1
# Copyright (C) 2005, 2006, 2007 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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
"""Tests for Knit data structure"""
19
from cStringIO import StringIO
32
from bzrlib.errors import (
33
RevisionAlreadyPresent,
38
from bzrlib.index import *
39
from bzrlib.knit import (
45
_VFContentMapGenerator,
52
from bzrlib.repofmt import pack_repo
53
from bzrlib.tests import (
57
TestCaseWithMemoryTransport,
58
TestCaseWithTransport,
61
from bzrlib.transport import get_transport
62
from bzrlib.transport.memory import MemoryTransport
63
from bzrlib.tuned_gzip import GzipFile
64
from bzrlib.versionedfile import (
67
network_bytes_to_kind_and_offset,
68
RecordingVersionedFilesDecorator,
72
class _CompiledKnitFeature(Feature):
76
import bzrlib._knit_load_data_c
81
def feature_name(self):
82
return 'bzrlib._knit_load_data_c'
84
CompiledKnitFeature = _CompiledKnitFeature()
87
class KnitContentTestsMixin(object):
89
def test_constructor(self):
90
content = self._make_content([])
93
content = self._make_content([])
94
self.assertEqual(content.text(), [])
96
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
97
self.assertEqual(content.text(), ["text1", "text2"])
100
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
101
copy = content.copy()
102
self.assertIsInstance(copy, content.__class__)
103
self.assertEqual(copy.annotate(), content.annotate())
105
def assertDerivedBlocksEqual(self, source, target, noeol=False):
106
"""Assert that the derived matching blocks match real output"""
107
source_lines = source.splitlines(True)
108
target_lines = target.splitlines(True)
110
if noeol and not line.endswith('\n'):
114
source_content = self._make_content([(None, nl(l)) for l in source_lines])
115
target_content = self._make_content([(None, nl(l)) for l in target_lines])
116
line_delta = source_content.line_delta(target_content)
117
delta_blocks = list(KnitContent.get_line_delta_blocks(line_delta,
118
source_lines, target_lines))
119
matcher = KnitSequenceMatcher(None, source_lines, target_lines)
120
matcher_blocks = list(list(matcher.get_matching_blocks()))
121
self.assertEqual(matcher_blocks, delta_blocks)
123
def test_get_line_delta_blocks(self):
124
self.assertDerivedBlocksEqual('a\nb\nc\n', 'q\nc\n')
125
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1)
126
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1A)
127
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1B)
128
self.assertDerivedBlocksEqual(TEXT_1B, TEXT_1A)
129
self.assertDerivedBlocksEqual(TEXT_1A, TEXT_1B)
130
self.assertDerivedBlocksEqual(TEXT_1A, '')
131
self.assertDerivedBlocksEqual('', TEXT_1A)
132
self.assertDerivedBlocksEqual('', '')
133
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd')
135
def test_get_line_delta_blocks_noeol(self):
136
"""Handle historical knit deltas safely
138
Some existing knit deltas don't consider the last line to differ
139
when the only difference whether it has a final newline.
141
New knit deltas appear to always consider the last line to differ
144
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd\n', noeol=True)
145
self.assertDerivedBlocksEqual('a\nb\nc\nd\n', 'a\nb\nc', noeol=True)
146
self.assertDerivedBlocksEqual('a\nb\nc\n', 'a\nb\nc', noeol=True)
147
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\n', noeol=True)
159
Banana cup cake recipe
169
Banana cup cake recipe
171
- bananas (do not use plantains!!!)
178
Banana cup cake recipe
195
class TestPlainKnitContent(TestCase, KnitContentTestsMixin):
197
def _make_content(self, lines):
198
annotated_content = AnnotatedKnitContent(lines)
199
return PlainKnitContent(annotated_content.text(), 'bogus')
201
def test_annotate(self):
202
content = self._make_content([])
203
self.assertEqual(content.annotate(), [])
205
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
206
self.assertEqual(content.annotate(),
207
[("bogus", "text1"), ("bogus", "text2")])
209
def test_line_delta(self):
210
content1 = self._make_content([("", "a"), ("", "b")])
211
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
212
self.assertEqual(content1.line_delta(content2),
213
[(1, 2, 2, ["a", "c"])])
215
def test_line_delta_iter(self):
216
content1 = self._make_content([("", "a"), ("", "b")])
217
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
218
it = content1.line_delta_iter(content2)
219
self.assertEqual(it.next(), (1, 2, 2, ["a", "c"]))
220
self.assertRaises(StopIteration, it.next)
223
class TestAnnotatedKnitContent(TestCase, KnitContentTestsMixin):
225
def _make_content(self, lines):
226
return AnnotatedKnitContent(lines)
228
def test_annotate(self):
229
content = self._make_content([])
230
self.assertEqual(content.annotate(), [])
232
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
233
self.assertEqual(content.annotate(),
234
[("origin1", "text1"), ("origin2", "text2")])
236
def test_line_delta(self):
237
content1 = self._make_content([("", "a"), ("", "b")])
238
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
239
self.assertEqual(content1.line_delta(content2),
240
[(1, 2, 2, [("", "a"), ("", "c")])])
242
def test_line_delta_iter(self):
243
content1 = self._make_content([("", "a"), ("", "b")])
244
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
245
it = content1.line_delta_iter(content2)
246
self.assertEqual(it.next(), (1, 2, 2, [("", "a"), ("", "c")]))
247
self.assertRaises(StopIteration, it.next)
250
class MockTransport(object):
252
def __init__(self, file_lines=None):
253
self.file_lines = file_lines
255
# We have no base directory for the MockTransport
258
def get(self, filename):
259
if self.file_lines is None:
260
raise NoSuchFile(filename)
262
return StringIO("\n".join(self.file_lines))
264
def readv(self, relpath, offsets):
265
fp = self.get(relpath)
266
for offset, size in offsets:
268
yield offset, fp.read(size)
270
def __getattr__(self, name):
271
def queue_call(*args, **kwargs):
272
self.calls.append((name, args, kwargs))
276
class MockReadvFailingTransport(MockTransport):
277
"""Fail in the middle of a readv() result.
279
This Transport will successfully yield the first two requested hunks, but
280
raise NoSuchFile for the rest.
283
def readv(self, relpath, offsets):
285
for result in MockTransport.readv(self, relpath, offsets):
287
# we use 2 because the first offset is the pack header, the second
288
# is the first actual content requset
290
raise errors.NoSuchFile(relpath)
294
class KnitRecordAccessTestsMixin(object):
295
"""Tests for getting and putting knit records."""
297
def test_add_raw_records(self):
298
"""Add_raw_records adds records retrievable later."""
299
access = self.get_access()
300
memos = access.add_raw_records([('key', 10)], '1234567890')
301
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
303
def test_add_several_raw_records(self):
304
"""add_raw_records with many records and read some back."""
305
access = self.get_access()
306
memos = access.add_raw_records([('key', 10), ('key2', 2), ('key3', 5)],
308
self.assertEqual(['1234567890', '12', '34567'],
309
list(access.get_raw_records(memos)))
310
self.assertEqual(['1234567890'],
311
list(access.get_raw_records(memos[0:1])))
312
self.assertEqual(['12'],
313
list(access.get_raw_records(memos[1:2])))
314
self.assertEqual(['34567'],
315
list(access.get_raw_records(memos[2:3])))
316
self.assertEqual(['1234567890', '34567'],
317
list(access.get_raw_records(memos[0:1] + memos[2:3])))
320
class TestKnitKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
321
"""Tests for the .kndx implementation."""
323
def get_access(self):
324
"""Get a .knit style access instance."""
325
mapper = ConstantMapper("foo")
326
access = _KnitKeyAccess(self.get_transport(), mapper)
330
class _TestException(Exception):
331
"""Just an exception for local tests to use."""
334
class TestPackKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
335
"""Tests for the pack based access."""
337
def get_access(self):
338
return self._get_access()[0]
340
def _get_access(self, packname='packfile', index='FOO'):
341
transport = self.get_transport()
342
def write_data(bytes):
343
transport.append_bytes(packname, bytes)
344
writer = pack.ContainerWriter(write_data)
346
access = _DirectPackAccess({})
347
access.set_writer(writer, index, (transport, packname))
348
return access, writer
350
def make_pack_file(self):
351
"""Create a pack file with 2 records."""
352
access, writer = self._get_access(packname='packname', index='foo')
354
memos.extend(access.add_raw_records([('key1', 10)], '1234567890'))
355
memos.extend(access.add_raw_records([('key2', 5)], '12345'))
359
def make_vf_for_retrying(self):
360
"""Create 3 packs and a reload function.
362
Originally, 2 pack files will have the data, but one will be missing.
363
And then the third will be used in place of the first two if reload()
366
:return: (versioned_file, reload_counter)
367
versioned_file a KnitVersionedFiles using the packs for access
369
tree = self.make_branch_and_memory_tree('tree')
372
tree.add([''], ['root-id'])
373
tree.commit('one', rev_id='rev-1')
374
tree.commit('two', rev_id='rev-2')
375
tree.commit('three', rev_id='rev-3')
376
# Pack these two revisions into another pack file, but don't remove
378
repo = tree.branch.repository
379
collection = repo._pack_collection
380
collection.ensure_loaded()
381
orig_packs = collection.packs
382
packer = pack_repo.Packer(collection, orig_packs, '.testpack')
383
new_pack = packer.pack()
385
vf = tree.branch.repository.revisions
388
tree.branch.repository.lock_read()
389
self.addCleanup(tree.branch.repository.unlock)
391
# Set up a reload() function that switches to using the new pack file
392
new_index = new_pack.revision_index
393
access_tuple = new_pack.access_tuple()
394
reload_counter = [0, 0, 0]
396
reload_counter[0] += 1
397
if reload_counter[1] > 0:
398
# We already reloaded, nothing more to do
399
reload_counter[2] += 1
401
reload_counter[1] += 1
402
vf._index._graph_index._indices[:] = [new_index]
403
vf._access._indices.clear()
404
vf._access._indices[new_index] = access_tuple
406
# Delete one of the pack files so the data will need to be reloaded. We
407
# will delete the file with 'rev-2' in it
408
trans, name = orig_packs[1].access_tuple()
410
# We don't have the index trigger reloading because we want to test
411
# that we reload when the .pack disappears
412
vf._access._reload_func = reload
413
return vf, reload_counter
415
def make_reload_func(self, return_val=True):
418
reload_called[0] += 1
420
return reload_called, reload
422
def make_retry_exception(self):
423
# We raise a real exception so that sys.exc_info() is properly
426
raise _TestException('foobar')
427
except _TestException, e:
428
retry_exc = errors.RetryWithNewPacks(None, reload_occurred=False,
429
exc_info=sys.exc_info())
432
def test_read_from_several_packs(self):
433
access, writer = self._get_access()
435
memos.extend(access.add_raw_records([('key', 10)], '1234567890'))
437
access, writer = self._get_access('pack2', 'FOOBAR')
438
memos.extend(access.add_raw_records([('key', 5)], '12345'))
440
access, writer = self._get_access('pack3', 'BAZ')
441
memos.extend(access.add_raw_records([('key', 5)], 'alpha'))
443
transport = self.get_transport()
444
access = _DirectPackAccess({"FOO":(transport, 'packfile'),
445
"FOOBAR":(transport, 'pack2'),
446
"BAZ":(transport, 'pack3')})
447
self.assertEqual(['1234567890', '12345', 'alpha'],
448
list(access.get_raw_records(memos)))
449
self.assertEqual(['1234567890'],
450
list(access.get_raw_records(memos[0:1])))
451
self.assertEqual(['12345'],
452
list(access.get_raw_records(memos[1:2])))
453
self.assertEqual(['alpha'],
454
list(access.get_raw_records(memos[2:3])))
455
self.assertEqual(['1234567890', 'alpha'],
456
list(access.get_raw_records(memos[0:1] + memos[2:3])))
458
def test_set_writer(self):
459
"""The writer should be settable post construction."""
460
access = _DirectPackAccess({})
461
transport = self.get_transport()
462
packname = 'packfile'
464
def write_data(bytes):
465
transport.append_bytes(packname, bytes)
466
writer = pack.ContainerWriter(write_data)
468
access.set_writer(writer, index, (transport, packname))
469
memos = access.add_raw_records([('key', 10)], '1234567890')
471
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
473
def test_missing_index_raises_retry(self):
474
memos = self.make_pack_file()
475
transport = self.get_transport()
476
reload_called, reload_func = self.make_reload_func()
477
# Note that the index key has changed from 'foo' to 'bar'
478
access = _DirectPackAccess({'bar':(transport, 'packname')},
479
reload_func=reload_func)
480
e = self.assertListRaises(errors.RetryWithNewPacks,
481
access.get_raw_records, memos)
482
# Because a key was passed in which does not match our index list, we
483
# assume that the listing was already reloaded
484
self.assertTrue(e.reload_occurred)
485
self.assertIsInstance(e.exc_info, tuple)
486
self.assertIs(e.exc_info[0], KeyError)
487
self.assertIsInstance(e.exc_info[1], KeyError)
489
def test_missing_index_raises_key_error_with_no_reload(self):
490
memos = self.make_pack_file()
491
transport = self.get_transport()
492
# Note that the index key has changed from 'foo' to 'bar'
493
access = _DirectPackAccess({'bar':(transport, 'packname')})
494
e = self.assertListRaises(KeyError, access.get_raw_records, memos)
496
def test_missing_file_raises_retry(self):
497
memos = self.make_pack_file()
498
transport = self.get_transport()
499
reload_called, reload_func = self.make_reload_func()
500
# Note that the 'filename' has been changed to 'different-packname'
501
access = _DirectPackAccess({'foo':(transport, 'different-packname')},
502
reload_func=reload_func)
503
e = self.assertListRaises(errors.RetryWithNewPacks,
504
access.get_raw_records, memos)
505
# The file has gone missing, so we assume we need to reload
506
self.assertFalse(e.reload_occurred)
507
self.assertIsInstance(e.exc_info, tuple)
508
self.assertIs(e.exc_info[0], errors.NoSuchFile)
509
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
510
self.assertEqual('different-packname', e.exc_info[1].path)
512
def test_missing_file_raises_no_such_file_with_no_reload(self):
513
memos = self.make_pack_file()
514
transport = self.get_transport()
515
# Note that the 'filename' has been changed to 'different-packname'
516
access = _DirectPackAccess({'foo':(transport, 'different-packname')})
517
e = self.assertListRaises(errors.NoSuchFile,
518
access.get_raw_records, memos)
520
def test_failing_readv_raises_retry(self):
521
memos = self.make_pack_file()
522
transport = self.get_transport()
523
failing_transport = MockReadvFailingTransport(
524
[transport.get_bytes('packname')])
525
reload_called, reload_func = self.make_reload_func()
526
access = _DirectPackAccess({'foo':(failing_transport, 'packname')},
527
reload_func=reload_func)
528
# Asking for a single record will not trigger the Mock failure
529
self.assertEqual(['1234567890'],
530
list(access.get_raw_records(memos[:1])))
531
self.assertEqual(['12345'],
532
list(access.get_raw_records(memos[1:2])))
533
# A multiple offset readv() will fail mid-way through
534
e = self.assertListRaises(errors.RetryWithNewPacks,
535
access.get_raw_records, memos)
536
# The file has gone missing, so we assume we need to reload
537
self.assertFalse(e.reload_occurred)
538
self.assertIsInstance(e.exc_info, tuple)
539
self.assertIs(e.exc_info[0], errors.NoSuchFile)
540
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
541
self.assertEqual('packname', e.exc_info[1].path)
543
def test_failing_readv_raises_no_such_file_with_no_reload(self):
544
memos = self.make_pack_file()
545
transport = self.get_transport()
546
failing_transport = MockReadvFailingTransport(
547
[transport.get_bytes('packname')])
548
reload_called, reload_func = self.make_reload_func()
549
access = _DirectPackAccess({'foo':(failing_transport, 'packname')})
550
# Asking for a single record will not trigger the Mock failure
551
self.assertEqual(['1234567890'],
552
list(access.get_raw_records(memos[:1])))
553
self.assertEqual(['12345'],
554
list(access.get_raw_records(memos[1:2])))
555
# A multiple offset readv() will fail mid-way through
556
e = self.assertListRaises(errors.NoSuchFile,
557
access.get_raw_records, memos)
559
def test_reload_or_raise_no_reload(self):
560
access = _DirectPackAccess({}, reload_func=None)
561
retry_exc = self.make_retry_exception()
562
# Without a reload_func, we will just re-raise the original exception
563
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
565
def test_reload_or_raise_reload_changed(self):
566
reload_called, reload_func = self.make_reload_func(return_val=True)
567
access = _DirectPackAccess({}, reload_func=reload_func)
568
retry_exc = self.make_retry_exception()
569
access.reload_or_raise(retry_exc)
570
self.assertEqual([1], reload_called)
571
retry_exc.reload_occurred=True
572
access.reload_or_raise(retry_exc)
573
self.assertEqual([2], reload_called)
575
def test_reload_or_raise_reload_no_change(self):
576
reload_called, reload_func = self.make_reload_func(return_val=False)
577
access = _DirectPackAccess({}, reload_func=reload_func)
578
retry_exc = self.make_retry_exception()
579
# If reload_occurred is False, then we consider it an error to have
580
# reload_func() return False (no changes).
581
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
582
self.assertEqual([1], reload_called)
583
retry_exc.reload_occurred=True
584
# If reload_occurred is True, then we assume nothing changed because
585
# it had changed earlier, but didn't change again
586
access.reload_or_raise(retry_exc)
587
self.assertEqual([2], reload_called)
589
def test_annotate_retries(self):
590
vf, reload_counter = self.make_vf_for_retrying()
591
# It is a little bit bogus to annotate the Revision VF, but it works,
592
# as we have ancestry stored there
594
reload_lines = vf.annotate(key)
595
self.assertEqual([1, 1, 0], reload_counter)
596
plain_lines = vf.annotate(key)
597
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
598
if reload_lines != plain_lines:
599
self.fail('Annotation was not identical with reloading.')
600
# Now delete the packs-in-use, which should trigger another reload, but
601
# this time we just raise an exception because we can't recover
602
for trans, name in vf._access._indices.itervalues():
604
self.assertRaises(errors.NoSuchFile, vf.annotate, key)
605
self.assertEqual([2, 1, 1], reload_counter)
607
def test__get_record_map_retries(self):
608
vf, reload_counter = self.make_vf_for_retrying()
609
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
610
records = vf._get_record_map(keys)
611
self.assertEqual(keys, sorted(records.keys()))
612
self.assertEqual([1, 1, 0], reload_counter)
613
# Now delete the packs-in-use, which should trigger another reload, but
614
# this time we just raise an exception because we can't recover
615
for trans, name in vf._access._indices.itervalues():
617
self.assertRaises(errors.NoSuchFile, vf._get_record_map, keys)
618
self.assertEqual([2, 1, 1], reload_counter)
620
def test_get_record_stream_retries(self):
621
vf, reload_counter = self.make_vf_for_retrying()
622
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
623
record_stream = vf.get_record_stream(keys, 'topological', False)
624
record = record_stream.next()
625
self.assertEqual(('rev-1',), record.key)
626
self.assertEqual([0, 0, 0], reload_counter)
627
record = record_stream.next()
628
self.assertEqual(('rev-2',), record.key)
629
self.assertEqual([1, 1, 0], reload_counter)
630
record = record_stream.next()
631
self.assertEqual(('rev-3',), record.key)
632
self.assertEqual([1, 1, 0], reload_counter)
633
# Now delete all pack files, and see that we raise the right error
634
for trans, name in vf._access._indices.itervalues():
636
self.assertListRaises(errors.NoSuchFile,
637
vf.get_record_stream, keys, 'topological', False)
639
def test_iter_lines_added_or_present_in_keys_retries(self):
640
vf, reload_counter = self.make_vf_for_retrying()
641
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
642
# Unfortunately, iter_lines_added_or_present_in_keys iterates the
643
# result in random order (determined by the iteration order from a
644
# set()), so we don't have any solid way to trigger whether data is
645
# read before or after. However we tried to delete the middle node to
646
# exercise the code well.
647
# What we care about is that all lines are always yielded, but not
650
reload_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
651
self.assertEqual([1, 1, 0], reload_counter)
652
# Now do it again, to make sure the result is equivalent
653
plain_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
654
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
655
self.assertEqual(plain_lines, reload_lines)
656
self.assertEqual(21, len(plain_lines))
657
# Now delete all pack files, and see that we raise the right error
658
for trans, name in vf._access._indices.itervalues():
660
self.assertListRaises(errors.NoSuchFile,
661
vf.iter_lines_added_or_present_in_keys, keys)
662
self.assertEqual([2, 1, 1], reload_counter)
664
def test_get_record_stream_yields_disk_sorted_order(self):
665
# if we get 'unordered' pick a semi-optimal order for reading. The
666
# order should be grouped by pack file, and then by position in file
667
repo = self.make_repository('test', format='pack-0.92')
669
self.addCleanup(repo.unlock)
670
repo.start_write_group()
672
vf.add_lines(('f-id', 'rev-5'), [('f-id', 'rev-4')], ['lines\n'])
673
vf.add_lines(('f-id', 'rev-1'), [], ['lines\n'])
674
vf.add_lines(('f-id', 'rev-2'), [('f-id', 'rev-1')], ['lines\n'])
675
repo.commit_write_group()
676
# We inserted them as rev-5, rev-1, rev-2, we should get them back in
678
stream = vf.get_record_stream([('f-id', 'rev-1'), ('f-id', 'rev-5'),
679
('f-id', 'rev-2')], 'unordered', False)
680
keys = [r.key for r in stream]
681
self.assertEqual([('f-id', 'rev-5'), ('f-id', 'rev-1'),
682
('f-id', 'rev-2')], keys)
683
repo.start_write_group()
684
vf.add_lines(('f-id', 'rev-4'), [('f-id', 'rev-3')], ['lines\n'])
685
vf.add_lines(('f-id', 'rev-3'), [('f-id', 'rev-2')], ['lines\n'])
686
vf.add_lines(('f-id', 'rev-6'), [('f-id', 'rev-5')], ['lines\n'])
687
repo.commit_write_group()
688
# Request in random order, to make sure the output order isn't based on
690
request_keys = set(('f-id', 'rev-%d' % i) for i in range(1, 7))
691
stream = vf.get_record_stream(request_keys, 'unordered', False)
692
keys = [r.key for r in stream]
693
# We want to get the keys back in disk order, but it doesn't matter
694
# which pack we read from first. So this can come back in 2 orders
695
alt1 = [('f-id', 'rev-%d' % i) for i in [4, 3, 6, 5, 1, 2]]
696
alt2 = [('f-id', 'rev-%d' % i) for i in [5, 1, 2, 4, 3, 6]]
697
if keys != alt1 and keys != alt2:
698
self.fail('Returned key order did not match either expected order.'
699
' expected %s or %s, not %s'
700
% (alt1, alt2, keys))
703
class LowLevelKnitDataTests(TestCase):
705
def create_gz_content(self, text):
707
gz_file = gzip.GzipFile(mode='wb', fileobj=sio)
710
return sio.getvalue()
712
def make_multiple_records(self):
713
"""Create the content for multiple records."""
714
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
716
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
721
record_1 = (0, len(gz_txt), sha1sum)
722
total_txt.append(gz_txt)
723
sha1sum = osutils.sha('baz\n').hexdigest()
724
gz_txt = self.create_gz_content('version rev-id-2 1 %s\n'
728
record_2 = (record_1[1], len(gz_txt), sha1sum)
729
total_txt.append(gz_txt)
730
return total_txt, record_1, record_2
732
def test_valid_knit_data(self):
733
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
734
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
739
transport = MockTransport([gz_txt])
740
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
741
knit = KnitVersionedFiles(None, access)
742
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
744
contents = list(knit._read_records_iter(records))
745
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'],
746
'4e48e2c9a3d2ca8a708cb0cc545700544efb5021')], contents)
748
raw_contents = list(knit._read_records_iter_raw(records))
749
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
751
def test_multiple_records_valid(self):
752
total_txt, record_1, record_2 = self.make_multiple_records()
753
transport = MockTransport([''.join(total_txt)])
754
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
755
knit = KnitVersionedFiles(None, access)
756
records = [(('rev-id-1',), (('rev-id-1',), record_1[0], record_1[1])),
757
(('rev-id-2',), (('rev-id-2',), record_2[0], record_2[1]))]
759
contents = list(knit._read_records_iter(records))
760
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'], record_1[2]),
761
(('rev-id-2',), ['baz\n'], record_2[2])],
764
raw_contents = list(knit._read_records_iter_raw(records))
765
self.assertEqual([(('rev-id-1',), total_txt[0], record_1[2]),
766
(('rev-id-2',), total_txt[1], record_2[2])],
769
def test_not_enough_lines(self):
770
sha1sum = osutils.sha('foo\n').hexdigest()
771
# record says 2 lines data says 1
772
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
776
transport = MockTransport([gz_txt])
777
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
778
knit = KnitVersionedFiles(None, access)
779
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
780
self.assertRaises(errors.KnitCorrupt, list,
781
knit._read_records_iter(records))
783
# read_records_iter_raw won't detect that sort of mismatch/corruption
784
raw_contents = list(knit._read_records_iter_raw(records))
785
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
787
def test_too_many_lines(self):
788
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
789
# record says 1 lines data says 2
790
gz_txt = self.create_gz_content('version rev-id-1 1 %s\n'
795
transport = MockTransport([gz_txt])
796
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
797
knit = KnitVersionedFiles(None, access)
798
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
799
self.assertRaises(errors.KnitCorrupt, list,
800
knit._read_records_iter(records))
802
# read_records_iter_raw won't detect that sort of mismatch/corruption
803
raw_contents = list(knit._read_records_iter_raw(records))
804
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
806
def test_mismatched_version_id(self):
807
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
808
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
813
transport = MockTransport([gz_txt])
814
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
815
knit = KnitVersionedFiles(None, access)
816
# We are asking for rev-id-2, but the data is rev-id-1
817
records = [(('rev-id-2',), (('rev-id-2',), 0, len(gz_txt)))]
818
self.assertRaises(errors.KnitCorrupt, list,
819
knit._read_records_iter(records))
821
# read_records_iter_raw detects mismatches in the header
822
self.assertRaises(errors.KnitCorrupt, list,
823
knit._read_records_iter_raw(records))
825
def test_uncompressed_data(self):
826
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
827
txt = ('version rev-id-1 2 %s\n'
832
transport = MockTransport([txt])
833
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
834
knit = KnitVersionedFiles(None, access)
835
records = [(('rev-id-1',), (('rev-id-1',), 0, len(txt)))]
837
# We don't have valid gzip data ==> corrupt
838
self.assertRaises(errors.KnitCorrupt, list,
839
knit._read_records_iter(records))
841
# read_records_iter_raw will notice the bad data
842
self.assertRaises(errors.KnitCorrupt, list,
843
knit._read_records_iter_raw(records))
845
def test_corrupted_data(self):
846
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
847
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
852
# Change 2 bytes in the middle to \xff
853
gz_txt = gz_txt[:10] + '\xff\xff' + gz_txt[12:]
854
transport = MockTransport([gz_txt])
855
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
856
knit = KnitVersionedFiles(None, access)
857
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
858
self.assertRaises(errors.KnitCorrupt, list,
859
knit._read_records_iter(records))
860
# read_records_iter_raw will barf on bad gz data
861
self.assertRaises(errors.KnitCorrupt, list,
862
knit._read_records_iter_raw(records))
865
class LowLevelKnitIndexTests(TestCase):
867
def get_knit_index(self, transport, name, mode):
868
mapper = ConstantMapper(name)
869
orig = knit._load_data
871
knit._load_data = orig
872
self.addCleanup(reset)
873
from bzrlib._knit_load_data_py import _load_data_py
874
knit._load_data = _load_data_py
875
allow_writes = lambda: 'w' in mode
876
return _KndxIndex(transport, mapper, lambda:None, allow_writes, lambda:True)
878
def test_create_file(self):
879
transport = MockTransport()
880
index = self.get_knit_index(transport, "filename", "w")
882
call = transport.calls.pop(0)
883
# call[1][1] is a StringIO - we can't test it by simple equality.
884
self.assertEqual('put_file_non_atomic', call[0])
885
self.assertEqual('filename.kndx', call[1][0])
886
# With no history, _KndxIndex writes a new index:
887
self.assertEqual(_KndxIndex.HEADER,
888
call[1][1].getvalue())
889
self.assertEqual({'create_parent_dir': True}, call[2])
891
def test_read_utf8_version_id(self):
892
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
893
utf8_revision_id = unicode_revision_id.encode('utf-8')
894
transport = MockTransport([
896
'%s option 0 1 :' % (utf8_revision_id,)
898
index = self.get_knit_index(transport, "filename", "r")
899
# _KndxIndex is a private class, and deals in utf8 revision_ids, not
900
# Unicode revision_ids.
901
self.assertEqual({(utf8_revision_id,):()},
902
index.get_parent_map(index.keys()))
903
self.assertFalse((unicode_revision_id,) in index.keys())
905
def test_read_utf8_parents(self):
906
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
907
utf8_revision_id = unicode_revision_id.encode('utf-8')
908
transport = MockTransport([
910
"version option 0 1 .%s :" % (utf8_revision_id,)
912
index = self.get_knit_index(transport, "filename", "r")
913
self.assertEqual({("version",):((utf8_revision_id,),)},
914
index.get_parent_map(index.keys()))
916
def test_read_ignore_corrupted_lines(self):
917
transport = MockTransport([
920
"corrupted options 0 1 .b .c ",
921
"version options 0 1 :"
923
index = self.get_knit_index(transport, "filename", "r")
924
self.assertEqual(1, len(index.keys()))
925
self.assertEqual(set([("version",)]), index.keys())
927
def test_read_corrupted_header(self):
928
transport = MockTransport(['not a bzr knit index header\n'])
929
index = self.get_knit_index(transport, "filename", "r")
930
self.assertRaises(KnitHeaderError, index.keys)
932
def test_read_duplicate_entries(self):
933
transport = MockTransport([
935
"parent options 0 1 :",
936
"version options1 0 1 0 :",
937
"version options2 1 2 .other :",
938
"version options3 3 4 0 .other :"
940
index = self.get_knit_index(transport, "filename", "r")
941
self.assertEqual(2, len(index.keys()))
942
# check that the index used is the first one written. (Specific
943
# to KnitIndex style indices.
944
self.assertEqual("1", index._dictionary_compress([("version",)]))
945
self.assertEqual((("version",), 3, 4), index.get_position(("version",)))
946
self.assertEqual(["options3"], index.get_options(("version",)))
947
self.assertEqual({("version",):(("parent",), ("other",))},
948
index.get_parent_map([("version",)]))
950
def test_read_compressed_parents(self):
951
transport = MockTransport([
955
"c option 0 1 1 0 :",
957
index = self.get_knit_index(transport, "filename", "r")
958
self.assertEqual({("b",):(("a",),), ("c",):(("b",), ("a",))},
959
index.get_parent_map([("b",), ("c",)]))
961
def test_write_utf8_version_id(self):
962
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
963
utf8_revision_id = unicode_revision_id.encode('utf-8')
964
transport = MockTransport([
967
index = self.get_knit_index(transport, "filename", "r")
969
((utf8_revision_id,), ["option"], ((utf8_revision_id,), 0, 1), [])])
970
call = transport.calls.pop(0)
971
# call[1][1] is a StringIO - we can't test it by simple equality.
972
self.assertEqual('put_file_non_atomic', call[0])
973
self.assertEqual('filename.kndx', call[1][0])
974
# With no history, _KndxIndex writes a new index:
975
self.assertEqual(_KndxIndex.HEADER +
976
"\n%s option 0 1 :" % (utf8_revision_id,),
977
call[1][1].getvalue())
978
self.assertEqual({'create_parent_dir': True}, call[2])
980
def test_write_utf8_parents(self):
981
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
982
utf8_revision_id = unicode_revision_id.encode('utf-8')
983
transport = MockTransport([
986
index = self.get_knit_index(transport, "filename", "r")
988
(("version",), ["option"], (("version",), 0, 1), [(utf8_revision_id,)])])
989
call = transport.calls.pop(0)
990
# call[1][1] is a StringIO - we can't test it by simple equality.
991
self.assertEqual('put_file_non_atomic', call[0])
992
self.assertEqual('filename.kndx', call[1][0])
993
# With no history, _KndxIndex writes a new index:
994
self.assertEqual(_KndxIndex.HEADER +
995
"\nversion option 0 1 .%s :" % (utf8_revision_id,),
996
call[1][1].getvalue())
997
self.assertEqual({'create_parent_dir': True}, call[2])
1000
transport = MockTransport([
1003
index = self.get_knit_index(transport, "filename", "r")
1005
self.assertEqual(set(), index.keys())
1007
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1008
self.assertEqual(set([("a",)]), index.keys())
1010
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1011
self.assertEqual(set([("a",)]), index.keys())
1013
index.add_records([(("b",), ["option"], (("b",), 0, 1), [])])
1014
self.assertEqual(set([("a",), ("b",)]), index.keys())
1016
def add_a_b(self, index, random_id=None):
1018
if random_id is not None:
1019
kwargs["random_id"] = random_id
1021
(("a",), ["option"], (("a",), 0, 1), [("b",)]),
1022
(("a",), ["opt"], (("a",), 1, 2), [("c",)]),
1023
(("b",), ["option"], (("b",), 2, 3), [("a",)])
1026
def assertIndexIsAB(self, index):
1031
index.get_parent_map(index.keys()))
1032
self.assertEqual((("a",), 1, 2), index.get_position(("a",)))
1033
self.assertEqual((("b",), 2, 3), index.get_position(("b",)))
1034
self.assertEqual(["opt"], index.get_options(("a",)))
1036
def test_add_versions(self):
1037
transport = MockTransport([
1040
index = self.get_knit_index(transport, "filename", "r")
1043
call = transport.calls.pop(0)
1044
# call[1][1] is a StringIO - we can't test it by simple equality.
1045
self.assertEqual('put_file_non_atomic', call[0])
1046
self.assertEqual('filename.kndx', call[1][0])
1047
# With no history, _KndxIndex writes a new index:
1050
"\na option 0 1 .b :"
1052
"\nb option 2 3 0 :",
1053
call[1][1].getvalue())
1054
self.assertEqual({'create_parent_dir': True}, call[2])
1055
self.assertIndexIsAB(index)
1057
def test_add_versions_random_id_is_accepted(self):
1058
transport = MockTransport([
1061
index = self.get_knit_index(transport, "filename", "r")
1062
self.add_a_b(index, random_id=True)
1064
def test_delay_create_and_add_versions(self):
1065
transport = MockTransport()
1067
index = self.get_knit_index(transport, "filename", "w")
1069
self.assertEqual([], transport.calls)
1072
#[ {"dir_mode": 0777, "create_parent_dir": True, "mode": "wb"},
1074
# Two calls: one during which we load the existing index (and when its
1075
# missing create it), then a second where we write the contents out.
1076
self.assertEqual(2, len(transport.calls))
1077
call = transport.calls.pop(0)
1078
self.assertEqual('put_file_non_atomic', call[0])
1079
self.assertEqual('filename.kndx', call[1][0])
1080
# With no history, _KndxIndex writes a new index:
1081
self.assertEqual(_KndxIndex.HEADER, call[1][1].getvalue())
1082
self.assertEqual({'create_parent_dir': True}, call[2])
1083
call = transport.calls.pop(0)
1084
# call[1][1] is a StringIO - we can't test it by simple equality.
1085
self.assertEqual('put_file_non_atomic', call[0])
1086
self.assertEqual('filename.kndx', call[1][0])
1087
# With no history, _KndxIndex writes a new index:
1090
"\na option 0 1 .b :"
1092
"\nb option 2 3 0 :",
1093
call[1][1].getvalue())
1094
self.assertEqual({'create_parent_dir': True}, call[2])
1096
def assertTotalBuildSize(self, size, keys, positions):
1097
self.assertEqual(size,
1098
knit._get_total_build_size(None, keys, positions))
1100
def test__get_total_build_size(self):
1102
('a',): (('fulltext', False), (('a',), 0, 100), None),
1103
('b',): (('line-delta', False), (('b',), 100, 21), ('a',)),
1104
('c',): (('line-delta', False), (('c',), 121, 35), ('b',)),
1105
('d',): (('line-delta', False), (('d',), 156, 12), ('b',)),
1107
self.assertTotalBuildSize(100, [('a',)], positions)
1108
self.assertTotalBuildSize(121, [('b',)], positions)
1109
# c needs both a & b
1110
self.assertTotalBuildSize(156, [('c',)], positions)
1111
# we shouldn't count 'b' twice
1112
self.assertTotalBuildSize(156, [('b',), ('c',)], positions)
1113
self.assertTotalBuildSize(133, [('d',)], positions)
1114
self.assertTotalBuildSize(168, [('c',), ('d',)], positions)
1116
def test_get_position(self):
1117
transport = MockTransport([
1122
index = self.get_knit_index(transport, "filename", "r")
1124
self.assertEqual((("a",), 0, 1), index.get_position(("a",)))
1125
self.assertEqual((("b",), 1, 2), index.get_position(("b",)))
1127
def test_get_method(self):
1128
transport = MockTransport([
1130
"a fulltext,unknown 0 1 :",
1131
"b unknown,line-delta 1 2 :",
1134
index = self.get_knit_index(transport, "filename", "r")
1136
self.assertEqual("fulltext", index.get_method("a"))
1137
self.assertEqual("line-delta", index.get_method("b"))
1138
self.assertRaises(errors.KnitIndexUnknownMethod, index.get_method, "c")
1140
def test_get_options(self):
1141
transport = MockTransport([
1146
index = self.get_knit_index(transport, "filename", "r")
1148
self.assertEqual(["opt1"], index.get_options("a"))
1149
self.assertEqual(["opt2", "opt3"], index.get_options("b"))
1151
def test_get_parent_map(self):
1152
transport = MockTransport([
1155
"b option 1 2 0 .c :",
1156
"c option 1 2 1 0 .e :"
1158
index = self.get_knit_index(transport, "filename", "r")
1162
("b",):(("a",), ("c",)),
1163
("c",):(("b",), ("a",), ("e",)),
1164
}, index.get_parent_map(index.keys()))
1166
def test_impossible_parent(self):
1167
"""Test we get KnitCorrupt if the parent couldn't possibly exist."""
1168
transport = MockTransport([
1171
"b option 0 1 4 :" # We don't have a 4th record
1173
index = self.get_knit_index(transport, 'filename', 'r')
1175
self.assertRaises(errors.KnitCorrupt, index.keys)
1176
except TypeError, e:
1177
if (str(e) == ('exceptions must be strings, classes, or instances,'
1178
' not exceptions.IndexError')
1179
and sys.version_info[0:2] >= (2,5)):
1180
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1181
' raising new style exceptions with python'
1186
def test_corrupted_parent(self):
1187
transport = MockTransport([
1191
"c option 0 1 1v :", # Can't have a parent of '1v'
1193
index = self.get_knit_index(transport, 'filename', 'r')
1195
self.assertRaises(errors.KnitCorrupt, index.keys)
1196
except TypeError, e:
1197
if (str(e) == ('exceptions must be strings, classes, or instances,'
1198
' not exceptions.ValueError')
1199
and sys.version_info[0:2] >= (2,5)):
1200
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1201
' raising new style exceptions with python'
1206
def test_corrupted_parent_in_list(self):
1207
transport = MockTransport([
1211
"c option 0 1 1 v :", # Can't have a parent of 'v'
1213
index = self.get_knit_index(transport, 'filename', 'r')
1215
self.assertRaises(errors.KnitCorrupt, index.keys)
1216
except TypeError, e:
1217
if (str(e) == ('exceptions must be strings, classes, or instances,'
1218
' not exceptions.ValueError')
1219
and sys.version_info[0:2] >= (2,5)):
1220
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1221
' raising new style exceptions with python'
1226
def test_invalid_position(self):
1227
transport = MockTransport([
1231
index = self.get_knit_index(transport, 'filename', 'r')
1233
self.assertRaises(errors.KnitCorrupt, index.keys)
1234
except TypeError, e:
1235
if (str(e) == ('exceptions must be strings, classes, or instances,'
1236
' not exceptions.ValueError')
1237
and sys.version_info[0:2] >= (2,5)):
1238
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1239
' raising new style exceptions with python'
1244
def test_invalid_size(self):
1245
transport = MockTransport([
1249
index = self.get_knit_index(transport, 'filename', 'r')
1251
self.assertRaises(errors.KnitCorrupt, index.keys)
1252
except TypeError, e:
1253
if (str(e) == ('exceptions must be strings, classes, or instances,'
1254
' not exceptions.ValueError')
1255
and sys.version_info[0:2] >= (2,5)):
1256
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1257
' raising new style exceptions with python'
1262
def test_scan_unvalidated_index_not_implemented(self):
1263
transport = MockTransport()
1264
index = self.get_knit_index(transport, 'filename', 'r')
1266
NotImplementedError, index.scan_unvalidated_index,
1267
'dummy graph_index')
1269
NotImplementedError, index.get_missing_compression_parents)
1271
def test_short_line(self):
1272
transport = MockTransport([
1275
"b option 10 10 0", # This line isn't terminated, ignored
1277
index = self.get_knit_index(transport, "filename", "r")
1278
self.assertEqual(set([('a',)]), index.keys())
1280
def test_skip_incomplete_record(self):
1281
# A line with bogus data should just be skipped
1282
transport = MockTransport([
1285
"b option 10 10 0", # This line isn't terminated, ignored
1286
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1288
index = self.get_knit_index(transport, "filename", "r")
1289
self.assertEqual(set([('a',), ('c',)]), index.keys())
1291
def test_trailing_characters(self):
1292
# A line with bogus data should just be skipped
1293
transport = MockTransport([
1296
"b option 10 10 0 :a", # This line has extra trailing characters
1297
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1299
index = self.get_knit_index(transport, "filename", "r")
1300
self.assertEqual(set([('a',), ('c',)]), index.keys())
1303
class LowLevelKnitIndexTests_c(LowLevelKnitIndexTests):
1305
_test_needs_features = [CompiledKnitFeature]
1307
def get_knit_index(self, transport, name, mode):
1308
mapper = ConstantMapper(name)
1309
orig = knit._load_data
1311
knit._load_data = orig
1312
self.addCleanup(reset)
1313
from bzrlib._knit_load_data_c import _load_data_c
1314
knit._load_data = _load_data_c
1315
allow_writes = lambda: mode == 'w'
1316
return _KndxIndex(transport, mapper, lambda:None, allow_writes, lambda:True)
1319
class KnitTests(TestCaseWithTransport):
1320
"""Class containing knit test helper routines."""
1322
def make_test_knit(self, annotate=False, name='test'):
1323
mapper = ConstantMapper(name)
1324
return make_file_factory(annotate, mapper)(self.get_transport())
1327
class TestBadShaError(KnitTests):
1328
"""Tests for handling of sha errors."""
1330
def test_sha_exception_has_text(self):
1331
# having the failed text included in the error allows for recovery.
1332
source = self.make_test_knit()
1333
target = self.make_test_knit(name="target")
1334
if not source._max_delta_chain:
1335
raise TestNotApplicable(
1336
"cannot get delta-caused sha failures without deltas.")
1339
broken = ('broken',)
1340
source.add_lines(basis, (), ['foo\n'])
1341
source.add_lines(broken, (basis,), ['foo\n', 'bar\n'])
1342
# Seed target with a bad basis text
1343
target.add_lines(basis, (), ['gam\n'])
1344
target.insert_record_stream(
1345
source.get_record_stream([broken], 'unordered', False))
1346
err = self.assertRaises(errors.KnitCorrupt,
1347
target.get_record_stream([broken], 'unordered', True
1348
).next().get_bytes_as, 'chunked')
1349
self.assertEqual(['gam\n', 'bar\n'], err.content)
1350
# Test for formatting with live data
1351
self.assertStartsWith(str(err), "Knit ")
1354
class TestKnitIndex(KnitTests):
1356
def test_add_versions_dictionary_compresses(self):
1357
"""Adding versions to the index should update the lookup dict"""
1358
knit = self.make_test_knit()
1360
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1361
self.check_file_contents('test.kndx',
1362
'# bzr knit index 8\n'
1364
'a-1 fulltext 0 0 :'
1367
(('a-2',), ['fulltext'], (('a-2',), 0, 0), [('a-1',)]),
1368
(('a-3',), ['fulltext'], (('a-3',), 0, 0), [('a-2',)]),
1370
self.check_file_contents('test.kndx',
1371
'# bzr knit index 8\n'
1373
'a-1 fulltext 0 0 :\n'
1374
'a-2 fulltext 0 0 0 :\n'
1375
'a-3 fulltext 0 0 1 :'
1377
self.assertEqual(set([('a-3',), ('a-1',), ('a-2',)]), idx.keys())
1379
('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False)),
1380
('a-2',): ((('a-2',), 0, 0), None, (('a-1',),), ('fulltext', False)),
1381
('a-3',): ((('a-3',), 0, 0), None, (('a-2',),), ('fulltext', False)),
1382
}, idx.get_build_details(idx.keys()))
1383
self.assertEqual({('a-1',):(),
1384
('a-2',):(('a-1',),),
1385
('a-3',):(('a-2',),),},
1386
idx.get_parent_map(idx.keys()))
1388
def test_add_versions_fails_clean(self):
1389
"""If add_versions fails in the middle, it restores a pristine state.
1391
Any modifications that are made to the index are reset if all versions
1394
# This cheats a little bit by passing in a generator which will
1395
# raise an exception before the processing finishes
1396
# Other possibilities would be to have an version with the wrong number
1397
# of entries, or to make the backing transport unable to write any
1400
knit = self.make_test_knit()
1402
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1404
class StopEarly(Exception):
1407
def generate_failure():
1408
"""Add some entries and then raise an exception"""
1409
yield (('a-2',), ['fulltext'], (None, 0, 0), ('a-1',))
1410
yield (('a-3',), ['fulltext'], (None, 0, 0), ('a-2',))
1413
# Assert the pre-condition
1415
self.assertEqual(set([('a-1',)]), set(idx.keys()))
1417
{('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False))},
1418
idx.get_build_details([('a-1',)]))
1419
self.assertEqual({('a-1',):()}, idx.get_parent_map(idx.keys()))
1422
self.assertRaises(StopEarly, idx.add_records, generate_failure())
1423
# And it shouldn't be modified
1426
def test_knit_index_ignores_empty_files(self):
1427
# There was a race condition in older bzr, where a ^C at the right time
1428
# could leave an empty .kndx file, which bzr would later claim was a
1429
# corrupted file since the header was not present. In reality, the file
1430
# just wasn't created, so it should be ignored.
1431
t = get_transport('.')
1432
t.put_bytes('test.kndx', '')
1434
knit = self.make_test_knit()
1436
def test_knit_index_checks_header(self):
1437
t = get_transport('.')
1438
t.put_bytes('test.kndx', '# not really a knit header\n\n')
1439
k = self.make_test_knit()
1440
self.assertRaises(KnitHeaderError, k.keys)
1443
class TestGraphIndexKnit(KnitTests):
1444
"""Tests for knits using a GraphIndex rather than a KnitIndex."""
1446
def make_g_index(self, name, ref_lists=0, nodes=[]):
1447
builder = GraphIndexBuilder(ref_lists)
1448
for node, references, value in nodes:
1449
builder.add_node(node, references, value)
1450
stream = builder.finish()
1451
trans = self.get_transport()
1452
size = trans.put_file(name, stream)
1453
return GraphIndex(trans, name, size)
1455
def two_graph_index(self, deltas=False, catch_adds=False):
1456
"""Build a two-graph index.
1458
:param deltas: If true, use underlying indices with two node-ref
1459
lists and 'parent' set to a delta-compressed against tail.
1461
# build a complex graph across several indices.
1463
# delta compression inn the index
1464
index1 = self.make_g_index('1', 2, [
1465
(('tip', ), 'N0 100', ([('parent', )], [], )),
1466
(('tail', ), '', ([], []))])
1467
index2 = self.make_g_index('2', 2, [
1468
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], [('tail', )])),
1469
(('separate', ), '', ([], []))])
1471
# just blob location and graph in the index.
1472
index1 = self.make_g_index('1', 1, [
1473
(('tip', ), 'N0 100', ([('parent', )], )),
1474
(('tail', ), '', ([], ))])
1475
index2 = self.make_g_index('2', 1, [
1476
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], )),
1477
(('separate', ), '', ([], ))])
1478
combined_index = CombinedGraphIndex([index1, index2])
1480
self.combined_index = combined_index
1481
self.caught_entries = []
1482
add_callback = self.catch_add
1485
return _KnitGraphIndex(combined_index, lambda:True, deltas=deltas,
1486
add_callback=add_callback)
1488
def test_keys(self):
1489
index = self.two_graph_index()
1490
self.assertEqual(set([('tail',), ('tip',), ('parent',), ('separate',)]),
1493
def test_get_position(self):
1494
index = self.two_graph_index()
1495
self.assertEqual((index._graph_index._indices[0], 0, 100), index.get_position(('tip',)))
1496
self.assertEqual((index._graph_index._indices[1], 100, 78), index.get_position(('parent',)))
1498
def test_get_method_deltas(self):
1499
index = self.two_graph_index(deltas=True)
1500
self.assertEqual('fulltext', index.get_method(('tip',)))
1501
self.assertEqual('line-delta', index.get_method(('parent',)))
1503
def test_get_method_no_deltas(self):
1504
# check that the parent-history lookup is ignored with deltas=False.
1505
index = self.two_graph_index(deltas=False)
1506
self.assertEqual('fulltext', index.get_method(('tip',)))
1507
self.assertEqual('fulltext', index.get_method(('parent',)))
1509
def test_get_options_deltas(self):
1510
index = self.two_graph_index(deltas=True)
1511
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1512
self.assertEqual(['line-delta'], index.get_options(('parent',)))
1514
def test_get_options_no_deltas(self):
1515
# check that the parent-history lookup is ignored with deltas=False.
1516
index = self.two_graph_index(deltas=False)
1517
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1518
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1520
def test_get_parent_map(self):
1521
index = self.two_graph_index()
1522
self.assertEqual({('parent',):(('tail',), ('ghost',))},
1523
index.get_parent_map([('parent',), ('ghost',)]))
1525
def catch_add(self, entries):
1526
self.caught_entries.append(entries)
1528
def test_add_no_callback_errors(self):
1529
index = self.two_graph_index()
1530
self.assertRaises(errors.ReadOnlyError, index.add_records,
1531
[(('new',), 'fulltext,no-eol', (None, 50, 60), ['separate'])])
1533
def test_add_version_smoke(self):
1534
index = self.two_graph_index(catch_adds=True)
1535
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60),
1537
self.assertEqual([[(('new', ), 'N50 60', ((('separate',),),))]],
1538
self.caught_entries)
1540
def test_add_version_delta_not_delta_index(self):
1541
index = self.two_graph_index(catch_adds=True)
1542
self.assertRaises(errors.KnitCorrupt, index.add_records,
1543
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1544
self.assertEqual([], self.caught_entries)
1546
def test_add_version_same_dup(self):
1547
index = self.two_graph_index(catch_adds=True)
1548
# options can be spelt two different ways
1549
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
1550
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
1551
# position/length are ignored (because each pack could have fulltext or
1552
# delta, and be at a different position.
1553
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1555
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1557
# but neither should have added data:
1558
self.assertEqual([[], [], [], []], self.caught_entries)
1560
def test_add_version_different_dup(self):
1561
index = self.two_graph_index(deltas=True, catch_adds=True)
1563
self.assertRaises(errors.KnitCorrupt, index.add_records,
1564
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1565
self.assertRaises(errors.KnitCorrupt, index.add_records,
1566
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1568
self.assertRaises(errors.KnitCorrupt, index.add_records,
1569
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1570
self.assertEqual([], self.caught_entries)
1572
def test_add_versions_nodeltas(self):
1573
index = self.two_graph_index(catch_adds=True)
1575
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1576
(('new2',), 'fulltext', (None, 0, 6), [('new',)]),
1578
self.assertEqual([(('new', ), 'N50 60', ((('separate',),),)),
1579
(('new2', ), ' 0 6', ((('new',),),))],
1580
sorted(self.caught_entries[0]))
1581
self.assertEqual(1, len(self.caught_entries))
1583
def test_add_versions_deltas(self):
1584
index = self.two_graph_index(deltas=True, catch_adds=True)
1586
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1587
(('new2',), 'line-delta', (None, 0, 6), [('new',)]),
1589
self.assertEqual([(('new', ), 'N50 60', ((('separate',),), ())),
1590
(('new2', ), ' 0 6', ((('new',),), (('new',),), ))],
1591
sorted(self.caught_entries[0]))
1592
self.assertEqual(1, len(self.caught_entries))
1594
def test_add_versions_delta_not_delta_index(self):
1595
index = self.two_graph_index(catch_adds=True)
1596
self.assertRaises(errors.KnitCorrupt, index.add_records,
1597
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1598
self.assertEqual([], self.caught_entries)
1600
def test_add_versions_random_id_accepted(self):
1601
index = self.two_graph_index(catch_adds=True)
1602
index.add_records([], random_id=True)
1604
def test_add_versions_same_dup(self):
1605
index = self.two_graph_index(catch_adds=True)
1606
# options can be spelt two different ways
1607
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100),
1609
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100),
1611
# position/length are ignored (because each pack could have fulltext or
1612
# delta, and be at a different position.
1613
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1615
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1617
# but neither should have added data.
1618
self.assertEqual([[], [], [], []], self.caught_entries)
1620
def test_add_versions_different_dup(self):
1621
index = self.two_graph_index(deltas=True, catch_adds=True)
1623
self.assertRaises(errors.KnitCorrupt, index.add_records,
1624
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1625
self.assertRaises(errors.KnitCorrupt, index.add_records,
1626
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1628
self.assertRaises(errors.KnitCorrupt, index.add_records,
1629
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1630
# change options in the second record
1631
self.assertRaises(errors.KnitCorrupt, index.add_records,
1632
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)]),
1633
(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1634
self.assertEqual([], self.caught_entries)
1636
def make_g_index_missing_compression_parent(self):
1637
graph_index = self.make_g_index('missing_comp', 2,
1638
[(('tip', ), ' 100 78',
1639
([('missing-parent', ), ('ghost', )], [('missing-parent', )]))])
1642
def make_g_index_no_external_refs(self):
1643
graph_index = self.make_g_index('no_external_refs', 2,
1644
[(('rev', ), ' 100 78',
1645
([('parent', ), ('ghost', )], []))])
1648
def test_add_good_unvalidated_index(self):
1649
unvalidated = self.make_g_index_no_external_refs()
1650
combined = CombinedGraphIndex([unvalidated])
1651
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1652
index.scan_unvalidated_index(unvalidated)
1653
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1655
def test_add_incomplete_unvalidated_index(self):
1656
unvalidated = self.make_g_index_missing_compression_parent()
1657
combined = CombinedGraphIndex([unvalidated])
1658
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1659
index.scan_unvalidated_index(unvalidated)
1660
# This also checks that its only the compression parent that is
1661
# examined, otherwise 'ghost' would also be reported as a missing
1664
frozenset([('missing-parent',)]),
1665
index.get_missing_compression_parents())
1667
def test_add_unvalidated_index_with_present_external_references(self):
1668
index = self.two_graph_index(deltas=True)
1669
# Ugly hack to get at one of the underlying GraphIndex objects that
1670
# two_graph_index built.
1671
unvalidated = index._graph_index._indices[1]
1672
# 'parent' is an external ref of _indices[1] (unvalidated), but is
1673
# present in _indices[0].
1674
index.scan_unvalidated_index(unvalidated)
1675
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1677
def make_new_missing_parent_g_index(self, name):
1678
missing_parent = name + '-missing-parent'
1679
graph_index = self.make_g_index(name, 2,
1680
[((name + 'tip', ), ' 100 78',
1681
([(missing_parent, ), ('ghost', )], [(missing_parent, )]))])
1684
def test_add_mulitiple_unvalidated_indices_with_missing_parents(self):
1685
g_index_1 = self.make_new_missing_parent_g_index('one')
1686
g_index_2 = self.make_new_missing_parent_g_index('two')
1687
combined = CombinedGraphIndex([g_index_1, g_index_2])
1688
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1689
index.scan_unvalidated_index(g_index_1)
1690
index.scan_unvalidated_index(g_index_2)
1692
frozenset([('one-missing-parent',), ('two-missing-parent',)]),
1693
index.get_missing_compression_parents())
1695
def test_add_mulitiple_unvalidated_indices_with_mutual_dependencies(self):
1696
graph_index_a = self.make_g_index('one', 2,
1697
[(('parent-one', ), ' 100 78', ([('non-compression-parent',)], [])),
1698
(('child-of-two', ), ' 100 78',
1699
([('parent-two',)], [('parent-two',)]))])
1700
graph_index_b = self.make_g_index('two', 2,
1701
[(('parent-two', ), ' 100 78', ([('non-compression-parent',)], [])),
1702
(('child-of-one', ), ' 100 78',
1703
([('parent-one',)], [('parent-one',)]))])
1704
combined = CombinedGraphIndex([graph_index_a, graph_index_b])
1705
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1706
index.scan_unvalidated_index(graph_index_a)
1707
index.scan_unvalidated_index(graph_index_b)
1709
frozenset([]), index.get_missing_compression_parents())
1712
class TestNoParentsGraphIndexKnit(KnitTests):
1713
"""Tests for knits using _KnitGraphIndex with no parents."""
1715
def make_g_index(self, name, ref_lists=0, nodes=[]):
1716
builder = GraphIndexBuilder(ref_lists)
1717
for node, references in nodes:
1718
builder.add_node(node, references)
1719
stream = builder.finish()
1720
trans = self.get_transport()
1721
size = trans.put_file(name, stream)
1722
return GraphIndex(trans, name, size)
1724
def test_add_good_unvalidated_index(self):
1725
unvalidated = self.make_g_index('unvalidated')
1726
combined = CombinedGraphIndex([unvalidated])
1727
index = _KnitGraphIndex(combined, lambda: True, parents=False)
1728
index.scan_unvalidated_index(unvalidated)
1729
self.assertEqual(frozenset(),
1730
index.get_missing_compression_parents())
1732
def test_parents_deltas_incompatible(self):
1733
index = CombinedGraphIndex([])
1734
self.assertRaises(errors.KnitError, _KnitGraphIndex, lambda:True,
1735
index, deltas=True, parents=False)
1737
def two_graph_index(self, catch_adds=False):
1738
"""Build a two-graph index.
1740
:param deltas: If true, use underlying indices with two node-ref
1741
lists and 'parent' set to a delta-compressed against tail.
1743
# put several versions in the index.
1744
index1 = self.make_g_index('1', 0, [
1745
(('tip', ), 'N0 100'),
1747
index2 = self.make_g_index('2', 0, [
1748
(('parent', ), ' 100 78'),
1749
(('separate', ), '')])
1750
combined_index = CombinedGraphIndex([index1, index2])
1752
self.combined_index = combined_index
1753
self.caught_entries = []
1754
add_callback = self.catch_add
1757
return _KnitGraphIndex(combined_index, lambda:True, parents=False,
1758
add_callback=add_callback)
1760
def test_keys(self):
1761
index = self.two_graph_index()
1762
self.assertEqual(set([('tail',), ('tip',), ('parent',), ('separate',)]),
1765
def test_get_position(self):
1766
index = self.two_graph_index()
1767
self.assertEqual((index._graph_index._indices[0], 0, 100),
1768
index.get_position(('tip',)))
1769
self.assertEqual((index._graph_index._indices[1], 100, 78),
1770
index.get_position(('parent',)))
1772
def test_get_method(self):
1773
index = self.two_graph_index()
1774
self.assertEqual('fulltext', index.get_method(('tip',)))
1775
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1777
def test_get_options(self):
1778
index = self.two_graph_index()
1779
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1780
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1782
def test_get_parent_map(self):
1783
index = self.two_graph_index()
1784
self.assertEqual({('parent',):None},
1785
index.get_parent_map([('parent',), ('ghost',)]))
1787
def catch_add(self, entries):
1788
self.caught_entries.append(entries)
1790
def test_add_no_callback_errors(self):
1791
index = self.two_graph_index()
1792
self.assertRaises(errors.ReadOnlyError, index.add_records,
1793
[(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)])])
1795
def test_add_version_smoke(self):
1796
index = self.two_graph_index(catch_adds=True)
1797
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60), [])])
1798
self.assertEqual([[(('new', ), 'N50 60')]],
1799
self.caught_entries)
1801
def test_add_version_delta_not_delta_index(self):
1802
index = self.two_graph_index(catch_adds=True)
1803
self.assertRaises(errors.KnitCorrupt, index.add_records,
1804
[(('new',), 'no-eol,line-delta', (None, 0, 100), [])])
1805
self.assertEqual([], self.caught_entries)
1807
def test_add_version_same_dup(self):
1808
index = self.two_graph_index(catch_adds=True)
1809
# options can be spelt two different ways
1810
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1811
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
1812
# position/length are ignored (because each pack could have fulltext or
1813
# delta, and be at a different position.
1814
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
1815
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
1816
# but neither should have added data.
1817
self.assertEqual([[], [], [], []], self.caught_entries)
1819
def test_add_version_different_dup(self):
1820
index = self.two_graph_index(catch_adds=True)
1822
self.assertRaises(errors.KnitCorrupt, index.add_records,
1823
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
1824
self.assertRaises(errors.KnitCorrupt, index.add_records,
1825
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
1826
self.assertRaises(errors.KnitCorrupt, index.add_records,
1827
[(('tip',), 'fulltext', (None, 0, 100), [])])
1829
self.assertRaises(errors.KnitCorrupt, index.add_records,
1830
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
1831
self.assertEqual([], self.caught_entries)
1833
def test_add_versions(self):
1834
index = self.two_graph_index(catch_adds=True)
1836
(('new',), 'fulltext,no-eol', (None, 50, 60), []),
1837
(('new2',), 'fulltext', (None, 0, 6), []),
1839
self.assertEqual([(('new', ), 'N50 60'), (('new2', ), ' 0 6')],
1840
sorted(self.caught_entries[0]))
1841
self.assertEqual(1, len(self.caught_entries))
1843
def test_add_versions_delta_not_delta_index(self):
1844
index = self.two_graph_index(catch_adds=True)
1845
self.assertRaises(errors.KnitCorrupt, index.add_records,
1846
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1847
self.assertEqual([], self.caught_entries)
1849
def test_add_versions_parents_not_parents_index(self):
1850
index = self.two_graph_index(catch_adds=True)
1851
self.assertRaises(errors.KnitCorrupt, index.add_records,
1852
[(('new',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
1853
self.assertEqual([], self.caught_entries)
1855
def test_add_versions_random_id_accepted(self):
1856
index = self.two_graph_index(catch_adds=True)
1857
index.add_records([], random_id=True)
1859
def test_add_versions_same_dup(self):
1860
index = self.two_graph_index(catch_adds=True)
1861
# options can be spelt two different ways
1862
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1863
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
1864
# position/length are ignored (because each pack could have fulltext or
1865
# delta, and be at a different position.
1866
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
1867
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
1868
# but neither should have added data.
1869
self.assertEqual([[], [], [], []], self.caught_entries)
1871
def test_add_versions_different_dup(self):
1872
index = self.two_graph_index(catch_adds=True)
1874
self.assertRaises(errors.KnitCorrupt, index.add_records,
1875
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
1876
self.assertRaises(errors.KnitCorrupt, index.add_records,
1877
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
1878
self.assertRaises(errors.KnitCorrupt, index.add_records,
1879
[(('tip',), 'fulltext', (None, 0, 100), [])])
1881
self.assertRaises(errors.KnitCorrupt, index.add_records,
1882
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
1883
# change options in the second record
1884
self.assertRaises(errors.KnitCorrupt, index.add_records,
1885
[(('tip',), 'fulltext,no-eol', (None, 0, 100), []),
1886
(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
1887
self.assertEqual([], self.caught_entries)
1890
class TestKnitVersionedFiles(KnitTests):
1892
def assertGroupKeysForIo(self, exp_groups, keys, non_local_keys,
1893
positions, _min_buffer_size=None):
1894
kvf = self.make_test_knit()
1895
if _min_buffer_size is None:
1896
_min_buffer_size = knit._STREAM_MIN_BUFFER_SIZE
1897
self.assertEqual(exp_groups, kvf._group_keys_for_io(keys,
1898
non_local_keys, positions,
1899
_min_buffer_size=_min_buffer_size))
1901
def assertSplitByPrefix(self, expected_map, expected_prefix_order,
1903
split, prefix_order = KnitVersionedFiles._split_by_prefix(keys)
1904
self.assertEqual(expected_map, split)
1905
self.assertEqual(expected_prefix_order, prefix_order)
1907
def test__group_keys_for_io(self):
1908
ft_detail = ('fulltext', False)
1909
ld_detail = ('line-delta', False)
1917
f_a: (ft_detail, (f_a, 0, 100), None),
1918
f_b: (ld_detail, (f_b, 100, 21), f_a),
1919
f_c: (ld_detail, (f_c, 180, 15), f_b),
1920
g_a: (ft_detail, (g_a, 121, 35), None),
1921
g_b: (ld_detail, (g_b, 156, 12), g_a),
1922
g_c: (ld_detail, (g_c, 195, 13), g_a),
1924
self.assertGroupKeysForIo([([f_a], set())],
1925
[f_a], [], positions)
1926
self.assertGroupKeysForIo([([f_a], set([f_a]))],
1927
[f_a], [f_a], positions)
1928
self.assertGroupKeysForIo([([f_a, f_b], set([]))],
1929
[f_a, f_b], [], positions)
1930
self.assertGroupKeysForIo([([f_a, f_b], set([f_b]))],
1931
[f_a, f_b], [f_b], positions)
1932
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
1933
[f_a, g_a, f_b, g_b], [], positions)
1934
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
1935
[f_a, g_a, f_b, g_b], [], positions,
1936
_min_buffer_size=150)
1937
self.assertGroupKeysForIo([([f_a, f_b], set()), ([g_a, g_b], set())],
1938
[f_a, g_a, f_b, g_b], [], positions,
1939
_min_buffer_size=100)
1940
self.assertGroupKeysForIo([([f_c], set()), ([g_b], set())],
1941
[f_c, g_b], [], positions,
1942
_min_buffer_size=125)
1943
self.assertGroupKeysForIo([([g_b, f_c], set())],
1944
[g_b, f_c], [], positions,
1945
_min_buffer_size=125)
1947
def test__split_by_prefix(self):
1948
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
1949
'g': [('g', 'b'), ('g', 'a')],
1951
[('f', 'a'), ('g', 'b'),
1952
('g', 'a'), ('f', 'b')])
1954
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
1955
'g': [('g', 'b'), ('g', 'a')],
1957
[('f', 'a'), ('f', 'b'),
1958
('g', 'b'), ('g', 'a')])
1960
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
1961
'g': [('g', 'b'), ('g', 'a')],
1963
[('f', 'a'), ('f', 'b'),
1964
('g', 'b'), ('g', 'a')])
1966
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
1967
'g': [('g', 'b'), ('g', 'a')],
1968
'': [('a',), ('b',)]
1970
[('f', 'a'), ('g', 'b'),
1972
('g', 'a'), ('f', 'b')])
1975
class TestStacking(KnitTests):
1977
def get_basis_and_test_knit(self):
1978
basis = self.make_test_knit(name='basis')
1979
basis = RecordingVersionedFilesDecorator(basis)
1980
test = self.make_test_knit(name='test')
1981
test.add_fallback_versioned_files(basis)
1984
def test_add_fallback_versioned_files(self):
1985
basis = self.make_test_knit(name='basis')
1986
test = self.make_test_knit(name='test')
1987
# It must not error; other tests test that the fallback is referred to
1988
# when accessing data.
1989
test.add_fallback_versioned_files(basis)
1991
def test_add_lines(self):
1992
# lines added to the test are not added to the basis
1993
basis, test = self.get_basis_and_test_knit()
1995
key_basis = ('bar',)
1996
key_cross_border = ('quux',)
1997
key_delta = ('zaphod',)
1998
test.add_lines(key, (), ['foo\n'])
1999
self.assertEqual({}, basis.get_parent_map([key]))
2000
# lines added to the test that reference across the stack do a
2002
basis.add_lines(key_basis, (), ['foo\n'])
2004
test.add_lines(key_cross_border, (key_basis,), ['foo\n'])
2005
self.assertEqual('fulltext', test._index.get_method(key_cross_border))
2006
# we don't even need to look at the basis to see that this should be
2007
# stored as a fulltext
2008
self.assertEqual([], basis.calls)
2009
# Subsequent adds do delta.
2011
test.add_lines(key_delta, (key_cross_border,), ['foo\n'])
2012
self.assertEqual('line-delta', test._index.get_method(key_delta))
2013
self.assertEqual([], basis.calls)
2015
def test_annotate(self):
2016
# annotations from the test knit are answered without asking the basis
2017
basis, test = self.get_basis_and_test_knit()
2019
key_basis = ('bar',)
2020
key_missing = ('missing',)
2021
test.add_lines(key, (), ['foo\n'])
2022
details = test.annotate(key)
2023
self.assertEqual([(key, 'foo\n')], details)
2024
self.assertEqual([], basis.calls)
2025
# But texts that are not in the test knit are looked for in the basis
2027
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2029
details = test.annotate(key_basis)
2030
self.assertEqual([(key_basis, 'foo\n'), (key_basis, 'bar\n')], details)
2031
# Not optimised to date:
2032
# self.assertEqual([("annotate", key_basis)], basis.calls)
2033
self.assertEqual([('get_parent_map', set([key_basis])),
2034
('get_parent_map', set([key_basis])),
2035
('get_parent_map', set([key_basis])),
2036
('get_record_stream', [key_basis], 'unordered', True)],
2039
def test_check(self):
2040
# At the moment checking a stacked knit does implicitly check the
2042
basis, test = self.get_basis_and_test_knit()
2045
def test_get_parent_map(self):
2046
# parents in the test knit are answered without asking the basis
2047
basis, test = self.get_basis_and_test_knit()
2049
key_basis = ('bar',)
2050
key_missing = ('missing',)
2051
test.add_lines(key, (), [])
2052
parent_map = test.get_parent_map([key])
2053
self.assertEqual({key: ()}, parent_map)
2054
self.assertEqual([], basis.calls)
2055
# But parents that are not in the test knit are looked for in the basis
2056
basis.add_lines(key_basis, (), [])
2058
parent_map = test.get_parent_map([key, key_basis, key_missing])
2059
self.assertEqual({key: (),
2060
key_basis: ()}, parent_map)
2061
self.assertEqual([("get_parent_map", set([key_basis, key_missing]))],
2064
def test_get_record_stream_unordered_fulltexts(self):
2065
# records from the test knit are answered without asking the basis:
2066
basis, test = self.get_basis_and_test_knit()
2068
key_basis = ('bar',)
2069
key_missing = ('missing',)
2070
test.add_lines(key, (), ['foo\n'])
2071
records = list(test.get_record_stream([key], 'unordered', True))
2072
self.assertEqual(1, len(records))
2073
self.assertEqual([], basis.calls)
2074
# Missing (from test knit) objects are retrieved from the basis:
2075
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2077
records = list(test.get_record_stream([key_basis, key_missing],
2079
self.assertEqual(2, len(records))
2080
calls = list(basis.calls)
2081
for record in records:
2082
self.assertSubset([record.key], (key_basis, key_missing))
2083
if record.key == key_missing:
2084
self.assertIsInstance(record, AbsentContentFactory)
2086
reference = list(basis.get_record_stream([key_basis],
2087
'unordered', True))[0]
2088
self.assertEqual(reference.key, record.key)
2089
self.assertEqual(reference.sha1, record.sha1)
2090
self.assertEqual(reference.storage_kind, record.storage_kind)
2091
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2092
record.get_bytes_as(record.storage_kind))
2093
self.assertEqual(reference.get_bytes_as('fulltext'),
2094
record.get_bytes_as('fulltext'))
2095
# It's not strictly minimal, but it seems reasonable for now for it to
2096
# ask which fallbacks have which parents.
2098
("get_parent_map", set([key_basis, key_missing])),
2099
("get_record_stream", [key_basis], 'unordered', True)],
2102
def test_get_record_stream_ordered_fulltexts(self):
2103
# ordering is preserved down into the fallback store.
2104
basis, test = self.get_basis_and_test_knit()
2106
key_basis = ('bar',)
2107
key_basis_2 = ('quux',)
2108
key_missing = ('missing',)
2109
test.add_lines(key, (key_basis,), ['foo\n'])
2110
# Missing (from test knit) objects are retrieved from the basis:
2111
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2112
basis.add_lines(key_basis_2, (), ['quux\n'])
2114
# ask for in non-topological order
2115
records = list(test.get_record_stream(
2116
[key, key_basis, key_missing, key_basis_2], 'topological', True))
2117
self.assertEqual(4, len(records))
2119
for record in records:
2120
self.assertSubset([record.key],
2121
(key_basis, key_missing, key_basis_2, key))
2122
if record.key == key_missing:
2123
self.assertIsInstance(record, AbsentContentFactory)
2125
results.append((record.key, record.sha1, record.storage_kind,
2126
record.get_bytes_as('fulltext')))
2127
calls = list(basis.calls)
2128
order = [record[0] for record in results]
2129
self.assertEqual([key_basis_2, key_basis, key], order)
2130
for result in results:
2131
if result[0] == key:
2135
record = source.get_record_stream([result[0]], 'unordered',
2137
self.assertEqual(record.key, result[0])
2138
self.assertEqual(record.sha1, result[1])
2139
# We used to check that the storage kind matched, but actually it
2140
# depends on whether it was sourced from the basis, or in a single
2141
# group, because asking for full texts returns proxy objects to a
2142
# _ContentMapGenerator object; so checking the kind is unneeded.
2143
self.assertEqual(record.get_bytes_as('fulltext'), result[3])
2144
# It's not strictly minimal, but it seems reasonable for now for it to
2145
# ask which fallbacks have which parents.
2147
("get_parent_map", set([key_basis, key_basis_2, key_missing])),
2148
# unordered is asked for by the underlying worker as it still
2149
# buffers everything while answering - which is a problem!
2150
("get_record_stream", [key_basis_2, key_basis], 'unordered', True)],
2153
def test_get_record_stream_unordered_deltas(self):
2154
# records from the test knit are answered without asking the basis:
2155
basis, test = self.get_basis_and_test_knit()
2157
key_basis = ('bar',)
2158
key_missing = ('missing',)
2159
test.add_lines(key, (), ['foo\n'])
2160
records = list(test.get_record_stream([key], 'unordered', False))
2161
self.assertEqual(1, len(records))
2162
self.assertEqual([], basis.calls)
2163
# Missing (from test knit) objects are retrieved from the basis:
2164
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2166
records = list(test.get_record_stream([key_basis, key_missing],
2167
'unordered', False))
2168
self.assertEqual(2, len(records))
2169
calls = list(basis.calls)
2170
for record in records:
2171
self.assertSubset([record.key], (key_basis, key_missing))
2172
if record.key == key_missing:
2173
self.assertIsInstance(record, AbsentContentFactory)
2175
reference = list(basis.get_record_stream([key_basis],
2176
'unordered', False))[0]
2177
self.assertEqual(reference.key, record.key)
2178
self.assertEqual(reference.sha1, record.sha1)
2179
self.assertEqual(reference.storage_kind, record.storage_kind)
2180
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2181
record.get_bytes_as(record.storage_kind))
2182
# It's not strictly minimal, but it seems reasonable for now for it to
2183
# ask which fallbacks have which parents.
2185
("get_parent_map", set([key_basis, key_missing])),
2186
("get_record_stream", [key_basis], 'unordered', False)],
2189
def test_get_record_stream_ordered_deltas(self):
2190
# ordering is preserved down into the fallback store.
2191
basis, test = self.get_basis_and_test_knit()
2193
key_basis = ('bar',)
2194
key_basis_2 = ('quux',)
2195
key_missing = ('missing',)
2196
test.add_lines(key, (key_basis,), ['foo\n'])
2197
# Missing (from test knit) objects are retrieved from the basis:
2198
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2199
basis.add_lines(key_basis_2, (), ['quux\n'])
2201
# ask for in non-topological order
2202
records = list(test.get_record_stream(
2203
[key, key_basis, key_missing, key_basis_2], 'topological', False))
2204
self.assertEqual(4, len(records))
2206
for record in records:
2207
self.assertSubset([record.key],
2208
(key_basis, key_missing, key_basis_2, key))
2209
if record.key == key_missing:
2210
self.assertIsInstance(record, AbsentContentFactory)
2212
results.append((record.key, record.sha1, record.storage_kind,
2213
record.get_bytes_as(record.storage_kind)))
2214
calls = list(basis.calls)
2215
order = [record[0] for record in results]
2216
self.assertEqual([key_basis_2, key_basis, key], order)
2217
for result in results:
2218
if result[0] == key:
2222
record = source.get_record_stream([result[0]], 'unordered',
2224
self.assertEqual(record.key, result[0])
2225
self.assertEqual(record.sha1, result[1])
2226
self.assertEqual(record.storage_kind, result[2])
2227
self.assertEqual(record.get_bytes_as(record.storage_kind), result[3])
2228
# It's not strictly minimal, but it seems reasonable for now for it to
2229
# ask which fallbacks have which parents.
2231
("get_parent_map", set([key_basis, key_basis_2, key_missing])),
2232
("get_record_stream", [key_basis_2, key_basis], 'topological', False)],
2235
def test_get_sha1s(self):
2236
# sha1's in the test knit are answered without asking the basis
2237
basis, test = self.get_basis_and_test_knit()
2239
key_basis = ('bar',)
2240
key_missing = ('missing',)
2241
test.add_lines(key, (), ['foo\n'])
2242
key_sha1sum = osutils.sha('foo\n').hexdigest()
2243
sha1s = test.get_sha1s([key])
2244
self.assertEqual({key: key_sha1sum}, sha1s)
2245
self.assertEqual([], basis.calls)
2246
# But texts that are not in the test knit are looked for in the basis
2247
# directly (rather than via text reconstruction) so that remote servers
2248
# etc don't have to answer with full content.
2249
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2250
basis_sha1sum = osutils.sha('foo\nbar\n').hexdigest()
2252
sha1s = test.get_sha1s([key, key_missing, key_basis])
2253
self.assertEqual({key: key_sha1sum,
2254
key_basis: basis_sha1sum}, sha1s)
2255
self.assertEqual([("get_sha1s", set([key_basis, key_missing]))],
2258
def test_insert_record_stream(self):
2259
# records are inserted as normal; insert_record_stream builds on
2260
# add_lines, so a smoke test should be all that's needed:
2262
key_basis = ('bar',)
2263
key_delta = ('zaphod',)
2264
basis, test = self.get_basis_and_test_knit()
2265
source = self.make_test_knit(name='source')
2266
basis.add_lines(key_basis, (), ['foo\n'])
2268
source.add_lines(key_basis, (), ['foo\n'])
2269
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2270
stream = source.get_record_stream([key_delta], 'unordered', False)
2271
test.insert_record_stream(stream)
2272
# XXX: this does somewhat too many calls in making sure of whether it
2273
# has to recreate the full text.
2274
self.assertEqual([("get_parent_map", set([key_basis])),
2275
('get_parent_map', set([key_basis])),
2276
('get_record_stream', [key_basis], 'unordered', True)],
2278
self.assertEqual({key_delta:(key_basis,)},
2279
test.get_parent_map([key_delta]))
2280
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2281
'unordered', True).next().get_bytes_as('fulltext'))
2283
def test_iter_lines_added_or_present_in_keys(self):
2284
# Lines from the basis are returned, and lines for a given key are only
2288
# all sources are asked for keys:
2289
basis, test = self.get_basis_and_test_knit()
2290
basis.add_lines(key1, (), ["foo"])
2292
lines = list(test.iter_lines_added_or_present_in_keys([key1]))
2293
self.assertEqual([("foo\n", key1)], lines)
2294
self.assertEqual([("iter_lines_added_or_present_in_keys", set([key1]))],
2296
# keys in both are not duplicated:
2297
test.add_lines(key2, (), ["bar\n"])
2298
basis.add_lines(key2, (), ["bar\n"])
2300
lines = list(test.iter_lines_added_or_present_in_keys([key2]))
2301
self.assertEqual([("bar\n", key2)], lines)
2302
self.assertEqual([], basis.calls)
2304
def test_keys(self):
2307
# all sources are asked for keys:
2308
basis, test = self.get_basis_and_test_knit()
2310
self.assertEqual(set(), set(keys))
2311
self.assertEqual([("keys",)], basis.calls)
2312
# keys from a basis are returned:
2313
basis.add_lines(key1, (), [])
2316
self.assertEqual(set([key1]), set(keys))
2317
self.assertEqual([("keys",)], basis.calls)
2318
# keys in both are not duplicated:
2319
test.add_lines(key2, (), [])
2320
basis.add_lines(key2, (), [])
2323
self.assertEqual(2, len(keys))
2324
self.assertEqual(set([key1, key2]), set(keys))
2325
self.assertEqual([("keys",)], basis.calls)
2327
def test_add_mpdiffs(self):
2328
# records are inserted as normal; add_mpdiff builds on
2329
# add_lines, so a smoke test should be all that's needed:
2331
key_basis = ('bar',)
2332
key_delta = ('zaphod',)
2333
basis, test = self.get_basis_and_test_knit()
2334
source = self.make_test_knit(name='source')
2335
basis.add_lines(key_basis, (), ['foo\n'])
2337
source.add_lines(key_basis, (), ['foo\n'])
2338
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2339
diffs = source.make_mpdiffs([key_delta])
2340
test.add_mpdiffs([(key_delta, (key_basis,),
2341
source.get_sha1s([key_delta])[key_delta], diffs[0])])
2342
self.assertEqual([("get_parent_map", set([key_basis])),
2343
('get_record_stream', [key_basis], 'unordered', True),],
2345
self.assertEqual({key_delta:(key_basis,)},
2346
test.get_parent_map([key_delta]))
2347
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2348
'unordered', True).next().get_bytes_as('fulltext'))
2350
def test_make_mpdiffs(self):
2351
# Generating an mpdiff across a stacking boundary should detect parent
2355
key_right = ('zaphod',)
2356
basis, test = self.get_basis_and_test_knit()
2357
basis.add_lines(key_left, (), ['bar\n'])
2358
basis.add_lines(key_right, (), ['zaphod\n'])
2360
test.add_lines(key, (key_left, key_right),
2361
['bar\n', 'foo\n', 'zaphod\n'])
2362
diffs = test.make_mpdiffs([key])
2364
multiparent.MultiParent([multiparent.ParentText(0, 0, 0, 1),
2365
multiparent.NewText(['foo\n']),
2366
multiparent.ParentText(1, 0, 2, 1)])],
2368
self.assertEqual(3, len(basis.calls))
2370
("get_parent_map", set([key_left, key_right])),
2371
("get_parent_map", set([key_left, key_right])),
2374
last_call = basis.calls[-1]
2375
self.assertEqual('get_record_stream', last_call[0])
2376
self.assertEqual(set([key_left, key_right]), set(last_call[1]))
2377
self.assertEqual('unordered', last_call[2])
2378
self.assertEqual(True, last_call[3])
2381
class TestNetworkBehaviour(KnitTests):
2382
"""Tests for getting data out of/into knits over the network."""
2384
def test_include_delta_closure_generates_a_knit_delta_closure(self):
2385
vf = self.make_test_knit(name='test')
2386
# put in three texts, giving ft, delta, delta
2387
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2388
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2389
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2390
# But heuristics could interfere, so check what happened:
2391
self.assertEqual(['knit-ft-gz', 'knit-delta-gz', 'knit-delta-gz'],
2392
[record.storage_kind for record in
2393
vf.get_record_stream([('base',), ('d1',), ('d2',)],
2394
'topological', False)])
2395
# generate a stream of just the deltas include_delta_closure=True,
2396
# serialise to the network, and check that we get a delta closure on the wire.
2397
stream = vf.get_record_stream([('d1',), ('d2',)], 'topological', True)
2398
netb = [record.get_bytes_as(record.storage_kind) for record in stream]
2399
# The first bytes should be a memo from _ContentMapGenerator, and the
2400
# second bytes should be empty (because its a API proxy not something
2401
# for wire serialisation.
2402
self.assertEqual('', netb[1])
2404
kind, line_end = network_bytes_to_kind_and_offset(bytes)
2405
self.assertEqual('knit-delta-closure', kind)
2408
class TestContentMapGenerator(KnitTests):
2409
"""Tests for ContentMapGenerator"""
2411
def test_get_record_stream_gives_records(self):
2412
vf = self.make_test_knit(name='test')
2413
# put in three texts, giving ft, delta, delta
2414
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2415
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2416
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2417
keys = [('d1',), ('d2',)]
2418
generator = _VFContentMapGenerator(vf, keys,
2419
global_map=vf.get_parent_map(keys))
2420
for record in generator.get_record_stream():
2421
if record.key == ('d1',):
2422
self.assertEqual('d1\n', record.get_bytes_as('fulltext'))
2424
self.assertEqual('d2\n', record.get_bytes_as('fulltext'))
2426
def test_get_record_stream_kinds_are_raw(self):
2427
vf = self.make_test_knit(name='test')
2428
# put in three texts, giving ft, delta, delta
2429
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2430
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2431
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2432
keys = [('base',), ('d1',), ('d2',)]
2433
generator = _VFContentMapGenerator(vf, keys,
2434
global_map=vf.get_parent_map(keys))
2435
kinds = {('base',): 'knit-delta-closure',
2436
('d1',): 'knit-delta-closure-ref',
2437
('d2',): 'knit-delta-closure-ref',
2439
for record in generator.get_record_stream():
2440
self.assertEqual(kinds[record.key], record.storage_kind)
added
removed
bzrlib/tests/test_knit.py
