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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 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')
371
self.addCleanup(tree.branch.repository.unlock)
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 three 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()
384
# forget about the new pack
387
vf = tree.branch.repository.revisions
389
# Set up a reload() function that switches to using the new pack file
390
new_index = new_pack.revision_index
391
access_tuple = new_pack.access_tuple()
392
reload_counter = [0, 0, 0]
394
reload_counter[0] += 1
395
if reload_counter[1] > 0:
396
# We already reloaded, nothing more to do
397
reload_counter[2] += 1
399
reload_counter[1] += 1
400
vf._index._graph_index._indices[:] = [new_index]
401
vf._access._indices.clear()
402
vf._access._indices[new_index] = access_tuple
404
# Delete one of the pack files so the data will need to be reloaded. We
405
# will delete the file with 'rev-2' in it
406
trans, name = orig_packs[1].access_tuple()
408
# We don't have the index trigger reloading because we want to test
409
# that we reload when the .pack disappears
410
vf._access._reload_func = reload
411
return vf, reload_counter
413
def make_reload_func(self, return_val=True):
416
reload_called[0] += 1
418
return reload_called, reload
420
def make_retry_exception(self):
421
# We raise a real exception so that sys.exc_info() is properly
424
raise _TestException('foobar')
425
except _TestException, e:
426
retry_exc = errors.RetryWithNewPacks(None, reload_occurred=False,
427
exc_info=sys.exc_info())
430
def test_read_from_several_packs(self):
431
access, writer = self._get_access()
433
memos.extend(access.add_raw_records([('key', 10)], '1234567890'))
435
access, writer = self._get_access('pack2', 'FOOBAR')
436
memos.extend(access.add_raw_records([('key', 5)], '12345'))
438
access, writer = self._get_access('pack3', 'BAZ')
439
memos.extend(access.add_raw_records([('key', 5)], 'alpha'))
441
transport = self.get_transport()
442
access = _DirectPackAccess({"FOO":(transport, 'packfile'),
443
"FOOBAR":(transport, 'pack2'),
444
"BAZ":(transport, 'pack3')})
445
self.assertEqual(['1234567890', '12345', 'alpha'],
446
list(access.get_raw_records(memos)))
447
self.assertEqual(['1234567890'],
448
list(access.get_raw_records(memos[0:1])))
449
self.assertEqual(['12345'],
450
list(access.get_raw_records(memos[1:2])))
451
self.assertEqual(['alpha'],
452
list(access.get_raw_records(memos[2:3])))
453
self.assertEqual(['1234567890', 'alpha'],
454
list(access.get_raw_records(memos[0:1] + memos[2:3])))
456
def test_set_writer(self):
457
"""The writer should be settable post construction."""
458
access = _DirectPackAccess({})
459
transport = self.get_transport()
460
packname = 'packfile'
462
def write_data(bytes):
463
transport.append_bytes(packname, bytes)
464
writer = pack.ContainerWriter(write_data)
466
access.set_writer(writer, index, (transport, packname))
467
memos = access.add_raw_records([('key', 10)], '1234567890')
469
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
471
def test_missing_index_raises_retry(self):
472
memos = self.make_pack_file()
473
transport = self.get_transport()
474
reload_called, reload_func = self.make_reload_func()
475
# Note that the index key has changed from 'foo' to 'bar'
476
access = _DirectPackAccess({'bar':(transport, 'packname')},
477
reload_func=reload_func)
478
e = self.assertListRaises(errors.RetryWithNewPacks,
479
access.get_raw_records, memos)
480
# Because a key was passed in which does not match our index list, we
481
# assume that the listing was already reloaded
482
self.assertTrue(e.reload_occurred)
483
self.assertIsInstance(e.exc_info, tuple)
484
self.assertIs(e.exc_info[0], KeyError)
485
self.assertIsInstance(e.exc_info[1], KeyError)
487
def test_missing_index_raises_key_error_with_no_reload(self):
488
memos = self.make_pack_file()
489
transport = self.get_transport()
490
# Note that the index key has changed from 'foo' to 'bar'
491
access = _DirectPackAccess({'bar':(transport, 'packname')})
492
e = self.assertListRaises(KeyError, access.get_raw_records, memos)
494
def test_missing_file_raises_retry(self):
495
memos = self.make_pack_file()
496
transport = self.get_transport()
497
reload_called, reload_func = self.make_reload_func()
498
# Note that the 'filename' has been changed to 'different-packname'
499
access = _DirectPackAccess({'foo':(transport, 'different-packname')},
500
reload_func=reload_func)
501
e = self.assertListRaises(errors.RetryWithNewPacks,
502
access.get_raw_records, memos)
503
# The file has gone missing, so we assume we need to reload
504
self.assertFalse(e.reload_occurred)
505
self.assertIsInstance(e.exc_info, tuple)
506
self.assertIs(e.exc_info[0], errors.NoSuchFile)
507
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
508
self.assertEqual('different-packname', e.exc_info[1].path)
510
def test_missing_file_raises_no_such_file_with_no_reload(self):
511
memos = self.make_pack_file()
512
transport = self.get_transport()
513
# Note that the 'filename' has been changed to 'different-packname'
514
access = _DirectPackAccess({'foo':(transport, 'different-packname')})
515
e = self.assertListRaises(errors.NoSuchFile,
516
access.get_raw_records, memos)
518
def test_failing_readv_raises_retry(self):
519
memos = self.make_pack_file()
520
transport = self.get_transport()
521
failing_transport = MockReadvFailingTransport(
522
[transport.get_bytes('packname')])
523
reload_called, reload_func = self.make_reload_func()
524
access = _DirectPackAccess({'foo':(failing_transport, 'packname')},
525
reload_func=reload_func)
526
# Asking for a single record will not trigger the Mock failure
527
self.assertEqual(['1234567890'],
528
list(access.get_raw_records(memos[:1])))
529
self.assertEqual(['12345'],
530
list(access.get_raw_records(memos[1:2])))
531
# A multiple offset readv() will fail mid-way through
532
e = self.assertListRaises(errors.RetryWithNewPacks,
533
access.get_raw_records, memos)
534
# The file has gone missing, so we assume we need to reload
535
self.assertFalse(e.reload_occurred)
536
self.assertIsInstance(e.exc_info, tuple)
537
self.assertIs(e.exc_info[0], errors.NoSuchFile)
538
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
539
self.assertEqual('packname', e.exc_info[1].path)
541
def test_failing_readv_raises_no_such_file_with_no_reload(self):
542
memos = self.make_pack_file()
543
transport = self.get_transport()
544
failing_transport = MockReadvFailingTransport(
545
[transport.get_bytes('packname')])
546
reload_called, reload_func = self.make_reload_func()
547
access = _DirectPackAccess({'foo':(failing_transport, 'packname')})
548
# Asking for a single record will not trigger the Mock failure
549
self.assertEqual(['1234567890'],
550
list(access.get_raw_records(memos[:1])))
551
self.assertEqual(['12345'],
552
list(access.get_raw_records(memos[1:2])))
553
# A multiple offset readv() will fail mid-way through
554
e = self.assertListRaises(errors.NoSuchFile,
555
access.get_raw_records, memos)
557
def test_reload_or_raise_no_reload(self):
558
access = _DirectPackAccess({}, reload_func=None)
559
retry_exc = self.make_retry_exception()
560
# Without a reload_func, we will just re-raise the original exception
561
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
563
def test_reload_or_raise_reload_changed(self):
564
reload_called, reload_func = self.make_reload_func(return_val=True)
565
access = _DirectPackAccess({}, reload_func=reload_func)
566
retry_exc = self.make_retry_exception()
567
access.reload_or_raise(retry_exc)
568
self.assertEqual([1], reload_called)
569
retry_exc.reload_occurred=True
570
access.reload_or_raise(retry_exc)
571
self.assertEqual([2], reload_called)
573
def test_reload_or_raise_reload_no_change(self):
574
reload_called, reload_func = self.make_reload_func(return_val=False)
575
access = _DirectPackAccess({}, reload_func=reload_func)
576
retry_exc = self.make_retry_exception()
577
# If reload_occurred is False, then we consider it an error to have
578
# reload_func() return False (no changes).
579
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
580
self.assertEqual([1], reload_called)
581
retry_exc.reload_occurred=True
582
# If reload_occurred is True, then we assume nothing changed because
583
# it had changed earlier, but didn't change again
584
access.reload_or_raise(retry_exc)
585
self.assertEqual([2], reload_called)
587
def test_annotate_retries(self):
588
vf, reload_counter = self.make_vf_for_retrying()
589
# It is a little bit bogus to annotate the Revision VF, but it works,
590
# as we have ancestry stored there
592
reload_lines = vf.annotate(key)
593
self.assertEqual([1, 1, 0], reload_counter)
594
plain_lines = vf.annotate(key)
595
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
596
if reload_lines != plain_lines:
597
self.fail('Annotation was not identical with reloading.')
598
# Now delete the packs-in-use, which should trigger another reload, but
599
# this time we just raise an exception because we can't recover
600
for trans, name in vf._access._indices.itervalues():
602
self.assertRaises(errors.NoSuchFile, vf.annotate, key)
603
self.assertEqual([2, 1, 1], reload_counter)
605
def test__get_record_map_retries(self):
606
vf, reload_counter = self.make_vf_for_retrying()
607
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
608
records = vf._get_record_map(keys)
609
self.assertEqual(keys, sorted(records.keys()))
610
self.assertEqual([1, 1, 0], reload_counter)
611
# Now delete the packs-in-use, which should trigger another reload, but
612
# this time we just raise an exception because we can't recover
613
for trans, name in vf._access._indices.itervalues():
615
self.assertRaises(errors.NoSuchFile, vf._get_record_map, keys)
616
self.assertEqual([2, 1, 1], reload_counter)
618
def test_get_record_stream_retries(self):
619
vf, reload_counter = self.make_vf_for_retrying()
620
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
621
record_stream = vf.get_record_stream(keys, 'topological', False)
622
record = record_stream.next()
623
self.assertEqual(('rev-1',), record.key)
624
self.assertEqual([0, 0, 0], reload_counter)
625
record = record_stream.next()
626
self.assertEqual(('rev-2',), record.key)
627
self.assertEqual([1, 1, 0], reload_counter)
628
record = record_stream.next()
629
self.assertEqual(('rev-3',), record.key)
630
self.assertEqual([1, 1, 0], reload_counter)
631
# Now delete all pack files, and see that we raise the right error
632
for trans, name in vf._access._indices.itervalues():
634
self.assertListRaises(errors.NoSuchFile,
635
vf.get_record_stream, keys, 'topological', False)
637
def test_iter_lines_added_or_present_in_keys_retries(self):
638
vf, reload_counter = self.make_vf_for_retrying()
639
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
640
# Unfortunately, iter_lines_added_or_present_in_keys iterates the
641
# result in random order (determined by the iteration order from a
642
# set()), so we don't have any solid way to trigger whether data is
643
# read before or after. However we tried to delete the middle node to
644
# exercise the code well.
645
# What we care about is that all lines are always yielded, but not
648
reload_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
649
self.assertEqual([1, 1, 0], reload_counter)
650
# Now do it again, to make sure the result is equivalent
651
plain_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
652
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
653
self.assertEqual(plain_lines, reload_lines)
654
self.assertEqual(21, len(plain_lines))
655
# Now delete all pack files, and see that we raise the right error
656
for trans, name in vf._access._indices.itervalues():
658
self.assertListRaises(errors.NoSuchFile,
659
vf.iter_lines_added_or_present_in_keys, keys)
660
self.assertEqual([2, 1, 1], reload_counter)
662
def test_get_record_stream_yields_disk_sorted_order(self):
663
# if we get 'unordered' pick a semi-optimal order for reading. The
664
# order should be grouped by pack file, and then by position in file
665
repo = self.make_repository('test', format='pack-0.92')
667
self.addCleanup(repo.unlock)
668
repo.start_write_group()
670
vf.add_lines(('f-id', 'rev-5'), [('f-id', 'rev-4')], ['lines\n'])
671
vf.add_lines(('f-id', 'rev-1'), [], ['lines\n'])
672
vf.add_lines(('f-id', 'rev-2'), [('f-id', 'rev-1')], ['lines\n'])
673
repo.commit_write_group()
674
# We inserted them as rev-5, rev-1, rev-2, we should get them back in
676
stream = vf.get_record_stream([('f-id', 'rev-1'), ('f-id', 'rev-5'),
677
('f-id', 'rev-2')], 'unordered', False)
678
keys = [r.key for r in stream]
679
self.assertEqual([('f-id', 'rev-5'), ('f-id', 'rev-1'),
680
('f-id', 'rev-2')], keys)
681
repo.start_write_group()
682
vf.add_lines(('f-id', 'rev-4'), [('f-id', 'rev-3')], ['lines\n'])
683
vf.add_lines(('f-id', 'rev-3'), [('f-id', 'rev-2')], ['lines\n'])
684
vf.add_lines(('f-id', 'rev-6'), [('f-id', 'rev-5')], ['lines\n'])
685
repo.commit_write_group()
686
# Request in random order, to make sure the output order isn't based on
688
request_keys = set(('f-id', 'rev-%d' % i) for i in range(1, 7))
689
stream = vf.get_record_stream(request_keys, 'unordered', False)
690
keys = [r.key for r in stream]
691
# We want to get the keys back in disk order, but it doesn't matter
692
# which pack we read from first. So this can come back in 2 orders
693
alt1 = [('f-id', 'rev-%d' % i) for i in [4, 3, 6, 5, 1, 2]]
694
alt2 = [('f-id', 'rev-%d' % i) for i in [5, 1, 2, 4, 3, 6]]
695
if keys != alt1 and keys != alt2:
696
self.fail('Returned key order did not match either expected order.'
697
' expected %s or %s, not %s'
698
% (alt1, alt2, keys))
701
class LowLevelKnitDataTests(TestCase):
703
def create_gz_content(self, text):
705
gz_file = gzip.GzipFile(mode='wb', fileobj=sio)
708
return sio.getvalue()
710
def make_multiple_records(self):
711
"""Create the content for multiple records."""
712
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
714
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
719
record_1 = (0, len(gz_txt), sha1sum)
720
total_txt.append(gz_txt)
721
sha1sum = osutils.sha('baz\n').hexdigest()
722
gz_txt = self.create_gz_content('version rev-id-2 1 %s\n'
726
record_2 = (record_1[1], len(gz_txt), sha1sum)
727
total_txt.append(gz_txt)
728
return total_txt, record_1, record_2
730
def test_valid_knit_data(self):
731
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
732
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
737
transport = MockTransport([gz_txt])
738
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
739
knit = KnitVersionedFiles(None, access)
740
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
742
contents = list(knit._read_records_iter(records))
743
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'],
744
'4e48e2c9a3d2ca8a708cb0cc545700544efb5021')], contents)
746
raw_contents = list(knit._read_records_iter_raw(records))
747
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
749
def test_multiple_records_valid(self):
750
total_txt, record_1, record_2 = self.make_multiple_records()
751
transport = MockTransport([''.join(total_txt)])
752
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
753
knit = KnitVersionedFiles(None, access)
754
records = [(('rev-id-1',), (('rev-id-1',), record_1[0], record_1[1])),
755
(('rev-id-2',), (('rev-id-2',), record_2[0], record_2[1]))]
757
contents = list(knit._read_records_iter(records))
758
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'], record_1[2]),
759
(('rev-id-2',), ['baz\n'], record_2[2])],
762
raw_contents = list(knit._read_records_iter_raw(records))
763
self.assertEqual([(('rev-id-1',), total_txt[0], record_1[2]),
764
(('rev-id-2',), total_txt[1], record_2[2])],
767
def test_not_enough_lines(self):
768
sha1sum = osutils.sha('foo\n').hexdigest()
769
# record says 2 lines data says 1
770
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
774
transport = MockTransport([gz_txt])
775
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
776
knit = KnitVersionedFiles(None, access)
777
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
778
self.assertRaises(errors.KnitCorrupt, list,
779
knit._read_records_iter(records))
781
# read_records_iter_raw won't detect that sort of mismatch/corruption
782
raw_contents = list(knit._read_records_iter_raw(records))
783
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
785
def test_too_many_lines(self):
786
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
787
# record says 1 lines data says 2
788
gz_txt = self.create_gz_content('version rev-id-1 1 %s\n'
793
transport = MockTransport([gz_txt])
794
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
795
knit = KnitVersionedFiles(None, access)
796
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
797
self.assertRaises(errors.KnitCorrupt, list,
798
knit._read_records_iter(records))
800
# read_records_iter_raw won't detect that sort of mismatch/corruption
801
raw_contents = list(knit._read_records_iter_raw(records))
802
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
804
def test_mismatched_version_id(self):
805
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
806
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
811
transport = MockTransport([gz_txt])
812
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
813
knit = KnitVersionedFiles(None, access)
814
# We are asking for rev-id-2, but the data is rev-id-1
815
records = [(('rev-id-2',), (('rev-id-2',), 0, len(gz_txt)))]
816
self.assertRaises(errors.KnitCorrupt, list,
817
knit._read_records_iter(records))
819
# read_records_iter_raw detects mismatches in the header
820
self.assertRaises(errors.KnitCorrupt, list,
821
knit._read_records_iter_raw(records))
823
def test_uncompressed_data(self):
824
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
825
txt = ('version rev-id-1 2 %s\n'
830
transport = MockTransport([txt])
831
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
832
knit = KnitVersionedFiles(None, access)
833
records = [(('rev-id-1',), (('rev-id-1',), 0, len(txt)))]
835
# We don't have valid gzip data ==> corrupt
836
self.assertRaises(errors.KnitCorrupt, list,
837
knit._read_records_iter(records))
839
# read_records_iter_raw will notice the bad data
840
self.assertRaises(errors.KnitCorrupt, list,
841
knit._read_records_iter_raw(records))
843
def test_corrupted_data(self):
844
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
845
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
850
# Change 2 bytes in the middle to \xff
851
gz_txt = gz_txt[:10] + '\xff\xff' + gz_txt[12:]
852
transport = MockTransport([gz_txt])
853
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
854
knit = KnitVersionedFiles(None, access)
855
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
856
self.assertRaises(errors.KnitCorrupt, list,
857
knit._read_records_iter(records))
858
# read_records_iter_raw will barf on bad gz data
859
self.assertRaises(errors.KnitCorrupt, list,
860
knit._read_records_iter_raw(records))
863
class LowLevelKnitIndexTests(TestCase):
865
def get_knit_index(self, transport, name, mode):
866
mapper = ConstantMapper(name)
867
orig = knit._load_data
869
knit._load_data = orig
870
self.addCleanup(reset)
871
from bzrlib._knit_load_data_py import _load_data_py
872
knit._load_data = _load_data_py
873
allow_writes = lambda: 'w' in mode
874
return _KndxIndex(transport, mapper, lambda:None, allow_writes, lambda:True)
876
def test_create_file(self):
877
transport = MockTransport()
878
index = self.get_knit_index(transport, "filename", "w")
880
call = transport.calls.pop(0)
881
# call[1][1] is a StringIO - we can't test it by simple equality.
882
self.assertEqual('put_file_non_atomic', call[0])
883
self.assertEqual('filename.kndx', call[1][0])
884
# With no history, _KndxIndex writes a new index:
885
self.assertEqual(_KndxIndex.HEADER,
886
call[1][1].getvalue())
887
self.assertEqual({'create_parent_dir': True}, call[2])
889
def test_read_utf8_version_id(self):
890
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
891
utf8_revision_id = unicode_revision_id.encode('utf-8')
892
transport = MockTransport([
894
'%s option 0 1 :' % (utf8_revision_id,)
896
index = self.get_knit_index(transport, "filename", "r")
897
# _KndxIndex is a private class, and deals in utf8 revision_ids, not
898
# Unicode revision_ids.
899
self.assertEqual({(utf8_revision_id,):()},
900
index.get_parent_map(index.keys()))
901
self.assertFalse((unicode_revision_id,) in index.keys())
903
def test_read_utf8_parents(self):
904
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
905
utf8_revision_id = unicode_revision_id.encode('utf-8')
906
transport = MockTransport([
908
"version option 0 1 .%s :" % (utf8_revision_id,)
910
index = self.get_knit_index(transport, "filename", "r")
911
self.assertEqual({("version",):((utf8_revision_id,),)},
912
index.get_parent_map(index.keys()))
914
def test_read_ignore_corrupted_lines(self):
915
transport = MockTransport([
918
"corrupted options 0 1 .b .c ",
919
"version options 0 1 :"
921
index = self.get_knit_index(transport, "filename", "r")
922
self.assertEqual(1, len(index.keys()))
923
self.assertEqual(set([("version",)]), index.keys())
925
def test_read_corrupted_header(self):
926
transport = MockTransport(['not a bzr knit index header\n'])
927
index = self.get_knit_index(transport, "filename", "r")
928
self.assertRaises(KnitHeaderError, index.keys)
930
def test_read_duplicate_entries(self):
931
transport = MockTransport([
933
"parent options 0 1 :",
934
"version options1 0 1 0 :",
935
"version options2 1 2 .other :",
936
"version options3 3 4 0 .other :"
938
index = self.get_knit_index(transport, "filename", "r")
939
self.assertEqual(2, len(index.keys()))
940
# check that the index used is the first one written. (Specific
941
# to KnitIndex style indices.
942
self.assertEqual("1", index._dictionary_compress([("version",)]))
943
self.assertEqual((("version",), 3, 4), index.get_position(("version",)))
944
self.assertEqual(["options3"], index.get_options(("version",)))
945
self.assertEqual({("version",):(("parent",), ("other",))},
946
index.get_parent_map([("version",)]))
948
def test_read_compressed_parents(self):
949
transport = MockTransport([
953
"c option 0 1 1 0 :",
955
index = self.get_knit_index(transport, "filename", "r")
956
self.assertEqual({("b",):(("a",),), ("c",):(("b",), ("a",))},
957
index.get_parent_map([("b",), ("c",)]))
959
def test_write_utf8_version_id(self):
960
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
961
utf8_revision_id = unicode_revision_id.encode('utf-8')
962
transport = MockTransport([
965
index = self.get_knit_index(transport, "filename", "r")
967
((utf8_revision_id,), ["option"], ((utf8_revision_id,), 0, 1), [])])
968
call = transport.calls.pop(0)
969
# call[1][1] is a StringIO - we can't test it by simple equality.
970
self.assertEqual('put_file_non_atomic', call[0])
971
self.assertEqual('filename.kndx', call[1][0])
972
# With no history, _KndxIndex writes a new index:
973
self.assertEqual(_KndxIndex.HEADER +
974
"\n%s option 0 1 :" % (utf8_revision_id,),
975
call[1][1].getvalue())
976
self.assertEqual({'create_parent_dir': True}, call[2])
978
def test_write_utf8_parents(self):
979
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
980
utf8_revision_id = unicode_revision_id.encode('utf-8')
981
transport = MockTransport([
984
index = self.get_knit_index(transport, "filename", "r")
986
(("version",), ["option"], (("version",), 0, 1), [(utf8_revision_id,)])])
987
call = transport.calls.pop(0)
988
# call[1][1] is a StringIO - we can't test it by simple equality.
989
self.assertEqual('put_file_non_atomic', call[0])
990
self.assertEqual('filename.kndx', call[1][0])
991
# With no history, _KndxIndex writes a new index:
992
self.assertEqual(_KndxIndex.HEADER +
993
"\nversion option 0 1 .%s :" % (utf8_revision_id,),
994
call[1][1].getvalue())
995
self.assertEqual({'create_parent_dir': True}, call[2])
998
transport = MockTransport([
1001
index = self.get_knit_index(transport, "filename", "r")
1003
self.assertEqual(set(), index.keys())
1005
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1006
self.assertEqual(set([("a",)]), index.keys())
1008
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1009
self.assertEqual(set([("a",)]), index.keys())
1011
index.add_records([(("b",), ["option"], (("b",), 0, 1), [])])
1012
self.assertEqual(set([("a",), ("b",)]), index.keys())
1014
def add_a_b(self, index, random_id=None):
1016
if random_id is not None:
1017
kwargs["random_id"] = random_id
1019
(("a",), ["option"], (("a",), 0, 1), [("b",)]),
1020
(("a",), ["opt"], (("a",), 1, 2), [("c",)]),
1021
(("b",), ["option"], (("b",), 2, 3), [("a",)])
1024
def assertIndexIsAB(self, index):
1029
index.get_parent_map(index.keys()))
1030
self.assertEqual((("a",), 1, 2), index.get_position(("a",)))
1031
self.assertEqual((("b",), 2, 3), index.get_position(("b",)))
1032
self.assertEqual(["opt"], index.get_options(("a",)))
1034
def test_add_versions(self):
1035
transport = MockTransport([
1038
index = self.get_knit_index(transport, "filename", "r")
1041
call = transport.calls.pop(0)
1042
# call[1][1] is a StringIO - we can't test it by simple equality.
1043
self.assertEqual('put_file_non_atomic', call[0])
1044
self.assertEqual('filename.kndx', call[1][0])
1045
# With no history, _KndxIndex writes a new index:
1048
"\na option 0 1 .b :"
1050
"\nb option 2 3 0 :",
1051
call[1][1].getvalue())
1052
self.assertEqual({'create_parent_dir': True}, call[2])
1053
self.assertIndexIsAB(index)
1055
def test_add_versions_random_id_is_accepted(self):
1056
transport = MockTransport([
1059
index = self.get_knit_index(transport, "filename", "r")
1060
self.add_a_b(index, random_id=True)
1062
def test_delay_create_and_add_versions(self):
1063
transport = MockTransport()
1065
index = self.get_knit_index(transport, "filename", "w")
1067
self.assertEqual([], transport.calls)
1070
#[ {"dir_mode": 0777, "create_parent_dir": True, "mode": "wb"},
1072
# Two calls: one during which we load the existing index (and when its
1073
# missing create it), then a second where we write the contents out.
1074
self.assertEqual(2, len(transport.calls))
1075
call = transport.calls.pop(0)
1076
self.assertEqual('put_file_non_atomic', call[0])
1077
self.assertEqual('filename.kndx', call[1][0])
1078
# With no history, _KndxIndex writes a new index:
1079
self.assertEqual(_KndxIndex.HEADER, call[1][1].getvalue())
1080
self.assertEqual({'create_parent_dir': True}, call[2])
1081
call = transport.calls.pop(0)
1082
# call[1][1] is a StringIO - we can't test it by simple equality.
1083
self.assertEqual('put_file_non_atomic', call[0])
1084
self.assertEqual('filename.kndx', call[1][0])
1085
# With no history, _KndxIndex writes a new index:
1088
"\na option 0 1 .b :"
1090
"\nb option 2 3 0 :",
1091
call[1][1].getvalue())
1092
self.assertEqual({'create_parent_dir': True}, call[2])
1094
def assertTotalBuildSize(self, size, keys, positions):
1095
self.assertEqual(size,
1096
knit._get_total_build_size(None, keys, positions))
1098
def test__get_total_build_size(self):
1100
('a',): (('fulltext', False), (('a',), 0, 100), None),
1101
('b',): (('line-delta', False), (('b',), 100, 21), ('a',)),
1102
('c',): (('line-delta', False), (('c',), 121, 35), ('b',)),
1103
('d',): (('line-delta', False), (('d',), 156, 12), ('b',)),
1105
self.assertTotalBuildSize(100, [('a',)], positions)
1106
self.assertTotalBuildSize(121, [('b',)], positions)
1107
# c needs both a & b
1108
self.assertTotalBuildSize(156, [('c',)], positions)
1109
# we shouldn't count 'b' twice
1110
self.assertTotalBuildSize(156, [('b',), ('c',)], positions)
1111
self.assertTotalBuildSize(133, [('d',)], positions)
1112
self.assertTotalBuildSize(168, [('c',), ('d',)], positions)
1114
def test_get_position(self):
1115
transport = MockTransport([
1120
index = self.get_knit_index(transport, "filename", "r")
1122
self.assertEqual((("a",), 0, 1), index.get_position(("a",)))
1123
self.assertEqual((("b",), 1, 2), index.get_position(("b",)))
1125
def test_get_method(self):
1126
transport = MockTransport([
1128
"a fulltext,unknown 0 1 :",
1129
"b unknown,line-delta 1 2 :",
1132
index = self.get_knit_index(transport, "filename", "r")
1134
self.assertEqual("fulltext", index.get_method("a"))
1135
self.assertEqual("line-delta", index.get_method("b"))
1136
self.assertRaises(errors.KnitIndexUnknownMethod, index.get_method, "c")
1138
def test_get_options(self):
1139
transport = MockTransport([
1144
index = self.get_knit_index(transport, "filename", "r")
1146
self.assertEqual(["opt1"], index.get_options("a"))
1147
self.assertEqual(["opt2", "opt3"], index.get_options("b"))
1149
def test_get_parent_map(self):
1150
transport = MockTransport([
1153
"b option 1 2 0 .c :",
1154
"c option 1 2 1 0 .e :"
1156
index = self.get_knit_index(transport, "filename", "r")
1160
("b",):(("a",), ("c",)),
1161
("c",):(("b",), ("a",), ("e",)),
1162
}, index.get_parent_map(index.keys()))
1164
def test_impossible_parent(self):
1165
"""Test we get KnitCorrupt if the parent couldn't possibly exist."""
1166
transport = MockTransport([
1169
"b option 0 1 4 :" # We don't have a 4th record
1171
index = self.get_knit_index(transport, 'filename', 'r')
1173
self.assertRaises(errors.KnitCorrupt, index.keys)
1174
except TypeError, e:
1175
if (str(e) == ('exceptions must be strings, classes, or instances,'
1176
' not exceptions.IndexError')
1177
and sys.version_info[0:2] >= (2,5)):
1178
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1179
' raising new style exceptions with python'
1184
def test_corrupted_parent(self):
1185
transport = MockTransport([
1189
"c option 0 1 1v :", # Can't have a parent of '1v'
1191
index = self.get_knit_index(transport, 'filename', 'r')
1193
self.assertRaises(errors.KnitCorrupt, index.keys)
1194
except TypeError, e:
1195
if (str(e) == ('exceptions must be strings, classes, or instances,'
1196
' not exceptions.ValueError')
1197
and sys.version_info[0:2] >= (2,5)):
1198
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1199
' raising new style exceptions with python'
1204
def test_corrupted_parent_in_list(self):
1205
transport = MockTransport([
1209
"c option 0 1 1 v :", # Can't have a parent of 'v'
1211
index = self.get_knit_index(transport, 'filename', 'r')
1213
self.assertRaises(errors.KnitCorrupt, index.keys)
1214
except TypeError, e:
1215
if (str(e) == ('exceptions must be strings, classes, or instances,'
1216
' not exceptions.ValueError')
1217
and sys.version_info[0:2] >= (2,5)):
1218
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1219
' raising new style exceptions with python'
1224
def test_invalid_position(self):
1225
transport = MockTransport([
1229
index = self.get_knit_index(transport, 'filename', 'r')
1231
self.assertRaises(errors.KnitCorrupt, index.keys)
1232
except TypeError, e:
1233
if (str(e) == ('exceptions must be strings, classes, or instances,'
1234
' not exceptions.ValueError')
1235
and sys.version_info[0:2] >= (2,5)):
1236
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1237
' raising new style exceptions with python'
1242
def test_invalid_size(self):
1243
transport = MockTransport([
1247
index = self.get_knit_index(transport, 'filename', 'r')
1249
self.assertRaises(errors.KnitCorrupt, index.keys)
1250
except TypeError, e:
1251
if (str(e) == ('exceptions must be strings, classes, or instances,'
1252
' not exceptions.ValueError')
1253
and sys.version_info[0:2] >= (2,5)):
1254
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1255
' raising new style exceptions with python'
1260
def test_scan_unvalidated_index_not_implemented(self):
1261
transport = MockTransport()
1262
index = self.get_knit_index(transport, 'filename', 'r')
1264
NotImplementedError, index.scan_unvalidated_index,
1265
'dummy graph_index')
1267
NotImplementedError, index.get_missing_compression_parents)
1269
def test_short_line(self):
1270
transport = MockTransport([
1273
"b option 10 10 0", # This line isn't terminated, ignored
1275
index = self.get_knit_index(transport, "filename", "r")
1276
self.assertEqual(set([('a',)]), index.keys())
1278
def test_skip_incomplete_record(self):
1279
# A line with bogus data should just be skipped
1280
transport = MockTransport([
1283
"b option 10 10 0", # This line isn't terminated, ignored
1284
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1286
index = self.get_knit_index(transport, "filename", "r")
1287
self.assertEqual(set([('a',), ('c',)]), index.keys())
1289
def test_trailing_characters(self):
1290
# A line with bogus data should just be skipped
1291
transport = MockTransport([
1294
"b option 10 10 0 :a", # This line has extra trailing characters
1295
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1297
index = self.get_knit_index(transport, "filename", "r")
1298
self.assertEqual(set([('a',), ('c',)]), index.keys())
1301
class LowLevelKnitIndexTests_c(LowLevelKnitIndexTests):
1303
_test_needs_features = [CompiledKnitFeature]
1305
def get_knit_index(self, transport, name, mode):
1306
mapper = ConstantMapper(name)
1307
orig = knit._load_data
1309
knit._load_data = orig
1310
self.addCleanup(reset)
1311
from bzrlib._knit_load_data_c import _load_data_c
1312
knit._load_data = _load_data_c
1313
allow_writes = lambda: mode == 'w'
1314
return _KndxIndex(transport, mapper, lambda:None, allow_writes, lambda:True)
1317
class KnitTests(TestCaseWithTransport):
1318
"""Class containing knit test helper routines."""
1320
def make_test_knit(self, annotate=False, name='test'):
1321
mapper = ConstantMapper(name)
1322
return make_file_factory(annotate, mapper)(self.get_transport())
1325
class TestBadShaError(KnitTests):
1326
"""Tests for handling of sha errors."""
1328
def test_sha_exception_has_text(self):
1329
# having the failed text included in the error allows for recovery.
1330
source = self.make_test_knit()
1331
target = self.make_test_knit(name="target")
1332
if not source._max_delta_chain:
1333
raise TestNotApplicable(
1334
"cannot get delta-caused sha failures without deltas.")
1337
broken = ('broken',)
1338
source.add_lines(basis, (), ['foo\n'])
1339
source.add_lines(broken, (basis,), ['foo\n', 'bar\n'])
1340
# Seed target with a bad basis text
1341
target.add_lines(basis, (), ['gam\n'])
1342
target.insert_record_stream(
1343
source.get_record_stream([broken], 'unordered', False))
1344
err = self.assertRaises(errors.KnitCorrupt,
1345
target.get_record_stream([broken], 'unordered', True
1346
).next().get_bytes_as, 'chunked')
1347
self.assertEqual(['gam\n', 'bar\n'], err.content)
1348
# Test for formatting with live data
1349
self.assertStartsWith(str(err), "Knit ")
1352
class TestKnitIndex(KnitTests):
1354
def test_add_versions_dictionary_compresses(self):
1355
"""Adding versions to the index should update the lookup dict"""
1356
knit = self.make_test_knit()
1358
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1359
self.check_file_contents('test.kndx',
1360
'# bzr knit index 8\n'
1362
'a-1 fulltext 0 0 :'
1365
(('a-2',), ['fulltext'], (('a-2',), 0, 0), [('a-1',)]),
1366
(('a-3',), ['fulltext'], (('a-3',), 0, 0), [('a-2',)]),
1368
self.check_file_contents('test.kndx',
1369
'# bzr knit index 8\n'
1371
'a-1 fulltext 0 0 :\n'
1372
'a-2 fulltext 0 0 0 :\n'
1373
'a-3 fulltext 0 0 1 :'
1375
self.assertEqual(set([('a-3',), ('a-1',), ('a-2',)]), idx.keys())
1377
('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False)),
1378
('a-2',): ((('a-2',), 0, 0), None, (('a-1',),), ('fulltext', False)),
1379
('a-3',): ((('a-3',), 0, 0), None, (('a-2',),), ('fulltext', False)),
1380
}, idx.get_build_details(idx.keys()))
1381
self.assertEqual({('a-1',):(),
1382
('a-2',):(('a-1',),),
1383
('a-3',):(('a-2',),),},
1384
idx.get_parent_map(idx.keys()))
1386
def test_add_versions_fails_clean(self):
1387
"""If add_versions fails in the middle, it restores a pristine state.
1389
Any modifications that are made to the index are reset if all versions
1392
# This cheats a little bit by passing in a generator which will
1393
# raise an exception before the processing finishes
1394
# Other possibilities would be to have an version with the wrong number
1395
# of entries, or to make the backing transport unable to write any
1398
knit = self.make_test_knit()
1400
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1402
class StopEarly(Exception):
1405
def generate_failure():
1406
"""Add some entries and then raise an exception"""
1407
yield (('a-2',), ['fulltext'], (None, 0, 0), ('a-1',))
1408
yield (('a-3',), ['fulltext'], (None, 0, 0), ('a-2',))
1411
# Assert the pre-condition
1413
self.assertEqual(set([('a-1',)]), set(idx.keys()))
1415
{('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False))},
1416
idx.get_build_details([('a-1',)]))
1417
self.assertEqual({('a-1',):()}, idx.get_parent_map(idx.keys()))
1420
self.assertRaises(StopEarly, idx.add_records, generate_failure())
1421
# And it shouldn't be modified
1424
def test_knit_index_ignores_empty_files(self):
1425
# There was a race condition in older bzr, where a ^C at the right time
1426
# could leave an empty .kndx file, which bzr would later claim was a
1427
# corrupted file since the header was not present. In reality, the file
1428
# just wasn't created, so it should be ignored.
1429
t = get_transport('.')
1430
t.put_bytes('test.kndx', '')
1432
knit = self.make_test_knit()
1434
def test_knit_index_checks_header(self):
1435
t = get_transport('.')
1436
t.put_bytes('test.kndx', '# not really a knit header\n\n')
1437
k = self.make_test_knit()
1438
self.assertRaises(KnitHeaderError, k.keys)
1441
class TestGraphIndexKnit(KnitTests):
1442
"""Tests for knits using a GraphIndex rather than a KnitIndex."""
1444
def make_g_index(self, name, ref_lists=0, nodes=[]):
1445
builder = GraphIndexBuilder(ref_lists)
1446
for node, references, value in nodes:
1447
builder.add_node(node, references, value)
1448
stream = builder.finish()
1449
trans = self.get_transport()
1450
size = trans.put_file(name, stream)
1451
return GraphIndex(trans, name, size)
1453
def two_graph_index(self, deltas=False, catch_adds=False):
1454
"""Build a two-graph index.
1456
:param deltas: If true, use underlying indices with two node-ref
1457
lists and 'parent' set to a delta-compressed against tail.
1459
# build a complex graph across several indices.
1461
# delta compression inn the index
1462
index1 = self.make_g_index('1', 2, [
1463
(('tip', ), 'N0 100', ([('parent', )], [], )),
1464
(('tail', ), '', ([], []))])
1465
index2 = self.make_g_index('2', 2, [
1466
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], [('tail', )])),
1467
(('separate', ), '', ([], []))])
1469
# just blob location and graph in the index.
1470
index1 = self.make_g_index('1', 1, [
1471
(('tip', ), 'N0 100', ([('parent', )], )),
1472
(('tail', ), '', ([], ))])
1473
index2 = self.make_g_index('2', 1, [
1474
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], )),
1475
(('separate', ), '', ([], ))])
1476
combined_index = CombinedGraphIndex([index1, index2])
1478
self.combined_index = combined_index
1479
self.caught_entries = []
1480
add_callback = self.catch_add
1483
return _KnitGraphIndex(combined_index, lambda:True, deltas=deltas,
1484
add_callback=add_callback)
1486
def test_keys(self):
1487
index = self.two_graph_index()
1488
self.assertEqual(set([('tail',), ('tip',), ('parent',), ('separate',)]),
1491
def test_get_position(self):
1492
index = self.two_graph_index()
1493
self.assertEqual((index._graph_index._indices[0], 0, 100), index.get_position(('tip',)))
1494
self.assertEqual((index._graph_index._indices[1], 100, 78), index.get_position(('parent',)))
1496
def test_get_method_deltas(self):
1497
index = self.two_graph_index(deltas=True)
1498
self.assertEqual('fulltext', index.get_method(('tip',)))
1499
self.assertEqual('line-delta', index.get_method(('parent',)))
1501
def test_get_method_no_deltas(self):
1502
# check that the parent-history lookup is ignored with deltas=False.
1503
index = self.two_graph_index(deltas=False)
1504
self.assertEqual('fulltext', index.get_method(('tip',)))
1505
self.assertEqual('fulltext', index.get_method(('parent',)))
1507
def test_get_options_deltas(self):
1508
index = self.two_graph_index(deltas=True)
1509
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1510
self.assertEqual(['line-delta'], index.get_options(('parent',)))
1512
def test_get_options_no_deltas(self):
1513
# check that the parent-history lookup is ignored with deltas=False.
1514
index = self.two_graph_index(deltas=False)
1515
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1516
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1518
def test_get_parent_map(self):
1519
index = self.two_graph_index()
1520
self.assertEqual({('parent',):(('tail',), ('ghost',))},
1521
index.get_parent_map([('parent',), ('ghost',)]))
1523
def catch_add(self, entries):
1524
self.caught_entries.append(entries)
1526
def test_add_no_callback_errors(self):
1527
index = self.two_graph_index()
1528
self.assertRaises(errors.ReadOnlyError, index.add_records,
1529
[(('new',), 'fulltext,no-eol', (None, 50, 60), ['separate'])])
1531
def test_add_version_smoke(self):
1532
index = self.two_graph_index(catch_adds=True)
1533
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60),
1535
self.assertEqual([[(('new', ), 'N50 60', ((('separate',),),))]],
1536
self.caught_entries)
1538
def test_add_version_delta_not_delta_index(self):
1539
index = self.two_graph_index(catch_adds=True)
1540
self.assertRaises(errors.KnitCorrupt, index.add_records,
1541
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1542
self.assertEqual([], self.caught_entries)
1544
def test_add_version_same_dup(self):
1545
index = self.two_graph_index(catch_adds=True)
1546
# options can be spelt two different ways
1547
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
1548
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
1549
# position/length are ignored (because each pack could have fulltext or
1550
# delta, and be at a different position.
1551
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1553
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1555
# but neither should have added data:
1556
self.assertEqual([[], [], [], []], self.caught_entries)
1558
def test_add_version_different_dup(self):
1559
index = self.two_graph_index(deltas=True, catch_adds=True)
1561
self.assertRaises(errors.KnitCorrupt, index.add_records,
1562
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1563
self.assertRaises(errors.KnitCorrupt, index.add_records,
1564
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1566
self.assertRaises(errors.KnitCorrupt, index.add_records,
1567
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1568
self.assertEqual([], self.caught_entries)
1570
def test_add_versions_nodeltas(self):
1571
index = self.two_graph_index(catch_adds=True)
1573
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1574
(('new2',), 'fulltext', (None, 0, 6), [('new',)]),
1576
self.assertEqual([(('new', ), 'N50 60', ((('separate',),),)),
1577
(('new2', ), ' 0 6', ((('new',),),))],
1578
sorted(self.caught_entries[0]))
1579
self.assertEqual(1, len(self.caught_entries))
1581
def test_add_versions_deltas(self):
1582
index = self.two_graph_index(deltas=True, catch_adds=True)
1584
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1585
(('new2',), 'line-delta', (None, 0, 6), [('new',)]),
1587
self.assertEqual([(('new', ), 'N50 60', ((('separate',),), ())),
1588
(('new2', ), ' 0 6', ((('new',),), (('new',),), ))],
1589
sorted(self.caught_entries[0]))
1590
self.assertEqual(1, len(self.caught_entries))
1592
def test_add_versions_delta_not_delta_index(self):
1593
index = self.two_graph_index(catch_adds=True)
1594
self.assertRaises(errors.KnitCorrupt, index.add_records,
1595
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1596
self.assertEqual([], self.caught_entries)
1598
def test_add_versions_random_id_accepted(self):
1599
index = self.two_graph_index(catch_adds=True)
1600
index.add_records([], random_id=True)
1602
def test_add_versions_same_dup(self):
1603
index = self.two_graph_index(catch_adds=True)
1604
# options can be spelt two different ways
1605
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100),
1607
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100),
1609
# position/length are ignored (because each pack could have fulltext or
1610
# delta, and be at a different position.
1611
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1613
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1615
# but neither should have added data.
1616
self.assertEqual([[], [], [], []], self.caught_entries)
1618
def test_add_versions_different_dup(self):
1619
index = self.two_graph_index(deltas=True, catch_adds=True)
1621
self.assertRaises(errors.KnitCorrupt, index.add_records,
1622
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1623
self.assertRaises(errors.KnitCorrupt, index.add_records,
1624
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1626
self.assertRaises(errors.KnitCorrupt, index.add_records,
1627
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1628
# change options in the second record
1629
self.assertRaises(errors.KnitCorrupt, index.add_records,
1630
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)]),
1631
(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1632
self.assertEqual([], self.caught_entries)
1634
def make_g_index_missing_compression_parent(self):
1635
graph_index = self.make_g_index('missing_comp', 2,
1636
[(('tip', ), ' 100 78',
1637
([('missing-parent', ), ('ghost', )], [('missing-parent', )]))])
1640
def make_g_index_no_external_refs(self):
1641
graph_index = self.make_g_index('no_external_refs', 2,
1642
[(('rev', ), ' 100 78',
1643
([('parent', ), ('ghost', )], []))])
1646
def test_add_good_unvalidated_index(self):
1647
unvalidated = self.make_g_index_no_external_refs()
1648
combined = CombinedGraphIndex([unvalidated])
1649
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1650
index.scan_unvalidated_index(unvalidated)
1651
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1653
def test_add_incomplete_unvalidated_index(self):
1654
unvalidated = self.make_g_index_missing_compression_parent()
1655
combined = CombinedGraphIndex([unvalidated])
1656
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1657
index.scan_unvalidated_index(unvalidated)
1658
# This also checks that its only the compression parent that is
1659
# examined, otherwise 'ghost' would also be reported as a missing
1662
frozenset([('missing-parent',)]),
1663
index.get_missing_compression_parents())
1665
def test_add_unvalidated_index_with_present_external_references(self):
1666
index = self.two_graph_index(deltas=True)
1667
# Ugly hack to get at one of the underlying GraphIndex objects that
1668
# two_graph_index built.
1669
unvalidated = index._graph_index._indices[1]
1670
# 'parent' is an external ref of _indices[1] (unvalidated), but is
1671
# present in _indices[0].
1672
index.scan_unvalidated_index(unvalidated)
1673
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1675
def make_new_missing_parent_g_index(self, name):
1676
missing_parent = name + '-missing-parent'
1677
graph_index = self.make_g_index(name, 2,
1678
[((name + 'tip', ), ' 100 78',
1679
([(missing_parent, ), ('ghost', )], [(missing_parent, )]))])
1682
def test_add_mulitiple_unvalidated_indices_with_missing_parents(self):
1683
g_index_1 = self.make_new_missing_parent_g_index('one')
1684
g_index_2 = self.make_new_missing_parent_g_index('two')
1685
combined = CombinedGraphIndex([g_index_1, g_index_2])
1686
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1687
index.scan_unvalidated_index(g_index_1)
1688
index.scan_unvalidated_index(g_index_2)
1690
frozenset([('one-missing-parent',), ('two-missing-parent',)]),
1691
index.get_missing_compression_parents())
1693
def test_add_mulitiple_unvalidated_indices_with_mutual_dependencies(self):
1694
graph_index_a = self.make_g_index('one', 2,
1695
[(('parent-one', ), ' 100 78', ([('non-compression-parent',)], [])),
1696
(('child-of-two', ), ' 100 78',
1697
([('parent-two',)], [('parent-two',)]))])
1698
graph_index_b = self.make_g_index('two', 2,
1699
[(('parent-two', ), ' 100 78', ([('non-compression-parent',)], [])),
1700
(('child-of-one', ), ' 100 78',
1701
([('parent-one',)], [('parent-one',)]))])
1702
combined = CombinedGraphIndex([graph_index_a, graph_index_b])
1703
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1704
index.scan_unvalidated_index(graph_index_a)
1705
index.scan_unvalidated_index(graph_index_b)
1707
frozenset([]), index.get_missing_compression_parents())
1710
class TestNoParentsGraphIndexKnit(KnitTests):
1711
"""Tests for knits using _KnitGraphIndex with no parents."""
1713
def make_g_index(self, name, ref_lists=0, nodes=[]):
1714
builder = GraphIndexBuilder(ref_lists)
1715
for node, references in nodes:
1716
builder.add_node(node, references)
1717
stream = builder.finish()
1718
trans = self.get_transport()
1719
size = trans.put_file(name, stream)
1720
return GraphIndex(trans, name, size)
1722
def test_add_good_unvalidated_index(self):
1723
unvalidated = self.make_g_index('unvalidated')
1724
combined = CombinedGraphIndex([unvalidated])
1725
index = _KnitGraphIndex(combined, lambda: True, parents=False)
1726
index.scan_unvalidated_index(unvalidated)
1727
self.assertEqual(frozenset(),
1728
index.get_missing_compression_parents())
1730
def test_parents_deltas_incompatible(self):
1731
index = CombinedGraphIndex([])
1732
self.assertRaises(errors.KnitError, _KnitGraphIndex, lambda:True,
1733
index, deltas=True, parents=False)
1735
def two_graph_index(self, catch_adds=False):
1736
"""Build a two-graph index.
1738
:param deltas: If true, use underlying indices with two node-ref
1739
lists and 'parent' set to a delta-compressed against tail.
1741
# put several versions in the index.
1742
index1 = self.make_g_index('1', 0, [
1743
(('tip', ), 'N0 100'),
1745
index2 = self.make_g_index('2', 0, [
1746
(('parent', ), ' 100 78'),
1747
(('separate', ), '')])
1748
combined_index = CombinedGraphIndex([index1, index2])
1750
self.combined_index = combined_index
1751
self.caught_entries = []
1752
add_callback = self.catch_add
1755
return _KnitGraphIndex(combined_index, lambda:True, parents=False,
1756
add_callback=add_callback)
1758
def test_keys(self):
1759
index = self.two_graph_index()
1760
self.assertEqual(set([('tail',), ('tip',), ('parent',), ('separate',)]),
1763
def test_get_position(self):
1764
index = self.two_graph_index()
1765
self.assertEqual((index._graph_index._indices[0], 0, 100),
1766
index.get_position(('tip',)))
1767
self.assertEqual((index._graph_index._indices[1], 100, 78),
1768
index.get_position(('parent',)))
1770
def test_get_method(self):
1771
index = self.two_graph_index()
1772
self.assertEqual('fulltext', index.get_method(('tip',)))
1773
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1775
def test_get_options(self):
1776
index = self.two_graph_index()
1777
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1778
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1780
def test_get_parent_map(self):
1781
index = self.two_graph_index()
1782
self.assertEqual({('parent',):None},
1783
index.get_parent_map([('parent',), ('ghost',)]))
1785
def catch_add(self, entries):
1786
self.caught_entries.append(entries)
1788
def test_add_no_callback_errors(self):
1789
index = self.two_graph_index()
1790
self.assertRaises(errors.ReadOnlyError, index.add_records,
1791
[(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)])])
1793
def test_add_version_smoke(self):
1794
index = self.two_graph_index(catch_adds=True)
1795
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60), [])])
1796
self.assertEqual([[(('new', ), 'N50 60')]],
1797
self.caught_entries)
1799
def test_add_version_delta_not_delta_index(self):
1800
index = self.two_graph_index(catch_adds=True)
1801
self.assertRaises(errors.KnitCorrupt, index.add_records,
1802
[(('new',), 'no-eol,line-delta', (None, 0, 100), [])])
1803
self.assertEqual([], self.caught_entries)
1805
def test_add_version_same_dup(self):
1806
index = self.two_graph_index(catch_adds=True)
1807
# options can be spelt two different ways
1808
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1809
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
1810
# position/length are ignored (because each pack could have fulltext or
1811
# delta, and be at a different position.
1812
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
1813
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
1814
# but neither should have added data.
1815
self.assertEqual([[], [], [], []], self.caught_entries)
1817
def test_add_version_different_dup(self):
1818
index = self.two_graph_index(catch_adds=True)
1820
self.assertRaises(errors.KnitCorrupt, index.add_records,
1821
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
1822
self.assertRaises(errors.KnitCorrupt, index.add_records,
1823
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
1824
self.assertRaises(errors.KnitCorrupt, index.add_records,
1825
[(('tip',), 'fulltext', (None, 0, 100), [])])
1827
self.assertRaises(errors.KnitCorrupt, index.add_records,
1828
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
1829
self.assertEqual([], self.caught_entries)
1831
def test_add_versions(self):
1832
index = self.two_graph_index(catch_adds=True)
1834
(('new',), 'fulltext,no-eol', (None, 50, 60), []),
1835
(('new2',), 'fulltext', (None, 0, 6), []),
1837
self.assertEqual([(('new', ), 'N50 60'), (('new2', ), ' 0 6')],
1838
sorted(self.caught_entries[0]))
1839
self.assertEqual(1, len(self.caught_entries))
1841
def test_add_versions_delta_not_delta_index(self):
1842
index = self.two_graph_index(catch_adds=True)
1843
self.assertRaises(errors.KnitCorrupt, index.add_records,
1844
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1845
self.assertEqual([], self.caught_entries)
1847
def test_add_versions_parents_not_parents_index(self):
1848
index = self.two_graph_index(catch_adds=True)
1849
self.assertRaises(errors.KnitCorrupt, index.add_records,
1850
[(('new',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
1851
self.assertEqual([], self.caught_entries)
1853
def test_add_versions_random_id_accepted(self):
1854
index = self.two_graph_index(catch_adds=True)
1855
index.add_records([], random_id=True)
1857
def test_add_versions_same_dup(self):
1858
index = self.two_graph_index(catch_adds=True)
1859
# options can be spelt two different ways
1860
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1861
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
1862
# position/length are ignored (because each pack could have fulltext or
1863
# delta, and be at a different position.
1864
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
1865
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
1866
# but neither should have added data.
1867
self.assertEqual([[], [], [], []], self.caught_entries)
1869
def test_add_versions_different_dup(self):
1870
index = self.two_graph_index(catch_adds=True)
1872
self.assertRaises(errors.KnitCorrupt, index.add_records,
1873
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
1874
self.assertRaises(errors.KnitCorrupt, index.add_records,
1875
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
1876
self.assertRaises(errors.KnitCorrupt, index.add_records,
1877
[(('tip',), 'fulltext', (None, 0, 100), [])])
1879
self.assertRaises(errors.KnitCorrupt, index.add_records,
1880
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
1881
# change options in the second record
1882
self.assertRaises(errors.KnitCorrupt, index.add_records,
1883
[(('tip',), 'fulltext,no-eol', (None, 0, 100), []),
1884
(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
1885
self.assertEqual([], self.caught_entries)
1888
class TestKnitVersionedFiles(KnitTests):
1890
def assertGroupKeysForIo(self, exp_groups, keys, non_local_keys,
1891
positions, _min_buffer_size=None):
1892
kvf = self.make_test_knit()
1893
if _min_buffer_size is None:
1894
_min_buffer_size = knit._STREAM_MIN_BUFFER_SIZE
1895
self.assertEqual(exp_groups, kvf._group_keys_for_io(keys,
1896
non_local_keys, positions,
1897
_min_buffer_size=_min_buffer_size))
1899
def assertSplitByPrefix(self, expected_map, expected_prefix_order,
1901
split, prefix_order = KnitVersionedFiles._split_by_prefix(keys)
1902
self.assertEqual(expected_map, split)
1903
self.assertEqual(expected_prefix_order, prefix_order)
1905
def test__group_keys_for_io(self):
1906
ft_detail = ('fulltext', False)
1907
ld_detail = ('line-delta', False)
1915
f_a: (ft_detail, (f_a, 0, 100), None),
1916
f_b: (ld_detail, (f_b, 100, 21), f_a),
1917
f_c: (ld_detail, (f_c, 180, 15), f_b),
1918
g_a: (ft_detail, (g_a, 121, 35), None),
1919
g_b: (ld_detail, (g_b, 156, 12), g_a),
1920
g_c: (ld_detail, (g_c, 195, 13), g_a),
1922
self.assertGroupKeysForIo([([f_a], set())],
1923
[f_a], [], positions)
1924
self.assertGroupKeysForIo([([f_a], set([f_a]))],
1925
[f_a], [f_a], positions)
1926
self.assertGroupKeysForIo([([f_a, f_b], set([]))],
1927
[f_a, f_b], [], positions)
1928
self.assertGroupKeysForIo([([f_a, f_b], set([f_b]))],
1929
[f_a, f_b], [f_b], positions)
1930
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
1931
[f_a, g_a, f_b, g_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
_min_buffer_size=150)
1935
self.assertGroupKeysForIo([([f_a, f_b], set()), ([g_a, g_b], set())],
1936
[f_a, g_a, f_b, g_b], [], positions,
1937
_min_buffer_size=100)
1938
self.assertGroupKeysForIo([([f_c], set()), ([g_b], set())],
1939
[f_c, g_b], [], positions,
1940
_min_buffer_size=125)
1941
self.assertGroupKeysForIo([([g_b, f_c], set())],
1942
[g_b, f_c], [], positions,
1943
_min_buffer_size=125)
1945
def test__split_by_prefix(self):
1946
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
1947
'g': [('g', 'b'), ('g', 'a')],
1949
[('f', 'a'), ('g', 'b'),
1950
('g', 'a'), ('f', 'b')])
1952
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
1953
'g': [('g', 'b'), ('g', 'a')],
1955
[('f', 'a'), ('f', 'b'),
1956
('g', 'b'), ('g', 'a')])
1958
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
1959
'g': [('g', 'b'), ('g', 'a')],
1961
[('f', 'a'), ('f', 'b'),
1962
('g', 'b'), ('g', 'a')])
1964
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
1965
'g': [('g', 'b'), ('g', 'a')],
1966
'': [('a',), ('b',)]
1968
[('f', 'a'), ('g', 'b'),
1970
('g', 'a'), ('f', 'b')])
1973
class TestStacking(KnitTests):
1975
def get_basis_and_test_knit(self):
1976
basis = self.make_test_knit(name='basis')
1977
basis = RecordingVersionedFilesDecorator(basis)
1978
test = self.make_test_knit(name='test')
1979
test.add_fallback_versioned_files(basis)
1982
def test_add_fallback_versioned_files(self):
1983
basis = self.make_test_knit(name='basis')
1984
test = self.make_test_knit(name='test')
1985
# It must not error; other tests test that the fallback is referred to
1986
# when accessing data.
1987
test.add_fallback_versioned_files(basis)
1989
def test_add_lines(self):
1990
# lines added to the test are not added to the basis
1991
basis, test = self.get_basis_and_test_knit()
1993
key_basis = ('bar',)
1994
key_cross_border = ('quux',)
1995
key_delta = ('zaphod',)
1996
test.add_lines(key, (), ['foo\n'])
1997
self.assertEqual({}, basis.get_parent_map([key]))
1998
# lines added to the test that reference across the stack do a
2000
basis.add_lines(key_basis, (), ['foo\n'])
2002
test.add_lines(key_cross_border, (key_basis,), ['foo\n'])
2003
self.assertEqual('fulltext', test._index.get_method(key_cross_border))
2004
# we don't even need to look at the basis to see that this should be
2005
# stored as a fulltext
2006
self.assertEqual([], basis.calls)
2007
# Subsequent adds do delta.
2009
test.add_lines(key_delta, (key_cross_border,), ['foo\n'])
2010
self.assertEqual('line-delta', test._index.get_method(key_delta))
2011
self.assertEqual([], basis.calls)
2013
def test_annotate(self):
2014
# annotations from the test knit are answered without asking the basis
2015
basis, test = self.get_basis_and_test_knit()
2017
key_basis = ('bar',)
2018
key_missing = ('missing',)
2019
test.add_lines(key, (), ['foo\n'])
2020
details = test.annotate(key)
2021
self.assertEqual([(key, 'foo\n')], details)
2022
self.assertEqual([], basis.calls)
2023
# But texts that are not in the test knit are looked for in the basis
2025
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2027
details = test.annotate(key_basis)
2028
self.assertEqual([(key_basis, 'foo\n'), (key_basis, 'bar\n')], details)
2029
# Not optimised to date:
2030
# self.assertEqual([("annotate", key_basis)], basis.calls)
2031
self.assertEqual([('get_parent_map', set([key_basis])),
2032
('get_parent_map', set([key_basis])),
2033
('get_parent_map', set([key_basis])),
2034
('get_record_stream', [key_basis], 'unordered', True)],
2037
def test_check(self):
2038
# At the moment checking a stacked knit does implicitly check the
2040
basis, test = self.get_basis_and_test_knit()
2043
def test_get_parent_map(self):
2044
# parents in the test knit are answered without asking the basis
2045
basis, test = self.get_basis_and_test_knit()
2047
key_basis = ('bar',)
2048
key_missing = ('missing',)
2049
test.add_lines(key, (), [])
2050
parent_map = test.get_parent_map([key])
2051
self.assertEqual({key: ()}, parent_map)
2052
self.assertEqual([], basis.calls)
2053
# But parents that are not in the test knit are looked for in the basis
2054
basis.add_lines(key_basis, (), [])
2056
parent_map = test.get_parent_map([key, key_basis, key_missing])
2057
self.assertEqual({key: (),
2058
key_basis: ()}, parent_map)
2059
self.assertEqual([("get_parent_map", set([key_basis, key_missing]))],
2062
def test_get_record_stream_unordered_fulltexts(self):
2063
# records from the test knit are answered without asking the basis:
2064
basis, test = self.get_basis_and_test_knit()
2066
key_basis = ('bar',)
2067
key_missing = ('missing',)
2068
test.add_lines(key, (), ['foo\n'])
2069
records = list(test.get_record_stream([key], 'unordered', True))
2070
self.assertEqual(1, len(records))
2071
self.assertEqual([], basis.calls)
2072
# Missing (from test knit) objects are retrieved from the basis:
2073
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2075
records = list(test.get_record_stream([key_basis, key_missing],
2077
self.assertEqual(2, len(records))
2078
calls = list(basis.calls)
2079
for record in records:
2080
self.assertSubset([record.key], (key_basis, key_missing))
2081
if record.key == key_missing:
2082
self.assertIsInstance(record, AbsentContentFactory)
2084
reference = list(basis.get_record_stream([key_basis],
2085
'unordered', True))[0]
2086
self.assertEqual(reference.key, record.key)
2087
self.assertEqual(reference.sha1, record.sha1)
2088
self.assertEqual(reference.storage_kind, record.storage_kind)
2089
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2090
record.get_bytes_as(record.storage_kind))
2091
self.assertEqual(reference.get_bytes_as('fulltext'),
2092
record.get_bytes_as('fulltext'))
2093
# It's not strictly minimal, but it seems reasonable for now for it to
2094
# ask which fallbacks have which parents.
2096
("get_parent_map", set([key_basis, key_missing])),
2097
("get_record_stream", [key_basis], 'unordered', True)],
2100
def test_get_record_stream_ordered_fulltexts(self):
2101
# ordering is preserved down into the fallback store.
2102
basis, test = self.get_basis_and_test_knit()
2104
key_basis = ('bar',)
2105
key_basis_2 = ('quux',)
2106
key_missing = ('missing',)
2107
test.add_lines(key, (key_basis,), ['foo\n'])
2108
# Missing (from test knit) objects are retrieved from the basis:
2109
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2110
basis.add_lines(key_basis_2, (), ['quux\n'])
2112
# ask for in non-topological order
2113
records = list(test.get_record_stream(
2114
[key, key_basis, key_missing, key_basis_2], 'topological', True))
2115
self.assertEqual(4, len(records))
2117
for record in records:
2118
self.assertSubset([record.key],
2119
(key_basis, key_missing, key_basis_2, key))
2120
if record.key == key_missing:
2121
self.assertIsInstance(record, AbsentContentFactory)
2123
results.append((record.key, record.sha1, record.storage_kind,
2124
record.get_bytes_as('fulltext')))
2125
calls = list(basis.calls)
2126
order = [record[0] for record in results]
2127
self.assertEqual([key_basis_2, key_basis, key], order)
2128
for result in results:
2129
if result[0] == key:
2133
record = source.get_record_stream([result[0]], 'unordered',
2135
self.assertEqual(record.key, result[0])
2136
self.assertEqual(record.sha1, result[1])
2137
# We used to check that the storage kind matched, but actually it
2138
# depends on whether it was sourced from the basis, or in a single
2139
# group, because asking for full texts returns proxy objects to a
2140
# _ContentMapGenerator object; so checking the kind is unneeded.
2141
self.assertEqual(record.get_bytes_as('fulltext'), result[3])
2142
# It's not strictly minimal, but it seems reasonable for now for it to
2143
# ask which fallbacks have which parents.
2145
("get_parent_map", set([key_basis, key_basis_2, key_missing])),
2146
# unordered is asked for by the underlying worker as it still
2147
# buffers everything while answering - which is a problem!
2148
("get_record_stream", [key_basis_2, key_basis], 'unordered', True)],
2151
def test_get_record_stream_unordered_deltas(self):
2152
# records from the test knit are answered without asking the basis:
2153
basis, test = self.get_basis_and_test_knit()
2155
key_basis = ('bar',)
2156
key_missing = ('missing',)
2157
test.add_lines(key, (), ['foo\n'])
2158
records = list(test.get_record_stream([key], 'unordered', False))
2159
self.assertEqual(1, len(records))
2160
self.assertEqual([], basis.calls)
2161
# Missing (from test knit) objects are retrieved from the basis:
2162
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2164
records = list(test.get_record_stream([key_basis, key_missing],
2165
'unordered', False))
2166
self.assertEqual(2, len(records))
2167
calls = list(basis.calls)
2168
for record in records:
2169
self.assertSubset([record.key], (key_basis, key_missing))
2170
if record.key == key_missing:
2171
self.assertIsInstance(record, AbsentContentFactory)
2173
reference = list(basis.get_record_stream([key_basis],
2174
'unordered', False))[0]
2175
self.assertEqual(reference.key, record.key)
2176
self.assertEqual(reference.sha1, record.sha1)
2177
self.assertEqual(reference.storage_kind, record.storage_kind)
2178
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2179
record.get_bytes_as(record.storage_kind))
2180
# It's not strictly minimal, but it seems reasonable for now for it to
2181
# ask which fallbacks have which parents.
2183
("get_parent_map", set([key_basis, key_missing])),
2184
("get_record_stream", [key_basis], 'unordered', False)],
2187
def test_get_record_stream_ordered_deltas(self):
2188
# ordering is preserved down into the fallback store.
2189
basis, test = self.get_basis_and_test_knit()
2191
key_basis = ('bar',)
2192
key_basis_2 = ('quux',)
2193
key_missing = ('missing',)
2194
test.add_lines(key, (key_basis,), ['foo\n'])
2195
# Missing (from test knit) objects are retrieved from the basis:
2196
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2197
basis.add_lines(key_basis_2, (), ['quux\n'])
2199
# ask for in non-topological order
2200
records = list(test.get_record_stream(
2201
[key, key_basis, key_missing, key_basis_2], 'topological', False))
2202
self.assertEqual(4, len(records))
2204
for record in records:
2205
self.assertSubset([record.key],
2206
(key_basis, key_missing, key_basis_2, key))
2207
if record.key == key_missing:
2208
self.assertIsInstance(record, AbsentContentFactory)
2210
results.append((record.key, record.sha1, record.storage_kind,
2211
record.get_bytes_as(record.storage_kind)))
2212
calls = list(basis.calls)
2213
order = [record[0] for record in results]
2214
self.assertEqual([key_basis_2, key_basis, key], order)
2215
for result in results:
2216
if result[0] == key:
2220
record = source.get_record_stream([result[0]], 'unordered',
2222
self.assertEqual(record.key, result[0])
2223
self.assertEqual(record.sha1, result[1])
2224
self.assertEqual(record.storage_kind, result[2])
2225
self.assertEqual(record.get_bytes_as(record.storage_kind), result[3])
2226
# It's not strictly minimal, but it seems reasonable for now for it to
2227
# ask which fallbacks have which parents.
2229
("get_parent_map", set([key_basis, key_basis_2, key_missing])),
2230
("get_record_stream", [key_basis_2, key_basis], 'topological', False)],
2233
def test_get_sha1s(self):
2234
# sha1's in the test knit are answered without asking the basis
2235
basis, test = self.get_basis_and_test_knit()
2237
key_basis = ('bar',)
2238
key_missing = ('missing',)
2239
test.add_lines(key, (), ['foo\n'])
2240
key_sha1sum = osutils.sha('foo\n').hexdigest()
2241
sha1s = test.get_sha1s([key])
2242
self.assertEqual({key: key_sha1sum}, sha1s)
2243
self.assertEqual([], basis.calls)
2244
# But texts that are not in the test knit are looked for in the basis
2245
# directly (rather than via text reconstruction) so that remote servers
2246
# etc don't have to answer with full content.
2247
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2248
basis_sha1sum = osutils.sha('foo\nbar\n').hexdigest()
2250
sha1s = test.get_sha1s([key, key_missing, key_basis])
2251
self.assertEqual({key: key_sha1sum,
2252
key_basis: basis_sha1sum}, sha1s)
2253
self.assertEqual([("get_sha1s", set([key_basis, key_missing]))],
2256
def test_insert_record_stream(self):
2257
# records are inserted as normal; insert_record_stream builds on
2258
# add_lines, so a smoke test should be all that's needed:
2260
key_basis = ('bar',)
2261
key_delta = ('zaphod',)
2262
basis, test = self.get_basis_and_test_knit()
2263
source = self.make_test_knit(name='source')
2264
basis.add_lines(key_basis, (), ['foo\n'])
2266
source.add_lines(key_basis, (), ['foo\n'])
2267
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2268
stream = source.get_record_stream([key_delta], 'unordered', False)
2269
test.insert_record_stream(stream)
2270
# XXX: this does somewhat too many calls in making sure of whether it
2271
# has to recreate the full text.
2272
self.assertEqual([("get_parent_map", set([key_basis])),
2273
('get_parent_map', set([key_basis])),
2274
('get_record_stream', [key_basis], 'unordered', True)],
2276
self.assertEqual({key_delta:(key_basis,)},
2277
test.get_parent_map([key_delta]))
2278
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2279
'unordered', True).next().get_bytes_as('fulltext'))
2281
def test_iter_lines_added_or_present_in_keys(self):
2282
# Lines from the basis are returned, and lines for a given key are only
2286
# all sources are asked for keys:
2287
basis, test = self.get_basis_and_test_knit()
2288
basis.add_lines(key1, (), ["foo"])
2290
lines = list(test.iter_lines_added_or_present_in_keys([key1]))
2291
self.assertEqual([("foo\n", key1)], lines)
2292
self.assertEqual([("iter_lines_added_or_present_in_keys", set([key1]))],
2294
# keys in both are not duplicated:
2295
test.add_lines(key2, (), ["bar\n"])
2296
basis.add_lines(key2, (), ["bar\n"])
2298
lines = list(test.iter_lines_added_or_present_in_keys([key2]))
2299
self.assertEqual([("bar\n", key2)], lines)
2300
self.assertEqual([], basis.calls)
2302
def test_keys(self):
2305
# all sources are asked for keys:
2306
basis, test = self.get_basis_and_test_knit()
2308
self.assertEqual(set(), set(keys))
2309
self.assertEqual([("keys",)], basis.calls)
2310
# keys from a basis are returned:
2311
basis.add_lines(key1, (), [])
2314
self.assertEqual(set([key1]), set(keys))
2315
self.assertEqual([("keys",)], basis.calls)
2316
# keys in both are not duplicated:
2317
test.add_lines(key2, (), [])
2318
basis.add_lines(key2, (), [])
2321
self.assertEqual(2, len(keys))
2322
self.assertEqual(set([key1, key2]), set(keys))
2323
self.assertEqual([("keys",)], basis.calls)
2325
def test_add_mpdiffs(self):
2326
# records are inserted as normal; add_mpdiff builds on
2327
# add_lines, so a smoke test should be all that's needed:
2329
key_basis = ('bar',)
2330
key_delta = ('zaphod',)
2331
basis, test = self.get_basis_and_test_knit()
2332
source = self.make_test_knit(name='source')
2333
basis.add_lines(key_basis, (), ['foo\n'])
2335
source.add_lines(key_basis, (), ['foo\n'])
2336
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2337
diffs = source.make_mpdiffs([key_delta])
2338
test.add_mpdiffs([(key_delta, (key_basis,),
2339
source.get_sha1s([key_delta])[key_delta], diffs[0])])
2340
self.assertEqual([("get_parent_map", set([key_basis])),
2341
('get_record_stream', [key_basis], 'unordered', True),],
2343
self.assertEqual({key_delta:(key_basis,)},
2344
test.get_parent_map([key_delta]))
2345
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2346
'unordered', True).next().get_bytes_as('fulltext'))
2348
def test_make_mpdiffs(self):
2349
# Generating an mpdiff across a stacking boundary should detect parent
2353
key_right = ('zaphod',)
2354
basis, test = self.get_basis_and_test_knit()
2355
basis.add_lines(key_left, (), ['bar\n'])
2356
basis.add_lines(key_right, (), ['zaphod\n'])
2358
test.add_lines(key, (key_left, key_right),
2359
['bar\n', 'foo\n', 'zaphod\n'])
2360
diffs = test.make_mpdiffs([key])
2362
multiparent.MultiParent([multiparent.ParentText(0, 0, 0, 1),
2363
multiparent.NewText(['foo\n']),
2364
multiparent.ParentText(1, 0, 2, 1)])],
2366
self.assertEqual(3, len(basis.calls))
2368
("get_parent_map", set([key_left, key_right])),
2369
("get_parent_map", set([key_left, key_right])),
2372
last_call = basis.calls[-1]
2373
self.assertEqual('get_record_stream', last_call[0])
2374
self.assertEqual(set([key_left, key_right]), set(last_call[1]))
2375
self.assertEqual('unordered', last_call[2])
2376
self.assertEqual(True, last_call[3])
2379
class TestNetworkBehaviour(KnitTests):
2380
"""Tests for getting data out of/into knits over the network."""
2382
def test_include_delta_closure_generates_a_knit_delta_closure(self):
2383
vf = self.make_test_knit(name='test')
2384
# put in three texts, giving ft, delta, delta
2385
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2386
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2387
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2388
# But heuristics could interfere, so check what happened:
2389
self.assertEqual(['knit-ft-gz', 'knit-delta-gz', 'knit-delta-gz'],
2390
[record.storage_kind for record in
2391
vf.get_record_stream([('base',), ('d1',), ('d2',)],
2392
'topological', False)])
2393
# generate a stream of just the deltas include_delta_closure=True,
2394
# serialise to the network, and check that we get a delta closure on the wire.
2395
stream = vf.get_record_stream([('d1',), ('d2',)], 'topological', True)
2396
netb = [record.get_bytes_as(record.storage_kind) for record in stream]
2397
# The first bytes should be a memo from _ContentMapGenerator, and the
2398
# second bytes should be empty (because its a API proxy not something
2399
# for wire serialisation.
2400
self.assertEqual('', netb[1])
2402
kind, line_end = network_bytes_to_kind_and_offset(bytes)
2403
self.assertEqual('knit-delta-closure', kind)
2406
class TestContentMapGenerator(KnitTests):
2407
"""Tests for ContentMapGenerator"""
2409
def test_get_record_stream_gives_records(self):
2410
vf = self.make_test_knit(name='test')
2411
# put in three texts, giving ft, delta, delta
2412
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2413
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2414
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2415
keys = [('d1',), ('d2',)]
2416
generator = _VFContentMapGenerator(vf, keys,
2417
global_map=vf.get_parent_map(keys))
2418
for record in generator.get_record_stream():
2419
if record.key == ('d1',):
2420
self.assertEqual('d1\n', record.get_bytes_as('fulltext'))
2422
self.assertEqual('d2\n', record.get_bytes_as('fulltext'))
2424
def test_get_record_stream_kinds_are_raw(self):
2425
vf = self.make_test_knit(name='test')
2426
# put in three texts, giving ft, delta, delta
2427
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2428
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2429
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2430
keys = [('base',), ('d1',), ('d2',)]
2431
generator = _VFContentMapGenerator(vf, keys,
2432
global_map=vf.get_parent_map(keys))
2433
kinds = {('base',): 'knit-delta-closure',
2434
('d1',): 'knit-delta-closure-ref',
2435
('d2',): 'knit-delta-closure-ref',
2437
for record in generator.get_record_stream():
2438
self.assertEqual(kinds[record.key], record.storage_kind)
added
removed
bzrlib/tests/test_knit.py
