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"""
20
from io import BytesIO
34
from ..bzr.index import *
35
from ..bzr.knit import (
39
KnitDataStreamIncompatible,
40
KnitDataStreamUnknown,
42
KnitIndexUnknownMethod,
45
_VFContentMapGenerator,
51
from ..patiencediff import PatienceSequenceMatcher
58
TestCaseWithMemoryTransport,
59
TestCaseWithTransport,
62
from ..bzr.versionedfile import (
65
network_bytes_to_kind_and_offset,
66
RecordingVersionedFilesDecorator,
73
compiled_knit_feature = features.ModuleAvailableFeature(
74
'breezy.bzr._knit_load_data_pyx')
77
class ErrorTests(TestCase):
79
def test_knit_data_stream_incompatible(self):
80
error = KnitDataStreamIncompatible(
81
'stream format', 'target format')
82
self.assertEqual('Cannot insert knit data stream of format '
83
'"stream format" into knit of format '
84
'"target format".', str(error))
86
def test_knit_data_stream_unknown(self):
87
error = KnitDataStreamUnknown(
89
self.assertEqual('Cannot parse knit data stream of format '
90
'"stream format".', str(error))
92
def test_knit_header_error(self):
93
error = KnitHeaderError('line foo\n', 'path/to/file')
94
self.assertEqual("Knit header error: 'line foo\\n' unexpected"
95
" for file \"path/to/file\".", str(error))
97
def test_knit_index_unknown_method(self):
98
error = KnitIndexUnknownMethod('http://host/foo.kndx',
100
self.assertEqual("Knit index http://host/foo.kndx does not have a"
101
" known method in options: ['bad', 'no-eol']",
105
class KnitContentTestsMixin(object):
107
def test_constructor(self):
108
content = self._make_content([])
111
content = self._make_content([])
112
self.assertEqual(content.text(), [])
114
content = self._make_content(
115
[(b"origin1", b"text1"), (b"origin2", b"text2")])
116
self.assertEqual(content.text(), [b"text1", b"text2"])
119
content = self._make_content(
120
[(b"origin1", b"text1"), (b"origin2", b"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)
131
if noeol and not line.endswith('\n'):
135
source_content = self._make_content(
136
[(None, nl(l)) for l in source_lines])
137
target_content = self._make_content(
138
[(None, nl(l)) for l in target_lines])
139
line_delta = source_content.line_delta(target_content)
140
delta_blocks = list(KnitContent.get_line_delta_blocks(line_delta,
141
source_lines, target_lines))
142
matcher = PatienceSequenceMatcher(None, source_lines, target_lines)
143
matcher_blocks = list(matcher.get_matching_blocks())
144
self.assertEqual(matcher_blocks, delta_blocks)
146
def test_get_line_delta_blocks(self):
147
self.assertDerivedBlocksEqual('a\nb\nc\n', 'q\nc\n')
148
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1)
149
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1A)
150
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1B)
151
self.assertDerivedBlocksEqual(TEXT_1B, TEXT_1A)
152
self.assertDerivedBlocksEqual(TEXT_1A, TEXT_1B)
153
self.assertDerivedBlocksEqual(TEXT_1A, '')
154
self.assertDerivedBlocksEqual('', TEXT_1A)
155
self.assertDerivedBlocksEqual('', '')
156
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd')
158
def test_get_line_delta_blocks_noeol(self):
159
"""Handle historical knit deltas safely
161
Some existing knit deltas don't consider the last line to differ
162
when the only difference whether it has a final newline.
164
New knit deltas appear to always consider the last line to differ
167
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd\n', noeol=True)
168
self.assertDerivedBlocksEqual('a\nb\nc\nd\n', 'a\nb\nc', noeol=True)
169
self.assertDerivedBlocksEqual('a\nb\nc\n', 'a\nb\nc', noeol=True)
170
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\n', noeol=True)
182
Banana cup cake recipe
192
Banana cup cake recipe
194
- bananas (do not use plantains!!!)
201
Banana cup cake recipe
218
class TestPlainKnitContent(TestCase, KnitContentTestsMixin):
220
def _make_content(self, lines):
221
annotated_content = AnnotatedKnitContent(lines)
222
return PlainKnitContent(annotated_content.text(), 'bogus')
224
def test_annotate(self):
225
content = self._make_content([])
226
self.assertEqual(content.annotate(), [])
228
content = self._make_content(
229
[("origin1", "text1"), ("origin2", "text2")])
230
self.assertEqual(content.annotate(),
231
[("bogus", "text1"), ("bogus", "text2")])
233
def test_line_delta(self):
234
content1 = self._make_content([("", "a"), ("", "b")])
235
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
236
self.assertEqual(content1.line_delta(content2),
237
[(1, 2, 2, ["a", "c"])])
239
def test_line_delta_iter(self):
240
content1 = self._make_content([("", "a"), ("", "b")])
241
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
242
it = content1.line_delta_iter(content2)
243
self.assertEqual(next(it), (1, 2, 2, ["a", "c"]))
244
self.assertRaises(StopIteration, next, it)
247
class TestAnnotatedKnitContent(TestCase, KnitContentTestsMixin):
249
def _make_content(self, lines):
250
return AnnotatedKnitContent(lines)
252
def test_annotate(self):
253
content = self._make_content([])
254
self.assertEqual(content.annotate(), [])
256
content = self._make_content(
257
[(b"origin1", b"text1"), (b"origin2", b"text2")])
258
self.assertEqual(content.annotate(),
259
[(b"origin1", b"text1"), (b"origin2", b"text2")])
261
def test_line_delta(self):
262
content1 = self._make_content([("", "a"), ("", "b")])
263
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
264
self.assertEqual(content1.line_delta(content2),
265
[(1, 2, 2, [("", "a"), ("", "c")])])
267
def test_line_delta_iter(self):
268
content1 = self._make_content([("", "a"), ("", "b")])
269
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
270
it = content1.line_delta_iter(content2)
271
self.assertEqual(next(it), (1, 2, 2, [("", "a"), ("", "c")]))
272
self.assertRaises(StopIteration, next, it)
275
class MockTransport(object):
277
def __init__(self, file_lines=None):
278
self.file_lines = file_lines
280
# We have no base directory for the MockTransport
283
def get(self, filename):
284
if self.file_lines is None:
285
raise errors.NoSuchFile(filename)
287
return BytesIO(b"\n".join(self.file_lines))
289
def readv(self, relpath, offsets):
290
fp = self.get(relpath)
291
for offset, size in offsets:
293
yield offset, fp.read(size)
295
def __getattr__(self, name):
296
def queue_call(*args, **kwargs):
297
self.calls.append((name, args, kwargs))
301
class MockReadvFailingTransport(MockTransport):
302
"""Fail in the middle of a readv() result.
304
This Transport will successfully yield the first two requested hunks, but
305
raise NoSuchFile for the rest.
308
def readv(self, relpath, offsets):
310
for result in MockTransport.readv(self, relpath, offsets):
312
# we use 2 because the first offset is the pack header, the second
313
# is the first actual content requset
315
raise errors.NoSuchFile(relpath)
319
class KnitRecordAccessTestsMixin(object):
320
"""Tests for getting and putting knit records."""
322
def test_add_raw_records(self):
323
"""Add_raw_records adds records retrievable later."""
324
access = self.get_access()
325
memos = access.add_raw_records([(b'key', 10)], b'1234567890')
326
self.assertEqual([b'1234567890'], list(access.get_raw_records(memos)))
328
def test_add_several_raw_records(self):
329
"""add_raw_records with many records and read some back."""
330
access = self.get_access()
331
memos = access.add_raw_records([(b'key', 10), (b'key2', 2), (b'key3', 5)],
332
b'12345678901234567')
333
self.assertEqual([b'1234567890', b'12', b'34567'],
334
list(access.get_raw_records(memos)))
335
self.assertEqual([b'1234567890'],
336
list(access.get_raw_records(memos[0:1])))
337
self.assertEqual([b'12'],
338
list(access.get_raw_records(memos[1:2])))
339
self.assertEqual([b'34567'],
340
list(access.get_raw_records(memos[2:3])))
341
self.assertEqual([b'1234567890', b'34567'],
342
list(access.get_raw_records(memos[0:1] + memos[2:3])))
345
class TestKnitKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
346
"""Tests for the .kndx implementation."""
348
def get_access(self):
349
"""Get a .knit style access instance."""
350
mapper = ConstantMapper("foo")
351
access = _KnitKeyAccess(self.get_transport(), mapper)
355
class _TestException(Exception):
356
"""Just an exception for local tests to use."""
359
class TestPackKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
360
"""Tests for the pack based access."""
362
def get_access(self):
363
return self._get_access()[0]
365
def _get_access(self, packname='packfile', index='FOO'):
366
transport = self.get_transport()
368
def write_data(bytes):
369
transport.append_bytes(packname, bytes)
370
writer = pack.ContainerWriter(write_data)
372
access = pack_repo._DirectPackAccess({})
373
access.set_writer(writer, index, (transport, packname))
374
return access, writer
376
def make_pack_file(self):
377
"""Create a pack file with 2 records."""
378
access, writer = self._get_access(packname='packname', index='foo')
380
memos.extend(access.add_raw_records([(b'key1', 10)], b'1234567890'))
381
memos.extend(access.add_raw_records([(b'key2', 5)], b'12345'))
385
def test_pack_collection_pack_retries(self):
386
"""An explicit pack of a pack collection succeeds even when a
387
concurrent pack happens.
389
builder = self.make_branch_builder('.')
390
builder.start_series()
391
builder.build_snapshot(None, [
392
('add', ('', b'root-id', 'directory', None)),
393
('add', ('file', b'file-id', 'file', b'content\nrev 1\n')),
394
], revision_id=b'rev-1')
395
builder.build_snapshot([b'rev-1'], [
396
('modify', ('file', b'content\nrev 2\n')),
397
], revision_id=b'rev-2')
398
builder.build_snapshot([b'rev-2'], [
399
('modify', ('file', b'content\nrev 3\n')),
400
], revision_id=b'rev-3')
401
self.addCleanup(builder.finish_series)
402
b = builder.get_branch()
403
self.addCleanup(b.lock_write().unlock)
405
collection = repo._pack_collection
406
# Concurrently repack the repo.
407
reopened_repo = repo.controldir.open_repository()
412
def make_vf_for_retrying(self):
413
"""Create 3 packs and a reload function.
415
Originally, 2 pack files will have the data, but one will be missing.
416
And then the third will be used in place of the first two if reload()
419
:return: (versioned_file, reload_counter)
420
versioned_file a KnitVersionedFiles using the packs for access
422
builder = self.make_branch_builder('.', format="1.9")
423
builder.start_series()
424
builder.build_snapshot(None, [
425
('add', ('', b'root-id', 'directory', None)),
426
('add', ('file', b'file-id', 'file', b'content\nrev 1\n')),
427
], revision_id=b'rev-1')
428
builder.build_snapshot([b'rev-1'], [
429
('modify', ('file', b'content\nrev 2\n')),
430
], revision_id=b'rev-2')
431
builder.build_snapshot([b'rev-2'], [
432
('modify', ('file', b'content\nrev 3\n')),
433
], revision_id=b'rev-3')
434
builder.finish_series()
435
b = builder.get_branch()
437
self.addCleanup(b.unlock)
438
# Pack these three revisions into another pack file, but don't remove
441
collection = repo._pack_collection
442
collection.ensure_loaded()
443
orig_packs = collection.packs
444
packer = knitpack_repo.KnitPacker(collection, orig_packs, '.testpack')
445
new_pack = packer.pack()
446
# forget about the new pack
450
# Set up a reload() function that switches to using the new pack file
451
new_index = new_pack.revision_index
452
access_tuple = new_pack.access_tuple()
453
reload_counter = [0, 0, 0]
456
reload_counter[0] += 1
457
if reload_counter[1] > 0:
458
# We already reloaded, nothing more to do
459
reload_counter[2] += 1
461
reload_counter[1] += 1
462
vf._index._graph_index._indices[:] = [new_index]
463
vf._access._indices.clear()
464
vf._access._indices[new_index] = access_tuple
466
# Delete one of the pack files so the data will need to be reloaded. We
467
# will delete the file with 'rev-2' in it
468
trans, name = orig_packs[1].access_tuple()
470
# We don't have the index trigger reloading because we want to test
471
# that we reload when the .pack disappears
472
vf._access._reload_func = reload
473
return vf, reload_counter
475
def make_reload_func(self, return_val=True):
479
reload_called[0] += 1
481
return reload_called, reload
483
def make_retry_exception(self):
484
# We raise a real exception so that sys.exc_info() is properly
487
raise _TestException('foobar')
488
except _TestException as e:
489
retry_exc = errors.RetryWithNewPacks(None, reload_occurred=False,
490
exc_info=sys.exc_info())
491
# GZ 2010-08-10: Cycle with exc_info affects 3 tests
494
def test_read_from_several_packs(self):
495
access, writer = self._get_access()
497
memos.extend(access.add_raw_records([(b'key', 10)], b'1234567890'))
499
access, writer = self._get_access('pack2', 'FOOBAR')
500
memos.extend(access.add_raw_records([(b'key', 5)], b'12345'))
502
access, writer = self._get_access('pack3', 'BAZ')
503
memos.extend(access.add_raw_records([(b'key', 5)], b'alpha'))
505
transport = self.get_transport()
506
access = pack_repo._DirectPackAccess({"FOO": (transport, 'packfile'),
507
"FOOBAR": (transport, 'pack2'),
508
"BAZ": (transport, 'pack3')})
509
self.assertEqual([b'1234567890', b'12345', b'alpha'],
510
list(access.get_raw_records(memos)))
511
self.assertEqual([b'1234567890'],
512
list(access.get_raw_records(memos[0:1])))
513
self.assertEqual([b'12345'],
514
list(access.get_raw_records(memos[1:2])))
515
self.assertEqual([b'alpha'],
516
list(access.get_raw_records(memos[2:3])))
517
self.assertEqual([b'1234567890', b'alpha'],
518
list(access.get_raw_records(memos[0:1] + memos[2:3])))
520
def test_set_writer(self):
521
"""The writer should be settable post construction."""
522
access = pack_repo._DirectPackAccess({})
523
transport = self.get_transport()
524
packname = 'packfile'
527
def write_data(bytes):
528
transport.append_bytes(packname, bytes)
529
writer = pack.ContainerWriter(write_data)
531
access.set_writer(writer, index, (transport, packname))
532
memos = access.add_raw_records([(b'key', 10)], b'1234567890')
534
self.assertEqual([b'1234567890'], list(access.get_raw_records(memos)))
536
def test_missing_index_raises_retry(self):
537
memos = self.make_pack_file()
538
transport = self.get_transport()
539
reload_called, reload_func = self.make_reload_func()
540
# Note that the index key has changed from 'foo' to 'bar'
541
access = pack_repo._DirectPackAccess({'bar': (transport, 'packname')},
542
reload_func=reload_func)
543
e = self.assertListRaises(errors.RetryWithNewPacks,
544
access.get_raw_records, memos)
545
# Because a key was passed in which does not match our index list, we
546
# assume that the listing was already reloaded
547
self.assertTrue(e.reload_occurred)
548
self.assertIsInstance(e.exc_info, tuple)
549
self.assertIs(e.exc_info[0], KeyError)
550
self.assertIsInstance(e.exc_info[1], KeyError)
552
def test_missing_index_raises_key_error_with_no_reload(self):
553
memos = self.make_pack_file()
554
transport = self.get_transport()
555
# Note that the index key has changed from 'foo' to 'bar'
556
access = pack_repo._DirectPackAccess({'bar': (transport, 'packname')})
557
e = self.assertListRaises(KeyError, access.get_raw_records, memos)
559
def test_missing_file_raises_retry(self):
560
memos = self.make_pack_file()
561
transport = self.get_transport()
562
reload_called, reload_func = self.make_reload_func()
563
# Note that the 'filename' has been changed to 'different-packname'
564
access = pack_repo._DirectPackAccess(
565
{'foo': (transport, 'different-packname')},
566
reload_func=reload_func)
567
e = self.assertListRaises(errors.RetryWithNewPacks,
568
access.get_raw_records, memos)
569
# The file has gone missing, so we assume we need to reload
570
self.assertFalse(e.reload_occurred)
571
self.assertIsInstance(e.exc_info, tuple)
572
self.assertIs(e.exc_info[0], errors.NoSuchFile)
573
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
574
self.assertEqual('different-packname', e.exc_info[1].path)
576
def test_missing_file_raises_no_such_file_with_no_reload(self):
577
memos = self.make_pack_file()
578
transport = self.get_transport()
579
# Note that the 'filename' has been changed to 'different-packname'
580
access = pack_repo._DirectPackAccess(
581
{'foo': (transport, 'different-packname')})
582
e = self.assertListRaises(errors.NoSuchFile,
583
access.get_raw_records, memos)
585
def test_failing_readv_raises_retry(self):
586
memos = self.make_pack_file()
587
transport = self.get_transport()
588
failing_transport = MockReadvFailingTransport(
589
[transport.get_bytes('packname')])
590
reload_called, reload_func = self.make_reload_func()
591
access = pack_repo._DirectPackAccess(
592
{'foo': (failing_transport, 'packname')},
593
reload_func=reload_func)
594
# Asking for a single record will not trigger the Mock failure
595
self.assertEqual([b'1234567890'],
596
list(access.get_raw_records(memos[:1])))
597
self.assertEqual([b'12345'],
598
list(access.get_raw_records(memos[1:2])))
599
# A multiple offset readv() will fail mid-way through
600
e = self.assertListRaises(errors.RetryWithNewPacks,
601
access.get_raw_records, memos)
602
# The file has gone missing, so we assume we need to reload
603
self.assertFalse(e.reload_occurred)
604
self.assertIsInstance(e.exc_info, tuple)
605
self.assertIs(e.exc_info[0], errors.NoSuchFile)
606
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
607
self.assertEqual('packname', e.exc_info[1].path)
609
def test_failing_readv_raises_no_such_file_with_no_reload(self):
610
memos = self.make_pack_file()
611
transport = self.get_transport()
612
failing_transport = MockReadvFailingTransport(
613
[transport.get_bytes('packname')])
614
reload_called, reload_func = self.make_reload_func()
615
access = pack_repo._DirectPackAccess(
616
{'foo': (failing_transport, 'packname')})
617
# Asking for a single record will not trigger the Mock failure
618
self.assertEqual([b'1234567890'],
619
list(access.get_raw_records(memos[:1])))
620
self.assertEqual([b'12345'],
621
list(access.get_raw_records(memos[1:2])))
622
# A multiple offset readv() will fail mid-way through
623
e = self.assertListRaises(errors.NoSuchFile,
624
access.get_raw_records, memos)
626
def test_reload_or_raise_no_reload(self):
627
access = pack_repo._DirectPackAccess({}, reload_func=None)
628
retry_exc = self.make_retry_exception()
629
# Without a reload_func, we will just re-raise the original exception
630
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
632
def test_reload_or_raise_reload_changed(self):
633
reload_called, reload_func = self.make_reload_func(return_val=True)
634
access = pack_repo._DirectPackAccess({}, reload_func=reload_func)
635
retry_exc = self.make_retry_exception()
636
access.reload_or_raise(retry_exc)
637
self.assertEqual([1], reload_called)
638
retry_exc.reload_occurred = True
639
access.reload_or_raise(retry_exc)
640
self.assertEqual([2], reload_called)
642
def test_reload_or_raise_reload_no_change(self):
643
reload_called, reload_func = self.make_reload_func(return_val=False)
644
access = pack_repo._DirectPackAccess({}, reload_func=reload_func)
645
retry_exc = self.make_retry_exception()
646
# If reload_occurred is False, then we consider it an error to have
647
# reload_func() return False (no changes).
648
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
649
self.assertEqual([1], reload_called)
650
retry_exc.reload_occurred = True
651
# If reload_occurred is True, then we assume nothing changed because
652
# it had changed earlier, but didn't change again
653
access.reload_or_raise(retry_exc)
654
self.assertEqual([2], reload_called)
656
def test_annotate_retries(self):
657
vf, reload_counter = self.make_vf_for_retrying()
658
# It is a little bit bogus to annotate the Revision VF, but it works,
659
# as we have ancestry stored there
661
reload_lines = vf.annotate(key)
662
self.assertEqual([1, 1, 0], reload_counter)
663
plain_lines = vf.annotate(key)
664
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
665
if reload_lines != plain_lines:
666
self.fail('Annotation was not identical with reloading.')
667
# Now delete the packs-in-use, which should trigger another reload, but
668
# this time we just raise an exception because we can't recover
669
for trans, name in vf._access._indices.values():
671
self.assertRaises(errors.NoSuchFile, vf.annotate, key)
672
self.assertEqual([2, 1, 1], reload_counter)
674
def test__get_record_map_retries(self):
675
vf, reload_counter = self.make_vf_for_retrying()
676
keys = [(b'rev-1',), (b'rev-2',), (b'rev-3',)]
677
records = vf._get_record_map(keys)
678
self.assertEqual(keys, sorted(records.keys()))
679
self.assertEqual([1, 1, 0], reload_counter)
680
# Now delete the packs-in-use, which should trigger another reload, but
681
# this time we just raise an exception because we can't recover
682
for trans, name in vf._access._indices.values():
684
self.assertRaises(errors.NoSuchFile, vf._get_record_map, keys)
685
self.assertEqual([2, 1, 1], reload_counter)
687
def test_get_record_stream_retries(self):
688
vf, reload_counter = self.make_vf_for_retrying()
689
keys = [(b'rev-1',), (b'rev-2',), (b'rev-3',)]
690
record_stream = vf.get_record_stream(keys, 'topological', False)
691
record = next(record_stream)
692
self.assertEqual((b'rev-1',), record.key)
693
self.assertEqual([0, 0, 0], reload_counter)
694
record = next(record_stream)
695
self.assertEqual((b'rev-2',), record.key)
696
self.assertEqual([1, 1, 0], reload_counter)
697
record = next(record_stream)
698
self.assertEqual((b'rev-3',), record.key)
699
self.assertEqual([1, 1, 0], reload_counter)
700
# Now delete all pack files, and see that we raise the right error
701
for trans, name in vf._access._indices.values():
703
self.assertListRaises(errors.NoSuchFile,
704
vf.get_record_stream, keys, 'topological', False)
706
def test_iter_lines_added_or_present_in_keys_retries(self):
707
vf, reload_counter = self.make_vf_for_retrying()
708
keys = [(b'rev-1',), (b'rev-2',), (b'rev-3',)]
709
# Unfortunately, iter_lines_added_or_present_in_keys iterates the
710
# result in random order (determined by the iteration order from a
711
# set()), so we don't have any solid way to trigger whether data is
712
# read before or after. However we tried to delete the middle node to
713
# exercise the code well.
714
# What we care about is that all lines are always yielded, but not
717
reload_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
718
self.assertEqual([1, 1, 0], reload_counter)
719
# Now do it again, to make sure the result is equivalent
720
plain_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
721
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
722
self.assertEqual(plain_lines, reload_lines)
723
self.assertEqual(21, len(plain_lines))
724
# Now delete all pack files, and see that we raise the right error
725
for trans, name in vf._access._indices.values():
727
self.assertListRaises(errors.NoSuchFile,
728
vf.iter_lines_added_or_present_in_keys, keys)
729
self.assertEqual([2, 1, 1], reload_counter)
731
def test_get_record_stream_yields_disk_sorted_order(self):
732
# if we get 'unordered' pick a semi-optimal order for reading. The
733
# order should be grouped by pack file, and then by position in file
734
repo = self.make_repository('test', format='pack-0.92')
736
self.addCleanup(repo.unlock)
737
repo.start_write_group()
739
vf.add_lines((b'f-id', b'rev-5'), [(b'f-id', b'rev-4')], [b'lines\n'])
740
vf.add_lines((b'f-id', b'rev-1'), [], [b'lines\n'])
741
vf.add_lines((b'f-id', b'rev-2'), [(b'f-id', b'rev-1')], [b'lines\n'])
742
repo.commit_write_group()
743
# We inserted them as rev-5, rev-1, rev-2, we should get them back in
745
stream = vf.get_record_stream([(b'f-id', b'rev-1'), (b'f-id', b'rev-5'),
746
(b'f-id', b'rev-2')], 'unordered', False)
747
keys = [r.key for r in stream]
748
self.assertEqual([(b'f-id', b'rev-5'), (b'f-id', b'rev-1'),
749
(b'f-id', b'rev-2')], keys)
750
repo.start_write_group()
751
vf.add_lines((b'f-id', b'rev-4'), [(b'f-id', b'rev-3')], [b'lines\n'])
752
vf.add_lines((b'f-id', b'rev-3'), [(b'f-id', b'rev-2')], [b'lines\n'])
753
vf.add_lines((b'f-id', b'rev-6'), [(b'f-id', b'rev-5')], [b'lines\n'])
754
repo.commit_write_group()
755
# Request in random order, to make sure the output order isn't based on
757
request_keys = set((b'f-id', b'rev-%d' % i) for i in range(1, 7))
758
stream = vf.get_record_stream(request_keys, 'unordered', False)
759
keys = [r.key for r in stream]
760
# We want to get the keys back in disk order, but it doesn't matter
761
# which pack we read from first. So this can come back in 2 orders
762
alt1 = [(b'f-id', b'rev-%d' % i) for i in [4, 3, 6, 5, 1, 2]]
763
alt2 = [(b'f-id', b'rev-%d' % i) for i in [5, 1, 2, 4, 3, 6]]
764
if keys != alt1 and keys != alt2:
765
self.fail('Returned key order did not match either expected order.'
766
' expected %s or %s, not %s'
767
% (alt1, alt2, keys))
770
class LowLevelKnitDataTests(TestCase):
772
def create_gz_content(self, text):
774
with gzip.GzipFile(mode='wb', fileobj=sio) as gz_file:
776
return sio.getvalue()
778
def make_multiple_records(self):
779
"""Create the content for multiple records."""
780
sha1sum = osutils.sha_string(b'foo\nbar\n')
782
gz_txt = self.create_gz_content(b'version rev-id-1 2 %s\n'
787
record_1 = (0, len(gz_txt), sha1sum)
788
total_txt.append(gz_txt)
789
sha1sum = osutils.sha_string(b'baz\n')
790
gz_txt = self.create_gz_content(b'version rev-id-2 1 %s\n'
794
record_2 = (record_1[1], len(gz_txt), sha1sum)
795
total_txt.append(gz_txt)
796
return total_txt, record_1, record_2
798
def test_valid_knit_data(self):
799
sha1sum = osutils.sha_string(b'foo\nbar\n')
800
gz_txt = self.create_gz_content(b'version rev-id-1 2 %s\n'
805
transport = MockTransport([gz_txt])
806
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
807
knit = KnitVersionedFiles(None, access)
808
records = [((b'rev-id-1',), ((b'rev-id-1',), 0, len(gz_txt)))]
810
contents = list(knit._read_records_iter(records))
811
self.assertEqual([((b'rev-id-1',), [b'foo\n', b'bar\n'],
812
b'4e48e2c9a3d2ca8a708cb0cc545700544efb5021')], contents)
814
raw_contents = list(knit._read_records_iter_raw(records))
815
self.assertEqual([((b'rev-id-1',), gz_txt, sha1sum)], raw_contents)
817
def test_multiple_records_valid(self):
818
total_txt, record_1, record_2 = self.make_multiple_records()
819
transport = MockTransport([b''.join(total_txt)])
820
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
821
knit = KnitVersionedFiles(None, access)
822
records = [((b'rev-id-1',), ((b'rev-id-1',), record_1[0], record_1[1])),
823
((b'rev-id-2',), ((b'rev-id-2',), record_2[0], record_2[1]))]
825
contents = list(knit._read_records_iter(records))
826
self.assertEqual([((b'rev-id-1',), [b'foo\n', b'bar\n'], record_1[2]),
827
((b'rev-id-2',), [b'baz\n'], record_2[2])],
830
raw_contents = list(knit._read_records_iter_raw(records))
831
self.assertEqual([((b'rev-id-1',), total_txt[0], record_1[2]),
832
((b'rev-id-2',), total_txt[1], record_2[2])],
835
def test_not_enough_lines(self):
836
sha1sum = osutils.sha_string(b'foo\n')
837
# record says 2 lines data says 1
838
gz_txt = self.create_gz_content(b'version rev-id-1 2 %s\n'
842
transport = MockTransport([gz_txt])
843
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
844
knit = KnitVersionedFiles(None, access)
845
records = [((b'rev-id-1',), ((b'rev-id-1',), 0, len(gz_txt)))]
846
self.assertRaises(KnitCorrupt, list,
847
knit._read_records_iter(records))
849
# read_records_iter_raw won't detect that sort of mismatch/corruption
850
raw_contents = list(knit._read_records_iter_raw(records))
851
self.assertEqual([((b'rev-id-1',), gz_txt, sha1sum)], raw_contents)
853
def test_too_many_lines(self):
854
sha1sum = osutils.sha_string(b'foo\nbar\n')
855
# record says 1 lines data says 2
856
gz_txt = self.create_gz_content(b'version rev-id-1 1 %s\n'
861
transport = MockTransport([gz_txt])
862
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
863
knit = KnitVersionedFiles(None, access)
864
records = [((b'rev-id-1',), ((b'rev-id-1',), 0, len(gz_txt)))]
865
self.assertRaises(KnitCorrupt, list,
866
knit._read_records_iter(records))
868
# read_records_iter_raw won't detect that sort of mismatch/corruption
869
raw_contents = list(knit._read_records_iter_raw(records))
870
self.assertEqual([((b'rev-id-1',), gz_txt, sha1sum)], raw_contents)
872
def test_mismatched_version_id(self):
873
sha1sum = osutils.sha_string(b'foo\nbar\n')
874
gz_txt = self.create_gz_content(b'version rev-id-1 2 %s\n'
879
transport = MockTransport([gz_txt])
880
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
881
knit = KnitVersionedFiles(None, access)
882
# We are asking for rev-id-2, but the data is rev-id-1
883
records = [((b'rev-id-2',), ((b'rev-id-2',), 0, len(gz_txt)))]
884
self.assertRaises(KnitCorrupt, list,
885
knit._read_records_iter(records))
887
# read_records_iter_raw detects mismatches in the header
888
self.assertRaises(KnitCorrupt, list,
889
knit._read_records_iter_raw(records))
891
def test_uncompressed_data(self):
892
sha1sum = osutils.sha_string(b'foo\nbar\n')
893
txt = (b'version rev-id-1 2 %s\n'
898
transport = MockTransport([txt])
899
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
900
knit = KnitVersionedFiles(None, access)
901
records = [((b'rev-id-1',), ((b'rev-id-1',), 0, len(txt)))]
903
# We don't have valid gzip data ==> corrupt
904
self.assertRaises(KnitCorrupt, list,
905
knit._read_records_iter(records))
907
# read_records_iter_raw will notice the bad data
908
self.assertRaises(KnitCorrupt, list,
909
knit._read_records_iter_raw(records))
911
def test_corrupted_data(self):
912
sha1sum = osutils.sha_string(b'foo\nbar\n')
913
gz_txt = self.create_gz_content(b'version rev-id-1 2 %s\n'
918
# Change 2 bytes in the middle to \xff
919
gz_txt = gz_txt[:10] + b'\xff\xff' + gz_txt[12:]
920
transport = MockTransport([gz_txt])
921
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
922
knit = KnitVersionedFiles(None, access)
923
records = [((b'rev-id-1',), ((b'rev-id-1',), 0, len(gz_txt)))]
924
self.assertRaises(KnitCorrupt, list,
925
knit._read_records_iter(records))
926
# read_records_iter_raw will barf on bad gz data
927
self.assertRaises(KnitCorrupt, list,
928
knit._read_records_iter_raw(records))
931
class LowLevelKnitIndexTests(TestCase):
934
def _load_data(self):
935
from ..bzr._knit_load_data_py import _load_data_py
938
def get_knit_index(self, transport, name, mode):
939
mapper = ConstantMapper(name)
940
self.overrideAttr(knit, '_load_data', self._load_data)
944
return _KndxIndex(transport, mapper, lambda: None, allow_writes, lambda: True)
946
def test_create_file(self):
947
transport = MockTransport()
948
index = self.get_knit_index(transport, "filename", "w")
950
call = transport.calls.pop(0)
951
# call[1][1] is a BytesIO - we can't test it by simple equality.
952
self.assertEqual('put_file_non_atomic', call[0])
953
self.assertEqual('filename.kndx', call[1][0])
954
# With no history, _KndxIndex writes a new index:
955
self.assertEqual(_KndxIndex.HEADER,
956
call[1][1].getvalue())
957
self.assertEqual({'create_parent_dir': True}, call[2])
959
def test_read_utf8_version_id(self):
960
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
961
utf8_revision_id = unicode_revision_id.encode('utf-8')
962
transport = MockTransport([
964
b'%s option 0 1 :' % (utf8_revision_id,)
966
index = self.get_knit_index(transport, "filename", "r")
967
# _KndxIndex is a private class, and deals in utf8 revision_ids, not
968
# Unicode revision_ids.
969
self.assertEqual({(utf8_revision_id,): ()},
970
index.get_parent_map(index.keys()))
971
self.assertFalse((unicode_revision_id,) in index.keys())
973
def test_read_utf8_parents(self):
974
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
975
utf8_revision_id = unicode_revision_id.encode('utf-8')
976
transport = MockTransport([
978
b"version option 0 1 .%s :" % (utf8_revision_id,)
980
index = self.get_knit_index(transport, "filename", "r")
981
self.assertEqual({(b"version",): ((utf8_revision_id,),)},
982
index.get_parent_map(index.keys()))
984
def test_read_ignore_corrupted_lines(self):
985
transport = MockTransport([
988
b"corrupted options 0 1 .b .c ",
989
b"version options 0 1 :"
991
index = self.get_knit_index(transport, "filename", "r")
992
self.assertEqual(1, len(index.keys()))
993
self.assertEqual({(b"version",)}, index.keys())
995
def test_read_corrupted_header(self):
996
transport = MockTransport([b'not a bzr knit index header\n'])
997
index = self.get_knit_index(transport, "filename", "r")
998
self.assertRaises(KnitHeaderError, index.keys)
1000
def test_read_duplicate_entries(self):
1001
transport = MockTransport([
1003
b"parent options 0 1 :",
1004
b"version options1 0 1 0 :",
1005
b"version options2 1 2 .other :",
1006
b"version options3 3 4 0 .other :"
1008
index = self.get_knit_index(transport, "filename", "r")
1009
self.assertEqual(2, len(index.keys()))
1010
# check that the index used is the first one written. (Specific
1011
# to KnitIndex style indices.
1012
self.assertEqual(b"1", index._dictionary_compress([(b"version",)]))
1013
self.assertEqual(((b"version",), 3, 4),
1014
index.get_position((b"version",)))
1015
self.assertEqual([b"options3"], index.get_options((b"version",)))
1016
self.assertEqual({(b"version",): ((b"parent",), (b"other",))},
1017
index.get_parent_map([(b"version",)]))
1019
def test_read_compressed_parents(self):
1020
transport = MockTransport([
1023
b"b option 0 1 0 :",
1024
b"c option 0 1 1 0 :",
1026
index = self.get_knit_index(transport, "filename", "r")
1027
self.assertEqual({(b"b",): ((b"a",),), (b"c",): ((b"b",), (b"a",))},
1028
index.get_parent_map([(b"b",), (b"c",)]))
1030
def test_write_utf8_version_id(self):
1031
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
1032
utf8_revision_id = unicode_revision_id.encode('utf-8')
1033
transport = MockTransport([
1036
index = self.get_knit_index(transport, "filename", "r")
1038
((utf8_revision_id,), [b"option"], ((utf8_revision_id,), 0, 1), [])])
1039
call = transport.calls.pop(0)
1040
# call[1][1] is a BytesIO - we can't test it by simple equality.
1041
self.assertEqual('put_file_non_atomic', call[0])
1042
self.assertEqual('filename.kndx', call[1][0])
1043
# With no history, _KndxIndex writes a new index:
1044
self.assertEqual(_KndxIndex.HEADER +
1045
b"\n%s option 0 1 :" % (utf8_revision_id,),
1046
call[1][1].getvalue())
1047
self.assertEqual({'create_parent_dir': True}, call[2])
1049
def test_write_utf8_parents(self):
1050
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
1051
utf8_revision_id = unicode_revision_id.encode('utf-8')
1052
transport = MockTransport([
1055
index = self.get_knit_index(transport, "filename", "r")
1057
((b"version",), [b"option"], ((b"version",), 0, 1), [(utf8_revision_id,)])])
1058
call = transport.calls.pop(0)
1059
# call[1][1] is a BytesIO - we can't test it by simple equality.
1060
self.assertEqual('put_file_non_atomic', call[0])
1061
self.assertEqual('filename.kndx', call[1][0])
1062
# With no history, _KndxIndex writes a new index:
1063
self.assertEqual(_KndxIndex.HEADER +
1064
b"\nversion option 0 1 .%s :" % (utf8_revision_id,),
1065
call[1][1].getvalue())
1066
self.assertEqual({'create_parent_dir': True}, call[2])
1068
def test_keys(self):
1069
transport = MockTransport([
1072
index = self.get_knit_index(transport, "filename", "r")
1074
self.assertEqual(set(), index.keys())
1076
index.add_records([((b"a",), [b"option"], ((b"a",), 0, 1), [])])
1077
self.assertEqual({(b"a",)}, index.keys())
1079
index.add_records([((b"a",), [b"option"], ((b"a",), 0, 1), [])])
1080
self.assertEqual({(b"a",)}, index.keys())
1082
index.add_records([((b"b",), [b"option"], ((b"b",), 0, 1), [])])
1083
self.assertEqual({(b"a",), (b"b",)}, index.keys())
1085
def add_a_b(self, index, random_id=None):
1087
if random_id is not None:
1088
kwargs["random_id"] = random_id
1090
((b"a",), [b"option"], ((b"a",), 0, 1), [(b"b",)]),
1091
((b"a",), [b"opt"], ((b"a",), 1, 2), [(b"c",)]),
1092
((b"b",), [b"option"], ((b"b",), 2, 3), [(b"a",)])
1095
def assertIndexIsAB(self, index):
1097
(b'a',): ((b'c',),),
1098
(b'b',): ((b'a',),),
1100
index.get_parent_map(index.keys()))
1101
self.assertEqual(((b"a",), 1, 2), index.get_position((b"a",)))
1102
self.assertEqual(((b"b",), 2, 3), index.get_position((b"b",)))
1103
self.assertEqual([b"opt"], index.get_options((b"a",)))
1105
def test_add_versions(self):
1106
transport = MockTransport([
1109
index = self.get_knit_index(transport, "filename", "r")
1112
call = transport.calls.pop(0)
1113
# call[1][1] is a BytesIO - we can't test it by simple equality.
1114
self.assertEqual('put_file_non_atomic', call[0])
1115
self.assertEqual('filename.kndx', call[1][0])
1116
# With no history, _KndxIndex writes a new index:
1119
b"\na option 0 1 .b :"
1121
b"\nb option 2 3 0 :",
1122
call[1][1].getvalue())
1123
self.assertEqual({'create_parent_dir': True}, call[2])
1124
self.assertIndexIsAB(index)
1126
def test_add_versions_random_id_is_accepted(self):
1127
transport = MockTransport([
1130
index = self.get_knit_index(transport, "filename", "r")
1131
self.add_a_b(index, random_id=True)
1133
def test_delay_create_and_add_versions(self):
1134
transport = MockTransport()
1136
index = self.get_knit_index(transport, "filename", "w")
1138
self.assertEqual([], transport.calls)
1141
# [ {"dir_mode": 0777, "create_parent_dir": True, "mode": "wb"},
1143
# Two calls: one during which we load the existing index (and when its
1144
# missing create it), then a second where we write the contents out.
1145
self.assertEqual(2, len(transport.calls))
1146
call = transport.calls.pop(0)
1147
self.assertEqual('put_file_non_atomic', call[0])
1148
self.assertEqual('filename.kndx', call[1][0])
1149
# With no history, _KndxIndex writes a new index:
1150
self.assertEqual(_KndxIndex.HEADER, call[1][1].getvalue())
1151
self.assertEqual({'create_parent_dir': True}, call[2])
1152
call = transport.calls.pop(0)
1153
# call[1][1] is a BytesIO - we can't test it by simple equality.
1154
self.assertEqual('put_file_non_atomic', call[0])
1155
self.assertEqual('filename.kndx', call[1][0])
1156
# With no history, _KndxIndex writes a new index:
1159
b"\na option 0 1 .b :"
1161
b"\nb option 2 3 0 :",
1162
call[1][1].getvalue())
1163
self.assertEqual({'create_parent_dir': True}, call[2])
1165
def assertTotalBuildSize(self, size, keys, positions):
1166
self.assertEqual(size,
1167
knit._get_total_build_size(None, keys, positions))
1169
def test__get_total_build_size(self):
1171
(b'a',): (('fulltext', False), ((b'a',), 0, 100), None),
1172
(b'b',): (('line-delta', False), ((b'b',), 100, 21), (b'a',)),
1173
(b'c',): (('line-delta', False), ((b'c',), 121, 35), (b'b',)),
1174
(b'd',): (('line-delta', False), ((b'd',), 156, 12), (b'b',)),
1176
self.assertTotalBuildSize(100, [(b'a',)], positions)
1177
self.assertTotalBuildSize(121, [(b'b',)], positions)
1178
# c needs both a & b
1179
self.assertTotalBuildSize(156, [(b'c',)], positions)
1180
# we shouldn't count 'b' twice
1181
self.assertTotalBuildSize(156, [(b'b',), (b'c',)], positions)
1182
self.assertTotalBuildSize(133, [(b'd',)], positions)
1183
self.assertTotalBuildSize(168, [(b'c',), (b'd',)], positions)
1185
def test_get_position(self):
1186
transport = MockTransport([
1191
index = self.get_knit_index(transport, "filename", "r")
1193
self.assertEqual(((b"a",), 0, 1), index.get_position((b"a",)))
1194
self.assertEqual(((b"b",), 1, 2), index.get_position((b"b",)))
1196
def test_get_method(self):
1197
transport = MockTransport([
1199
b"a fulltext,unknown 0 1 :",
1200
b"b unknown,line-delta 1 2 :",
1203
index = self.get_knit_index(transport, "filename", "r")
1205
self.assertEqual("fulltext", index.get_method(b"a"))
1206
self.assertEqual("line-delta", index.get_method(b"b"))
1207
self.assertRaises(knit.KnitIndexUnknownMethod, index.get_method, b"c")
1209
def test_get_options(self):
1210
transport = MockTransport([
1213
b"b opt2,opt3 1 2 :"
1215
index = self.get_knit_index(transport, "filename", "r")
1217
self.assertEqual([b"opt1"], index.get_options(b"a"))
1218
self.assertEqual([b"opt2", b"opt3"], index.get_options(b"b"))
1220
def test_get_parent_map(self):
1221
transport = MockTransport([
1224
b"b option 1 2 0 .c :",
1225
b"c option 1 2 1 0 .e :"
1227
index = self.get_knit_index(transport, "filename", "r")
1231
(b"b",): ((b"a",), (b"c",)),
1232
(b"c",): ((b"b",), (b"a",), (b"e",)),
1233
}, index.get_parent_map(index.keys()))
1235
def test_impossible_parent(self):
1236
"""Test we get KnitCorrupt if the parent couldn't possibly exist."""
1237
transport = MockTransport([
1240
b"b option 0 1 4 :" # We don't have a 4th record
1242
index = self.get_knit_index(transport, 'filename', 'r')
1243
self.assertRaises(KnitCorrupt, index.keys)
1245
def test_corrupted_parent(self):
1246
transport = MockTransport([
1250
b"c option 0 1 1v :", # Can't have a parent of '1v'
1252
index = self.get_knit_index(transport, 'filename', 'r')
1253
self.assertRaises(KnitCorrupt, index.keys)
1255
def test_corrupted_parent_in_list(self):
1256
transport = MockTransport([
1260
b"c option 0 1 1 v :", # Can't have a parent of 'v'
1262
index = self.get_knit_index(transport, 'filename', 'r')
1263
self.assertRaises(KnitCorrupt, index.keys)
1265
def test_invalid_position(self):
1266
transport = MockTransport([
1270
index = self.get_knit_index(transport, 'filename', 'r')
1271
self.assertRaises(KnitCorrupt, index.keys)
1273
def test_invalid_size(self):
1274
transport = MockTransport([
1278
index = self.get_knit_index(transport, 'filename', 'r')
1279
self.assertRaises(KnitCorrupt, index.keys)
1281
def test_scan_unvalidated_index_not_implemented(self):
1282
transport = MockTransport()
1283
index = self.get_knit_index(transport, 'filename', 'r')
1285
NotImplementedError, index.scan_unvalidated_index,
1286
'dummy graph_index')
1288
NotImplementedError, index.get_missing_compression_parents)
1290
def test_short_line(self):
1291
transport = MockTransport([
1294
b"b option 10 10 0", # This line isn't terminated, ignored
1296
index = self.get_knit_index(transport, "filename", "r")
1297
self.assertEqual({(b'a',)}, index.keys())
1299
def test_skip_incomplete_record(self):
1300
# A line with bogus data should just be skipped
1301
transport = MockTransport([
1304
b"b option 10 10 0", # This line isn't terminated, ignored
1305
b"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1307
index = self.get_knit_index(transport, "filename", "r")
1308
self.assertEqual({(b'a',), (b'c',)}, index.keys())
1310
def test_trailing_characters(self):
1311
# A line with bogus data should just be skipped
1312
transport = MockTransport([
1315
b"b option 10 10 0 :a", # This line has extra trailing characters
1316
b"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1318
index = self.get_knit_index(transport, "filename", "r")
1319
self.assertEqual({(b'a',), (b'c',)}, index.keys())
1322
class LowLevelKnitIndexTests_c(LowLevelKnitIndexTests):
1324
_test_needs_features = [compiled_knit_feature]
1327
def _load_data(self):
1328
from ..bzr._knit_load_data_pyx import _load_data_c
1332
class Test_KnitAnnotator(TestCaseWithMemoryTransport):
1334
def make_annotator(self):
1335
factory = knit.make_pack_factory(True, True, 1)
1336
vf = factory(self.get_transport())
1337
return knit._KnitAnnotator(vf)
1339
def test__expand_fulltext(self):
1340
ann = self.make_annotator()
1341
rev_key = (b'rev-id',)
1342
ann._num_compression_children[rev_key] = 1
1343
res = ann._expand_record(rev_key, ((b'parent-id',),), None,
1344
[b'line1\n', b'line2\n'], ('fulltext', True))
1345
# The content object and text lines should be cached appropriately
1346
self.assertEqual([b'line1\n', b'line2'], res)
1347
content_obj = ann._content_objects[rev_key]
1348
self.assertEqual([b'line1\n', b'line2\n'], content_obj._lines)
1349
self.assertEqual(res, content_obj.text())
1350
self.assertEqual(res, ann._text_cache[rev_key])
1352
def test__expand_delta_comp_parent_not_available(self):
1353
# Parent isn't available yet, so we return nothing, but queue up this
1354
# node for later processing
1355
ann = self.make_annotator()
1356
rev_key = (b'rev-id',)
1357
parent_key = (b'parent-id',)
1358
record = [b'0,1,1\n', b'new-line\n']
1359
details = ('line-delta', False)
1360
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1362
self.assertEqual(None, res)
1363
self.assertTrue(parent_key in ann._pending_deltas)
1364
pending = ann._pending_deltas[parent_key]
1365
self.assertEqual(1, len(pending))
1366
self.assertEqual((rev_key, (parent_key,), record, details), pending[0])
1368
def test__expand_record_tracks_num_children(self):
1369
ann = self.make_annotator()
1370
rev_key = (b'rev-id',)
1371
rev2_key = (b'rev2-id',)
1372
parent_key = (b'parent-id',)
1373
record = [b'0,1,1\n', b'new-line\n']
1374
details = ('line-delta', False)
1375
ann._num_compression_children[parent_key] = 2
1376
ann._expand_record(parent_key, (), None, [b'line1\n', b'line2\n'],
1377
('fulltext', False))
1378
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1380
self.assertEqual({parent_key: 1}, ann._num_compression_children)
1381
# Expanding the second child should remove the content object, and the
1382
# num_compression_children entry
1383
res = ann._expand_record(rev2_key, (parent_key,), parent_key,
1385
self.assertFalse(parent_key in ann._content_objects)
1386
self.assertEqual({}, ann._num_compression_children)
1387
# We should not cache the content_objects for rev2 and rev, because
1388
# they do not have compression children of their own.
1389
self.assertEqual({}, ann._content_objects)
1391
def test__expand_delta_records_blocks(self):
1392
ann = self.make_annotator()
1393
rev_key = (b'rev-id',)
1394
parent_key = (b'parent-id',)
1395
record = [b'0,1,1\n', b'new-line\n']
1396
details = ('line-delta', True)
1397
ann._num_compression_children[parent_key] = 2
1398
ann._expand_record(parent_key, (), None,
1399
[b'line1\n', b'line2\n', b'line3\n'],
1400
('fulltext', False))
1401
ann._expand_record(rev_key, (parent_key,), parent_key, record, details)
1402
self.assertEqual({(rev_key, parent_key): [(1, 1, 1), (3, 3, 0)]},
1403
ann._matching_blocks)
1404
rev2_key = (b'rev2-id',)
1405
record = [b'0,1,1\n', b'new-line\n']
1406
details = ('line-delta', False)
1407
ann._expand_record(rev2_key, (parent_key,),
1408
parent_key, record, details)
1409
self.assertEqual([(1, 1, 2), (3, 3, 0)],
1410
ann._matching_blocks[(rev2_key, parent_key)])
1412
def test__get_parent_ann_uses_matching_blocks(self):
1413
ann = self.make_annotator()
1414
rev_key = (b'rev-id',)
1415
parent_key = (b'parent-id',)
1416
parent_ann = [(parent_key,)] * 3
1417
block_key = (rev_key, parent_key)
1418
ann._annotations_cache[parent_key] = parent_ann
1419
ann._matching_blocks[block_key] = [(0, 1, 1), (3, 3, 0)]
1420
# We should not try to access any parent_lines content, because we know
1421
# we already have the matching blocks
1422
par_ann, blocks = ann._get_parent_annotations_and_matches(rev_key,
1423
[b'1\n', b'2\n', b'3\n'], parent_key)
1424
self.assertEqual(parent_ann, par_ann)
1425
self.assertEqual([(0, 1, 1), (3, 3, 0)], blocks)
1426
self.assertEqual({}, ann._matching_blocks)
1428
def test__process_pending(self):
1429
ann = self.make_annotator()
1430
rev_key = (b'rev-id',)
1431
p1_key = (b'p1-id',)
1432
p2_key = (b'p2-id',)
1433
record = [b'0,1,1\n', b'new-line\n']
1434
details = ('line-delta', False)
1435
p1_record = [b'line1\n', b'line2\n']
1436
ann._num_compression_children[p1_key] = 1
1437
res = ann._expand_record(rev_key, (p1_key, p2_key), p1_key,
1439
self.assertEqual(None, res)
1440
# self.assertTrue(p1_key in ann._pending_deltas)
1441
self.assertEqual({}, ann._pending_annotation)
1442
# Now insert p1, and we should be able to expand the delta
1443
res = ann._expand_record(p1_key, (), None, p1_record,
1444
('fulltext', False))
1445
self.assertEqual(p1_record, res)
1446
ann._annotations_cache[p1_key] = [(p1_key,)] * 2
1447
res = ann._process_pending(p1_key)
1448
self.assertEqual([], res)
1449
self.assertFalse(p1_key in ann._pending_deltas)
1450
self.assertTrue(p2_key in ann._pending_annotation)
1451
self.assertEqual({p2_key: [(rev_key, (p1_key, p2_key))]},
1452
ann._pending_annotation)
1453
# Now fill in parent 2, and pending annotation should be satisfied
1454
res = ann._expand_record(p2_key, (), None, [], ('fulltext', False))
1455
ann._annotations_cache[p2_key] = []
1456
res = ann._process_pending(p2_key)
1457
self.assertEqual([rev_key], res)
1458
self.assertEqual({}, ann._pending_annotation)
1459
self.assertEqual({}, ann._pending_deltas)
1461
def test_record_delta_removes_basis(self):
1462
ann = self.make_annotator()
1463
ann._expand_record((b'parent-id',), (), None,
1464
[b'line1\n', b'line2\n'], ('fulltext', False))
1465
ann._num_compression_children[b'parent-id'] = 2
1467
def test_annotate_special_text(self):
1468
ann = self.make_annotator()
1470
rev1_key = (b'rev-1',)
1471
rev2_key = (b'rev-2',)
1472
rev3_key = (b'rev-3',)
1473
spec_key = (b'special:',)
1474
vf.add_lines(rev1_key, [], [b'initial content\n'])
1475
vf.add_lines(rev2_key, [rev1_key], [b'initial content\n',
1476
b'common content\n',
1478
vf.add_lines(rev3_key, [rev1_key], [b'initial content\n',
1479
b'common content\n',
1481
spec_text = (b'initial content\n'
1485
ann.add_special_text(spec_key, [rev2_key, rev3_key], spec_text)
1486
anns, lines = ann.annotate(spec_key)
1487
self.assertEqual([(rev1_key,),
1488
(rev2_key, rev3_key),
1492
self.assertEqualDiff(spec_text, b''.join(lines))
1495
class KnitTests(TestCaseWithTransport):
1496
"""Class containing knit test helper routines."""
1498
def make_test_knit(self, annotate=False, name='test'):
1499
mapper = ConstantMapper(name)
1500
return make_file_factory(annotate, mapper)(self.get_transport())
1503
class TestBadShaError(KnitTests):
1504
"""Tests for handling of sha errors."""
1506
def test_sha_exception_has_text(self):
1507
# having the failed text included in the error allows for recovery.
1508
source = self.make_test_knit()
1509
target = self.make_test_knit(name="target")
1510
if not source._max_delta_chain:
1511
raise TestNotApplicable(
1512
"cannot get delta-caused sha failures without deltas.")
1515
broken = (b'broken',)
1516
source.add_lines(basis, (), [b'foo\n'])
1517
source.add_lines(broken, (basis,), [b'foo\n', b'bar\n'])
1518
# Seed target with a bad basis text
1519
target.add_lines(basis, (), [b'gam\n'])
1520
target.insert_record_stream(
1521
source.get_record_stream([broken], 'unordered', False))
1522
err = self.assertRaises(KnitCorrupt,
1523
next(target.get_record_stream([broken], 'unordered', True
1524
)).get_bytes_as, 'chunked')
1525
self.assertEqual([b'gam\n', b'bar\n'], err.content)
1526
# Test for formatting with live data
1527
self.assertStartsWith(str(err), "Knit ")
1530
class TestKnitIndex(KnitTests):
1532
def test_add_versions_dictionary_compresses(self):
1533
"""Adding versions to the index should update the lookup dict"""
1534
knit = self.make_test_knit()
1536
idx.add_records([((b'a-1',), [b'fulltext'], ((b'a-1',), 0, 0), [])])
1537
self.check_file_contents('test.kndx',
1538
b'# bzr knit index 8\n'
1540
b'a-1 fulltext 0 0 :'
1543
((b'a-2',), [b'fulltext'], ((b'a-2',), 0, 0), [(b'a-1',)]),
1544
((b'a-3',), [b'fulltext'], ((b'a-3',), 0, 0), [(b'a-2',)]),
1546
self.check_file_contents('test.kndx',
1547
b'# bzr knit index 8\n'
1549
b'a-1 fulltext 0 0 :\n'
1550
b'a-2 fulltext 0 0 0 :\n'
1551
b'a-3 fulltext 0 0 1 :'
1553
self.assertEqual({(b'a-3',), (b'a-1',), (b'a-2',)}, idx.keys())
1555
(b'a-1',): (((b'a-1',), 0, 0), None, (), ('fulltext', False)),
1556
(b'a-2',): (((b'a-2',), 0, 0), None, ((b'a-1',),), ('fulltext', False)),
1557
(b'a-3',): (((b'a-3',), 0, 0), None, ((b'a-2',),), ('fulltext', False)),
1558
}, idx.get_build_details(idx.keys()))
1559
self.assertEqual({(b'a-1',): (),
1560
(b'a-2',): ((b'a-1',),),
1561
(b'a-3',): ((b'a-2',),), },
1562
idx.get_parent_map(idx.keys()))
1564
def test_add_versions_fails_clean(self):
1565
"""If add_versions fails in the middle, it restores a pristine state.
1567
Any modifications that are made to the index are reset if all versions
1570
# This cheats a little bit by passing in a generator which will
1571
# raise an exception before the processing finishes
1572
# Other possibilities would be to have an version with the wrong number
1573
# of entries, or to make the backing transport unable to write any
1576
knit = self.make_test_knit()
1578
idx.add_records([((b'a-1',), [b'fulltext'], ((b'a-1',), 0, 0), [])])
1580
class StopEarly(Exception):
1583
def generate_failure():
1584
"""Add some entries and then raise an exception"""
1585
yield ((b'a-2',), [b'fulltext'], (None, 0, 0), (b'a-1',))
1586
yield ((b'a-3',), [b'fulltext'], (None, 0, 0), (b'a-2',))
1589
# Assert the pre-condition
1591
self.assertEqual({(b'a-1',)}, set(idx.keys()))
1593
{(b'a-1',): (((b'a-1',), 0, 0), None, (), ('fulltext', False))},
1594
idx.get_build_details([(b'a-1',)]))
1595
self.assertEqual({(b'a-1',): ()}, idx.get_parent_map(idx.keys()))
1598
self.assertRaises(StopEarly, idx.add_records, generate_failure())
1599
# And it shouldn't be modified
1602
def test_knit_index_ignores_empty_files(self):
1603
# There was a race condition in older bzr, where a ^C at the right time
1604
# could leave an empty .kndx file, which bzr would later claim was a
1605
# corrupted file since the header was not present. In reality, the file
1606
# just wasn't created, so it should be ignored.
1607
t = transport.get_transport_from_path('.')
1608
t.put_bytes('test.kndx', b'')
1610
knit = self.make_test_knit()
1612
def test_knit_index_checks_header(self):
1613
t = transport.get_transport_from_path('.')
1614
t.put_bytes('test.kndx', b'# not really a knit header\n\n')
1615
k = self.make_test_knit()
1616
self.assertRaises(KnitHeaderError, k.keys)
1619
class TestGraphIndexKnit(KnitTests):
1620
"""Tests for knits using a GraphIndex rather than a KnitIndex."""
1622
def make_g_index(self, name, ref_lists=0, nodes=[]):
1623
builder = GraphIndexBuilder(ref_lists)
1624
for node, references, value in nodes:
1625
builder.add_node(node, references, value)
1626
stream = builder.finish()
1627
trans = self.get_transport()
1628
size = trans.put_file(name, stream)
1629
return GraphIndex(trans, name, size)
1631
def two_graph_index(self, deltas=False, catch_adds=False):
1632
"""Build a two-graph index.
1634
:param deltas: If true, use underlying indices with two node-ref
1635
lists and 'parent' set to a delta-compressed against tail.
1637
# build a complex graph across several indices.
1639
# delta compression inn the index
1640
index1 = self.make_g_index('1', 2, [
1641
((b'tip', ), b'N0 100', ([(b'parent', )], [], )),
1642
((b'tail', ), b'', ([], []))])
1643
index2 = self.make_g_index('2', 2, [
1644
((b'parent', ), b' 100 78',
1645
([(b'tail', ), (b'ghost', )], [(b'tail', )])),
1646
((b'separate', ), b'', ([], []))])
1648
# just blob location and graph in the index.
1649
index1 = self.make_g_index('1', 1, [
1650
((b'tip', ), b'N0 100', ([(b'parent', )], )),
1651
((b'tail', ), b'', ([], ))])
1652
index2 = self.make_g_index('2', 1, [
1653
((b'parent', ), b' 100 78', ([(b'tail', ), (b'ghost', )], )),
1654
((b'separate', ), b'', ([], ))])
1655
combined_index = CombinedGraphIndex([index1, index2])
1657
self.combined_index = combined_index
1658
self.caught_entries = []
1659
add_callback = self.catch_add
1662
return _KnitGraphIndex(combined_index, lambda: True, deltas=deltas,
1663
add_callback=add_callback)
1665
def test_keys(self):
1666
index = self.two_graph_index()
1667
self.assertEqual({(b'tail',), (b'tip',), (b'parent',), (b'separate',)},
1670
def test_get_position(self):
1671
index = self.two_graph_index()
1673
(index._graph_index._indices[0], 0, 100), index.get_position((b'tip',)))
1675
(index._graph_index._indices[1], 100, 78), index.get_position((b'parent',)))
1677
def test_get_method_deltas(self):
1678
index = self.two_graph_index(deltas=True)
1679
self.assertEqual('fulltext', index.get_method((b'tip',)))
1680
self.assertEqual('line-delta', index.get_method((b'parent',)))
1682
def test_get_method_no_deltas(self):
1683
# check that the parent-history lookup is ignored with deltas=False.
1684
index = self.two_graph_index(deltas=False)
1685
self.assertEqual('fulltext', index.get_method((b'tip',)))
1686
self.assertEqual('fulltext', index.get_method((b'parent',)))
1688
def test_get_options_deltas(self):
1689
index = self.two_graph_index(deltas=True)
1690
self.assertEqual([b'fulltext', b'no-eol'],
1691
index.get_options((b'tip',)))
1692
self.assertEqual([b'line-delta'], index.get_options((b'parent',)))
1694
def test_get_options_no_deltas(self):
1695
# check that the parent-history lookup is ignored with deltas=False.
1696
index = self.two_graph_index(deltas=False)
1697
self.assertEqual([b'fulltext', b'no-eol'],
1698
index.get_options((b'tip',)))
1699
self.assertEqual([b'fulltext'], index.get_options((b'parent',)))
1701
def test_get_parent_map(self):
1702
index = self.two_graph_index()
1703
self.assertEqual({(b'parent',): ((b'tail',), (b'ghost',))},
1704
index.get_parent_map([(b'parent',), (b'ghost',)]))
1706
def catch_add(self, entries):
1707
self.caught_entries.append(entries)
1709
def test_add_no_callback_errors(self):
1710
index = self.two_graph_index()
1711
self.assertRaises(errors.ReadOnlyError, index.add_records,
1712
[((b'new',), b'fulltext,no-eol', (None, 50, 60), [b'separate'])])
1714
def test_add_version_smoke(self):
1715
index = self.two_graph_index(catch_adds=True)
1716
index.add_records([((b'new',), b'fulltext,no-eol', (None, 50, 60),
1718
self.assertEqual([[((b'new', ), b'N50 60', (((b'separate',),),))]],
1719
self.caught_entries)
1721
def test_add_version_delta_not_delta_index(self):
1722
index = self.two_graph_index(catch_adds=True)
1723
self.assertRaises(KnitCorrupt, index.add_records,
1724
[((b'new',), b'no-eol,line-delta', (None, 0, 100), [(b'parent',)])])
1725
self.assertEqual([], self.caught_entries)
1727
def test_add_version_same_dup(self):
1728
index = self.two_graph_index(catch_adds=True)
1729
# options can be spelt two different ways
1731
[((b'tip',), b'fulltext,no-eol', (None, 0, 100), [(b'parent',)])])
1733
[((b'tip',), b'no-eol,fulltext', (None, 0, 100), [(b'parent',)])])
1734
# position/length are ignored (because each pack could have fulltext or
1735
# delta, and be at a different position.
1736
index.add_records([((b'tip',), b'fulltext,no-eol', (None, 50, 100),
1738
index.add_records([((b'tip',), b'fulltext,no-eol', (None, 0, 1000),
1740
# but neither should have added data:
1741
self.assertEqual([[], [], [], []], self.caught_entries)
1743
def test_add_version_different_dup(self):
1744
index = self.two_graph_index(deltas=True, catch_adds=True)
1746
self.assertRaises(KnitCorrupt, index.add_records,
1747
[((b'tip',), b'line-delta', (None, 0, 100), [(b'parent',)])])
1748
self.assertRaises(KnitCorrupt, index.add_records,
1749
[((b'tip',), b'fulltext', (None, 0, 100), [(b'parent',)])])
1751
self.assertRaises(KnitCorrupt, index.add_records,
1752
[((b'tip',), b'fulltext,no-eol', (None, 0, 100), [])])
1753
self.assertEqual([], self.caught_entries)
1755
def test_add_versions_nodeltas(self):
1756
index = self.two_graph_index(catch_adds=True)
1758
((b'new',), b'fulltext,no-eol', (None, 50, 60), [(b'separate',)]),
1759
((b'new2',), b'fulltext', (None, 0, 6), [(b'new',)]),
1761
self.assertEqual([((b'new', ), b'N50 60', (((b'separate',),),)),
1762
((b'new2', ), b' 0 6', (((b'new',),),))],
1763
sorted(self.caught_entries[0]))
1764
self.assertEqual(1, len(self.caught_entries))
1766
def test_add_versions_deltas(self):
1767
index = self.two_graph_index(deltas=True, catch_adds=True)
1769
((b'new',), b'fulltext,no-eol', (None, 50, 60), [(b'separate',)]),
1770
((b'new2',), b'line-delta', (None, 0, 6), [(b'new',)]),
1772
self.assertEqual([((b'new', ), b'N50 60', (((b'separate',),), ())),
1773
((b'new2', ), b' 0 6', (((b'new',),), ((b'new',),), ))],
1774
sorted(self.caught_entries[0]))
1775
self.assertEqual(1, len(self.caught_entries))
1777
def test_add_versions_delta_not_delta_index(self):
1778
index = self.two_graph_index(catch_adds=True)
1779
self.assertRaises(KnitCorrupt, index.add_records,
1780
[((b'new',), b'no-eol,line-delta', (None, 0, 100), [(b'parent',)])])
1781
self.assertEqual([], self.caught_entries)
1783
def test_add_versions_random_id_accepted(self):
1784
index = self.two_graph_index(catch_adds=True)
1785
index.add_records([], random_id=True)
1787
def test_add_versions_same_dup(self):
1788
index = self.two_graph_index(catch_adds=True)
1789
# options can be spelt two different ways
1790
index.add_records([((b'tip',), b'fulltext,no-eol', (None, 0, 100),
1792
index.add_records([((b'tip',), b'no-eol,fulltext', (None, 0, 100),
1794
# position/length are ignored (because each pack could have fulltext or
1795
# delta, and be at a different position.
1796
index.add_records([((b'tip',), b'fulltext,no-eol', (None, 50, 100),
1798
index.add_records([((b'tip',), b'fulltext,no-eol', (None, 0, 1000),
1800
# but neither should have added data.
1801
self.assertEqual([[], [], [], []], self.caught_entries)
1803
def test_add_versions_different_dup(self):
1804
index = self.two_graph_index(deltas=True, catch_adds=True)
1806
self.assertRaises(KnitCorrupt, index.add_records,
1807
[((b'tip',), b'line-delta', (None, 0, 100), [(b'parent',)])])
1808
self.assertRaises(KnitCorrupt, index.add_records,
1809
[((b'tip',), b'fulltext', (None, 0, 100), [(b'parent',)])])
1811
self.assertRaises(KnitCorrupt, index.add_records,
1812
[((b'tip',), b'fulltext,no-eol', (None, 0, 100), [])])
1813
# change options in the second record
1814
self.assertRaises(KnitCorrupt, index.add_records,
1815
[((b'tip',), b'fulltext,no-eol', (None, 0, 100), [(b'parent',)]),
1816
((b'tip',), b'line-delta', (None, 0, 100), [(b'parent',)])])
1817
self.assertEqual([], self.caught_entries)
1819
def make_g_index_missing_compression_parent(self):
1820
graph_index = self.make_g_index('missing_comp', 2,
1821
[((b'tip', ), b' 100 78',
1822
([(b'missing-parent', ), (b'ghost', )], [(b'missing-parent', )]))])
1825
def make_g_index_missing_parent(self):
1826
graph_index = self.make_g_index('missing_parent', 2,
1827
[((b'parent', ), b' 100 78', ([], [])),
1828
((b'tip', ), b' 100 78',
1829
([(b'parent', ), (b'missing-parent', )], [(b'parent', )])),
1833
def make_g_index_no_external_refs(self):
1834
graph_index = self.make_g_index('no_external_refs', 2,
1835
[((b'rev', ), b' 100 78',
1836
([(b'parent', ), (b'ghost', )], []))])
1839
def test_add_good_unvalidated_index(self):
1840
unvalidated = self.make_g_index_no_external_refs()
1841
combined = CombinedGraphIndex([unvalidated])
1842
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1843
index.scan_unvalidated_index(unvalidated)
1844
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1846
def test_add_missing_compression_parent_unvalidated_index(self):
1847
unvalidated = self.make_g_index_missing_compression_parent()
1848
combined = CombinedGraphIndex([unvalidated])
1849
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1850
index.scan_unvalidated_index(unvalidated)
1851
# This also checks that its only the compression parent that is
1852
# examined, otherwise 'ghost' would also be reported as a missing
1855
frozenset([(b'missing-parent',)]),
1856
index.get_missing_compression_parents())
1858
def test_add_missing_noncompression_parent_unvalidated_index(self):
1859
unvalidated = self.make_g_index_missing_parent()
1860
combined = CombinedGraphIndex([unvalidated])
1861
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1862
track_external_parent_refs=True)
1863
index.scan_unvalidated_index(unvalidated)
1865
frozenset([(b'missing-parent',)]), index.get_missing_parents())
1867
def test_track_external_parent_refs(self):
1868
g_index = self.make_g_index('empty', 2, [])
1869
combined = CombinedGraphIndex([g_index])
1870
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1871
add_callback=self.catch_add, track_external_parent_refs=True)
1872
self.caught_entries = []
1874
((b'new-key',), b'fulltext,no-eol', (None, 50, 60),
1875
[(b'parent-1',), (b'parent-2',)])])
1877
frozenset([(b'parent-1',), (b'parent-2',)]),
1878
index.get_missing_parents())
1880
def test_add_unvalidated_index_with_present_external_references(self):
1881
index = self.two_graph_index(deltas=True)
1882
# Ugly hack to get at one of the underlying GraphIndex objects that
1883
# two_graph_index built.
1884
unvalidated = index._graph_index._indices[1]
1885
# 'parent' is an external ref of _indices[1] (unvalidated), but is
1886
# present in _indices[0].
1887
index.scan_unvalidated_index(unvalidated)
1888
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1890
def make_new_missing_parent_g_index(self, name):
1891
missing_parent = name.encode('ascii') + b'-missing-parent'
1892
graph_index = self.make_g_index(name, 2,
1893
[((name.encode('ascii') + b'tip', ), b' 100 78',
1894
([(missing_parent, ), (b'ghost', )], [(missing_parent, )]))])
1897
def test_add_mulitiple_unvalidated_indices_with_missing_parents(self):
1898
g_index_1 = self.make_new_missing_parent_g_index('one')
1899
g_index_2 = self.make_new_missing_parent_g_index('two')
1900
combined = CombinedGraphIndex([g_index_1, g_index_2])
1901
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1902
index.scan_unvalidated_index(g_index_1)
1903
index.scan_unvalidated_index(g_index_2)
1905
frozenset([(b'one-missing-parent',), (b'two-missing-parent',)]),
1906
index.get_missing_compression_parents())
1908
def test_add_mulitiple_unvalidated_indices_with_mutual_dependencies(self):
1909
graph_index_a = self.make_g_index('one', 2,
1910
[((b'parent-one', ), b' 100 78', ([(b'non-compression-parent',)], [])),
1911
((b'child-of-two', ), b' 100 78',
1912
([(b'parent-two',)], [(b'parent-two',)]))])
1913
graph_index_b = self.make_g_index('two', 2,
1914
[((b'parent-two', ), b' 100 78', ([(b'non-compression-parent',)], [])),
1915
((b'child-of-one', ), b' 100 78',
1916
([(b'parent-one',)], [(b'parent-one',)]))])
1917
combined = CombinedGraphIndex([graph_index_a, graph_index_b])
1918
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1919
index.scan_unvalidated_index(graph_index_a)
1920
index.scan_unvalidated_index(graph_index_b)
1922
frozenset([]), index.get_missing_compression_parents())
1925
class TestNoParentsGraphIndexKnit(KnitTests):
1926
"""Tests for knits using _KnitGraphIndex with no parents."""
1928
def make_g_index(self, name, ref_lists=0, nodes=[]):
1929
builder = GraphIndexBuilder(ref_lists)
1930
for node, references in nodes:
1931
builder.add_node(node, references)
1932
stream = builder.finish()
1933
trans = self.get_transport()
1934
size = trans.put_file(name, stream)
1935
return GraphIndex(trans, name, size)
1937
def test_add_good_unvalidated_index(self):
1938
unvalidated = self.make_g_index('unvalidated')
1939
combined = CombinedGraphIndex([unvalidated])
1940
index = _KnitGraphIndex(combined, lambda: True, parents=False)
1941
index.scan_unvalidated_index(unvalidated)
1942
self.assertEqual(frozenset(),
1943
index.get_missing_compression_parents())
1945
def test_parents_deltas_incompatible(self):
1946
index = CombinedGraphIndex([])
1947
self.assertRaises(knit.KnitError, _KnitGraphIndex, lambda: True,
1948
index, deltas=True, parents=False)
1950
def two_graph_index(self, catch_adds=False):
1951
"""Build a two-graph index.
1953
:param deltas: If true, use underlying indices with two node-ref
1954
lists and 'parent' set to a delta-compressed against tail.
1956
# put several versions in the index.
1957
index1 = self.make_g_index('1', 0, [
1958
((b'tip', ), b'N0 100'),
1959
((b'tail', ), b'')])
1960
index2 = self.make_g_index('2', 0, [
1961
((b'parent', ), b' 100 78'),
1962
((b'separate', ), b'')])
1963
combined_index = CombinedGraphIndex([index1, index2])
1965
self.combined_index = combined_index
1966
self.caught_entries = []
1967
add_callback = self.catch_add
1970
return _KnitGraphIndex(combined_index, lambda: True, parents=False,
1971
add_callback=add_callback)
1973
def test_keys(self):
1974
index = self.two_graph_index()
1975
self.assertEqual({(b'tail',), (b'tip',), (b'parent',), (b'separate',)},
1978
def test_get_position(self):
1979
index = self.two_graph_index()
1980
self.assertEqual((index._graph_index._indices[0], 0, 100),
1981
index.get_position((b'tip',)))
1982
self.assertEqual((index._graph_index._indices[1], 100, 78),
1983
index.get_position((b'parent',)))
1985
def test_get_method(self):
1986
index = self.two_graph_index()
1987
self.assertEqual('fulltext', index.get_method((b'tip',)))
1988
self.assertEqual([b'fulltext'], index.get_options((b'parent',)))
1990
def test_get_options(self):
1991
index = self.two_graph_index()
1992
self.assertEqual([b'fulltext', b'no-eol'],
1993
index.get_options((b'tip',)))
1994
self.assertEqual([b'fulltext'], index.get_options((b'parent',)))
1996
def test_get_parent_map(self):
1997
index = self.two_graph_index()
1998
self.assertEqual({(b'parent',): None},
1999
index.get_parent_map([(b'parent',), (b'ghost',)]))
2001
def catch_add(self, entries):
2002
self.caught_entries.append(entries)
2004
def test_add_no_callback_errors(self):
2005
index = self.two_graph_index()
2006
self.assertRaises(errors.ReadOnlyError, index.add_records,
2007
[((b'new',), b'fulltext,no-eol', (None, 50, 60), [(b'separate',)])])
2009
def test_add_version_smoke(self):
2010
index = self.two_graph_index(catch_adds=True)
2012
[((b'new',), b'fulltext,no-eol', (None, 50, 60), [])])
2013
self.assertEqual([[((b'new', ), b'N50 60')]],
2014
self.caught_entries)
2016
def test_add_version_delta_not_delta_index(self):
2017
index = self.two_graph_index(catch_adds=True)
2018
self.assertRaises(KnitCorrupt, index.add_records,
2019
[((b'new',), b'no-eol,line-delta', (None, 0, 100), [])])
2020
self.assertEqual([], self.caught_entries)
2022
def test_add_version_same_dup(self):
2023
index = self.two_graph_index(catch_adds=True)
2024
# options can be spelt two different ways
2026
[((b'tip',), b'fulltext,no-eol', (None, 0, 100), [])])
2028
[((b'tip',), b'no-eol,fulltext', (None, 0, 100), [])])
2029
# position/length are ignored (because each pack could have fulltext or
2030
# delta, and be at a different position.
2032
[((b'tip',), b'fulltext,no-eol', (None, 50, 100), [])])
2034
[((b'tip',), b'fulltext,no-eol', (None, 0, 1000), [])])
2035
# but neither should have added data.
2036
self.assertEqual([[], [], [], []], self.caught_entries)
2038
def test_add_version_different_dup(self):
2039
index = self.two_graph_index(catch_adds=True)
2041
self.assertRaises(KnitCorrupt, index.add_records,
2042
[((b'tip',), b'no-eol,line-delta', (None, 0, 100), [])])
2043
self.assertRaises(KnitCorrupt, index.add_records,
2044
[((b'tip',), b'line-delta,no-eol', (None, 0, 100), [])])
2045
self.assertRaises(KnitCorrupt, index.add_records,
2046
[((b'tip',), b'fulltext', (None, 0, 100), [])])
2048
self.assertRaises(KnitCorrupt, index.add_records,
2049
[((b'tip',), b'fulltext,no-eol', (None, 0, 100), [(b'parent',)])])
2050
self.assertEqual([], self.caught_entries)
2052
def test_add_versions(self):
2053
index = self.two_graph_index(catch_adds=True)
2055
((b'new',), b'fulltext,no-eol', (None, 50, 60), []),
2056
((b'new2',), b'fulltext', (None, 0, 6), []),
2058
self.assertEqual([((b'new', ), b'N50 60'), ((b'new2', ), b' 0 6')],
2059
sorted(self.caught_entries[0]))
2060
self.assertEqual(1, len(self.caught_entries))
2062
def test_add_versions_delta_not_delta_index(self):
2063
index = self.two_graph_index(catch_adds=True)
2064
self.assertRaises(KnitCorrupt, index.add_records,
2065
[((b'new',), b'no-eol,line-delta', (None, 0, 100), [(b'parent',)])])
2066
self.assertEqual([], self.caught_entries)
2068
def test_add_versions_parents_not_parents_index(self):
2069
index = self.two_graph_index(catch_adds=True)
2070
self.assertRaises(KnitCorrupt, index.add_records,
2071
[((b'new',), b'no-eol,fulltext', (None, 0, 100), [(b'parent',)])])
2072
self.assertEqual([], self.caught_entries)
2074
def test_add_versions_random_id_accepted(self):
2075
index = self.two_graph_index(catch_adds=True)
2076
index.add_records([], random_id=True)
2078
def test_add_versions_same_dup(self):
2079
index = self.two_graph_index(catch_adds=True)
2080
# options can be spelt two different ways
2082
[((b'tip',), b'fulltext,no-eol', (None, 0, 100), [])])
2084
[((b'tip',), b'no-eol,fulltext', (None, 0, 100), [])])
2085
# position/length are ignored (because each pack could have fulltext or
2086
# delta, and be at a different position.
2088
[((b'tip',), b'fulltext,no-eol', (None, 50, 100), [])])
2090
[((b'tip',), b'fulltext,no-eol', (None, 0, 1000), [])])
2091
# but neither should have added data.
2092
self.assertEqual([[], [], [], []], self.caught_entries)
2094
def test_add_versions_different_dup(self):
2095
index = self.two_graph_index(catch_adds=True)
2097
self.assertRaises(KnitCorrupt, index.add_records,
2098
[((b'tip',), b'no-eol,line-delta', (None, 0, 100), [])])
2099
self.assertRaises(KnitCorrupt, index.add_records,
2100
[((b'tip',), b'line-delta,no-eol', (None, 0, 100), [])])
2101
self.assertRaises(KnitCorrupt, index.add_records,
2102
[((b'tip',), b'fulltext', (None, 0, 100), [])])
2104
self.assertRaises(KnitCorrupt, index.add_records,
2105
[((b'tip',), b'fulltext,no-eol', (None, 0, 100), [(b'parent',)])])
2106
# change options in the second record
2107
self.assertRaises(KnitCorrupt, index.add_records,
2108
[((b'tip',), b'fulltext,no-eol', (None, 0, 100), []),
2109
((b'tip',), b'no-eol,line-delta', (None, 0, 100), [])])
2110
self.assertEqual([], self.caught_entries)
2113
class TestKnitVersionedFiles(KnitTests):
2115
def assertGroupKeysForIo(self, exp_groups, keys, non_local_keys,
2116
positions, _min_buffer_size=None):
2117
kvf = self.make_test_knit()
2118
if _min_buffer_size is None:
2119
_min_buffer_size = knit._STREAM_MIN_BUFFER_SIZE
2120
self.assertEqual(exp_groups, kvf._group_keys_for_io(keys,
2121
non_local_keys, positions,
2122
_min_buffer_size=_min_buffer_size))
2124
def assertSplitByPrefix(self, expected_map, expected_prefix_order,
2126
split, prefix_order = KnitVersionedFiles._split_by_prefix(keys)
2127
self.assertEqual(expected_map, split)
2128
self.assertEqual(expected_prefix_order, prefix_order)
2130
def test__group_keys_for_io(self):
2131
ft_detail = ('fulltext', False)
2132
ld_detail = ('line-delta', False)
2140
f_a: (ft_detail, (f_a, 0, 100), None),
2141
f_b: (ld_detail, (f_b, 100, 21), f_a),
2142
f_c: (ld_detail, (f_c, 180, 15), f_b),
2143
g_a: (ft_detail, (g_a, 121, 35), None),
2144
g_b: (ld_detail, (g_b, 156, 12), g_a),
2145
g_c: (ld_detail, (g_c, 195, 13), g_a),
2147
self.assertGroupKeysForIo([([f_a], set())],
2148
[f_a], [], positions)
2149
self.assertGroupKeysForIo([([f_a], {f_a})],
2150
[f_a], [f_a], positions)
2151
self.assertGroupKeysForIo([([f_a, f_b], set([]))],
2152
[f_a, f_b], [], positions)
2153
self.assertGroupKeysForIo([([f_a, f_b], {f_b})],
2154
[f_a, f_b], [f_b], positions)
2155
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2156
[f_a, g_a, f_b, g_b], [], positions)
2157
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2158
[f_a, g_a, f_b, g_b], [], positions,
2159
_min_buffer_size=150)
2160
self.assertGroupKeysForIo([([f_a, f_b], set()), ([g_a, g_b], set())],
2161
[f_a, g_a, f_b, g_b], [], positions,
2162
_min_buffer_size=100)
2163
self.assertGroupKeysForIo([([f_c], set()), ([g_b], set())],
2164
[f_c, g_b], [], positions,
2165
_min_buffer_size=125)
2166
self.assertGroupKeysForIo([([g_b, f_c], set())],
2167
[g_b, f_c], [], positions,
2168
_min_buffer_size=125)
2170
def test__split_by_prefix(self):
2171
self.assertSplitByPrefix({b'f': [(b'f', b'a'), (b'f', b'b')],
2172
b'g': [(b'g', b'b'), (b'g', b'a')],
2174
[(b'f', b'a'), (b'g', b'b'),
2175
(b'g', b'a'), (b'f', b'b')])
2177
self.assertSplitByPrefix({b'f': [(b'f', b'a'), (b'f', b'b')],
2178
b'g': [(b'g', b'b'), (b'g', b'a')],
2180
[(b'f', b'a'), (b'f', b'b'),
2181
(b'g', b'b'), (b'g', b'a')])
2183
self.assertSplitByPrefix({b'f': [(b'f', b'a'), (b'f', b'b')],
2184
b'g': [(b'g', b'b'), (b'g', b'a')],
2186
[(b'f', b'a'), (b'f', b'b'),
2187
(b'g', b'b'), (b'g', b'a')])
2189
self.assertSplitByPrefix({b'f': [(b'f', b'a'), (b'f', b'b')],
2190
b'g': [(b'g', b'b'), (b'g', b'a')],
2191
b'': [(b'a',), (b'b',)]
2192
}, [b'f', b'g', b''],
2193
[(b'f', b'a'), (b'g', b'b'),
2195
(b'g', b'a'), (b'f', b'b')])
2198
class TestStacking(KnitTests):
2200
def get_basis_and_test_knit(self):
2201
basis = self.make_test_knit(name='basis')
2202
basis = RecordingVersionedFilesDecorator(basis)
2203
test = self.make_test_knit(name='test')
2204
test.add_fallback_versioned_files(basis)
2207
def test_add_fallback_versioned_files(self):
2208
basis = self.make_test_knit(name='basis')
2209
test = self.make_test_knit(name='test')
2210
# It must not error; other tests test that the fallback is referred to
2211
# when accessing data.
2212
test.add_fallback_versioned_files(basis)
2214
def test_add_lines(self):
2215
# lines added to the test are not added to the basis
2216
basis, test = self.get_basis_and_test_knit()
2218
key_basis = (b'bar',)
2219
key_cross_border = (b'quux',)
2220
key_delta = (b'zaphod',)
2221
test.add_lines(key, (), [b'foo\n'])
2222
self.assertEqual({}, basis.get_parent_map([key]))
2223
# lines added to the test that reference across the stack do a
2225
basis.add_lines(key_basis, (), [b'foo\n'])
2227
test.add_lines(key_cross_border, (key_basis,), [b'foo\n'])
2228
self.assertEqual('fulltext', test._index.get_method(key_cross_border))
2229
# we don't even need to look at the basis to see that this should be
2230
# stored as a fulltext
2231
self.assertEqual([], basis.calls)
2232
# Subsequent adds do delta.
2234
test.add_lines(key_delta, (key_cross_border,), [b'foo\n'])
2235
self.assertEqual('line-delta', test._index.get_method(key_delta))
2236
self.assertEqual([], basis.calls)
2238
def test_annotate(self):
2239
# annotations from the test knit are answered without asking the basis
2240
basis, test = self.get_basis_and_test_knit()
2242
key_basis = (b'bar',)
2243
test.add_lines(key, (), [b'foo\n'])
2244
details = test.annotate(key)
2245
self.assertEqual([(key, b'foo\n')], details)
2246
self.assertEqual([], basis.calls)
2247
# But texts that are not in the test knit are looked for in the basis
2249
basis.add_lines(key_basis, (), [b'foo\n', b'bar\n'])
2251
details = test.annotate(key_basis)
2253
[(key_basis, b'foo\n'), (key_basis, b'bar\n')], details)
2254
# Not optimised to date:
2255
# self.assertEqual([("annotate", key_basis)], basis.calls)
2256
self.assertEqual([('get_parent_map', {key_basis}),
2257
('get_parent_map', {key_basis}),
2258
('get_record_stream', [key_basis], 'topological', True)],
2261
def test_check(self):
2262
# At the moment checking a stacked knit does implicitly check the
2264
basis, test = self.get_basis_and_test_knit()
2267
def test_get_parent_map(self):
2268
# parents in the test knit are answered without asking the basis
2269
basis, test = self.get_basis_and_test_knit()
2271
key_basis = (b'bar',)
2272
key_missing = (b'missing',)
2273
test.add_lines(key, (), [])
2274
parent_map = test.get_parent_map([key])
2275
self.assertEqual({key: ()}, parent_map)
2276
self.assertEqual([], basis.calls)
2277
# But parents that are not in the test knit are looked for in the basis
2278
basis.add_lines(key_basis, (), [])
2280
parent_map = test.get_parent_map([key, key_basis, key_missing])
2281
self.assertEqual({key: (),
2282
key_basis: ()}, parent_map)
2283
self.assertEqual([("get_parent_map", {key_basis, key_missing})],
2286
def test_get_record_stream_unordered_fulltexts(self):
2287
# records from the test knit are answered without asking the basis:
2288
basis, test = self.get_basis_and_test_knit()
2290
key_basis = (b'bar',)
2291
key_missing = (b'missing',)
2292
test.add_lines(key, (), [b'foo\n'])
2293
records = list(test.get_record_stream([key], 'unordered', True))
2294
self.assertEqual(1, len(records))
2295
self.assertEqual([], basis.calls)
2296
# Missing (from test knit) objects are retrieved from the basis:
2297
basis.add_lines(key_basis, (), [b'foo\n', b'bar\n'])
2299
records = list(test.get_record_stream([key_basis, key_missing],
2301
self.assertEqual(2, len(records))
2302
calls = list(basis.calls)
2303
for record in records:
2304
self.assertSubset([record.key], (key_basis, key_missing))
2305
if record.key == key_missing:
2306
self.assertIsInstance(record, AbsentContentFactory)
2308
reference = list(basis.get_record_stream([key_basis],
2309
'unordered', True))[0]
2310
self.assertEqual(reference.key, record.key)
2311
self.assertEqual(reference.sha1, record.sha1)
2312
self.assertEqual(reference.storage_kind, record.storage_kind)
2313
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2314
record.get_bytes_as(record.storage_kind))
2315
self.assertEqual(reference.get_bytes_as('fulltext'),
2316
record.get_bytes_as('fulltext'))
2317
# It's not strictly minimal, but it seems reasonable for now for it to
2318
# ask which fallbacks have which parents.
2320
("get_parent_map", {key_basis, key_missing}),
2321
("get_record_stream", [key_basis], 'unordered', True)],
2324
def test_get_record_stream_ordered_fulltexts(self):
2325
# ordering is preserved down into the fallback store.
2326
basis, test = self.get_basis_and_test_knit()
2328
key_basis = (b'bar',)
2329
key_basis_2 = (b'quux',)
2330
key_missing = (b'missing',)
2331
test.add_lines(key, (key_basis,), [b'foo\n'])
2332
# Missing (from test knit) objects are retrieved from the basis:
2333
basis.add_lines(key_basis, (key_basis_2,), [b'foo\n', b'bar\n'])
2334
basis.add_lines(key_basis_2, (), [b'quux\n'])
2336
# ask for in non-topological order
2337
records = list(test.get_record_stream(
2338
[key, key_basis, key_missing, key_basis_2], 'topological', True))
2339
self.assertEqual(4, len(records))
2341
for record in records:
2342
self.assertSubset([record.key],
2343
(key_basis, key_missing, key_basis_2, key))
2344
if record.key == key_missing:
2345
self.assertIsInstance(record, AbsentContentFactory)
2347
results.append((record.key, record.sha1, record.storage_kind,
2348
record.get_bytes_as('fulltext')))
2349
calls = list(basis.calls)
2350
order = [record[0] for record in results]
2351
self.assertEqual([key_basis_2, key_basis, key], order)
2352
for result in results:
2353
if result[0] == key:
2357
record = next(source.get_record_stream([result[0]], 'unordered',
2359
self.assertEqual(record.key, result[0])
2360
self.assertEqual(record.sha1, result[1])
2361
# We used to check that the storage kind matched, but actually it
2362
# depends on whether it was sourced from the basis, or in a single
2363
# group, because asking for full texts returns proxy objects to a
2364
# _ContentMapGenerator object; so checking the kind is unneeded.
2365
self.assertEqual(record.get_bytes_as('fulltext'), result[3])
2366
# It's not strictly minimal, but it seems reasonable for now for it to
2367
# ask which fallbacks have which parents.
2368
self.assertEqual(2, len(calls))
2370
("get_parent_map", {key_basis, key_basis_2, key_missing}),
2372
# topological is requested from the fallback, because that is what
2373
# was requested at the top level.
2376
("get_record_stream", [key_basis_2,
2377
key_basis], 'topological', True),
2378
("get_record_stream", [key_basis, key_basis_2], 'topological', True)])
2380
def test_get_record_stream_unordered_deltas(self):
2381
# records from the test knit are answered without asking the basis:
2382
basis, test = self.get_basis_and_test_knit()
2384
key_basis = (b'bar',)
2385
key_missing = (b'missing',)
2386
test.add_lines(key, (), [b'foo\n'])
2387
records = list(test.get_record_stream([key], 'unordered', False))
2388
self.assertEqual(1, len(records))
2389
self.assertEqual([], basis.calls)
2390
# Missing (from test knit) objects are retrieved from the basis:
2391
basis.add_lines(key_basis, (), [b'foo\n', b'bar\n'])
2393
records = list(test.get_record_stream([key_basis, key_missing],
2394
'unordered', False))
2395
self.assertEqual(2, len(records))
2396
calls = list(basis.calls)
2397
for record in records:
2398
self.assertSubset([record.key], (key_basis, key_missing))
2399
if record.key == key_missing:
2400
self.assertIsInstance(record, AbsentContentFactory)
2402
reference = list(basis.get_record_stream([key_basis],
2403
'unordered', False))[0]
2404
self.assertEqual(reference.key, record.key)
2405
self.assertEqual(reference.sha1, record.sha1)
2406
self.assertEqual(reference.storage_kind, record.storage_kind)
2407
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2408
record.get_bytes_as(record.storage_kind))
2409
# It's not strictly minimal, but it seems reasonable for now for it to
2410
# ask which fallbacks have which parents.
2412
("get_parent_map", {key_basis, key_missing}),
2413
("get_record_stream", [key_basis], 'unordered', False)],
2416
def test_get_record_stream_ordered_deltas(self):
2417
# ordering is preserved down into the fallback store.
2418
basis, test = self.get_basis_and_test_knit()
2420
key_basis = (b'bar',)
2421
key_basis_2 = (b'quux',)
2422
key_missing = (b'missing',)
2423
test.add_lines(key, (key_basis,), [b'foo\n'])
2424
# Missing (from test knit) objects are retrieved from the basis:
2425
basis.add_lines(key_basis, (key_basis_2,), [b'foo\n', b'bar\n'])
2426
basis.add_lines(key_basis_2, (), [b'quux\n'])
2428
# ask for in non-topological order
2429
records = list(test.get_record_stream(
2430
[key, key_basis, key_missing, key_basis_2], 'topological', False))
2431
self.assertEqual(4, len(records))
2433
for record in records:
2434
self.assertSubset([record.key],
2435
(key_basis, key_missing, key_basis_2, key))
2436
if record.key == key_missing:
2437
self.assertIsInstance(record, AbsentContentFactory)
2439
results.append((record.key, record.sha1, record.storage_kind,
2440
record.get_bytes_as(record.storage_kind)))
2441
calls = list(basis.calls)
2442
order = [record[0] for record in results]
2443
self.assertEqual([key_basis_2, key_basis, key], order)
2444
for result in results:
2445
if result[0] == key:
2449
record = next(source.get_record_stream([result[0]], 'unordered',
2451
self.assertEqual(record.key, result[0])
2452
self.assertEqual(record.sha1, result[1])
2453
self.assertEqual(record.storage_kind, result[2])
2454
self.assertEqual(record.get_bytes_as(
2455
record.storage_kind), result[3])
2456
# It's not strictly minimal, but it seems reasonable for now for it to
2457
# ask which fallbacks have which parents.
2459
("get_parent_map", {key_basis, key_basis_2, key_missing}),
2460
("get_record_stream", [key_basis_2, key_basis], 'topological', False)],
2463
def test_get_sha1s(self):
2464
# sha1's in the test knit are answered without asking the basis
2465
basis, test = self.get_basis_and_test_knit()
2467
key_basis = (b'bar',)
2468
key_missing = (b'missing',)
2469
test.add_lines(key, (), [b'foo\n'])
2470
key_sha1sum = osutils.sha_string(b'foo\n')
2471
sha1s = test.get_sha1s([key])
2472
self.assertEqual({key: key_sha1sum}, sha1s)
2473
self.assertEqual([], basis.calls)
2474
# But texts that are not in the test knit are looked for in the basis
2475
# directly (rather than via text reconstruction) so that remote servers
2476
# etc don't have to answer with full content.
2477
basis.add_lines(key_basis, (), [b'foo\n', b'bar\n'])
2478
basis_sha1sum = osutils.sha_string(b'foo\nbar\n')
2480
sha1s = test.get_sha1s([key, key_missing, key_basis])
2481
self.assertEqual({key: key_sha1sum,
2482
key_basis: basis_sha1sum}, sha1s)
2483
self.assertEqual([("get_sha1s", {key_basis, key_missing})],
2486
def test_insert_record_stream(self):
2487
# records are inserted as normal; insert_record_stream builds on
2488
# add_lines, so a smoke test should be all that's needed:
2489
key_basis = (b'bar',)
2490
key_delta = (b'zaphod',)
2491
basis, test = self.get_basis_and_test_knit()
2492
source = self.make_test_knit(name='source')
2493
basis.add_lines(key_basis, (), [b'foo\n'])
2495
source.add_lines(key_basis, (), [b'foo\n'])
2496
source.add_lines(key_delta, (key_basis,), [b'bar\n'])
2497
stream = source.get_record_stream([key_delta], 'unordered', False)
2498
test.insert_record_stream(stream)
2499
# XXX: this does somewhat too many calls in making sure of whether it
2500
# has to recreate the full text.
2501
self.assertEqual([("get_parent_map", {key_basis}),
2502
('get_parent_map', {key_basis}),
2503
('get_record_stream', [key_basis], 'unordered', True)],
2505
self.assertEqual({key_delta: (key_basis,)},
2506
test.get_parent_map([key_delta]))
2507
self.assertEqual(b'bar\n', next(test.get_record_stream([key_delta],
2508
'unordered', True)).get_bytes_as('fulltext'))
2510
def test_iter_lines_added_or_present_in_keys(self):
2511
# Lines from the basis are returned, and lines for a given key are only
2515
# all sources are asked for keys:
2516
basis, test = self.get_basis_and_test_knit()
2517
basis.add_lines(key1, (), [b"foo"])
2519
lines = list(test.iter_lines_added_or_present_in_keys([key1]))
2520
self.assertEqual([(b"foo\n", key1)], lines)
2521
self.assertEqual([("iter_lines_added_or_present_in_keys", {key1})],
2523
# keys in both are not duplicated:
2524
test.add_lines(key2, (), [b"bar\n"])
2525
basis.add_lines(key2, (), [b"bar\n"])
2527
lines = list(test.iter_lines_added_or_present_in_keys([key2]))
2528
self.assertEqual([(b"bar\n", key2)], lines)
2529
self.assertEqual([], basis.calls)
2531
def test_keys(self):
2534
# all sources are asked for keys:
2535
basis, test = self.get_basis_and_test_knit()
2537
self.assertEqual(set(), set(keys))
2538
self.assertEqual([("keys",)], basis.calls)
2539
# keys from a basis are returned:
2540
basis.add_lines(key1, (), [])
2543
self.assertEqual({key1}, set(keys))
2544
self.assertEqual([("keys",)], basis.calls)
2545
# keys in both are not duplicated:
2546
test.add_lines(key2, (), [])
2547
basis.add_lines(key2, (), [])
2550
self.assertEqual(2, len(keys))
2551
self.assertEqual({key1, key2}, set(keys))
2552
self.assertEqual([("keys",)], basis.calls)
2554
def test_add_mpdiffs(self):
2555
# records are inserted as normal; add_mpdiff builds on
2556
# add_lines, so a smoke test should be all that's needed:
2557
key_basis = (b'bar',)
2558
key_delta = (b'zaphod',)
2559
basis, test = self.get_basis_and_test_knit()
2560
source = self.make_test_knit(name='source')
2561
basis.add_lines(key_basis, (), [b'foo\n'])
2563
source.add_lines(key_basis, (), [b'foo\n'])
2564
source.add_lines(key_delta, (key_basis,), [b'bar\n'])
2565
diffs = source.make_mpdiffs([key_delta])
2566
test.add_mpdiffs([(key_delta, (key_basis,),
2567
source.get_sha1s([key_delta])[key_delta], diffs[0])])
2568
self.assertEqual([("get_parent_map", {key_basis}),
2569
('get_record_stream', [key_basis], 'unordered', True), ],
2571
self.assertEqual({key_delta: (key_basis,)},
2572
test.get_parent_map([key_delta]))
2573
self.assertEqual(b'bar\n', next(test.get_record_stream([key_delta],
2574
'unordered', True)).get_bytes_as('fulltext'))
2576
def test_make_mpdiffs(self):
2577
# Generating an mpdiff across a stacking boundary should detect parent
2580
key_left = (b'bar',)
2581
key_right = (b'zaphod',)
2582
basis, test = self.get_basis_and_test_knit()
2583
basis.add_lines(key_left, (), [b'bar\n'])
2584
basis.add_lines(key_right, (), [b'zaphod\n'])
2586
test.add_lines(key, (key_left, key_right),
2587
[b'bar\n', b'foo\n', b'zaphod\n'])
2588
diffs = test.make_mpdiffs([key])
2590
multiparent.MultiParent([multiparent.ParentText(0, 0, 0, 1),
2591
multiparent.NewText([b'foo\n']),
2592
multiparent.ParentText(1, 0, 2, 1)])],
2594
self.assertEqual(3, len(basis.calls))
2596
("get_parent_map", {key_left, key_right}),
2597
("get_parent_map", {key_left, key_right}),
2600
last_call = basis.calls[-1]
2601
self.assertEqual('get_record_stream', last_call[0])
2602
self.assertEqual({key_left, key_right}, set(last_call[1]))
2603
self.assertEqual('topological', last_call[2])
2604
self.assertEqual(True, last_call[3])
2607
class TestNetworkBehaviour(KnitTests):
2608
"""Tests for getting data out of/into knits over the network."""
2610
def test_include_delta_closure_generates_a_knit_delta_closure(self):
2611
vf = self.make_test_knit(name='test')
2612
# put in three texts, giving ft, delta, delta
2613
vf.add_lines((b'base',), (), [b'base\n', b'content\n'])
2614
vf.add_lines((b'd1',), ((b'base',),), [b'd1\n'])
2615
vf.add_lines((b'd2',), ((b'd1',),), [b'd2\n'])
2616
# But heuristics could interfere, so check what happened:
2617
self.assertEqual(['knit-ft-gz', 'knit-delta-gz', 'knit-delta-gz'],
2618
[record.storage_kind for record in
2619
vf.get_record_stream([(b'base',), (b'd1',), (b'd2',)],
2620
'topological', False)])
2621
# generate a stream of just the deltas include_delta_closure=True,
2622
# serialise to the network, and check that we get a delta closure on the wire.
2623
stream = vf.get_record_stream(
2624
[(b'd1',), (b'd2',)], 'topological', True)
2625
netb = [record.get_bytes_as(record.storage_kind) for record in stream]
2626
# The first bytes should be a memo from _ContentMapGenerator, and the
2627
# second bytes should be empty (because its a API proxy not something
2628
# for wire serialisation.
2629
self.assertEqual(b'', netb[1])
2631
kind, line_end = network_bytes_to_kind_and_offset(bytes)
2632
self.assertEqual('knit-delta-closure', kind)
2635
class TestContentMapGenerator(KnitTests):
2636
"""Tests for ContentMapGenerator"""
2638
def test_get_record_stream_gives_records(self):
2639
vf = self.make_test_knit(name='test')
2640
# put in three texts, giving ft, delta, delta
2641
vf.add_lines((b'base',), (), [b'base\n', b'content\n'])
2642
vf.add_lines((b'd1',), ((b'base',),), [b'd1\n'])
2643
vf.add_lines((b'd2',), ((b'd1',),), [b'd2\n'])
2644
keys = [(b'd1',), (b'd2',)]
2645
generator = _VFContentMapGenerator(vf, keys,
2646
global_map=vf.get_parent_map(keys))
2647
for record in generator.get_record_stream():
2648
if record.key == (b'd1',):
2649
self.assertEqual(b'd1\n', record.get_bytes_as('fulltext'))
2651
self.assertEqual(b'd2\n', record.get_bytes_as('fulltext'))
2653
def test_get_record_stream_kinds_are_raw(self):
2654
vf = self.make_test_knit(name='test')
2655
# put in three texts, giving ft, delta, delta
2656
vf.add_lines((b'base',), (), [b'base\n', b'content\n'])
2657
vf.add_lines((b'd1',), ((b'base',),), [b'd1\n'])
2658
vf.add_lines((b'd2',), ((b'd1',),), [b'd2\n'])
2659
keys = [(b'base',), (b'd1',), (b'd2',)]
2660
generator = _VFContentMapGenerator(vf, keys,
2661
global_map=vf.get_parent_map(keys))
2662
kinds = {(b'base',): 'knit-delta-closure',
2663
(b'd1',): 'knit-delta-closure-ref',
2664
(b'd2',): 'knit-delta-closure-ref',
2666
for record in generator.get_record_stream():
2667
self.assertEqual(kinds[record.key], record.storage_kind)
added
removed
breezy/tests/test_knit.py
