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([(b"origin1", b"text1"), (b"origin2", b"text2")])
115
self.assertEqual(content.text(), [b"text1", b"text2"])
118
content = self._make_content([(b"origin1", b"text1"), (b"origin2", b"text2")])
119
copy = content.copy()
120
self.assertIsInstance(copy, content.__class__)
121
self.assertEqual(copy.annotate(), content.annotate())
123
def assertDerivedBlocksEqual(self, source, target, noeol=False):
124
"""Assert that the derived matching blocks match real output"""
125
source_lines = source.splitlines(True)
126
target_lines = target.splitlines(True)
128
if noeol and not line.endswith('\n'):
132
source_content = self._make_content([(None, nl(l)) for l in source_lines])
133
target_content = self._make_content([(None, nl(l)) for l in target_lines])
134
line_delta = source_content.line_delta(target_content)
135
delta_blocks = list(KnitContent.get_line_delta_blocks(line_delta,
136
source_lines, target_lines))
137
matcher = PatienceSequenceMatcher(None, source_lines, target_lines)
138
matcher_blocks = list(matcher.get_matching_blocks())
139
self.assertEqual(matcher_blocks, delta_blocks)
141
def test_get_line_delta_blocks(self):
142
self.assertDerivedBlocksEqual('a\nb\nc\n', 'q\nc\n')
143
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1)
144
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1A)
145
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1B)
146
self.assertDerivedBlocksEqual(TEXT_1B, TEXT_1A)
147
self.assertDerivedBlocksEqual(TEXT_1A, TEXT_1B)
148
self.assertDerivedBlocksEqual(TEXT_1A, '')
149
self.assertDerivedBlocksEqual('', TEXT_1A)
150
self.assertDerivedBlocksEqual('', '')
151
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd')
153
def test_get_line_delta_blocks_noeol(self):
154
"""Handle historical knit deltas safely
156
Some existing knit deltas don't consider the last line to differ
157
when the only difference whether it has a final newline.
159
New knit deltas appear to always consider the last line to differ
162
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd\n', noeol=True)
163
self.assertDerivedBlocksEqual('a\nb\nc\nd\n', 'a\nb\nc', noeol=True)
164
self.assertDerivedBlocksEqual('a\nb\nc\n', 'a\nb\nc', noeol=True)
165
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\n', noeol=True)
177
Banana cup cake recipe
187
Banana cup cake recipe
189
- bananas (do not use plantains!!!)
196
Banana cup cake recipe
213
class TestPlainKnitContent(TestCase, KnitContentTestsMixin):
215
def _make_content(self, lines):
216
annotated_content = AnnotatedKnitContent(lines)
217
return PlainKnitContent(annotated_content.text(), 'bogus')
219
def test_annotate(self):
220
content = self._make_content([])
221
self.assertEqual(content.annotate(), [])
223
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
224
self.assertEqual(content.annotate(),
225
[("bogus", "text1"), ("bogus", "text2")])
227
def test_line_delta(self):
228
content1 = self._make_content([("", "a"), ("", "b")])
229
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
230
self.assertEqual(content1.line_delta(content2),
231
[(1, 2, 2, ["a", "c"])])
233
def test_line_delta_iter(self):
234
content1 = self._make_content([("", "a"), ("", "b")])
235
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
236
it = content1.line_delta_iter(content2)
237
self.assertEqual(next(it), (1, 2, 2, ["a", "c"]))
238
self.assertRaises(StopIteration, next, it)
241
class TestAnnotatedKnitContent(TestCase, KnitContentTestsMixin):
243
def _make_content(self, lines):
244
return AnnotatedKnitContent(lines)
246
def test_annotate(self):
247
content = self._make_content([])
248
self.assertEqual(content.annotate(), [])
250
content = self._make_content([(b"origin1", b"text1"), (b"origin2", b"text2")])
251
self.assertEqual(content.annotate(),
252
[(b"origin1", b"text1"), (b"origin2", b"text2")])
254
def test_line_delta(self):
255
content1 = self._make_content([("", "a"), ("", "b")])
256
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
257
self.assertEqual(content1.line_delta(content2),
258
[(1, 2, 2, [("", "a"), ("", "c")])])
260
def test_line_delta_iter(self):
261
content1 = self._make_content([("", "a"), ("", "b")])
262
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
263
it = content1.line_delta_iter(content2)
264
self.assertEqual(next(it), (1, 2, 2, [("", "a"), ("", "c")]))
265
self.assertRaises(StopIteration, next, it)
268
class MockTransport(object):
270
def __init__(self, file_lines=None):
271
self.file_lines = file_lines
273
# We have no base directory for the MockTransport
276
def get(self, filename):
277
if self.file_lines is None:
278
raise errors.NoSuchFile(filename)
280
return BytesIO(b"\n".join(self.file_lines))
282
def readv(self, relpath, offsets):
283
fp = self.get(relpath)
284
for offset, size in offsets:
286
yield offset, fp.read(size)
288
def __getattr__(self, name):
289
def queue_call(*args, **kwargs):
290
self.calls.append((name, args, kwargs))
294
class MockReadvFailingTransport(MockTransport):
295
"""Fail in the middle of a readv() result.
297
This Transport will successfully yield the first two requested hunks, but
298
raise NoSuchFile for the rest.
301
def readv(self, relpath, offsets):
303
for result in MockTransport.readv(self, relpath, offsets):
305
# we use 2 because the first offset is the pack header, the second
306
# is the first actual content requset
308
raise errors.NoSuchFile(relpath)
312
class KnitRecordAccessTestsMixin(object):
313
"""Tests for getting and putting knit records."""
315
def test_add_raw_records(self):
316
"""Add_raw_records adds records retrievable later."""
317
access = self.get_access()
318
memos = access.add_raw_records([(b'key', 10)], b'1234567890')
319
self.assertEqual([b'1234567890'], list(access.get_raw_records(memos)))
321
def test_add_several_raw_records(self):
322
"""add_raw_records with many records and read some back."""
323
access = self.get_access()
324
memos = access.add_raw_records([(b'key', 10), (b'key2', 2), (b'key3', 5)],
325
b'12345678901234567')
326
self.assertEqual([b'1234567890', b'12', b'34567'],
327
list(access.get_raw_records(memos)))
328
self.assertEqual([b'1234567890'],
329
list(access.get_raw_records(memos[0:1])))
330
self.assertEqual([b'12'],
331
list(access.get_raw_records(memos[1:2])))
332
self.assertEqual([b'34567'],
333
list(access.get_raw_records(memos[2:3])))
334
self.assertEqual([b'1234567890', b'34567'],
335
list(access.get_raw_records(memos[0:1] + memos[2:3])))
338
class TestKnitKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
339
"""Tests for the .kndx implementation."""
341
def get_access(self):
342
"""Get a .knit style access instance."""
343
mapper = ConstantMapper("foo")
344
access = _KnitKeyAccess(self.get_transport(), mapper)
348
class _TestException(Exception):
349
"""Just an exception for local tests to use."""
352
class TestPackKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
353
"""Tests for the pack based access."""
355
def get_access(self):
356
return self._get_access()[0]
358
def _get_access(self, packname='packfile', index='FOO'):
359
transport = self.get_transport()
360
def write_data(bytes):
361
transport.append_bytes(packname, bytes)
362
writer = pack.ContainerWriter(write_data)
364
access = pack_repo._DirectPackAccess({})
365
access.set_writer(writer, index, (transport, packname))
366
return access, writer
368
def make_pack_file(self):
369
"""Create a pack file with 2 records."""
370
access, writer = self._get_access(packname='packname', index='foo')
372
memos.extend(access.add_raw_records([(b'key1', 10)], b'1234567890'))
373
memos.extend(access.add_raw_records([(b'key2', 5)], b'12345'))
377
def test_pack_collection_pack_retries(self):
378
"""An explicit pack of a pack collection succeeds even when a
379
concurrent pack happens.
381
builder = self.make_branch_builder('.')
382
builder.start_series()
383
builder.build_snapshot(None, [
384
('add', ('', b'root-id', 'directory', None)),
385
('add', ('file', b'file-id', 'file', b'content\nrev 1\n')),
386
], revision_id=b'rev-1')
387
builder.build_snapshot([b'rev-1'], [
388
('modify', ('file', b'content\nrev 2\n')),
389
], revision_id=b'rev-2')
390
builder.build_snapshot([b'rev-2'], [
391
('modify', ('file', b'content\nrev 3\n')),
392
], revision_id=b'rev-3')
393
self.addCleanup(builder.finish_series)
394
b = builder.get_branch()
395
self.addCleanup(b.lock_write().unlock)
397
collection = repo._pack_collection
398
# Concurrently repack the repo.
399
reopened_repo = repo.controldir.open_repository()
404
def make_vf_for_retrying(self):
405
"""Create 3 packs and a reload function.
407
Originally, 2 pack files will have the data, but one will be missing.
408
And then the third will be used in place of the first two if reload()
411
:return: (versioned_file, reload_counter)
412
versioned_file a KnitVersionedFiles using the packs for access
414
builder = self.make_branch_builder('.', format="1.9")
415
builder.start_series()
416
builder.build_snapshot(None, [
417
('add', ('', b'root-id', 'directory', None)),
418
('add', ('file', b'file-id', 'file', b'content\nrev 1\n')),
419
], revision_id=b'rev-1')
420
builder.build_snapshot([b'rev-1'], [
421
('modify', ('file', b'content\nrev 2\n')),
422
], revision_id=b'rev-2')
423
builder.build_snapshot([b'rev-2'], [
424
('modify', ('file', b'content\nrev 3\n')),
425
], revision_id=b'rev-3')
426
builder.finish_series()
427
b = builder.get_branch()
429
self.addCleanup(b.unlock)
430
# Pack these three revisions into another pack file, but don't remove
433
collection = repo._pack_collection
434
collection.ensure_loaded()
435
orig_packs = collection.packs
436
packer = knitpack_repo.KnitPacker(collection, orig_packs, '.testpack')
437
new_pack = packer.pack()
438
# forget about the new pack
442
# Set up a reload() function that switches to using the new pack file
443
new_index = new_pack.revision_index
444
access_tuple = new_pack.access_tuple()
445
reload_counter = [0, 0, 0]
447
reload_counter[0] += 1
448
if reload_counter[1] > 0:
449
# We already reloaded, nothing more to do
450
reload_counter[2] += 1
452
reload_counter[1] += 1
453
vf._index._graph_index._indices[:] = [new_index]
454
vf._access._indices.clear()
455
vf._access._indices[new_index] = access_tuple
457
# Delete one of the pack files so the data will need to be reloaded. We
458
# will delete the file with 'rev-2' in it
459
trans, name = orig_packs[1].access_tuple()
461
# We don't have the index trigger reloading because we want to test
462
# that we reload when the .pack disappears
463
vf._access._reload_func = reload
464
return vf, reload_counter
466
def make_reload_func(self, return_val=True):
469
reload_called[0] += 1
471
return reload_called, reload
473
def make_retry_exception(self):
474
# We raise a real exception so that sys.exc_info() is properly
477
raise _TestException('foobar')
478
except _TestException as e:
479
retry_exc = errors.RetryWithNewPacks(None, reload_occurred=False,
480
exc_info=sys.exc_info())
481
# GZ 2010-08-10: Cycle with exc_info affects 3 tests
484
def test_read_from_several_packs(self):
485
access, writer = self._get_access()
487
memos.extend(access.add_raw_records([(b'key', 10)], b'1234567890'))
489
access, writer = self._get_access('pack2', 'FOOBAR')
490
memos.extend(access.add_raw_records([(b'key', 5)], b'12345'))
492
access, writer = self._get_access('pack3', 'BAZ')
493
memos.extend(access.add_raw_records([(b'key', 5)], b'alpha'))
495
transport = self.get_transport()
496
access = pack_repo._DirectPackAccess({"FOO":(transport, 'packfile'),
497
"FOOBAR":(transport, 'pack2'),
498
"BAZ":(transport, 'pack3')})
499
self.assertEqual([b'1234567890', b'12345', b'alpha'],
500
list(access.get_raw_records(memos)))
501
self.assertEqual([b'1234567890'],
502
list(access.get_raw_records(memos[0:1])))
503
self.assertEqual([b'12345'],
504
list(access.get_raw_records(memos[1:2])))
505
self.assertEqual([b'alpha'],
506
list(access.get_raw_records(memos[2:3])))
507
self.assertEqual([b'1234567890', b'alpha'],
508
list(access.get_raw_records(memos[0:1] + memos[2:3])))
510
def test_set_writer(self):
511
"""The writer should be settable post construction."""
512
access = pack_repo._DirectPackAccess({})
513
transport = self.get_transport()
514
packname = 'packfile'
516
def write_data(bytes):
517
transport.append_bytes(packname, bytes)
518
writer = pack.ContainerWriter(write_data)
520
access.set_writer(writer, index, (transport, packname))
521
memos = access.add_raw_records([(b'key', 10)], b'1234567890')
523
self.assertEqual([b'1234567890'], list(access.get_raw_records(memos)))
525
def test_missing_index_raises_retry(self):
526
memos = self.make_pack_file()
527
transport = self.get_transport()
528
reload_called, reload_func = self.make_reload_func()
529
# Note that the index key has changed from 'foo' to 'bar'
530
access = pack_repo._DirectPackAccess({'bar':(transport, 'packname')},
531
reload_func=reload_func)
532
e = self.assertListRaises(errors.RetryWithNewPacks,
533
access.get_raw_records, memos)
534
# Because a key was passed in which does not match our index list, we
535
# assume that the listing was already reloaded
536
self.assertTrue(e.reload_occurred)
537
self.assertIsInstance(e.exc_info, tuple)
538
self.assertIs(e.exc_info[0], KeyError)
539
self.assertIsInstance(e.exc_info[1], KeyError)
541
def test_missing_index_raises_key_error_with_no_reload(self):
542
memos = self.make_pack_file()
543
transport = self.get_transport()
544
# Note that the index key has changed from 'foo' to 'bar'
545
access = pack_repo._DirectPackAccess({'bar':(transport, 'packname')})
546
e = self.assertListRaises(KeyError, access.get_raw_records, memos)
548
def test_missing_file_raises_retry(self):
549
memos = self.make_pack_file()
550
transport = self.get_transport()
551
reload_called, reload_func = self.make_reload_func()
552
# Note that the 'filename' has been changed to 'different-packname'
553
access = pack_repo._DirectPackAccess(
554
{'foo':(transport, 'different-packname')},
555
reload_func=reload_func)
556
e = self.assertListRaises(errors.RetryWithNewPacks,
557
access.get_raw_records, memos)
558
# The file has gone missing, so we assume we need to reload
559
self.assertFalse(e.reload_occurred)
560
self.assertIsInstance(e.exc_info, tuple)
561
self.assertIs(e.exc_info[0], errors.NoSuchFile)
562
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
563
self.assertEqual('different-packname', e.exc_info[1].path)
565
def test_missing_file_raises_no_such_file_with_no_reload(self):
566
memos = self.make_pack_file()
567
transport = self.get_transport()
568
# Note that the 'filename' has been changed to 'different-packname'
569
access = pack_repo._DirectPackAccess(
570
{'foo': (transport, 'different-packname')})
571
e = self.assertListRaises(errors.NoSuchFile,
572
access.get_raw_records, memos)
574
def test_failing_readv_raises_retry(self):
575
memos = self.make_pack_file()
576
transport = self.get_transport()
577
failing_transport = MockReadvFailingTransport(
578
[transport.get_bytes('packname')])
579
reload_called, reload_func = self.make_reload_func()
580
access = pack_repo._DirectPackAccess(
581
{'foo': (failing_transport, 'packname')},
582
reload_func=reload_func)
583
# Asking for a single record will not trigger the Mock failure
584
self.assertEqual([b'1234567890'],
585
list(access.get_raw_records(memos[:1])))
586
self.assertEqual([b'12345'],
587
list(access.get_raw_records(memos[1:2])))
588
# A multiple offset readv() will fail mid-way through
589
e = self.assertListRaises(errors.RetryWithNewPacks,
590
access.get_raw_records, memos)
591
# The file has gone missing, so we assume we need to reload
592
self.assertFalse(e.reload_occurred)
593
self.assertIsInstance(e.exc_info, tuple)
594
self.assertIs(e.exc_info[0], errors.NoSuchFile)
595
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
596
self.assertEqual('packname', e.exc_info[1].path)
598
def test_failing_readv_raises_no_such_file_with_no_reload(self):
599
memos = self.make_pack_file()
600
transport = self.get_transport()
601
failing_transport = MockReadvFailingTransport(
602
[transport.get_bytes('packname')])
603
reload_called, reload_func = self.make_reload_func()
604
access = pack_repo._DirectPackAccess(
605
{'foo':(failing_transport, 'packname')})
606
# Asking for a single record will not trigger the Mock failure
607
self.assertEqual([b'1234567890'],
608
list(access.get_raw_records(memos[:1])))
609
self.assertEqual([b'12345'],
610
list(access.get_raw_records(memos[1:2])))
611
# A multiple offset readv() will fail mid-way through
612
e = self.assertListRaises(errors.NoSuchFile,
613
access.get_raw_records, memos)
615
def test_reload_or_raise_no_reload(self):
616
access = pack_repo._DirectPackAccess({}, reload_func=None)
617
retry_exc = self.make_retry_exception()
618
# Without a reload_func, we will just re-raise the original exception
619
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
621
def test_reload_or_raise_reload_changed(self):
622
reload_called, reload_func = self.make_reload_func(return_val=True)
623
access = pack_repo._DirectPackAccess({}, reload_func=reload_func)
624
retry_exc = self.make_retry_exception()
625
access.reload_or_raise(retry_exc)
626
self.assertEqual([1], reload_called)
627
retry_exc.reload_occurred=True
628
access.reload_or_raise(retry_exc)
629
self.assertEqual([2], reload_called)
631
def test_reload_or_raise_reload_no_change(self):
632
reload_called, reload_func = self.make_reload_func(return_val=False)
633
access = pack_repo._DirectPackAccess({}, reload_func=reload_func)
634
retry_exc = self.make_retry_exception()
635
# If reload_occurred is False, then we consider it an error to have
636
# reload_func() return False (no changes).
637
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
638
self.assertEqual([1], reload_called)
639
retry_exc.reload_occurred=True
640
# If reload_occurred is True, then we assume nothing changed because
641
# it had changed earlier, but didn't change again
642
access.reload_or_raise(retry_exc)
643
self.assertEqual([2], reload_called)
645
def test_annotate_retries(self):
646
vf, reload_counter = self.make_vf_for_retrying()
647
# It is a little bit bogus to annotate the Revision VF, but it works,
648
# as we have ancestry stored there
650
reload_lines = vf.annotate(key)
651
self.assertEqual([1, 1, 0], reload_counter)
652
plain_lines = vf.annotate(key)
653
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
654
if reload_lines != plain_lines:
655
self.fail('Annotation was not identical with reloading.')
656
# Now delete the packs-in-use, which should trigger another reload, but
657
# this time we just raise an exception because we can't recover
658
for trans, name in vf._access._indices.values():
660
self.assertRaises(errors.NoSuchFile, vf.annotate, key)
661
self.assertEqual([2, 1, 1], reload_counter)
663
def test__get_record_map_retries(self):
664
vf, reload_counter = self.make_vf_for_retrying()
665
keys = [(b'rev-1',), (b'rev-2',), (b'rev-3',)]
666
records = vf._get_record_map(keys)
667
self.assertEqual(keys, sorted(records.keys()))
668
self.assertEqual([1, 1, 0], reload_counter)
669
# Now delete the packs-in-use, which should trigger another reload, but
670
# this time we just raise an exception because we can't recover
671
for trans, name in vf._access._indices.values():
673
self.assertRaises(errors.NoSuchFile, vf._get_record_map, keys)
674
self.assertEqual([2, 1, 1], reload_counter)
676
def test_get_record_stream_retries(self):
677
vf, reload_counter = self.make_vf_for_retrying()
678
keys = [(b'rev-1',), (b'rev-2',), (b'rev-3',)]
679
record_stream = vf.get_record_stream(keys, 'topological', False)
680
record = next(record_stream)
681
self.assertEqual((b'rev-1',), record.key)
682
self.assertEqual([0, 0, 0], reload_counter)
683
record = next(record_stream)
684
self.assertEqual((b'rev-2',), record.key)
685
self.assertEqual([1, 1, 0], reload_counter)
686
record = next(record_stream)
687
self.assertEqual((b'rev-3',), record.key)
688
self.assertEqual([1, 1, 0], reload_counter)
689
# Now delete all pack files, and see that we raise the right error
690
for trans, name in vf._access._indices.values():
692
self.assertListRaises(errors.NoSuchFile,
693
vf.get_record_stream, keys, 'topological', False)
695
def test_iter_lines_added_or_present_in_keys_retries(self):
696
vf, reload_counter = self.make_vf_for_retrying()
697
keys = [(b'rev-1',), (b'rev-2',), (b'rev-3',)]
698
# Unfortunately, iter_lines_added_or_present_in_keys iterates the
699
# result in random order (determined by the iteration order from a
700
# set()), so we don't have any solid way to trigger whether data is
701
# read before or after. However we tried to delete the middle node to
702
# exercise the code well.
703
# What we care about is that all lines are always yielded, but not
706
reload_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
707
self.assertEqual([1, 1, 0], reload_counter)
708
# Now do it again, to make sure the result is equivalent
709
plain_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
710
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
711
self.assertEqual(plain_lines, reload_lines)
712
self.assertEqual(21, len(plain_lines))
713
# Now delete all pack files, and see that we raise the right error
714
for trans, name in vf._access._indices.values():
716
self.assertListRaises(errors.NoSuchFile,
717
vf.iter_lines_added_or_present_in_keys, keys)
718
self.assertEqual([2, 1, 1], reload_counter)
720
def test_get_record_stream_yields_disk_sorted_order(self):
721
# if we get 'unordered' pick a semi-optimal order for reading. The
722
# order should be grouped by pack file, and then by position in file
723
repo = self.make_repository('test', format='pack-0.92')
725
self.addCleanup(repo.unlock)
726
repo.start_write_group()
728
vf.add_lines((b'f-id', b'rev-5'), [(b'f-id', b'rev-4')], [b'lines\n'])
729
vf.add_lines((b'f-id', b'rev-1'), [], [b'lines\n'])
730
vf.add_lines((b'f-id', b'rev-2'), [(b'f-id', b'rev-1')], [b'lines\n'])
731
repo.commit_write_group()
732
# We inserted them as rev-5, rev-1, rev-2, we should get them back in
734
stream = vf.get_record_stream([(b'f-id', b'rev-1'), (b'f-id', b'rev-5'),
735
(b'f-id', b'rev-2')], 'unordered', False)
736
keys = [r.key for r in stream]
737
self.assertEqual([(b'f-id', b'rev-5'), (b'f-id', b'rev-1'),
738
(b'f-id', b'rev-2')], keys)
739
repo.start_write_group()
740
vf.add_lines((b'f-id', b'rev-4'), [(b'f-id', b'rev-3')], [b'lines\n'])
741
vf.add_lines((b'f-id', b'rev-3'), [(b'f-id', b'rev-2')], [b'lines\n'])
742
vf.add_lines((b'f-id', b'rev-6'), [(b'f-id', b'rev-5')], [b'lines\n'])
743
repo.commit_write_group()
744
# Request in random order, to make sure the output order isn't based on
746
request_keys = set((b'f-id', b'rev-%d' % i) for i in range(1, 7))
747
stream = vf.get_record_stream(request_keys, 'unordered', False)
748
keys = [r.key for r in stream]
749
# We want to get the keys back in disk order, but it doesn't matter
750
# which pack we read from first. So this can come back in 2 orders
751
alt1 = [(b'f-id', b'rev-%d' % i) for i in [4, 3, 6, 5, 1, 2]]
752
alt2 = [(b'f-id', b'rev-%d' % i) for i in [5, 1, 2, 4, 3, 6]]
753
if keys != alt1 and keys != alt2:
754
self.fail('Returned key order did not match either expected order.'
755
' expected %s or %s, not %s'
756
% (alt1, alt2, keys))
759
class LowLevelKnitDataTests(TestCase):
761
def create_gz_content(self, text):
763
with gzip.GzipFile(mode='wb', fileobj=sio) as gz_file:
765
return sio.getvalue()
767
def make_multiple_records(self):
768
"""Create the content for multiple records."""
769
sha1sum = osutils.sha_string(b'foo\nbar\n')
771
gz_txt = self.create_gz_content(b'version rev-id-1 2 %s\n'
776
record_1 = (0, len(gz_txt), sha1sum)
777
total_txt.append(gz_txt)
778
sha1sum = osutils.sha_string(b'baz\n')
779
gz_txt = self.create_gz_content(b'version rev-id-2 1 %s\n'
783
record_2 = (record_1[1], len(gz_txt), sha1sum)
784
total_txt.append(gz_txt)
785
return total_txt, record_1, record_2
787
def test_valid_knit_data(self):
788
sha1sum = osutils.sha_string(b'foo\nbar\n')
789
gz_txt = self.create_gz_content(b'version rev-id-1 2 %s\n'
794
transport = MockTransport([gz_txt])
795
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
796
knit = KnitVersionedFiles(None, access)
797
records = [((b'rev-id-1',), ((b'rev-id-1',), 0, len(gz_txt)))]
799
contents = list(knit._read_records_iter(records))
800
self.assertEqual([((b'rev-id-1',), [b'foo\n', b'bar\n'],
801
b'4e48e2c9a3d2ca8a708cb0cc545700544efb5021')], contents)
803
raw_contents = list(knit._read_records_iter_raw(records))
804
self.assertEqual([((b'rev-id-1',), gz_txt, sha1sum)], raw_contents)
806
def test_multiple_records_valid(self):
807
total_txt, record_1, record_2 = self.make_multiple_records()
808
transport = MockTransport([b''.join(total_txt)])
809
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
810
knit = KnitVersionedFiles(None, access)
811
records = [((b'rev-id-1',), ((b'rev-id-1',), record_1[0], record_1[1])),
812
((b'rev-id-2',), ((b'rev-id-2',), record_2[0], record_2[1]))]
814
contents = list(knit._read_records_iter(records))
815
self.assertEqual([((b'rev-id-1',), [b'foo\n', b'bar\n'], record_1[2]),
816
((b'rev-id-2',), [b'baz\n'], record_2[2])],
819
raw_contents = list(knit._read_records_iter_raw(records))
820
self.assertEqual([((b'rev-id-1',), total_txt[0], record_1[2]),
821
((b'rev-id-2',), total_txt[1], record_2[2])],
824
def test_not_enough_lines(self):
825
sha1sum = osutils.sha_string(b'foo\n')
826
# record says 2 lines data says 1
827
gz_txt = self.create_gz_content(b'version rev-id-1 2 %s\n'
831
transport = MockTransport([gz_txt])
832
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
833
knit = KnitVersionedFiles(None, access)
834
records = [((b'rev-id-1',), ((b'rev-id-1',), 0, len(gz_txt)))]
835
self.assertRaises(KnitCorrupt, list,
836
knit._read_records_iter(records))
838
# read_records_iter_raw won't detect that sort of mismatch/corruption
839
raw_contents = list(knit._read_records_iter_raw(records))
840
self.assertEqual([((b'rev-id-1',), gz_txt, sha1sum)], raw_contents)
842
def test_too_many_lines(self):
843
sha1sum = osutils.sha_string(b'foo\nbar\n')
844
# record says 1 lines data says 2
845
gz_txt = self.create_gz_content(b'version rev-id-1 1 %s\n'
850
transport = MockTransport([gz_txt])
851
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
852
knit = KnitVersionedFiles(None, access)
853
records = [((b'rev-id-1',), ((b'rev-id-1',), 0, len(gz_txt)))]
854
self.assertRaises(KnitCorrupt, list,
855
knit._read_records_iter(records))
857
# read_records_iter_raw won't detect that sort of mismatch/corruption
858
raw_contents = list(knit._read_records_iter_raw(records))
859
self.assertEqual([((b'rev-id-1',), gz_txt, sha1sum)], raw_contents)
861
def test_mismatched_version_id(self):
862
sha1sum = osutils.sha_string(b'foo\nbar\n')
863
gz_txt = self.create_gz_content(b'version rev-id-1 2 %s\n'
868
transport = MockTransport([gz_txt])
869
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
870
knit = KnitVersionedFiles(None, access)
871
# We are asking for rev-id-2, but the data is rev-id-1
872
records = [((b'rev-id-2',), ((b'rev-id-2',), 0, len(gz_txt)))]
873
self.assertRaises(KnitCorrupt, list,
874
knit._read_records_iter(records))
876
# read_records_iter_raw detects mismatches in the header
877
self.assertRaises(KnitCorrupt, list,
878
knit._read_records_iter_raw(records))
880
def test_uncompressed_data(self):
881
sha1sum = osutils.sha_string(b'foo\nbar\n')
882
txt = (b'version rev-id-1 2 %s\n'
887
transport = MockTransport([txt])
888
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
889
knit = KnitVersionedFiles(None, access)
890
records = [((b'rev-id-1',), ((b'rev-id-1',), 0, len(txt)))]
892
# We don't have valid gzip data ==> corrupt
893
self.assertRaises(KnitCorrupt, list,
894
knit._read_records_iter(records))
896
# read_records_iter_raw will notice the bad data
897
self.assertRaises(KnitCorrupt, list,
898
knit._read_records_iter_raw(records))
900
def test_corrupted_data(self):
901
sha1sum = osutils.sha_string(b'foo\nbar\n')
902
gz_txt = self.create_gz_content(b'version rev-id-1 2 %s\n'
907
# Change 2 bytes in the middle to \xff
908
gz_txt = gz_txt[:10] + b'\xff\xff' + gz_txt[12:]
909
transport = MockTransport([gz_txt])
910
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
911
knit = KnitVersionedFiles(None, access)
912
records = [((b'rev-id-1',), ((b'rev-id-1',), 0, len(gz_txt)))]
913
self.assertRaises(KnitCorrupt, list,
914
knit._read_records_iter(records))
915
# read_records_iter_raw will barf on bad gz data
916
self.assertRaises(KnitCorrupt, list,
917
knit._read_records_iter_raw(records))
920
class LowLevelKnitIndexTests(TestCase):
923
def _load_data(self):
924
from ..bzr._knit_load_data_py import _load_data_py
927
def get_knit_index(self, transport, name, mode):
928
mapper = ConstantMapper(name)
929
self.overrideAttr(knit, '_load_data', self._load_data)
930
allow_writes = lambda: 'w' in mode
931
return _KndxIndex(transport, mapper, lambda: None, allow_writes, lambda: True)
933
def test_create_file(self):
934
transport = MockTransport()
935
index = self.get_knit_index(transport, "filename", "w")
937
call = transport.calls.pop(0)
938
# call[1][1] is a BytesIO - we can't test it by simple equality.
939
self.assertEqual('put_file_non_atomic', call[0])
940
self.assertEqual('filename.kndx', call[1][0])
941
# With no history, _KndxIndex writes a new index:
942
self.assertEqual(_KndxIndex.HEADER,
943
call[1][1].getvalue())
944
self.assertEqual({'create_parent_dir': True}, call[2])
946
def test_read_utf8_version_id(self):
947
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
948
utf8_revision_id = unicode_revision_id.encode('utf-8')
949
transport = MockTransport([
951
b'%s option 0 1 :' % (utf8_revision_id,)
953
index = self.get_knit_index(transport, "filename", "r")
954
# _KndxIndex is a private class, and deals in utf8 revision_ids, not
955
# Unicode revision_ids.
956
self.assertEqual({(utf8_revision_id,):()},
957
index.get_parent_map(index.keys()))
958
self.assertFalse((unicode_revision_id,) in index.keys())
960
def test_read_utf8_parents(self):
961
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
962
utf8_revision_id = unicode_revision_id.encode('utf-8')
963
transport = MockTransport([
965
b"version option 0 1 .%s :" % (utf8_revision_id,)
967
index = self.get_knit_index(transport, "filename", "r")
968
self.assertEqual({(b"version",): ((utf8_revision_id,),)},
969
index.get_parent_map(index.keys()))
971
def test_read_ignore_corrupted_lines(self):
972
transport = MockTransport([
975
b"corrupted options 0 1 .b .c ",
976
b"version options 0 1 :"
978
index = self.get_knit_index(transport, "filename", "r")
979
self.assertEqual(1, len(index.keys()))
980
self.assertEqual({(b"version",)}, index.keys())
982
def test_read_corrupted_header(self):
983
transport = MockTransport([b'not a bzr knit index header\n'])
984
index = self.get_knit_index(transport, "filename", "r")
985
self.assertRaises(KnitHeaderError, index.keys)
987
def test_read_duplicate_entries(self):
988
transport = MockTransport([
990
b"parent options 0 1 :",
991
b"version options1 0 1 0 :",
992
b"version options2 1 2 .other :",
993
b"version options3 3 4 0 .other :"
995
index = self.get_knit_index(transport, "filename", "r")
996
self.assertEqual(2, len(index.keys()))
997
# check that the index used is the first one written. (Specific
998
# to KnitIndex style indices.
999
self.assertEqual(b"1", index._dictionary_compress([(b"version",)]))
1000
self.assertEqual(((b"version",), 3, 4), index.get_position((b"version",)))
1001
self.assertEqual([b"options3"], index.get_options((b"version",)))
1002
self.assertEqual({(b"version",): ((b"parent",), (b"other",))},
1003
index.get_parent_map([(b"version",)]))
1005
def test_read_compressed_parents(self):
1006
transport = MockTransport([
1009
b"b option 0 1 0 :",
1010
b"c option 0 1 1 0 :",
1012
index = self.get_knit_index(transport, "filename", "r")
1013
self.assertEqual({(b"b",):((b"a",),), (b"c",):((b"b",), (b"a",))},
1014
index.get_parent_map([(b"b",), (b"c",)]))
1016
def test_write_utf8_version_id(self):
1017
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
1018
utf8_revision_id = unicode_revision_id.encode('utf-8')
1019
transport = MockTransport([
1022
index = self.get_knit_index(transport, "filename", "r")
1024
((utf8_revision_id,), [b"option"], ((utf8_revision_id,), 0, 1), [])])
1025
call = transport.calls.pop(0)
1026
# call[1][1] is a BytesIO - we can't test it by simple equality.
1027
self.assertEqual('put_file_non_atomic', call[0])
1028
self.assertEqual('filename.kndx', call[1][0])
1029
# With no history, _KndxIndex writes a new index:
1030
self.assertEqual(_KndxIndex.HEADER +
1031
b"\n%s option 0 1 :" % (utf8_revision_id,),
1032
call[1][1].getvalue())
1033
self.assertEqual({'create_parent_dir': True}, call[2])
1035
def test_write_utf8_parents(self):
1036
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
1037
utf8_revision_id = unicode_revision_id.encode('utf-8')
1038
transport = MockTransport([
1041
index = self.get_knit_index(transport, "filename", "r")
1043
((b"version",), [b"option"], ((b"version",), 0, 1), [(utf8_revision_id,)])])
1044
call = transport.calls.pop(0)
1045
# call[1][1] is a BytesIO - we can't test it by simple equality.
1046
self.assertEqual('put_file_non_atomic', call[0])
1047
self.assertEqual('filename.kndx', call[1][0])
1048
# With no history, _KndxIndex writes a new index:
1049
self.assertEqual(_KndxIndex.HEADER +
1050
b"\nversion option 0 1 .%s :" % (utf8_revision_id,),
1051
call[1][1].getvalue())
1052
self.assertEqual({'create_parent_dir': True}, call[2])
1054
def test_keys(self):
1055
transport = MockTransport([
1058
index = self.get_knit_index(transport, "filename", "r")
1060
self.assertEqual(set(), index.keys())
1062
index.add_records([((b"a",), [b"option"], ((b"a",), 0, 1), [])])
1063
self.assertEqual({(b"a",)}, index.keys())
1065
index.add_records([((b"a",), [b"option"], ((b"a",), 0, 1), [])])
1066
self.assertEqual({(b"a",)}, index.keys())
1068
index.add_records([((b"b",), [b"option"], ((b"b",), 0, 1), [])])
1069
self.assertEqual({(b"a",), (b"b",)}, index.keys())
1071
def add_a_b(self, index, random_id=None):
1073
if random_id is not None:
1074
kwargs["random_id"] = random_id
1076
((b"a",), [b"option"], ((b"a",), 0, 1), [(b"b",)]),
1077
((b"a",), [b"opt"], ((b"a",), 1, 2), [(b"c",)]),
1078
((b"b",), [b"option"], ((b"b",), 2, 3), [(b"a",)])
1081
def assertIndexIsAB(self, index):
1083
(b'a',): ((b'c',),),
1084
(b'b',): ((b'a',),),
1086
index.get_parent_map(index.keys()))
1087
self.assertEqual(((b"a",), 1, 2), index.get_position((b"a",)))
1088
self.assertEqual(((b"b",), 2, 3), index.get_position((b"b",)))
1089
self.assertEqual([b"opt"], index.get_options((b"a",)))
1091
def test_add_versions(self):
1092
transport = MockTransport([
1095
index = self.get_knit_index(transport, "filename", "r")
1098
call = transport.calls.pop(0)
1099
# call[1][1] is a BytesIO - we can't test it by simple equality.
1100
self.assertEqual('put_file_non_atomic', call[0])
1101
self.assertEqual('filename.kndx', call[1][0])
1102
# With no history, _KndxIndex writes a new index:
1105
b"\na option 0 1 .b :"
1107
b"\nb option 2 3 0 :",
1108
call[1][1].getvalue())
1109
self.assertEqual({'create_parent_dir': True}, call[2])
1110
self.assertIndexIsAB(index)
1112
def test_add_versions_random_id_is_accepted(self):
1113
transport = MockTransport([
1116
index = self.get_knit_index(transport, "filename", "r")
1117
self.add_a_b(index, random_id=True)
1119
def test_delay_create_and_add_versions(self):
1120
transport = MockTransport()
1122
index = self.get_knit_index(transport, "filename", "w")
1124
self.assertEqual([], transport.calls)
1127
#[ {"dir_mode": 0777, "create_parent_dir": True, "mode": "wb"},
1129
# Two calls: one during which we load the existing index (and when its
1130
# missing create it), then a second where we write the contents out.
1131
self.assertEqual(2, len(transport.calls))
1132
call = transport.calls.pop(0)
1133
self.assertEqual('put_file_non_atomic', call[0])
1134
self.assertEqual('filename.kndx', call[1][0])
1135
# With no history, _KndxIndex writes a new index:
1136
self.assertEqual(_KndxIndex.HEADER, call[1][1].getvalue())
1137
self.assertEqual({'create_parent_dir': True}, call[2])
1138
call = transport.calls.pop(0)
1139
# call[1][1] is a BytesIO - we can't test it by simple equality.
1140
self.assertEqual('put_file_non_atomic', call[0])
1141
self.assertEqual('filename.kndx', call[1][0])
1142
# With no history, _KndxIndex writes a new index:
1145
b"\na option 0 1 .b :"
1147
b"\nb option 2 3 0 :",
1148
call[1][1].getvalue())
1149
self.assertEqual({'create_parent_dir': True}, call[2])
1151
def assertTotalBuildSize(self, size, keys, positions):
1152
self.assertEqual(size,
1153
knit._get_total_build_size(None, keys, positions))
1155
def test__get_total_build_size(self):
1157
(b'a',): (('fulltext', False), ((b'a',), 0, 100), None),
1158
(b'b',): (('line-delta', False), ((b'b',), 100, 21), (b'a',)),
1159
(b'c',): (('line-delta', False), ((b'c',), 121, 35), (b'b',)),
1160
(b'd',): (('line-delta', False), ((b'd',), 156, 12), (b'b',)),
1162
self.assertTotalBuildSize(100, [(b'a',)], positions)
1163
self.assertTotalBuildSize(121, [(b'b',)], positions)
1164
# c needs both a & b
1165
self.assertTotalBuildSize(156, [(b'c',)], positions)
1166
# we shouldn't count 'b' twice
1167
self.assertTotalBuildSize(156, [(b'b',), (b'c',)], positions)
1168
self.assertTotalBuildSize(133, [(b'd',)], positions)
1169
self.assertTotalBuildSize(168, [(b'c',), (b'd',)], positions)
1171
def test_get_position(self):
1172
transport = MockTransport([
1177
index = self.get_knit_index(transport, "filename", "r")
1179
self.assertEqual(((b"a",), 0, 1), index.get_position((b"a",)))
1180
self.assertEqual(((b"b",), 1, 2), index.get_position((b"b",)))
1182
def test_get_method(self):
1183
transport = MockTransport([
1185
b"a fulltext,unknown 0 1 :",
1186
b"b unknown,line-delta 1 2 :",
1189
index = self.get_knit_index(transport, "filename", "r")
1191
self.assertEqual("fulltext", index.get_method(b"a"))
1192
self.assertEqual("line-delta", index.get_method(b"b"))
1193
self.assertRaises(knit.KnitIndexUnknownMethod, index.get_method, b"c")
1195
def test_get_options(self):
1196
transport = MockTransport([
1199
b"b opt2,opt3 1 2 :"
1201
index = self.get_knit_index(transport, "filename", "r")
1203
self.assertEqual([b"opt1"], index.get_options(b"a"))
1204
self.assertEqual([b"opt2", b"opt3"], index.get_options(b"b"))
1206
def test_get_parent_map(self):
1207
transport = MockTransport([
1210
b"b option 1 2 0 .c :",
1211
b"c option 1 2 1 0 .e :"
1213
index = self.get_knit_index(transport, "filename", "r")
1217
(b"b",): ((b"a",), (b"c",)),
1218
(b"c",): ((b"b",), (b"a",), (b"e",)),
1219
}, index.get_parent_map(index.keys()))
1221
def test_impossible_parent(self):
1222
"""Test we get KnitCorrupt if the parent couldn't possibly exist."""
1223
transport = MockTransport([
1226
b"b option 0 1 4 :" # We don't have a 4th record
1228
index = self.get_knit_index(transport, 'filename', 'r')
1229
self.assertRaises(KnitCorrupt, index.keys)
1231
def test_corrupted_parent(self):
1232
transport = MockTransport([
1236
b"c option 0 1 1v :", # Can't have a parent of '1v'
1238
index = self.get_knit_index(transport, 'filename', 'r')
1239
self.assertRaises(KnitCorrupt, index.keys)
1241
def test_corrupted_parent_in_list(self):
1242
transport = MockTransport([
1246
b"c option 0 1 1 v :", # Can't have a parent of 'v'
1248
index = self.get_knit_index(transport, 'filename', 'r')
1249
self.assertRaises(KnitCorrupt, index.keys)
1251
def test_invalid_position(self):
1252
transport = MockTransport([
1256
index = self.get_knit_index(transport, 'filename', 'r')
1257
self.assertRaises(KnitCorrupt, index.keys)
1259
def test_invalid_size(self):
1260
transport = MockTransport([
1264
index = self.get_knit_index(transport, 'filename', 'r')
1265
self.assertRaises(KnitCorrupt, index.keys)
1267
def test_scan_unvalidated_index_not_implemented(self):
1268
transport = MockTransport()
1269
index = self.get_knit_index(transport, 'filename', 'r')
1271
NotImplementedError, index.scan_unvalidated_index,
1272
'dummy graph_index')
1274
NotImplementedError, index.get_missing_compression_parents)
1276
def test_short_line(self):
1277
transport = MockTransport([
1280
b"b option 10 10 0", # This line isn't terminated, ignored
1282
index = self.get_knit_index(transport, "filename", "r")
1283
self.assertEqual({(b'a',)}, index.keys())
1285
def test_skip_incomplete_record(self):
1286
# A line with bogus data should just be skipped
1287
transport = MockTransport([
1290
b"b option 10 10 0", # This line isn't terminated, ignored
1291
b"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1293
index = self.get_knit_index(transport, "filename", "r")
1294
self.assertEqual({(b'a',), (b'c',)}, index.keys())
1296
def test_trailing_characters(self):
1297
# A line with bogus data should just be skipped
1298
transport = MockTransport([
1301
b"b option 10 10 0 :a", # This line has extra trailing characters
1302
b"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1304
index = self.get_knit_index(transport, "filename", "r")
1305
self.assertEqual({(b'a',), (b'c',)}, index.keys())
1308
class LowLevelKnitIndexTests_c(LowLevelKnitIndexTests):
1310
_test_needs_features = [compiled_knit_feature]
1313
def _load_data(self):
1314
from ..bzr._knit_load_data_pyx import _load_data_c
1318
class Test_KnitAnnotator(TestCaseWithMemoryTransport):
1320
def make_annotator(self):
1321
factory = knit.make_pack_factory(True, True, 1)
1322
vf = factory(self.get_transport())
1323
return knit._KnitAnnotator(vf)
1325
def test__expand_fulltext(self):
1326
ann = self.make_annotator()
1327
rev_key = (b'rev-id',)
1328
ann._num_compression_children[rev_key] = 1
1329
res = ann._expand_record(rev_key, ((b'parent-id',),), None,
1330
[b'line1\n', b'line2\n'], ('fulltext', True))
1331
# The content object and text lines should be cached appropriately
1332
self.assertEqual([b'line1\n', b'line2'], res)
1333
content_obj = ann._content_objects[rev_key]
1334
self.assertEqual([b'line1\n', b'line2\n'], content_obj._lines)
1335
self.assertEqual(res, content_obj.text())
1336
self.assertEqual(res, ann._text_cache[rev_key])
1338
def test__expand_delta_comp_parent_not_available(self):
1339
# Parent isn't available yet, so we return nothing, but queue up this
1340
# node for later processing
1341
ann = self.make_annotator()
1342
rev_key = (b'rev-id',)
1343
parent_key = (b'parent-id',)
1344
record = [b'0,1,1\n', b'new-line\n']
1345
details = ('line-delta', False)
1346
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1348
self.assertEqual(None, res)
1349
self.assertTrue(parent_key in ann._pending_deltas)
1350
pending = ann._pending_deltas[parent_key]
1351
self.assertEqual(1, len(pending))
1352
self.assertEqual((rev_key, (parent_key,), record, details), pending[0])
1354
def test__expand_record_tracks_num_children(self):
1355
ann = self.make_annotator()
1356
rev_key = (b'rev-id',)
1357
rev2_key = (b'rev2-id',)
1358
parent_key = (b'parent-id',)
1359
record = [b'0,1,1\n', b'new-line\n']
1360
details = ('line-delta', False)
1361
ann._num_compression_children[parent_key] = 2
1362
ann._expand_record(parent_key, (), None, [b'line1\n', b'line2\n'],
1363
('fulltext', False))
1364
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1366
self.assertEqual({parent_key: 1}, ann._num_compression_children)
1367
# Expanding the second child should remove the content object, and the
1368
# num_compression_children entry
1369
res = ann._expand_record(rev2_key, (parent_key,), parent_key,
1371
self.assertFalse(parent_key in ann._content_objects)
1372
self.assertEqual({}, ann._num_compression_children)
1373
# We should not cache the content_objects for rev2 and rev, because
1374
# they do not have compression children of their own.
1375
self.assertEqual({}, ann._content_objects)
1377
def test__expand_delta_records_blocks(self):
1378
ann = self.make_annotator()
1379
rev_key = (b'rev-id',)
1380
parent_key = (b'parent-id',)
1381
record = [b'0,1,1\n', b'new-line\n']
1382
details = ('line-delta', True)
1383
ann._num_compression_children[parent_key] = 2
1384
ann._expand_record(parent_key, (), None,
1385
[b'line1\n', b'line2\n', b'line3\n'],
1386
('fulltext', False))
1387
ann._expand_record(rev_key, (parent_key,), parent_key, record, details)
1388
self.assertEqual({(rev_key, parent_key): [(1, 1, 1), (3, 3, 0)]},
1389
ann._matching_blocks)
1390
rev2_key = (b'rev2-id',)
1391
record = [b'0,1,1\n', b'new-line\n']
1392
details = ('line-delta', False)
1393
ann._expand_record(rev2_key, (parent_key,), parent_key, record, details)
1394
self.assertEqual([(1, 1, 2), (3, 3, 0)],
1395
ann._matching_blocks[(rev2_key, parent_key)])
1397
def test__get_parent_ann_uses_matching_blocks(self):
1398
ann = self.make_annotator()
1399
rev_key = (b'rev-id',)
1400
parent_key = (b'parent-id',)
1401
parent_ann = [(parent_key,)]*3
1402
block_key = (rev_key, parent_key)
1403
ann._annotations_cache[parent_key] = parent_ann
1404
ann._matching_blocks[block_key] = [(0, 1, 1), (3, 3, 0)]
1405
# We should not try to access any parent_lines content, because we know
1406
# we already have the matching blocks
1407
par_ann, blocks = ann._get_parent_annotations_and_matches(rev_key,
1408
[b'1\n', b'2\n', b'3\n'], parent_key)
1409
self.assertEqual(parent_ann, par_ann)
1410
self.assertEqual([(0, 1, 1), (3, 3, 0)], blocks)
1411
self.assertEqual({}, ann._matching_blocks)
1413
def test__process_pending(self):
1414
ann = self.make_annotator()
1415
rev_key = (b'rev-id',)
1416
p1_key = (b'p1-id',)
1417
p2_key = (b'p2-id',)
1418
record = [b'0,1,1\n', b'new-line\n']
1419
details = ('line-delta', False)
1420
p1_record = [b'line1\n', b'line2\n']
1421
ann._num_compression_children[p1_key] = 1
1422
res = ann._expand_record(rev_key, (p1_key, p2_key), p1_key,
1424
self.assertEqual(None, res)
1425
# self.assertTrue(p1_key in ann._pending_deltas)
1426
self.assertEqual({}, ann._pending_annotation)
1427
# Now insert p1, and we should be able to expand the delta
1428
res = ann._expand_record(p1_key, (), None, p1_record,
1429
('fulltext', False))
1430
self.assertEqual(p1_record, res)
1431
ann._annotations_cache[p1_key] = [(p1_key,)]*2
1432
res = ann._process_pending(p1_key)
1433
self.assertEqual([], res)
1434
self.assertFalse(p1_key in ann._pending_deltas)
1435
self.assertTrue(p2_key in ann._pending_annotation)
1436
self.assertEqual({p2_key: [(rev_key, (p1_key, p2_key))]},
1437
ann._pending_annotation)
1438
# Now fill in parent 2, and pending annotation should be satisfied
1439
res = ann._expand_record(p2_key, (), None, [], ('fulltext', False))
1440
ann._annotations_cache[p2_key] = []
1441
res = ann._process_pending(p2_key)
1442
self.assertEqual([rev_key], res)
1443
self.assertEqual({}, ann._pending_annotation)
1444
self.assertEqual({}, ann._pending_deltas)
1446
def test_record_delta_removes_basis(self):
1447
ann = self.make_annotator()
1448
ann._expand_record((b'parent-id',), (), None,
1449
[b'line1\n', b'line2\n'], ('fulltext', False))
1450
ann._num_compression_children[b'parent-id'] = 2
1452
def test_annotate_special_text(self):
1453
ann = self.make_annotator()
1455
rev1_key = (b'rev-1',)
1456
rev2_key = (b'rev-2',)
1457
rev3_key = (b'rev-3',)
1458
spec_key = (b'special:',)
1459
vf.add_lines(rev1_key, [], [b'initial content\n'])
1460
vf.add_lines(rev2_key, [rev1_key], [b'initial content\n',
1461
b'common content\n',
1463
vf.add_lines(rev3_key, [rev1_key], [b'initial content\n',
1464
b'common content\n',
1466
spec_text = (b'initial content\n'
1470
ann.add_special_text(spec_key, [rev2_key, rev3_key], spec_text)
1471
anns, lines = ann.annotate(spec_key)
1472
self.assertEqual([(rev1_key,),
1473
(rev2_key, rev3_key),
1477
self.assertEqualDiff(spec_text, b''.join(lines))
1480
class KnitTests(TestCaseWithTransport):
1481
"""Class containing knit test helper routines."""
1483
def make_test_knit(self, annotate=False, name='test'):
1484
mapper = ConstantMapper(name)
1485
return make_file_factory(annotate, mapper)(self.get_transport())
1488
class TestBadShaError(KnitTests):
1489
"""Tests for handling of sha errors."""
1491
def test_sha_exception_has_text(self):
1492
# having the failed text included in the error allows for recovery.
1493
source = self.make_test_knit()
1494
target = self.make_test_knit(name="target")
1495
if not source._max_delta_chain:
1496
raise TestNotApplicable(
1497
"cannot get delta-caused sha failures without deltas.")
1500
broken = (b'broken',)
1501
source.add_lines(basis, (), [b'foo\n'])
1502
source.add_lines(broken, (basis,), [b'foo\n', b'bar\n'])
1503
# Seed target with a bad basis text
1504
target.add_lines(basis, (), [b'gam\n'])
1505
target.insert_record_stream(
1506
source.get_record_stream([broken], 'unordered', False))
1507
err = self.assertRaises(KnitCorrupt,
1508
next(target.get_record_stream([broken], 'unordered', True
1509
)).get_bytes_as, 'chunked')
1510
self.assertEqual([b'gam\n', b'bar\n'], err.content)
1511
# Test for formatting with live data
1512
self.assertStartsWith(str(err), "Knit ")
1515
class TestKnitIndex(KnitTests):
1517
def test_add_versions_dictionary_compresses(self):
1518
"""Adding versions to the index should update the lookup dict"""
1519
knit = self.make_test_knit()
1521
idx.add_records([((b'a-1',), [b'fulltext'], ((b'a-1',), 0, 0), [])])
1522
self.check_file_contents('test.kndx',
1523
b'# bzr knit index 8\n'
1525
b'a-1 fulltext 0 0 :'
1528
((b'a-2',), [b'fulltext'], ((b'a-2',), 0, 0), [(b'a-1',)]),
1529
((b'a-3',), [b'fulltext'], ((b'a-3',), 0, 0), [(b'a-2',)]),
1531
self.check_file_contents('test.kndx',
1532
b'# bzr knit index 8\n'
1534
b'a-1 fulltext 0 0 :\n'
1535
b'a-2 fulltext 0 0 0 :\n'
1536
b'a-3 fulltext 0 0 1 :'
1538
self.assertEqual({(b'a-3',), (b'a-1',), (b'a-2',)}, idx.keys())
1540
(b'a-1',): (((b'a-1',), 0, 0), None, (), ('fulltext', False)),
1541
(b'a-2',): (((b'a-2',), 0, 0), None, ((b'a-1',),), ('fulltext', False)),
1542
(b'a-3',): (((b'a-3',), 0, 0), None, ((b'a-2',),), ('fulltext', False)),
1543
}, idx.get_build_details(idx.keys()))
1544
self.assertEqual({(b'a-1',): (),
1545
(b'a-2',): ((b'a-1',),),
1546
(b'a-3',): ((b'a-2',),),},
1547
idx.get_parent_map(idx.keys()))
1549
def test_add_versions_fails_clean(self):
1550
"""If add_versions fails in the middle, it restores a pristine state.
1552
Any modifications that are made to the index are reset if all versions
1555
# This cheats a little bit by passing in a generator which will
1556
# raise an exception before the processing finishes
1557
# Other possibilities would be to have an version with the wrong number
1558
# of entries, or to make the backing transport unable to write any
1561
knit = self.make_test_knit()
1563
idx.add_records([((b'a-1',), [b'fulltext'], ((b'a-1',), 0, 0), [])])
1565
class StopEarly(Exception):
1568
def generate_failure():
1569
"""Add some entries and then raise an exception"""
1570
yield ((b'a-2',), [b'fulltext'], (None, 0, 0), (b'a-1',))
1571
yield ((b'a-3',), [b'fulltext'], (None, 0, 0), (b'a-2',))
1574
# Assert the pre-condition
1576
self.assertEqual({(b'a-1',)}, set(idx.keys()))
1578
{(b'a-1',): (((b'a-1',), 0, 0), None, (), ('fulltext', False))},
1579
idx.get_build_details([(b'a-1',)]))
1580
self.assertEqual({(b'a-1',):()}, idx.get_parent_map(idx.keys()))
1583
self.assertRaises(StopEarly, idx.add_records, generate_failure())
1584
# And it shouldn't be modified
1587
def test_knit_index_ignores_empty_files(self):
1588
# There was a race condition in older bzr, where a ^C at the right time
1589
# could leave an empty .kndx file, which bzr would later claim was a
1590
# corrupted file since the header was not present. In reality, the file
1591
# just wasn't created, so it should be ignored.
1592
t = transport.get_transport_from_path('.')
1593
t.put_bytes('test.kndx', b'')
1595
knit = self.make_test_knit()
1597
def test_knit_index_checks_header(self):
1598
t = transport.get_transport_from_path('.')
1599
t.put_bytes('test.kndx', b'# not really a knit header\n\n')
1600
k = self.make_test_knit()
1601
self.assertRaises(KnitHeaderError, k.keys)
1604
class TestGraphIndexKnit(KnitTests):
1605
"""Tests for knits using a GraphIndex rather than a KnitIndex."""
1607
def make_g_index(self, name, ref_lists=0, nodes=[]):
1608
builder = GraphIndexBuilder(ref_lists)
1609
for node, references, value in nodes:
1610
builder.add_node(node, references, value)
1611
stream = builder.finish()
1612
trans = self.get_transport()
1613
size = trans.put_file(name, stream)
1614
return GraphIndex(trans, name, size)
1616
def two_graph_index(self, deltas=False, catch_adds=False):
1617
"""Build a two-graph index.
1619
:param deltas: If true, use underlying indices with two node-ref
1620
lists and 'parent' set to a delta-compressed against tail.
1622
# build a complex graph across several indices.
1624
# delta compression inn the index
1625
index1 = self.make_g_index('1', 2, [
1626
((b'tip', ), b'N0 100', ([(b'parent', )], [], )),
1627
((b'tail', ), b'', ([], []))])
1628
index2 = self.make_g_index('2', 2, [
1629
((b'parent', ), b' 100 78', ([(b'tail', ), (b'ghost', )], [(b'tail', )])),
1630
((b'separate', ), b'', ([], []))])
1632
# just blob location and graph in the index.
1633
index1 = self.make_g_index('1', 1, [
1634
((b'tip', ), b'N0 100', ([(b'parent', )], )),
1635
((b'tail', ), b'', ([], ))])
1636
index2 = self.make_g_index('2', 1, [
1637
((b'parent', ), b' 100 78', ([(b'tail', ), (b'ghost', )], )),
1638
((b'separate', ), b'', ([], ))])
1639
combined_index = CombinedGraphIndex([index1, index2])
1641
self.combined_index = combined_index
1642
self.caught_entries = []
1643
add_callback = self.catch_add
1646
return _KnitGraphIndex(combined_index, lambda:True, deltas=deltas,
1647
add_callback=add_callback)
1649
def test_keys(self):
1650
index = self.two_graph_index()
1651
self.assertEqual({(b'tail',), (b'tip',), (b'parent',), (b'separate',)},
1654
def test_get_position(self):
1655
index = self.two_graph_index()
1656
self.assertEqual((index._graph_index._indices[0], 0, 100), index.get_position((b'tip',)))
1657
self.assertEqual((index._graph_index._indices[1], 100, 78), index.get_position((b'parent',)))
1659
def test_get_method_deltas(self):
1660
index = self.two_graph_index(deltas=True)
1661
self.assertEqual('fulltext', index.get_method((b'tip',)))
1662
self.assertEqual('line-delta', index.get_method((b'parent',)))
1664
def test_get_method_no_deltas(self):
1665
# check that the parent-history lookup is ignored with deltas=False.
1666
index = self.two_graph_index(deltas=False)
1667
self.assertEqual('fulltext', index.get_method((b'tip',)))
1668
self.assertEqual('fulltext', index.get_method((b'parent',)))
1670
def test_get_options_deltas(self):
1671
index = self.two_graph_index(deltas=True)
1672
self.assertEqual([b'fulltext', b'no-eol'], index.get_options((b'tip',)))
1673
self.assertEqual([b'line-delta'], index.get_options((b'parent',)))
1675
def test_get_options_no_deltas(self):
1676
# check that the parent-history lookup is ignored with deltas=False.
1677
index = self.two_graph_index(deltas=False)
1678
self.assertEqual([b'fulltext', b'no-eol'], index.get_options((b'tip',)))
1679
self.assertEqual([b'fulltext'], index.get_options((b'parent',)))
1681
def test_get_parent_map(self):
1682
index = self.two_graph_index()
1683
self.assertEqual({(b'parent',):((b'tail',), (b'ghost',))},
1684
index.get_parent_map([(b'parent',), (b'ghost',)]))
1686
def catch_add(self, entries):
1687
self.caught_entries.append(entries)
1689
def test_add_no_callback_errors(self):
1690
index = self.two_graph_index()
1691
self.assertRaises(errors.ReadOnlyError, index.add_records,
1692
[((b'new',), b'fulltext,no-eol', (None, 50, 60), [b'separate'])])
1694
def test_add_version_smoke(self):
1695
index = self.two_graph_index(catch_adds=True)
1696
index.add_records([((b'new',), b'fulltext,no-eol', (None, 50, 60),
1698
self.assertEqual([[((b'new', ), b'N50 60', (((b'separate',),),))]],
1699
self.caught_entries)
1701
def test_add_version_delta_not_delta_index(self):
1702
index = self.two_graph_index(catch_adds=True)
1703
self.assertRaises(KnitCorrupt, index.add_records,
1704
[((b'new',), b'no-eol,line-delta', (None, 0, 100), [(b'parent',)])])
1705
self.assertEqual([], self.caught_entries)
1707
def test_add_version_same_dup(self):
1708
index = self.two_graph_index(catch_adds=True)
1709
# options can be spelt two different ways
1710
index.add_records([((b'tip',), b'fulltext,no-eol', (None, 0, 100), [(b'parent',)])])
1711
index.add_records([((b'tip',), b'no-eol,fulltext', (None, 0, 100), [(b'parent',)])])
1712
# position/length are ignored (because each pack could have fulltext or
1713
# delta, and be at a different position.
1714
index.add_records([((b'tip',), b'fulltext,no-eol', (None, 50, 100),
1716
index.add_records([((b'tip',), b'fulltext,no-eol', (None, 0, 1000),
1718
# but neither should have added data:
1719
self.assertEqual([[], [], [], []], self.caught_entries)
1721
def test_add_version_different_dup(self):
1722
index = self.two_graph_index(deltas=True, catch_adds=True)
1724
self.assertRaises(KnitCorrupt, index.add_records,
1725
[((b'tip',), b'line-delta', (None, 0, 100), [(b'parent',)])])
1726
self.assertRaises(KnitCorrupt, index.add_records,
1727
[((b'tip',), b'fulltext', (None, 0, 100), [(b'parent',)])])
1729
self.assertRaises(KnitCorrupt, index.add_records,
1730
[((b'tip',), b'fulltext,no-eol', (None, 0, 100), [])])
1731
self.assertEqual([], self.caught_entries)
1733
def test_add_versions_nodeltas(self):
1734
index = self.two_graph_index(catch_adds=True)
1736
((b'new',), b'fulltext,no-eol', (None, 50, 60), [(b'separate',)]),
1737
((b'new2',), b'fulltext', (None, 0, 6), [(b'new',)]),
1739
self.assertEqual([((b'new', ), b'N50 60', (((b'separate',),),)),
1740
((b'new2', ), b' 0 6', (((b'new',),),))],
1741
sorted(self.caught_entries[0]))
1742
self.assertEqual(1, len(self.caught_entries))
1744
def test_add_versions_deltas(self):
1745
index = self.two_graph_index(deltas=True, catch_adds=True)
1747
((b'new',), b'fulltext,no-eol', (None, 50, 60), [(b'separate',)]),
1748
((b'new2',), b'line-delta', (None, 0, 6), [(b'new',)]),
1750
self.assertEqual([((b'new', ), b'N50 60', (((b'separate',),), ())),
1751
((b'new2', ), b' 0 6', (((b'new',),), ((b'new',),), ))],
1752
sorted(self.caught_entries[0]))
1753
self.assertEqual(1, len(self.caught_entries))
1755
def test_add_versions_delta_not_delta_index(self):
1756
index = self.two_graph_index(catch_adds=True)
1757
self.assertRaises(KnitCorrupt, index.add_records,
1758
[((b'new',), b'no-eol,line-delta', (None, 0, 100), [(b'parent',)])])
1759
self.assertEqual([], self.caught_entries)
1761
def test_add_versions_random_id_accepted(self):
1762
index = self.two_graph_index(catch_adds=True)
1763
index.add_records([], random_id=True)
1765
def test_add_versions_same_dup(self):
1766
index = self.two_graph_index(catch_adds=True)
1767
# options can be spelt two different ways
1768
index.add_records([((b'tip',), b'fulltext,no-eol', (None, 0, 100),
1770
index.add_records([((b'tip',), b'no-eol,fulltext', (None, 0, 100),
1772
# position/length are ignored (because each pack could have fulltext or
1773
# delta, and be at a different position.
1774
index.add_records([((b'tip',), b'fulltext,no-eol', (None, 50, 100),
1776
index.add_records([((b'tip',), b'fulltext,no-eol', (None, 0, 1000),
1778
# but neither should have added data.
1779
self.assertEqual([[], [], [], []], self.caught_entries)
1781
def test_add_versions_different_dup(self):
1782
index = self.two_graph_index(deltas=True, catch_adds=True)
1784
self.assertRaises(KnitCorrupt, index.add_records,
1785
[((b'tip',), b'line-delta', (None, 0, 100), [(b'parent',)])])
1786
self.assertRaises(KnitCorrupt, index.add_records,
1787
[((b'tip',), b'fulltext', (None, 0, 100), [(b'parent',)])])
1789
self.assertRaises(KnitCorrupt, index.add_records,
1790
[((b'tip',), b'fulltext,no-eol', (None, 0, 100), [])])
1791
# change options in the second record
1792
self.assertRaises(KnitCorrupt, index.add_records,
1793
[((b'tip',), b'fulltext,no-eol', (None, 0, 100), [(b'parent',)]),
1794
((b'tip',), b'line-delta', (None, 0, 100), [(b'parent',)])])
1795
self.assertEqual([], self.caught_entries)
1797
def make_g_index_missing_compression_parent(self):
1798
graph_index = self.make_g_index('missing_comp', 2,
1799
[((b'tip', ), b' 100 78',
1800
([(b'missing-parent', ), (b'ghost', )], [(b'missing-parent', )]))])
1803
def make_g_index_missing_parent(self):
1804
graph_index = self.make_g_index('missing_parent', 2,
1805
[((b'parent', ), b' 100 78', ([], [])),
1806
((b'tip', ), b' 100 78',
1807
([(b'parent', ), (b'missing-parent', )], [(b'parent', )])),
1811
def make_g_index_no_external_refs(self):
1812
graph_index = self.make_g_index('no_external_refs', 2,
1813
[((b'rev', ), b' 100 78',
1814
([(b'parent', ), (b'ghost', )], []))])
1817
def test_add_good_unvalidated_index(self):
1818
unvalidated = self.make_g_index_no_external_refs()
1819
combined = CombinedGraphIndex([unvalidated])
1820
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1821
index.scan_unvalidated_index(unvalidated)
1822
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1824
def test_add_missing_compression_parent_unvalidated_index(self):
1825
unvalidated = self.make_g_index_missing_compression_parent()
1826
combined = CombinedGraphIndex([unvalidated])
1827
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1828
index.scan_unvalidated_index(unvalidated)
1829
# This also checks that its only the compression parent that is
1830
# examined, otherwise 'ghost' would also be reported as a missing
1833
frozenset([(b'missing-parent',)]),
1834
index.get_missing_compression_parents())
1836
def test_add_missing_noncompression_parent_unvalidated_index(self):
1837
unvalidated = self.make_g_index_missing_parent()
1838
combined = CombinedGraphIndex([unvalidated])
1839
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1840
track_external_parent_refs=True)
1841
index.scan_unvalidated_index(unvalidated)
1843
frozenset([(b'missing-parent',)]), index.get_missing_parents())
1845
def test_track_external_parent_refs(self):
1846
g_index = self.make_g_index('empty', 2, [])
1847
combined = CombinedGraphIndex([g_index])
1848
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1849
add_callback=self.catch_add, track_external_parent_refs=True)
1850
self.caught_entries = []
1852
((b'new-key',), b'fulltext,no-eol', (None, 50, 60),
1853
[(b'parent-1',), (b'parent-2',)])])
1855
frozenset([(b'parent-1',), (b'parent-2',)]),
1856
index.get_missing_parents())
1858
def test_add_unvalidated_index_with_present_external_references(self):
1859
index = self.two_graph_index(deltas=True)
1860
# Ugly hack to get at one of the underlying GraphIndex objects that
1861
# two_graph_index built.
1862
unvalidated = index._graph_index._indices[1]
1863
# 'parent' is an external ref of _indices[1] (unvalidated), but is
1864
# present in _indices[0].
1865
index.scan_unvalidated_index(unvalidated)
1866
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1868
def make_new_missing_parent_g_index(self, name):
1869
missing_parent = name.encode('ascii') + b'-missing-parent'
1870
graph_index = self.make_g_index(name, 2,
1871
[((name.encode('ascii') + b'tip', ), b' 100 78',
1872
([(missing_parent, ), (b'ghost', )], [(missing_parent, )]))])
1875
def test_add_mulitiple_unvalidated_indices_with_missing_parents(self):
1876
g_index_1 = self.make_new_missing_parent_g_index('one')
1877
g_index_2 = self.make_new_missing_parent_g_index('two')
1878
combined = CombinedGraphIndex([g_index_1, g_index_2])
1879
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1880
index.scan_unvalidated_index(g_index_1)
1881
index.scan_unvalidated_index(g_index_2)
1883
frozenset([(b'one-missing-parent',), (b'two-missing-parent',)]),
1884
index.get_missing_compression_parents())
1886
def test_add_mulitiple_unvalidated_indices_with_mutual_dependencies(self):
1887
graph_index_a = self.make_g_index('one', 2,
1888
[((b'parent-one', ), b' 100 78', ([(b'non-compression-parent',)], [])),
1889
((b'child-of-two', ), b' 100 78',
1890
([(b'parent-two',)], [(b'parent-two',)]))])
1891
graph_index_b = self.make_g_index('two', 2,
1892
[((b'parent-two', ), b' 100 78', ([(b'non-compression-parent',)], [])),
1893
((b'child-of-one', ), b' 100 78',
1894
([(b'parent-one',)], [(b'parent-one',)]))])
1895
combined = CombinedGraphIndex([graph_index_a, graph_index_b])
1896
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1897
index.scan_unvalidated_index(graph_index_a)
1898
index.scan_unvalidated_index(graph_index_b)
1900
frozenset([]), index.get_missing_compression_parents())
1903
class TestNoParentsGraphIndexKnit(KnitTests):
1904
"""Tests for knits using _KnitGraphIndex with no parents."""
1906
def make_g_index(self, name, ref_lists=0, nodes=[]):
1907
builder = GraphIndexBuilder(ref_lists)
1908
for node, references in nodes:
1909
builder.add_node(node, references)
1910
stream = builder.finish()
1911
trans = self.get_transport()
1912
size = trans.put_file(name, stream)
1913
return GraphIndex(trans, name, size)
1915
def test_add_good_unvalidated_index(self):
1916
unvalidated = self.make_g_index('unvalidated')
1917
combined = CombinedGraphIndex([unvalidated])
1918
index = _KnitGraphIndex(combined, lambda: True, parents=False)
1919
index.scan_unvalidated_index(unvalidated)
1920
self.assertEqual(frozenset(),
1921
index.get_missing_compression_parents())
1923
def test_parents_deltas_incompatible(self):
1924
index = CombinedGraphIndex([])
1925
self.assertRaises(knit.KnitError, _KnitGraphIndex, lambda:True,
1926
index, deltas=True, parents=False)
1928
def two_graph_index(self, catch_adds=False):
1929
"""Build a two-graph index.
1931
:param deltas: If true, use underlying indices with two node-ref
1932
lists and 'parent' set to a delta-compressed against tail.
1934
# put several versions in the index.
1935
index1 = self.make_g_index('1', 0, [
1936
((b'tip', ), b'N0 100'),
1937
((b'tail', ), b'')])
1938
index2 = self.make_g_index('2', 0, [
1939
((b'parent', ), b' 100 78'),
1940
((b'separate', ), b'')])
1941
combined_index = CombinedGraphIndex([index1, index2])
1943
self.combined_index = combined_index
1944
self.caught_entries = []
1945
add_callback = self.catch_add
1948
return _KnitGraphIndex(combined_index, lambda:True, parents=False,
1949
add_callback=add_callback)
1951
def test_keys(self):
1952
index = self.two_graph_index()
1953
self.assertEqual({(b'tail',), (b'tip',), (b'parent',), (b'separate',)},
1956
def test_get_position(self):
1957
index = self.two_graph_index()
1958
self.assertEqual((index._graph_index._indices[0], 0, 100),
1959
index.get_position((b'tip',)))
1960
self.assertEqual((index._graph_index._indices[1], 100, 78),
1961
index.get_position((b'parent',)))
1963
def test_get_method(self):
1964
index = self.two_graph_index()
1965
self.assertEqual('fulltext', index.get_method((b'tip',)))
1966
self.assertEqual([b'fulltext'], index.get_options((b'parent',)))
1968
def test_get_options(self):
1969
index = self.two_graph_index()
1970
self.assertEqual([b'fulltext', b'no-eol'], index.get_options((b'tip',)))
1971
self.assertEqual([b'fulltext'], index.get_options((b'parent',)))
1973
def test_get_parent_map(self):
1974
index = self.two_graph_index()
1975
self.assertEqual({(b'parent',):None},
1976
index.get_parent_map([(b'parent',), (b'ghost',)]))
1978
def catch_add(self, entries):
1979
self.caught_entries.append(entries)
1981
def test_add_no_callback_errors(self):
1982
index = self.two_graph_index()
1983
self.assertRaises(errors.ReadOnlyError, index.add_records,
1984
[((b'new',), b'fulltext,no-eol', (None, 50, 60), [(b'separate',)])])
1986
def test_add_version_smoke(self):
1987
index = self.two_graph_index(catch_adds=True)
1988
index.add_records([((b'new',), b'fulltext,no-eol', (None, 50, 60), [])])
1989
self.assertEqual([[((b'new', ), b'N50 60')]],
1990
self.caught_entries)
1992
def test_add_version_delta_not_delta_index(self):
1993
index = self.two_graph_index(catch_adds=True)
1994
self.assertRaises(KnitCorrupt, index.add_records,
1995
[((b'new',), b'no-eol,line-delta', (None, 0, 100), [])])
1996
self.assertEqual([], self.caught_entries)
1998
def test_add_version_same_dup(self):
1999
index = self.two_graph_index(catch_adds=True)
2000
# options can be spelt two different ways
2001
index.add_records([((b'tip',), b'fulltext,no-eol', (None, 0, 100), [])])
2002
index.add_records([((b'tip',), b'no-eol,fulltext', (None, 0, 100), [])])
2003
# position/length are ignored (because each pack could have fulltext or
2004
# delta, and be at a different position.
2005
index.add_records([((b'tip',), b'fulltext,no-eol', (None, 50, 100), [])])
2006
index.add_records([((b'tip',), b'fulltext,no-eol', (None, 0, 1000), [])])
2007
# but neither should have added data.
2008
self.assertEqual([[], [], [], []], self.caught_entries)
2010
def test_add_version_different_dup(self):
2011
index = self.two_graph_index(catch_adds=True)
2013
self.assertRaises(KnitCorrupt, index.add_records,
2014
[((b'tip',), b'no-eol,line-delta', (None, 0, 100), [])])
2015
self.assertRaises(KnitCorrupt, index.add_records,
2016
[((b'tip',), b'line-delta,no-eol', (None, 0, 100), [])])
2017
self.assertRaises(KnitCorrupt, index.add_records,
2018
[((b'tip',), b'fulltext', (None, 0, 100), [])])
2020
self.assertRaises(KnitCorrupt, index.add_records,
2021
[((b'tip',), b'fulltext,no-eol', (None, 0, 100), [(b'parent',)])])
2022
self.assertEqual([], self.caught_entries)
2024
def test_add_versions(self):
2025
index = self.two_graph_index(catch_adds=True)
2027
((b'new',), b'fulltext,no-eol', (None, 50, 60), []),
2028
((b'new2',), b'fulltext', (None, 0, 6), []),
2030
self.assertEqual([((b'new', ), b'N50 60'), ((b'new2', ), b' 0 6')],
2031
sorted(self.caught_entries[0]))
2032
self.assertEqual(1, len(self.caught_entries))
2034
def test_add_versions_delta_not_delta_index(self):
2035
index = self.two_graph_index(catch_adds=True)
2036
self.assertRaises(KnitCorrupt, index.add_records,
2037
[((b'new',), b'no-eol,line-delta', (None, 0, 100), [(b'parent',)])])
2038
self.assertEqual([], self.caught_entries)
2040
def test_add_versions_parents_not_parents_index(self):
2041
index = self.two_graph_index(catch_adds=True)
2042
self.assertRaises(KnitCorrupt, index.add_records,
2043
[((b'new',), b'no-eol,fulltext', (None, 0, 100), [(b'parent',)])])
2044
self.assertEqual([], self.caught_entries)
2046
def test_add_versions_random_id_accepted(self):
2047
index = self.two_graph_index(catch_adds=True)
2048
index.add_records([], random_id=True)
2050
def test_add_versions_same_dup(self):
2051
index = self.two_graph_index(catch_adds=True)
2052
# options can be spelt two different ways
2053
index.add_records([((b'tip',), b'fulltext,no-eol', (None, 0, 100), [])])
2054
index.add_records([((b'tip',), b'no-eol,fulltext', (None, 0, 100), [])])
2055
# position/length are ignored (because each pack could have fulltext or
2056
# delta, and be at a different position.
2057
index.add_records([((b'tip',), b'fulltext,no-eol', (None, 50, 100), [])])
2058
index.add_records([((b'tip',), b'fulltext,no-eol', (None, 0, 1000), [])])
2059
# but neither should have added data.
2060
self.assertEqual([[], [], [], []], self.caught_entries)
2062
def test_add_versions_different_dup(self):
2063
index = self.two_graph_index(catch_adds=True)
2065
self.assertRaises(KnitCorrupt, index.add_records,
2066
[((b'tip',), b'no-eol,line-delta', (None, 0, 100), [])])
2067
self.assertRaises(KnitCorrupt, index.add_records,
2068
[((b'tip',), b'line-delta,no-eol', (None, 0, 100), [])])
2069
self.assertRaises(KnitCorrupt, index.add_records,
2070
[((b'tip',), b'fulltext', (None, 0, 100), [])])
2072
self.assertRaises(KnitCorrupt, index.add_records,
2073
[((b'tip',), b'fulltext,no-eol', (None, 0, 100), [(b'parent',)])])
2074
# change options in the second record
2075
self.assertRaises(KnitCorrupt, index.add_records,
2076
[((b'tip',), b'fulltext,no-eol', (None, 0, 100), []),
2077
((b'tip',), b'no-eol,line-delta', (None, 0, 100), [])])
2078
self.assertEqual([], self.caught_entries)
2081
class TestKnitVersionedFiles(KnitTests):
2083
def assertGroupKeysForIo(self, exp_groups, keys, non_local_keys,
2084
positions, _min_buffer_size=None):
2085
kvf = self.make_test_knit()
2086
if _min_buffer_size is None:
2087
_min_buffer_size = knit._STREAM_MIN_BUFFER_SIZE
2088
self.assertEqual(exp_groups, kvf._group_keys_for_io(keys,
2089
non_local_keys, positions,
2090
_min_buffer_size=_min_buffer_size))
2092
def assertSplitByPrefix(self, expected_map, expected_prefix_order,
2094
split, prefix_order = KnitVersionedFiles._split_by_prefix(keys)
2095
self.assertEqual(expected_map, split)
2096
self.assertEqual(expected_prefix_order, prefix_order)
2098
def test__group_keys_for_io(self):
2099
ft_detail = ('fulltext', False)
2100
ld_detail = ('line-delta', False)
2108
f_a: (ft_detail, (f_a, 0, 100), None),
2109
f_b: (ld_detail, (f_b, 100, 21), f_a),
2110
f_c: (ld_detail, (f_c, 180, 15), f_b),
2111
g_a: (ft_detail, (g_a, 121, 35), None),
2112
g_b: (ld_detail, (g_b, 156, 12), g_a),
2113
g_c: (ld_detail, (g_c, 195, 13), g_a),
2115
self.assertGroupKeysForIo([([f_a], set())],
2116
[f_a], [], positions)
2117
self.assertGroupKeysForIo([([f_a], {f_a})],
2118
[f_a], [f_a], positions)
2119
self.assertGroupKeysForIo([([f_a, f_b], set([]))],
2120
[f_a, f_b], [], positions)
2121
self.assertGroupKeysForIo([([f_a, f_b], {f_b})],
2122
[f_a, f_b], [f_b], positions)
2123
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2124
[f_a, g_a, f_b, g_b], [], positions)
2125
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2126
[f_a, g_a, f_b, g_b], [], positions,
2127
_min_buffer_size=150)
2128
self.assertGroupKeysForIo([([f_a, f_b], set()), ([g_a, g_b], set())],
2129
[f_a, g_a, f_b, g_b], [], positions,
2130
_min_buffer_size=100)
2131
self.assertGroupKeysForIo([([f_c], set()), ([g_b], set())],
2132
[f_c, g_b], [], positions,
2133
_min_buffer_size=125)
2134
self.assertGroupKeysForIo([([g_b, f_c], set())],
2135
[g_b, f_c], [], positions,
2136
_min_buffer_size=125)
2138
def test__split_by_prefix(self):
2139
self.assertSplitByPrefix({b'f': [(b'f', b'a'), (b'f', b'b')],
2140
b'g': [(b'g', b'b'), (b'g', b'a')],
2142
[(b'f', b'a'), (b'g', b'b'),
2143
(b'g', b'a'), (b'f', b'b')])
2145
self.assertSplitByPrefix({b'f': [(b'f', b'a'), (b'f', b'b')],
2146
b'g': [(b'g', b'b'), (b'g', b'a')],
2148
[(b'f', b'a'), (b'f', b'b'),
2149
(b'g', b'b'), (b'g', b'a')])
2151
self.assertSplitByPrefix({b'f': [(b'f', b'a'), (b'f', b'b')],
2152
b'g': [(b'g', b'b'), (b'g', b'a')],
2154
[(b'f', b'a'), (b'f', b'b'),
2155
(b'g', b'b'), (b'g', b'a')])
2157
self.assertSplitByPrefix({b'f': [(b'f', b'a'), (b'f', b'b')],
2158
b'g': [(b'g', b'b'), (b'g', b'a')],
2159
b'': [(b'a',), (b'b',)]
2160
}, [b'f', b'g', b''],
2161
[(b'f', b'a'), (b'g', b'b'),
2163
(b'g', b'a'), (b'f', b'b')])
2166
class TestStacking(KnitTests):
2168
def get_basis_and_test_knit(self):
2169
basis = self.make_test_knit(name='basis')
2170
basis = RecordingVersionedFilesDecorator(basis)
2171
test = self.make_test_knit(name='test')
2172
test.add_fallback_versioned_files(basis)
2175
def test_add_fallback_versioned_files(self):
2176
basis = self.make_test_knit(name='basis')
2177
test = self.make_test_knit(name='test')
2178
# It must not error; other tests test that the fallback is referred to
2179
# when accessing data.
2180
test.add_fallback_versioned_files(basis)
2182
def test_add_lines(self):
2183
# lines added to the test are not added to the basis
2184
basis, test = self.get_basis_and_test_knit()
2186
key_basis = (b'bar',)
2187
key_cross_border = (b'quux',)
2188
key_delta = (b'zaphod',)
2189
test.add_lines(key, (), [b'foo\n'])
2190
self.assertEqual({}, basis.get_parent_map([key]))
2191
# lines added to the test that reference across the stack do a
2193
basis.add_lines(key_basis, (), [b'foo\n'])
2195
test.add_lines(key_cross_border, (key_basis,), [b'foo\n'])
2196
self.assertEqual('fulltext', test._index.get_method(key_cross_border))
2197
# we don't even need to look at the basis to see that this should be
2198
# stored as a fulltext
2199
self.assertEqual([], basis.calls)
2200
# Subsequent adds do delta.
2202
test.add_lines(key_delta, (key_cross_border,), [b'foo\n'])
2203
self.assertEqual('line-delta', test._index.get_method(key_delta))
2204
self.assertEqual([], basis.calls)
2206
def test_annotate(self):
2207
# annotations from the test knit are answered without asking the basis
2208
basis, test = self.get_basis_and_test_knit()
2210
key_basis = (b'bar',)
2211
key_missing = (b'missing',)
2212
test.add_lines(key, (), [b'foo\n'])
2213
details = test.annotate(key)
2214
self.assertEqual([(key, b'foo\n')], details)
2215
self.assertEqual([], basis.calls)
2216
# But texts that are not in the test knit are looked for in the basis
2218
basis.add_lines(key_basis, (), [b'foo\n', b'bar\n'])
2220
details = test.annotate(key_basis)
2221
self.assertEqual([(key_basis, b'foo\n'), (key_basis, b'bar\n')], details)
2222
# Not optimised to date:
2223
# self.assertEqual([("annotate", key_basis)], basis.calls)
2224
self.assertEqual([('get_parent_map', {key_basis}),
2225
('get_parent_map', {key_basis}),
2226
('get_record_stream', [key_basis], 'topological', True)],
2229
def test_check(self):
2230
# At the moment checking a stacked knit does implicitly check the
2232
basis, test = self.get_basis_and_test_knit()
2235
def test_get_parent_map(self):
2236
# parents in the test knit are answered without asking the basis
2237
basis, test = self.get_basis_and_test_knit()
2239
key_basis = (b'bar',)
2240
key_missing = (b'missing',)
2241
test.add_lines(key, (), [])
2242
parent_map = test.get_parent_map([key])
2243
self.assertEqual({key: ()}, parent_map)
2244
self.assertEqual([], basis.calls)
2245
# But parents that are not in the test knit are looked for in the basis
2246
basis.add_lines(key_basis, (), [])
2248
parent_map = test.get_parent_map([key, key_basis, key_missing])
2249
self.assertEqual({key: (),
2250
key_basis: ()}, parent_map)
2251
self.assertEqual([("get_parent_map", {key_basis, key_missing})],
2254
def test_get_record_stream_unordered_fulltexts(self):
2255
# records from the test knit are answered without asking the basis:
2256
basis, test = self.get_basis_and_test_knit()
2258
key_basis = (b'bar',)
2259
key_missing = (b'missing',)
2260
test.add_lines(key, (), [b'foo\n'])
2261
records = list(test.get_record_stream([key], 'unordered', True))
2262
self.assertEqual(1, len(records))
2263
self.assertEqual([], basis.calls)
2264
# Missing (from test knit) objects are retrieved from the basis:
2265
basis.add_lines(key_basis, (), [b'foo\n', b'bar\n'])
2267
records = list(test.get_record_stream([key_basis, key_missing],
2269
self.assertEqual(2, len(records))
2270
calls = list(basis.calls)
2271
for record in records:
2272
self.assertSubset([record.key], (key_basis, key_missing))
2273
if record.key == key_missing:
2274
self.assertIsInstance(record, AbsentContentFactory)
2276
reference = list(basis.get_record_stream([key_basis],
2277
'unordered', True))[0]
2278
self.assertEqual(reference.key, record.key)
2279
self.assertEqual(reference.sha1, record.sha1)
2280
self.assertEqual(reference.storage_kind, record.storage_kind)
2281
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2282
record.get_bytes_as(record.storage_kind))
2283
self.assertEqual(reference.get_bytes_as('fulltext'),
2284
record.get_bytes_as('fulltext'))
2285
# It's not strictly minimal, but it seems reasonable for now for it to
2286
# ask which fallbacks have which parents.
2288
("get_parent_map", {key_basis, key_missing}),
2289
("get_record_stream", [key_basis], 'unordered', True)],
2292
def test_get_record_stream_ordered_fulltexts(self):
2293
# ordering is preserved down into the fallback store.
2294
basis, test = self.get_basis_and_test_knit()
2296
key_basis = (b'bar',)
2297
key_basis_2 = (b'quux',)
2298
key_missing = (b'missing',)
2299
test.add_lines(key, (key_basis,), [b'foo\n'])
2300
# Missing (from test knit) objects are retrieved from the basis:
2301
basis.add_lines(key_basis, (key_basis_2,), [b'foo\n', b'bar\n'])
2302
basis.add_lines(key_basis_2, (), [b'quux\n'])
2304
# ask for in non-topological order
2305
records = list(test.get_record_stream(
2306
[key, key_basis, key_missing, key_basis_2], 'topological', True))
2307
self.assertEqual(4, len(records))
2309
for record in records:
2310
self.assertSubset([record.key],
2311
(key_basis, key_missing, key_basis_2, key))
2312
if record.key == key_missing:
2313
self.assertIsInstance(record, AbsentContentFactory)
2315
results.append((record.key, record.sha1, record.storage_kind,
2316
record.get_bytes_as('fulltext')))
2317
calls = list(basis.calls)
2318
order = [record[0] for record in results]
2319
self.assertEqual([key_basis_2, key_basis, key], order)
2320
for result in results:
2321
if result[0] == key:
2325
record = next(source.get_record_stream([result[0]], 'unordered',
2327
self.assertEqual(record.key, result[0])
2328
self.assertEqual(record.sha1, result[1])
2329
# We used to check that the storage kind matched, but actually it
2330
# depends on whether it was sourced from the basis, or in a single
2331
# group, because asking for full texts returns proxy objects to a
2332
# _ContentMapGenerator object; so checking the kind is unneeded.
2333
self.assertEqual(record.get_bytes_as('fulltext'), result[3])
2334
# It's not strictly minimal, but it seems reasonable for now for it to
2335
# ask which fallbacks have which parents.
2336
self.assertEqual(2, len(calls))
2338
("get_parent_map", {key_basis, key_basis_2, key_missing}),
2340
# topological is requested from the fallback, because that is what
2341
# was requested at the top level.
2344
("get_record_stream", [key_basis_2, key_basis], 'topological', True),
2345
("get_record_stream", [key_basis, key_basis_2], 'topological', True)])
2347
def test_get_record_stream_unordered_deltas(self):
2348
# records from the test knit are answered without asking the basis:
2349
basis, test = self.get_basis_and_test_knit()
2351
key_basis = (b'bar',)
2352
key_missing = (b'missing',)
2353
test.add_lines(key, (), [b'foo\n'])
2354
records = list(test.get_record_stream([key], 'unordered', False))
2355
self.assertEqual(1, len(records))
2356
self.assertEqual([], basis.calls)
2357
# Missing (from test knit) objects are retrieved from the basis:
2358
basis.add_lines(key_basis, (), [b'foo\n', b'bar\n'])
2360
records = list(test.get_record_stream([key_basis, key_missing],
2361
'unordered', False))
2362
self.assertEqual(2, len(records))
2363
calls = list(basis.calls)
2364
for record in records:
2365
self.assertSubset([record.key], (key_basis, key_missing))
2366
if record.key == key_missing:
2367
self.assertIsInstance(record, AbsentContentFactory)
2369
reference = list(basis.get_record_stream([key_basis],
2370
'unordered', False))[0]
2371
self.assertEqual(reference.key, record.key)
2372
self.assertEqual(reference.sha1, record.sha1)
2373
self.assertEqual(reference.storage_kind, record.storage_kind)
2374
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2375
record.get_bytes_as(record.storage_kind))
2376
# It's not strictly minimal, but it seems reasonable for now for it to
2377
# ask which fallbacks have which parents.
2379
("get_parent_map", {key_basis, key_missing}),
2380
("get_record_stream", [key_basis], 'unordered', False)],
2383
def test_get_record_stream_ordered_deltas(self):
2384
# ordering is preserved down into the fallback store.
2385
basis, test = self.get_basis_and_test_knit()
2387
key_basis = (b'bar',)
2388
key_basis_2 = (b'quux',)
2389
key_missing = (b'missing',)
2390
test.add_lines(key, (key_basis,), [b'foo\n'])
2391
# Missing (from test knit) objects are retrieved from the basis:
2392
basis.add_lines(key_basis, (key_basis_2,), [b'foo\n', b'bar\n'])
2393
basis.add_lines(key_basis_2, (), [b'quux\n'])
2395
# ask for in non-topological order
2396
records = list(test.get_record_stream(
2397
[key, key_basis, key_missing, key_basis_2], 'topological', False))
2398
self.assertEqual(4, len(records))
2400
for record in records:
2401
self.assertSubset([record.key],
2402
(key_basis, key_missing, key_basis_2, key))
2403
if record.key == key_missing:
2404
self.assertIsInstance(record, AbsentContentFactory)
2406
results.append((record.key, record.sha1, record.storage_kind,
2407
record.get_bytes_as(record.storage_kind)))
2408
calls = list(basis.calls)
2409
order = [record[0] for record in results]
2410
self.assertEqual([key_basis_2, key_basis, key], order)
2411
for result in results:
2412
if result[0] == key:
2416
record = next(source.get_record_stream([result[0]], 'unordered',
2418
self.assertEqual(record.key, result[0])
2419
self.assertEqual(record.sha1, result[1])
2420
self.assertEqual(record.storage_kind, result[2])
2421
self.assertEqual(record.get_bytes_as(record.storage_kind), result[3])
2422
# It's not strictly minimal, but it seems reasonable for now for it to
2423
# ask which fallbacks have which parents.
2425
("get_parent_map", {key_basis, key_basis_2, key_missing}),
2426
("get_record_stream", [key_basis_2, key_basis], 'topological', False)],
2429
def test_get_sha1s(self):
2430
# sha1's in the test knit are answered without asking the basis
2431
basis, test = self.get_basis_and_test_knit()
2433
key_basis = (b'bar',)
2434
key_missing = (b'missing',)
2435
test.add_lines(key, (), [b'foo\n'])
2436
key_sha1sum = osutils.sha_string(b'foo\n')
2437
sha1s = test.get_sha1s([key])
2438
self.assertEqual({key: key_sha1sum}, sha1s)
2439
self.assertEqual([], basis.calls)
2440
# But texts that are not in the test knit are looked for in the basis
2441
# directly (rather than via text reconstruction) so that remote servers
2442
# etc don't have to answer with full content.
2443
basis.add_lines(key_basis, (), [b'foo\n', b'bar\n'])
2444
basis_sha1sum = osutils.sha_string(b'foo\nbar\n')
2446
sha1s = test.get_sha1s([key, key_missing, key_basis])
2447
self.assertEqual({key: key_sha1sum,
2448
key_basis: basis_sha1sum}, sha1s)
2449
self.assertEqual([("get_sha1s", {key_basis, key_missing})],
2452
def test_insert_record_stream(self):
2453
# records are inserted as normal; insert_record_stream builds on
2454
# add_lines, so a smoke test should be all that's needed:
2456
key_basis = (b'bar',)
2457
key_delta = (b'zaphod',)
2458
basis, test = self.get_basis_and_test_knit()
2459
source = self.make_test_knit(name='source')
2460
basis.add_lines(key_basis, (), [b'foo\n'])
2462
source.add_lines(key_basis, (), [b'foo\n'])
2463
source.add_lines(key_delta, (key_basis,), [b'bar\n'])
2464
stream = source.get_record_stream([key_delta], 'unordered', False)
2465
test.insert_record_stream(stream)
2466
# XXX: this does somewhat too many calls in making sure of whether it
2467
# has to recreate the full text.
2468
self.assertEqual([("get_parent_map", {key_basis}),
2469
('get_parent_map', {key_basis}),
2470
('get_record_stream', [key_basis], 'unordered', True)],
2472
self.assertEqual({key_delta:(key_basis,)},
2473
test.get_parent_map([key_delta]))
2474
self.assertEqual(b'bar\n', next(test.get_record_stream([key_delta],
2475
'unordered', True)).get_bytes_as('fulltext'))
2477
def test_iter_lines_added_or_present_in_keys(self):
2478
# Lines from the basis are returned, and lines for a given key are only
2482
# all sources are asked for keys:
2483
basis, test = self.get_basis_and_test_knit()
2484
basis.add_lines(key1, (), [b"foo"])
2486
lines = list(test.iter_lines_added_or_present_in_keys([key1]))
2487
self.assertEqual([(b"foo\n", key1)], lines)
2488
self.assertEqual([("iter_lines_added_or_present_in_keys", {key1})],
2490
# keys in both are not duplicated:
2491
test.add_lines(key2, (), [b"bar\n"])
2492
basis.add_lines(key2, (), [b"bar\n"])
2494
lines = list(test.iter_lines_added_or_present_in_keys([key2]))
2495
self.assertEqual([(b"bar\n", key2)], lines)
2496
self.assertEqual([], basis.calls)
2498
def test_keys(self):
2501
# all sources are asked for keys:
2502
basis, test = self.get_basis_and_test_knit()
2504
self.assertEqual(set(), set(keys))
2505
self.assertEqual([("keys",)], basis.calls)
2506
# keys from a basis are returned:
2507
basis.add_lines(key1, (), [])
2510
self.assertEqual({key1}, set(keys))
2511
self.assertEqual([("keys",)], basis.calls)
2512
# keys in both are not duplicated:
2513
test.add_lines(key2, (), [])
2514
basis.add_lines(key2, (), [])
2517
self.assertEqual(2, len(keys))
2518
self.assertEqual({key1, key2}, set(keys))
2519
self.assertEqual([("keys",)], basis.calls)
2521
def test_add_mpdiffs(self):
2522
# records are inserted as normal; add_mpdiff builds on
2523
# add_lines, so a smoke test should be all that's needed:
2525
key_basis = (b'bar',)
2526
key_delta = (b'zaphod',)
2527
basis, test = self.get_basis_and_test_knit()
2528
source = self.make_test_knit(name='source')
2529
basis.add_lines(key_basis, (), [b'foo\n'])
2531
source.add_lines(key_basis, (), [b'foo\n'])
2532
source.add_lines(key_delta, (key_basis,), [b'bar\n'])
2533
diffs = source.make_mpdiffs([key_delta])
2534
test.add_mpdiffs([(key_delta, (key_basis,),
2535
source.get_sha1s([key_delta])[key_delta], diffs[0])])
2536
self.assertEqual([("get_parent_map", {key_basis}),
2537
('get_record_stream', [key_basis], 'unordered', True),],
2539
self.assertEqual({key_delta:(key_basis,)},
2540
test.get_parent_map([key_delta]))
2541
self.assertEqual(b'bar\n', next(test.get_record_stream([key_delta],
2542
'unordered', True)).get_bytes_as('fulltext'))
2544
def test_make_mpdiffs(self):
2545
# Generating an mpdiff across a stacking boundary should detect parent
2548
key_left = (b'bar',)
2549
key_right = (b'zaphod',)
2550
basis, test = self.get_basis_and_test_knit()
2551
basis.add_lines(key_left, (), [b'bar\n'])
2552
basis.add_lines(key_right, (), [b'zaphod\n'])
2554
test.add_lines(key, (key_left, key_right),
2555
[b'bar\n', b'foo\n', b'zaphod\n'])
2556
diffs = test.make_mpdiffs([key])
2558
multiparent.MultiParent([multiparent.ParentText(0, 0, 0, 1),
2559
multiparent.NewText([b'foo\n']),
2560
multiparent.ParentText(1, 0, 2, 1)])],
2562
self.assertEqual(3, len(basis.calls))
2564
("get_parent_map", {key_left, key_right}),
2565
("get_parent_map", {key_left, key_right}),
2568
last_call = basis.calls[-1]
2569
self.assertEqual('get_record_stream', last_call[0])
2570
self.assertEqual({key_left, key_right}, set(last_call[1]))
2571
self.assertEqual('topological', last_call[2])
2572
self.assertEqual(True, last_call[3])
2575
class TestNetworkBehaviour(KnitTests):
2576
"""Tests for getting data out of/into knits over the network."""
2578
def test_include_delta_closure_generates_a_knit_delta_closure(self):
2579
vf = self.make_test_knit(name='test')
2580
# put in three texts, giving ft, delta, delta
2581
vf.add_lines((b'base',), (), [b'base\n', b'content\n'])
2582
vf.add_lines((b'd1',), ((b'base',),), [b'd1\n'])
2583
vf.add_lines((b'd2',), ((b'd1',),), [b'd2\n'])
2584
# But heuristics could interfere, so check what happened:
2585
self.assertEqual(['knit-ft-gz', 'knit-delta-gz', 'knit-delta-gz'],
2586
[record.storage_kind for record in
2587
vf.get_record_stream([(b'base',), (b'd1',), (b'd2',)],
2588
'topological', False)])
2589
# generate a stream of just the deltas include_delta_closure=True,
2590
# serialise to the network, and check that we get a delta closure on the wire.
2591
stream = vf.get_record_stream([(b'd1',), (b'd2',)], 'topological', True)
2592
netb = [record.get_bytes_as(record.storage_kind) for record in stream]
2593
# The first bytes should be a memo from _ContentMapGenerator, and the
2594
# second bytes should be empty (because its a API proxy not something
2595
# for wire serialisation.
2596
self.assertEqual(b'', netb[1])
2598
kind, line_end = network_bytes_to_kind_and_offset(bytes)
2599
self.assertEqual('knit-delta-closure', kind)
2602
class TestContentMapGenerator(KnitTests):
2603
"""Tests for ContentMapGenerator"""
2605
def test_get_record_stream_gives_records(self):
2606
vf = self.make_test_knit(name='test')
2607
# put in three texts, giving ft, delta, delta
2608
vf.add_lines((b'base',), (), [b'base\n', b'content\n'])
2609
vf.add_lines((b'd1',), ((b'base',),), [b'd1\n'])
2610
vf.add_lines((b'd2',), ((b'd1',),), [b'd2\n'])
2611
keys = [(b'd1',), (b'd2',)]
2612
generator = _VFContentMapGenerator(vf, keys,
2613
global_map=vf.get_parent_map(keys))
2614
for record in generator.get_record_stream():
2615
if record.key == (b'd1',):
2616
self.assertEqual(b'd1\n', record.get_bytes_as('fulltext'))
2618
self.assertEqual(b'd2\n', record.get_bytes_as('fulltext'))
2620
def test_get_record_stream_kinds_are_raw(self):
2621
vf = self.make_test_knit(name='test')
2622
# put in three texts, giving ft, delta, delta
2623
vf.add_lines((b'base',), (), [b'base\n', b'content\n'])
2624
vf.add_lines((b'd1',), ((b'base',),), [b'd1\n'])
2625
vf.add_lines((b'd2',), ((b'd1',),), [b'd2\n'])
2626
keys = [(b'base',), (b'd1',), (b'd2',)]
2627
generator = _VFContentMapGenerator(vf, keys,
2628
global_map=vf.get_parent_map(keys))
2629
kinds = {(b'base',): 'knit-delta-closure',
2630
(b'd1',): 'knit-delta-closure-ref',
2631
(b'd2',): 'knit-delta-closure-ref',
2633
for record in generator.get_record_stream():
2634
self.assertEqual(kinds[record.key], record.storage_kind)
added
removed
breezy/tests/test_knit.py
