1
# Copyright (C) 2008, 2009, 2010 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
"""Tests for group compression."""
31
from bzrlib.osutils import sha_string
32
from bzrlib.tests.test__groupcompress import compiled_groupcompress_feature
35
def load_tests(standard_tests, module, loader):
36
"""Parameterize tests for all versions of groupcompress."""
37
to_adapt, result = tests.split_suite_by_condition(
38
standard_tests, tests.condition_isinstance(TestAllGroupCompressors))
40
('python', {'compressor': groupcompress.PythonGroupCompressor}),
42
if compiled_groupcompress_feature.available():
43
scenarios.append(('C',
44
{'compressor': groupcompress.PyrexGroupCompressor}))
45
return tests.multiply_tests(to_adapt, scenarios, result)
48
class TestGroupCompressor(tests.TestCase):
50
def _chunks_to_repr_lines(self, chunks):
51
return '\n'.join(map(repr, ''.join(chunks).split('\n')))
53
def assertEqualDiffEncoded(self, expected, actual):
54
"""Compare the actual content to the expected content.
56
:param expected: A group of chunks that we expect to see
57
:param actual: The measured 'chunks'
59
We will transform the chunks back into lines, and then run 'repr()'
60
over them to handle non-ascii characters.
62
self.assertEqualDiff(self._chunks_to_repr_lines(expected),
63
self._chunks_to_repr_lines(actual))
66
class TestAllGroupCompressors(TestGroupCompressor):
67
"""Tests for GroupCompressor"""
69
compressor = None # Set by multiply_tests
71
def test_empty_delta(self):
72
compressor = self.compressor()
73
self.assertEqual([], compressor.chunks)
75
def test_one_nosha_delta(self):
77
compressor = self.compressor()
78
sha1, start_point, end_point, _ = compressor.compress(('label',),
79
'strange\ncommon\n', None)
80
self.assertEqual(sha_string('strange\ncommon\n'), sha1)
81
expected_lines = 'f' '\x0f' 'strange\ncommon\n'
82
self.assertEqual(expected_lines, ''.join(compressor.chunks))
83
self.assertEqual(0, start_point)
84
self.assertEqual(sum(map(len, expected_lines)), end_point)
86
def test_empty_content(self):
87
compressor = self.compressor()
88
# Adding empty bytes should return the 'null' record
89
sha1, start_point, end_point, kind = compressor.compress(('empty',),
91
self.assertEqual(0, start_point)
92
self.assertEqual(0, end_point)
93
self.assertEqual('fulltext', kind)
94
self.assertEqual(groupcompress._null_sha1, sha1)
95
self.assertEqual(0, compressor.endpoint)
96
self.assertEqual([], compressor.chunks)
97
# Even after adding some content
98
compressor.compress(('content',), 'some\nbytes\n', None)
99
self.assertTrue(compressor.endpoint > 0)
100
sha1, start_point, end_point, kind = compressor.compress(('empty2',),
102
self.assertEqual(0, start_point)
103
self.assertEqual(0, end_point)
104
self.assertEqual('fulltext', kind)
105
self.assertEqual(groupcompress._null_sha1, sha1)
107
def test_extract_from_compressor(self):
108
# Knit fetching will try to reconstruct texts locally which results in
109
# reading something that is in the compressor stream already.
110
compressor = self.compressor()
111
sha1_1, _, _, _ = compressor.compress(('label',),
112
'strange\ncommon long line\nthat needs a 16 byte match\n', None)
113
expected_lines = list(compressor.chunks)
114
sha1_2, _, end_point, _ = compressor.compress(('newlabel',),
115
'common long line\nthat needs a 16 byte match\ndifferent\n', None)
117
self.assertEqual(('strange\ncommon long line\n'
118
'that needs a 16 byte match\n', sha1_1),
119
compressor.extract(('label',)))
121
self.assertEqual(('common long line\nthat needs a 16 byte match\n'
122
'different\n', sha1_2),
123
compressor.extract(('newlabel',)))
125
def test_pop_last(self):
126
compressor = self.compressor()
127
_, _, _, _ = compressor.compress(('key1',),
128
'some text\nfor the first entry\n', None)
129
expected_lines = list(compressor.chunks)
130
_, _, _, _ = compressor.compress(('key2',),
131
'some text\nfor the second entry\n', None)
132
compressor.pop_last()
133
self.assertEqual(expected_lines, compressor.chunks)
136
class TestPyrexGroupCompressor(TestGroupCompressor):
138
_test_needs_features = [compiled_groupcompress_feature]
139
compressor = groupcompress.PyrexGroupCompressor
141
def test_stats(self):
142
compressor = self.compressor()
143
compressor.compress(('label',),
145
'common very very long line\n'
146
'plus more text\n', None)
147
compressor.compress(('newlabel',),
148
'common very very long line\n'
151
'moredifferent\n', None)
152
compressor.compress(('label3',),
154
'common very very long line\n'
157
'moredifferent\n', None)
158
self.assertAlmostEqual(1.9, compressor.ratio(), 1)
160
def test_two_nosha_delta(self):
161
compressor = self.compressor()
162
sha1_1, _, _, _ = compressor.compress(('label',),
163
'strange\ncommon long line\nthat needs a 16 byte match\n', None)
164
expected_lines = list(compressor.chunks)
165
sha1_2, start_point, end_point, _ = compressor.compress(('newlabel',),
166
'common long line\nthat needs a 16 byte match\ndifferent\n', None)
167
self.assertEqual(sha_string('common long line\n'
168
'that needs a 16 byte match\n'
169
'different\n'), sha1_2)
170
expected_lines.extend([
171
# 'delta', delta length
173
# source and target length
175
# copy the line common
176
'\x91\x0a\x2c', #copy, offset 0x0a, len 0x2c
177
# add the line different, and the trailing newline
178
'\x0adifferent\n', # insert 10 bytes
180
self.assertEqualDiffEncoded(expected_lines, compressor.chunks)
181
self.assertEqual(sum(map(len, expected_lines)), end_point)
183
def test_three_nosha_delta(self):
184
# The first interesting test: make a change that should use lines from
186
compressor = self.compressor()
187
sha1_1, _, _, _ = compressor.compress(('label',),
188
'strange\ncommon very very long line\nwith some extra text\n', None)
189
sha1_2, _, _, _ = compressor.compress(('newlabel',),
190
'different\nmoredifferent\nand then some more\n', None)
191
expected_lines = list(compressor.chunks)
192
sha1_3, start_point, end_point, _ = compressor.compress(('label3',),
193
'new\ncommon very very long line\nwith some extra text\n'
194
'different\nmoredifferent\nand then some more\n',
197
sha_string('new\ncommon very very long line\nwith some extra text\n'
198
'different\nmoredifferent\nand then some more\n'),
200
expected_lines.extend([
201
# 'delta', delta length
203
# source and target length
207
# Copy of first parent 'common' range
208
'\x91\x09\x31' # copy, offset 0x09, 0x31 bytes
209
# Copy of second parent 'different' range
210
'\x91\x3c\x2b' # copy, offset 0x3c, 0x2b bytes
212
self.assertEqualDiffEncoded(expected_lines, compressor.chunks)
213
self.assertEqual(sum(map(len, expected_lines)), end_point)
216
class TestPythonGroupCompressor(TestGroupCompressor):
218
compressor = groupcompress.PythonGroupCompressor
220
def test_stats(self):
221
compressor = self.compressor()
222
compressor.compress(('label',),
224
'common very very long line\n'
225
'plus more text\n', None)
226
compressor.compress(('newlabel',),
227
'common very very long line\n'
230
'moredifferent\n', None)
231
compressor.compress(('label3',),
233
'common very very long line\n'
236
'moredifferent\n', None)
237
self.assertAlmostEqual(1.9, compressor.ratio(), 1)
239
def test_two_nosha_delta(self):
240
compressor = self.compressor()
241
sha1_1, _, _, _ = compressor.compress(('label',),
242
'strange\ncommon long line\nthat needs a 16 byte match\n', None)
243
expected_lines = list(compressor.chunks)
244
sha1_2, start_point, end_point, _ = compressor.compress(('newlabel',),
245
'common long line\nthat needs a 16 byte match\ndifferent\n', None)
246
self.assertEqual(sha_string('common long line\n'
247
'that needs a 16 byte match\n'
248
'different\n'), sha1_2)
249
expected_lines.extend([
250
# 'delta', delta length
254
# copy the line common
255
'\x91\x0a\x2c', #copy, offset 0x0a, len 0x2c
256
# add the line different, and the trailing newline
257
'\x0adifferent\n', # insert 10 bytes
259
self.assertEqualDiffEncoded(expected_lines, compressor.chunks)
260
self.assertEqual(sum(map(len, expected_lines)), end_point)
262
def test_three_nosha_delta(self):
263
# The first interesting test: make a change that should use lines from
265
compressor = self.compressor()
266
sha1_1, _, _, _ = compressor.compress(('label',),
267
'strange\ncommon very very long line\nwith some extra text\n', None)
268
sha1_2, _, _, _ = compressor.compress(('newlabel',),
269
'different\nmoredifferent\nand then some more\n', None)
270
expected_lines = list(compressor.chunks)
271
sha1_3, start_point, end_point, _ = compressor.compress(('label3',),
272
'new\ncommon very very long line\nwith some extra text\n'
273
'different\nmoredifferent\nand then some more\n',
276
sha_string('new\ncommon very very long line\nwith some extra text\n'
277
'different\nmoredifferent\nand then some more\n'),
279
expected_lines.extend([
280
# 'delta', delta length
286
# Copy of first parent 'common' range
287
'\x91\x0a\x30' # copy, offset 0x0a, 0x30 bytes
288
# Copy of second parent 'different' range
289
'\x91\x3c\x2b' # copy, offset 0x3c, 0x2b bytes
291
self.assertEqualDiffEncoded(expected_lines, compressor.chunks)
292
self.assertEqual(sum(map(len, expected_lines)), end_point)
295
class TestGroupCompressBlock(tests.TestCase):
297
def make_block(self, key_to_text):
298
"""Create a GroupCompressBlock, filling it with the given texts."""
299
compressor = groupcompress.GroupCompressor()
301
for key in sorted(key_to_text):
302
compressor.compress(key, key_to_text[key], None)
303
locs = dict((key, (start, end)) for key, (start, _, end, _)
304
in compressor.labels_deltas.iteritems())
305
block = compressor.flush()
306
raw_bytes = block.to_bytes()
307
# Go through from_bytes(to_bytes()) so that we start with a compressed
309
return locs, groupcompress.GroupCompressBlock.from_bytes(raw_bytes)
311
def test_from_empty_bytes(self):
312
self.assertRaises(ValueError,
313
groupcompress.GroupCompressBlock.from_bytes, '')
315
def test_from_minimal_bytes(self):
316
block = groupcompress.GroupCompressBlock.from_bytes(
318
self.assertIsInstance(block, groupcompress.GroupCompressBlock)
319
self.assertIs(None, block._content)
320
self.assertEqual('', block._z_content)
321
block._ensure_content()
322
self.assertEqual('', block._content)
323
self.assertEqual('', block._z_content)
324
block._ensure_content() # Ensure content is safe to call 2x
326
def test_from_invalid(self):
327
self.assertRaises(ValueError,
328
groupcompress.GroupCompressBlock.from_bytes,
329
'this is not a valid header')
331
def test_from_bytes(self):
332
content = ('a tiny bit of content\n')
333
z_content = zlib.compress(content)
335
'gcb1z\n' # group compress block v1 plain
336
'%d\n' # Length of compressed content
337
'%d\n' # Length of uncompressed content
338
'%s' # Compressed content
339
) % (len(z_content), len(content), z_content)
340
block = groupcompress.GroupCompressBlock.from_bytes(
342
self.assertEqual(z_content, block._z_content)
343
self.assertIs(None, block._content)
344
self.assertEqual(len(z_content), block._z_content_length)
345
self.assertEqual(len(content), block._content_length)
346
block._ensure_content()
347
self.assertEqual(z_content, block._z_content)
348
self.assertEqual(content, block._content)
350
def test_to_bytes(self):
351
content = ('this is some content\n'
352
'this content will be compressed\n')
353
gcb = groupcompress.GroupCompressBlock()
354
gcb.set_content(content)
355
bytes = gcb.to_bytes()
356
self.assertEqual(gcb._z_content_length, len(gcb._z_content))
357
self.assertEqual(gcb._content_length, len(content))
358
expected_header =('gcb1z\n' # group compress block v1 zlib
359
'%d\n' # Length of compressed content
360
'%d\n' # Length of uncompressed content
361
) % (gcb._z_content_length, gcb._content_length)
362
self.assertStartsWith(bytes, expected_header)
363
remaining_bytes = bytes[len(expected_header):]
364
raw_bytes = zlib.decompress(remaining_bytes)
365
self.assertEqual(content, raw_bytes)
367
# we should get the same results if using the chunked version
368
gcb = groupcompress.GroupCompressBlock()
369
gcb.set_chunked_content(['this is some content\n'
370
'this content will be compressed\n'],
373
bytes = gcb.to_bytes()
374
self.assertEqual(old_bytes, bytes)
376
def test_partial_decomp(self):
378
# We need a sufficient amount of data so that zlib.decompress has
379
# partial decompression to work with. Most auto-generated data
380
# compresses a bit too well, we want a combination, so we combine a sha
381
# hash with compressible data.
382
for i in xrange(2048):
383
next_content = '%d\nThis is a bit of duplicate text\n' % (i,)
384
content_chunks.append(next_content)
385
next_sha1 = osutils.sha_string(next_content)
386
content_chunks.append(next_sha1 + '\n')
387
content = ''.join(content_chunks)
388
self.assertEqual(158634, len(content))
389
z_content = zlib.compress(content)
390
self.assertEqual(57182, len(z_content))
391
block = groupcompress.GroupCompressBlock()
392
block._z_content = z_content
393
block._z_content_length = len(z_content)
394
block._compressor_name = 'zlib'
395
block._content_length = 158634
396
self.assertIs(None, block._content)
397
block._ensure_content(100)
398
self.assertIsNot(None, block._content)
399
# We have decompressed at least 100 bytes
400
self.assertTrue(len(block._content) >= 100)
401
# We have not decompressed the whole content
402
self.assertTrue(len(block._content) < 158634)
403
self.assertEqualDiff(content[:len(block._content)], block._content)
404
# ensuring content that we already have shouldn't cause any more data
406
cur_len = len(block._content)
407
block._ensure_content(cur_len - 10)
408
self.assertEqual(cur_len, len(block._content))
409
# Now we want a bit more content
411
block._ensure_content(cur_len)
412
self.assertTrue(len(block._content) >= cur_len)
413
self.assertTrue(len(block._content) < 158634)
414
self.assertEqualDiff(content[:len(block._content)], block._content)
415
# And now lets finish
416
block._ensure_content(158634)
417
self.assertEqualDiff(content, block._content)
418
# And the decompressor is finalized
419
self.assertIs(None, block._z_content_decompressor)
421
def test__ensure_all_content(self):
423
# We need a sufficient amount of data so that zlib.decompress has
424
# partial decompression to work with. Most auto-generated data
425
# compresses a bit too well, we want a combination, so we combine a sha
426
# hash with compressible data.
427
for i in xrange(2048):
428
next_content = '%d\nThis is a bit of duplicate text\n' % (i,)
429
content_chunks.append(next_content)
430
next_sha1 = osutils.sha_string(next_content)
431
content_chunks.append(next_sha1 + '\n')
432
content = ''.join(content_chunks)
433
self.assertEqual(158634, len(content))
434
z_content = zlib.compress(content)
435
self.assertEqual(57182, len(z_content))
436
block = groupcompress.GroupCompressBlock()
437
block._z_content = z_content
438
block._z_content_length = len(z_content)
439
block._compressor_name = 'zlib'
440
block._content_length = 158634
441
self.assertIs(None, block._content)
442
# The first _ensure_content got all of the required data
443
block._ensure_content(158634)
444
self.assertEqualDiff(content, block._content)
445
# And we should have released the _z_content_decompressor since it was
447
self.assertIs(None, block._z_content_decompressor)
449
def test__dump(self):
450
dup_content = 'some duplicate content\nwhich is sufficiently long\n'
451
key_to_text = {('1',): dup_content + '1 unique\n',
452
('2',): dup_content + '2 extra special\n'}
453
locs, block = self.make_block(key_to_text)
454
self.assertEqual([('f', len(key_to_text[('1',)])),
455
('d', 21, len(key_to_text[('2',)]),
456
[('c', 2, len(dup_content)),
457
('i', len('2 extra special\n'), '')
462
class TestCaseWithGroupCompressVersionedFiles(
463
tests.TestCaseWithMemoryTransport):
465
def make_test_vf(self, create_graph, keylength=1, do_cleanup=True,
466
dir='.', inconsistency_fatal=True):
467
t = self.get_transport(dir)
469
vf = groupcompress.make_pack_factory(graph=create_graph,
470
delta=False, keylength=keylength,
471
inconsistency_fatal=inconsistency_fatal)(t)
473
self.addCleanup(groupcompress.cleanup_pack_group, vf)
477
class TestGroupCompressVersionedFiles(TestCaseWithGroupCompressVersionedFiles):
479
def make_g_index(self, name, ref_lists=0, nodes=[]):
480
builder = btree_index.BTreeBuilder(ref_lists)
481
for node, references, value in nodes:
482
builder.add_node(node, references, value)
483
stream = builder.finish()
484
trans = self.get_transport()
485
size = trans.put_file(name, stream)
486
return btree_index.BTreeGraphIndex(trans, name, size)
488
def make_g_index_missing_parent(self):
489
graph_index = self.make_g_index('missing_parent', 1,
490
[(('parent', ), '2 78 2 10', ([],)),
491
(('tip', ), '2 78 2 10',
492
([('parent', ), ('missing-parent', )],)),
496
def test_get_record_stream_as_requested(self):
497
# Consider promoting 'as-requested' to general availability, and
498
# make this a VF interface test
499
vf = self.make_test_vf(False, dir='source')
500
vf.add_lines(('a',), (), ['lines\n'])
501
vf.add_lines(('b',), (), ['lines\n'])
502
vf.add_lines(('c',), (), ['lines\n'])
503
vf.add_lines(('d',), (), ['lines\n'])
505
keys = [record.key for record in vf.get_record_stream(
506
[('a',), ('b',), ('c',), ('d',)],
507
'as-requested', False)]
508
self.assertEqual([('a',), ('b',), ('c',), ('d',)], keys)
509
keys = [record.key for record in vf.get_record_stream(
510
[('b',), ('a',), ('d',), ('c',)],
511
'as-requested', False)]
512
self.assertEqual([('b',), ('a',), ('d',), ('c',)], keys)
514
# It should work even after being repacked into another VF
515
vf2 = self.make_test_vf(False, dir='target')
516
vf2.insert_record_stream(vf.get_record_stream(
517
[('b',), ('a',), ('d',), ('c',)], 'as-requested', False))
520
keys = [record.key for record in vf2.get_record_stream(
521
[('a',), ('b',), ('c',), ('d',)],
522
'as-requested', False)]
523
self.assertEqual([('a',), ('b',), ('c',), ('d',)], keys)
524
keys = [record.key for record in vf2.get_record_stream(
525
[('b',), ('a',), ('d',), ('c',)],
526
'as-requested', False)]
527
self.assertEqual([('b',), ('a',), ('d',), ('c',)], keys)
529
def test_insert_record_stream_reuses_blocks(self):
530
vf = self.make_test_vf(True, dir='source')
531
def grouped_stream(revision_ids, first_parents=()):
532
parents = first_parents
533
for revision_id in revision_ids:
535
record = versionedfile.FulltextContentFactory(
537
'some content that is\n'
538
'identical except for\n'
539
'revision_id:%s\n' % (revision_id,))
543
vf.insert_record_stream(grouped_stream(['a', 'b', 'c', 'd']))
545
vf.insert_record_stream(grouped_stream(['e', 'f', 'g', 'h'],
546
first_parents=(('d',),)))
548
stream = vf.get_record_stream([(r,) for r in 'abcdefgh'],
551
for record in stream:
552
if record.key in [('a',), ('e',)]:
553
self.assertEqual('groupcompress-block', record.storage_kind)
555
self.assertEqual('groupcompress-block-ref',
557
block_bytes[record.key] = record._manager._block._z_content
559
self.assertEqual(8, num_records)
562
self.assertIs(block_bytes[key], block_bytes[('a',)])
563
self.assertNotEqual(block_bytes[key], block_bytes[('e',)])
566
self.assertIs(block_bytes[key], block_bytes[('e',)])
567
self.assertNotEqual(block_bytes[key], block_bytes[('a',)])
568
# Now copy the blocks into another vf, and ensure that the blocks are
569
# preserved without creating new entries
570
vf2 = self.make_test_vf(True, dir='target')
571
# ordering in 'groupcompress' order, should actually swap the groups in
572
# the target vf, but the groups themselves should not be disturbed.
573
def small_size_stream():
574
for record in vf.get_record_stream([(r,) for r in 'abcdefgh'],
575
'groupcompress', False):
576
record._manager._full_enough_block_size = \
577
record._manager._block._content_length
580
vf2.insert_record_stream(small_size_stream())
581
stream = vf2.get_record_stream([(r,) for r in 'abcdefgh'],
582
'groupcompress', False)
585
for record in stream:
587
self.assertEqual(block_bytes[record.key],
588
record._manager._block._z_content)
589
self.assertEqual(8, num_records)
591
def test_insert_record_stream_packs_on_the_fly(self):
592
vf = self.make_test_vf(True, dir='source')
593
def grouped_stream(revision_ids, first_parents=()):
594
parents = first_parents
595
for revision_id in revision_ids:
597
record = versionedfile.FulltextContentFactory(
599
'some content that is\n'
600
'identical except for\n'
601
'revision_id:%s\n' % (revision_id,))
605
vf.insert_record_stream(grouped_stream(['a', 'b', 'c', 'd']))
607
vf.insert_record_stream(grouped_stream(['e', 'f', 'g', 'h'],
608
first_parents=(('d',),)))
609
# Now copy the blocks into another vf, and see that the
610
# insert_record_stream rebuilt a new block on-the-fly because of
612
vf2 = self.make_test_vf(True, dir='target')
613
vf2.insert_record_stream(vf.get_record_stream(
614
[(r,) for r in 'abcdefgh'], 'groupcompress', False))
615
stream = vf2.get_record_stream([(r,) for r in 'abcdefgh'],
616
'groupcompress', False)
619
# All of the records should be recombined into a single block
621
for record in stream:
624
block = record._manager._block
626
self.assertIs(block, record._manager._block)
627
self.assertEqual(8, num_records)
629
def test__insert_record_stream_no_reuse_block(self):
630
vf = self.make_test_vf(True, dir='source')
631
def grouped_stream(revision_ids, first_parents=()):
632
parents = first_parents
633
for revision_id in revision_ids:
635
record = versionedfile.FulltextContentFactory(
637
'some content that is\n'
638
'identical except for\n'
639
'revision_id:%s\n' % (revision_id,))
643
vf.insert_record_stream(grouped_stream(['a', 'b', 'c', 'd']))
645
vf.insert_record_stream(grouped_stream(['e', 'f', 'g', 'h'],
646
first_parents=(('d',),)))
648
self.assertEqual(8, len(list(vf.get_record_stream(
649
[(r,) for r in 'abcdefgh'],
650
'unordered', False))))
651
# Now copy the blocks into another vf, and ensure that the blocks are
652
# preserved without creating new entries
653
vf2 = self.make_test_vf(True, dir='target')
654
# ordering in 'groupcompress' order, should actually swap the groups in
655
# the target vf, but the groups themselves should not be disturbed.
656
list(vf2._insert_record_stream(vf.get_record_stream(
657
[(r,) for r in 'abcdefgh'], 'groupcompress', False),
660
# After inserting with reuse_blocks=False, we should have everything in
661
# a single new block.
662
stream = vf2.get_record_stream([(r,) for r in 'abcdefgh'],
663
'groupcompress', False)
665
for record in stream:
667
block = record._manager._block
669
self.assertIs(block, record._manager._block)
671
def test_add_missing_noncompression_parent_unvalidated_index(self):
672
unvalidated = self.make_g_index_missing_parent()
673
combined = _mod_index.CombinedGraphIndex([unvalidated])
674
index = groupcompress._GCGraphIndex(combined,
675
is_locked=lambda: True, parents=True,
676
track_external_parent_refs=True)
677
index.scan_unvalidated_index(unvalidated)
679
frozenset([('missing-parent',)]), index.get_missing_parents())
681
def test_track_external_parent_refs(self):
682
g_index = self.make_g_index('empty', 1, [])
683
mod_index = btree_index.BTreeBuilder(1, 1)
684
combined = _mod_index.CombinedGraphIndex([g_index, mod_index])
685
index = groupcompress._GCGraphIndex(combined,
686
is_locked=lambda: True, parents=True,
687
add_callback=mod_index.add_nodes,
688
track_external_parent_refs=True)
690
(('new-key',), '2 10 2 10', [(('parent-1',), ('parent-2',))])])
692
frozenset([('parent-1',), ('parent-2',)]),
693
index.get_missing_parents())
695
def make_source_with_b(self, a_parent, path):
696
source = self.make_test_vf(True, dir=path)
697
source.add_lines(('a',), (), ['lines\n'])
699
b_parents = (('a',),)
702
source.add_lines(('b',), b_parents, ['lines\n'])
705
def do_inconsistent_inserts(self, inconsistency_fatal):
706
target = self.make_test_vf(True, dir='target',
707
inconsistency_fatal=inconsistency_fatal)
709
source = self.make_source_with_b(x==1, 'source%s' % x)
710
target.insert_record_stream(source.get_record_stream(
711
[('b',)], 'unordered', False))
713
def test_inconsistent_redundant_inserts_warn(self):
714
"""Should not insert a record that is already present."""
716
def warning(template, args):
717
warnings.append(template % args)
718
_trace_warning = trace.warning
719
trace.warning = warning
721
self.do_inconsistent_inserts(inconsistency_fatal=False)
723
trace.warning = _trace_warning
724
self.assertEqual(["inconsistent details in skipped record: ('b',)"
725
" ('42 32 0 8', ((),)) ('74 32 0 8', ((('a',),),))"],
728
def test_inconsistent_redundant_inserts_raises(self):
729
e = self.assertRaises(errors.KnitCorrupt, self.do_inconsistent_inserts,
730
inconsistency_fatal=True)
731
self.assertContainsRe(str(e), "Knit.* corrupt: inconsistent details"
733
" \('b',\) \('42 32 0 8', \(\(\),\)\) \('74 32"
734
" 0 8', \(\(\('a',\),\),\)\)")
736
def test_clear_cache(self):
737
vf = self.make_source_with_b(True, 'source')
739
for record in vf.get_record_stream([('a',), ('b',)], 'unordered',
742
self.assertTrue(len(vf._group_cache) > 0)
744
self.assertEqual(0, len(vf._group_cache))
748
class StubGCVF(object):
749
def __init__(self, canned_get_blocks=None):
750
self._group_cache = {}
751
self._canned_get_blocks = canned_get_blocks or []
752
def _get_blocks(self, read_memos):
753
return iter(self._canned_get_blocks)
756
class Test_BatchingBlockFetcher(TestCaseWithGroupCompressVersionedFiles):
757
"""Simple whitebox unit tests for _BatchingBlockFetcher."""
759
def test_add_key_new_read_memo(self):
760
"""Adding a key with an uncached read_memo new to this batch adds that
761
read_memo to the list of memos to fetch.
763
# locations are: index_memo, ignored, parents, ignored
764
# where index_memo is: (idx, offset, len, factory_start, factory_end)
765
# and (idx, offset, size) is known as the 'read_memo', identifying the
767
read_memo = ('fake index', 100, 50)
769
('key',): (read_memo + (None, None), None, None, None)}
770
batcher = groupcompress._BatchingBlockFetcher(StubGCVF(), locations)
771
total_size = batcher.add_key(('key',))
772
self.assertEqual(50, total_size)
773
self.assertEqual([('key',)], batcher.keys)
774
self.assertEqual([read_memo], batcher.memos_to_get)
776
def test_add_key_duplicate_read_memo(self):
777
"""read_memos that occur multiple times in a batch will only be fetched
780
read_memo = ('fake index', 100, 50)
781
# Two keys, both sharing the same read memo (but different overall
784
('key1',): (read_memo + (0, 1), None, None, None),
785
('key2',): (read_memo + (1, 2), None, None, None)}
786
batcher = groupcompress._BatchingBlockFetcher(StubGCVF(), locations)
787
total_size = batcher.add_key(('key1',))
788
total_size = batcher.add_key(('key2',))
789
self.assertEqual(50, total_size)
790
self.assertEqual([('key1',), ('key2',)], batcher.keys)
791
self.assertEqual([read_memo], batcher.memos_to_get)
793
def test_add_key_cached_read_memo(self):
794
"""Adding a key with a cached read_memo will not cause that read_memo
795
to be added to the list to fetch.
797
read_memo = ('fake index', 100, 50)
799
gcvf._group_cache[read_memo] = 'fake block'
801
('key',): (read_memo + (None, None), None, None, None)}
802
batcher = groupcompress._BatchingBlockFetcher(gcvf, locations)
803
total_size = batcher.add_key(('key',))
804
self.assertEqual(0, total_size)
805
self.assertEqual([('key',)], batcher.keys)
806
self.assertEqual([], batcher.memos_to_get)
808
def test_yield_factories_empty(self):
809
"""An empty batch yields no factories."""
810
batcher = groupcompress._BatchingBlockFetcher(StubGCVF(), {})
811
self.assertEqual([], list(batcher.yield_factories()))
813
def test_yield_factories_calls_get_blocks(self):
814
"""Uncached memos are retrieved via get_blocks."""
815
read_memo1 = ('fake index', 100, 50)
816
read_memo2 = ('fake index', 150, 40)
819
(read_memo1, groupcompress.GroupCompressBlock()),
820
(read_memo2, groupcompress.GroupCompressBlock())])
822
('key1',): (read_memo1 + (None, None), None, None, None),
823
('key2',): (read_memo2 + (None, None), None, None, None)}
824
batcher = groupcompress._BatchingBlockFetcher(gcvf, locations)
825
batcher.add_key(('key1',))
826
batcher.add_key(('key2',))
827
factories = list(batcher.yield_factories(full_flush=True))
828
self.assertLength(2, factories)
829
keys = [f.key for f in factories]
830
kinds = [f.storage_kind for f in factories]
831
self.assertEqual([('key1',), ('key2',)], keys)
832
self.assertEqual(['groupcompress-block', 'groupcompress-block'], kinds)
834
def test_yield_factories_flushing(self):
835
"""yield_factories holds back on yielding results from the final block
836
unless passed full_flush=True.
838
fake_block = groupcompress.GroupCompressBlock()
839
read_memo = ('fake index', 100, 50)
841
gcvf._group_cache[read_memo] = fake_block
843
('key',): (read_memo + (None, None), None, None, None)}
844
batcher = groupcompress._BatchingBlockFetcher(gcvf, locations)
845
batcher.add_key(('key',))
846
self.assertEqual([], list(batcher.yield_factories()))
847
factories = list(batcher.yield_factories(full_flush=True))
848
self.assertLength(1, factories)
849
self.assertEqual(('key',), factories[0].key)
850
self.assertEqual('groupcompress-block', factories[0].storage_kind)
853
class TestLazyGroupCompress(tests.TestCaseWithTransport):
856
('key1',): "this is a text\n"
857
"with a reasonable amount of compressible bytes\n"
858
"which can be shared between various other texts\n",
859
('key2',): "another text\n"
860
"with a reasonable amount of compressible bytes\n"
861
"which can be shared between various other texts\n",
862
('key3',): "yet another text which won't be extracted\n"
863
"with a reasonable amount of compressible bytes\n"
864
"which can be shared between various other texts\n",
865
('key4',): "this will be extracted\n"
866
"but references most of its bytes from\n"
867
"yet another text which won't be extracted\n"
868
"with a reasonable amount of compressible bytes\n"
869
"which can be shared between various other texts\n",
871
def make_block(self, key_to_text):
872
"""Create a GroupCompressBlock, filling it with the given texts."""
873
compressor = groupcompress.GroupCompressor()
875
for key in sorted(key_to_text):
876
compressor.compress(key, key_to_text[key], None)
877
locs = dict((key, (start, end)) for key, (start, _, end, _)
878
in compressor.labels_deltas.iteritems())
879
block = compressor.flush()
880
raw_bytes = block.to_bytes()
881
return locs, groupcompress.GroupCompressBlock.from_bytes(raw_bytes)
883
def add_key_to_manager(self, key, locations, block, manager):
884
start, end = locations[key]
885
manager.add_factory(key, (), start, end)
887
def make_block_and_full_manager(self, texts):
888
locations, block = self.make_block(texts)
889
manager = groupcompress._LazyGroupContentManager(block)
890
for key in sorted(texts):
891
self.add_key_to_manager(key, locations, block, manager)
892
return block, manager
894
def test_get_fulltexts(self):
895
locations, block = self.make_block(self._texts)
896
manager = groupcompress._LazyGroupContentManager(block)
897
self.add_key_to_manager(('key1',), locations, block, manager)
898
self.add_key_to_manager(('key2',), locations, block, manager)
900
for record in manager.get_record_stream():
901
result_order.append(record.key)
902
text = self._texts[record.key]
903
self.assertEqual(text, record.get_bytes_as('fulltext'))
904
self.assertEqual([('key1',), ('key2',)], result_order)
906
# If we build the manager in the opposite order, we should get them
907
# back in the opposite order
908
manager = groupcompress._LazyGroupContentManager(block)
909
self.add_key_to_manager(('key2',), locations, block, manager)
910
self.add_key_to_manager(('key1',), locations, block, manager)
912
for record in manager.get_record_stream():
913
result_order.append(record.key)
914
text = self._texts[record.key]
915
self.assertEqual(text, record.get_bytes_as('fulltext'))
916
self.assertEqual([('key2',), ('key1',)], result_order)
918
def test__wire_bytes_no_keys(self):
919
locations, block = self.make_block(self._texts)
920
manager = groupcompress._LazyGroupContentManager(block)
921
wire_bytes = manager._wire_bytes()
922
block_length = len(block.to_bytes())
923
# We should have triggered a strip, since we aren't using any content
924
stripped_block = manager._block.to_bytes()
925
self.assertTrue(block_length > len(stripped_block))
926
empty_z_header = zlib.compress('')
927
self.assertEqual('groupcompress-block\n'
928
'8\n' # len(compress(''))
930
'%d\n'# compressed block len
933
% (len(stripped_block), empty_z_header,
937
def test__wire_bytes(self):
938
locations, block = self.make_block(self._texts)
939
manager = groupcompress._LazyGroupContentManager(block)
940
self.add_key_to_manager(('key1',), locations, block, manager)
941
self.add_key_to_manager(('key4',), locations, block, manager)
942
block_bytes = block.to_bytes()
943
wire_bytes = manager._wire_bytes()
944
(storage_kind, z_header_len, header_len,
945
block_len, rest) = wire_bytes.split('\n', 4)
946
z_header_len = int(z_header_len)
947
header_len = int(header_len)
948
block_len = int(block_len)
949
self.assertEqual('groupcompress-block', storage_kind)
950
self.assertEqual(34, z_header_len)
951
self.assertEqual(26, header_len)
952
self.assertEqual(len(block_bytes), block_len)
953
z_header = rest[:z_header_len]
954
header = zlib.decompress(z_header)
955
self.assertEqual(header_len, len(header))
956
entry1 = locations[('key1',)]
957
entry4 = locations[('key4',)]
958
self.assertEqualDiff('key1\n'
960
'%d\n' # start offset
966
% (entry1[0], entry1[1],
967
entry4[0], entry4[1]),
969
z_block = rest[z_header_len:]
970
self.assertEqual(block_bytes, z_block)
972
def test_from_bytes(self):
973
locations, block = self.make_block(self._texts)
974
manager = groupcompress._LazyGroupContentManager(block)
975
self.add_key_to_manager(('key1',), locations, block, manager)
976
self.add_key_to_manager(('key4',), locations, block, manager)
977
wire_bytes = manager._wire_bytes()
978
self.assertStartsWith(wire_bytes, 'groupcompress-block\n')
979
manager = groupcompress._LazyGroupContentManager.from_bytes(wire_bytes)
980
self.assertIsInstance(manager, groupcompress._LazyGroupContentManager)
981
self.assertEqual(2, len(manager._factories))
982
self.assertEqual(block._z_content, manager._block._z_content)
984
for record in manager.get_record_stream():
985
result_order.append(record.key)
986
text = self._texts[record.key]
987
self.assertEqual(text, record.get_bytes_as('fulltext'))
988
self.assertEqual([('key1',), ('key4',)], result_order)
990
def test__check_rebuild_no_changes(self):
991
block, manager = self.make_block_and_full_manager(self._texts)
992
manager._check_rebuild_block()
993
self.assertIs(block, manager._block)
995
def test__check_rebuild_only_one(self):
996
locations, block = self.make_block(self._texts)
997
manager = groupcompress._LazyGroupContentManager(block)
998
# Request just the first key, which should trigger a 'strip' action
999
self.add_key_to_manager(('key1',), locations, block, manager)
1000
manager._check_rebuild_block()
1001
self.assertIsNot(block, manager._block)
1002
self.assertTrue(block._content_length > manager._block._content_length)
1003
# We should be able to still get the content out of this block, though
1004
# it should only have 1 entry
1005
for record in manager.get_record_stream():
1006
self.assertEqual(('key1',), record.key)
1007
self.assertEqual(self._texts[record.key],
1008
record.get_bytes_as('fulltext'))
1010
def test__check_rebuild_middle(self):
1011
locations, block = self.make_block(self._texts)
1012
manager = groupcompress._LazyGroupContentManager(block)
1013
# Request a small key in the middle should trigger a 'rebuild'
1014
self.add_key_to_manager(('key4',), locations, block, manager)
1015
manager._check_rebuild_block()
1016
self.assertIsNot(block, manager._block)
1017
self.assertTrue(block._content_length > manager._block._content_length)
1018
for record in manager.get_record_stream():
1019
self.assertEqual(('key4',), record.key)
1020
self.assertEqual(self._texts[record.key],
1021
record.get_bytes_as('fulltext'))
1023
def test_check_is_well_utilized_all_keys(self):
1024
block, manager = self.make_block_and_full_manager(self._texts)
1025
self.assertFalse(manager.check_is_well_utilized())
1026
# Though we can fake it by changing the recommended minimum size
1027
manager._full_enough_block_size = block._content_length
1028
self.assertTrue(manager.check_is_well_utilized())
1029
# Setting it just above causes it to fail
1030
manager._full_enough_block_size = block._content_length + 1
1031
self.assertFalse(manager.check_is_well_utilized())
1032
# Setting the mixed-block size doesn't do anything, because the content
1033
# is considered to not be 'mixed'
1034
manager._full_enough_mixed_block_size = block._content_length
1035
self.assertFalse(manager.check_is_well_utilized())
1037
def test_check_is_well_utilized_mixed_keys(self):
1043
texts[f1k1] = self._texts[('key1',)]
1044
texts[f1k2] = self._texts[('key2',)]
1045
texts[f2k1] = self._texts[('key3',)]
1046
texts[f2k2] = self._texts[('key4',)]
1047
block, manager = self.make_block_and_full_manager(texts)
1048
self.assertFalse(manager.check_is_well_utilized())
1049
manager._full_enough_block_size = block._content_length
1050
self.assertTrue(manager.check_is_well_utilized())
1051
manager._full_enough_block_size = block._content_length + 1
1052
self.assertFalse(manager.check_is_well_utilized())
1053
manager._full_enough_mixed_block_size = block._content_length
1054
self.assertTrue(manager.check_is_well_utilized())
1056
def test_check_is_well_utilized_partial_use(self):
1057
locations, block = self.make_block(self._texts)
1058
manager = groupcompress._LazyGroupContentManager(block)
1059
manager._full_enough_block_size = block._content_length
1060
self.add_key_to_manager(('key1',), locations, block, manager)
1061
self.add_key_to_manager(('key2',), locations, block, manager)
1062
# Just using the content from key1 and 2 is not enough to be considered
1064
self.assertFalse(manager.check_is_well_utilized())
1065
# However if we add key3, then we have enough, as we only require 75%
1067
self.add_key_to_manager(('key4',), locations, block, manager)
1068
self.assertTrue(manager.check_is_well_utilized())
added
removed
bzrlib/tests/test_groupcompress.py
