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"""
31
from ..errors import (
41
_VFContentMapGenerator,
47
from ..patiencediff import PatienceSequenceMatcher
48
from ..repofmt import (
52
from ..sixish import (
57
TestCaseWithMemoryTransport,
58
TestCaseWithTransport,
61
from ..versionedfile import (
64
network_bytes_to_kind_and_offset,
65
RecordingVersionedFilesDecorator,
72
compiled_knit_feature = features.ModuleAvailableFeature(
73
'breezy._knit_load_data_pyx')
76
class KnitContentTestsMixin(object):
78
def test_constructor(self):
79
content = self._make_content([])
82
content = self._make_content([])
83
self.assertEqual(content.text(), [])
85
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
86
self.assertEqual(content.text(), ["text1", "text2"])
89
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
91
self.assertIsInstance(copy, content.__class__)
92
self.assertEqual(copy.annotate(), content.annotate())
94
def assertDerivedBlocksEqual(self, source, target, noeol=False):
95
"""Assert that the derived matching blocks match real output"""
96
source_lines = source.splitlines(True)
97
target_lines = target.splitlines(True)
99
if noeol and not line.endswith('\n'):
103
source_content = self._make_content([(None, nl(l)) for l in source_lines])
104
target_content = self._make_content([(None, nl(l)) for l in target_lines])
105
line_delta = source_content.line_delta(target_content)
106
delta_blocks = list(KnitContent.get_line_delta_blocks(line_delta,
107
source_lines, target_lines))
108
matcher = PatienceSequenceMatcher(None, source_lines, target_lines)
109
matcher_blocks = list(matcher.get_matching_blocks())
110
self.assertEqual(matcher_blocks, delta_blocks)
112
def test_get_line_delta_blocks(self):
113
self.assertDerivedBlocksEqual('a\nb\nc\n', 'q\nc\n')
114
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1)
115
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1A)
116
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1B)
117
self.assertDerivedBlocksEqual(TEXT_1B, TEXT_1A)
118
self.assertDerivedBlocksEqual(TEXT_1A, TEXT_1B)
119
self.assertDerivedBlocksEqual(TEXT_1A, '')
120
self.assertDerivedBlocksEqual('', TEXT_1A)
121
self.assertDerivedBlocksEqual('', '')
122
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd')
124
def test_get_line_delta_blocks_noeol(self):
125
"""Handle historical knit deltas safely
127
Some existing knit deltas don't consider the last line to differ
128
when the only difference whether it has a final newline.
130
New knit deltas appear to always consider the last line to differ
133
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd\n', noeol=True)
134
self.assertDerivedBlocksEqual('a\nb\nc\nd\n', 'a\nb\nc', noeol=True)
135
self.assertDerivedBlocksEqual('a\nb\nc\n', 'a\nb\nc', noeol=True)
136
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\n', noeol=True)
148
Banana cup cake recipe
158
Banana cup cake recipe
160
- bananas (do not use plantains!!!)
167
Banana cup cake recipe
184
class TestPlainKnitContent(TestCase, KnitContentTestsMixin):
186
def _make_content(self, lines):
187
annotated_content = AnnotatedKnitContent(lines)
188
return PlainKnitContent(annotated_content.text(), 'bogus')
190
def test_annotate(self):
191
content = self._make_content([])
192
self.assertEqual(content.annotate(), [])
194
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
195
self.assertEqual(content.annotate(),
196
[("bogus", "text1"), ("bogus", "text2")])
198
def test_line_delta(self):
199
content1 = self._make_content([("", "a"), ("", "b")])
200
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
201
self.assertEqual(content1.line_delta(content2),
202
[(1, 2, 2, ["a", "c"])])
204
def test_line_delta_iter(self):
205
content1 = self._make_content([("", "a"), ("", "b")])
206
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
207
it = content1.line_delta_iter(content2)
208
self.assertEqual(next(it), (1, 2, 2, ["a", "c"]))
209
self.assertRaises(StopIteration, next, it)
212
class TestAnnotatedKnitContent(TestCase, KnitContentTestsMixin):
214
def _make_content(self, lines):
215
return AnnotatedKnitContent(lines)
217
def test_annotate(self):
218
content = self._make_content([])
219
self.assertEqual(content.annotate(), [])
221
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
222
self.assertEqual(content.annotate(),
223
[("origin1", "text1"), ("origin2", "text2")])
225
def test_line_delta(self):
226
content1 = self._make_content([("", "a"), ("", "b")])
227
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
228
self.assertEqual(content1.line_delta(content2),
229
[(1, 2, 2, [("", "a"), ("", "c")])])
231
def test_line_delta_iter(self):
232
content1 = self._make_content([("", "a"), ("", "b")])
233
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
234
it = content1.line_delta_iter(content2)
235
self.assertEqual(next(it), (1, 2, 2, [("", "a"), ("", "c")]))
236
self.assertRaises(StopIteration, next, it)
239
class MockTransport(object):
241
def __init__(self, file_lines=None):
242
self.file_lines = file_lines
244
# We have no base directory for the MockTransport
247
def get(self, filename):
248
if self.file_lines is None:
249
raise NoSuchFile(filename)
251
return BytesIO(b"\n".join(self.file_lines))
253
def readv(self, relpath, offsets):
254
fp = self.get(relpath)
255
for offset, size in offsets:
257
yield offset, fp.read(size)
259
def __getattr__(self, name):
260
def queue_call(*args, **kwargs):
261
self.calls.append((name, args, kwargs))
265
class MockReadvFailingTransport(MockTransport):
266
"""Fail in the middle of a readv() result.
268
This Transport will successfully yield the first two requested hunks, but
269
raise NoSuchFile for the rest.
272
def readv(self, relpath, offsets):
274
for result in MockTransport.readv(self, relpath, offsets):
276
# we use 2 because the first offset is the pack header, the second
277
# is the first actual content requset
279
raise errors.NoSuchFile(relpath)
283
class KnitRecordAccessTestsMixin(object):
284
"""Tests for getting and putting knit records."""
286
def test_add_raw_records(self):
287
"""Add_raw_records adds records retrievable later."""
288
access = self.get_access()
289
memos = access.add_raw_records([('key', 10)], '1234567890')
290
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
292
def test_add_several_raw_records(self):
293
"""add_raw_records with many records and read some back."""
294
access = self.get_access()
295
memos = access.add_raw_records([('key', 10), ('key2', 2), ('key3', 5)],
297
self.assertEqual(['1234567890', '12', '34567'],
298
list(access.get_raw_records(memos)))
299
self.assertEqual(['1234567890'],
300
list(access.get_raw_records(memos[0:1])))
301
self.assertEqual(['12'],
302
list(access.get_raw_records(memos[1:2])))
303
self.assertEqual(['34567'],
304
list(access.get_raw_records(memos[2:3])))
305
self.assertEqual(['1234567890', '34567'],
306
list(access.get_raw_records(memos[0:1] + memos[2:3])))
309
class TestKnitKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
310
"""Tests for the .kndx implementation."""
312
def get_access(self):
313
"""Get a .knit style access instance."""
314
mapper = ConstantMapper("foo")
315
access = _KnitKeyAccess(self.get_transport(), mapper)
319
class _TestException(Exception):
320
"""Just an exception for local tests to use."""
323
class TestPackKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
324
"""Tests for the pack based access."""
326
def get_access(self):
327
return self._get_access()[0]
329
def _get_access(self, packname='packfile', index='FOO'):
330
transport = self.get_transport()
331
def write_data(bytes):
332
transport.append_bytes(packname, bytes)
333
writer = pack.ContainerWriter(write_data)
335
access = pack_repo._DirectPackAccess({})
336
access.set_writer(writer, index, (transport, packname))
337
return access, writer
339
def make_pack_file(self):
340
"""Create a pack file with 2 records."""
341
access, writer = self._get_access(packname='packname', index='foo')
343
memos.extend(access.add_raw_records([('key1', 10)], '1234567890'))
344
memos.extend(access.add_raw_records([('key2', 5)], '12345'))
348
def test_pack_collection_pack_retries(self):
349
"""An explicit pack of a pack collection succeeds even when a
350
concurrent pack happens.
352
builder = self.make_branch_builder('.')
353
builder.start_series()
354
builder.build_snapshot('rev-1', None, [
355
('add', ('', 'root-id', 'directory', None)),
356
('add', ('file', 'file-id', 'file', 'content\nrev 1\n')),
358
builder.build_snapshot('rev-2', ['rev-1'], [
359
('modify', ('file-id', 'content\nrev 2\n')),
361
builder.build_snapshot('rev-3', ['rev-2'], [
362
('modify', ('file-id', 'content\nrev 3\n')),
364
self.addCleanup(builder.finish_series)
365
b = builder.get_branch()
366
self.addCleanup(b.lock_write().unlock)
368
collection = repo._pack_collection
369
# Concurrently repack the repo.
370
reopened_repo = repo.bzrdir.open_repository()
375
def make_vf_for_retrying(self):
376
"""Create 3 packs and a reload function.
378
Originally, 2 pack files will have the data, but one will be missing.
379
And then the third will be used in place of the first two if reload()
382
:return: (versioned_file, reload_counter)
383
versioned_file a KnitVersionedFiles using the packs for access
385
builder = self.make_branch_builder('.', format="1.9")
386
builder.start_series()
387
builder.build_snapshot('rev-1', None, [
388
('add', ('', 'root-id', 'directory', None)),
389
('add', ('file', 'file-id', 'file', 'content\nrev 1\n')),
391
builder.build_snapshot('rev-2', ['rev-1'], [
392
('modify', ('file-id', 'content\nrev 2\n')),
394
builder.build_snapshot('rev-3', ['rev-2'], [
395
('modify', ('file-id', 'content\nrev 3\n')),
397
builder.finish_series()
398
b = builder.get_branch()
400
self.addCleanup(b.unlock)
401
# Pack these three revisions into another pack file, but don't remove
404
collection = repo._pack_collection
405
collection.ensure_loaded()
406
orig_packs = collection.packs
407
packer = knitpack_repo.KnitPacker(collection, orig_packs, '.testpack')
408
new_pack = packer.pack()
409
# forget about the new pack
413
# Set up a reload() function that switches to using the new pack file
414
new_index = new_pack.revision_index
415
access_tuple = new_pack.access_tuple()
416
reload_counter = [0, 0, 0]
418
reload_counter[0] += 1
419
if reload_counter[1] > 0:
420
# We already reloaded, nothing more to do
421
reload_counter[2] += 1
423
reload_counter[1] += 1
424
vf._index._graph_index._indices[:] = [new_index]
425
vf._access._indices.clear()
426
vf._access._indices[new_index] = access_tuple
428
# Delete one of the pack files so the data will need to be reloaded. We
429
# will delete the file with 'rev-2' in it
430
trans, name = orig_packs[1].access_tuple()
432
# We don't have the index trigger reloading because we want to test
433
# that we reload when the .pack disappears
434
vf._access._reload_func = reload
435
return vf, reload_counter
437
def make_reload_func(self, return_val=True):
440
reload_called[0] += 1
442
return reload_called, reload
444
def make_retry_exception(self):
445
# We raise a real exception so that sys.exc_info() is properly
448
raise _TestException('foobar')
449
except _TestException as e:
450
retry_exc = errors.RetryWithNewPacks(None, reload_occurred=False,
451
exc_info=sys.exc_info())
452
# GZ 2010-08-10: Cycle with exc_info affects 3 tests
455
def test_read_from_several_packs(self):
456
access, writer = self._get_access()
458
memos.extend(access.add_raw_records([('key', 10)], '1234567890'))
460
access, writer = self._get_access('pack2', 'FOOBAR')
461
memos.extend(access.add_raw_records([('key', 5)], '12345'))
463
access, writer = self._get_access('pack3', 'BAZ')
464
memos.extend(access.add_raw_records([('key', 5)], 'alpha'))
466
transport = self.get_transport()
467
access = pack_repo._DirectPackAccess({"FOO":(transport, 'packfile'),
468
"FOOBAR":(transport, 'pack2'),
469
"BAZ":(transport, 'pack3')})
470
self.assertEqual(['1234567890', '12345', 'alpha'],
471
list(access.get_raw_records(memos)))
472
self.assertEqual(['1234567890'],
473
list(access.get_raw_records(memos[0:1])))
474
self.assertEqual(['12345'],
475
list(access.get_raw_records(memos[1:2])))
476
self.assertEqual(['alpha'],
477
list(access.get_raw_records(memos[2:3])))
478
self.assertEqual(['1234567890', 'alpha'],
479
list(access.get_raw_records(memos[0:1] + memos[2:3])))
481
def test_set_writer(self):
482
"""The writer should be settable post construction."""
483
access = pack_repo._DirectPackAccess({})
484
transport = self.get_transport()
485
packname = 'packfile'
487
def write_data(bytes):
488
transport.append_bytes(packname, bytes)
489
writer = pack.ContainerWriter(write_data)
491
access.set_writer(writer, index, (transport, packname))
492
memos = access.add_raw_records([('key', 10)], '1234567890')
494
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
496
def test_missing_index_raises_retry(self):
497
memos = self.make_pack_file()
498
transport = self.get_transport()
499
reload_called, reload_func = self.make_reload_func()
500
# Note that the index key has changed from 'foo' to 'bar'
501
access = pack_repo._DirectPackAccess({'bar':(transport, 'packname')},
502
reload_func=reload_func)
503
e = self.assertListRaises(errors.RetryWithNewPacks,
504
access.get_raw_records, memos)
505
# Because a key was passed in which does not match our index list, we
506
# assume that the listing was already reloaded
507
self.assertTrue(e.reload_occurred)
508
self.assertIsInstance(e.exc_info, tuple)
509
self.assertIs(e.exc_info[0], KeyError)
510
self.assertIsInstance(e.exc_info[1], KeyError)
512
def test_missing_index_raises_key_error_with_no_reload(self):
513
memos = self.make_pack_file()
514
transport = self.get_transport()
515
# Note that the index key has changed from 'foo' to 'bar'
516
access = pack_repo._DirectPackAccess({'bar':(transport, 'packname')})
517
e = self.assertListRaises(KeyError, access.get_raw_records, memos)
519
def test_missing_file_raises_retry(self):
520
memos = self.make_pack_file()
521
transport = self.get_transport()
522
reload_called, reload_func = self.make_reload_func()
523
# Note that the 'filename' has been changed to 'different-packname'
524
access = pack_repo._DirectPackAccess(
525
{'foo':(transport, 'different-packname')},
526
reload_func=reload_func)
527
e = self.assertListRaises(errors.RetryWithNewPacks,
528
access.get_raw_records, memos)
529
# The file has gone missing, so we assume we need to reload
530
self.assertFalse(e.reload_occurred)
531
self.assertIsInstance(e.exc_info, tuple)
532
self.assertIs(e.exc_info[0], errors.NoSuchFile)
533
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
534
self.assertEqual('different-packname', e.exc_info[1].path)
536
def test_missing_file_raises_no_such_file_with_no_reload(self):
537
memos = self.make_pack_file()
538
transport = self.get_transport()
539
# Note that the 'filename' has been changed to 'different-packname'
540
access = pack_repo._DirectPackAccess(
541
{'foo': (transport, 'different-packname')})
542
e = self.assertListRaises(errors.NoSuchFile,
543
access.get_raw_records, memos)
545
def test_failing_readv_raises_retry(self):
546
memos = self.make_pack_file()
547
transport = self.get_transport()
548
failing_transport = MockReadvFailingTransport(
549
[transport.get_bytes('packname')])
550
reload_called, reload_func = self.make_reload_func()
551
access = pack_repo._DirectPackAccess(
552
{'foo': (failing_transport, 'packname')},
553
reload_func=reload_func)
554
# Asking for a single record will not trigger the Mock failure
555
self.assertEqual(['1234567890'],
556
list(access.get_raw_records(memos[:1])))
557
self.assertEqual(['12345'],
558
list(access.get_raw_records(memos[1:2])))
559
# A multiple offset readv() will fail mid-way through
560
e = self.assertListRaises(errors.RetryWithNewPacks,
561
access.get_raw_records, memos)
562
# The file has gone missing, so we assume we need to reload
563
self.assertFalse(e.reload_occurred)
564
self.assertIsInstance(e.exc_info, tuple)
565
self.assertIs(e.exc_info[0], errors.NoSuchFile)
566
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
567
self.assertEqual('packname', e.exc_info[1].path)
569
def test_failing_readv_raises_no_such_file_with_no_reload(self):
570
memos = self.make_pack_file()
571
transport = self.get_transport()
572
failing_transport = MockReadvFailingTransport(
573
[transport.get_bytes('packname')])
574
reload_called, reload_func = self.make_reload_func()
575
access = pack_repo._DirectPackAccess(
576
{'foo':(failing_transport, 'packname')})
577
# Asking for a single record will not trigger the Mock failure
578
self.assertEqual(['1234567890'],
579
list(access.get_raw_records(memos[:1])))
580
self.assertEqual(['12345'],
581
list(access.get_raw_records(memos[1:2])))
582
# A multiple offset readv() will fail mid-way through
583
e = self.assertListRaises(errors.NoSuchFile,
584
access.get_raw_records, memos)
586
def test_reload_or_raise_no_reload(self):
587
access = pack_repo._DirectPackAccess({}, reload_func=None)
588
retry_exc = self.make_retry_exception()
589
# Without a reload_func, we will just re-raise the original exception
590
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
592
def test_reload_or_raise_reload_changed(self):
593
reload_called, reload_func = self.make_reload_func(return_val=True)
594
access = pack_repo._DirectPackAccess({}, reload_func=reload_func)
595
retry_exc = self.make_retry_exception()
596
access.reload_or_raise(retry_exc)
597
self.assertEqual([1], reload_called)
598
retry_exc.reload_occurred=True
599
access.reload_or_raise(retry_exc)
600
self.assertEqual([2], reload_called)
602
def test_reload_or_raise_reload_no_change(self):
603
reload_called, reload_func = self.make_reload_func(return_val=False)
604
access = pack_repo._DirectPackAccess({}, reload_func=reload_func)
605
retry_exc = self.make_retry_exception()
606
# If reload_occurred is False, then we consider it an error to have
607
# reload_func() return False (no changes).
608
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
609
self.assertEqual([1], reload_called)
610
retry_exc.reload_occurred=True
611
# If reload_occurred is True, then we assume nothing changed because
612
# it had changed earlier, but didn't change again
613
access.reload_or_raise(retry_exc)
614
self.assertEqual([2], reload_called)
616
def test_annotate_retries(self):
617
vf, reload_counter = self.make_vf_for_retrying()
618
# It is a little bit bogus to annotate the Revision VF, but it works,
619
# as we have ancestry stored there
621
reload_lines = vf.annotate(key)
622
self.assertEqual([1, 1, 0], reload_counter)
623
plain_lines = vf.annotate(key)
624
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
625
if reload_lines != plain_lines:
626
self.fail('Annotation was not identical with reloading.')
627
# Now delete the packs-in-use, which should trigger another reload, but
628
# this time we just raise an exception because we can't recover
629
for trans, name in vf._access._indices.values():
631
self.assertRaises(errors.NoSuchFile, vf.annotate, key)
632
self.assertEqual([2, 1, 1], reload_counter)
634
def test__get_record_map_retries(self):
635
vf, reload_counter = self.make_vf_for_retrying()
636
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
637
records = vf._get_record_map(keys)
638
self.assertEqual(keys, sorted(records.keys()))
639
self.assertEqual([1, 1, 0], reload_counter)
640
# Now delete the packs-in-use, which should trigger another reload, but
641
# this time we just raise an exception because we can't recover
642
for trans, name in vf._access._indices.values():
644
self.assertRaises(errors.NoSuchFile, vf._get_record_map, keys)
645
self.assertEqual([2, 1, 1], reload_counter)
647
def test_get_record_stream_retries(self):
648
vf, reload_counter = self.make_vf_for_retrying()
649
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
650
record_stream = vf.get_record_stream(keys, 'topological', False)
651
record = next(record_stream)
652
self.assertEqual(('rev-1',), record.key)
653
self.assertEqual([0, 0, 0], reload_counter)
654
record = next(record_stream)
655
self.assertEqual(('rev-2',), record.key)
656
self.assertEqual([1, 1, 0], reload_counter)
657
record = next(record_stream)
658
self.assertEqual(('rev-3',), record.key)
659
self.assertEqual([1, 1, 0], reload_counter)
660
# Now delete all pack files, and see that we raise the right error
661
for trans, name in vf._access._indices.values():
663
self.assertListRaises(errors.NoSuchFile,
664
vf.get_record_stream, keys, 'topological', False)
666
def test_iter_lines_added_or_present_in_keys_retries(self):
667
vf, reload_counter = self.make_vf_for_retrying()
668
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
669
# Unfortunately, iter_lines_added_or_present_in_keys iterates the
670
# result in random order (determined by the iteration order from a
671
# set()), so we don't have any solid way to trigger whether data is
672
# read before or after. However we tried to delete the middle node to
673
# exercise the code well.
674
# What we care about is that all lines are always yielded, but not
677
reload_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
678
self.assertEqual([1, 1, 0], reload_counter)
679
# Now do it again, to make sure the result is equivalent
680
plain_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
681
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
682
self.assertEqual(plain_lines, reload_lines)
683
self.assertEqual(21, len(plain_lines))
684
# Now delete all pack files, and see that we raise the right error
685
for trans, name in vf._access._indices.values():
687
self.assertListRaises(errors.NoSuchFile,
688
vf.iter_lines_added_or_present_in_keys, keys)
689
self.assertEqual([2, 1, 1], reload_counter)
691
def test_get_record_stream_yields_disk_sorted_order(self):
692
# if we get 'unordered' pick a semi-optimal order for reading. The
693
# order should be grouped by pack file, and then by position in file
694
repo = self.make_repository('test', format='pack-0.92')
696
self.addCleanup(repo.unlock)
697
repo.start_write_group()
699
vf.add_lines(('f-id', 'rev-5'), [('f-id', 'rev-4')], ['lines\n'])
700
vf.add_lines(('f-id', 'rev-1'), [], ['lines\n'])
701
vf.add_lines(('f-id', 'rev-2'), [('f-id', 'rev-1')], ['lines\n'])
702
repo.commit_write_group()
703
# We inserted them as rev-5, rev-1, rev-2, we should get them back in
705
stream = vf.get_record_stream([('f-id', 'rev-1'), ('f-id', 'rev-5'),
706
('f-id', 'rev-2')], 'unordered', False)
707
keys = [r.key for r in stream]
708
self.assertEqual([('f-id', 'rev-5'), ('f-id', 'rev-1'),
709
('f-id', 'rev-2')], keys)
710
repo.start_write_group()
711
vf.add_lines(('f-id', 'rev-4'), [('f-id', 'rev-3')], ['lines\n'])
712
vf.add_lines(('f-id', 'rev-3'), [('f-id', 'rev-2')], ['lines\n'])
713
vf.add_lines(('f-id', 'rev-6'), [('f-id', 'rev-5')], ['lines\n'])
714
repo.commit_write_group()
715
# Request in random order, to make sure the output order isn't based on
717
request_keys = set(('f-id', 'rev-%d' % i) for i in range(1, 7))
718
stream = vf.get_record_stream(request_keys, 'unordered', False)
719
keys = [r.key for r in stream]
720
# We want to get the keys back in disk order, but it doesn't matter
721
# which pack we read from first. So this can come back in 2 orders
722
alt1 = [('f-id', 'rev-%d' % i) for i in [4, 3, 6, 5, 1, 2]]
723
alt2 = [('f-id', 'rev-%d' % i) for i in [5, 1, 2, 4, 3, 6]]
724
if keys != alt1 and keys != alt2:
725
self.fail('Returned key order did not match either expected order.'
726
' expected %s or %s, not %s'
727
% (alt1, alt2, keys))
730
class LowLevelKnitDataTests(TestCase):
732
def create_gz_content(self, text):
734
gz_file = gzip.GzipFile(mode='wb', fileobj=sio)
737
return sio.getvalue()
739
def make_multiple_records(self):
740
"""Create the content for multiple records."""
741
sha1sum = osutils.sha_string('foo\nbar\n')
743
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
748
record_1 = (0, len(gz_txt), sha1sum)
749
total_txt.append(gz_txt)
750
sha1sum = osutils.sha_string('baz\n')
751
gz_txt = self.create_gz_content('version rev-id-2 1 %s\n'
755
record_2 = (record_1[1], len(gz_txt), sha1sum)
756
total_txt.append(gz_txt)
757
return total_txt, record_1, record_2
759
def test_valid_knit_data(self):
760
sha1sum = osutils.sha_string('foo\nbar\n')
761
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
766
transport = MockTransport([gz_txt])
767
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
768
knit = KnitVersionedFiles(None, access)
769
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
771
contents = list(knit._read_records_iter(records))
772
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'],
773
'4e48e2c9a3d2ca8a708cb0cc545700544efb5021')], contents)
775
raw_contents = list(knit._read_records_iter_raw(records))
776
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
778
def test_multiple_records_valid(self):
779
total_txt, record_1, record_2 = self.make_multiple_records()
780
transport = MockTransport([''.join(total_txt)])
781
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
782
knit = KnitVersionedFiles(None, access)
783
records = [(('rev-id-1',), (('rev-id-1',), record_1[0], record_1[1])),
784
(('rev-id-2',), (('rev-id-2',), record_2[0], record_2[1]))]
786
contents = list(knit._read_records_iter(records))
787
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'], record_1[2]),
788
(('rev-id-2',), ['baz\n'], record_2[2])],
791
raw_contents = list(knit._read_records_iter_raw(records))
792
self.assertEqual([(('rev-id-1',), total_txt[0], record_1[2]),
793
(('rev-id-2',), total_txt[1], record_2[2])],
796
def test_not_enough_lines(self):
797
sha1sum = osutils.sha_string('foo\n')
798
# record says 2 lines data says 1
799
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
803
transport = MockTransport([gz_txt])
804
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
805
knit = KnitVersionedFiles(None, access)
806
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
807
self.assertRaises(errors.KnitCorrupt, list,
808
knit._read_records_iter(records))
810
# read_records_iter_raw won't detect that sort of mismatch/corruption
811
raw_contents = list(knit._read_records_iter_raw(records))
812
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
814
def test_too_many_lines(self):
815
sha1sum = osutils.sha_string('foo\nbar\n')
816
# record says 1 lines data says 2
817
gz_txt = self.create_gz_content('version rev-id-1 1 %s\n'
822
transport = MockTransport([gz_txt])
823
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
824
knit = KnitVersionedFiles(None, access)
825
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
826
self.assertRaises(errors.KnitCorrupt, list,
827
knit._read_records_iter(records))
829
# read_records_iter_raw won't detect that sort of mismatch/corruption
830
raw_contents = list(knit._read_records_iter_raw(records))
831
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
833
def test_mismatched_version_id(self):
834
sha1sum = osutils.sha_string('foo\nbar\n')
835
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
840
transport = MockTransport([gz_txt])
841
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
842
knit = KnitVersionedFiles(None, access)
843
# We are asking for rev-id-2, but the data is rev-id-1
844
records = [(('rev-id-2',), (('rev-id-2',), 0, len(gz_txt)))]
845
self.assertRaises(errors.KnitCorrupt, list,
846
knit._read_records_iter(records))
848
# read_records_iter_raw detects mismatches in the header
849
self.assertRaises(errors.KnitCorrupt, list,
850
knit._read_records_iter_raw(records))
852
def test_uncompressed_data(self):
853
sha1sum = osutils.sha_string('foo\nbar\n')
854
txt = ('version rev-id-1 2 %s\n'
859
transport = MockTransport([txt])
860
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
861
knit = KnitVersionedFiles(None, access)
862
records = [(('rev-id-1',), (('rev-id-1',), 0, len(txt)))]
864
# We don't have valid gzip data ==> corrupt
865
self.assertRaises(errors.KnitCorrupt, list,
866
knit._read_records_iter(records))
868
# read_records_iter_raw will notice the bad data
869
self.assertRaises(errors.KnitCorrupt, list,
870
knit._read_records_iter_raw(records))
872
def test_corrupted_data(self):
873
sha1sum = osutils.sha_string('foo\nbar\n')
874
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
879
# Change 2 bytes in the middle to \xff
880
gz_txt = gz_txt[:10] + '\xff\xff' + gz_txt[12:]
881
transport = MockTransport([gz_txt])
882
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
883
knit = KnitVersionedFiles(None, access)
884
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
885
self.assertRaises(errors.KnitCorrupt, list,
886
knit._read_records_iter(records))
887
# read_records_iter_raw will barf on bad gz data
888
self.assertRaises(errors.KnitCorrupt, list,
889
knit._read_records_iter_raw(records))
892
class LowLevelKnitIndexTests(TestCase):
894
def get_knit_index(self, transport, name, mode):
895
mapper = ConstantMapper(name)
896
from breezy._knit_load_data_py import _load_data_py
897
self.overrideAttr(knit, '_load_data', _load_data_py)
898
allow_writes = lambda: 'w' in mode
899
return _KndxIndex(transport, mapper, lambda:None, allow_writes, lambda:True)
901
def test_create_file(self):
902
transport = MockTransport()
903
index = self.get_knit_index(transport, "filename", "w")
905
call = transport.calls.pop(0)
906
# call[1][1] is a BytesIO - we can't test it by simple equality.
907
self.assertEqual('put_file_non_atomic', call[0])
908
self.assertEqual('filename.kndx', call[1][0])
909
# With no history, _KndxIndex writes a new index:
910
self.assertEqual(_KndxIndex.HEADER,
911
call[1][1].getvalue())
912
self.assertEqual({'create_parent_dir': True}, call[2])
914
def test_read_utf8_version_id(self):
915
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
916
utf8_revision_id = unicode_revision_id.encode('utf-8')
917
transport = MockTransport([
919
'%s option 0 1 :' % (utf8_revision_id,)
921
index = self.get_knit_index(transport, "filename", "r")
922
# _KndxIndex is a private class, and deals in utf8 revision_ids, not
923
# Unicode revision_ids.
924
self.assertEqual({(utf8_revision_id,):()},
925
index.get_parent_map(index.keys()))
926
self.assertFalse((unicode_revision_id,) in index.keys())
928
def test_read_utf8_parents(self):
929
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
930
utf8_revision_id = unicode_revision_id.encode('utf-8')
931
transport = MockTransport([
933
"version option 0 1 .%s :" % (utf8_revision_id,)
935
index = self.get_knit_index(transport, "filename", "r")
936
self.assertEqual({("version",):((utf8_revision_id,),)},
937
index.get_parent_map(index.keys()))
939
def test_read_ignore_corrupted_lines(self):
940
transport = MockTransport([
943
"corrupted options 0 1 .b .c ",
944
"version options 0 1 :"
946
index = self.get_knit_index(transport, "filename", "r")
947
self.assertEqual(1, len(index.keys()))
948
self.assertEqual({("version",)}, index.keys())
950
def test_read_corrupted_header(self):
951
transport = MockTransport(['not a bzr knit index header\n'])
952
index = self.get_knit_index(transport, "filename", "r")
953
self.assertRaises(KnitHeaderError, index.keys)
955
def test_read_duplicate_entries(self):
956
transport = MockTransport([
958
"parent options 0 1 :",
959
"version options1 0 1 0 :",
960
"version options2 1 2 .other :",
961
"version options3 3 4 0 .other :"
963
index = self.get_knit_index(transport, "filename", "r")
964
self.assertEqual(2, len(index.keys()))
965
# check that the index used is the first one written. (Specific
966
# to KnitIndex style indices.
967
self.assertEqual("1", index._dictionary_compress([("version",)]))
968
self.assertEqual((("version",), 3, 4), index.get_position(("version",)))
969
self.assertEqual(["options3"], index.get_options(("version",)))
970
self.assertEqual({("version",):(("parent",), ("other",))},
971
index.get_parent_map([("version",)]))
973
def test_read_compressed_parents(self):
974
transport = MockTransport([
978
"c option 0 1 1 0 :",
980
index = self.get_knit_index(transport, "filename", "r")
981
self.assertEqual({("b",):(("a",),), ("c",):(("b",), ("a",))},
982
index.get_parent_map([("b",), ("c",)]))
984
def test_write_utf8_version_id(self):
985
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
986
utf8_revision_id = unicode_revision_id.encode('utf-8')
987
transport = MockTransport([
990
index = self.get_knit_index(transport, "filename", "r")
992
((utf8_revision_id,), ["option"], ((utf8_revision_id,), 0, 1), [])])
993
call = transport.calls.pop(0)
994
# call[1][1] is a BytesIO - we can't test it by simple equality.
995
self.assertEqual('put_file_non_atomic', call[0])
996
self.assertEqual('filename.kndx', call[1][0])
997
# With no history, _KndxIndex writes a new index:
998
self.assertEqual(_KndxIndex.HEADER +
999
"\n%s option 0 1 :" % (utf8_revision_id,),
1000
call[1][1].getvalue())
1001
self.assertEqual({'create_parent_dir': True}, call[2])
1003
def test_write_utf8_parents(self):
1004
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
1005
utf8_revision_id = unicode_revision_id.encode('utf-8')
1006
transport = MockTransport([
1009
index = self.get_knit_index(transport, "filename", "r")
1011
(("version",), ["option"], (("version",), 0, 1), [(utf8_revision_id,)])])
1012
call = transport.calls.pop(0)
1013
# call[1][1] is a BytesIO - we can't test it by simple equality.
1014
self.assertEqual('put_file_non_atomic', call[0])
1015
self.assertEqual('filename.kndx', call[1][0])
1016
# With no history, _KndxIndex writes a new index:
1017
self.assertEqual(_KndxIndex.HEADER +
1018
"\nversion option 0 1 .%s :" % (utf8_revision_id,),
1019
call[1][1].getvalue())
1020
self.assertEqual({'create_parent_dir': True}, call[2])
1022
def test_keys(self):
1023
transport = MockTransport([
1026
index = self.get_knit_index(transport, "filename", "r")
1028
self.assertEqual(set(), index.keys())
1030
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1031
self.assertEqual({("a",)}, index.keys())
1033
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1034
self.assertEqual({("a",)}, index.keys())
1036
index.add_records([(("b",), ["option"], (("b",), 0, 1), [])])
1037
self.assertEqual({("a",), ("b",)}, index.keys())
1039
def add_a_b(self, index, random_id=None):
1041
if random_id is not None:
1042
kwargs["random_id"] = random_id
1044
(("a",), ["option"], (("a",), 0, 1), [("b",)]),
1045
(("a",), ["opt"], (("a",), 1, 2), [("c",)]),
1046
(("b",), ["option"], (("b",), 2, 3), [("a",)])
1049
def assertIndexIsAB(self, index):
1054
index.get_parent_map(index.keys()))
1055
self.assertEqual((("a",), 1, 2), index.get_position(("a",)))
1056
self.assertEqual((("b",), 2, 3), index.get_position(("b",)))
1057
self.assertEqual(["opt"], index.get_options(("a",)))
1059
def test_add_versions(self):
1060
transport = MockTransport([
1063
index = self.get_knit_index(transport, "filename", "r")
1066
call = transport.calls.pop(0)
1067
# call[1][1] is a BytesIO - we can't test it by simple equality.
1068
self.assertEqual('put_file_non_atomic', call[0])
1069
self.assertEqual('filename.kndx', call[1][0])
1070
# With no history, _KndxIndex writes a new index:
1073
"\na option 0 1 .b :"
1075
"\nb option 2 3 0 :",
1076
call[1][1].getvalue())
1077
self.assertEqual({'create_parent_dir': True}, call[2])
1078
self.assertIndexIsAB(index)
1080
def test_add_versions_random_id_is_accepted(self):
1081
transport = MockTransport([
1084
index = self.get_knit_index(transport, "filename", "r")
1085
self.add_a_b(index, random_id=True)
1087
def test_delay_create_and_add_versions(self):
1088
transport = MockTransport()
1090
index = self.get_knit_index(transport, "filename", "w")
1092
self.assertEqual([], transport.calls)
1095
#[ {"dir_mode": 0777, "create_parent_dir": True, "mode": "wb"},
1097
# Two calls: one during which we load the existing index (and when its
1098
# missing create it), then a second where we write the contents out.
1099
self.assertEqual(2, len(transport.calls))
1100
call = transport.calls.pop(0)
1101
self.assertEqual('put_file_non_atomic', call[0])
1102
self.assertEqual('filename.kndx', call[1][0])
1103
# With no history, _KndxIndex writes a new index:
1104
self.assertEqual(_KndxIndex.HEADER, call[1][1].getvalue())
1105
self.assertEqual({'create_parent_dir': True}, call[2])
1106
call = transport.calls.pop(0)
1107
# call[1][1] is a BytesIO - we can't test it by simple equality.
1108
self.assertEqual('put_file_non_atomic', call[0])
1109
self.assertEqual('filename.kndx', call[1][0])
1110
# With no history, _KndxIndex writes a new index:
1113
"\na option 0 1 .b :"
1115
"\nb option 2 3 0 :",
1116
call[1][1].getvalue())
1117
self.assertEqual({'create_parent_dir': True}, call[2])
1119
def assertTotalBuildSize(self, size, keys, positions):
1120
self.assertEqual(size,
1121
knit._get_total_build_size(None, keys, positions))
1123
def test__get_total_build_size(self):
1125
('a',): (('fulltext', False), (('a',), 0, 100), None),
1126
('b',): (('line-delta', False), (('b',), 100, 21), ('a',)),
1127
('c',): (('line-delta', False), (('c',), 121, 35), ('b',)),
1128
('d',): (('line-delta', False), (('d',), 156, 12), ('b',)),
1130
self.assertTotalBuildSize(100, [('a',)], positions)
1131
self.assertTotalBuildSize(121, [('b',)], positions)
1132
# c needs both a & b
1133
self.assertTotalBuildSize(156, [('c',)], positions)
1134
# we shouldn't count 'b' twice
1135
self.assertTotalBuildSize(156, [('b',), ('c',)], positions)
1136
self.assertTotalBuildSize(133, [('d',)], positions)
1137
self.assertTotalBuildSize(168, [('c',), ('d',)], positions)
1139
def test_get_position(self):
1140
transport = MockTransport([
1145
index = self.get_knit_index(transport, "filename", "r")
1147
self.assertEqual((("a",), 0, 1), index.get_position(("a",)))
1148
self.assertEqual((("b",), 1, 2), index.get_position(("b",)))
1150
def test_get_method(self):
1151
transport = MockTransport([
1153
"a fulltext,unknown 0 1 :",
1154
"b unknown,line-delta 1 2 :",
1157
index = self.get_knit_index(transport, "filename", "r")
1159
self.assertEqual("fulltext", index.get_method("a"))
1160
self.assertEqual("line-delta", index.get_method("b"))
1161
self.assertRaises(errors.KnitIndexUnknownMethod, index.get_method, "c")
1163
def test_get_options(self):
1164
transport = MockTransport([
1169
index = self.get_knit_index(transport, "filename", "r")
1171
self.assertEqual(["opt1"], index.get_options("a"))
1172
self.assertEqual(["opt2", "opt3"], index.get_options("b"))
1174
def test_get_parent_map(self):
1175
transport = MockTransport([
1178
"b option 1 2 0 .c :",
1179
"c option 1 2 1 0 .e :"
1181
index = self.get_knit_index(transport, "filename", "r")
1185
("b",):(("a",), ("c",)),
1186
("c",):(("b",), ("a",), ("e",)),
1187
}, index.get_parent_map(index.keys()))
1189
def test_impossible_parent(self):
1190
"""Test we get KnitCorrupt if the parent couldn't possibly exist."""
1191
transport = MockTransport([
1194
"b option 0 1 4 :" # We don't have a 4th record
1196
index = self.get_knit_index(transport, 'filename', 'r')
1198
self.assertRaises(errors.KnitCorrupt, index.keys)
1199
except TypeError as e:
1200
if (str(e) == ('exceptions must be strings, classes, or instances,'
1201
' not exceptions.IndexError')):
1202
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1203
' raising new style exceptions with python'
1208
def test_corrupted_parent(self):
1209
transport = MockTransport([
1213
"c option 0 1 1v :", # Can't have a parent of '1v'
1215
index = self.get_knit_index(transport, 'filename', 'r')
1217
self.assertRaises(errors.KnitCorrupt, index.keys)
1218
except TypeError as e:
1219
if (str(e) == ('exceptions must be strings, classes, or instances,'
1220
' not exceptions.ValueError')):
1221
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1222
' raising new style exceptions with python'
1227
def test_corrupted_parent_in_list(self):
1228
transport = MockTransport([
1232
"c option 0 1 1 v :", # Can't have a parent of 'v'
1234
index = self.get_knit_index(transport, 'filename', 'r')
1236
self.assertRaises(errors.KnitCorrupt, index.keys)
1237
except TypeError as e:
1238
if (str(e) == ('exceptions must be strings, classes, or instances,'
1239
' not exceptions.ValueError')):
1240
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1241
' raising new style exceptions with python'
1246
def test_invalid_position(self):
1247
transport = MockTransport([
1251
index = self.get_knit_index(transport, 'filename', 'r')
1253
self.assertRaises(errors.KnitCorrupt, index.keys)
1254
except TypeError as e:
1255
if (str(e) == ('exceptions must be strings, classes, or instances,'
1256
' not exceptions.ValueError')):
1257
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1258
' raising new style exceptions with python'
1263
def test_invalid_size(self):
1264
transport = MockTransport([
1268
index = self.get_knit_index(transport, 'filename', 'r')
1270
self.assertRaises(errors.KnitCorrupt, index.keys)
1271
except TypeError as e:
1272
if (str(e) == ('exceptions must be strings, classes, or instances,'
1273
' not exceptions.ValueError')):
1274
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1275
' raising new style exceptions with python'
1280
def test_scan_unvalidated_index_not_implemented(self):
1281
transport = MockTransport()
1282
index = self.get_knit_index(transport, 'filename', 'r')
1284
NotImplementedError, index.scan_unvalidated_index,
1285
'dummy graph_index')
1287
NotImplementedError, index.get_missing_compression_parents)
1289
def test_short_line(self):
1290
transport = MockTransport([
1293
"b option 10 10 0", # This line isn't terminated, ignored
1295
index = self.get_knit_index(transport, "filename", "r")
1296
self.assertEqual({('a',)}, index.keys())
1298
def test_skip_incomplete_record(self):
1299
# A line with bogus data should just be skipped
1300
transport = MockTransport([
1303
"b option 10 10 0", # This line isn't terminated, ignored
1304
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1306
index = self.get_knit_index(transport, "filename", "r")
1307
self.assertEqual({('a',), ('c',)}, index.keys())
1309
def test_trailing_characters(self):
1310
# A line with bogus data should just be skipped
1311
transport = MockTransport([
1314
"b option 10 10 0 :a", # This line has extra trailing characters
1315
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1317
index = self.get_knit_index(transport, "filename", "r")
1318
self.assertEqual({('a',), ('c',)}, index.keys())
1321
class LowLevelKnitIndexTests_c(LowLevelKnitIndexTests):
1323
_test_needs_features = [compiled_knit_feature]
1325
def get_knit_index(self, transport, name, mode):
1326
mapper = ConstantMapper(name)
1327
from breezy._knit_load_data_pyx import _load_data_c
1328
self.overrideAttr(knit, '_load_data', _load_data_c)
1329
allow_writes = lambda: mode == 'w'
1330
return _KndxIndex(transport, mapper, lambda:None,
1331
allow_writes, lambda:True)
1334
class Test_KnitAnnotator(TestCaseWithMemoryTransport):
1336
def make_annotator(self):
1337
factory = knit.make_pack_factory(True, True, 1)
1338
vf = factory(self.get_transport())
1339
return knit._KnitAnnotator(vf)
1341
def test__expand_fulltext(self):
1342
ann = self.make_annotator()
1343
rev_key = ('rev-id',)
1344
ann._num_compression_children[rev_key] = 1
1345
res = ann._expand_record(rev_key, (('parent-id',),), None,
1346
['line1\n', 'line2\n'], ('fulltext', True))
1347
# The content object and text lines should be cached appropriately
1348
self.assertEqual(['line1\n', 'line2'], res)
1349
content_obj = ann._content_objects[rev_key]
1350
self.assertEqual(['line1\n', 'line2\n'], content_obj._lines)
1351
self.assertEqual(res, content_obj.text())
1352
self.assertEqual(res, ann._text_cache[rev_key])
1354
def test__expand_delta_comp_parent_not_available(self):
1355
# Parent isn't available yet, so we return nothing, but queue up this
1356
# node for later processing
1357
ann = self.make_annotator()
1358
rev_key = ('rev-id',)
1359
parent_key = ('parent-id',)
1360
record = ['0,1,1\n', 'new-line\n']
1361
details = ('line-delta', False)
1362
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1364
self.assertEqual(None, res)
1365
self.assertTrue(parent_key in ann._pending_deltas)
1366
pending = ann._pending_deltas[parent_key]
1367
self.assertEqual(1, len(pending))
1368
self.assertEqual((rev_key, (parent_key,), record, details), pending[0])
1370
def test__expand_record_tracks_num_children(self):
1371
ann = self.make_annotator()
1372
rev_key = ('rev-id',)
1373
rev2_key = ('rev2-id',)
1374
parent_key = ('parent-id',)
1375
record = ['0,1,1\n', 'new-line\n']
1376
details = ('line-delta', False)
1377
ann._num_compression_children[parent_key] = 2
1378
ann._expand_record(parent_key, (), None, ['line1\n', 'line2\n'],
1379
('fulltext', False))
1380
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1382
self.assertEqual({parent_key: 1}, ann._num_compression_children)
1383
# Expanding the second child should remove the content object, and the
1384
# num_compression_children entry
1385
res = ann._expand_record(rev2_key, (parent_key,), parent_key,
1387
self.assertFalse(parent_key in ann._content_objects)
1388
self.assertEqual({}, ann._num_compression_children)
1389
# We should not cache the content_objects for rev2 and rev, because
1390
# they do not have compression children of their own.
1391
self.assertEqual({}, ann._content_objects)
1393
def test__expand_delta_records_blocks(self):
1394
ann = self.make_annotator()
1395
rev_key = ('rev-id',)
1396
parent_key = ('parent-id',)
1397
record = ['0,1,1\n', 'new-line\n']
1398
details = ('line-delta', True)
1399
ann._num_compression_children[parent_key] = 2
1400
ann._expand_record(parent_key, (), None,
1401
['line1\n', 'line2\n', 'line3\n'],
1402
('fulltext', False))
1403
ann._expand_record(rev_key, (parent_key,), parent_key, record, details)
1404
self.assertEqual({(rev_key, parent_key): [(1, 1, 1), (3, 3, 0)]},
1405
ann._matching_blocks)
1406
rev2_key = ('rev2-id',)
1407
record = ['0,1,1\n', 'new-line\n']
1408
details = ('line-delta', False)
1409
ann._expand_record(rev2_key, (parent_key,), parent_key, record, details)
1410
self.assertEqual([(1, 1, 2), (3, 3, 0)],
1411
ann._matching_blocks[(rev2_key, parent_key)])
1413
def test__get_parent_ann_uses_matching_blocks(self):
1414
ann = self.make_annotator()
1415
rev_key = ('rev-id',)
1416
parent_key = ('parent-id',)
1417
parent_ann = [(parent_key,)]*3
1418
block_key = (rev_key, parent_key)
1419
ann._annotations_cache[parent_key] = parent_ann
1420
ann._matching_blocks[block_key] = [(0, 1, 1), (3, 3, 0)]
1421
# We should not try to access any parent_lines content, because we know
1422
# we already have the matching blocks
1423
par_ann, blocks = ann._get_parent_annotations_and_matches(rev_key,
1424
['1\n', '2\n', '3\n'], parent_key)
1425
self.assertEqual(parent_ann, par_ann)
1426
self.assertEqual([(0, 1, 1), (3, 3, 0)], blocks)
1427
self.assertEqual({}, ann._matching_blocks)
1429
def test__process_pending(self):
1430
ann = self.make_annotator()
1431
rev_key = ('rev-id',)
1434
record = ['0,1,1\n', 'new-line\n']
1435
details = ('line-delta', False)
1436
p1_record = ['line1\n', 'line2\n']
1437
ann._num_compression_children[p1_key] = 1
1438
res = ann._expand_record(rev_key, (p1_key,p2_key), p1_key,
1440
self.assertEqual(None, res)
1441
# self.assertTrue(p1_key in ann._pending_deltas)
1442
self.assertEqual({}, ann._pending_annotation)
1443
# Now insert p1, and we should be able to expand the delta
1444
res = ann._expand_record(p1_key, (), None, p1_record,
1445
('fulltext', False))
1446
self.assertEqual(p1_record, res)
1447
ann._annotations_cache[p1_key] = [(p1_key,)]*2
1448
res = ann._process_pending(p1_key)
1449
self.assertEqual([], res)
1450
self.assertFalse(p1_key in ann._pending_deltas)
1451
self.assertTrue(p2_key in ann._pending_annotation)
1452
self.assertEqual({p2_key: [(rev_key, (p1_key, p2_key))]},
1453
ann._pending_annotation)
1454
# Now fill in parent 2, and pending annotation should be satisfied
1455
res = ann._expand_record(p2_key, (), None, [], ('fulltext', False))
1456
ann._annotations_cache[p2_key] = []
1457
res = ann._process_pending(p2_key)
1458
self.assertEqual([rev_key], res)
1459
self.assertEqual({}, ann._pending_annotation)
1460
self.assertEqual({}, ann._pending_deltas)
1462
def test_record_delta_removes_basis(self):
1463
ann = self.make_annotator()
1464
ann._expand_record(('parent-id',), (), None,
1465
['line1\n', 'line2\n'], ('fulltext', False))
1466
ann._num_compression_children['parent-id'] = 2
1468
def test_annotate_special_text(self):
1469
ann = self.make_annotator()
1471
rev1_key = ('rev-1',)
1472
rev2_key = ('rev-2',)
1473
rev3_key = ('rev-3',)
1474
spec_key = ('special:',)
1475
vf.add_lines(rev1_key, [], ['initial content\n'])
1476
vf.add_lines(rev2_key, [rev1_key], ['initial content\n',
1479
vf.add_lines(rev3_key, [rev1_key], ['initial content\n',
1482
spec_text = ('initial content\n'
1486
ann.add_special_text(spec_key, [rev2_key, rev3_key], spec_text)
1487
anns, lines = ann.annotate(spec_key)
1488
self.assertEqual([(rev1_key,),
1489
(rev2_key, rev3_key),
1493
self.assertEqualDiff(spec_text, ''.join(lines))
1496
class KnitTests(TestCaseWithTransport):
1497
"""Class containing knit test helper routines."""
1499
def make_test_knit(self, annotate=False, name='test'):
1500
mapper = ConstantMapper(name)
1501
return make_file_factory(annotate, mapper)(self.get_transport())
1504
class TestBadShaError(KnitTests):
1505
"""Tests for handling of sha errors."""
1507
def test_sha_exception_has_text(self):
1508
# having the failed text included in the error allows for recovery.
1509
source = self.make_test_knit()
1510
target = self.make_test_knit(name="target")
1511
if not source._max_delta_chain:
1512
raise TestNotApplicable(
1513
"cannot get delta-caused sha failures without deltas.")
1516
broken = ('broken',)
1517
source.add_lines(basis, (), ['foo\n'])
1518
source.add_lines(broken, (basis,), ['foo\n', 'bar\n'])
1519
# Seed target with a bad basis text
1520
target.add_lines(basis, (), ['gam\n'])
1521
target.insert_record_stream(
1522
source.get_record_stream([broken], 'unordered', False))
1523
err = self.assertRaises(errors.KnitCorrupt,
1524
target.get_record_stream([broken], 'unordered', True
1525
).next().get_bytes_as, 'chunked')
1526
self.assertEqual(['gam\n', 'bar\n'], err.content)
1527
# Test for formatting with live data
1528
self.assertStartsWith(str(err), "Knit ")
1531
class TestKnitIndex(KnitTests):
1533
def test_add_versions_dictionary_compresses(self):
1534
"""Adding versions to the index should update the lookup dict"""
1535
knit = self.make_test_knit()
1537
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1538
self.check_file_contents('test.kndx',
1539
'# bzr knit index 8\n'
1541
'a-1 fulltext 0 0 :'
1544
(('a-2',), ['fulltext'], (('a-2',), 0, 0), [('a-1',)]),
1545
(('a-3',), ['fulltext'], (('a-3',), 0, 0), [('a-2',)]),
1547
self.check_file_contents('test.kndx',
1548
'# bzr knit index 8\n'
1550
'a-1 fulltext 0 0 :\n'
1551
'a-2 fulltext 0 0 0 :\n'
1552
'a-3 fulltext 0 0 1 :'
1554
self.assertEqual({('a-3',), ('a-1',), ('a-2',)}, idx.keys())
1556
('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False)),
1557
('a-2',): ((('a-2',), 0, 0), None, (('a-1',),), ('fulltext', False)),
1558
('a-3',): ((('a-3',), 0, 0), None, (('a-2',),), ('fulltext', False)),
1559
}, idx.get_build_details(idx.keys()))
1560
self.assertEqual({('a-1',):(),
1561
('a-2',):(('a-1',),),
1562
('a-3',):(('a-2',),),},
1563
idx.get_parent_map(idx.keys()))
1565
def test_add_versions_fails_clean(self):
1566
"""If add_versions fails in the middle, it restores a pristine state.
1568
Any modifications that are made to the index are reset if all versions
1571
# This cheats a little bit by passing in a generator which will
1572
# raise an exception before the processing finishes
1573
# Other possibilities would be to have an version with the wrong number
1574
# of entries, or to make the backing transport unable to write any
1577
knit = self.make_test_knit()
1579
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1581
class StopEarly(Exception):
1584
def generate_failure():
1585
"""Add some entries and then raise an exception"""
1586
yield (('a-2',), ['fulltext'], (None, 0, 0), ('a-1',))
1587
yield (('a-3',), ['fulltext'], (None, 0, 0), ('a-2',))
1590
# Assert the pre-condition
1592
self.assertEqual({('a-1',)}, set(idx.keys()))
1594
{('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False))},
1595
idx.get_build_details([('a-1',)]))
1596
self.assertEqual({('a-1',):()}, idx.get_parent_map(idx.keys()))
1599
self.assertRaises(StopEarly, idx.add_records, generate_failure())
1600
# And it shouldn't be modified
1603
def test_knit_index_ignores_empty_files(self):
1604
# There was a race condition in older bzr, where a ^C at the right time
1605
# could leave an empty .kndx file, which bzr would later claim was a
1606
# corrupted file since the header was not present. In reality, the file
1607
# just wasn't created, so it should be ignored.
1608
t = transport.get_transport_from_path('.')
1609
t.put_bytes('test.kndx', '')
1611
knit = self.make_test_knit()
1613
def test_knit_index_checks_header(self):
1614
t = transport.get_transport_from_path('.')
1615
t.put_bytes('test.kndx', '# not really a knit header\n\n')
1616
k = self.make_test_knit()
1617
self.assertRaises(KnitHeaderError, k.keys)
1620
class TestGraphIndexKnit(KnitTests):
1621
"""Tests for knits using a GraphIndex rather than a KnitIndex."""
1623
def make_g_index(self, name, ref_lists=0, nodes=[]):
1624
builder = GraphIndexBuilder(ref_lists)
1625
for node, references, value in nodes:
1626
builder.add_node(node, references, value)
1627
stream = builder.finish()
1628
trans = self.get_transport()
1629
size = trans.put_file(name, stream)
1630
return GraphIndex(trans, name, size)
1632
def two_graph_index(self, deltas=False, catch_adds=False):
1633
"""Build a two-graph index.
1635
:param deltas: If true, use underlying indices with two node-ref
1636
lists and 'parent' set to a delta-compressed against tail.
1638
# build a complex graph across several indices.
1640
# delta compression inn the index
1641
index1 = self.make_g_index('1', 2, [
1642
(('tip', ), 'N0 100', ([('parent', )], [], )),
1643
(('tail', ), '', ([], []))])
1644
index2 = self.make_g_index('2', 2, [
1645
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], [('tail', )])),
1646
(('separate', ), '', ([], []))])
1648
# just blob location and graph in the index.
1649
index1 = self.make_g_index('1', 1, [
1650
(('tip', ), 'N0 100', ([('parent', )], )),
1651
(('tail', ), '', ([], ))])
1652
index2 = self.make_g_index('2', 1, [
1653
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], )),
1654
(('separate', ), '', ([], ))])
1655
combined_index = CombinedGraphIndex([index1, index2])
1657
self.combined_index = combined_index
1658
self.caught_entries = []
1659
add_callback = self.catch_add
1662
return _KnitGraphIndex(combined_index, lambda:True, deltas=deltas,
1663
add_callback=add_callback)
1665
def test_keys(self):
1666
index = self.two_graph_index()
1667
self.assertEqual({('tail',), ('tip',), ('parent',), ('separate',)},
1670
def test_get_position(self):
1671
index = self.two_graph_index()
1672
self.assertEqual((index._graph_index._indices[0], 0, 100), index.get_position(('tip',)))
1673
self.assertEqual((index._graph_index._indices[1], 100, 78), index.get_position(('parent',)))
1675
def test_get_method_deltas(self):
1676
index = self.two_graph_index(deltas=True)
1677
self.assertEqual('fulltext', index.get_method(('tip',)))
1678
self.assertEqual('line-delta', index.get_method(('parent',)))
1680
def test_get_method_no_deltas(self):
1681
# check that the parent-history lookup is ignored with deltas=False.
1682
index = self.two_graph_index(deltas=False)
1683
self.assertEqual('fulltext', index.get_method(('tip',)))
1684
self.assertEqual('fulltext', index.get_method(('parent',)))
1686
def test_get_options_deltas(self):
1687
index = self.two_graph_index(deltas=True)
1688
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1689
self.assertEqual(['line-delta'], index.get_options(('parent',)))
1691
def test_get_options_no_deltas(self):
1692
# check that the parent-history lookup is ignored with deltas=False.
1693
index = self.two_graph_index(deltas=False)
1694
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1695
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1697
def test_get_parent_map(self):
1698
index = self.two_graph_index()
1699
self.assertEqual({('parent',):(('tail',), ('ghost',))},
1700
index.get_parent_map([('parent',), ('ghost',)]))
1702
def catch_add(self, entries):
1703
self.caught_entries.append(entries)
1705
def test_add_no_callback_errors(self):
1706
index = self.two_graph_index()
1707
self.assertRaises(errors.ReadOnlyError, index.add_records,
1708
[(('new',), 'fulltext,no-eol', (None, 50, 60), ['separate'])])
1710
def test_add_version_smoke(self):
1711
index = self.two_graph_index(catch_adds=True)
1712
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60),
1714
self.assertEqual([[(('new', ), 'N50 60', ((('separate',),),))]],
1715
self.caught_entries)
1717
def test_add_version_delta_not_delta_index(self):
1718
index = self.two_graph_index(catch_adds=True)
1719
self.assertRaises(errors.KnitCorrupt, index.add_records,
1720
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1721
self.assertEqual([], self.caught_entries)
1723
def test_add_version_same_dup(self):
1724
index = self.two_graph_index(catch_adds=True)
1725
# options can be spelt two different ways
1726
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
1727
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
1728
# position/length are ignored (because each pack could have fulltext or
1729
# delta, and be at a different position.
1730
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1732
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1734
# but neither should have added data:
1735
self.assertEqual([[], [], [], []], self.caught_entries)
1737
def test_add_version_different_dup(self):
1738
index = self.two_graph_index(deltas=True, catch_adds=True)
1740
self.assertRaises(errors.KnitCorrupt, index.add_records,
1741
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1742
self.assertRaises(errors.KnitCorrupt, index.add_records,
1743
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1745
self.assertRaises(errors.KnitCorrupt, index.add_records,
1746
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1747
self.assertEqual([], self.caught_entries)
1749
def test_add_versions_nodeltas(self):
1750
index = self.two_graph_index(catch_adds=True)
1752
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1753
(('new2',), 'fulltext', (None, 0, 6), [('new',)]),
1755
self.assertEqual([(('new', ), 'N50 60', ((('separate',),),)),
1756
(('new2', ), ' 0 6', ((('new',),),))],
1757
sorted(self.caught_entries[0]))
1758
self.assertEqual(1, len(self.caught_entries))
1760
def test_add_versions_deltas(self):
1761
index = self.two_graph_index(deltas=True, catch_adds=True)
1763
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1764
(('new2',), 'line-delta', (None, 0, 6), [('new',)]),
1766
self.assertEqual([(('new', ), 'N50 60', ((('separate',),), ())),
1767
(('new2', ), ' 0 6', ((('new',),), (('new',),), ))],
1768
sorted(self.caught_entries[0]))
1769
self.assertEqual(1, len(self.caught_entries))
1771
def test_add_versions_delta_not_delta_index(self):
1772
index = self.two_graph_index(catch_adds=True)
1773
self.assertRaises(errors.KnitCorrupt, index.add_records,
1774
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1775
self.assertEqual([], self.caught_entries)
1777
def test_add_versions_random_id_accepted(self):
1778
index = self.two_graph_index(catch_adds=True)
1779
index.add_records([], random_id=True)
1781
def test_add_versions_same_dup(self):
1782
index = self.two_graph_index(catch_adds=True)
1783
# options can be spelt two different ways
1784
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100),
1786
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100),
1788
# position/length are ignored (because each pack could have fulltext or
1789
# delta, and be at a different position.
1790
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1792
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1794
# but neither should have added data.
1795
self.assertEqual([[], [], [], []], self.caught_entries)
1797
def test_add_versions_different_dup(self):
1798
index = self.two_graph_index(deltas=True, catch_adds=True)
1800
self.assertRaises(errors.KnitCorrupt, index.add_records,
1801
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1802
self.assertRaises(errors.KnitCorrupt, index.add_records,
1803
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1805
self.assertRaises(errors.KnitCorrupt, index.add_records,
1806
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1807
# change options in the second record
1808
self.assertRaises(errors.KnitCorrupt, index.add_records,
1809
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)]),
1810
(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1811
self.assertEqual([], self.caught_entries)
1813
def make_g_index_missing_compression_parent(self):
1814
graph_index = self.make_g_index('missing_comp', 2,
1815
[(('tip', ), ' 100 78',
1816
([('missing-parent', ), ('ghost', )], [('missing-parent', )]))])
1819
def make_g_index_missing_parent(self):
1820
graph_index = self.make_g_index('missing_parent', 2,
1821
[(('parent', ), ' 100 78', ([], [])),
1822
(('tip', ), ' 100 78',
1823
([('parent', ), ('missing-parent', )], [('parent', )])),
1827
def make_g_index_no_external_refs(self):
1828
graph_index = self.make_g_index('no_external_refs', 2,
1829
[(('rev', ), ' 100 78',
1830
([('parent', ), ('ghost', )], []))])
1833
def test_add_good_unvalidated_index(self):
1834
unvalidated = self.make_g_index_no_external_refs()
1835
combined = CombinedGraphIndex([unvalidated])
1836
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1837
index.scan_unvalidated_index(unvalidated)
1838
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1840
def test_add_missing_compression_parent_unvalidated_index(self):
1841
unvalidated = self.make_g_index_missing_compression_parent()
1842
combined = CombinedGraphIndex([unvalidated])
1843
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1844
index.scan_unvalidated_index(unvalidated)
1845
# This also checks that its only the compression parent that is
1846
# examined, otherwise 'ghost' would also be reported as a missing
1849
frozenset([('missing-parent',)]),
1850
index.get_missing_compression_parents())
1852
def test_add_missing_noncompression_parent_unvalidated_index(self):
1853
unvalidated = self.make_g_index_missing_parent()
1854
combined = CombinedGraphIndex([unvalidated])
1855
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1856
track_external_parent_refs=True)
1857
index.scan_unvalidated_index(unvalidated)
1859
frozenset([('missing-parent',)]), index.get_missing_parents())
1861
def test_track_external_parent_refs(self):
1862
g_index = self.make_g_index('empty', 2, [])
1863
combined = CombinedGraphIndex([g_index])
1864
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1865
add_callback=self.catch_add, track_external_parent_refs=True)
1866
self.caught_entries = []
1868
(('new-key',), 'fulltext,no-eol', (None, 50, 60),
1869
[('parent-1',), ('parent-2',)])])
1871
frozenset([('parent-1',), ('parent-2',)]),
1872
index.get_missing_parents())
1874
def test_add_unvalidated_index_with_present_external_references(self):
1875
index = self.two_graph_index(deltas=True)
1876
# Ugly hack to get at one of the underlying GraphIndex objects that
1877
# two_graph_index built.
1878
unvalidated = index._graph_index._indices[1]
1879
# 'parent' is an external ref of _indices[1] (unvalidated), but is
1880
# present in _indices[0].
1881
index.scan_unvalidated_index(unvalidated)
1882
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1884
def make_new_missing_parent_g_index(self, name):
1885
missing_parent = name + '-missing-parent'
1886
graph_index = self.make_g_index(name, 2,
1887
[((name + 'tip', ), ' 100 78',
1888
([(missing_parent, ), ('ghost', )], [(missing_parent, )]))])
1891
def test_add_mulitiple_unvalidated_indices_with_missing_parents(self):
1892
g_index_1 = self.make_new_missing_parent_g_index('one')
1893
g_index_2 = self.make_new_missing_parent_g_index('two')
1894
combined = CombinedGraphIndex([g_index_1, g_index_2])
1895
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1896
index.scan_unvalidated_index(g_index_1)
1897
index.scan_unvalidated_index(g_index_2)
1899
frozenset([('one-missing-parent',), ('two-missing-parent',)]),
1900
index.get_missing_compression_parents())
1902
def test_add_mulitiple_unvalidated_indices_with_mutual_dependencies(self):
1903
graph_index_a = self.make_g_index('one', 2,
1904
[(('parent-one', ), ' 100 78', ([('non-compression-parent',)], [])),
1905
(('child-of-two', ), ' 100 78',
1906
([('parent-two',)], [('parent-two',)]))])
1907
graph_index_b = self.make_g_index('two', 2,
1908
[(('parent-two', ), ' 100 78', ([('non-compression-parent',)], [])),
1909
(('child-of-one', ), ' 100 78',
1910
([('parent-one',)], [('parent-one',)]))])
1911
combined = CombinedGraphIndex([graph_index_a, graph_index_b])
1912
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1913
index.scan_unvalidated_index(graph_index_a)
1914
index.scan_unvalidated_index(graph_index_b)
1916
frozenset([]), index.get_missing_compression_parents())
1919
class TestNoParentsGraphIndexKnit(KnitTests):
1920
"""Tests for knits using _KnitGraphIndex with no parents."""
1922
def make_g_index(self, name, ref_lists=0, nodes=[]):
1923
builder = GraphIndexBuilder(ref_lists)
1924
for node, references in nodes:
1925
builder.add_node(node, references)
1926
stream = builder.finish()
1927
trans = self.get_transport()
1928
size = trans.put_file(name, stream)
1929
return GraphIndex(trans, name, size)
1931
def test_add_good_unvalidated_index(self):
1932
unvalidated = self.make_g_index('unvalidated')
1933
combined = CombinedGraphIndex([unvalidated])
1934
index = _KnitGraphIndex(combined, lambda: True, parents=False)
1935
index.scan_unvalidated_index(unvalidated)
1936
self.assertEqual(frozenset(),
1937
index.get_missing_compression_parents())
1939
def test_parents_deltas_incompatible(self):
1940
index = CombinedGraphIndex([])
1941
self.assertRaises(errors.KnitError, _KnitGraphIndex, lambda:True,
1942
index, deltas=True, parents=False)
1944
def two_graph_index(self, catch_adds=False):
1945
"""Build a two-graph index.
1947
:param deltas: If true, use underlying indices with two node-ref
1948
lists and 'parent' set to a delta-compressed against tail.
1950
# put several versions in the index.
1951
index1 = self.make_g_index('1', 0, [
1952
(('tip', ), 'N0 100'),
1954
index2 = self.make_g_index('2', 0, [
1955
(('parent', ), ' 100 78'),
1956
(('separate', ), '')])
1957
combined_index = CombinedGraphIndex([index1, index2])
1959
self.combined_index = combined_index
1960
self.caught_entries = []
1961
add_callback = self.catch_add
1964
return _KnitGraphIndex(combined_index, lambda:True, parents=False,
1965
add_callback=add_callback)
1967
def test_keys(self):
1968
index = self.two_graph_index()
1969
self.assertEqual({('tail',), ('tip',), ('parent',), ('separate',)},
1972
def test_get_position(self):
1973
index = self.two_graph_index()
1974
self.assertEqual((index._graph_index._indices[0], 0, 100),
1975
index.get_position(('tip',)))
1976
self.assertEqual((index._graph_index._indices[1], 100, 78),
1977
index.get_position(('parent',)))
1979
def test_get_method(self):
1980
index = self.two_graph_index()
1981
self.assertEqual('fulltext', index.get_method(('tip',)))
1982
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1984
def test_get_options(self):
1985
index = self.two_graph_index()
1986
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1987
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1989
def test_get_parent_map(self):
1990
index = self.two_graph_index()
1991
self.assertEqual({('parent',):None},
1992
index.get_parent_map([('parent',), ('ghost',)]))
1994
def catch_add(self, entries):
1995
self.caught_entries.append(entries)
1997
def test_add_no_callback_errors(self):
1998
index = self.two_graph_index()
1999
self.assertRaises(errors.ReadOnlyError, index.add_records,
2000
[(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)])])
2002
def test_add_version_smoke(self):
2003
index = self.two_graph_index(catch_adds=True)
2004
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60), [])])
2005
self.assertEqual([[(('new', ), 'N50 60')]],
2006
self.caught_entries)
2008
def test_add_version_delta_not_delta_index(self):
2009
index = self.two_graph_index(catch_adds=True)
2010
self.assertRaises(errors.KnitCorrupt, index.add_records,
2011
[(('new',), 'no-eol,line-delta', (None, 0, 100), [])])
2012
self.assertEqual([], self.caught_entries)
2014
def test_add_version_same_dup(self):
2015
index = self.two_graph_index(catch_adds=True)
2016
# options can be spelt two different ways
2017
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
2018
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
2019
# position/length are ignored (because each pack could have fulltext or
2020
# delta, and be at a different position.
2021
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
2022
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
2023
# but neither should have added data.
2024
self.assertEqual([[], [], [], []], self.caught_entries)
2026
def test_add_version_different_dup(self):
2027
index = self.two_graph_index(catch_adds=True)
2029
self.assertRaises(errors.KnitCorrupt, index.add_records,
2030
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2031
self.assertRaises(errors.KnitCorrupt, index.add_records,
2032
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
2033
self.assertRaises(errors.KnitCorrupt, index.add_records,
2034
[(('tip',), 'fulltext', (None, 0, 100), [])])
2036
self.assertRaises(errors.KnitCorrupt, index.add_records,
2037
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
2038
self.assertEqual([], self.caught_entries)
2040
def test_add_versions(self):
2041
index = self.two_graph_index(catch_adds=True)
2043
(('new',), 'fulltext,no-eol', (None, 50, 60), []),
2044
(('new2',), 'fulltext', (None, 0, 6), []),
2046
self.assertEqual([(('new', ), 'N50 60'), (('new2', ), ' 0 6')],
2047
sorted(self.caught_entries[0]))
2048
self.assertEqual(1, len(self.caught_entries))
2050
def test_add_versions_delta_not_delta_index(self):
2051
index = self.two_graph_index(catch_adds=True)
2052
self.assertRaises(errors.KnitCorrupt, index.add_records,
2053
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
2054
self.assertEqual([], self.caught_entries)
2056
def test_add_versions_parents_not_parents_index(self):
2057
index = self.two_graph_index(catch_adds=True)
2058
self.assertRaises(errors.KnitCorrupt, index.add_records,
2059
[(('new',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
2060
self.assertEqual([], self.caught_entries)
2062
def test_add_versions_random_id_accepted(self):
2063
index = self.two_graph_index(catch_adds=True)
2064
index.add_records([], random_id=True)
2066
def test_add_versions_same_dup(self):
2067
index = self.two_graph_index(catch_adds=True)
2068
# options can be spelt two different ways
2069
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
2070
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
2071
# position/length are ignored (because each pack could have fulltext or
2072
# delta, and be at a different position.
2073
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
2074
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
2075
# but neither should have added data.
2076
self.assertEqual([[], [], [], []], self.caught_entries)
2078
def test_add_versions_different_dup(self):
2079
index = self.two_graph_index(catch_adds=True)
2081
self.assertRaises(errors.KnitCorrupt, index.add_records,
2082
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2083
self.assertRaises(errors.KnitCorrupt, index.add_records,
2084
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
2085
self.assertRaises(errors.KnitCorrupt, index.add_records,
2086
[(('tip',), 'fulltext', (None, 0, 100), [])])
2088
self.assertRaises(errors.KnitCorrupt, index.add_records,
2089
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
2090
# change options in the second record
2091
self.assertRaises(errors.KnitCorrupt, index.add_records,
2092
[(('tip',), 'fulltext,no-eol', (None, 0, 100), []),
2093
(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2094
self.assertEqual([], self.caught_entries)
2097
class TestKnitVersionedFiles(KnitTests):
2099
def assertGroupKeysForIo(self, exp_groups, keys, non_local_keys,
2100
positions, _min_buffer_size=None):
2101
kvf = self.make_test_knit()
2102
if _min_buffer_size is None:
2103
_min_buffer_size = knit._STREAM_MIN_BUFFER_SIZE
2104
self.assertEqual(exp_groups, kvf._group_keys_for_io(keys,
2105
non_local_keys, positions,
2106
_min_buffer_size=_min_buffer_size))
2108
def assertSplitByPrefix(self, expected_map, expected_prefix_order,
2110
split, prefix_order = KnitVersionedFiles._split_by_prefix(keys)
2111
self.assertEqual(expected_map, split)
2112
self.assertEqual(expected_prefix_order, prefix_order)
2114
def test__group_keys_for_io(self):
2115
ft_detail = ('fulltext', False)
2116
ld_detail = ('line-delta', False)
2124
f_a: (ft_detail, (f_a, 0, 100), None),
2125
f_b: (ld_detail, (f_b, 100, 21), f_a),
2126
f_c: (ld_detail, (f_c, 180, 15), f_b),
2127
g_a: (ft_detail, (g_a, 121, 35), None),
2128
g_b: (ld_detail, (g_b, 156, 12), g_a),
2129
g_c: (ld_detail, (g_c, 195, 13), g_a),
2131
self.assertGroupKeysForIo([([f_a], set())],
2132
[f_a], [], positions)
2133
self.assertGroupKeysForIo([([f_a], {f_a})],
2134
[f_a], [f_a], positions)
2135
self.assertGroupKeysForIo([([f_a, f_b], set([]))],
2136
[f_a, f_b], [], positions)
2137
self.assertGroupKeysForIo([([f_a, f_b], {f_b})],
2138
[f_a, f_b], [f_b], positions)
2139
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2140
[f_a, g_a, f_b, g_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
_min_buffer_size=150)
2144
self.assertGroupKeysForIo([([f_a, f_b], set()), ([g_a, g_b], set())],
2145
[f_a, g_a, f_b, g_b], [], positions,
2146
_min_buffer_size=100)
2147
self.assertGroupKeysForIo([([f_c], set()), ([g_b], set())],
2148
[f_c, g_b], [], positions,
2149
_min_buffer_size=125)
2150
self.assertGroupKeysForIo([([g_b, f_c], set())],
2151
[g_b, f_c], [], positions,
2152
_min_buffer_size=125)
2154
def test__split_by_prefix(self):
2155
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2156
'g': [('g', 'b'), ('g', 'a')],
2158
[('f', 'a'), ('g', 'b'),
2159
('g', 'a'), ('f', 'b')])
2161
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2162
'g': [('g', 'b'), ('g', 'a')],
2164
[('f', 'a'), ('f', 'b'),
2165
('g', 'b'), ('g', 'a')])
2167
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2168
'g': [('g', 'b'), ('g', 'a')],
2170
[('f', 'a'), ('f', 'b'),
2171
('g', 'b'), ('g', 'a')])
2173
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2174
'g': [('g', 'b'), ('g', 'a')],
2175
'': [('a',), ('b',)]
2177
[('f', 'a'), ('g', 'b'),
2179
('g', 'a'), ('f', 'b')])
2182
class TestStacking(KnitTests):
2184
def get_basis_and_test_knit(self):
2185
basis = self.make_test_knit(name='basis')
2186
basis = RecordingVersionedFilesDecorator(basis)
2187
test = self.make_test_knit(name='test')
2188
test.add_fallback_versioned_files(basis)
2191
def test_add_fallback_versioned_files(self):
2192
basis = self.make_test_knit(name='basis')
2193
test = self.make_test_knit(name='test')
2194
# It must not error; other tests test that the fallback is referred to
2195
# when accessing data.
2196
test.add_fallback_versioned_files(basis)
2198
def test_add_lines(self):
2199
# lines added to the test are not added to the basis
2200
basis, test = self.get_basis_and_test_knit()
2202
key_basis = ('bar',)
2203
key_cross_border = ('quux',)
2204
key_delta = ('zaphod',)
2205
test.add_lines(key, (), ['foo\n'])
2206
self.assertEqual({}, basis.get_parent_map([key]))
2207
# lines added to the test that reference across the stack do a
2209
basis.add_lines(key_basis, (), ['foo\n'])
2211
test.add_lines(key_cross_border, (key_basis,), ['foo\n'])
2212
self.assertEqual('fulltext', test._index.get_method(key_cross_border))
2213
# we don't even need to look at the basis to see that this should be
2214
# stored as a fulltext
2215
self.assertEqual([], basis.calls)
2216
# Subsequent adds do delta.
2218
test.add_lines(key_delta, (key_cross_border,), ['foo\n'])
2219
self.assertEqual('line-delta', test._index.get_method(key_delta))
2220
self.assertEqual([], basis.calls)
2222
def test_annotate(self):
2223
# annotations from the test knit are answered without asking the basis
2224
basis, test = self.get_basis_and_test_knit()
2226
key_basis = ('bar',)
2227
key_missing = ('missing',)
2228
test.add_lines(key, (), ['foo\n'])
2229
details = test.annotate(key)
2230
self.assertEqual([(key, 'foo\n')], details)
2231
self.assertEqual([], basis.calls)
2232
# But texts that are not in the test knit are looked for in the basis
2234
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2236
details = test.annotate(key_basis)
2237
self.assertEqual([(key_basis, 'foo\n'), (key_basis, 'bar\n')], details)
2238
# Not optimised to date:
2239
# self.assertEqual([("annotate", key_basis)], basis.calls)
2240
self.assertEqual([('get_parent_map', {key_basis}),
2241
('get_parent_map', {key_basis}),
2242
('get_record_stream', [key_basis], 'topological', True)],
2245
def test_check(self):
2246
# At the moment checking a stacked knit does implicitly check the
2248
basis, test = self.get_basis_and_test_knit()
2251
def test_get_parent_map(self):
2252
# parents in the test knit are answered without asking the basis
2253
basis, test = self.get_basis_and_test_knit()
2255
key_basis = ('bar',)
2256
key_missing = ('missing',)
2257
test.add_lines(key, (), [])
2258
parent_map = test.get_parent_map([key])
2259
self.assertEqual({key: ()}, parent_map)
2260
self.assertEqual([], basis.calls)
2261
# But parents that are not in the test knit are looked for in the basis
2262
basis.add_lines(key_basis, (), [])
2264
parent_map = test.get_parent_map([key, key_basis, key_missing])
2265
self.assertEqual({key: (),
2266
key_basis: ()}, parent_map)
2267
self.assertEqual([("get_parent_map", {key_basis, key_missing})],
2270
def test_get_record_stream_unordered_fulltexts(self):
2271
# records from the test knit are answered without asking the basis:
2272
basis, test = self.get_basis_and_test_knit()
2274
key_basis = ('bar',)
2275
key_missing = ('missing',)
2276
test.add_lines(key, (), ['foo\n'])
2277
records = list(test.get_record_stream([key], 'unordered', True))
2278
self.assertEqual(1, len(records))
2279
self.assertEqual([], basis.calls)
2280
# Missing (from test knit) objects are retrieved from the basis:
2281
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2283
records = list(test.get_record_stream([key_basis, key_missing],
2285
self.assertEqual(2, len(records))
2286
calls = list(basis.calls)
2287
for record in records:
2288
self.assertSubset([record.key], (key_basis, key_missing))
2289
if record.key == key_missing:
2290
self.assertIsInstance(record, AbsentContentFactory)
2292
reference = list(basis.get_record_stream([key_basis],
2293
'unordered', True))[0]
2294
self.assertEqual(reference.key, record.key)
2295
self.assertEqual(reference.sha1, record.sha1)
2296
self.assertEqual(reference.storage_kind, record.storage_kind)
2297
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2298
record.get_bytes_as(record.storage_kind))
2299
self.assertEqual(reference.get_bytes_as('fulltext'),
2300
record.get_bytes_as('fulltext'))
2301
# It's not strictly minimal, but it seems reasonable for now for it to
2302
# ask which fallbacks have which parents.
2304
("get_parent_map", {key_basis, key_missing}),
2305
("get_record_stream", [key_basis], 'unordered', True)],
2308
def test_get_record_stream_ordered_fulltexts(self):
2309
# ordering is preserved down into the fallback store.
2310
basis, test = self.get_basis_and_test_knit()
2312
key_basis = ('bar',)
2313
key_basis_2 = ('quux',)
2314
key_missing = ('missing',)
2315
test.add_lines(key, (key_basis,), ['foo\n'])
2316
# Missing (from test knit) objects are retrieved from the basis:
2317
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2318
basis.add_lines(key_basis_2, (), ['quux\n'])
2320
# ask for in non-topological order
2321
records = list(test.get_record_stream(
2322
[key, key_basis, key_missing, key_basis_2], 'topological', True))
2323
self.assertEqual(4, len(records))
2325
for record in records:
2326
self.assertSubset([record.key],
2327
(key_basis, key_missing, key_basis_2, key))
2328
if record.key == key_missing:
2329
self.assertIsInstance(record, AbsentContentFactory)
2331
results.append((record.key, record.sha1, record.storage_kind,
2332
record.get_bytes_as('fulltext')))
2333
calls = list(basis.calls)
2334
order = [record[0] for record in results]
2335
self.assertEqual([key_basis_2, key_basis, key], order)
2336
for result in results:
2337
if result[0] == key:
2341
record = next(source.get_record_stream([result[0]], 'unordered',
2343
self.assertEqual(record.key, result[0])
2344
self.assertEqual(record.sha1, result[1])
2345
# We used to check that the storage kind matched, but actually it
2346
# depends on whether it was sourced from the basis, or in a single
2347
# group, because asking for full texts returns proxy objects to a
2348
# _ContentMapGenerator object; so checking the kind is unneeded.
2349
self.assertEqual(record.get_bytes_as('fulltext'), result[3])
2350
# It's not strictly minimal, but it seems reasonable for now for it to
2351
# ask which fallbacks have which parents.
2353
("get_parent_map", {key_basis, key_basis_2, key_missing}),
2354
# topological is requested from the fallback, because that is what
2355
# was requested at the top level.
2356
("get_record_stream", [key_basis_2, key_basis], 'topological', True)],
2359
def test_get_record_stream_unordered_deltas(self):
2360
# records from the test knit are answered without asking the basis:
2361
basis, test = self.get_basis_and_test_knit()
2363
key_basis = ('bar',)
2364
key_missing = ('missing',)
2365
test.add_lines(key, (), ['foo\n'])
2366
records = list(test.get_record_stream([key], 'unordered', False))
2367
self.assertEqual(1, len(records))
2368
self.assertEqual([], basis.calls)
2369
# Missing (from test knit) objects are retrieved from the basis:
2370
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2372
records = list(test.get_record_stream([key_basis, key_missing],
2373
'unordered', False))
2374
self.assertEqual(2, len(records))
2375
calls = list(basis.calls)
2376
for record in records:
2377
self.assertSubset([record.key], (key_basis, key_missing))
2378
if record.key == key_missing:
2379
self.assertIsInstance(record, AbsentContentFactory)
2381
reference = list(basis.get_record_stream([key_basis],
2382
'unordered', False))[0]
2383
self.assertEqual(reference.key, record.key)
2384
self.assertEqual(reference.sha1, record.sha1)
2385
self.assertEqual(reference.storage_kind, record.storage_kind)
2386
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2387
record.get_bytes_as(record.storage_kind))
2388
# It's not strictly minimal, but it seems reasonable for now for it to
2389
# ask which fallbacks have which parents.
2391
("get_parent_map", {key_basis, key_missing}),
2392
("get_record_stream", [key_basis], 'unordered', False)],
2395
def test_get_record_stream_ordered_deltas(self):
2396
# ordering is preserved down into the fallback store.
2397
basis, test = self.get_basis_and_test_knit()
2399
key_basis = ('bar',)
2400
key_basis_2 = ('quux',)
2401
key_missing = ('missing',)
2402
test.add_lines(key, (key_basis,), ['foo\n'])
2403
# Missing (from test knit) objects are retrieved from the basis:
2404
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2405
basis.add_lines(key_basis_2, (), ['quux\n'])
2407
# ask for in non-topological order
2408
records = list(test.get_record_stream(
2409
[key, key_basis, key_missing, key_basis_2], 'topological', False))
2410
self.assertEqual(4, len(records))
2412
for record in records:
2413
self.assertSubset([record.key],
2414
(key_basis, key_missing, key_basis_2, key))
2415
if record.key == key_missing:
2416
self.assertIsInstance(record, AbsentContentFactory)
2418
results.append((record.key, record.sha1, record.storage_kind,
2419
record.get_bytes_as(record.storage_kind)))
2420
calls = list(basis.calls)
2421
order = [record[0] for record in results]
2422
self.assertEqual([key_basis_2, key_basis, key], order)
2423
for result in results:
2424
if result[0] == key:
2428
record = next(source.get_record_stream([result[0]], 'unordered',
2430
self.assertEqual(record.key, result[0])
2431
self.assertEqual(record.sha1, result[1])
2432
self.assertEqual(record.storage_kind, result[2])
2433
self.assertEqual(record.get_bytes_as(record.storage_kind), result[3])
2434
# It's not strictly minimal, but it seems reasonable for now for it to
2435
# ask which fallbacks have which parents.
2437
("get_parent_map", {key_basis, key_basis_2, key_missing}),
2438
("get_record_stream", [key_basis_2, key_basis], 'topological', False)],
2441
def test_get_sha1s(self):
2442
# sha1's in the test knit are answered without asking the basis
2443
basis, test = self.get_basis_and_test_knit()
2445
key_basis = ('bar',)
2446
key_missing = ('missing',)
2447
test.add_lines(key, (), ['foo\n'])
2448
key_sha1sum = osutils.sha_string('foo\n')
2449
sha1s = test.get_sha1s([key])
2450
self.assertEqual({key: key_sha1sum}, sha1s)
2451
self.assertEqual([], basis.calls)
2452
# But texts that are not in the test knit are looked for in the basis
2453
# directly (rather than via text reconstruction) so that remote servers
2454
# etc don't have to answer with full content.
2455
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2456
basis_sha1sum = osutils.sha_string('foo\nbar\n')
2458
sha1s = test.get_sha1s([key, key_missing, key_basis])
2459
self.assertEqual({key: key_sha1sum,
2460
key_basis: basis_sha1sum}, sha1s)
2461
self.assertEqual([("get_sha1s", {key_basis, key_missing})],
2464
def test_insert_record_stream(self):
2465
# records are inserted as normal; insert_record_stream builds on
2466
# add_lines, so a smoke test should be all that's needed:
2468
key_basis = ('bar',)
2469
key_delta = ('zaphod',)
2470
basis, test = self.get_basis_and_test_knit()
2471
source = self.make_test_knit(name='source')
2472
basis.add_lines(key_basis, (), ['foo\n'])
2474
source.add_lines(key_basis, (), ['foo\n'])
2475
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2476
stream = source.get_record_stream([key_delta], 'unordered', False)
2477
test.insert_record_stream(stream)
2478
# XXX: this does somewhat too many calls in making sure of whether it
2479
# has to recreate the full text.
2480
self.assertEqual([("get_parent_map", {key_basis}),
2481
('get_parent_map', {key_basis}),
2482
('get_record_stream', [key_basis], 'unordered', True)],
2484
self.assertEqual({key_delta:(key_basis,)},
2485
test.get_parent_map([key_delta]))
2486
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2487
'unordered', True).next().get_bytes_as('fulltext'))
2489
def test_iter_lines_added_or_present_in_keys(self):
2490
# Lines from the basis are returned, and lines for a given key are only
2494
# all sources are asked for keys:
2495
basis, test = self.get_basis_and_test_knit()
2496
basis.add_lines(key1, (), ["foo"])
2498
lines = list(test.iter_lines_added_or_present_in_keys([key1]))
2499
self.assertEqual([("foo\n", key1)], lines)
2500
self.assertEqual([("iter_lines_added_or_present_in_keys", {key1})],
2502
# keys in both are not duplicated:
2503
test.add_lines(key2, (), ["bar\n"])
2504
basis.add_lines(key2, (), ["bar\n"])
2506
lines = list(test.iter_lines_added_or_present_in_keys([key2]))
2507
self.assertEqual([("bar\n", key2)], lines)
2508
self.assertEqual([], basis.calls)
2510
def test_keys(self):
2513
# all sources are asked for keys:
2514
basis, test = self.get_basis_and_test_knit()
2516
self.assertEqual(set(), set(keys))
2517
self.assertEqual([("keys",)], basis.calls)
2518
# keys from a basis are returned:
2519
basis.add_lines(key1, (), [])
2522
self.assertEqual({key1}, set(keys))
2523
self.assertEqual([("keys",)], basis.calls)
2524
# keys in both are not duplicated:
2525
test.add_lines(key2, (), [])
2526
basis.add_lines(key2, (), [])
2529
self.assertEqual(2, len(keys))
2530
self.assertEqual({key1, key2}, set(keys))
2531
self.assertEqual([("keys",)], basis.calls)
2533
def test_add_mpdiffs(self):
2534
# records are inserted as normal; add_mpdiff builds on
2535
# add_lines, so a smoke test should be all that's needed:
2537
key_basis = ('bar',)
2538
key_delta = ('zaphod',)
2539
basis, test = self.get_basis_and_test_knit()
2540
source = self.make_test_knit(name='source')
2541
basis.add_lines(key_basis, (), ['foo\n'])
2543
source.add_lines(key_basis, (), ['foo\n'])
2544
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2545
diffs = source.make_mpdiffs([key_delta])
2546
test.add_mpdiffs([(key_delta, (key_basis,),
2547
source.get_sha1s([key_delta])[key_delta], diffs[0])])
2548
self.assertEqual([("get_parent_map", {key_basis}),
2549
('get_record_stream', [key_basis], 'unordered', True),],
2551
self.assertEqual({key_delta:(key_basis,)},
2552
test.get_parent_map([key_delta]))
2553
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2554
'unordered', True).next().get_bytes_as('fulltext'))
2556
def test_make_mpdiffs(self):
2557
# Generating an mpdiff across a stacking boundary should detect parent
2561
key_right = ('zaphod',)
2562
basis, test = self.get_basis_and_test_knit()
2563
basis.add_lines(key_left, (), ['bar\n'])
2564
basis.add_lines(key_right, (), ['zaphod\n'])
2566
test.add_lines(key, (key_left, key_right),
2567
['bar\n', 'foo\n', 'zaphod\n'])
2568
diffs = test.make_mpdiffs([key])
2570
multiparent.MultiParent([multiparent.ParentText(0, 0, 0, 1),
2571
multiparent.NewText(['foo\n']),
2572
multiparent.ParentText(1, 0, 2, 1)])],
2574
self.assertEqual(3, len(basis.calls))
2576
("get_parent_map", {key_left, key_right}),
2577
("get_parent_map", {key_left, key_right}),
2580
last_call = basis.calls[-1]
2581
self.assertEqual('get_record_stream', last_call[0])
2582
self.assertEqual({key_left, key_right}, set(last_call[1]))
2583
self.assertEqual('topological', last_call[2])
2584
self.assertEqual(True, last_call[3])
2587
class TestNetworkBehaviour(KnitTests):
2588
"""Tests for getting data out of/into knits over the network."""
2590
def test_include_delta_closure_generates_a_knit_delta_closure(self):
2591
vf = self.make_test_knit(name='test')
2592
# put in three texts, giving ft, delta, delta
2593
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2594
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2595
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2596
# But heuristics could interfere, so check what happened:
2597
self.assertEqual(['knit-ft-gz', 'knit-delta-gz', 'knit-delta-gz'],
2598
[record.storage_kind for record in
2599
vf.get_record_stream([('base',), ('d1',), ('d2',)],
2600
'topological', False)])
2601
# generate a stream of just the deltas include_delta_closure=True,
2602
# serialise to the network, and check that we get a delta closure on the wire.
2603
stream = vf.get_record_stream([('d1',), ('d2',)], 'topological', True)
2604
netb = [record.get_bytes_as(record.storage_kind) for record in stream]
2605
# The first bytes should be a memo from _ContentMapGenerator, and the
2606
# second bytes should be empty (because its a API proxy not something
2607
# for wire serialisation.
2608
self.assertEqual('', netb[1])
2610
kind, line_end = network_bytes_to_kind_and_offset(bytes)
2611
self.assertEqual('knit-delta-closure', kind)
2614
class TestContentMapGenerator(KnitTests):
2615
"""Tests for ContentMapGenerator"""
2617
def test_get_record_stream_gives_records(self):
2618
vf = self.make_test_knit(name='test')
2619
# put in three texts, giving ft, delta, delta
2620
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2621
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2622
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2623
keys = [('d1',), ('d2',)]
2624
generator = _VFContentMapGenerator(vf, keys,
2625
global_map=vf.get_parent_map(keys))
2626
for record in generator.get_record_stream():
2627
if record.key == ('d1',):
2628
self.assertEqual('d1\n', record.get_bytes_as('fulltext'))
2630
self.assertEqual('d2\n', record.get_bytes_as('fulltext'))
2632
def test_get_record_stream_kinds_are_raw(self):
2633
vf = self.make_test_knit(name='test')
2634
# put in three texts, giving ft, delta, delta
2635
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2636
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2637
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2638
keys = [('base',), ('d1',), ('d2',)]
2639
generator = _VFContentMapGenerator(vf, keys,
2640
global_map=vf.get_parent_map(keys))
2641
kinds = {('base',): 'knit-delta-closure',
2642
('d1',): 'knit-delta-closure-ref',
2643
('d2',): 'knit-delta-closure-ref',
2645
for record in generator.get_record_stream():
2646
self.assertEqual(kinds[record.key], record.storage_kind)
added
removed
breezy/tests/test_knit.py
