO_RECORD:
276
318
return self._add_full_text(text, text_sha, compress)
296
338
text = self._get_patched(idx, idxrec, recursion_limit)
298
340
if sha.new(text).digest() != idxrec[I_SHA]:
299
raise RevfileError("corrupt SHA-1 digest on record %d"
341
raise RevfileError("corrupt SHA-1 digest on record %d in %s"
342
% (idx, self.basename))
388
430
"""Read back all index records.
390
432
Do not seek the index file while this is underway!"""
391
sys.stderr.write(" ** iter called ** \n")
433
## sys.stderr.write(" ** iter called ** \n")
392
434
self._seek_index(0)
394
436
idxrec = self._read_next_index()
436
478
for idx in range(len(self)):
437
479
t += len(self.get(idx))
483
def check(self, pb=None):
484
"""Extract every version and check its hash."""
486
for i in range(total):
488
pb.update("check revision", i, total)
489
# the get method implicitly checks the SHA-1
445
500
except IndexError:
446
sys.stderr.write("usage: revfile dump\n"
448
" revfile add-delta BASE\n"
450
" revfile find-sha HEX\n"
451
" revfile total-text-size\n"
501
sys.stderr.write("usage: revfile dump REVFILE\n"
502
" revfile add REVFILE < INPUT\n"
503
" revfile add-delta REVFILE BASE < INPUT\n"
504
" revfile add-series REVFILE BASE FILE...\n"
505
" revfile get REVFILE IDX\n"
506
" revfile find-sha REVFILE HEX\n"
507
" revfile total-text-size REVFILE\n"
508
" revfile last REVFILE\n")
511
if filename.endswith('.drev') or filename.endswith('.irev'):
512
filename = filename[:-5]
456
return Revfile('testrev', 'w')
515
return Revfile(filename, 'w')
459
return Revfile('testrev', 'r')
518
return Revfile(filename, 'r')
462
521
print rw().add(sys.stdin.read())
463
522
elif cmd == 'add-delta':
464
print rw().add(sys.stdin.read(), int(argv[2]))
523
print rw().add(sys.stdin.read(), int(argv[3]))
524
elif cmd == 'add-series':
529
rev = r.add(file(fn).read(), rev)
465
530
elif cmd == 'dump':
467
532
elif cmd == 'get':
470
535
except IndexError:
471
sys.stderr.write("usage: revfile get IDX\n")
536
sys.stderr.write("usage: revfile get FILE IDX\n")
474
541
if idx < 0 or idx >= len(r):
475
542
sys.stderr.write("invalid index %r\n" % idx)
478
sys.stdout.write(ro().get(idx))
545
sys.stdout.write(r.get(idx))
479
546
elif cmd == 'find-sha':
481
s = unhexlify(argv[2])
548
s = unhexlify(argv[3])
482
549
except IndexError:
483
sys.stderr.write("usage: revfile find-sha HEX\n")
550
sys.stderr.write("usage: revfile find-sha FILE HEX\n")
486
553
idx = ro().find_sha(s)
added
removed
bzrlib/revfile.py
