1
# Copyright (C) 2006-2011 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"""
33
from ..bzr.index import *
34
from ..bzr.knit import (
38
KnitDataStreamIncompatible,
39
KnitDataStreamUnknown,
41
KnitIndexUnknownMethod,
44
_VFContentMapGenerator,
50
from ..patiencediff import PatienceSequenceMatcher
55
from ..sixish import (
60
TestCaseWithMemoryTransport,
61
TestCaseWithTransport,
64
from ..bzr.versionedfile import (
67
network_bytes_to_kind_and_offset,
68
RecordingVersionedFilesDecorator,
75
compiled_knit_feature = features.ModuleAvailableFeature(
76
'breezy.bzr._knit_load_data_pyx')
79
class ErrorTests(TestCase):
81
def test_knit_data_stream_incompatible(self):
82
error = KnitDataStreamIncompatible(
83
'stream format', 'target format')
84
self.assertEqual('Cannot insert knit data stream of format '
85
'"stream format" into knit of format '
86
'"target format".', str(error))
88
def test_knit_data_stream_unknown(self):
89
error = KnitDataStreamUnknown(
91
self.assertEqual('Cannot parse knit data stream of format '
92
'"stream format".', str(error))
94
def test_knit_header_error(self):
95
error = KnitHeaderError('line foo\n', 'path/to/file')
96
self.assertEqual("Knit header error: 'line foo\\n' unexpected"
97
" for file \"path/to/file\".", str(error))
99
def test_knit_index_unknown_method(self):
100
error = KnitIndexUnknownMethod('http://host/foo.kndx',
102
self.assertEqual("Knit index http://host/foo.kndx does not have a"
103
" known method in options: ['bad', 'no-eol']",
107
class KnitContentTestsMixin(object):
109
def test_constructor(self):
110
content = self._make_content([])
113
content = self._make_content([])
114
self.assertEqual(content.text(), [])
116
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
117
self.assertEqual(content.text(), ["text1", "text2"])
120
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
121
copy = content.copy()
122
self.assertIsInstance(copy, content.__class__)
123
self.assertEqual(copy.annotate(), content.annotate())
125
def assertDerivedBlocksEqual(self, source, target, noeol=False):
126
"""Assert that the derived matching blocks match real output"""
127
source_lines = source.splitlines(True)
128
target_lines = target.splitlines(True)
130
if noeol and not line.endswith('\n'):
134
source_content = self._make_content([(None, nl(l)) for l in source_lines])
135
target_content = self._make_content([(None, nl(l)) for l in target_lines])
136
line_delta = source_content.line_delta(target_content)
137
delta_blocks = list(KnitContent.get_line_delta_blocks(line_delta,
138
source_lines, target_lines))
139
matcher = PatienceSequenceMatcher(None, source_lines, target_lines)
140
matcher_blocks = list(matcher.get_matching_blocks())
141
self.assertEqual(matcher_blocks, delta_blocks)
143
def test_get_line_delta_blocks(self):
144
self.assertDerivedBlocksEqual('a\nb\nc\n', 'q\nc\n')
145
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1)
146
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1A)
147
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1B)
148
self.assertDerivedBlocksEqual(TEXT_1B, TEXT_1A)
149
self.assertDerivedBlocksEqual(TEXT_1A, TEXT_1B)
150
self.assertDerivedBlocksEqual(TEXT_1A, '')
151
self.assertDerivedBlocksEqual('', TEXT_1A)
152
self.assertDerivedBlocksEqual('', '')
153
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd')
155
def test_get_line_delta_blocks_noeol(self):
156
"""Handle historical knit deltas safely
158
Some existing knit deltas don't consider the last line to differ
159
when the only difference whether it has a final newline.
161
New knit deltas appear to always consider the last line to differ
164
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd\n', noeol=True)
165
self.assertDerivedBlocksEqual('a\nb\nc\nd\n', 'a\nb\nc', noeol=True)
166
self.assertDerivedBlocksEqual('a\nb\nc\n', 'a\nb\nc', noeol=True)
167
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\n', noeol=True)
179
Banana cup cake recipe
189
Banana cup cake recipe
191
- bananas (do not use plantains!!!)
198
Banana cup cake recipe
215
class TestPlainKnitContent(TestCase, KnitContentTestsMixin):
217
def _make_content(self, lines):
218
annotated_content = AnnotatedKnitContent(lines)
219
return PlainKnitContent(annotated_content.text(), 'bogus')
221
def test_annotate(self):
222
content = self._make_content([])
223
self.assertEqual(content.annotate(), [])
225
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
226
self.assertEqual(content.annotate(),
227
[("bogus", "text1"), ("bogus", "text2")])
229
def test_line_delta(self):
230
content1 = self._make_content([("", "a"), ("", "b")])
231
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
232
self.assertEqual(content1.line_delta(content2),
233
[(1, 2, 2, ["a", "c"])])
235
def test_line_delta_iter(self):
236
content1 = self._make_content([("", "a"), ("", "b")])
237
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
238
it = content1.line_delta_iter(content2)
239
self.assertEqual(next(it), (1, 2, 2, ["a", "c"]))
240
self.assertRaises(StopIteration, next, it)
243
class TestAnnotatedKnitContent(TestCase, KnitContentTestsMixin):
245
def _make_content(self, lines):
246
return AnnotatedKnitContent(lines)
248
def test_annotate(self):
249
content = self._make_content([])
250
self.assertEqual(content.annotate(), [])
252
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
253
self.assertEqual(content.annotate(),
254
[("origin1", "text1"), ("origin2", "text2")])
256
def test_line_delta(self):
257
content1 = self._make_content([("", "a"), ("", "b")])
258
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
259
self.assertEqual(content1.line_delta(content2),
260
[(1, 2, 2, [("", "a"), ("", "c")])])
262
def test_line_delta_iter(self):
263
content1 = self._make_content([("", "a"), ("", "b")])
264
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
265
it = content1.line_delta_iter(content2)
266
self.assertEqual(next(it), (1, 2, 2, [("", "a"), ("", "c")]))
267
self.assertRaises(StopIteration, next, it)
270
class MockTransport(object):
272
def __init__(self, file_lines=None):
273
self.file_lines = file_lines
275
# We have no base directory for the MockTransport
278
def get(self, filename):
279
if self.file_lines is None:
280
raise errors.NoSuchFile(filename)
282
return BytesIO(b"\n".join(self.file_lines))
284
def readv(self, relpath, offsets):
285
fp = self.get(relpath)
286
for offset, size in offsets:
288
yield offset, fp.read(size)
290
def __getattr__(self, name):
291
def queue_call(*args, **kwargs):
292
self.calls.append((name, args, kwargs))
296
class MockReadvFailingTransport(MockTransport):
297
"""Fail in the middle of a readv() result.
299
This Transport will successfully yield the first two requested hunks, but
300
raise NoSuchFile for the rest.
303
def readv(self, relpath, offsets):
305
for result in MockTransport.readv(self, relpath, offsets):
307
# we use 2 because the first offset is the pack header, the second
308
# is the first actual content requset
310
raise errors.NoSuchFile(relpath)
314
class KnitRecordAccessTestsMixin(object):
315
"""Tests for getting and putting knit records."""
317
def test_add_raw_records(self):
318
"""Add_raw_records adds records retrievable later."""
319
access = self.get_access()
320
memos = access.add_raw_records([('key', 10)], '1234567890')
321
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
323
def test_add_several_raw_records(self):
324
"""add_raw_records with many records and read some back."""
325
access = self.get_access()
326
memos = access.add_raw_records([('key', 10), ('key2', 2), ('key3', 5)],
328
self.assertEqual(['1234567890', '12', '34567'],
329
list(access.get_raw_records(memos)))
330
self.assertEqual(['1234567890'],
331
list(access.get_raw_records(memos[0:1])))
332
self.assertEqual(['12'],
333
list(access.get_raw_records(memos[1:2])))
334
self.assertEqual(['34567'],
335
list(access.get_raw_records(memos[2:3])))
336
self.assertEqual(['1234567890', '34567'],
337
list(access.get_raw_records(memos[0:1] + memos[2:3])))
340
class TestKnitKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
341
"""Tests for the .kndx implementation."""
343
def get_access(self):
344
"""Get a .knit style access instance."""
345
mapper = ConstantMapper("foo")
346
access = _KnitKeyAccess(self.get_transport(), mapper)
350
class _TestException(Exception):
351
"""Just an exception for local tests to use."""
354
class TestPackKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
355
"""Tests for the pack based access."""
357
def get_access(self):
358
return self._get_access()[0]
360
def _get_access(self, packname='packfile', index='FOO'):
361
transport = self.get_transport()
362
def write_data(bytes):
363
transport.append_bytes(packname, bytes)
364
writer = pack.ContainerWriter(write_data)
366
access = pack_repo._DirectPackAccess({})
367
access.set_writer(writer, index, (transport, packname))
368
return access, writer
370
def make_pack_file(self):
371
"""Create a pack file with 2 records."""
372
access, writer = self._get_access(packname='packname', index='foo')
374
memos.extend(access.add_raw_records([('key1', 10)], '1234567890'))
375
memos.extend(access.add_raw_records([('key2', 5)], '12345'))
379
def test_pack_collection_pack_retries(self):
380
"""An explicit pack of a pack collection succeeds even when a
381
concurrent pack happens.
383
builder = self.make_branch_builder('.')
384
builder.start_series()
385
builder.build_snapshot('rev-1', None, [
386
('add', ('', 'root-id', 'directory', None)),
387
('add', ('file', 'file-id', 'file', 'content\nrev 1\n')),
389
builder.build_snapshot('rev-2', ['rev-1'], [
390
('modify', ('file-id', 'content\nrev 2\n')),
392
builder.build_snapshot('rev-3', ['rev-2'], [
393
('modify', ('file-id', 'content\nrev 3\n')),
395
self.addCleanup(builder.finish_series)
396
b = builder.get_branch()
397
self.addCleanup(b.lock_write().unlock)
399
collection = repo._pack_collection
400
# Concurrently repack the repo.
401
reopened_repo = repo.controldir.open_repository()
406
def make_vf_for_retrying(self):
407
"""Create 3 packs and a reload function.
409
Originally, 2 pack files will have the data, but one will be missing.
410
And then the third will be used in place of the first two if reload()
413
:return: (versioned_file, reload_counter)
414
versioned_file a KnitVersionedFiles using the packs for access
416
builder = self.make_branch_builder('.', format="1.9")
417
builder.start_series()
418
builder.build_snapshot('rev-1', None, [
419
('add', ('', 'root-id', 'directory', None)),
420
('add', ('file', 'file-id', 'file', 'content\nrev 1\n')),
422
builder.build_snapshot('rev-2', ['rev-1'], [
423
('modify', ('file-id', 'content\nrev 2\n')),
425
builder.build_snapshot('rev-3', ['rev-2'], [
426
('modify', ('file-id', 'content\nrev 3\n')),
428
builder.finish_series()
429
b = builder.get_branch()
431
self.addCleanup(b.unlock)
432
# Pack these three revisions into another pack file, but don't remove
435
collection = repo._pack_collection
436
collection.ensure_loaded()
437
orig_packs = collection.packs
438
packer = knitpack_repo.KnitPacker(collection, orig_packs, '.testpack')
439
new_pack = packer.pack()
440
# forget about the new pack
444
# Set up a reload() function that switches to using the new pack file
445
new_index = new_pack.revision_index
446
access_tuple = new_pack.access_tuple()
447
reload_counter = [0, 0, 0]
449
reload_counter[0] += 1
450
if reload_counter[1] > 0:
451
# We already reloaded, nothing more to do
452
reload_counter[2] += 1
454
reload_counter[1] += 1
455
vf._index._graph_index._indices[:] = [new_index]
456
vf._access._indices.clear()
457
vf._access._indices[new_index] = access_tuple
459
# Delete one of the pack files so the data will need to be reloaded. We
460
# will delete the file with 'rev-2' in it
461
trans, name = orig_packs[1].access_tuple()
463
# We don't have the index trigger reloading because we want to test
464
# that we reload when the .pack disappears
465
vf._access._reload_func = reload
466
return vf, reload_counter
468
def make_reload_func(self, return_val=True):
471
reload_called[0] += 1
473
return reload_called, reload
475
def make_retry_exception(self):
476
# We raise a real exception so that sys.exc_info() is properly
479
raise _TestException('foobar')
480
except _TestException as e:
481
retry_exc = errors.RetryWithNewPacks(None, reload_occurred=False,
482
exc_info=sys.exc_info())
483
# GZ 2010-08-10: Cycle with exc_info affects 3 tests
486
def test_read_from_several_packs(self):
487
access, writer = self._get_access()
489
memos.extend(access.add_raw_records([('key', 10)], '1234567890'))
491
access, writer = self._get_access('pack2', 'FOOBAR')
492
memos.extend(access.add_raw_records([('key', 5)], '12345'))
494
access, writer = self._get_access('pack3', 'BAZ')
495
memos.extend(access.add_raw_records([('key', 5)], 'alpha'))
497
transport = self.get_transport()
498
access = pack_repo._DirectPackAccess({"FOO":(transport, 'packfile'),
499
"FOOBAR":(transport, 'pack2'),
500
"BAZ":(transport, 'pack3')})
501
self.assertEqual(['1234567890', '12345', 'alpha'],
502
list(access.get_raw_records(memos)))
503
self.assertEqual(['1234567890'],
504
list(access.get_raw_records(memos[0:1])))
505
self.assertEqual(['12345'],
506
list(access.get_raw_records(memos[1:2])))
507
self.assertEqual(['alpha'],
508
list(access.get_raw_records(memos[2:3])))
509
self.assertEqual(['1234567890', 'alpha'],
510
list(access.get_raw_records(memos[0:1] + memos[2:3])))
512
def test_set_writer(self):
513
"""The writer should be settable post construction."""
514
access = pack_repo._DirectPackAccess({})
515
transport = self.get_transport()
516
packname = 'packfile'
518
def write_data(bytes):
519
transport.append_bytes(packname, bytes)
520
writer = pack.ContainerWriter(write_data)
522
access.set_writer(writer, index, (transport, packname))
523
memos = access.add_raw_records([('key', 10)], '1234567890')
525
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
527
def test_missing_index_raises_retry(self):
528
memos = self.make_pack_file()
529
transport = self.get_transport()
530
reload_called, reload_func = self.make_reload_func()
531
# Note that the index key has changed from 'foo' to 'bar'
532
access = pack_repo._DirectPackAccess({'bar':(transport, 'packname')},
533
reload_func=reload_func)
534
e = self.assertListRaises(errors.RetryWithNewPacks,
535
access.get_raw_records, memos)
536
# Because a key was passed in which does not match our index list, we
537
# assume that the listing was already reloaded
538
self.assertTrue(e.reload_occurred)
539
self.assertIsInstance(e.exc_info, tuple)
540
self.assertIs(e.exc_info[0], KeyError)
541
self.assertIsInstance(e.exc_info[1], KeyError)
543
def test_missing_index_raises_key_error_with_no_reload(self):
544
memos = self.make_pack_file()
545
transport = self.get_transport()
546
# Note that the index key has changed from 'foo' to 'bar'
547
access = pack_repo._DirectPackAccess({'bar':(transport, 'packname')})
548
e = self.assertListRaises(KeyError, access.get_raw_records, memos)
550
def test_missing_file_raises_retry(self):
551
memos = self.make_pack_file()
552
transport = self.get_transport()
553
reload_called, reload_func = self.make_reload_func()
554
# Note that the 'filename' has been changed to 'different-packname'
555
access = pack_repo._DirectPackAccess(
556
{'foo':(transport, 'different-packname')},
557
reload_func=reload_func)
558
e = self.assertListRaises(errors.RetryWithNewPacks,
559
access.get_raw_records, memos)
560
# The file has gone missing, so we assume we need to reload
561
self.assertFalse(e.reload_occurred)
562
self.assertIsInstance(e.exc_info, tuple)
563
self.assertIs(e.exc_info[0], errors.NoSuchFile)
564
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
565
self.assertEqual('different-packname', e.exc_info[1].path)
567
def test_missing_file_raises_no_such_file_with_no_reload(self):
568
memos = self.make_pack_file()
569
transport = self.get_transport()
570
# Note that the 'filename' has been changed to 'different-packname'
571
access = pack_repo._DirectPackAccess(
572
{'foo': (transport, 'different-packname')})
573
e = self.assertListRaises(errors.NoSuchFile,
574
access.get_raw_records, memos)
576
def test_failing_readv_raises_retry(self):
577
memos = self.make_pack_file()
578
transport = self.get_transport()
579
failing_transport = MockReadvFailingTransport(
580
[transport.get_bytes('packname')])
581
reload_called, reload_func = self.make_reload_func()
582
access = pack_repo._DirectPackAccess(
583
{'foo': (failing_transport, 'packname')},
584
reload_func=reload_func)
585
# Asking for a single record will not trigger the Mock failure
586
self.assertEqual(['1234567890'],
587
list(access.get_raw_records(memos[:1])))
588
self.assertEqual(['12345'],
589
list(access.get_raw_records(memos[1:2])))
590
# A multiple offset readv() will fail mid-way through
591
e = self.assertListRaises(errors.RetryWithNewPacks,
592
access.get_raw_records, memos)
593
# The file has gone missing, so we assume we need to reload
594
self.assertFalse(e.reload_occurred)
595
self.assertIsInstance(e.exc_info, tuple)
596
self.assertIs(e.exc_info[0], errors.NoSuchFile)
597
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
598
self.assertEqual('packname', e.exc_info[1].path)
600
def test_failing_readv_raises_no_such_file_with_no_reload(self):
601
memos = self.make_pack_file()
602
transport = self.get_transport()
603
failing_transport = MockReadvFailingTransport(
604
[transport.get_bytes('packname')])
605
reload_called, reload_func = self.make_reload_func()
606
access = pack_repo._DirectPackAccess(
607
{'foo':(failing_transport, 'packname')})
608
# Asking for a single record will not trigger the Mock failure
609
self.assertEqual(['1234567890'],
610
list(access.get_raw_records(memos[:1])))
611
self.assertEqual(['12345'],
612
list(access.get_raw_records(memos[1:2])))
613
# A multiple offset readv() will fail mid-way through
614
e = self.assertListRaises(errors.NoSuchFile,
615
access.get_raw_records, memos)
617
def test_reload_or_raise_no_reload(self):
618
access = pack_repo._DirectPackAccess({}, reload_func=None)
619
retry_exc = self.make_retry_exception()
620
# Without a reload_func, we will just re-raise the original exception
621
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
623
def test_reload_or_raise_reload_changed(self):
624
reload_called, reload_func = self.make_reload_func(return_val=True)
625
access = pack_repo._DirectPackAccess({}, reload_func=reload_func)
626
retry_exc = self.make_retry_exception()
627
access.reload_or_raise(retry_exc)
628
self.assertEqual([1], reload_called)
629
retry_exc.reload_occurred=True
630
access.reload_or_raise(retry_exc)
631
self.assertEqual([2], reload_called)
633
def test_reload_or_raise_reload_no_change(self):
634
reload_called, reload_func = self.make_reload_func(return_val=False)
635
access = pack_repo._DirectPackAccess({}, reload_func=reload_func)
636
retry_exc = self.make_retry_exception()
637
# If reload_occurred is False, then we consider it an error to have
638
# reload_func() return False (no changes).
639
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
640
self.assertEqual([1], reload_called)
641
retry_exc.reload_occurred=True
642
# If reload_occurred is True, then we assume nothing changed because
643
# it had changed earlier, but didn't change again
644
access.reload_or_raise(retry_exc)
645
self.assertEqual([2], reload_called)
647
def test_annotate_retries(self):
648
vf, reload_counter = self.make_vf_for_retrying()
649
# It is a little bit bogus to annotate the Revision VF, but it works,
650
# as we have ancestry stored there
652
reload_lines = vf.annotate(key)
653
self.assertEqual([1, 1, 0], reload_counter)
654
plain_lines = vf.annotate(key)
655
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
656
if reload_lines != plain_lines:
657
self.fail('Annotation was not identical with reloading.')
658
# Now delete the packs-in-use, which should trigger another reload, but
659
# this time we just raise an exception because we can't recover
660
for trans, name in vf._access._indices.values():
662
self.assertRaises(errors.NoSuchFile, vf.annotate, key)
663
self.assertEqual([2, 1, 1], reload_counter)
665
def test__get_record_map_retries(self):
666
vf, reload_counter = self.make_vf_for_retrying()
667
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
668
records = vf._get_record_map(keys)
669
self.assertEqual(keys, sorted(records.keys()))
670
self.assertEqual([1, 1, 0], reload_counter)
671
# Now delete the packs-in-use, which should trigger another reload, but
672
# this time we just raise an exception because we can't recover
673
for trans, name in vf._access._indices.values():
675
self.assertRaises(errors.NoSuchFile, vf._get_record_map, keys)
676
self.assertEqual([2, 1, 1], reload_counter)
678
def test_get_record_stream_retries(self):
679
vf, reload_counter = self.make_vf_for_retrying()
680
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
681
record_stream = vf.get_record_stream(keys, 'topological', False)
682
record = next(record_stream)
683
self.assertEqual(('rev-1',), record.key)
684
self.assertEqual([0, 0, 0], reload_counter)
685
record = next(record_stream)
686
self.assertEqual(('rev-2',), record.key)
687
self.assertEqual([1, 1, 0], reload_counter)
688
record = next(record_stream)
689
self.assertEqual(('rev-3',), record.key)
690
self.assertEqual([1, 1, 0], reload_counter)
691
# Now delete all pack files, and see that we raise the right error
692
for trans, name in vf._access._indices.values():
694
self.assertListRaises(errors.NoSuchFile,
695
vf.get_record_stream, keys, 'topological', False)
697
def test_iter_lines_added_or_present_in_keys_retries(self):
698
vf, reload_counter = self.make_vf_for_retrying()
699
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
700
# Unfortunately, iter_lines_added_or_present_in_keys iterates the
701
# result in random order (determined by the iteration order from a
702
# set()), so we don't have any solid way to trigger whether data is
703
# read before or after. However we tried to delete the middle node to
704
# exercise the code well.
705
# What we care about is that all lines are always yielded, but not
708
reload_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
709
self.assertEqual([1, 1, 0], reload_counter)
710
# Now do it again, to make sure the result is equivalent
711
plain_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
712
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
713
self.assertEqual(plain_lines, reload_lines)
714
self.assertEqual(21, len(plain_lines))
715
# Now delete all pack files, and see that we raise the right error
716
for trans, name in vf._access._indices.values():
718
self.assertListRaises(errors.NoSuchFile,
719
vf.iter_lines_added_or_present_in_keys, keys)
720
self.assertEqual([2, 1, 1], reload_counter)
722
def test_get_record_stream_yields_disk_sorted_order(self):
723
# if we get 'unordered' pick a semi-optimal order for reading. The
724
# order should be grouped by pack file, and then by position in file
725
repo = self.make_repository('test', format='pack-0.92')
727
self.addCleanup(repo.unlock)
728
repo.start_write_group()
730
vf.add_lines(('f-id', 'rev-5'), [('f-id', 'rev-4')], ['lines\n'])
731
vf.add_lines(('f-id', 'rev-1'), [], ['lines\n'])
732
vf.add_lines(('f-id', 'rev-2'), [('f-id', 'rev-1')], ['lines\n'])
733
repo.commit_write_group()
734
# We inserted them as rev-5, rev-1, rev-2, we should get them back in
736
stream = vf.get_record_stream([('f-id', 'rev-1'), ('f-id', 'rev-5'),
737
('f-id', 'rev-2')], 'unordered', False)
738
keys = [r.key for r in stream]
739
self.assertEqual([('f-id', 'rev-5'), ('f-id', 'rev-1'),
740
('f-id', 'rev-2')], keys)
741
repo.start_write_group()
742
vf.add_lines(('f-id', 'rev-4'), [('f-id', 'rev-3')], ['lines\n'])
743
vf.add_lines(('f-id', 'rev-3'), [('f-id', 'rev-2')], ['lines\n'])
744
vf.add_lines(('f-id', 'rev-6'), [('f-id', 'rev-5')], ['lines\n'])
745
repo.commit_write_group()
746
# Request in random order, to make sure the output order isn't based on
748
request_keys = set(('f-id', 'rev-%d' % i) for i in range(1, 7))
749
stream = vf.get_record_stream(request_keys, 'unordered', False)
750
keys = [r.key for r in stream]
751
# We want to get the keys back in disk order, but it doesn't matter
752
# which pack we read from first. So this can come back in 2 orders
753
alt1 = [('f-id', 'rev-%d' % i) for i in [4, 3, 6, 5, 1, 2]]
754
alt2 = [('f-id', 'rev-%d' % i) for i in [5, 1, 2, 4, 3, 6]]
755
if keys != alt1 and keys != alt2:
756
self.fail('Returned key order did not match either expected order.'
757
' expected %s or %s, not %s'
758
% (alt1, alt2, keys))
761
class LowLevelKnitDataTests(TestCase):
763
def create_gz_content(self, text):
765
gz_file = gzip.GzipFile(mode='wb', fileobj=sio)
768
return sio.getvalue()
770
def make_multiple_records(self):
771
"""Create the content for multiple records."""
772
sha1sum = osutils.sha_string('foo\nbar\n')
774
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
779
record_1 = (0, len(gz_txt), sha1sum)
780
total_txt.append(gz_txt)
781
sha1sum = osutils.sha_string('baz\n')
782
gz_txt = self.create_gz_content('version rev-id-2 1 %s\n'
786
record_2 = (record_1[1], len(gz_txt), sha1sum)
787
total_txt.append(gz_txt)
788
return total_txt, record_1, record_2
790
def test_valid_knit_data(self):
791
sha1sum = osutils.sha_string('foo\nbar\n')
792
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
797
transport = MockTransport([gz_txt])
798
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
799
knit = KnitVersionedFiles(None, access)
800
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
802
contents = list(knit._read_records_iter(records))
803
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'],
804
'4e48e2c9a3d2ca8a708cb0cc545700544efb5021')], contents)
806
raw_contents = list(knit._read_records_iter_raw(records))
807
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
809
def test_multiple_records_valid(self):
810
total_txt, record_1, record_2 = self.make_multiple_records()
811
transport = MockTransport([''.join(total_txt)])
812
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
813
knit = KnitVersionedFiles(None, access)
814
records = [(('rev-id-1',), (('rev-id-1',), record_1[0], record_1[1])),
815
(('rev-id-2',), (('rev-id-2',), record_2[0], record_2[1]))]
817
contents = list(knit._read_records_iter(records))
818
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'], record_1[2]),
819
(('rev-id-2',), ['baz\n'], record_2[2])],
822
raw_contents = list(knit._read_records_iter_raw(records))
823
self.assertEqual([(('rev-id-1',), total_txt[0], record_1[2]),
824
(('rev-id-2',), total_txt[1], record_2[2])],
827
def test_not_enough_lines(self):
828
sha1sum = osutils.sha_string('foo\n')
829
# record says 2 lines data says 1
830
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
834
transport = MockTransport([gz_txt])
835
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
836
knit = KnitVersionedFiles(None, access)
837
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
838
self.assertRaises(KnitCorrupt, list,
839
knit._read_records_iter(records))
841
# read_records_iter_raw won't detect that sort of mismatch/corruption
842
raw_contents = list(knit._read_records_iter_raw(records))
843
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
845
def test_too_many_lines(self):
846
sha1sum = osutils.sha_string('foo\nbar\n')
847
# record says 1 lines data says 2
848
gz_txt = self.create_gz_content('version rev-id-1 1 %s\n'
853
transport = MockTransport([gz_txt])
854
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
855
knit = KnitVersionedFiles(None, access)
856
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
857
self.assertRaises(KnitCorrupt, list,
858
knit._read_records_iter(records))
860
# read_records_iter_raw won't detect that sort of mismatch/corruption
861
raw_contents = list(knit._read_records_iter_raw(records))
862
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
864
def test_mismatched_version_id(self):
865
sha1sum = osutils.sha_string('foo\nbar\n')
866
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
871
transport = MockTransport([gz_txt])
872
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
873
knit = KnitVersionedFiles(None, access)
874
# We are asking for rev-id-2, but the data is rev-id-1
875
records = [(('rev-id-2',), (('rev-id-2',), 0, len(gz_txt)))]
876
self.assertRaises(KnitCorrupt, list,
877
knit._read_records_iter(records))
879
# read_records_iter_raw detects mismatches in the header
880
self.assertRaises(KnitCorrupt, list,
881
knit._read_records_iter_raw(records))
883
def test_uncompressed_data(self):
884
sha1sum = osutils.sha_string('foo\nbar\n')
885
txt = ('version rev-id-1 2 %s\n'
890
transport = MockTransport([txt])
891
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
892
knit = KnitVersionedFiles(None, access)
893
records = [(('rev-id-1',), (('rev-id-1',), 0, len(txt)))]
895
# We don't have valid gzip data ==> corrupt
896
self.assertRaises(KnitCorrupt, list,
897
knit._read_records_iter(records))
899
# read_records_iter_raw will notice the bad data
900
self.assertRaises(KnitCorrupt, list,
901
knit._read_records_iter_raw(records))
903
def test_corrupted_data(self):
904
sha1sum = osutils.sha_string('foo\nbar\n')
905
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
910
# Change 2 bytes in the middle to \xff
911
gz_txt = gz_txt[:10] + '\xff\xff' + gz_txt[12:]
912
transport = MockTransport([gz_txt])
913
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
914
knit = KnitVersionedFiles(None, access)
915
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
916
self.assertRaises(KnitCorrupt, list,
917
knit._read_records_iter(records))
918
# read_records_iter_raw will barf on bad gz data
919
self.assertRaises(KnitCorrupt, list,
920
knit._read_records_iter_raw(records))
923
class LowLevelKnitIndexTests(TestCase):
925
def get_knit_index(self, transport, name, mode):
926
mapper = ConstantMapper(name)
927
from ..bzr._knit_load_data_py import _load_data_py
928
self.overrideAttr(knit, '_load_data', _load_data_py)
929
allow_writes = lambda: 'w' in mode
930
return _KndxIndex(transport, mapper, lambda:None, allow_writes, lambda:True)
932
def test_create_file(self):
933
transport = MockTransport()
934
index = self.get_knit_index(transport, "filename", "w")
936
call = transport.calls.pop(0)
937
# call[1][1] is a BytesIO - we can't test it by simple equality.
938
self.assertEqual('put_file_non_atomic', call[0])
939
self.assertEqual('filename.kndx', call[1][0])
940
# With no history, _KndxIndex writes a new index:
941
self.assertEqual(_KndxIndex.HEADER,
942
call[1][1].getvalue())
943
self.assertEqual({'create_parent_dir': True}, call[2])
945
def test_read_utf8_version_id(self):
946
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
947
utf8_revision_id = unicode_revision_id.encode('utf-8')
948
transport = MockTransport([
950
'%s option 0 1 :' % (utf8_revision_id,)
952
index = self.get_knit_index(transport, "filename", "r")
953
# _KndxIndex is a private class, and deals in utf8 revision_ids, not
954
# Unicode revision_ids.
955
self.assertEqual({(utf8_revision_id,):()},
956
index.get_parent_map(index.keys()))
957
self.assertFalse((unicode_revision_id,) in index.keys())
959
def test_read_utf8_parents(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([
964
"version option 0 1 .%s :" % (utf8_revision_id,)
966
index = self.get_knit_index(transport, "filename", "r")
967
self.assertEqual({("version",):((utf8_revision_id,),)},
968
index.get_parent_map(index.keys()))
970
def test_read_ignore_corrupted_lines(self):
971
transport = MockTransport([
974
"corrupted options 0 1 .b .c ",
975
"version options 0 1 :"
977
index = self.get_knit_index(transport, "filename", "r")
978
self.assertEqual(1, len(index.keys()))
979
self.assertEqual({("version",)}, index.keys())
981
def test_read_corrupted_header(self):
982
transport = MockTransport(['not a bzr knit index header\n'])
983
index = self.get_knit_index(transport, "filename", "r")
984
self.assertRaises(KnitHeaderError, index.keys)
986
def test_read_duplicate_entries(self):
987
transport = MockTransport([
989
"parent options 0 1 :",
990
"version options1 0 1 0 :",
991
"version options2 1 2 .other :",
992
"version options3 3 4 0 .other :"
994
index = self.get_knit_index(transport, "filename", "r")
995
self.assertEqual(2, len(index.keys()))
996
# check that the index used is the first one written. (Specific
997
# to KnitIndex style indices.
998
self.assertEqual("1", index._dictionary_compress([("version",)]))
999
self.assertEqual((("version",), 3, 4), index.get_position(("version",)))
1000
self.assertEqual(["options3"], index.get_options(("version",)))
1001
self.assertEqual({("version",):(("parent",), ("other",))},
1002
index.get_parent_map([("version",)]))
1004
def test_read_compressed_parents(self):
1005
transport = MockTransport([
1009
"c option 0 1 1 0 :",
1011
index = self.get_knit_index(transport, "filename", "r")
1012
self.assertEqual({("b",):(("a",),), ("c",):(("b",), ("a",))},
1013
index.get_parent_map([("b",), ("c",)]))
1015
def test_write_utf8_version_id(self):
1016
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
1017
utf8_revision_id = unicode_revision_id.encode('utf-8')
1018
transport = MockTransport([
1021
index = self.get_knit_index(transport, "filename", "r")
1023
((utf8_revision_id,), ["option"], ((utf8_revision_id,), 0, 1), [])])
1024
call = transport.calls.pop(0)
1025
# call[1][1] is a BytesIO - we can't test it by simple equality.
1026
self.assertEqual('put_file_non_atomic', call[0])
1027
self.assertEqual('filename.kndx', call[1][0])
1028
# With no history, _KndxIndex writes a new index:
1029
self.assertEqual(_KndxIndex.HEADER +
1030
"\n%s option 0 1 :" % (utf8_revision_id,),
1031
call[1][1].getvalue())
1032
self.assertEqual({'create_parent_dir': True}, call[2])
1034
def test_write_utf8_parents(self):
1035
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
1036
utf8_revision_id = unicode_revision_id.encode('utf-8')
1037
transport = MockTransport([
1040
index = self.get_knit_index(transport, "filename", "r")
1042
(("version",), ["option"], (("version",), 0, 1), [(utf8_revision_id,)])])
1043
call = transport.calls.pop(0)
1044
# call[1][1] is a BytesIO - 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:
1048
self.assertEqual(_KndxIndex.HEADER +
1049
"\nversion option 0 1 .%s :" % (utf8_revision_id,),
1050
call[1][1].getvalue())
1051
self.assertEqual({'create_parent_dir': True}, call[2])
1053
def test_keys(self):
1054
transport = MockTransport([
1057
index = self.get_knit_index(transport, "filename", "r")
1059
self.assertEqual(set(), index.keys())
1061
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1062
self.assertEqual({("a",)}, index.keys())
1064
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1065
self.assertEqual({("a",)}, index.keys())
1067
index.add_records([(("b",), ["option"], (("b",), 0, 1), [])])
1068
self.assertEqual({("a",), ("b",)}, index.keys())
1070
def add_a_b(self, index, random_id=None):
1072
if random_id is not None:
1073
kwargs["random_id"] = random_id
1075
(("a",), ["option"], (("a",), 0, 1), [("b",)]),
1076
(("a",), ["opt"], (("a",), 1, 2), [("c",)]),
1077
(("b",), ["option"], (("b",), 2, 3), [("a",)])
1080
def assertIndexIsAB(self, index):
1085
index.get_parent_map(index.keys()))
1086
self.assertEqual((("a",), 1, 2), index.get_position(("a",)))
1087
self.assertEqual((("b",), 2, 3), index.get_position(("b",)))
1088
self.assertEqual(["opt"], index.get_options(("a",)))
1090
def test_add_versions(self):
1091
transport = MockTransport([
1094
index = self.get_knit_index(transport, "filename", "r")
1097
call = transport.calls.pop(0)
1098
# call[1][1] is a BytesIO - we can't test it by simple equality.
1099
self.assertEqual('put_file_non_atomic', call[0])
1100
self.assertEqual('filename.kndx', call[1][0])
1101
# With no history, _KndxIndex writes a new index:
1104
"\na option 0 1 .b :"
1106
"\nb option 2 3 0 :",
1107
call[1][1].getvalue())
1108
self.assertEqual({'create_parent_dir': True}, call[2])
1109
self.assertIndexIsAB(index)
1111
def test_add_versions_random_id_is_accepted(self):
1112
transport = MockTransport([
1115
index = self.get_knit_index(transport, "filename", "r")
1116
self.add_a_b(index, random_id=True)
1118
def test_delay_create_and_add_versions(self):
1119
transport = MockTransport()
1121
index = self.get_knit_index(transport, "filename", "w")
1123
self.assertEqual([], transport.calls)
1126
#[ {"dir_mode": 0777, "create_parent_dir": True, "mode": "wb"},
1128
# Two calls: one during which we load the existing index (and when its
1129
# missing create it), then a second where we write the contents out.
1130
self.assertEqual(2, len(transport.calls))
1131
call = transport.calls.pop(0)
1132
self.assertEqual('put_file_non_atomic', call[0])
1133
self.assertEqual('filename.kndx', call[1][0])
1134
# With no history, _KndxIndex writes a new index:
1135
self.assertEqual(_KndxIndex.HEADER, call[1][1].getvalue())
1136
self.assertEqual({'create_parent_dir': True}, call[2])
1137
call = transport.calls.pop(0)
1138
# call[1][1] is a BytesIO - we can't test it by simple equality.
1139
self.assertEqual('put_file_non_atomic', call[0])
1140
self.assertEqual('filename.kndx', call[1][0])
1141
# With no history, _KndxIndex writes a new index:
1144
"\na option 0 1 .b :"
1146
"\nb option 2 3 0 :",
1147
call[1][1].getvalue())
1148
self.assertEqual({'create_parent_dir': True}, call[2])
1150
def assertTotalBuildSize(self, size, keys, positions):
1151
self.assertEqual(size,
1152
knit._get_total_build_size(None, keys, positions))
1154
def test__get_total_build_size(self):
1156
('a',): (('fulltext', False), (('a',), 0, 100), None),
1157
('b',): (('line-delta', False), (('b',), 100, 21), ('a',)),
1158
('c',): (('line-delta', False), (('c',), 121, 35), ('b',)),
1159
('d',): (('line-delta', False), (('d',), 156, 12), ('b',)),
1161
self.assertTotalBuildSize(100, [('a',)], positions)
1162
self.assertTotalBuildSize(121, [('b',)], positions)
1163
# c needs both a & b
1164
self.assertTotalBuildSize(156, [('c',)], positions)
1165
# we shouldn't count 'b' twice
1166
self.assertTotalBuildSize(156, [('b',), ('c',)], positions)
1167
self.assertTotalBuildSize(133, [('d',)], positions)
1168
self.assertTotalBuildSize(168, [('c',), ('d',)], positions)
1170
def test_get_position(self):
1171
transport = MockTransport([
1176
index = self.get_knit_index(transport, "filename", "r")
1178
self.assertEqual((("a",), 0, 1), index.get_position(("a",)))
1179
self.assertEqual((("b",), 1, 2), index.get_position(("b",)))
1181
def test_get_method(self):
1182
transport = MockTransport([
1184
"a fulltext,unknown 0 1 :",
1185
"b unknown,line-delta 1 2 :",
1188
index = self.get_knit_index(transport, "filename", "r")
1190
self.assertEqual("fulltext", index.get_method("a"))
1191
self.assertEqual("line-delta", index.get_method("b"))
1192
self.assertRaises(knit.KnitIndexUnknownMethod, index.get_method, "c")
1194
def test_get_options(self):
1195
transport = MockTransport([
1200
index = self.get_knit_index(transport, "filename", "r")
1202
self.assertEqual(["opt1"], index.get_options("a"))
1203
self.assertEqual(["opt2", "opt3"], index.get_options("b"))
1205
def test_get_parent_map(self):
1206
transport = MockTransport([
1209
"b option 1 2 0 .c :",
1210
"c option 1 2 1 0 .e :"
1212
index = self.get_knit_index(transport, "filename", "r")
1216
("b",):(("a",), ("c",)),
1217
("c",):(("b",), ("a",), ("e",)),
1218
}, index.get_parent_map(index.keys()))
1220
def test_impossible_parent(self):
1221
"""Test we get KnitCorrupt if the parent couldn't possibly exist."""
1222
transport = MockTransport([
1225
"b option 0 1 4 :" # We don't have a 4th record
1227
index = self.get_knit_index(transport, 'filename', 'r')
1228
self.assertRaises(KnitCorrupt, index.keys)
1230
def test_corrupted_parent(self):
1231
transport = MockTransport([
1235
"c option 0 1 1v :", # Can't have a parent of '1v'
1237
index = self.get_knit_index(transport, 'filename', 'r')
1238
self.assertRaises(KnitCorrupt, index.keys)
1240
def test_corrupted_parent_in_list(self):
1241
transport = MockTransport([
1245
"c option 0 1 1 v :", # Can't have a parent of 'v'
1247
index = self.get_knit_index(transport, 'filename', 'r')
1248
self.assertRaises(KnitCorrupt, index.keys)
1250
def test_invalid_position(self):
1251
transport = MockTransport([
1255
index = self.get_knit_index(transport, 'filename', 'r')
1256
self.assertRaises(KnitCorrupt, index.keys)
1258
def test_invalid_size(self):
1259
transport = MockTransport([
1263
index = self.get_knit_index(transport, 'filename', 'r')
1264
self.assertRaises(KnitCorrupt, index.keys)
1266
def test_scan_unvalidated_index_not_implemented(self):
1267
transport = MockTransport()
1268
index = self.get_knit_index(transport, 'filename', 'r')
1270
NotImplementedError, index.scan_unvalidated_index,
1271
'dummy graph_index')
1273
NotImplementedError, index.get_missing_compression_parents)
1275
def test_short_line(self):
1276
transport = MockTransport([
1279
"b option 10 10 0", # This line isn't terminated, ignored
1281
index = self.get_knit_index(transport, "filename", "r")
1282
self.assertEqual({('a',)}, index.keys())
1284
def test_skip_incomplete_record(self):
1285
# A line with bogus data should just be skipped
1286
transport = MockTransport([
1289
"b option 10 10 0", # This line isn't terminated, ignored
1290
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1292
index = self.get_knit_index(transport, "filename", "r")
1293
self.assertEqual({('a',), ('c',)}, index.keys())
1295
def test_trailing_characters(self):
1296
# A line with bogus data should just be skipped
1297
transport = MockTransport([
1300
"b option 10 10 0 :a", # This line has extra trailing characters
1301
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1303
index = self.get_knit_index(transport, "filename", "r")
1304
self.assertEqual({('a',), ('c',)}, index.keys())
1307
class LowLevelKnitIndexTests_c(LowLevelKnitIndexTests):
1309
_test_needs_features = [compiled_knit_feature]
1311
def get_knit_index(self, transport, name, mode):
1312
mapper = ConstantMapper(name)
1313
from ..bzr._knit_load_data_pyx import _load_data_c
1314
self.overrideAttr(knit, '_load_data', _load_data_c)
1315
allow_writes = lambda: mode == 'w'
1316
return _KndxIndex(transport, mapper, lambda:None,
1317
allow_writes, lambda:True)
1320
class Test_KnitAnnotator(TestCaseWithMemoryTransport):
1322
def make_annotator(self):
1323
factory = knit.make_pack_factory(True, True, 1)
1324
vf = factory(self.get_transport())
1325
return knit._KnitAnnotator(vf)
1327
def test__expand_fulltext(self):
1328
ann = self.make_annotator()
1329
rev_key = ('rev-id',)
1330
ann._num_compression_children[rev_key] = 1
1331
res = ann._expand_record(rev_key, (('parent-id',),), None,
1332
['line1\n', 'line2\n'], ('fulltext', True))
1333
# The content object and text lines should be cached appropriately
1334
self.assertEqual(['line1\n', 'line2'], res)
1335
content_obj = ann._content_objects[rev_key]
1336
self.assertEqual(['line1\n', 'line2\n'], content_obj._lines)
1337
self.assertEqual(res, content_obj.text())
1338
self.assertEqual(res, ann._text_cache[rev_key])
1340
def test__expand_delta_comp_parent_not_available(self):
1341
# Parent isn't available yet, so we return nothing, but queue up this
1342
# node for later processing
1343
ann = self.make_annotator()
1344
rev_key = ('rev-id',)
1345
parent_key = ('parent-id',)
1346
record = ['0,1,1\n', 'new-line\n']
1347
details = ('line-delta', False)
1348
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1350
self.assertEqual(None, res)
1351
self.assertTrue(parent_key in ann._pending_deltas)
1352
pending = ann._pending_deltas[parent_key]
1353
self.assertEqual(1, len(pending))
1354
self.assertEqual((rev_key, (parent_key,), record, details), pending[0])
1356
def test__expand_record_tracks_num_children(self):
1357
ann = self.make_annotator()
1358
rev_key = ('rev-id',)
1359
rev2_key = ('rev2-id',)
1360
parent_key = ('parent-id',)
1361
record = ['0,1,1\n', 'new-line\n']
1362
details = ('line-delta', False)
1363
ann._num_compression_children[parent_key] = 2
1364
ann._expand_record(parent_key, (), None, ['line1\n', 'line2\n'],
1365
('fulltext', False))
1366
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1368
self.assertEqual({parent_key: 1}, ann._num_compression_children)
1369
# Expanding the second child should remove the content object, and the
1370
# num_compression_children entry
1371
res = ann._expand_record(rev2_key, (parent_key,), parent_key,
1373
self.assertFalse(parent_key in ann._content_objects)
1374
self.assertEqual({}, ann._num_compression_children)
1375
# We should not cache the content_objects for rev2 and rev, because
1376
# they do not have compression children of their own.
1377
self.assertEqual({}, ann._content_objects)
1379
def test__expand_delta_records_blocks(self):
1380
ann = self.make_annotator()
1381
rev_key = ('rev-id',)
1382
parent_key = ('parent-id',)
1383
record = ['0,1,1\n', 'new-line\n']
1384
details = ('line-delta', True)
1385
ann._num_compression_children[parent_key] = 2
1386
ann._expand_record(parent_key, (), None,
1387
['line1\n', 'line2\n', 'line3\n'],
1388
('fulltext', False))
1389
ann._expand_record(rev_key, (parent_key,), parent_key, record, details)
1390
self.assertEqual({(rev_key, parent_key): [(1, 1, 1), (3, 3, 0)]},
1391
ann._matching_blocks)
1392
rev2_key = ('rev2-id',)
1393
record = ['0,1,1\n', 'new-line\n']
1394
details = ('line-delta', False)
1395
ann._expand_record(rev2_key, (parent_key,), parent_key, record, details)
1396
self.assertEqual([(1, 1, 2), (3, 3, 0)],
1397
ann._matching_blocks[(rev2_key, parent_key)])
1399
def test__get_parent_ann_uses_matching_blocks(self):
1400
ann = self.make_annotator()
1401
rev_key = ('rev-id',)
1402
parent_key = ('parent-id',)
1403
parent_ann = [(parent_key,)]*3
1404
block_key = (rev_key, parent_key)
1405
ann._annotations_cache[parent_key] = parent_ann
1406
ann._matching_blocks[block_key] = [(0, 1, 1), (3, 3, 0)]
1407
# We should not try to access any parent_lines content, because we know
1408
# we already have the matching blocks
1409
par_ann, blocks = ann._get_parent_annotations_and_matches(rev_key,
1410
['1\n', '2\n', '3\n'], parent_key)
1411
self.assertEqual(parent_ann, par_ann)
1412
self.assertEqual([(0, 1, 1), (3, 3, 0)], blocks)
1413
self.assertEqual({}, ann._matching_blocks)
1415
def test__process_pending(self):
1416
ann = self.make_annotator()
1417
rev_key = ('rev-id',)
1420
record = ['0,1,1\n', 'new-line\n']
1421
details = ('line-delta', False)
1422
p1_record = ['line1\n', 'line2\n']
1423
ann._num_compression_children[p1_key] = 1
1424
res = ann._expand_record(rev_key, (p1_key,p2_key), p1_key,
1426
self.assertEqual(None, res)
1427
# self.assertTrue(p1_key in ann._pending_deltas)
1428
self.assertEqual({}, ann._pending_annotation)
1429
# Now insert p1, and we should be able to expand the delta
1430
res = ann._expand_record(p1_key, (), None, p1_record,
1431
('fulltext', False))
1432
self.assertEqual(p1_record, res)
1433
ann._annotations_cache[p1_key] = [(p1_key,)]*2
1434
res = ann._process_pending(p1_key)
1435
self.assertEqual([], res)
1436
self.assertFalse(p1_key in ann._pending_deltas)
1437
self.assertTrue(p2_key in ann._pending_annotation)
1438
self.assertEqual({p2_key: [(rev_key, (p1_key, p2_key))]},
1439
ann._pending_annotation)
1440
# Now fill in parent 2, and pending annotation should be satisfied
1441
res = ann._expand_record(p2_key, (), None, [], ('fulltext', False))
1442
ann._annotations_cache[p2_key] = []
1443
res = ann._process_pending(p2_key)
1444
self.assertEqual([rev_key], res)
1445
self.assertEqual({}, ann._pending_annotation)
1446
self.assertEqual({}, ann._pending_deltas)
1448
def test_record_delta_removes_basis(self):
1449
ann = self.make_annotator()
1450
ann._expand_record(('parent-id',), (), None,
1451
['line1\n', 'line2\n'], ('fulltext', False))
1452
ann._num_compression_children['parent-id'] = 2
1454
def test_annotate_special_text(self):
1455
ann = self.make_annotator()
1457
rev1_key = ('rev-1',)
1458
rev2_key = ('rev-2',)
1459
rev3_key = ('rev-3',)
1460
spec_key = ('special:',)
1461
vf.add_lines(rev1_key, [], ['initial content\n'])
1462
vf.add_lines(rev2_key, [rev1_key], ['initial content\n',
1465
vf.add_lines(rev3_key, [rev1_key], ['initial content\n',
1468
spec_text = ('initial content\n'
1472
ann.add_special_text(spec_key, [rev2_key, rev3_key], spec_text)
1473
anns, lines = ann.annotate(spec_key)
1474
self.assertEqual([(rev1_key,),
1475
(rev2_key, rev3_key),
1479
self.assertEqualDiff(spec_text, ''.join(lines))
1482
class KnitTests(TestCaseWithTransport):
1483
"""Class containing knit test helper routines."""
1485
def make_test_knit(self, annotate=False, name='test'):
1486
mapper = ConstantMapper(name)
1487
return make_file_factory(annotate, mapper)(self.get_transport())
1490
class TestBadShaError(KnitTests):
1491
"""Tests for handling of sha errors."""
1493
def test_sha_exception_has_text(self):
1494
# having the failed text included in the error allows for recovery.
1495
source = self.make_test_knit()
1496
target = self.make_test_knit(name="target")
1497
if not source._max_delta_chain:
1498
raise TestNotApplicable(
1499
"cannot get delta-caused sha failures without deltas.")
1502
broken = ('broken',)
1503
source.add_lines(basis, (), ['foo\n'])
1504
source.add_lines(broken, (basis,), ['foo\n', 'bar\n'])
1505
# Seed target with a bad basis text
1506
target.add_lines(basis, (), ['gam\n'])
1507
target.insert_record_stream(
1508
source.get_record_stream([broken], 'unordered', False))
1509
err = self.assertRaises(KnitCorrupt,
1510
target.get_record_stream([broken], 'unordered', True
1511
).next().get_bytes_as, 'chunked')
1512
self.assertEqual(['gam\n', 'bar\n'], err.content)
1513
# Test for formatting with live data
1514
self.assertStartsWith(str(err), "Knit ")
1517
class TestKnitIndex(KnitTests):
1519
def test_add_versions_dictionary_compresses(self):
1520
"""Adding versions to the index should update the lookup dict"""
1521
knit = self.make_test_knit()
1523
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1524
self.check_file_contents('test.kndx',
1525
'# bzr knit index 8\n'
1527
'a-1 fulltext 0 0 :'
1530
(('a-2',), ['fulltext'], (('a-2',), 0, 0), [('a-1',)]),
1531
(('a-3',), ['fulltext'], (('a-3',), 0, 0), [('a-2',)]),
1533
self.check_file_contents('test.kndx',
1534
'# bzr knit index 8\n'
1536
'a-1 fulltext 0 0 :\n'
1537
'a-2 fulltext 0 0 0 :\n'
1538
'a-3 fulltext 0 0 1 :'
1540
self.assertEqual({('a-3',), ('a-1',), ('a-2',)}, idx.keys())
1542
('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False)),
1543
('a-2',): ((('a-2',), 0, 0), None, (('a-1',),), ('fulltext', False)),
1544
('a-3',): ((('a-3',), 0, 0), None, (('a-2',),), ('fulltext', False)),
1545
}, idx.get_build_details(idx.keys()))
1546
self.assertEqual({('a-1',):(),
1547
('a-2',):(('a-1',),),
1548
('a-3',):(('a-2',),),},
1549
idx.get_parent_map(idx.keys()))
1551
def test_add_versions_fails_clean(self):
1552
"""If add_versions fails in the middle, it restores a pristine state.
1554
Any modifications that are made to the index are reset if all versions
1557
# This cheats a little bit by passing in a generator which will
1558
# raise an exception before the processing finishes
1559
# Other possibilities would be to have an version with the wrong number
1560
# of entries, or to make the backing transport unable to write any
1563
knit = self.make_test_knit()
1565
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1567
class StopEarly(Exception):
1570
def generate_failure():
1571
"""Add some entries and then raise an exception"""
1572
yield (('a-2',), ['fulltext'], (None, 0, 0), ('a-1',))
1573
yield (('a-3',), ['fulltext'], (None, 0, 0), ('a-2',))
1576
# Assert the pre-condition
1578
self.assertEqual({('a-1',)}, set(idx.keys()))
1580
{('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False))},
1581
idx.get_build_details([('a-1',)]))
1582
self.assertEqual({('a-1',):()}, idx.get_parent_map(idx.keys()))
1585
self.assertRaises(StopEarly, idx.add_records, generate_failure())
1586
# And it shouldn't be modified
1589
def test_knit_index_ignores_empty_files(self):
1590
# There was a race condition in older bzr, where a ^C at the right time
1591
# could leave an empty .kndx file, which bzr would later claim was a
1592
# corrupted file since the header was not present. In reality, the file
1593
# just wasn't created, so it should be ignored.
1594
t = transport.get_transport_from_path('.')
1595
t.put_bytes('test.kndx', '')
1597
knit = self.make_test_knit()
1599
def test_knit_index_checks_header(self):
1600
t = transport.get_transport_from_path('.')
1601
t.put_bytes('test.kndx', '# not really a knit header\n\n')
1602
k = self.make_test_knit()
1603
self.assertRaises(KnitHeaderError, k.keys)
1606
class TestGraphIndexKnit(KnitTests):
1607
"""Tests for knits using a GraphIndex rather than a KnitIndex."""
1609
def make_g_index(self, name, ref_lists=0, nodes=[]):
1610
builder = GraphIndexBuilder(ref_lists)
1611
for node, references, value in nodes:
1612
builder.add_node(node, references, value)
1613
stream = builder.finish()
1614
trans = self.get_transport()
1615
size = trans.put_file(name, stream)
1616
return GraphIndex(trans, name, size)
1618
def two_graph_index(self, deltas=False, catch_adds=False):
1619
"""Build a two-graph index.
1621
:param deltas: If true, use underlying indices with two node-ref
1622
lists and 'parent' set to a delta-compressed against tail.
1624
# build a complex graph across several indices.
1626
# delta compression inn the index
1627
index1 = self.make_g_index('1', 2, [
1628
(('tip', ), 'N0 100', ([('parent', )], [], )),
1629
(('tail', ), '', ([], []))])
1630
index2 = self.make_g_index('2', 2, [
1631
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], [('tail', )])),
1632
(('separate', ), '', ([], []))])
1634
# just blob location and graph in the index.
1635
index1 = self.make_g_index('1', 1, [
1636
(('tip', ), 'N0 100', ([('parent', )], )),
1637
(('tail', ), '', ([], ))])
1638
index2 = self.make_g_index('2', 1, [
1639
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], )),
1640
(('separate', ), '', ([], ))])
1641
combined_index = CombinedGraphIndex([index1, index2])
1643
self.combined_index = combined_index
1644
self.caught_entries = []
1645
add_callback = self.catch_add
1648
return _KnitGraphIndex(combined_index, lambda:True, deltas=deltas,
1649
add_callback=add_callback)
1651
def test_keys(self):
1652
index = self.two_graph_index()
1653
self.assertEqual({('tail',), ('tip',), ('parent',), ('separate',)},
1656
def test_get_position(self):
1657
index = self.two_graph_index()
1658
self.assertEqual((index._graph_index._indices[0], 0, 100), index.get_position(('tip',)))
1659
self.assertEqual((index._graph_index._indices[1], 100, 78), index.get_position(('parent',)))
1661
def test_get_method_deltas(self):
1662
index = self.two_graph_index(deltas=True)
1663
self.assertEqual('fulltext', index.get_method(('tip',)))
1664
self.assertEqual('line-delta', index.get_method(('parent',)))
1666
def test_get_method_no_deltas(self):
1667
# check that the parent-history lookup is ignored with deltas=False.
1668
index = self.two_graph_index(deltas=False)
1669
self.assertEqual('fulltext', index.get_method(('tip',)))
1670
self.assertEqual('fulltext', index.get_method(('parent',)))
1672
def test_get_options_deltas(self):
1673
index = self.two_graph_index(deltas=True)
1674
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1675
self.assertEqual(['line-delta'], index.get_options(('parent',)))
1677
def test_get_options_no_deltas(self):
1678
# check that the parent-history lookup is ignored with deltas=False.
1679
index = self.two_graph_index(deltas=False)
1680
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1681
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1683
def test_get_parent_map(self):
1684
index = self.two_graph_index()
1685
self.assertEqual({('parent',):(('tail',), ('ghost',))},
1686
index.get_parent_map([('parent',), ('ghost',)]))
1688
def catch_add(self, entries):
1689
self.caught_entries.append(entries)
1691
def test_add_no_callback_errors(self):
1692
index = self.two_graph_index()
1693
self.assertRaises(errors.ReadOnlyError, index.add_records,
1694
[(('new',), 'fulltext,no-eol', (None, 50, 60), ['separate'])])
1696
def test_add_version_smoke(self):
1697
index = self.two_graph_index(catch_adds=True)
1698
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60),
1700
self.assertEqual([[(('new', ), 'N50 60', ((('separate',),),))]],
1701
self.caught_entries)
1703
def test_add_version_delta_not_delta_index(self):
1704
index = self.two_graph_index(catch_adds=True)
1705
self.assertRaises(KnitCorrupt, index.add_records,
1706
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1707
self.assertEqual([], self.caught_entries)
1709
def test_add_version_same_dup(self):
1710
index = self.two_graph_index(catch_adds=True)
1711
# options can be spelt two different ways
1712
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
1713
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
1714
# position/length are ignored (because each pack could have fulltext or
1715
# delta, and be at a different position.
1716
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1718
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1720
# but neither should have added data:
1721
self.assertEqual([[], [], [], []], self.caught_entries)
1723
def test_add_version_different_dup(self):
1724
index = self.two_graph_index(deltas=True, catch_adds=True)
1726
self.assertRaises(KnitCorrupt, index.add_records,
1727
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1728
self.assertRaises(KnitCorrupt, index.add_records,
1729
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1731
self.assertRaises(KnitCorrupt, index.add_records,
1732
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1733
self.assertEqual([], self.caught_entries)
1735
def test_add_versions_nodeltas(self):
1736
index = self.two_graph_index(catch_adds=True)
1738
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1739
(('new2',), 'fulltext', (None, 0, 6), [('new',)]),
1741
self.assertEqual([(('new', ), 'N50 60', ((('separate',),),)),
1742
(('new2', ), ' 0 6', ((('new',),),))],
1743
sorted(self.caught_entries[0]))
1744
self.assertEqual(1, len(self.caught_entries))
1746
def test_add_versions_deltas(self):
1747
index = self.two_graph_index(deltas=True, catch_adds=True)
1749
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1750
(('new2',), 'line-delta', (None, 0, 6), [('new',)]),
1752
self.assertEqual([(('new', ), 'N50 60', ((('separate',),), ())),
1753
(('new2', ), ' 0 6', ((('new',),), (('new',),), ))],
1754
sorted(self.caught_entries[0]))
1755
self.assertEqual(1, len(self.caught_entries))
1757
def test_add_versions_delta_not_delta_index(self):
1758
index = self.two_graph_index(catch_adds=True)
1759
self.assertRaises(KnitCorrupt, index.add_records,
1760
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1761
self.assertEqual([], self.caught_entries)
1763
def test_add_versions_random_id_accepted(self):
1764
index = self.two_graph_index(catch_adds=True)
1765
index.add_records([], random_id=True)
1767
def test_add_versions_same_dup(self):
1768
index = self.two_graph_index(catch_adds=True)
1769
# options can be spelt two different ways
1770
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100),
1772
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100),
1774
# position/length are ignored (because each pack could have fulltext or
1775
# delta, and be at a different position.
1776
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1778
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1780
# but neither should have added data.
1781
self.assertEqual([[], [], [], []], self.caught_entries)
1783
def test_add_versions_different_dup(self):
1784
index = self.two_graph_index(deltas=True, catch_adds=True)
1786
self.assertRaises(KnitCorrupt, index.add_records,
1787
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1788
self.assertRaises(KnitCorrupt, index.add_records,
1789
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1791
self.assertRaises(KnitCorrupt, index.add_records,
1792
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1793
# change options in the second record
1794
self.assertRaises(KnitCorrupt, index.add_records,
1795
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)]),
1796
(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1797
self.assertEqual([], self.caught_entries)
1799
def make_g_index_missing_compression_parent(self):
1800
graph_index = self.make_g_index('missing_comp', 2,
1801
[(('tip', ), ' 100 78',
1802
([('missing-parent', ), ('ghost', )], [('missing-parent', )]))])
1805
def make_g_index_missing_parent(self):
1806
graph_index = self.make_g_index('missing_parent', 2,
1807
[(('parent', ), ' 100 78', ([], [])),
1808
(('tip', ), ' 100 78',
1809
([('parent', ), ('missing-parent', )], [('parent', )])),
1813
def make_g_index_no_external_refs(self):
1814
graph_index = self.make_g_index('no_external_refs', 2,
1815
[(('rev', ), ' 100 78',
1816
([('parent', ), ('ghost', )], []))])
1819
def test_add_good_unvalidated_index(self):
1820
unvalidated = self.make_g_index_no_external_refs()
1821
combined = CombinedGraphIndex([unvalidated])
1822
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1823
index.scan_unvalidated_index(unvalidated)
1824
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1826
def test_add_missing_compression_parent_unvalidated_index(self):
1827
unvalidated = self.make_g_index_missing_compression_parent()
1828
combined = CombinedGraphIndex([unvalidated])
1829
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1830
index.scan_unvalidated_index(unvalidated)
1831
# This also checks that its only the compression parent that is
1832
# examined, otherwise 'ghost' would also be reported as a missing
1835
frozenset([('missing-parent',)]),
1836
index.get_missing_compression_parents())
1838
def test_add_missing_noncompression_parent_unvalidated_index(self):
1839
unvalidated = self.make_g_index_missing_parent()
1840
combined = CombinedGraphIndex([unvalidated])
1841
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1842
track_external_parent_refs=True)
1843
index.scan_unvalidated_index(unvalidated)
1845
frozenset([('missing-parent',)]), index.get_missing_parents())
1847
def test_track_external_parent_refs(self):
1848
g_index = self.make_g_index('empty', 2, [])
1849
combined = CombinedGraphIndex([g_index])
1850
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1851
add_callback=self.catch_add, track_external_parent_refs=True)
1852
self.caught_entries = []
1854
(('new-key',), 'fulltext,no-eol', (None, 50, 60),
1855
[('parent-1',), ('parent-2',)])])
1857
frozenset([('parent-1',), ('parent-2',)]),
1858
index.get_missing_parents())
1860
def test_add_unvalidated_index_with_present_external_references(self):
1861
index = self.two_graph_index(deltas=True)
1862
# Ugly hack to get at one of the underlying GraphIndex objects that
1863
# two_graph_index built.
1864
unvalidated = index._graph_index._indices[1]
1865
# 'parent' is an external ref of _indices[1] (unvalidated), but is
1866
# present in _indices[0].
1867
index.scan_unvalidated_index(unvalidated)
1868
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1870
def make_new_missing_parent_g_index(self, name):
1871
missing_parent = name + '-missing-parent'
1872
graph_index = self.make_g_index(name, 2,
1873
[((name + 'tip', ), ' 100 78',
1874
([(missing_parent, ), ('ghost', )], [(missing_parent, )]))])
1877
def test_add_mulitiple_unvalidated_indices_with_missing_parents(self):
1878
g_index_1 = self.make_new_missing_parent_g_index('one')
1879
g_index_2 = self.make_new_missing_parent_g_index('two')
1880
combined = CombinedGraphIndex([g_index_1, g_index_2])
1881
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1882
index.scan_unvalidated_index(g_index_1)
1883
index.scan_unvalidated_index(g_index_2)
1885
frozenset([('one-missing-parent',), ('two-missing-parent',)]),
1886
index.get_missing_compression_parents())
1888
def test_add_mulitiple_unvalidated_indices_with_mutual_dependencies(self):
1889
graph_index_a = self.make_g_index('one', 2,
1890
[(('parent-one', ), ' 100 78', ([('non-compression-parent',)], [])),
1891
(('child-of-two', ), ' 100 78',
1892
([('parent-two',)], [('parent-two',)]))])
1893
graph_index_b = self.make_g_index('two', 2,
1894
[(('parent-two', ), ' 100 78', ([('non-compression-parent',)], [])),
1895
(('child-of-one', ), ' 100 78',
1896
([('parent-one',)], [('parent-one',)]))])
1897
combined = CombinedGraphIndex([graph_index_a, graph_index_b])
1898
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1899
index.scan_unvalidated_index(graph_index_a)
1900
index.scan_unvalidated_index(graph_index_b)
1902
frozenset([]), index.get_missing_compression_parents())
1905
class TestNoParentsGraphIndexKnit(KnitTests):
1906
"""Tests for knits using _KnitGraphIndex with no parents."""
1908
def make_g_index(self, name, ref_lists=0, nodes=[]):
1909
builder = GraphIndexBuilder(ref_lists)
1910
for node, references in nodes:
1911
builder.add_node(node, references)
1912
stream = builder.finish()
1913
trans = self.get_transport()
1914
size = trans.put_file(name, stream)
1915
return GraphIndex(trans, name, size)
1917
def test_add_good_unvalidated_index(self):
1918
unvalidated = self.make_g_index('unvalidated')
1919
combined = CombinedGraphIndex([unvalidated])
1920
index = _KnitGraphIndex(combined, lambda: True, parents=False)
1921
index.scan_unvalidated_index(unvalidated)
1922
self.assertEqual(frozenset(),
1923
index.get_missing_compression_parents())
1925
def test_parents_deltas_incompatible(self):
1926
index = CombinedGraphIndex([])
1927
self.assertRaises(knit.KnitError, _KnitGraphIndex, lambda:True,
1928
index, deltas=True, parents=False)
1930
def two_graph_index(self, catch_adds=False):
1931
"""Build a two-graph index.
1933
:param deltas: If true, use underlying indices with two node-ref
1934
lists and 'parent' set to a delta-compressed against tail.
1936
# put several versions in the index.
1937
index1 = self.make_g_index('1', 0, [
1938
(('tip', ), 'N0 100'),
1940
index2 = self.make_g_index('2', 0, [
1941
(('parent', ), ' 100 78'),
1942
(('separate', ), '')])
1943
combined_index = CombinedGraphIndex([index1, index2])
1945
self.combined_index = combined_index
1946
self.caught_entries = []
1947
add_callback = self.catch_add
1950
return _KnitGraphIndex(combined_index, lambda:True, parents=False,
1951
add_callback=add_callback)
1953
def test_keys(self):
1954
index = self.two_graph_index()
1955
self.assertEqual({('tail',), ('tip',), ('parent',), ('separate',)},
1958
def test_get_position(self):
1959
index = self.two_graph_index()
1960
self.assertEqual((index._graph_index._indices[0], 0, 100),
1961
index.get_position(('tip',)))
1962
self.assertEqual((index._graph_index._indices[1], 100, 78),
1963
index.get_position(('parent',)))
1965
def test_get_method(self):
1966
index = self.two_graph_index()
1967
self.assertEqual('fulltext', index.get_method(('tip',)))
1968
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1970
def test_get_options(self):
1971
index = self.two_graph_index()
1972
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1973
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1975
def test_get_parent_map(self):
1976
index = self.two_graph_index()
1977
self.assertEqual({('parent',):None},
1978
index.get_parent_map([('parent',), ('ghost',)]))
1980
def catch_add(self, entries):
1981
self.caught_entries.append(entries)
1983
def test_add_no_callback_errors(self):
1984
index = self.two_graph_index()
1985
self.assertRaises(errors.ReadOnlyError, index.add_records,
1986
[(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)])])
1988
def test_add_version_smoke(self):
1989
index = self.two_graph_index(catch_adds=True)
1990
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60), [])])
1991
self.assertEqual([[(('new', ), 'N50 60')]],
1992
self.caught_entries)
1994
def test_add_version_delta_not_delta_index(self):
1995
index = self.two_graph_index(catch_adds=True)
1996
self.assertRaises(KnitCorrupt, index.add_records,
1997
[(('new',), 'no-eol,line-delta', (None, 0, 100), [])])
1998
self.assertEqual([], self.caught_entries)
2000
def test_add_version_same_dup(self):
2001
index = self.two_graph_index(catch_adds=True)
2002
# options can be spelt two different ways
2003
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
2004
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
2005
# position/length are ignored (because each pack could have fulltext or
2006
# delta, and be at a different position.
2007
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
2008
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
2009
# but neither should have added data.
2010
self.assertEqual([[], [], [], []], self.caught_entries)
2012
def test_add_version_different_dup(self):
2013
index = self.two_graph_index(catch_adds=True)
2015
self.assertRaises(KnitCorrupt, index.add_records,
2016
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2017
self.assertRaises(KnitCorrupt, index.add_records,
2018
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
2019
self.assertRaises(KnitCorrupt, index.add_records,
2020
[(('tip',), 'fulltext', (None, 0, 100), [])])
2022
self.assertRaises(KnitCorrupt, index.add_records,
2023
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
2024
self.assertEqual([], self.caught_entries)
2026
def test_add_versions(self):
2027
index = self.two_graph_index(catch_adds=True)
2029
(('new',), 'fulltext,no-eol', (None, 50, 60), []),
2030
(('new2',), 'fulltext', (None, 0, 6), []),
2032
self.assertEqual([(('new', ), 'N50 60'), (('new2', ), ' 0 6')],
2033
sorted(self.caught_entries[0]))
2034
self.assertEqual(1, len(self.caught_entries))
2036
def test_add_versions_delta_not_delta_index(self):
2037
index = self.two_graph_index(catch_adds=True)
2038
self.assertRaises(KnitCorrupt, index.add_records,
2039
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
2040
self.assertEqual([], self.caught_entries)
2042
def test_add_versions_parents_not_parents_index(self):
2043
index = self.two_graph_index(catch_adds=True)
2044
self.assertRaises(KnitCorrupt, index.add_records,
2045
[(('new',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
2046
self.assertEqual([], self.caught_entries)
2048
def test_add_versions_random_id_accepted(self):
2049
index = self.two_graph_index(catch_adds=True)
2050
index.add_records([], random_id=True)
2052
def test_add_versions_same_dup(self):
2053
index = self.two_graph_index(catch_adds=True)
2054
# options can be spelt two different ways
2055
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
2056
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
2057
# position/length are ignored (because each pack could have fulltext or
2058
# delta, and be at a different position.
2059
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
2060
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
2061
# but neither should have added data.
2062
self.assertEqual([[], [], [], []], self.caught_entries)
2064
def test_add_versions_different_dup(self):
2065
index = self.two_graph_index(catch_adds=True)
2067
self.assertRaises(KnitCorrupt, index.add_records,
2068
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2069
self.assertRaises(KnitCorrupt, index.add_records,
2070
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
2071
self.assertRaises(KnitCorrupt, index.add_records,
2072
[(('tip',), 'fulltext', (None, 0, 100), [])])
2074
self.assertRaises(KnitCorrupt, index.add_records,
2075
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
2076
# change options in the second record
2077
self.assertRaises(KnitCorrupt, index.add_records,
2078
[(('tip',), 'fulltext,no-eol', (None, 0, 100), []),
2079
(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2080
self.assertEqual([], self.caught_entries)
2083
class TestKnitVersionedFiles(KnitTests):
2085
def assertGroupKeysForIo(self, exp_groups, keys, non_local_keys,
2086
positions, _min_buffer_size=None):
2087
kvf = self.make_test_knit()
2088
if _min_buffer_size is None:
2089
_min_buffer_size = knit._STREAM_MIN_BUFFER_SIZE
2090
self.assertEqual(exp_groups, kvf._group_keys_for_io(keys,
2091
non_local_keys, positions,
2092
_min_buffer_size=_min_buffer_size))
2094
def assertSplitByPrefix(self, expected_map, expected_prefix_order,
2096
split, prefix_order = KnitVersionedFiles._split_by_prefix(keys)
2097
self.assertEqual(expected_map, split)
2098
self.assertEqual(expected_prefix_order, prefix_order)
2100
def test__group_keys_for_io(self):
2101
ft_detail = ('fulltext', False)
2102
ld_detail = ('line-delta', False)
2110
f_a: (ft_detail, (f_a, 0, 100), None),
2111
f_b: (ld_detail, (f_b, 100, 21), f_a),
2112
f_c: (ld_detail, (f_c, 180, 15), f_b),
2113
g_a: (ft_detail, (g_a, 121, 35), None),
2114
g_b: (ld_detail, (g_b, 156, 12), g_a),
2115
g_c: (ld_detail, (g_c, 195, 13), g_a),
2117
self.assertGroupKeysForIo([([f_a], set())],
2118
[f_a], [], positions)
2119
self.assertGroupKeysForIo([([f_a], {f_a})],
2120
[f_a], [f_a], positions)
2121
self.assertGroupKeysForIo([([f_a, f_b], set([]))],
2122
[f_a, f_b], [], positions)
2123
self.assertGroupKeysForIo([([f_a, f_b], {f_b})],
2124
[f_a, f_b], [f_b], positions)
2125
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2126
[f_a, g_a, f_b, g_b], [], positions)
2127
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2128
[f_a, g_a, f_b, g_b], [], positions,
2129
_min_buffer_size=150)
2130
self.assertGroupKeysForIo([([f_a, f_b], set()), ([g_a, g_b], set())],
2131
[f_a, g_a, f_b, g_b], [], positions,
2132
_min_buffer_size=100)
2133
self.assertGroupKeysForIo([([f_c], set()), ([g_b], set())],
2134
[f_c, g_b], [], positions,
2135
_min_buffer_size=125)
2136
self.assertGroupKeysForIo([([g_b, f_c], set())],
2137
[g_b, f_c], [], positions,
2138
_min_buffer_size=125)
2140
def test__split_by_prefix(self):
2141
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2142
'g': [('g', 'b'), ('g', 'a')],
2144
[('f', 'a'), ('g', 'b'),
2145
('g', 'a'), ('f', 'b')])
2147
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2148
'g': [('g', 'b'), ('g', 'a')],
2150
[('f', 'a'), ('f', 'b'),
2151
('g', 'b'), ('g', 'a')])
2153
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2154
'g': [('g', 'b'), ('g', 'a')],
2156
[('f', 'a'), ('f', 'b'),
2157
('g', 'b'), ('g', 'a')])
2159
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2160
'g': [('g', 'b'), ('g', 'a')],
2161
'': [('a',), ('b',)]
2163
[('f', 'a'), ('g', 'b'),
2165
('g', 'a'), ('f', 'b')])
2168
class TestStacking(KnitTests):
2170
def get_basis_and_test_knit(self):
2171
basis = self.make_test_knit(name='basis')
2172
basis = RecordingVersionedFilesDecorator(basis)
2173
test = self.make_test_knit(name='test')
2174
test.add_fallback_versioned_files(basis)
2177
def test_add_fallback_versioned_files(self):
2178
basis = self.make_test_knit(name='basis')
2179
test = self.make_test_knit(name='test')
2180
# It must not error; other tests test that the fallback is referred to
2181
# when accessing data.
2182
test.add_fallback_versioned_files(basis)
2184
def test_add_lines(self):
2185
# lines added to the test are not added to the basis
2186
basis, test = self.get_basis_and_test_knit()
2188
key_basis = ('bar',)
2189
key_cross_border = ('quux',)
2190
key_delta = ('zaphod',)
2191
test.add_lines(key, (), ['foo\n'])
2192
self.assertEqual({}, basis.get_parent_map([key]))
2193
# lines added to the test that reference across the stack do a
2195
basis.add_lines(key_basis, (), ['foo\n'])
2197
test.add_lines(key_cross_border, (key_basis,), ['foo\n'])
2198
self.assertEqual('fulltext', test._index.get_method(key_cross_border))
2199
# we don't even need to look at the basis to see that this should be
2200
# stored as a fulltext
2201
self.assertEqual([], basis.calls)
2202
# Subsequent adds do delta.
2204
test.add_lines(key_delta, (key_cross_border,), ['foo\n'])
2205
self.assertEqual('line-delta', test._index.get_method(key_delta))
2206
self.assertEqual([], basis.calls)
2208
def test_annotate(self):
2209
# annotations from the test knit are answered without asking the basis
2210
basis, test = self.get_basis_and_test_knit()
2212
key_basis = ('bar',)
2213
key_missing = ('missing',)
2214
test.add_lines(key, (), ['foo\n'])
2215
details = test.annotate(key)
2216
self.assertEqual([(key, 'foo\n')], details)
2217
self.assertEqual([], basis.calls)
2218
# But texts that are not in the test knit are looked for in the basis
2220
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2222
details = test.annotate(key_basis)
2223
self.assertEqual([(key_basis, 'foo\n'), (key_basis, 'bar\n')], details)
2224
# Not optimised to date:
2225
# self.assertEqual([("annotate", key_basis)], basis.calls)
2226
self.assertEqual([('get_parent_map', {key_basis}),
2227
('get_parent_map', {key_basis}),
2228
('get_record_stream', [key_basis], 'topological', True)],
2231
def test_check(self):
2232
# At the moment checking a stacked knit does implicitly check the
2234
basis, test = self.get_basis_and_test_knit()
2237
def test_get_parent_map(self):
2238
# parents in the test knit are answered without asking the basis
2239
basis, test = self.get_basis_and_test_knit()
2241
key_basis = ('bar',)
2242
key_missing = ('missing',)
2243
test.add_lines(key, (), [])
2244
parent_map = test.get_parent_map([key])
2245
self.assertEqual({key: ()}, parent_map)
2246
self.assertEqual([], basis.calls)
2247
# But parents that are not in the test knit are looked for in the basis
2248
basis.add_lines(key_basis, (), [])
2250
parent_map = test.get_parent_map([key, key_basis, key_missing])
2251
self.assertEqual({key: (),
2252
key_basis: ()}, parent_map)
2253
self.assertEqual([("get_parent_map", {key_basis, key_missing})],
2256
def test_get_record_stream_unordered_fulltexts(self):
2257
# records from the test knit are answered without asking the basis:
2258
basis, test = self.get_basis_and_test_knit()
2260
key_basis = ('bar',)
2261
key_missing = ('missing',)
2262
test.add_lines(key, (), ['foo\n'])
2263
records = list(test.get_record_stream([key], 'unordered', True))
2264
self.assertEqual(1, len(records))
2265
self.assertEqual([], basis.calls)
2266
# Missing (from test knit) objects are retrieved from the basis:
2267
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2269
records = list(test.get_record_stream([key_basis, key_missing],
2271
self.assertEqual(2, len(records))
2272
calls = list(basis.calls)
2273
for record in records:
2274
self.assertSubset([record.key], (key_basis, key_missing))
2275
if record.key == key_missing:
2276
self.assertIsInstance(record, AbsentContentFactory)
2278
reference = list(basis.get_record_stream([key_basis],
2279
'unordered', True))[0]
2280
self.assertEqual(reference.key, record.key)
2281
self.assertEqual(reference.sha1, record.sha1)
2282
self.assertEqual(reference.storage_kind, record.storage_kind)
2283
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2284
record.get_bytes_as(record.storage_kind))
2285
self.assertEqual(reference.get_bytes_as('fulltext'),
2286
record.get_bytes_as('fulltext'))
2287
# It's not strictly minimal, but it seems reasonable for now for it to
2288
# ask which fallbacks have which parents.
2290
("get_parent_map", {key_basis, key_missing}),
2291
("get_record_stream", [key_basis], 'unordered', True)],
2294
def test_get_record_stream_ordered_fulltexts(self):
2295
# ordering is preserved down into the fallback store.
2296
basis, test = self.get_basis_and_test_knit()
2298
key_basis = ('bar',)
2299
key_basis_2 = ('quux',)
2300
key_missing = ('missing',)
2301
test.add_lines(key, (key_basis,), ['foo\n'])
2302
# Missing (from test knit) objects are retrieved from the basis:
2303
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2304
basis.add_lines(key_basis_2, (), ['quux\n'])
2306
# ask for in non-topological order
2307
records = list(test.get_record_stream(
2308
[key, key_basis, key_missing, key_basis_2], 'topological', True))
2309
self.assertEqual(4, len(records))
2311
for record in records:
2312
self.assertSubset([record.key],
2313
(key_basis, key_missing, key_basis_2, key))
2314
if record.key == key_missing:
2315
self.assertIsInstance(record, AbsentContentFactory)
2317
results.append((record.key, record.sha1, record.storage_kind,
2318
record.get_bytes_as('fulltext')))
2319
calls = list(basis.calls)
2320
order = [record[0] for record in results]
2321
self.assertEqual([key_basis_2, key_basis, key], order)
2322
for result in results:
2323
if result[0] == key:
2327
record = next(source.get_record_stream([result[0]], 'unordered',
2329
self.assertEqual(record.key, result[0])
2330
self.assertEqual(record.sha1, result[1])
2331
# We used to check that the storage kind matched, but actually it
2332
# depends on whether it was sourced from the basis, or in a single
2333
# group, because asking for full texts returns proxy objects to a
2334
# _ContentMapGenerator object; so checking the kind is unneeded.
2335
self.assertEqual(record.get_bytes_as('fulltext'), result[3])
2336
# It's not strictly minimal, but it seems reasonable for now for it to
2337
# ask which fallbacks have which parents.
2339
("get_parent_map", {key_basis, key_basis_2, key_missing}),
2340
# topological is requested from the fallback, because that is what
2341
# was requested at the top level.
2342
("get_record_stream", [key_basis_2, key_basis], 'topological', True)],
2345
def test_get_record_stream_unordered_deltas(self):
2346
# records from the test knit are answered without asking the basis:
2347
basis, test = self.get_basis_and_test_knit()
2349
key_basis = ('bar',)
2350
key_missing = ('missing',)
2351
test.add_lines(key, (), ['foo\n'])
2352
records = list(test.get_record_stream([key], 'unordered', False))
2353
self.assertEqual(1, len(records))
2354
self.assertEqual([], basis.calls)
2355
# Missing (from test knit) objects are retrieved from the basis:
2356
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2358
records = list(test.get_record_stream([key_basis, key_missing],
2359
'unordered', False))
2360
self.assertEqual(2, len(records))
2361
calls = list(basis.calls)
2362
for record in records:
2363
self.assertSubset([record.key], (key_basis, key_missing))
2364
if record.key == key_missing:
2365
self.assertIsInstance(record, AbsentContentFactory)
2367
reference = list(basis.get_record_stream([key_basis],
2368
'unordered', False))[0]
2369
self.assertEqual(reference.key, record.key)
2370
self.assertEqual(reference.sha1, record.sha1)
2371
self.assertEqual(reference.storage_kind, record.storage_kind)
2372
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2373
record.get_bytes_as(record.storage_kind))
2374
# It's not strictly minimal, but it seems reasonable for now for it to
2375
# ask which fallbacks have which parents.
2377
("get_parent_map", {key_basis, key_missing}),
2378
("get_record_stream", [key_basis], 'unordered', False)],
2381
def test_get_record_stream_ordered_deltas(self):
2382
# ordering is preserved down into the fallback store.
2383
basis, test = self.get_basis_and_test_knit()
2385
key_basis = ('bar',)
2386
key_basis_2 = ('quux',)
2387
key_missing = ('missing',)
2388
test.add_lines(key, (key_basis,), ['foo\n'])
2389
# Missing (from test knit) objects are retrieved from the basis:
2390
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2391
basis.add_lines(key_basis_2, (), ['quux\n'])
2393
# ask for in non-topological order
2394
records = list(test.get_record_stream(
2395
[key, key_basis, key_missing, key_basis_2], 'topological', False))
2396
self.assertEqual(4, len(records))
2398
for record in records:
2399
self.assertSubset([record.key],
2400
(key_basis, key_missing, key_basis_2, key))
2401
if record.key == key_missing:
2402
self.assertIsInstance(record, AbsentContentFactory)
2404
results.append((record.key, record.sha1, record.storage_kind,
2405
record.get_bytes_as(record.storage_kind)))
2406
calls = list(basis.calls)
2407
order = [record[0] for record in results]
2408
self.assertEqual([key_basis_2, key_basis, key], order)
2409
for result in results:
2410
if result[0] == key:
2414
record = next(source.get_record_stream([result[0]], 'unordered',
2416
self.assertEqual(record.key, result[0])
2417
self.assertEqual(record.sha1, result[1])
2418
self.assertEqual(record.storage_kind, result[2])
2419
self.assertEqual(record.get_bytes_as(record.storage_kind), result[3])
2420
# It's not strictly minimal, but it seems reasonable for now for it to
2421
# ask which fallbacks have which parents.
2423
("get_parent_map", {key_basis, key_basis_2, key_missing}),
2424
("get_record_stream", [key_basis_2, key_basis], 'topological', False)],
2427
def test_get_sha1s(self):
2428
# sha1's in the test knit are answered without asking the basis
2429
basis, test = self.get_basis_and_test_knit()
2431
key_basis = ('bar',)
2432
key_missing = ('missing',)
2433
test.add_lines(key, (), ['foo\n'])
2434
key_sha1sum = osutils.sha_string('foo\n')
2435
sha1s = test.get_sha1s([key])
2436
self.assertEqual({key: key_sha1sum}, sha1s)
2437
self.assertEqual([], basis.calls)
2438
# But texts that are not in the test knit are looked for in the basis
2439
# directly (rather than via text reconstruction) so that remote servers
2440
# etc don't have to answer with full content.
2441
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2442
basis_sha1sum = osutils.sha_string('foo\nbar\n')
2444
sha1s = test.get_sha1s([key, key_missing, key_basis])
2445
self.assertEqual({key: key_sha1sum,
2446
key_basis: basis_sha1sum}, sha1s)
2447
self.assertEqual([("get_sha1s", {key_basis, key_missing})],
2450
def test_insert_record_stream(self):
2451
# records are inserted as normal; insert_record_stream builds on
2452
# add_lines, so a smoke test should be all that's needed:
2454
key_basis = ('bar',)
2455
key_delta = ('zaphod',)
2456
basis, test = self.get_basis_and_test_knit()
2457
source = self.make_test_knit(name='source')
2458
basis.add_lines(key_basis, (), ['foo\n'])
2460
source.add_lines(key_basis, (), ['foo\n'])
2461
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2462
stream = source.get_record_stream([key_delta], 'unordered', False)
2463
test.insert_record_stream(stream)
2464
# XXX: this does somewhat too many calls in making sure of whether it
2465
# has to recreate the full text.
2466
self.assertEqual([("get_parent_map", {key_basis}),
2467
('get_parent_map', {key_basis}),
2468
('get_record_stream', [key_basis], 'unordered', True)],
2470
self.assertEqual({key_delta:(key_basis,)},
2471
test.get_parent_map([key_delta]))
2472
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2473
'unordered', True).next().get_bytes_as('fulltext'))
2475
def test_iter_lines_added_or_present_in_keys(self):
2476
# Lines from the basis are returned, and lines for a given key are only
2480
# all sources are asked for keys:
2481
basis, test = self.get_basis_and_test_knit()
2482
basis.add_lines(key1, (), ["foo"])
2484
lines = list(test.iter_lines_added_or_present_in_keys([key1]))
2485
self.assertEqual([("foo\n", key1)], lines)
2486
self.assertEqual([("iter_lines_added_or_present_in_keys", {key1})],
2488
# keys in both are not duplicated:
2489
test.add_lines(key2, (), ["bar\n"])
2490
basis.add_lines(key2, (), ["bar\n"])
2492
lines = list(test.iter_lines_added_or_present_in_keys([key2]))
2493
self.assertEqual([("bar\n", key2)], lines)
2494
self.assertEqual([], basis.calls)
2496
def test_keys(self):
2499
# all sources are asked for keys:
2500
basis, test = self.get_basis_and_test_knit()
2502
self.assertEqual(set(), set(keys))
2503
self.assertEqual([("keys",)], basis.calls)
2504
# keys from a basis are returned:
2505
basis.add_lines(key1, (), [])
2508
self.assertEqual({key1}, set(keys))
2509
self.assertEqual([("keys",)], basis.calls)
2510
# keys in both are not duplicated:
2511
test.add_lines(key2, (), [])
2512
basis.add_lines(key2, (), [])
2515
self.assertEqual(2, len(keys))
2516
self.assertEqual({key1, key2}, set(keys))
2517
self.assertEqual([("keys",)], basis.calls)
2519
def test_add_mpdiffs(self):
2520
# records are inserted as normal; add_mpdiff builds on
2521
# add_lines, so a smoke test should be all that's needed:
2523
key_basis = ('bar',)
2524
key_delta = ('zaphod',)
2525
basis, test = self.get_basis_and_test_knit()
2526
source = self.make_test_knit(name='source')
2527
basis.add_lines(key_basis, (), ['foo\n'])
2529
source.add_lines(key_basis, (), ['foo\n'])
2530
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2531
diffs = source.make_mpdiffs([key_delta])
2532
test.add_mpdiffs([(key_delta, (key_basis,),
2533
source.get_sha1s([key_delta])[key_delta], diffs[0])])
2534
self.assertEqual([("get_parent_map", {key_basis}),
2535
('get_record_stream', [key_basis], 'unordered', True),],
2537
self.assertEqual({key_delta:(key_basis,)},
2538
test.get_parent_map([key_delta]))
2539
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2540
'unordered', True).next().get_bytes_as('fulltext'))
2542
def test_make_mpdiffs(self):
2543
# Generating an mpdiff across a stacking boundary should detect parent
2547
key_right = ('zaphod',)
2548
basis, test = self.get_basis_and_test_knit()
2549
basis.add_lines(key_left, (), ['bar\n'])
2550
basis.add_lines(key_right, (), ['zaphod\n'])
2552
test.add_lines(key, (key_left, key_right),
2553
['bar\n', 'foo\n', 'zaphod\n'])
2554
diffs = test.make_mpdiffs([key])
2556
multiparent.MultiParent([multiparent.ParentText(0, 0, 0, 1),
2557
multiparent.NewText(['foo\n']),
2558
multiparent.ParentText(1, 0, 2, 1)])],
2560
self.assertEqual(3, len(basis.calls))
2562
("get_parent_map", {key_left, key_right}),
2563
("get_parent_map", {key_left, key_right}),
2566
last_call = basis.calls[-1]
2567
self.assertEqual('get_record_stream', last_call[0])
2568
self.assertEqual({key_left, key_right}, set(last_call[1]))
2569
self.assertEqual('topological', last_call[2])
2570
self.assertEqual(True, last_call[3])
2573
class TestNetworkBehaviour(KnitTests):
2574
"""Tests for getting data out of/into knits over the network."""
2576
def test_include_delta_closure_generates_a_knit_delta_closure(self):
2577
vf = self.make_test_knit(name='test')
2578
# put in three texts, giving ft, delta, delta
2579
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2580
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2581
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2582
# But heuristics could interfere, so check what happened:
2583
self.assertEqual(['knit-ft-gz', 'knit-delta-gz', 'knit-delta-gz'],
2584
[record.storage_kind for record in
2585
vf.get_record_stream([('base',), ('d1',), ('d2',)],
2586
'topological', False)])
2587
# generate a stream of just the deltas include_delta_closure=True,
2588
# serialise to the network, and check that we get a delta closure on the wire.
2589
stream = vf.get_record_stream([('d1',), ('d2',)], 'topological', True)
2590
netb = [record.get_bytes_as(record.storage_kind) for record in stream]
2591
# The first bytes should be a memo from _ContentMapGenerator, and the
2592
# second bytes should be empty (because its a API proxy not something
2593
# for wire serialisation.
2594
self.assertEqual('', netb[1])
2596
kind, line_end = network_bytes_to_kind_and_offset(bytes)
2597
self.assertEqual('knit-delta-closure', kind)
2600
class TestContentMapGenerator(KnitTests):
2601
"""Tests for ContentMapGenerator"""
2603
def test_get_record_stream_gives_records(self):
2604
vf = self.make_test_knit(name='test')
2605
# put in three texts, giving ft, delta, delta
2606
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2607
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2608
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2609
keys = [('d1',), ('d2',)]
2610
generator = _VFContentMapGenerator(vf, keys,
2611
global_map=vf.get_parent_map(keys))
2612
for record in generator.get_record_stream():
2613
if record.key == ('d1',):
2614
self.assertEqual('d1\n', record.get_bytes_as('fulltext'))
2616
self.assertEqual('d2\n', record.get_bytes_as('fulltext'))
2618
def test_get_record_stream_kinds_are_raw(self):
2619
vf = self.make_test_knit(name='test')
2620
# put in three texts, giving ft, delta, delta
2621
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2622
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2623
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2624
keys = [('base',), ('d1',), ('d2',)]
2625
generator = _VFContentMapGenerator(vf, keys,
2626
global_map=vf.get_parent_map(keys))
2627
kinds = {('base',): 'knit-delta-closure',
2628
('d1',): 'knit-delta-closure-ref',
2629
('d2',): 'knit-delta-closure-ref',
2631
for record in generator.get_record_stream():
2632
self.assertEqual(kinds[record.key], record.storage_kind)
added
removed
breezy/tests/test_knit.py
