1
# Copyright (C) 2006-2011 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
"""Tests for Knit data structure"""
33
from ..errors import (
37
from ..bzr.index import *
38
from ..bzr.knit import (
43
_VFContentMapGenerator,
49
from ..patiencediff import PatienceSequenceMatcher
54
from ..sixish import (
59
TestCaseWithMemoryTransport,
60
TestCaseWithTransport,
63
from ..bzr.versionedfile import (
66
network_bytes_to_kind_and_offset,
67
RecordingVersionedFilesDecorator,
74
compiled_knit_feature = features.ModuleAvailableFeature(
75
'breezy.bzr._knit_load_data_pyx')
78
class KnitContentTestsMixin(object):
80
def test_constructor(self):
81
content = self._make_content([])
84
content = self._make_content([])
85
self.assertEqual(content.text(), [])
87
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
88
self.assertEqual(content.text(), ["text1", "text2"])
91
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
93
self.assertIsInstance(copy, content.__class__)
94
self.assertEqual(copy.annotate(), content.annotate())
96
def assertDerivedBlocksEqual(self, source, target, noeol=False):
97
"""Assert that the derived matching blocks match real output"""
98
source_lines = source.splitlines(True)
99
target_lines = target.splitlines(True)
101
if noeol and not line.endswith('\n'):
105
source_content = self._make_content([(None, nl(l)) for l in source_lines])
106
target_content = self._make_content([(None, nl(l)) for l in target_lines])
107
line_delta = source_content.line_delta(target_content)
108
delta_blocks = list(KnitContent.get_line_delta_blocks(line_delta,
109
source_lines, target_lines))
110
matcher = PatienceSequenceMatcher(None, source_lines, target_lines)
111
matcher_blocks = list(matcher.get_matching_blocks())
112
self.assertEqual(matcher_blocks, delta_blocks)
114
def test_get_line_delta_blocks(self):
115
self.assertDerivedBlocksEqual('a\nb\nc\n', 'q\nc\n')
116
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1)
117
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1A)
118
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1B)
119
self.assertDerivedBlocksEqual(TEXT_1B, TEXT_1A)
120
self.assertDerivedBlocksEqual(TEXT_1A, TEXT_1B)
121
self.assertDerivedBlocksEqual(TEXT_1A, '')
122
self.assertDerivedBlocksEqual('', TEXT_1A)
123
self.assertDerivedBlocksEqual('', '')
124
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd')
126
def test_get_line_delta_blocks_noeol(self):
127
"""Handle historical knit deltas safely
129
Some existing knit deltas don't consider the last line to differ
130
when the only difference whether it has a final newline.
132
New knit deltas appear to always consider the last line to differ
135
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd\n', noeol=True)
136
self.assertDerivedBlocksEqual('a\nb\nc\nd\n', 'a\nb\nc', noeol=True)
137
self.assertDerivedBlocksEqual('a\nb\nc\n', 'a\nb\nc', noeol=True)
138
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\n', noeol=True)
150
Banana cup cake recipe
160
Banana cup cake recipe
162
- bananas (do not use plantains!!!)
169
Banana cup cake recipe
186
class TestPlainKnitContent(TestCase, KnitContentTestsMixin):
188
def _make_content(self, lines):
189
annotated_content = AnnotatedKnitContent(lines)
190
return PlainKnitContent(annotated_content.text(), 'bogus')
192
def test_annotate(self):
193
content = self._make_content([])
194
self.assertEqual(content.annotate(), [])
196
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
197
self.assertEqual(content.annotate(),
198
[("bogus", "text1"), ("bogus", "text2")])
200
def test_line_delta(self):
201
content1 = self._make_content([("", "a"), ("", "b")])
202
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
203
self.assertEqual(content1.line_delta(content2),
204
[(1, 2, 2, ["a", "c"])])
206
def test_line_delta_iter(self):
207
content1 = self._make_content([("", "a"), ("", "b")])
208
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
209
it = content1.line_delta_iter(content2)
210
self.assertEqual(next(it), (1, 2, 2, ["a", "c"]))
211
self.assertRaises(StopIteration, next, it)
214
class TestAnnotatedKnitContent(TestCase, KnitContentTestsMixin):
216
def _make_content(self, lines):
217
return AnnotatedKnitContent(lines)
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
[("origin1", "text1"), ("origin2", "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 MockTransport(object):
243
def __init__(self, file_lines=None):
244
self.file_lines = file_lines
246
# We have no base directory for the MockTransport
249
def get(self, filename):
250
if self.file_lines is None:
251
raise NoSuchFile(filename)
253
return BytesIO(b"\n".join(self.file_lines))
255
def readv(self, relpath, offsets):
256
fp = self.get(relpath)
257
for offset, size in offsets:
259
yield offset, fp.read(size)
261
def __getattr__(self, name):
262
def queue_call(*args, **kwargs):
263
self.calls.append((name, args, kwargs))
267
class MockReadvFailingTransport(MockTransport):
268
"""Fail in the middle of a readv() result.
270
This Transport will successfully yield the first two requested hunks, but
271
raise NoSuchFile for the rest.
274
def readv(self, relpath, offsets):
276
for result in MockTransport.readv(self, relpath, offsets):
278
# we use 2 because the first offset is the pack header, the second
279
# is the first actual content requset
281
raise errors.NoSuchFile(relpath)
285
class KnitRecordAccessTestsMixin(object):
286
"""Tests for getting and putting knit records."""
288
def test_add_raw_records(self):
289
"""Add_raw_records adds records retrievable later."""
290
access = self.get_access()
291
memos = access.add_raw_records([('key', 10)], '1234567890')
292
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
294
def test_add_several_raw_records(self):
295
"""add_raw_records with many records and read some back."""
296
access = self.get_access()
297
memos = access.add_raw_records([('key', 10), ('key2', 2), ('key3', 5)],
299
self.assertEqual(['1234567890', '12', '34567'],
300
list(access.get_raw_records(memos)))
301
self.assertEqual(['1234567890'],
302
list(access.get_raw_records(memos[0:1])))
303
self.assertEqual(['12'],
304
list(access.get_raw_records(memos[1:2])))
305
self.assertEqual(['34567'],
306
list(access.get_raw_records(memos[2:3])))
307
self.assertEqual(['1234567890', '34567'],
308
list(access.get_raw_records(memos[0:1] + memos[2:3])))
311
class TestKnitKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
312
"""Tests for the .kndx implementation."""
314
def get_access(self):
315
"""Get a .knit style access instance."""
316
mapper = ConstantMapper("foo")
317
access = _KnitKeyAccess(self.get_transport(), mapper)
321
class _TestException(Exception):
322
"""Just an exception for local tests to use."""
325
class TestPackKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
326
"""Tests for the pack based access."""
328
def get_access(self):
329
return self._get_access()[0]
331
def _get_access(self, packname='packfile', index='FOO'):
332
transport = self.get_transport()
333
def write_data(bytes):
334
transport.append_bytes(packname, bytes)
335
writer = pack.ContainerWriter(write_data)
337
access = pack_repo._DirectPackAccess({})
338
access.set_writer(writer, index, (transport, packname))
339
return access, writer
341
def make_pack_file(self):
342
"""Create a pack file with 2 records."""
343
access, writer = self._get_access(packname='packname', index='foo')
345
memos.extend(access.add_raw_records([('key1', 10)], '1234567890'))
346
memos.extend(access.add_raw_records([('key2', 5)], '12345'))
350
def test_pack_collection_pack_retries(self):
351
"""An explicit pack of a pack collection succeeds even when a
352
concurrent pack happens.
354
builder = self.make_branch_builder('.')
355
builder.start_series()
356
builder.build_snapshot('rev-1', None, [
357
('add', ('', 'root-id', 'directory', None)),
358
('add', ('file', 'file-id', 'file', 'content\nrev 1\n')),
360
builder.build_snapshot('rev-2', ['rev-1'], [
361
('modify', ('file-id', 'content\nrev 2\n')),
363
builder.build_snapshot('rev-3', ['rev-2'], [
364
('modify', ('file-id', 'content\nrev 3\n')),
366
self.addCleanup(builder.finish_series)
367
b = builder.get_branch()
368
self.addCleanup(b.lock_write().unlock)
370
collection = repo._pack_collection
371
# Concurrently repack the repo.
372
reopened_repo = repo.controldir.open_repository()
377
def make_vf_for_retrying(self):
378
"""Create 3 packs and a reload function.
380
Originally, 2 pack files will have the data, but one will be missing.
381
And then the third will be used in place of the first two if reload()
384
:return: (versioned_file, reload_counter)
385
versioned_file a KnitVersionedFiles using the packs for access
387
builder = self.make_branch_builder('.', format="1.9")
388
builder.start_series()
389
builder.build_snapshot('rev-1', None, [
390
('add', ('', 'root-id', 'directory', None)),
391
('add', ('file', 'file-id', 'file', 'content\nrev 1\n')),
393
builder.build_snapshot('rev-2', ['rev-1'], [
394
('modify', ('file-id', 'content\nrev 2\n')),
396
builder.build_snapshot('rev-3', ['rev-2'], [
397
('modify', ('file-id', 'content\nrev 3\n')),
399
builder.finish_series()
400
b = builder.get_branch()
402
self.addCleanup(b.unlock)
403
# Pack these three revisions into another pack file, but don't remove
406
collection = repo._pack_collection
407
collection.ensure_loaded()
408
orig_packs = collection.packs
409
packer = knitpack_repo.KnitPacker(collection, orig_packs, '.testpack')
410
new_pack = packer.pack()
411
# forget about the new pack
415
# Set up a reload() function that switches to using the new pack file
416
new_index = new_pack.revision_index
417
access_tuple = new_pack.access_tuple()
418
reload_counter = [0, 0, 0]
420
reload_counter[0] += 1
421
if reload_counter[1] > 0:
422
# We already reloaded, nothing more to do
423
reload_counter[2] += 1
425
reload_counter[1] += 1
426
vf._index._graph_index._indices[:] = [new_index]
427
vf._access._indices.clear()
428
vf._access._indices[new_index] = access_tuple
430
# Delete one of the pack files so the data will need to be reloaded. We
431
# will delete the file with 'rev-2' in it
432
trans, name = orig_packs[1].access_tuple()
434
# We don't have the index trigger reloading because we want to test
435
# that we reload when the .pack disappears
436
vf._access._reload_func = reload
437
return vf, reload_counter
439
def make_reload_func(self, return_val=True):
442
reload_called[0] += 1
444
return reload_called, reload
446
def make_retry_exception(self):
447
# We raise a real exception so that sys.exc_info() is properly
450
raise _TestException('foobar')
451
except _TestException as e:
452
retry_exc = errors.RetryWithNewPacks(None, reload_occurred=False,
453
exc_info=sys.exc_info())
454
# GZ 2010-08-10: Cycle with exc_info affects 3 tests
457
def test_read_from_several_packs(self):
458
access, writer = self._get_access()
460
memos.extend(access.add_raw_records([('key', 10)], '1234567890'))
462
access, writer = self._get_access('pack2', 'FOOBAR')
463
memos.extend(access.add_raw_records([('key', 5)], '12345'))
465
access, writer = self._get_access('pack3', 'BAZ')
466
memos.extend(access.add_raw_records([('key', 5)], 'alpha'))
468
transport = self.get_transport()
469
access = pack_repo._DirectPackAccess({"FOO":(transport, 'packfile'),
470
"FOOBAR":(transport, 'pack2'),
471
"BAZ":(transport, 'pack3')})
472
self.assertEqual(['1234567890', '12345', 'alpha'],
473
list(access.get_raw_records(memos)))
474
self.assertEqual(['1234567890'],
475
list(access.get_raw_records(memos[0:1])))
476
self.assertEqual(['12345'],
477
list(access.get_raw_records(memos[1:2])))
478
self.assertEqual(['alpha'],
479
list(access.get_raw_records(memos[2:3])))
480
self.assertEqual(['1234567890', 'alpha'],
481
list(access.get_raw_records(memos[0:1] + memos[2:3])))
483
def test_set_writer(self):
484
"""The writer should be settable post construction."""
485
access = pack_repo._DirectPackAccess({})
486
transport = self.get_transport()
487
packname = 'packfile'
489
def write_data(bytes):
490
transport.append_bytes(packname, bytes)
491
writer = pack.ContainerWriter(write_data)
493
access.set_writer(writer, index, (transport, packname))
494
memos = access.add_raw_records([('key', 10)], '1234567890')
496
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
498
def test_missing_index_raises_retry(self):
499
memos = self.make_pack_file()
500
transport = self.get_transport()
501
reload_called, reload_func = self.make_reload_func()
502
# Note that the index key has changed from 'foo' to 'bar'
503
access = pack_repo._DirectPackAccess({'bar':(transport, 'packname')},
504
reload_func=reload_func)
505
e = self.assertListRaises(errors.RetryWithNewPacks,
506
access.get_raw_records, memos)
507
# Because a key was passed in which does not match our index list, we
508
# assume that the listing was already reloaded
509
self.assertTrue(e.reload_occurred)
510
self.assertIsInstance(e.exc_info, tuple)
511
self.assertIs(e.exc_info[0], KeyError)
512
self.assertIsInstance(e.exc_info[1], KeyError)
514
def test_missing_index_raises_key_error_with_no_reload(self):
515
memos = self.make_pack_file()
516
transport = self.get_transport()
517
# Note that the index key has changed from 'foo' to 'bar'
518
access = pack_repo._DirectPackAccess({'bar':(transport, 'packname')})
519
e = self.assertListRaises(KeyError, access.get_raw_records, memos)
521
def test_missing_file_raises_retry(self):
522
memos = self.make_pack_file()
523
transport = self.get_transport()
524
reload_called, reload_func = self.make_reload_func()
525
# Note that the 'filename' has been changed to 'different-packname'
526
access = pack_repo._DirectPackAccess(
527
{'foo':(transport, 'different-packname')},
528
reload_func=reload_func)
529
e = self.assertListRaises(errors.RetryWithNewPacks,
530
access.get_raw_records, memos)
531
# The file has gone missing, so we assume we need to reload
532
self.assertFalse(e.reload_occurred)
533
self.assertIsInstance(e.exc_info, tuple)
534
self.assertIs(e.exc_info[0], errors.NoSuchFile)
535
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
536
self.assertEqual('different-packname', e.exc_info[1].path)
538
def test_missing_file_raises_no_such_file_with_no_reload(self):
539
memos = self.make_pack_file()
540
transport = self.get_transport()
541
# Note that the 'filename' has been changed to 'different-packname'
542
access = pack_repo._DirectPackAccess(
543
{'foo': (transport, 'different-packname')})
544
e = self.assertListRaises(errors.NoSuchFile,
545
access.get_raw_records, memos)
547
def test_failing_readv_raises_retry(self):
548
memos = self.make_pack_file()
549
transport = self.get_transport()
550
failing_transport = MockReadvFailingTransport(
551
[transport.get_bytes('packname')])
552
reload_called, reload_func = self.make_reload_func()
553
access = pack_repo._DirectPackAccess(
554
{'foo': (failing_transport, 'packname')},
555
reload_func=reload_func)
556
# Asking for a single record will not trigger the Mock failure
557
self.assertEqual(['1234567890'],
558
list(access.get_raw_records(memos[:1])))
559
self.assertEqual(['12345'],
560
list(access.get_raw_records(memos[1:2])))
561
# A multiple offset readv() will fail mid-way through
562
e = self.assertListRaises(errors.RetryWithNewPacks,
563
access.get_raw_records, memos)
564
# The file has gone missing, so we assume we need to reload
565
self.assertFalse(e.reload_occurred)
566
self.assertIsInstance(e.exc_info, tuple)
567
self.assertIs(e.exc_info[0], errors.NoSuchFile)
568
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
569
self.assertEqual('packname', e.exc_info[1].path)
571
def test_failing_readv_raises_no_such_file_with_no_reload(self):
572
memos = self.make_pack_file()
573
transport = self.get_transport()
574
failing_transport = MockReadvFailingTransport(
575
[transport.get_bytes('packname')])
576
reload_called, reload_func = self.make_reload_func()
577
access = pack_repo._DirectPackAccess(
578
{'foo':(failing_transport, 'packname')})
579
# Asking for a single record will not trigger the Mock failure
580
self.assertEqual(['1234567890'],
581
list(access.get_raw_records(memos[:1])))
582
self.assertEqual(['12345'],
583
list(access.get_raw_records(memos[1:2])))
584
# A multiple offset readv() will fail mid-way through
585
e = self.assertListRaises(errors.NoSuchFile,
586
access.get_raw_records, memos)
588
def test_reload_or_raise_no_reload(self):
589
access = pack_repo._DirectPackAccess({}, reload_func=None)
590
retry_exc = self.make_retry_exception()
591
# Without a reload_func, we will just re-raise the original exception
592
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
594
def test_reload_or_raise_reload_changed(self):
595
reload_called, reload_func = self.make_reload_func(return_val=True)
596
access = pack_repo._DirectPackAccess({}, reload_func=reload_func)
597
retry_exc = self.make_retry_exception()
598
access.reload_or_raise(retry_exc)
599
self.assertEqual([1], reload_called)
600
retry_exc.reload_occurred=True
601
access.reload_or_raise(retry_exc)
602
self.assertEqual([2], reload_called)
604
def test_reload_or_raise_reload_no_change(self):
605
reload_called, reload_func = self.make_reload_func(return_val=False)
606
access = pack_repo._DirectPackAccess({}, reload_func=reload_func)
607
retry_exc = self.make_retry_exception()
608
# If reload_occurred is False, then we consider it an error to have
609
# reload_func() return False (no changes).
610
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
611
self.assertEqual([1], reload_called)
612
retry_exc.reload_occurred=True
613
# If reload_occurred is True, then we assume nothing changed because
614
# it had changed earlier, but didn't change again
615
access.reload_or_raise(retry_exc)
616
self.assertEqual([2], reload_called)
618
def test_annotate_retries(self):
619
vf, reload_counter = self.make_vf_for_retrying()
620
# It is a little bit bogus to annotate the Revision VF, but it works,
621
# as we have ancestry stored there
623
reload_lines = vf.annotate(key)
624
self.assertEqual([1, 1, 0], reload_counter)
625
plain_lines = vf.annotate(key)
626
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
627
if reload_lines != plain_lines:
628
self.fail('Annotation was not identical with reloading.')
629
# Now delete the packs-in-use, which should trigger another reload, but
630
# this time we just raise an exception because we can't recover
631
for trans, name in vf._access._indices.values():
633
self.assertRaises(errors.NoSuchFile, vf.annotate, key)
634
self.assertEqual([2, 1, 1], reload_counter)
636
def test__get_record_map_retries(self):
637
vf, reload_counter = self.make_vf_for_retrying()
638
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
639
records = vf._get_record_map(keys)
640
self.assertEqual(keys, sorted(records.keys()))
641
self.assertEqual([1, 1, 0], reload_counter)
642
# Now delete the packs-in-use, which should trigger another reload, but
643
# this time we just raise an exception because we can't recover
644
for trans, name in vf._access._indices.values():
646
self.assertRaises(errors.NoSuchFile, vf._get_record_map, keys)
647
self.assertEqual([2, 1, 1], reload_counter)
649
def test_get_record_stream_retries(self):
650
vf, reload_counter = self.make_vf_for_retrying()
651
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
652
record_stream = vf.get_record_stream(keys, 'topological', False)
653
record = next(record_stream)
654
self.assertEqual(('rev-1',), record.key)
655
self.assertEqual([0, 0, 0], reload_counter)
656
record = next(record_stream)
657
self.assertEqual(('rev-2',), record.key)
658
self.assertEqual([1, 1, 0], reload_counter)
659
record = next(record_stream)
660
self.assertEqual(('rev-3',), record.key)
661
self.assertEqual([1, 1, 0], reload_counter)
662
# Now delete all pack files, and see that we raise the right error
663
for trans, name in vf._access._indices.values():
665
self.assertListRaises(errors.NoSuchFile,
666
vf.get_record_stream, keys, 'topological', False)
668
def test_iter_lines_added_or_present_in_keys_retries(self):
669
vf, reload_counter = self.make_vf_for_retrying()
670
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
671
# Unfortunately, iter_lines_added_or_present_in_keys iterates the
672
# result in random order (determined by the iteration order from a
673
# set()), so we don't have any solid way to trigger whether data is
674
# read before or after. However we tried to delete the middle node to
675
# exercise the code well.
676
# What we care about is that all lines are always yielded, but not
679
reload_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
680
self.assertEqual([1, 1, 0], reload_counter)
681
# Now do it again, to make sure the result is equivalent
682
plain_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
683
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
684
self.assertEqual(plain_lines, reload_lines)
685
self.assertEqual(21, len(plain_lines))
686
# Now delete all pack files, and see that we raise the right error
687
for trans, name in vf._access._indices.values():
689
self.assertListRaises(errors.NoSuchFile,
690
vf.iter_lines_added_or_present_in_keys, keys)
691
self.assertEqual([2, 1, 1], reload_counter)
693
def test_get_record_stream_yields_disk_sorted_order(self):
694
# if we get 'unordered' pick a semi-optimal order for reading. The
695
# order should be grouped by pack file, and then by position in file
696
repo = self.make_repository('test', format='pack-0.92')
698
self.addCleanup(repo.unlock)
699
repo.start_write_group()
701
vf.add_lines(('f-id', 'rev-5'), [('f-id', 'rev-4')], ['lines\n'])
702
vf.add_lines(('f-id', 'rev-1'), [], ['lines\n'])
703
vf.add_lines(('f-id', 'rev-2'), [('f-id', 'rev-1')], ['lines\n'])
704
repo.commit_write_group()
705
# We inserted them as rev-5, rev-1, rev-2, we should get them back in
707
stream = vf.get_record_stream([('f-id', 'rev-1'), ('f-id', 'rev-5'),
708
('f-id', 'rev-2')], 'unordered', False)
709
keys = [r.key for r in stream]
710
self.assertEqual([('f-id', 'rev-5'), ('f-id', 'rev-1'),
711
('f-id', 'rev-2')], keys)
712
repo.start_write_group()
713
vf.add_lines(('f-id', 'rev-4'), [('f-id', 'rev-3')], ['lines\n'])
714
vf.add_lines(('f-id', 'rev-3'), [('f-id', 'rev-2')], ['lines\n'])
715
vf.add_lines(('f-id', 'rev-6'), [('f-id', 'rev-5')], ['lines\n'])
716
repo.commit_write_group()
717
# Request in random order, to make sure the output order isn't based on
719
request_keys = set(('f-id', 'rev-%d' % i) for i in range(1, 7))
720
stream = vf.get_record_stream(request_keys, 'unordered', False)
721
keys = [r.key for r in stream]
722
# We want to get the keys back in disk order, but it doesn't matter
723
# which pack we read from first. So this can come back in 2 orders
724
alt1 = [('f-id', 'rev-%d' % i) for i in [4, 3, 6, 5, 1, 2]]
725
alt2 = [('f-id', 'rev-%d' % i) for i in [5, 1, 2, 4, 3, 6]]
726
if keys != alt1 and keys != alt2:
727
self.fail('Returned key order did not match either expected order.'
728
' expected %s or %s, not %s'
729
% (alt1, alt2, keys))
732
class LowLevelKnitDataTests(TestCase):
734
def create_gz_content(self, text):
736
gz_file = gzip.GzipFile(mode='wb', fileobj=sio)
739
return sio.getvalue()
741
def make_multiple_records(self):
742
"""Create the content for multiple records."""
743
sha1sum = osutils.sha_string('foo\nbar\n')
745
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
750
record_1 = (0, len(gz_txt), sha1sum)
751
total_txt.append(gz_txt)
752
sha1sum = osutils.sha_string('baz\n')
753
gz_txt = self.create_gz_content('version rev-id-2 1 %s\n'
757
record_2 = (record_1[1], len(gz_txt), sha1sum)
758
total_txt.append(gz_txt)
759
return total_txt, record_1, record_2
761
def test_valid_knit_data(self):
762
sha1sum = osutils.sha_string('foo\nbar\n')
763
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
768
transport = MockTransport([gz_txt])
769
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
770
knit = KnitVersionedFiles(None, access)
771
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
773
contents = list(knit._read_records_iter(records))
774
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'],
775
'4e48e2c9a3d2ca8a708cb0cc545700544efb5021')], contents)
777
raw_contents = list(knit._read_records_iter_raw(records))
778
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
780
def test_multiple_records_valid(self):
781
total_txt, record_1, record_2 = self.make_multiple_records()
782
transport = MockTransport([''.join(total_txt)])
783
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
784
knit = KnitVersionedFiles(None, access)
785
records = [(('rev-id-1',), (('rev-id-1',), record_1[0], record_1[1])),
786
(('rev-id-2',), (('rev-id-2',), record_2[0], record_2[1]))]
788
contents = list(knit._read_records_iter(records))
789
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'], record_1[2]),
790
(('rev-id-2',), ['baz\n'], record_2[2])],
793
raw_contents = list(knit._read_records_iter_raw(records))
794
self.assertEqual([(('rev-id-1',), total_txt[0], record_1[2]),
795
(('rev-id-2',), total_txt[1], record_2[2])],
798
def test_not_enough_lines(self):
799
sha1sum = osutils.sha_string('foo\n')
800
# record says 2 lines data says 1
801
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
805
transport = MockTransport([gz_txt])
806
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
807
knit = KnitVersionedFiles(None, access)
808
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
809
self.assertRaises(errors.KnitCorrupt, list,
810
knit._read_records_iter(records))
812
# read_records_iter_raw won't detect that sort of mismatch/corruption
813
raw_contents = list(knit._read_records_iter_raw(records))
814
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
816
def test_too_many_lines(self):
817
sha1sum = osutils.sha_string('foo\nbar\n')
818
# record says 1 lines data says 2
819
gz_txt = self.create_gz_content('version rev-id-1 1 %s\n'
824
transport = MockTransport([gz_txt])
825
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
826
knit = KnitVersionedFiles(None, access)
827
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
828
self.assertRaises(errors.KnitCorrupt, list,
829
knit._read_records_iter(records))
831
# read_records_iter_raw won't detect that sort of mismatch/corruption
832
raw_contents = list(knit._read_records_iter_raw(records))
833
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
835
def test_mismatched_version_id(self):
836
sha1sum = osutils.sha_string('foo\nbar\n')
837
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
842
transport = MockTransport([gz_txt])
843
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
844
knit = KnitVersionedFiles(None, access)
845
# We are asking for rev-id-2, but the data is rev-id-1
846
records = [(('rev-id-2',), (('rev-id-2',), 0, len(gz_txt)))]
847
self.assertRaises(errors.KnitCorrupt, list,
848
knit._read_records_iter(records))
850
# read_records_iter_raw detects mismatches in the header
851
self.assertRaises(errors.KnitCorrupt, list,
852
knit._read_records_iter_raw(records))
854
def test_uncompressed_data(self):
855
sha1sum = osutils.sha_string('foo\nbar\n')
856
txt = ('version rev-id-1 2 %s\n'
861
transport = MockTransport([txt])
862
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
863
knit = KnitVersionedFiles(None, access)
864
records = [(('rev-id-1',), (('rev-id-1',), 0, len(txt)))]
866
# We don't have valid gzip data ==> corrupt
867
self.assertRaises(errors.KnitCorrupt, list,
868
knit._read_records_iter(records))
870
# read_records_iter_raw will notice the bad data
871
self.assertRaises(errors.KnitCorrupt, list,
872
knit._read_records_iter_raw(records))
874
def test_corrupted_data(self):
875
sha1sum = osutils.sha_string('foo\nbar\n')
876
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
881
# Change 2 bytes in the middle to \xff
882
gz_txt = gz_txt[:10] + '\xff\xff' + gz_txt[12:]
883
transport = MockTransport([gz_txt])
884
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
885
knit = KnitVersionedFiles(None, access)
886
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
887
self.assertRaises(errors.KnitCorrupt, list,
888
knit._read_records_iter(records))
889
# read_records_iter_raw will barf on bad gz data
890
self.assertRaises(errors.KnitCorrupt, list,
891
knit._read_records_iter_raw(records))
894
class LowLevelKnitIndexTests(TestCase):
896
def get_knit_index(self, transport, name, mode):
897
mapper = ConstantMapper(name)
898
from ..bzr._knit_load_data_py import _load_data_py
899
self.overrideAttr(knit, '_load_data', _load_data_py)
900
allow_writes = lambda: 'w' in mode
901
return _KndxIndex(transport, mapper, lambda:None, allow_writes, lambda:True)
903
def test_create_file(self):
904
transport = MockTransport()
905
index = self.get_knit_index(transport, "filename", "w")
907
call = transport.calls.pop(0)
908
# call[1][1] is a BytesIO - we can't test it by simple equality.
909
self.assertEqual('put_file_non_atomic', call[0])
910
self.assertEqual('filename.kndx', call[1][0])
911
# With no history, _KndxIndex writes a new index:
912
self.assertEqual(_KndxIndex.HEADER,
913
call[1][1].getvalue())
914
self.assertEqual({'create_parent_dir': True}, call[2])
916
def test_read_utf8_version_id(self):
917
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
918
utf8_revision_id = unicode_revision_id.encode('utf-8')
919
transport = MockTransport([
921
'%s option 0 1 :' % (utf8_revision_id,)
923
index = self.get_knit_index(transport, "filename", "r")
924
# _KndxIndex is a private class, and deals in utf8 revision_ids, not
925
# Unicode revision_ids.
926
self.assertEqual({(utf8_revision_id,):()},
927
index.get_parent_map(index.keys()))
928
self.assertFalse((unicode_revision_id,) in index.keys())
930
def test_read_utf8_parents(self):
931
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
932
utf8_revision_id = unicode_revision_id.encode('utf-8')
933
transport = MockTransport([
935
"version option 0 1 .%s :" % (utf8_revision_id,)
937
index = self.get_knit_index(transport, "filename", "r")
938
self.assertEqual({("version",):((utf8_revision_id,),)},
939
index.get_parent_map(index.keys()))
941
def test_read_ignore_corrupted_lines(self):
942
transport = MockTransport([
945
"corrupted options 0 1 .b .c ",
946
"version options 0 1 :"
948
index = self.get_knit_index(transport, "filename", "r")
949
self.assertEqual(1, len(index.keys()))
950
self.assertEqual({("version",)}, index.keys())
952
def test_read_corrupted_header(self):
953
transport = MockTransport(['not a bzr knit index header\n'])
954
index = self.get_knit_index(transport, "filename", "r")
955
self.assertRaises(KnitHeaderError, index.keys)
957
def test_read_duplicate_entries(self):
958
transport = MockTransport([
960
"parent options 0 1 :",
961
"version options1 0 1 0 :",
962
"version options2 1 2 .other :",
963
"version options3 3 4 0 .other :"
965
index = self.get_knit_index(transport, "filename", "r")
966
self.assertEqual(2, len(index.keys()))
967
# check that the index used is the first one written. (Specific
968
# to KnitIndex style indices.
969
self.assertEqual("1", index._dictionary_compress([("version",)]))
970
self.assertEqual((("version",), 3, 4), index.get_position(("version",)))
971
self.assertEqual(["options3"], index.get_options(("version",)))
972
self.assertEqual({("version",):(("parent",), ("other",))},
973
index.get_parent_map([("version",)]))
975
def test_read_compressed_parents(self):
976
transport = MockTransport([
980
"c option 0 1 1 0 :",
982
index = self.get_knit_index(transport, "filename", "r")
983
self.assertEqual({("b",):(("a",),), ("c",):(("b",), ("a",))},
984
index.get_parent_map([("b",), ("c",)]))
986
def test_write_utf8_version_id(self):
987
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
988
utf8_revision_id = unicode_revision_id.encode('utf-8')
989
transport = MockTransport([
992
index = self.get_knit_index(transport, "filename", "r")
994
((utf8_revision_id,), ["option"], ((utf8_revision_id,), 0, 1), [])])
995
call = transport.calls.pop(0)
996
# call[1][1] is a BytesIO - we can't test it by simple equality.
997
self.assertEqual('put_file_non_atomic', call[0])
998
self.assertEqual('filename.kndx', call[1][0])
999
# With no history, _KndxIndex writes a new index:
1000
self.assertEqual(_KndxIndex.HEADER +
1001
"\n%s option 0 1 :" % (utf8_revision_id,),
1002
call[1][1].getvalue())
1003
self.assertEqual({'create_parent_dir': True}, call[2])
1005
def test_write_utf8_parents(self):
1006
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
1007
utf8_revision_id = unicode_revision_id.encode('utf-8')
1008
transport = MockTransport([
1011
index = self.get_knit_index(transport, "filename", "r")
1013
(("version",), ["option"], (("version",), 0, 1), [(utf8_revision_id,)])])
1014
call = transport.calls.pop(0)
1015
# call[1][1] is a BytesIO - we can't test it by simple equality.
1016
self.assertEqual('put_file_non_atomic', call[0])
1017
self.assertEqual('filename.kndx', call[1][0])
1018
# With no history, _KndxIndex writes a new index:
1019
self.assertEqual(_KndxIndex.HEADER +
1020
"\nversion option 0 1 .%s :" % (utf8_revision_id,),
1021
call[1][1].getvalue())
1022
self.assertEqual({'create_parent_dir': True}, call[2])
1024
def test_keys(self):
1025
transport = MockTransport([
1028
index = self.get_knit_index(transport, "filename", "r")
1030
self.assertEqual(set(), index.keys())
1032
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1033
self.assertEqual({("a",)}, index.keys())
1035
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1036
self.assertEqual({("a",)}, index.keys())
1038
index.add_records([(("b",), ["option"], (("b",), 0, 1), [])])
1039
self.assertEqual({("a",), ("b",)}, index.keys())
1041
def add_a_b(self, index, random_id=None):
1043
if random_id is not None:
1044
kwargs["random_id"] = random_id
1046
(("a",), ["option"], (("a",), 0, 1), [("b",)]),
1047
(("a",), ["opt"], (("a",), 1, 2), [("c",)]),
1048
(("b",), ["option"], (("b",), 2, 3), [("a",)])
1051
def assertIndexIsAB(self, index):
1056
index.get_parent_map(index.keys()))
1057
self.assertEqual((("a",), 1, 2), index.get_position(("a",)))
1058
self.assertEqual((("b",), 2, 3), index.get_position(("b",)))
1059
self.assertEqual(["opt"], index.get_options(("a",)))
1061
def test_add_versions(self):
1062
transport = MockTransport([
1065
index = self.get_knit_index(transport, "filename", "r")
1068
call = transport.calls.pop(0)
1069
# call[1][1] is a BytesIO - we can't test it by simple equality.
1070
self.assertEqual('put_file_non_atomic', call[0])
1071
self.assertEqual('filename.kndx', call[1][0])
1072
# With no history, _KndxIndex writes a new index:
1075
"\na option 0 1 .b :"
1077
"\nb option 2 3 0 :",
1078
call[1][1].getvalue())
1079
self.assertEqual({'create_parent_dir': True}, call[2])
1080
self.assertIndexIsAB(index)
1082
def test_add_versions_random_id_is_accepted(self):
1083
transport = MockTransport([
1086
index = self.get_knit_index(transport, "filename", "r")
1087
self.add_a_b(index, random_id=True)
1089
def test_delay_create_and_add_versions(self):
1090
transport = MockTransport()
1092
index = self.get_knit_index(transport, "filename", "w")
1094
self.assertEqual([], transport.calls)
1097
#[ {"dir_mode": 0777, "create_parent_dir": True, "mode": "wb"},
1099
# Two calls: one during which we load the existing index (and when its
1100
# missing create it), then a second where we write the contents out.
1101
self.assertEqual(2, len(transport.calls))
1102
call = transport.calls.pop(0)
1103
self.assertEqual('put_file_non_atomic', call[0])
1104
self.assertEqual('filename.kndx', call[1][0])
1105
# With no history, _KndxIndex writes a new index:
1106
self.assertEqual(_KndxIndex.HEADER, call[1][1].getvalue())
1107
self.assertEqual({'create_parent_dir': True}, call[2])
1108
call = transport.calls.pop(0)
1109
# call[1][1] is a BytesIO - we can't test it by simple equality.
1110
self.assertEqual('put_file_non_atomic', call[0])
1111
self.assertEqual('filename.kndx', call[1][0])
1112
# With no history, _KndxIndex writes a new index:
1115
"\na option 0 1 .b :"
1117
"\nb option 2 3 0 :",
1118
call[1][1].getvalue())
1119
self.assertEqual({'create_parent_dir': True}, call[2])
1121
def assertTotalBuildSize(self, size, keys, positions):
1122
self.assertEqual(size,
1123
knit._get_total_build_size(None, keys, positions))
1125
def test__get_total_build_size(self):
1127
('a',): (('fulltext', False), (('a',), 0, 100), None),
1128
('b',): (('line-delta', False), (('b',), 100, 21), ('a',)),
1129
('c',): (('line-delta', False), (('c',), 121, 35), ('b',)),
1130
('d',): (('line-delta', False), (('d',), 156, 12), ('b',)),
1132
self.assertTotalBuildSize(100, [('a',)], positions)
1133
self.assertTotalBuildSize(121, [('b',)], positions)
1134
# c needs both a & b
1135
self.assertTotalBuildSize(156, [('c',)], positions)
1136
# we shouldn't count 'b' twice
1137
self.assertTotalBuildSize(156, [('b',), ('c',)], positions)
1138
self.assertTotalBuildSize(133, [('d',)], positions)
1139
self.assertTotalBuildSize(168, [('c',), ('d',)], positions)
1141
def test_get_position(self):
1142
transport = MockTransport([
1147
index = self.get_knit_index(transport, "filename", "r")
1149
self.assertEqual((("a",), 0, 1), index.get_position(("a",)))
1150
self.assertEqual((("b",), 1, 2), index.get_position(("b",)))
1152
def test_get_method(self):
1153
transport = MockTransport([
1155
"a fulltext,unknown 0 1 :",
1156
"b unknown,line-delta 1 2 :",
1159
index = self.get_knit_index(transport, "filename", "r")
1161
self.assertEqual("fulltext", index.get_method("a"))
1162
self.assertEqual("line-delta", index.get_method("b"))
1163
self.assertRaises(errors.KnitIndexUnknownMethod, index.get_method, "c")
1165
def test_get_options(self):
1166
transport = MockTransport([
1171
index = self.get_knit_index(transport, "filename", "r")
1173
self.assertEqual(["opt1"], index.get_options("a"))
1174
self.assertEqual(["opt2", "opt3"], index.get_options("b"))
1176
def test_get_parent_map(self):
1177
transport = MockTransport([
1180
"b option 1 2 0 .c :",
1181
"c option 1 2 1 0 .e :"
1183
index = self.get_knit_index(transport, "filename", "r")
1187
("b",):(("a",), ("c",)),
1188
("c",):(("b",), ("a",), ("e",)),
1189
}, index.get_parent_map(index.keys()))
1191
def test_impossible_parent(self):
1192
"""Test we get KnitCorrupt if the parent couldn't possibly exist."""
1193
transport = MockTransport([
1196
"b option 0 1 4 :" # We don't have a 4th record
1198
index = self.get_knit_index(transport, 'filename', 'r')
1200
self.assertRaises(errors.KnitCorrupt, index.keys)
1201
except TypeError as e:
1202
if (str(e) == ('exceptions must be strings, classes, or instances,'
1203
' not exceptions.IndexError')):
1204
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1205
' raising new style exceptions with python'
1210
def test_corrupted_parent(self):
1211
transport = MockTransport([
1215
"c option 0 1 1v :", # Can't have a parent of '1v'
1217
index = self.get_knit_index(transport, 'filename', 'r')
1219
self.assertRaises(errors.KnitCorrupt, index.keys)
1220
except TypeError as e:
1221
if (str(e) == ('exceptions must be strings, classes, or instances,'
1222
' not exceptions.ValueError')):
1223
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1224
' raising new style exceptions with python'
1229
def test_corrupted_parent_in_list(self):
1230
transport = MockTransport([
1234
"c option 0 1 1 v :", # Can't have a parent of 'v'
1236
index = self.get_knit_index(transport, 'filename', 'r')
1238
self.assertRaises(errors.KnitCorrupt, index.keys)
1239
except TypeError as e:
1240
if (str(e) == ('exceptions must be strings, classes, or instances,'
1241
' not exceptions.ValueError')):
1242
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1243
' raising new style exceptions with python'
1248
def test_invalid_position(self):
1249
transport = MockTransport([
1253
index = self.get_knit_index(transport, 'filename', 'r')
1255
self.assertRaises(errors.KnitCorrupt, index.keys)
1256
except TypeError as e:
1257
if (str(e) == ('exceptions must be strings, classes, or instances,'
1258
' not exceptions.ValueError')):
1259
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1260
' raising new style exceptions with python'
1265
def test_invalid_size(self):
1266
transport = MockTransport([
1270
index = self.get_knit_index(transport, 'filename', 'r')
1272
self.assertRaises(errors.KnitCorrupt, index.keys)
1273
except TypeError as e:
1274
if (str(e) == ('exceptions must be strings, classes, or instances,'
1275
' not exceptions.ValueError')):
1276
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1277
' raising new style exceptions with python'
1282
def test_scan_unvalidated_index_not_implemented(self):
1283
transport = MockTransport()
1284
index = self.get_knit_index(transport, 'filename', 'r')
1286
NotImplementedError, index.scan_unvalidated_index,
1287
'dummy graph_index')
1289
NotImplementedError, index.get_missing_compression_parents)
1291
def test_short_line(self):
1292
transport = MockTransport([
1295
"b option 10 10 0", # This line isn't terminated, ignored
1297
index = self.get_knit_index(transport, "filename", "r")
1298
self.assertEqual({('a',)}, index.keys())
1300
def test_skip_incomplete_record(self):
1301
# A line with bogus data should just be skipped
1302
transport = MockTransport([
1305
"b option 10 10 0", # This line isn't terminated, ignored
1306
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1308
index = self.get_knit_index(transport, "filename", "r")
1309
self.assertEqual({('a',), ('c',)}, index.keys())
1311
def test_trailing_characters(self):
1312
# A line with bogus data should just be skipped
1313
transport = MockTransport([
1316
"b option 10 10 0 :a", # This line has extra trailing characters
1317
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1319
index = self.get_knit_index(transport, "filename", "r")
1320
self.assertEqual({('a',), ('c',)}, index.keys())
1323
class LowLevelKnitIndexTests_c(LowLevelKnitIndexTests):
1325
_test_needs_features = [compiled_knit_feature]
1327
def get_knit_index(self, transport, name, mode):
1328
mapper = ConstantMapper(name)
1329
from ..bzr._knit_load_data_pyx import _load_data_c
1330
self.overrideAttr(knit, '_load_data', _load_data_c)
1331
allow_writes = lambda: mode == 'w'
1332
return _KndxIndex(transport, mapper, lambda:None,
1333
allow_writes, lambda:True)
1336
class Test_KnitAnnotator(TestCaseWithMemoryTransport):
1338
def make_annotator(self):
1339
factory = knit.make_pack_factory(True, True, 1)
1340
vf = factory(self.get_transport())
1341
return knit._KnitAnnotator(vf)
1343
def test__expand_fulltext(self):
1344
ann = self.make_annotator()
1345
rev_key = ('rev-id',)
1346
ann._num_compression_children[rev_key] = 1
1347
res = ann._expand_record(rev_key, (('parent-id',),), None,
1348
['line1\n', 'line2\n'], ('fulltext', True))
1349
# The content object and text lines should be cached appropriately
1350
self.assertEqual(['line1\n', 'line2'], res)
1351
content_obj = ann._content_objects[rev_key]
1352
self.assertEqual(['line1\n', 'line2\n'], content_obj._lines)
1353
self.assertEqual(res, content_obj.text())
1354
self.assertEqual(res, ann._text_cache[rev_key])
1356
def test__expand_delta_comp_parent_not_available(self):
1357
# Parent isn't available yet, so we return nothing, but queue up this
1358
# node for later processing
1359
ann = self.make_annotator()
1360
rev_key = ('rev-id',)
1361
parent_key = ('parent-id',)
1362
record = ['0,1,1\n', 'new-line\n']
1363
details = ('line-delta', False)
1364
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1366
self.assertEqual(None, res)
1367
self.assertTrue(parent_key in ann._pending_deltas)
1368
pending = ann._pending_deltas[parent_key]
1369
self.assertEqual(1, len(pending))
1370
self.assertEqual((rev_key, (parent_key,), record, details), pending[0])
1372
def test__expand_record_tracks_num_children(self):
1373
ann = self.make_annotator()
1374
rev_key = ('rev-id',)
1375
rev2_key = ('rev2-id',)
1376
parent_key = ('parent-id',)
1377
record = ['0,1,1\n', 'new-line\n']
1378
details = ('line-delta', False)
1379
ann._num_compression_children[parent_key] = 2
1380
ann._expand_record(parent_key, (), None, ['line1\n', 'line2\n'],
1381
('fulltext', False))
1382
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1384
self.assertEqual({parent_key: 1}, ann._num_compression_children)
1385
# Expanding the second child should remove the content object, and the
1386
# num_compression_children entry
1387
res = ann._expand_record(rev2_key, (parent_key,), parent_key,
1389
self.assertFalse(parent_key in ann._content_objects)
1390
self.assertEqual({}, ann._num_compression_children)
1391
# We should not cache the content_objects for rev2 and rev, because
1392
# they do not have compression children of their own.
1393
self.assertEqual({}, ann._content_objects)
1395
def test__expand_delta_records_blocks(self):
1396
ann = self.make_annotator()
1397
rev_key = ('rev-id',)
1398
parent_key = ('parent-id',)
1399
record = ['0,1,1\n', 'new-line\n']
1400
details = ('line-delta', True)
1401
ann._num_compression_children[parent_key] = 2
1402
ann._expand_record(parent_key, (), None,
1403
['line1\n', 'line2\n', 'line3\n'],
1404
('fulltext', False))
1405
ann._expand_record(rev_key, (parent_key,), parent_key, record, details)
1406
self.assertEqual({(rev_key, parent_key): [(1, 1, 1), (3, 3, 0)]},
1407
ann._matching_blocks)
1408
rev2_key = ('rev2-id',)
1409
record = ['0,1,1\n', 'new-line\n']
1410
details = ('line-delta', False)
1411
ann._expand_record(rev2_key, (parent_key,), parent_key, record, details)
1412
self.assertEqual([(1, 1, 2), (3, 3, 0)],
1413
ann._matching_blocks[(rev2_key, parent_key)])
1415
def test__get_parent_ann_uses_matching_blocks(self):
1416
ann = self.make_annotator()
1417
rev_key = ('rev-id',)
1418
parent_key = ('parent-id',)
1419
parent_ann = [(parent_key,)]*3
1420
block_key = (rev_key, parent_key)
1421
ann._annotations_cache[parent_key] = parent_ann
1422
ann._matching_blocks[block_key] = [(0, 1, 1), (3, 3, 0)]
1423
# We should not try to access any parent_lines content, because we know
1424
# we already have the matching blocks
1425
par_ann, blocks = ann._get_parent_annotations_and_matches(rev_key,
1426
['1\n', '2\n', '3\n'], parent_key)
1427
self.assertEqual(parent_ann, par_ann)
1428
self.assertEqual([(0, 1, 1), (3, 3, 0)], blocks)
1429
self.assertEqual({}, ann._matching_blocks)
1431
def test__process_pending(self):
1432
ann = self.make_annotator()
1433
rev_key = ('rev-id',)
1436
record = ['0,1,1\n', 'new-line\n']
1437
details = ('line-delta', False)
1438
p1_record = ['line1\n', 'line2\n']
1439
ann._num_compression_children[p1_key] = 1
1440
res = ann._expand_record(rev_key, (p1_key,p2_key), p1_key,
1442
self.assertEqual(None, res)
1443
# self.assertTrue(p1_key in ann._pending_deltas)
1444
self.assertEqual({}, ann._pending_annotation)
1445
# Now insert p1, and we should be able to expand the delta
1446
res = ann._expand_record(p1_key, (), None, p1_record,
1447
('fulltext', False))
1448
self.assertEqual(p1_record, res)
1449
ann._annotations_cache[p1_key] = [(p1_key,)]*2
1450
res = ann._process_pending(p1_key)
1451
self.assertEqual([], res)
1452
self.assertFalse(p1_key in ann._pending_deltas)
1453
self.assertTrue(p2_key in ann._pending_annotation)
1454
self.assertEqual({p2_key: [(rev_key, (p1_key, p2_key))]},
1455
ann._pending_annotation)
1456
# Now fill in parent 2, and pending annotation should be satisfied
1457
res = ann._expand_record(p2_key, (), None, [], ('fulltext', False))
1458
ann._annotations_cache[p2_key] = []
1459
res = ann._process_pending(p2_key)
1460
self.assertEqual([rev_key], res)
1461
self.assertEqual({}, ann._pending_annotation)
1462
self.assertEqual({}, ann._pending_deltas)
1464
def test_record_delta_removes_basis(self):
1465
ann = self.make_annotator()
1466
ann._expand_record(('parent-id',), (), None,
1467
['line1\n', 'line2\n'], ('fulltext', False))
1468
ann._num_compression_children['parent-id'] = 2
1470
def test_annotate_special_text(self):
1471
ann = self.make_annotator()
1473
rev1_key = ('rev-1',)
1474
rev2_key = ('rev-2',)
1475
rev3_key = ('rev-3',)
1476
spec_key = ('special:',)
1477
vf.add_lines(rev1_key, [], ['initial content\n'])
1478
vf.add_lines(rev2_key, [rev1_key], ['initial content\n',
1481
vf.add_lines(rev3_key, [rev1_key], ['initial content\n',
1484
spec_text = ('initial content\n'
1488
ann.add_special_text(spec_key, [rev2_key, rev3_key], spec_text)
1489
anns, lines = ann.annotate(spec_key)
1490
self.assertEqual([(rev1_key,),
1491
(rev2_key, rev3_key),
1495
self.assertEqualDiff(spec_text, ''.join(lines))
1498
class KnitTests(TestCaseWithTransport):
1499
"""Class containing knit test helper routines."""
1501
def make_test_knit(self, annotate=False, name='test'):
1502
mapper = ConstantMapper(name)
1503
return make_file_factory(annotate, mapper)(self.get_transport())
1506
class TestBadShaError(KnitTests):
1507
"""Tests for handling of sha errors."""
1509
def test_sha_exception_has_text(self):
1510
# having the failed text included in the error allows for recovery.
1511
source = self.make_test_knit()
1512
target = self.make_test_knit(name="target")
1513
if not source._max_delta_chain:
1514
raise TestNotApplicable(
1515
"cannot get delta-caused sha failures without deltas.")
1518
broken = ('broken',)
1519
source.add_lines(basis, (), ['foo\n'])
1520
source.add_lines(broken, (basis,), ['foo\n', 'bar\n'])
1521
# Seed target with a bad basis text
1522
target.add_lines(basis, (), ['gam\n'])
1523
target.insert_record_stream(
1524
source.get_record_stream([broken], 'unordered', False))
1525
err = self.assertRaises(errors.KnitCorrupt,
1526
target.get_record_stream([broken], 'unordered', True
1527
).next().get_bytes_as, 'chunked')
1528
self.assertEqual(['gam\n', 'bar\n'], err.content)
1529
# Test for formatting with live data
1530
self.assertStartsWith(str(err), "Knit ")
1533
class TestKnitIndex(KnitTests):
1535
def test_add_versions_dictionary_compresses(self):
1536
"""Adding versions to the index should update the lookup dict"""
1537
knit = self.make_test_knit()
1539
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1540
self.check_file_contents('test.kndx',
1541
'# bzr knit index 8\n'
1543
'a-1 fulltext 0 0 :'
1546
(('a-2',), ['fulltext'], (('a-2',), 0, 0), [('a-1',)]),
1547
(('a-3',), ['fulltext'], (('a-3',), 0, 0), [('a-2',)]),
1549
self.check_file_contents('test.kndx',
1550
'# bzr knit index 8\n'
1552
'a-1 fulltext 0 0 :\n'
1553
'a-2 fulltext 0 0 0 :\n'
1554
'a-3 fulltext 0 0 1 :'
1556
self.assertEqual({('a-3',), ('a-1',), ('a-2',)}, idx.keys())
1558
('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False)),
1559
('a-2',): ((('a-2',), 0, 0), None, (('a-1',),), ('fulltext', False)),
1560
('a-3',): ((('a-3',), 0, 0), None, (('a-2',),), ('fulltext', False)),
1561
}, idx.get_build_details(idx.keys()))
1562
self.assertEqual({('a-1',):(),
1563
('a-2',):(('a-1',),),
1564
('a-3',):(('a-2',),),},
1565
idx.get_parent_map(idx.keys()))
1567
def test_add_versions_fails_clean(self):
1568
"""If add_versions fails in the middle, it restores a pristine state.
1570
Any modifications that are made to the index are reset if all versions
1573
# This cheats a little bit by passing in a generator which will
1574
# raise an exception before the processing finishes
1575
# Other possibilities would be to have an version with the wrong number
1576
# of entries, or to make the backing transport unable to write any
1579
knit = self.make_test_knit()
1581
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1583
class StopEarly(Exception):
1586
def generate_failure():
1587
"""Add some entries and then raise an exception"""
1588
yield (('a-2',), ['fulltext'], (None, 0, 0), ('a-1',))
1589
yield (('a-3',), ['fulltext'], (None, 0, 0), ('a-2',))
1592
# Assert the pre-condition
1594
self.assertEqual({('a-1',)}, set(idx.keys()))
1596
{('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False))},
1597
idx.get_build_details([('a-1',)]))
1598
self.assertEqual({('a-1',):()}, idx.get_parent_map(idx.keys()))
1601
self.assertRaises(StopEarly, idx.add_records, generate_failure())
1602
# And it shouldn't be modified
1605
def test_knit_index_ignores_empty_files(self):
1606
# There was a race condition in older bzr, where a ^C at the right time
1607
# could leave an empty .kndx file, which bzr would later claim was a
1608
# corrupted file since the header was not present. In reality, the file
1609
# just wasn't created, so it should be ignored.
1610
t = transport.get_transport_from_path('.')
1611
t.put_bytes('test.kndx', '')
1613
knit = self.make_test_knit()
1615
def test_knit_index_checks_header(self):
1616
t = transport.get_transport_from_path('.')
1617
t.put_bytes('test.kndx', '# not really a knit header\n\n')
1618
k = self.make_test_knit()
1619
self.assertRaises(KnitHeaderError, k.keys)
1622
class TestGraphIndexKnit(KnitTests):
1623
"""Tests for knits using a GraphIndex rather than a KnitIndex."""
1625
def make_g_index(self, name, ref_lists=0, nodes=[]):
1626
builder = GraphIndexBuilder(ref_lists)
1627
for node, references, value in nodes:
1628
builder.add_node(node, references, value)
1629
stream = builder.finish()
1630
trans = self.get_transport()
1631
size = trans.put_file(name, stream)
1632
return GraphIndex(trans, name, size)
1634
def two_graph_index(self, deltas=False, catch_adds=False):
1635
"""Build a two-graph index.
1637
:param deltas: If true, use underlying indices with two node-ref
1638
lists and 'parent' set to a delta-compressed against tail.
1640
# build a complex graph across several indices.
1642
# delta compression inn the index
1643
index1 = self.make_g_index('1', 2, [
1644
(('tip', ), 'N0 100', ([('parent', )], [], )),
1645
(('tail', ), '', ([], []))])
1646
index2 = self.make_g_index('2', 2, [
1647
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], [('tail', )])),
1648
(('separate', ), '', ([], []))])
1650
# just blob location and graph in the index.
1651
index1 = self.make_g_index('1', 1, [
1652
(('tip', ), 'N0 100', ([('parent', )], )),
1653
(('tail', ), '', ([], ))])
1654
index2 = self.make_g_index('2', 1, [
1655
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], )),
1656
(('separate', ), '', ([], ))])
1657
combined_index = CombinedGraphIndex([index1, index2])
1659
self.combined_index = combined_index
1660
self.caught_entries = []
1661
add_callback = self.catch_add
1664
return _KnitGraphIndex(combined_index, lambda:True, deltas=deltas,
1665
add_callback=add_callback)
1667
def test_keys(self):
1668
index = self.two_graph_index()
1669
self.assertEqual({('tail',), ('tip',), ('parent',), ('separate',)},
1672
def test_get_position(self):
1673
index = self.two_graph_index()
1674
self.assertEqual((index._graph_index._indices[0], 0, 100), index.get_position(('tip',)))
1675
self.assertEqual((index._graph_index._indices[1], 100, 78), index.get_position(('parent',)))
1677
def test_get_method_deltas(self):
1678
index = self.two_graph_index(deltas=True)
1679
self.assertEqual('fulltext', index.get_method(('tip',)))
1680
self.assertEqual('line-delta', index.get_method(('parent',)))
1682
def test_get_method_no_deltas(self):
1683
# check that the parent-history lookup is ignored with deltas=False.
1684
index = self.two_graph_index(deltas=False)
1685
self.assertEqual('fulltext', index.get_method(('tip',)))
1686
self.assertEqual('fulltext', index.get_method(('parent',)))
1688
def test_get_options_deltas(self):
1689
index = self.two_graph_index(deltas=True)
1690
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1691
self.assertEqual(['line-delta'], index.get_options(('parent',)))
1693
def test_get_options_no_deltas(self):
1694
# check that the parent-history lookup is ignored with deltas=False.
1695
index = self.two_graph_index(deltas=False)
1696
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1697
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1699
def test_get_parent_map(self):
1700
index = self.two_graph_index()
1701
self.assertEqual({('parent',):(('tail',), ('ghost',))},
1702
index.get_parent_map([('parent',), ('ghost',)]))
1704
def catch_add(self, entries):
1705
self.caught_entries.append(entries)
1707
def test_add_no_callback_errors(self):
1708
index = self.two_graph_index()
1709
self.assertRaises(errors.ReadOnlyError, index.add_records,
1710
[(('new',), 'fulltext,no-eol', (None, 50, 60), ['separate'])])
1712
def test_add_version_smoke(self):
1713
index = self.two_graph_index(catch_adds=True)
1714
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60),
1716
self.assertEqual([[(('new', ), 'N50 60', ((('separate',),),))]],
1717
self.caught_entries)
1719
def test_add_version_delta_not_delta_index(self):
1720
index = self.two_graph_index(catch_adds=True)
1721
self.assertRaises(errors.KnitCorrupt, index.add_records,
1722
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1723
self.assertEqual([], self.caught_entries)
1725
def test_add_version_same_dup(self):
1726
index = self.two_graph_index(catch_adds=True)
1727
# options can be spelt two different ways
1728
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
1729
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
1730
# position/length are ignored (because each pack could have fulltext or
1731
# delta, and be at a different position.
1732
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1734
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1736
# but neither should have added data:
1737
self.assertEqual([[], [], [], []], self.caught_entries)
1739
def test_add_version_different_dup(self):
1740
index = self.two_graph_index(deltas=True, catch_adds=True)
1742
self.assertRaises(errors.KnitCorrupt, index.add_records,
1743
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1744
self.assertRaises(errors.KnitCorrupt, index.add_records,
1745
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1747
self.assertRaises(errors.KnitCorrupt, index.add_records,
1748
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1749
self.assertEqual([], self.caught_entries)
1751
def test_add_versions_nodeltas(self):
1752
index = self.two_graph_index(catch_adds=True)
1754
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1755
(('new2',), 'fulltext', (None, 0, 6), [('new',)]),
1757
self.assertEqual([(('new', ), 'N50 60', ((('separate',),),)),
1758
(('new2', ), ' 0 6', ((('new',),),))],
1759
sorted(self.caught_entries[0]))
1760
self.assertEqual(1, len(self.caught_entries))
1762
def test_add_versions_deltas(self):
1763
index = self.two_graph_index(deltas=True, catch_adds=True)
1765
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1766
(('new2',), 'line-delta', (None, 0, 6), [('new',)]),
1768
self.assertEqual([(('new', ), 'N50 60', ((('separate',),), ())),
1769
(('new2', ), ' 0 6', ((('new',),), (('new',),), ))],
1770
sorted(self.caught_entries[0]))
1771
self.assertEqual(1, len(self.caught_entries))
1773
def test_add_versions_delta_not_delta_index(self):
1774
index = self.two_graph_index(catch_adds=True)
1775
self.assertRaises(errors.KnitCorrupt, index.add_records,
1776
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1777
self.assertEqual([], self.caught_entries)
1779
def test_add_versions_random_id_accepted(self):
1780
index = self.two_graph_index(catch_adds=True)
1781
index.add_records([], random_id=True)
1783
def test_add_versions_same_dup(self):
1784
index = self.two_graph_index(catch_adds=True)
1785
# options can be spelt two different ways
1786
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100),
1788
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100),
1790
# position/length are ignored (because each pack could have fulltext or
1791
# delta, and be at a different position.
1792
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1794
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1796
# but neither should have added data.
1797
self.assertEqual([[], [], [], []], self.caught_entries)
1799
def test_add_versions_different_dup(self):
1800
index = self.two_graph_index(deltas=True, catch_adds=True)
1802
self.assertRaises(errors.KnitCorrupt, index.add_records,
1803
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1804
self.assertRaises(errors.KnitCorrupt, index.add_records,
1805
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1807
self.assertRaises(errors.KnitCorrupt, index.add_records,
1808
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1809
# change options in the second record
1810
self.assertRaises(errors.KnitCorrupt, index.add_records,
1811
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)]),
1812
(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1813
self.assertEqual([], self.caught_entries)
1815
def make_g_index_missing_compression_parent(self):
1816
graph_index = self.make_g_index('missing_comp', 2,
1817
[(('tip', ), ' 100 78',
1818
([('missing-parent', ), ('ghost', )], [('missing-parent', )]))])
1821
def make_g_index_missing_parent(self):
1822
graph_index = self.make_g_index('missing_parent', 2,
1823
[(('parent', ), ' 100 78', ([], [])),
1824
(('tip', ), ' 100 78',
1825
([('parent', ), ('missing-parent', )], [('parent', )])),
1829
def make_g_index_no_external_refs(self):
1830
graph_index = self.make_g_index('no_external_refs', 2,
1831
[(('rev', ), ' 100 78',
1832
([('parent', ), ('ghost', )], []))])
1835
def test_add_good_unvalidated_index(self):
1836
unvalidated = self.make_g_index_no_external_refs()
1837
combined = CombinedGraphIndex([unvalidated])
1838
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1839
index.scan_unvalidated_index(unvalidated)
1840
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1842
def test_add_missing_compression_parent_unvalidated_index(self):
1843
unvalidated = self.make_g_index_missing_compression_parent()
1844
combined = CombinedGraphIndex([unvalidated])
1845
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1846
index.scan_unvalidated_index(unvalidated)
1847
# This also checks that its only the compression parent that is
1848
# examined, otherwise 'ghost' would also be reported as a missing
1851
frozenset([('missing-parent',)]),
1852
index.get_missing_compression_parents())
1854
def test_add_missing_noncompression_parent_unvalidated_index(self):
1855
unvalidated = self.make_g_index_missing_parent()
1856
combined = CombinedGraphIndex([unvalidated])
1857
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1858
track_external_parent_refs=True)
1859
index.scan_unvalidated_index(unvalidated)
1861
frozenset([('missing-parent',)]), index.get_missing_parents())
1863
def test_track_external_parent_refs(self):
1864
g_index = self.make_g_index('empty', 2, [])
1865
combined = CombinedGraphIndex([g_index])
1866
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1867
add_callback=self.catch_add, track_external_parent_refs=True)
1868
self.caught_entries = []
1870
(('new-key',), 'fulltext,no-eol', (None, 50, 60),
1871
[('parent-1',), ('parent-2',)])])
1873
frozenset([('parent-1',), ('parent-2',)]),
1874
index.get_missing_parents())
1876
def test_add_unvalidated_index_with_present_external_references(self):
1877
index = self.two_graph_index(deltas=True)
1878
# Ugly hack to get at one of the underlying GraphIndex objects that
1879
# two_graph_index built.
1880
unvalidated = index._graph_index._indices[1]
1881
# 'parent' is an external ref of _indices[1] (unvalidated), but is
1882
# present in _indices[0].
1883
index.scan_unvalidated_index(unvalidated)
1884
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1886
def make_new_missing_parent_g_index(self, name):
1887
missing_parent = name + '-missing-parent'
1888
graph_index = self.make_g_index(name, 2,
1889
[((name + 'tip', ), ' 100 78',
1890
([(missing_parent, ), ('ghost', )], [(missing_parent, )]))])
1893
def test_add_mulitiple_unvalidated_indices_with_missing_parents(self):
1894
g_index_1 = self.make_new_missing_parent_g_index('one')
1895
g_index_2 = self.make_new_missing_parent_g_index('two')
1896
combined = CombinedGraphIndex([g_index_1, g_index_2])
1897
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1898
index.scan_unvalidated_index(g_index_1)
1899
index.scan_unvalidated_index(g_index_2)
1901
frozenset([('one-missing-parent',), ('two-missing-parent',)]),
1902
index.get_missing_compression_parents())
1904
def test_add_mulitiple_unvalidated_indices_with_mutual_dependencies(self):
1905
graph_index_a = self.make_g_index('one', 2,
1906
[(('parent-one', ), ' 100 78', ([('non-compression-parent',)], [])),
1907
(('child-of-two', ), ' 100 78',
1908
([('parent-two',)], [('parent-two',)]))])
1909
graph_index_b = self.make_g_index('two', 2,
1910
[(('parent-two', ), ' 100 78', ([('non-compression-parent',)], [])),
1911
(('child-of-one', ), ' 100 78',
1912
([('parent-one',)], [('parent-one',)]))])
1913
combined = CombinedGraphIndex([graph_index_a, graph_index_b])
1914
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1915
index.scan_unvalidated_index(graph_index_a)
1916
index.scan_unvalidated_index(graph_index_b)
1918
frozenset([]), index.get_missing_compression_parents())
1921
class TestNoParentsGraphIndexKnit(KnitTests):
1922
"""Tests for knits using _KnitGraphIndex with no parents."""
1924
def make_g_index(self, name, ref_lists=0, nodes=[]):
1925
builder = GraphIndexBuilder(ref_lists)
1926
for node, references in nodes:
1927
builder.add_node(node, references)
1928
stream = builder.finish()
1929
trans = self.get_transport()
1930
size = trans.put_file(name, stream)
1931
return GraphIndex(trans, name, size)
1933
def test_add_good_unvalidated_index(self):
1934
unvalidated = self.make_g_index('unvalidated')
1935
combined = CombinedGraphIndex([unvalidated])
1936
index = _KnitGraphIndex(combined, lambda: True, parents=False)
1937
index.scan_unvalidated_index(unvalidated)
1938
self.assertEqual(frozenset(),
1939
index.get_missing_compression_parents())
1941
def test_parents_deltas_incompatible(self):
1942
index = CombinedGraphIndex([])
1943
self.assertRaises(errors.KnitError, _KnitGraphIndex, lambda:True,
1944
index, deltas=True, parents=False)
1946
def two_graph_index(self, catch_adds=False):
1947
"""Build a two-graph index.
1949
:param deltas: If true, use underlying indices with two node-ref
1950
lists and 'parent' set to a delta-compressed against tail.
1952
# put several versions in the index.
1953
index1 = self.make_g_index('1', 0, [
1954
(('tip', ), 'N0 100'),
1956
index2 = self.make_g_index('2', 0, [
1957
(('parent', ), ' 100 78'),
1958
(('separate', ), '')])
1959
combined_index = CombinedGraphIndex([index1, index2])
1961
self.combined_index = combined_index
1962
self.caught_entries = []
1963
add_callback = self.catch_add
1966
return _KnitGraphIndex(combined_index, lambda:True, parents=False,
1967
add_callback=add_callback)
1969
def test_keys(self):
1970
index = self.two_graph_index()
1971
self.assertEqual({('tail',), ('tip',), ('parent',), ('separate',)},
1974
def test_get_position(self):
1975
index = self.two_graph_index()
1976
self.assertEqual((index._graph_index._indices[0], 0, 100),
1977
index.get_position(('tip',)))
1978
self.assertEqual((index._graph_index._indices[1], 100, 78),
1979
index.get_position(('parent',)))
1981
def test_get_method(self):
1982
index = self.two_graph_index()
1983
self.assertEqual('fulltext', index.get_method(('tip',)))
1984
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1986
def test_get_options(self):
1987
index = self.two_graph_index()
1988
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1989
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1991
def test_get_parent_map(self):
1992
index = self.two_graph_index()
1993
self.assertEqual({('parent',):None},
1994
index.get_parent_map([('parent',), ('ghost',)]))
1996
def catch_add(self, entries):
1997
self.caught_entries.append(entries)
1999
def test_add_no_callback_errors(self):
2000
index = self.two_graph_index()
2001
self.assertRaises(errors.ReadOnlyError, index.add_records,
2002
[(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)])])
2004
def test_add_version_smoke(self):
2005
index = self.two_graph_index(catch_adds=True)
2006
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60), [])])
2007
self.assertEqual([[(('new', ), 'N50 60')]],
2008
self.caught_entries)
2010
def test_add_version_delta_not_delta_index(self):
2011
index = self.two_graph_index(catch_adds=True)
2012
self.assertRaises(errors.KnitCorrupt, index.add_records,
2013
[(('new',), 'no-eol,line-delta', (None, 0, 100), [])])
2014
self.assertEqual([], self.caught_entries)
2016
def test_add_version_same_dup(self):
2017
index = self.two_graph_index(catch_adds=True)
2018
# options can be spelt two different ways
2019
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
2020
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
2021
# position/length are ignored (because each pack could have fulltext or
2022
# delta, and be at a different position.
2023
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
2024
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
2025
# but neither should have added data.
2026
self.assertEqual([[], [], [], []], self.caught_entries)
2028
def test_add_version_different_dup(self):
2029
index = self.two_graph_index(catch_adds=True)
2031
self.assertRaises(errors.KnitCorrupt, index.add_records,
2032
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2033
self.assertRaises(errors.KnitCorrupt, index.add_records,
2034
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
2035
self.assertRaises(errors.KnitCorrupt, index.add_records,
2036
[(('tip',), 'fulltext', (None, 0, 100), [])])
2038
self.assertRaises(errors.KnitCorrupt, index.add_records,
2039
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
2040
self.assertEqual([], self.caught_entries)
2042
def test_add_versions(self):
2043
index = self.two_graph_index(catch_adds=True)
2045
(('new',), 'fulltext,no-eol', (None, 50, 60), []),
2046
(('new2',), 'fulltext', (None, 0, 6), []),
2048
self.assertEqual([(('new', ), 'N50 60'), (('new2', ), ' 0 6')],
2049
sorted(self.caught_entries[0]))
2050
self.assertEqual(1, len(self.caught_entries))
2052
def test_add_versions_delta_not_delta_index(self):
2053
index = self.two_graph_index(catch_adds=True)
2054
self.assertRaises(errors.KnitCorrupt, index.add_records,
2055
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
2056
self.assertEqual([], self.caught_entries)
2058
def test_add_versions_parents_not_parents_index(self):
2059
index = self.two_graph_index(catch_adds=True)
2060
self.assertRaises(errors.KnitCorrupt, index.add_records,
2061
[(('new',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
2062
self.assertEqual([], self.caught_entries)
2064
def test_add_versions_random_id_accepted(self):
2065
index = self.two_graph_index(catch_adds=True)
2066
index.add_records([], random_id=True)
2068
def test_add_versions_same_dup(self):
2069
index = self.two_graph_index(catch_adds=True)
2070
# options can be spelt two different ways
2071
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
2072
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
2073
# position/length are ignored (because each pack could have fulltext or
2074
# delta, and be at a different position.
2075
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
2076
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
2077
# but neither should have added data.
2078
self.assertEqual([[], [], [], []], self.caught_entries)
2080
def test_add_versions_different_dup(self):
2081
index = self.two_graph_index(catch_adds=True)
2083
self.assertRaises(errors.KnitCorrupt, index.add_records,
2084
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2085
self.assertRaises(errors.KnitCorrupt, index.add_records,
2086
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
2087
self.assertRaises(errors.KnitCorrupt, index.add_records,
2088
[(('tip',), 'fulltext', (None, 0, 100), [])])
2090
self.assertRaises(errors.KnitCorrupt, index.add_records,
2091
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
2092
# change options in the second record
2093
self.assertRaises(errors.KnitCorrupt, index.add_records,
2094
[(('tip',), 'fulltext,no-eol', (None, 0, 100), []),
2095
(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2096
self.assertEqual([], self.caught_entries)
2099
class TestKnitVersionedFiles(KnitTests):
2101
def assertGroupKeysForIo(self, exp_groups, keys, non_local_keys,
2102
positions, _min_buffer_size=None):
2103
kvf = self.make_test_knit()
2104
if _min_buffer_size is None:
2105
_min_buffer_size = knit._STREAM_MIN_BUFFER_SIZE
2106
self.assertEqual(exp_groups, kvf._group_keys_for_io(keys,
2107
non_local_keys, positions,
2108
_min_buffer_size=_min_buffer_size))
2110
def assertSplitByPrefix(self, expected_map, expected_prefix_order,
2112
split, prefix_order = KnitVersionedFiles._split_by_prefix(keys)
2113
self.assertEqual(expected_map, split)
2114
self.assertEqual(expected_prefix_order, prefix_order)
2116
def test__group_keys_for_io(self):
2117
ft_detail = ('fulltext', False)
2118
ld_detail = ('line-delta', False)
2126
f_a: (ft_detail, (f_a, 0, 100), None),
2127
f_b: (ld_detail, (f_b, 100, 21), f_a),
2128
f_c: (ld_detail, (f_c, 180, 15), f_b),
2129
g_a: (ft_detail, (g_a, 121, 35), None),
2130
g_b: (ld_detail, (g_b, 156, 12), g_a),
2131
g_c: (ld_detail, (g_c, 195, 13), g_a),
2133
self.assertGroupKeysForIo([([f_a], set())],
2134
[f_a], [], positions)
2135
self.assertGroupKeysForIo([([f_a], {f_a})],
2136
[f_a], [f_a], positions)
2137
self.assertGroupKeysForIo([([f_a, f_b], set([]))],
2138
[f_a, f_b], [], positions)
2139
self.assertGroupKeysForIo([([f_a, f_b], {f_b})],
2140
[f_a, f_b], [f_b], positions)
2141
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2142
[f_a, g_a, f_b, g_b], [], positions)
2143
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2144
[f_a, g_a, f_b, g_b], [], positions,
2145
_min_buffer_size=150)
2146
self.assertGroupKeysForIo([([f_a, f_b], set()), ([g_a, g_b], set())],
2147
[f_a, g_a, f_b, g_b], [], positions,
2148
_min_buffer_size=100)
2149
self.assertGroupKeysForIo([([f_c], set()), ([g_b], set())],
2150
[f_c, g_b], [], positions,
2151
_min_buffer_size=125)
2152
self.assertGroupKeysForIo([([g_b, f_c], set())],
2153
[g_b, f_c], [], positions,
2154
_min_buffer_size=125)
2156
def test__split_by_prefix(self):
2157
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2158
'g': [('g', 'b'), ('g', 'a')],
2160
[('f', 'a'), ('g', 'b'),
2161
('g', 'a'), ('f', 'b')])
2163
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2164
'g': [('g', 'b'), ('g', 'a')],
2166
[('f', 'a'), ('f', 'b'),
2167
('g', 'b'), ('g', 'a')])
2169
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2170
'g': [('g', 'b'), ('g', 'a')],
2172
[('f', 'a'), ('f', 'b'),
2173
('g', 'b'), ('g', 'a')])
2175
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2176
'g': [('g', 'b'), ('g', 'a')],
2177
'': [('a',), ('b',)]
2179
[('f', 'a'), ('g', 'b'),
2181
('g', 'a'), ('f', 'b')])
2184
class TestStacking(KnitTests):
2186
def get_basis_and_test_knit(self):
2187
basis = self.make_test_knit(name='basis')
2188
basis = RecordingVersionedFilesDecorator(basis)
2189
test = self.make_test_knit(name='test')
2190
test.add_fallback_versioned_files(basis)
2193
def test_add_fallback_versioned_files(self):
2194
basis = self.make_test_knit(name='basis')
2195
test = self.make_test_knit(name='test')
2196
# It must not error; other tests test that the fallback is referred to
2197
# when accessing data.
2198
test.add_fallback_versioned_files(basis)
2200
def test_add_lines(self):
2201
# lines added to the test are not added to the basis
2202
basis, test = self.get_basis_and_test_knit()
2204
key_basis = ('bar',)
2205
key_cross_border = ('quux',)
2206
key_delta = ('zaphod',)
2207
test.add_lines(key, (), ['foo\n'])
2208
self.assertEqual({}, basis.get_parent_map([key]))
2209
# lines added to the test that reference across the stack do a
2211
basis.add_lines(key_basis, (), ['foo\n'])
2213
test.add_lines(key_cross_border, (key_basis,), ['foo\n'])
2214
self.assertEqual('fulltext', test._index.get_method(key_cross_border))
2215
# we don't even need to look at the basis to see that this should be
2216
# stored as a fulltext
2217
self.assertEqual([], basis.calls)
2218
# Subsequent adds do delta.
2220
test.add_lines(key_delta, (key_cross_border,), ['foo\n'])
2221
self.assertEqual('line-delta', test._index.get_method(key_delta))
2222
self.assertEqual([], basis.calls)
2224
def test_annotate(self):
2225
# annotations from the test knit are answered without asking the basis
2226
basis, test = self.get_basis_and_test_knit()
2228
key_basis = ('bar',)
2229
key_missing = ('missing',)
2230
test.add_lines(key, (), ['foo\n'])
2231
details = test.annotate(key)
2232
self.assertEqual([(key, 'foo\n')], details)
2233
self.assertEqual([], basis.calls)
2234
# But texts that are not in the test knit are looked for in the basis
2236
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2238
details = test.annotate(key_basis)
2239
self.assertEqual([(key_basis, 'foo\n'), (key_basis, 'bar\n')], details)
2240
# Not optimised to date:
2241
# self.assertEqual([("annotate", key_basis)], basis.calls)
2242
self.assertEqual([('get_parent_map', {key_basis}),
2243
('get_parent_map', {key_basis}),
2244
('get_record_stream', [key_basis], 'topological', True)],
2247
def test_check(self):
2248
# At the moment checking a stacked knit does implicitly check the
2250
basis, test = self.get_basis_and_test_knit()
2253
def test_get_parent_map(self):
2254
# parents in the test knit are answered without asking the basis
2255
basis, test = self.get_basis_and_test_knit()
2257
key_basis = ('bar',)
2258
key_missing = ('missing',)
2259
test.add_lines(key, (), [])
2260
parent_map = test.get_parent_map([key])
2261
self.assertEqual({key: ()}, parent_map)
2262
self.assertEqual([], basis.calls)
2263
# But parents that are not in the test knit are looked for in the basis
2264
basis.add_lines(key_basis, (), [])
2266
parent_map = test.get_parent_map([key, key_basis, key_missing])
2267
self.assertEqual({key: (),
2268
key_basis: ()}, parent_map)
2269
self.assertEqual([("get_parent_map", {key_basis, key_missing})],
2272
def test_get_record_stream_unordered_fulltexts(self):
2273
# records from the test knit are answered without asking the basis:
2274
basis, test = self.get_basis_and_test_knit()
2276
key_basis = ('bar',)
2277
key_missing = ('missing',)
2278
test.add_lines(key, (), ['foo\n'])
2279
records = list(test.get_record_stream([key], 'unordered', True))
2280
self.assertEqual(1, len(records))
2281
self.assertEqual([], basis.calls)
2282
# Missing (from test knit) objects are retrieved from the basis:
2283
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2285
records = list(test.get_record_stream([key_basis, key_missing],
2287
self.assertEqual(2, len(records))
2288
calls = list(basis.calls)
2289
for record in records:
2290
self.assertSubset([record.key], (key_basis, key_missing))
2291
if record.key == key_missing:
2292
self.assertIsInstance(record, AbsentContentFactory)
2294
reference = list(basis.get_record_stream([key_basis],
2295
'unordered', True))[0]
2296
self.assertEqual(reference.key, record.key)
2297
self.assertEqual(reference.sha1, record.sha1)
2298
self.assertEqual(reference.storage_kind, record.storage_kind)
2299
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2300
record.get_bytes_as(record.storage_kind))
2301
self.assertEqual(reference.get_bytes_as('fulltext'),
2302
record.get_bytes_as('fulltext'))
2303
# It's not strictly minimal, but it seems reasonable for now for it to
2304
# ask which fallbacks have which parents.
2306
("get_parent_map", {key_basis, key_missing}),
2307
("get_record_stream", [key_basis], 'unordered', True)],
2310
def test_get_record_stream_ordered_fulltexts(self):
2311
# ordering is preserved down into the fallback store.
2312
basis, test = self.get_basis_and_test_knit()
2314
key_basis = ('bar',)
2315
key_basis_2 = ('quux',)
2316
key_missing = ('missing',)
2317
test.add_lines(key, (key_basis,), ['foo\n'])
2318
# Missing (from test knit) objects are retrieved from the basis:
2319
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2320
basis.add_lines(key_basis_2, (), ['quux\n'])
2322
# ask for in non-topological order
2323
records = list(test.get_record_stream(
2324
[key, key_basis, key_missing, key_basis_2], 'topological', True))
2325
self.assertEqual(4, len(records))
2327
for record in records:
2328
self.assertSubset([record.key],
2329
(key_basis, key_missing, key_basis_2, key))
2330
if record.key == key_missing:
2331
self.assertIsInstance(record, AbsentContentFactory)
2333
results.append((record.key, record.sha1, record.storage_kind,
2334
record.get_bytes_as('fulltext')))
2335
calls = list(basis.calls)
2336
order = [record[0] for record in results]
2337
self.assertEqual([key_basis_2, key_basis, key], order)
2338
for result in results:
2339
if result[0] == key:
2343
record = next(source.get_record_stream([result[0]], 'unordered',
2345
self.assertEqual(record.key, result[0])
2346
self.assertEqual(record.sha1, result[1])
2347
# We used to check that the storage kind matched, but actually it
2348
# depends on whether it was sourced from the basis, or in a single
2349
# group, because asking for full texts returns proxy objects to a
2350
# _ContentMapGenerator object; so checking the kind is unneeded.
2351
self.assertEqual(record.get_bytes_as('fulltext'), result[3])
2352
# It's not strictly minimal, but it seems reasonable for now for it to
2353
# ask which fallbacks have which parents.
2355
("get_parent_map", {key_basis, key_basis_2, key_missing}),
2356
# topological is requested from the fallback, because that is what
2357
# was requested at the top level.
2358
("get_record_stream", [key_basis_2, key_basis], 'topological', True)],
2361
def test_get_record_stream_unordered_deltas(self):
2362
# records from the test knit are answered without asking the basis:
2363
basis, test = self.get_basis_and_test_knit()
2365
key_basis = ('bar',)
2366
key_missing = ('missing',)
2367
test.add_lines(key, (), ['foo\n'])
2368
records = list(test.get_record_stream([key], 'unordered', False))
2369
self.assertEqual(1, len(records))
2370
self.assertEqual([], basis.calls)
2371
# Missing (from test knit) objects are retrieved from the basis:
2372
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2374
records = list(test.get_record_stream([key_basis, key_missing],
2375
'unordered', False))
2376
self.assertEqual(2, len(records))
2377
calls = list(basis.calls)
2378
for record in records:
2379
self.assertSubset([record.key], (key_basis, key_missing))
2380
if record.key == key_missing:
2381
self.assertIsInstance(record, AbsentContentFactory)
2383
reference = list(basis.get_record_stream([key_basis],
2384
'unordered', False))[0]
2385
self.assertEqual(reference.key, record.key)
2386
self.assertEqual(reference.sha1, record.sha1)
2387
self.assertEqual(reference.storage_kind, record.storage_kind)
2388
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2389
record.get_bytes_as(record.storage_kind))
2390
# It's not strictly minimal, but it seems reasonable for now for it to
2391
# ask which fallbacks have which parents.
2393
("get_parent_map", {key_basis, key_missing}),
2394
("get_record_stream", [key_basis], 'unordered', False)],
2397
def test_get_record_stream_ordered_deltas(self):
2398
# ordering is preserved down into the fallback store.
2399
basis, test = self.get_basis_and_test_knit()
2401
key_basis = ('bar',)
2402
key_basis_2 = ('quux',)
2403
key_missing = ('missing',)
2404
test.add_lines(key, (key_basis,), ['foo\n'])
2405
# Missing (from test knit) objects are retrieved from the basis:
2406
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2407
basis.add_lines(key_basis_2, (), ['quux\n'])
2409
# ask for in non-topological order
2410
records = list(test.get_record_stream(
2411
[key, key_basis, key_missing, key_basis_2], 'topological', False))
2412
self.assertEqual(4, len(records))
2414
for record in records:
2415
self.assertSubset([record.key],
2416
(key_basis, key_missing, key_basis_2, key))
2417
if record.key == key_missing:
2418
self.assertIsInstance(record, AbsentContentFactory)
2420
results.append((record.key, record.sha1, record.storage_kind,
2421
record.get_bytes_as(record.storage_kind)))
2422
calls = list(basis.calls)
2423
order = [record[0] for record in results]
2424
self.assertEqual([key_basis_2, key_basis, key], order)
2425
for result in results:
2426
if result[0] == key:
2430
record = next(source.get_record_stream([result[0]], 'unordered',
2432
self.assertEqual(record.key, result[0])
2433
self.assertEqual(record.sha1, result[1])
2434
self.assertEqual(record.storage_kind, result[2])
2435
self.assertEqual(record.get_bytes_as(record.storage_kind), result[3])
2436
# It's not strictly minimal, but it seems reasonable for now for it to
2437
# ask which fallbacks have which parents.
2439
("get_parent_map", {key_basis, key_basis_2, key_missing}),
2440
("get_record_stream", [key_basis_2, key_basis], 'topological', False)],
2443
def test_get_sha1s(self):
2444
# sha1's in the test knit are answered without asking the basis
2445
basis, test = self.get_basis_and_test_knit()
2447
key_basis = ('bar',)
2448
key_missing = ('missing',)
2449
test.add_lines(key, (), ['foo\n'])
2450
key_sha1sum = osutils.sha_string('foo\n')
2451
sha1s = test.get_sha1s([key])
2452
self.assertEqual({key: key_sha1sum}, sha1s)
2453
self.assertEqual([], basis.calls)
2454
# But texts that are not in the test knit are looked for in the basis
2455
# directly (rather than via text reconstruction) so that remote servers
2456
# etc don't have to answer with full content.
2457
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2458
basis_sha1sum = osutils.sha_string('foo\nbar\n')
2460
sha1s = test.get_sha1s([key, key_missing, key_basis])
2461
self.assertEqual({key: key_sha1sum,
2462
key_basis: basis_sha1sum}, sha1s)
2463
self.assertEqual([("get_sha1s", {key_basis, key_missing})],
2466
def test_insert_record_stream(self):
2467
# records are inserted as normal; insert_record_stream builds on
2468
# add_lines, so a smoke test should be all that's needed:
2470
key_basis = ('bar',)
2471
key_delta = ('zaphod',)
2472
basis, test = self.get_basis_and_test_knit()
2473
source = self.make_test_knit(name='source')
2474
basis.add_lines(key_basis, (), ['foo\n'])
2476
source.add_lines(key_basis, (), ['foo\n'])
2477
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2478
stream = source.get_record_stream([key_delta], 'unordered', False)
2479
test.insert_record_stream(stream)
2480
# XXX: this does somewhat too many calls in making sure of whether it
2481
# has to recreate the full text.
2482
self.assertEqual([("get_parent_map", {key_basis}),
2483
('get_parent_map', {key_basis}),
2484
('get_record_stream', [key_basis], 'unordered', True)],
2486
self.assertEqual({key_delta:(key_basis,)},
2487
test.get_parent_map([key_delta]))
2488
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2489
'unordered', True).next().get_bytes_as('fulltext'))
2491
def test_iter_lines_added_or_present_in_keys(self):
2492
# Lines from the basis are returned, and lines for a given key are only
2496
# all sources are asked for keys:
2497
basis, test = self.get_basis_and_test_knit()
2498
basis.add_lines(key1, (), ["foo"])
2500
lines = list(test.iter_lines_added_or_present_in_keys([key1]))
2501
self.assertEqual([("foo\n", key1)], lines)
2502
self.assertEqual([("iter_lines_added_or_present_in_keys", {key1})],
2504
# keys in both are not duplicated:
2505
test.add_lines(key2, (), ["bar\n"])
2506
basis.add_lines(key2, (), ["bar\n"])
2508
lines = list(test.iter_lines_added_or_present_in_keys([key2]))
2509
self.assertEqual([("bar\n", key2)], lines)
2510
self.assertEqual([], basis.calls)
2512
def test_keys(self):
2515
# all sources are asked for keys:
2516
basis, test = self.get_basis_and_test_knit()
2518
self.assertEqual(set(), set(keys))
2519
self.assertEqual([("keys",)], basis.calls)
2520
# keys from a basis are returned:
2521
basis.add_lines(key1, (), [])
2524
self.assertEqual({key1}, set(keys))
2525
self.assertEqual([("keys",)], basis.calls)
2526
# keys in both are not duplicated:
2527
test.add_lines(key2, (), [])
2528
basis.add_lines(key2, (), [])
2531
self.assertEqual(2, len(keys))
2532
self.assertEqual({key1, key2}, set(keys))
2533
self.assertEqual([("keys",)], basis.calls)
2535
def test_add_mpdiffs(self):
2536
# records are inserted as normal; add_mpdiff builds on
2537
# add_lines, so a smoke test should be all that's needed:
2539
key_basis = ('bar',)
2540
key_delta = ('zaphod',)
2541
basis, test = self.get_basis_and_test_knit()
2542
source = self.make_test_knit(name='source')
2543
basis.add_lines(key_basis, (), ['foo\n'])
2545
source.add_lines(key_basis, (), ['foo\n'])
2546
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2547
diffs = source.make_mpdiffs([key_delta])
2548
test.add_mpdiffs([(key_delta, (key_basis,),
2549
source.get_sha1s([key_delta])[key_delta], diffs[0])])
2550
self.assertEqual([("get_parent_map", {key_basis}),
2551
('get_record_stream', [key_basis], 'unordered', True),],
2553
self.assertEqual({key_delta:(key_basis,)},
2554
test.get_parent_map([key_delta]))
2555
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2556
'unordered', True).next().get_bytes_as('fulltext'))
2558
def test_make_mpdiffs(self):
2559
# Generating an mpdiff across a stacking boundary should detect parent
2563
key_right = ('zaphod',)
2564
basis, test = self.get_basis_and_test_knit()
2565
basis.add_lines(key_left, (), ['bar\n'])
2566
basis.add_lines(key_right, (), ['zaphod\n'])
2568
test.add_lines(key, (key_left, key_right),
2569
['bar\n', 'foo\n', 'zaphod\n'])
2570
diffs = test.make_mpdiffs([key])
2572
multiparent.MultiParent([multiparent.ParentText(0, 0, 0, 1),
2573
multiparent.NewText(['foo\n']),
2574
multiparent.ParentText(1, 0, 2, 1)])],
2576
self.assertEqual(3, len(basis.calls))
2578
("get_parent_map", {key_left, key_right}),
2579
("get_parent_map", {key_left, key_right}),
2582
last_call = basis.calls[-1]
2583
self.assertEqual('get_record_stream', last_call[0])
2584
self.assertEqual({key_left, key_right}, set(last_call[1]))
2585
self.assertEqual('topological', last_call[2])
2586
self.assertEqual(True, last_call[3])
2589
class TestNetworkBehaviour(KnitTests):
2590
"""Tests for getting data out of/into knits over the network."""
2592
def test_include_delta_closure_generates_a_knit_delta_closure(self):
2593
vf = self.make_test_knit(name='test')
2594
# put in three texts, giving ft, delta, delta
2595
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2596
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2597
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2598
# But heuristics could interfere, so check what happened:
2599
self.assertEqual(['knit-ft-gz', 'knit-delta-gz', 'knit-delta-gz'],
2600
[record.storage_kind for record in
2601
vf.get_record_stream([('base',), ('d1',), ('d2',)],
2602
'topological', False)])
2603
# generate a stream of just the deltas include_delta_closure=True,
2604
# serialise to the network, and check that we get a delta closure on the wire.
2605
stream = vf.get_record_stream([('d1',), ('d2',)], 'topological', True)
2606
netb = [record.get_bytes_as(record.storage_kind) for record in stream]
2607
# The first bytes should be a memo from _ContentMapGenerator, and the
2608
# second bytes should be empty (because its a API proxy not something
2609
# for wire serialisation.
2610
self.assertEqual('', netb[1])
2612
kind, line_end = network_bytes_to_kind_and_offset(bytes)
2613
self.assertEqual('knit-delta-closure', kind)
2616
class TestContentMapGenerator(KnitTests):
2617
"""Tests for ContentMapGenerator"""
2619
def test_get_record_stream_gives_records(self):
2620
vf = self.make_test_knit(name='test')
2621
# put in three texts, giving ft, delta, delta
2622
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2623
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2624
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2625
keys = [('d1',), ('d2',)]
2626
generator = _VFContentMapGenerator(vf, keys,
2627
global_map=vf.get_parent_map(keys))
2628
for record in generator.get_record_stream():
2629
if record.key == ('d1',):
2630
self.assertEqual('d1\n', record.get_bytes_as('fulltext'))
2632
self.assertEqual('d2\n', record.get_bytes_as('fulltext'))
2634
def test_get_record_stream_kinds_are_raw(self):
2635
vf = self.make_test_knit(name='test')
2636
# put in three texts, giving ft, delta, delta
2637
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2638
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2639
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2640
keys = [('base',), ('d1',), ('d2',)]
2641
generator = _VFContentMapGenerator(vf, keys,
2642
global_map=vf.get_parent_map(keys))
2643
kinds = {('base',): 'knit-delta-closure',
2644
('d1',): 'knit-delta-closure-ref',
2645
('d2',): 'knit-delta-closure-ref',
2647
for record in generator.get_record_stream():
2648
self.assertEqual(kinds[record.key], record.storage_kind)
added
removed
breezy/tests/test_knit.py
