1
# Copyright (C) 2005, 2006, 2007 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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
"""Tests for Knit data structure"""
19
from cStringIO import StringIO
32
from bzrlib.errors import (
33
RevisionAlreadyPresent,
38
from bzrlib.index import *
39
from bzrlib.knit import (
51
from bzrlib.repofmt import pack_repo
52
from bzrlib.tests import (
56
TestCaseWithMemoryTransport,
57
TestCaseWithTransport,
60
from bzrlib.transport import get_transport
61
from bzrlib.transport.memory import MemoryTransport
62
from bzrlib.tuned_gzip import GzipFile
63
from bzrlib.versionedfile import (
66
RecordingVersionedFilesDecorator,
70
class _CompiledKnitFeature(Feature):
74
import bzrlib._knit_load_data_c
79
def feature_name(self):
80
return 'bzrlib._knit_load_data_c'
82
CompiledKnitFeature = _CompiledKnitFeature()
85
class KnitContentTestsMixin(object):
87
def test_constructor(self):
88
content = self._make_content([])
91
content = self._make_content([])
92
self.assertEqual(content.text(), [])
94
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
95
self.assertEqual(content.text(), ["text1", "text2"])
98
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
100
self.assertIsInstance(copy, content.__class__)
101
self.assertEqual(copy.annotate(), content.annotate())
103
def assertDerivedBlocksEqual(self, source, target, noeol=False):
104
"""Assert that the derived matching blocks match real output"""
105
source_lines = source.splitlines(True)
106
target_lines = target.splitlines(True)
108
if noeol and not line.endswith('\n'):
112
source_content = self._make_content([(None, nl(l)) for l in source_lines])
113
target_content = self._make_content([(None, nl(l)) for l in target_lines])
114
line_delta = source_content.line_delta(target_content)
115
delta_blocks = list(KnitContent.get_line_delta_blocks(line_delta,
116
source_lines, target_lines))
117
matcher = KnitSequenceMatcher(None, source_lines, target_lines)
118
matcher_blocks = list(list(matcher.get_matching_blocks()))
119
self.assertEqual(matcher_blocks, delta_blocks)
121
def test_get_line_delta_blocks(self):
122
self.assertDerivedBlocksEqual('a\nb\nc\n', 'q\nc\n')
123
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1)
124
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1A)
125
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1B)
126
self.assertDerivedBlocksEqual(TEXT_1B, TEXT_1A)
127
self.assertDerivedBlocksEqual(TEXT_1A, TEXT_1B)
128
self.assertDerivedBlocksEqual(TEXT_1A, '')
129
self.assertDerivedBlocksEqual('', TEXT_1A)
130
self.assertDerivedBlocksEqual('', '')
131
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd')
133
def test_get_line_delta_blocks_noeol(self):
134
"""Handle historical knit deltas safely
136
Some existing knit deltas don't consider the last line to differ
137
when the only difference whether it has a final newline.
139
New knit deltas appear to always consider the last line to differ
142
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd\n', noeol=True)
143
self.assertDerivedBlocksEqual('a\nb\nc\nd\n', 'a\nb\nc', noeol=True)
144
self.assertDerivedBlocksEqual('a\nb\nc\n', 'a\nb\nc', noeol=True)
145
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\n', noeol=True)
157
Banana cup cake recipe
167
Banana cup cake recipe
169
- bananas (do not use plantains!!!)
176
Banana cup cake recipe
193
class TestPlainKnitContent(TestCase, KnitContentTestsMixin):
195
def _make_content(self, lines):
196
annotated_content = AnnotatedKnitContent(lines)
197
return PlainKnitContent(annotated_content.text(), 'bogus')
199
def test_annotate(self):
200
content = self._make_content([])
201
self.assertEqual(content.annotate(), [])
203
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
204
self.assertEqual(content.annotate(),
205
[("bogus", "text1"), ("bogus", "text2")])
207
def test_line_delta(self):
208
content1 = self._make_content([("", "a"), ("", "b")])
209
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
210
self.assertEqual(content1.line_delta(content2),
211
[(1, 2, 2, ["a", "c"])])
213
def test_line_delta_iter(self):
214
content1 = self._make_content([("", "a"), ("", "b")])
215
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
216
it = content1.line_delta_iter(content2)
217
self.assertEqual(it.next(), (1, 2, 2, ["a", "c"]))
218
self.assertRaises(StopIteration, it.next)
221
class TestAnnotatedKnitContent(TestCase, KnitContentTestsMixin):
223
def _make_content(self, lines):
224
return AnnotatedKnitContent(lines)
226
def test_annotate(self):
227
content = self._make_content([])
228
self.assertEqual(content.annotate(), [])
230
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
231
self.assertEqual(content.annotate(),
232
[("origin1", "text1"), ("origin2", "text2")])
234
def test_line_delta(self):
235
content1 = self._make_content([("", "a"), ("", "b")])
236
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
237
self.assertEqual(content1.line_delta(content2),
238
[(1, 2, 2, [("", "a"), ("", "c")])])
240
def test_line_delta_iter(self):
241
content1 = self._make_content([("", "a"), ("", "b")])
242
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
243
it = content1.line_delta_iter(content2)
244
self.assertEqual(it.next(), (1, 2, 2, [("", "a"), ("", "c")]))
245
self.assertRaises(StopIteration, it.next)
248
class MockTransport(object):
250
def __init__(self, file_lines=None):
251
self.file_lines = file_lines
253
# We have no base directory for the MockTransport
256
def get(self, filename):
257
if self.file_lines is None:
258
raise NoSuchFile(filename)
260
return StringIO("\n".join(self.file_lines))
262
def readv(self, relpath, offsets):
263
fp = self.get(relpath)
264
for offset, size in offsets:
266
yield offset, fp.read(size)
268
def __getattr__(self, name):
269
def queue_call(*args, **kwargs):
270
self.calls.append((name, args, kwargs))
274
class MockReadvFailingTransport(MockTransport):
275
"""Fail in the middle of a readv() result.
277
This Transport will successfully yield the first two requested hunks, but
278
raise NoSuchFile for the rest.
281
def readv(self, relpath, offsets):
283
for result in MockTransport.readv(self, relpath, offsets):
285
# we use 2 because the first offset is the pack header, the second
286
# is the first actual content requset
288
raise errors.NoSuchFile(relpath)
292
class KnitRecordAccessTestsMixin(object):
293
"""Tests for getting and putting knit records."""
295
def test_add_raw_records(self):
296
"""Add_raw_records adds records retrievable later."""
297
access = self.get_access()
298
memos = access.add_raw_records([('key', 10)], '1234567890')
299
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
301
def test_add_several_raw_records(self):
302
"""add_raw_records with many records and read some back."""
303
access = self.get_access()
304
memos = access.add_raw_records([('key', 10), ('key2', 2), ('key3', 5)],
306
self.assertEqual(['1234567890', '12', '34567'],
307
list(access.get_raw_records(memos)))
308
self.assertEqual(['1234567890'],
309
list(access.get_raw_records(memos[0:1])))
310
self.assertEqual(['12'],
311
list(access.get_raw_records(memos[1:2])))
312
self.assertEqual(['34567'],
313
list(access.get_raw_records(memos[2:3])))
314
self.assertEqual(['1234567890', '34567'],
315
list(access.get_raw_records(memos[0:1] + memos[2:3])))
318
class TestKnitKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
319
"""Tests for the .kndx implementation."""
321
def get_access(self):
322
"""Get a .knit style access instance."""
323
mapper = ConstantMapper("foo")
324
access = _KnitKeyAccess(self.get_transport(), mapper)
328
class _TestException(Exception):
329
"""Just an exception for local tests to use."""
332
class TestPackKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
333
"""Tests for the pack based access."""
335
def get_access(self):
336
return self._get_access()[0]
338
def _get_access(self, packname='packfile', index='FOO'):
339
transport = self.get_transport()
340
def write_data(bytes):
341
transport.append_bytes(packname, bytes)
342
writer = pack.ContainerWriter(write_data)
344
access = _DirectPackAccess({})
345
access.set_writer(writer, index, (transport, packname))
346
return access, writer
348
def make_pack_file(self):
349
"""Create a pack file with 2 records."""
350
access, writer = self._get_access(packname='packname', index='foo')
352
memos.extend(access.add_raw_records([('key1', 10)], '1234567890'))
353
memos.extend(access.add_raw_records([('key2', 5)], '12345'))
357
def make_vf_for_retrying(self):
358
"""Create 3 packs and a reload function.
360
Originally, 2 pack files will have the data, but one will be missing.
361
And then the third will be used in place of the first two if reload()
364
:return: (versioned_file, reload_counter)
365
versioned_file a KnitVersionedFiles using the packs for access
367
tree = self.make_branch_and_memory_tree('tree')
370
tree.add([''], ['root-id'])
371
tree.commit('one', rev_id='rev-1')
372
tree.commit('two', rev_id='rev-2')
373
tree.commit('three', rev_id='rev-3')
374
# Pack these two revisions into another pack file, but don't remove
376
repo = tree.branch.repository
377
collection = repo._pack_collection
378
collection.ensure_loaded()
379
orig_packs = collection.packs
380
packer = pack_repo.Packer(collection, orig_packs, '.testpack')
381
new_pack = packer.pack()
383
vf = tree.branch.repository.revisions
386
tree.branch.repository.lock_read()
387
self.addCleanup(tree.branch.repository.unlock)
389
# Set up a reload() function that switches to using the new pack file
390
new_index = new_pack.revision_index
391
access_tuple = new_pack.access_tuple()
392
reload_counter = [0, 0, 0]
394
reload_counter[0] += 1
395
if reload_counter[1] > 0:
396
# We already reloaded, nothing more to do
397
reload_counter[2] += 1
399
reload_counter[1] += 1
400
vf._index._graph_index._indices[:] = [new_index]
401
vf._access._indices.clear()
402
vf._access._indices[new_index] = access_tuple
404
# Delete one of the pack files so the data will need to be reloaded. We
405
# will delete the file with 'rev-2' in it
406
trans, name = orig_packs[1].access_tuple()
408
# We don't have the index trigger reloading because we want to test
409
# that we reload when the .pack disappears
410
vf._access._reload_func = reload
411
return vf, reload_counter
413
def make_reload_func(self, return_val=True):
416
reload_called[0] += 1
418
return reload_called, reload
420
def make_retry_exception(self):
421
# We raise a real exception so that sys.exc_info() is properly
424
raise _TestException('foobar')
425
except _TestException, e:
426
retry_exc = errors.RetryWithNewPacks(None, reload_occurred=False,
427
exc_info=sys.exc_info())
430
def test_read_from_several_packs(self):
431
access, writer = self._get_access()
433
memos.extend(access.add_raw_records([('key', 10)], '1234567890'))
435
access, writer = self._get_access('pack2', 'FOOBAR')
436
memos.extend(access.add_raw_records([('key', 5)], '12345'))
438
access, writer = self._get_access('pack3', 'BAZ')
439
memos.extend(access.add_raw_records([('key', 5)], 'alpha'))
441
transport = self.get_transport()
442
access = _DirectPackAccess({"FOO":(transport, 'packfile'),
443
"FOOBAR":(transport, 'pack2'),
444
"BAZ":(transport, 'pack3')})
445
self.assertEqual(['1234567890', '12345', 'alpha'],
446
list(access.get_raw_records(memos)))
447
self.assertEqual(['1234567890'],
448
list(access.get_raw_records(memos[0:1])))
449
self.assertEqual(['12345'],
450
list(access.get_raw_records(memos[1:2])))
451
self.assertEqual(['alpha'],
452
list(access.get_raw_records(memos[2:3])))
453
self.assertEqual(['1234567890', 'alpha'],
454
list(access.get_raw_records(memos[0:1] + memos[2:3])))
456
def test_set_writer(self):
457
"""The writer should be settable post construction."""
458
access = _DirectPackAccess({})
459
transport = self.get_transport()
460
packname = 'packfile'
462
def write_data(bytes):
463
transport.append_bytes(packname, bytes)
464
writer = pack.ContainerWriter(write_data)
466
access.set_writer(writer, index, (transport, packname))
467
memos = access.add_raw_records([('key', 10)], '1234567890')
469
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
471
def test_missing_index_raises_retry(self):
472
memos = self.make_pack_file()
473
transport = self.get_transport()
474
reload_called, reload_func = self.make_reload_func()
475
# Note that the index key has changed from 'foo' to 'bar'
476
access = _DirectPackAccess({'bar':(transport, 'packname')},
477
reload_func=reload_func)
478
e = self.assertListRaises(errors.RetryWithNewPacks,
479
access.get_raw_records, memos)
480
# Because a key was passed in which does not match our index list, we
481
# assume that the listing was already reloaded
482
self.assertTrue(e.reload_occurred)
483
self.assertIsInstance(e.exc_info, tuple)
484
self.assertIs(e.exc_info[0], KeyError)
485
self.assertIsInstance(e.exc_info[1], KeyError)
487
def test_missing_index_raises_key_error_with_no_reload(self):
488
memos = self.make_pack_file()
489
transport = self.get_transport()
490
# Note that the index key has changed from 'foo' to 'bar'
491
access = _DirectPackAccess({'bar':(transport, 'packname')})
492
e = self.assertListRaises(KeyError, access.get_raw_records, memos)
494
def test_missing_file_raises_retry(self):
495
memos = self.make_pack_file()
496
transport = self.get_transport()
497
reload_called, reload_func = self.make_reload_func()
498
# Note that the 'filename' has been changed to 'different-packname'
499
access = _DirectPackAccess({'foo':(transport, 'different-packname')},
500
reload_func=reload_func)
501
e = self.assertListRaises(errors.RetryWithNewPacks,
502
access.get_raw_records, memos)
503
# The file has gone missing, so we assume we need to reload
504
self.assertFalse(e.reload_occurred)
505
self.assertIsInstance(e.exc_info, tuple)
506
self.assertIs(e.exc_info[0], errors.NoSuchFile)
507
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
508
self.assertEqual('different-packname', e.exc_info[1].path)
510
def test_missing_file_raises_no_such_file_with_no_reload(self):
511
memos = self.make_pack_file()
512
transport = self.get_transport()
513
# Note that the 'filename' has been changed to 'different-packname'
514
access = _DirectPackAccess({'foo':(transport, 'different-packname')})
515
e = self.assertListRaises(errors.NoSuchFile,
516
access.get_raw_records, memos)
518
def test_failing_readv_raises_retry(self):
519
memos = self.make_pack_file()
520
transport = self.get_transport()
521
failing_transport = MockReadvFailingTransport(
522
[transport.get_bytes('packname')])
523
reload_called, reload_func = self.make_reload_func()
524
access = _DirectPackAccess({'foo':(failing_transport, 'packname')},
525
reload_func=reload_func)
526
# Asking for a single record will not trigger the Mock failure
527
self.assertEqual(['1234567890'],
528
list(access.get_raw_records(memos[:1])))
529
self.assertEqual(['12345'],
530
list(access.get_raw_records(memos[1:2])))
531
# A multiple offset readv() will fail mid-way through
532
e = self.assertListRaises(errors.RetryWithNewPacks,
533
access.get_raw_records, memos)
534
# The file has gone missing, so we assume we need to reload
535
self.assertFalse(e.reload_occurred)
536
self.assertIsInstance(e.exc_info, tuple)
537
self.assertIs(e.exc_info[0], errors.NoSuchFile)
538
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
539
self.assertEqual('packname', e.exc_info[1].path)
541
def test_failing_readv_raises_no_such_file_with_no_reload(self):
542
memos = self.make_pack_file()
543
transport = self.get_transport()
544
failing_transport = MockReadvFailingTransport(
545
[transport.get_bytes('packname')])
546
reload_called, reload_func = self.make_reload_func()
547
access = _DirectPackAccess({'foo':(failing_transport, 'packname')})
548
# Asking for a single record will not trigger the Mock failure
549
self.assertEqual(['1234567890'],
550
list(access.get_raw_records(memos[:1])))
551
self.assertEqual(['12345'],
552
list(access.get_raw_records(memos[1:2])))
553
# A multiple offset readv() will fail mid-way through
554
e = self.assertListRaises(errors.NoSuchFile,
555
access.get_raw_records, memos)
557
def test_reload_or_raise_no_reload(self):
558
access = _DirectPackAccess({}, reload_func=None)
559
retry_exc = self.make_retry_exception()
560
# Without a reload_func, we will just re-raise the original exception
561
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
563
def test_reload_or_raise_reload_changed(self):
564
reload_called, reload_func = self.make_reload_func(return_val=True)
565
access = _DirectPackAccess({}, reload_func=reload_func)
566
retry_exc = self.make_retry_exception()
567
access.reload_or_raise(retry_exc)
568
self.assertEqual([1], reload_called)
569
retry_exc.reload_occurred=True
570
access.reload_or_raise(retry_exc)
571
self.assertEqual([2], reload_called)
573
def test_reload_or_raise_reload_no_change(self):
574
reload_called, reload_func = self.make_reload_func(return_val=False)
575
access = _DirectPackAccess({}, reload_func=reload_func)
576
retry_exc = self.make_retry_exception()
577
# If reload_occurred is False, then we consider it an error to have
578
# reload_func() return False (no changes).
579
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
580
self.assertEqual([1], reload_called)
581
retry_exc.reload_occurred=True
582
# If reload_occurred is True, then we assume nothing changed because
583
# it had changed earlier, but didn't change again
584
access.reload_or_raise(retry_exc)
585
self.assertEqual([2], reload_called)
587
def test_annotate_retries(self):
588
vf, reload_counter = self.make_vf_for_retrying()
589
# It is a little bit bogus to annotate the Revision VF, but it works,
590
# as we have ancestry stored there
592
reload_lines = vf.annotate(key)
593
self.assertEqual([1, 1, 0], reload_counter)
594
plain_lines = vf.annotate(key)
595
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
596
if reload_lines != plain_lines:
597
self.fail('Annotation was not identical with reloading.')
598
# Now delete the packs-in-use, which should trigger another reload, but
599
# this time we just raise an exception because we can't recover
600
for trans, name in vf._access._indices.itervalues():
602
self.assertRaises(errors.NoSuchFile, vf.annotate, key)
603
self.assertEqual([2, 1, 1], reload_counter)
605
def test__get_record_map_retries(self):
606
vf, reload_counter = self.make_vf_for_retrying()
607
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
608
records = vf._get_record_map(keys)
609
self.assertEqual(keys, sorted(records.keys()))
610
self.assertEqual([1, 1, 0], reload_counter)
611
# Now delete the packs-in-use, which should trigger another reload, but
612
# this time we just raise an exception because we can't recover
613
for trans, name in vf._access._indices.itervalues():
615
self.assertRaises(errors.NoSuchFile, vf._get_record_map, keys)
616
self.assertEqual([2, 1, 1], reload_counter)
618
def test_get_record_stream_retries(self):
619
vf, reload_counter = self.make_vf_for_retrying()
620
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
621
record_stream = vf.get_record_stream(keys, 'topological', False)
622
record = record_stream.next()
623
self.assertEqual(('rev-1',), record.key)
624
self.assertEqual([0, 0, 0], reload_counter)
625
record = record_stream.next()
626
self.assertEqual(('rev-2',), record.key)
627
self.assertEqual([1, 1, 0], reload_counter)
628
record = record_stream.next()
629
self.assertEqual(('rev-3',), record.key)
630
self.assertEqual([1, 1, 0], reload_counter)
631
# Now delete all pack files, and see that we raise the right error
632
for trans, name in vf._access._indices.itervalues():
634
self.assertListRaises(errors.NoSuchFile,
635
vf.get_record_stream, keys, 'topological', False)
637
def test_iter_lines_added_or_present_in_keys_retries(self):
638
vf, reload_counter = self.make_vf_for_retrying()
639
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
640
# Unfortunately, iter_lines_added_or_present_in_keys iterates the
641
# result in random order (determined by the iteration order from a
642
# set()), so we don't have any solid way to trigger whether data is
643
# read before or after. However we tried to delete the middle node to
644
# exercise the code well.
645
# What we care about is that all lines are always yielded, but not
648
reload_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
649
self.assertEqual([1, 1, 0], reload_counter)
650
# Now do it again, to make sure the result is equivalent
651
plain_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
652
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
653
self.assertEqual(plain_lines, reload_lines)
654
self.assertEqual(21, len(plain_lines))
655
# Now delete all pack files, and see that we raise the right error
656
for trans, name in vf._access._indices.itervalues():
658
self.assertListRaises(errors.NoSuchFile,
659
vf.iter_lines_added_or_present_in_keys, keys)
660
self.assertEqual([2, 1, 1], reload_counter)
662
def test_get_record_stream_yields_disk_sorted_order(self):
663
# if we get 'unordered' pick a semi-optimal order for reading. The
664
# order should be grouped by pack file, and then by position in file
665
repo = self.make_repository('test', format='pack-0.92')
667
self.addCleanup(repo.unlock)
668
repo.start_write_group()
670
vf.add_lines(('f-id', 'rev-5'), [('f-id', 'rev-4')], ['lines\n'])
671
vf.add_lines(('f-id', 'rev-1'), [], ['lines\n'])
672
vf.add_lines(('f-id', 'rev-2'), [('f-id', 'rev-1')], ['lines\n'])
673
repo.commit_write_group()
674
# We inserted them as rev-5, rev-1, rev-2, we should get them back in
676
stream = vf.get_record_stream([('f-id', 'rev-1'), ('f-id', 'rev-5'),
677
('f-id', 'rev-2')], 'unordered', False)
678
keys = [r.key for r in stream]
679
self.assertEqual([('f-id', 'rev-5'), ('f-id', 'rev-1'),
680
('f-id', 'rev-2')], keys)
681
repo.start_write_group()
682
vf.add_lines(('f-id', 'rev-4'), [('f-id', 'rev-3')], ['lines\n'])
683
vf.add_lines(('f-id', 'rev-3'), [('f-id', 'rev-2')], ['lines\n'])
684
vf.add_lines(('f-id', 'rev-6'), [('f-id', 'rev-5')], ['lines\n'])
685
repo.commit_write_group()
686
# Request in random order, to make sure the output order isn't based on
688
request_keys = set(('f-id', 'rev-%d' % i) for i in range(1, 7))
689
stream = vf.get_record_stream(request_keys, 'unordered', False)
690
keys = [r.key for r in stream]
691
# We want to get the keys back in disk order, but it doesn't matter
692
# which pack we read from first. So this can come back in 2 orders
693
alt1 = [('f-id', 'rev-%d' % i) for i in [4, 3, 6, 5, 1, 2]]
694
alt2 = [('f-id', 'rev-%d' % i) for i in [5, 1, 2, 4, 3, 6]]
695
if keys != alt1 and keys != alt2:
696
self.fail('Returned key order did not match either expected order.'
697
' expected %s or %s, not %s'
698
% (alt1, alt2, keys))
701
class LowLevelKnitDataTests(TestCase):
703
def create_gz_content(self, text):
705
gz_file = gzip.GzipFile(mode='wb', fileobj=sio)
708
return sio.getvalue()
710
def make_multiple_records(self):
711
"""Create the content for multiple records."""
712
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
714
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
719
record_1 = (0, len(gz_txt), sha1sum)
720
total_txt.append(gz_txt)
721
sha1sum = osutils.sha('baz\n').hexdigest()
722
gz_txt = self.create_gz_content('version rev-id-2 1 %s\n'
726
record_2 = (record_1[1], len(gz_txt), sha1sum)
727
total_txt.append(gz_txt)
728
return total_txt, record_1, record_2
730
def test_valid_knit_data(self):
731
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
732
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
737
transport = MockTransport([gz_txt])
738
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
739
knit = KnitVersionedFiles(None, access)
740
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
742
contents = list(knit._read_records_iter(records))
743
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'],
744
'4e48e2c9a3d2ca8a708cb0cc545700544efb5021')], contents)
746
raw_contents = list(knit._read_records_iter_raw(records))
747
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
749
def test_multiple_records_valid(self):
750
total_txt, record_1, record_2 = self.make_multiple_records()
751
transport = MockTransport([''.join(total_txt)])
752
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
753
knit = KnitVersionedFiles(None, access)
754
records = [(('rev-id-1',), (('rev-id-1',), record_1[0], record_1[1])),
755
(('rev-id-2',), (('rev-id-2',), record_2[0], record_2[1]))]
757
contents = list(knit._read_records_iter(records))
758
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'], record_1[2]),
759
(('rev-id-2',), ['baz\n'], record_2[2])],
762
raw_contents = list(knit._read_records_iter_raw(records))
763
self.assertEqual([(('rev-id-1',), total_txt[0], record_1[2]),
764
(('rev-id-2',), total_txt[1], record_2[2])],
767
def test_not_enough_lines(self):
768
sha1sum = osutils.sha('foo\n').hexdigest()
769
# record says 2 lines data says 1
770
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
774
transport = MockTransport([gz_txt])
775
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
776
knit = KnitVersionedFiles(None, access)
777
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
778
self.assertRaises(errors.KnitCorrupt, list,
779
knit._read_records_iter(records))
781
# read_records_iter_raw won't detect that sort of mismatch/corruption
782
raw_contents = list(knit._read_records_iter_raw(records))
783
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
785
def test_too_many_lines(self):
786
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
787
# record says 1 lines data says 2
788
gz_txt = self.create_gz_content('version rev-id-1 1 %s\n'
793
transport = MockTransport([gz_txt])
794
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
795
knit = KnitVersionedFiles(None, access)
796
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
797
self.assertRaises(errors.KnitCorrupt, list,
798
knit._read_records_iter(records))
800
# read_records_iter_raw won't detect that sort of mismatch/corruption
801
raw_contents = list(knit._read_records_iter_raw(records))
802
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
804
def test_mismatched_version_id(self):
805
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
806
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
811
transport = MockTransport([gz_txt])
812
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
813
knit = KnitVersionedFiles(None, access)
814
# We are asking for rev-id-2, but the data is rev-id-1
815
records = [(('rev-id-2',), (('rev-id-2',), 0, len(gz_txt)))]
816
self.assertRaises(errors.KnitCorrupt, list,
817
knit._read_records_iter(records))
819
# read_records_iter_raw detects mismatches in the header
820
self.assertRaises(errors.KnitCorrupt, list,
821
knit._read_records_iter_raw(records))
823
def test_uncompressed_data(self):
824
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
825
txt = ('version rev-id-1 2 %s\n'
830
transport = MockTransport([txt])
831
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
832
knit = KnitVersionedFiles(None, access)
833
records = [(('rev-id-1',), (('rev-id-1',), 0, len(txt)))]
835
# We don't have valid gzip data ==> corrupt
836
self.assertRaises(errors.KnitCorrupt, list,
837
knit._read_records_iter(records))
839
# read_records_iter_raw will notice the bad data
840
self.assertRaises(errors.KnitCorrupt, list,
841
knit._read_records_iter_raw(records))
843
def test_corrupted_data(self):
844
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
845
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
850
# Change 2 bytes in the middle to \xff
851
gz_txt = gz_txt[:10] + '\xff\xff' + gz_txt[12:]
852
transport = MockTransport([gz_txt])
853
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
854
knit = KnitVersionedFiles(None, access)
855
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
856
self.assertRaises(errors.KnitCorrupt, list,
857
knit._read_records_iter(records))
858
# read_records_iter_raw will barf on bad gz data
859
self.assertRaises(errors.KnitCorrupt, list,
860
knit._read_records_iter_raw(records))
863
class LowLevelKnitIndexTests(TestCase):
865
def get_knit_index(self, transport, name, mode):
866
mapper = ConstantMapper(name)
867
orig = knit._load_data
869
knit._load_data = orig
870
self.addCleanup(reset)
871
from bzrlib._knit_load_data_py import _load_data_py
872
knit._load_data = _load_data_py
873
allow_writes = lambda: 'w' in mode
874
return _KndxIndex(transport, mapper, lambda:None, allow_writes, lambda:True)
876
def test_create_file(self):
877
transport = MockTransport()
878
index = self.get_knit_index(transport, "filename", "w")
880
call = transport.calls.pop(0)
881
# call[1][1] is a StringIO - we can't test it by simple equality.
882
self.assertEqual('put_file_non_atomic', call[0])
883
self.assertEqual('filename.kndx', call[1][0])
884
# With no history, _KndxIndex writes a new index:
885
self.assertEqual(_KndxIndex.HEADER,
886
call[1][1].getvalue())
887
self.assertEqual({'create_parent_dir': True}, call[2])
889
def test_read_utf8_version_id(self):
890
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
891
utf8_revision_id = unicode_revision_id.encode('utf-8')
892
transport = MockTransport([
894
'%s option 0 1 :' % (utf8_revision_id,)
896
index = self.get_knit_index(transport, "filename", "r")
897
# _KndxIndex is a private class, and deals in utf8 revision_ids, not
898
# Unicode revision_ids.
899
self.assertEqual({(utf8_revision_id,):()},
900
index.get_parent_map(index.keys()))
901
self.assertFalse((unicode_revision_id,) in index.keys())
903
def test_read_utf8_parents(self):
904
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
905
utf8_revision_id = unicode_revision_id.encode('utf-8')
906
transport = MockTransport([
908
"version option 0 1 .%s :" % (utf8_revision_id,)
910
index = self.get_knit_index(transport, "filename", "r")
911
self.assertEqual({("version",):((utf8_revision_id,),)},
912
index.get_parent_map(index.keys()))
914
def test_read_ignore_corrupted_lines(self):
915
transport = MockTransport([
918
"corrupted options 0 1 .b .c ",
919
"version options 0 1 :"
921
index = self.get_knit_index(transport, "filename", "r")
922
self.assertEqual(1, len(index.keys()))
923
self.assertEqual(set([("version",)]), index.keys())
925
def test_read_corrupted_header(self):
926
transport = MockTransport(['not a bzr knit index header\n'])
927
index = self.get_knit_index(transport, "filename", "r")
928
self.assertRaises(KnitHeaderError, index.keys)
930
def test_read_duplicate_entries(self):
931
transport = MockTransport([
933
"parent options 0 1 :",
934
"version options1 0 1 0 :",
935
"version options2 1 2 .other :",
936
"version options3 3 4 0 .other :"
938
index = self.get_knit_index(transport, "filename", "r")
939
self.assertEqual(2, len(index.keys()))
940
# check that the index used is the first one written. (Specific
941
# to KnitIndex style indices.
942
self.assertEqual("1", index._dictionary_compress([("version",)]))
943
self.assertEqual((("version",), 3, 4), index.get_position(("version",)))
944
self.assertEqual(["options3"], index.get_options(("version",)))
945
self.assertEqual({("version",):(("parent",), ("other",))},
946
index.get_parent_map([("version",)]))
948
def test_read_compressed_parents(self):
949
transport = MockTransport([
953
"c option 0 1 1 0 :",
955
index = self.get_knit_index(transport, "filename", "r")
956
self.assertEqual({("b",):(("a",),), ("c",):(("b",), ("a",))},
957
index.get_parent_map([("b",), ("c",)]))
959
def test_write_utf8_version_id(self):
960
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
961
utf8_revision_id = unicode_revision_id.encode('utf-8')
962
transport = MockTransport([
965
index = self.get_knit_index(transport, "filename", "r")
967
((utf8_revision_id,), ["option"], ((utf8_revision_id,), 0, 1), [])])
968
call = transport.calls.pop(0)
969
# call[1][1] is a StringIO - we can't test it by simple equality.
970
self.assertEqual('put_file_non_atomic', call[0])
971
self.assertEqual('filename.kndx', call[1][0])
972
# With no history, _KndxIndex writes a new index:
973
self.assertEqual(_KndxIndex.HEADER +
974
"\n%s option 0 1 :" % (utf8_revision_id,),
975
call[1][1].getvalue())
976
self.assertEqual({'create_parent_dir': True}, call[2])
978
def test_write_utf8_parents(self):
979
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
980
utf8_revision_id = unicode_revision_id.encode('utf-8')
981
transport = MockTransport([
984
index = self.get_knit_index(transport, "filename", "r")
986
(("version",), ["option"], (("version",), 0, 1), [(utf8_revision_id,)])])
987
call = transport.calls.pop(0)
988
# call[1][1] is a StringIO - we can't test it by simple equality.
989
self.assertEqual('put_file_non_atomic', call[0])
990
self.assertEqual('filename.kndx', call[1][0])
991
# With no history, _KndxIndex writes a new index:
992
self.assertEqual(_KndxIndex.HEADER +
993
"\nversion option 0 1 .%s :" % (utf8_revision_id,),
994
call[1][1].getvalue())
995
self.assertEqual({'create_parent_dir': True}, call[2])
998
transport = MockTransport([
1001
index = self.get_knit_index(transport, "filename", "r")
1003
self.assertEqual(set(), index.keys())
1005
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1006
self.assertEqual(set([("a",)]), index.keys())
1008
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1009
self.assertEqual(set([("a",)]), index.keys())
1011
index.add_records([(("b",), ["option"], (("b",), 0, 1), [])])
1012
self.assertEqual(set([("a",), ("b",)]), index.keys())
1014
def add_a_b(self, index, random_id=None):
1016
if random_id is not None:
1017
kwargs["random_id"] = random_id
1019
(("a",), ["option"], (("a",), 0, 1), [("b",)]),
1020
(("a",), ["opt"], (("a",), 1, 2), [("c",)]),
1021
(("b",), ["option"], (("b",), 2, 3), [("a",)])
1024
def assertIndexIsAB(self, index):
1029
index.get_parent_map(index.keys()))
1030
self.assertEqual((("a",), 1, 2), index.get_position(("a",)))
1031
self.assertEqual((("b",), 2, 3), index.get_position(("b",)))
1032
self.assertEqual(["opt"], index.get_options(("a",)))
1034
def test_add_versions(self):
1035
transport = MockTransport([
1038
index = self.get_knit_index(transport, "filename", "r")
1041
call = transport.calls.pop(0)
1042
# call[1][1] is a StringIO - we can't test it by simple equality.
1043
self.assertEqual('put_file_non_atomic', call[0])
1044
self.assertEqual('filename.kndx', call[1][0])
1045
# With no history, _KndxIndex writes a new index:
1048
"\na option 0 1 .b :"
1050
"\nb option 2 3 0 :",
1051
call[1][1].getvalue())
1052
self.assertEqual({'create_parent_dir': True}, call[2])
1053
self.assertIndexIsAB(index)
1055
def test_add_versions_random_id_is_accepted(self):
1056
transport = MockTransport([
1059
index = self.get_knit_index(transport, "filename", "r")
1060
self.add_a_b(index, random_id=True)
1062
def test_delay_create_and_add_versions(self):
1063
transport = MockTransport()
1065
index = self.get_knit_index(transport, "filename", "w")
1067
self.assertEqual([], transport.calls)
1070
#[ {"dir_mode": 0777, "create_parent_dir": True, "mode": "wb"},
1072
# Two calls: one during which we load the existing index (and when its
1073
# missing create it), then a second where we write the contents out.
1074
self.assertEqual(2, len(transport.calls))
1075
call = transport.calls.pop(0)
1076
self.assertEqual('put_file_non_atomic', call[0])
1077
self.assertEqual('filename.kndx', call[1][0])
1078
# With no history, _KndxIndex writes a new index:
1079
self.assertEqual(_KndxIndex.HEADER, call[1][1].getvalue())
1080
self.assertEqual({'create_parent_dir': True}, call[2])
1081
call = transport.calls.pop(0)
1082
# call[1][1] is a StringIO - we can't test it by simple equality.
1083
self.assertEqual('put_file_non_atomic', call[0])
1084
self.assertEqual('filename.kndx', call[1][0])
1085
# With no history, _KndxIndex writes a new index:
1088
"\na option 0 1 .b :"
1090
"\nb option 2 3 0 :",
1091
call[1][1].getvalue())
1092
self.assertEqual({'create_parent_dir': True}, call[2])
1094
def test_get_position(self):
1095
transport = MockTransport([
1100
index = self.get_knit_index(transport, "filename", "r")
1102
self.assertEqual((("a",), 0, 1), index.get_position(("a",)))
1103
self.assertEqual((("b",), 1, 2), index.get_position(("b",)))
1105
def test_get_method(self):
1106
transport = MockTransport([
1108
"a fulltext,unknown 0 1 :",
1109
"b unknown,line-delta 1 2 :",
1112
index = self.get_knit_index(transport, "filename", "r")
1114
self.assertEqual("fulltext", index.get_method("a"))
1115
self.assertEqual("line-delta", index.get_method("b"))
1116
self.assertRaises(errors.KnitIndexUnknownMethod, index.get_method, "c")
1118
def test_get_options(self):
1119
transport = MockTransport([
1124
index = self.get_knit_index(transport, "filename", "r")
1126
self.assertEqual(["opt1"], index.get_options("a"))
1127
self.assertEqual(["opt2", "opt3"], index.get_options("b"))
1129
def test_get_parent_map(self):
1130
transport = MockTransport([
1133
"b option 1 2 0 .c :",
1134
"c option 1 2 1 0 .e :"
1136
index = self.get_knit_index(transport, "filename", "r")
1140
("b",):(("a",), ("c",)),
1141
("c",):(("b",), ("a",), ("e",)),
1142
}, index.get_parent_map(index.keys()))
1144
def test_impossible_parent(self):
1145
"""Test we get KnitCorrupt if the parent couldn't possibly exist."""
1146
transport = MockTransport([
1149
"b option 0 1 4 :" # We don't have a 4th record
1151
index = self.get_knit_index(transport, 'filename', 'r')
1153
self.assertRaises(errors.KnitCorrupt, index.keys)
1154
except TypeError, e:
1155
if (str(e) == ('exceptions must be strings, classes, or instances,'
1156
' not exceptions.IndexError')
1157
and sys.version_info[0:2] >= (2,5)):
1158
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1159
' raising new style exceptions with python'
1164
def test_corrupted_parent(self):
1165
transport = MockTransport([
1169
"c option 0 1 1v :", # Can't have a parent of '1v'
1171
index = self.get_knit_index(transport, 'filename', 'r')
1173
self.assertRaises(errors.KnitCorrupt, index.keys)
1174
except TypeError, e:
1175
if (str(e) == ('exceptions must be strings, classes, or instances,'
1176
' not exceptions.ValueError')
1177
and sys.version_info[0:2] >= (2,5)):
1178
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1179
' raising new style exceptions with python'
1184
def test_corrupted_parent_in_list(self):
1185
transport = MockTransport([
1189
"c option 0 1 1 v :", # Can't have a parent of 'v'
1191
index = self.get_knit_index(transport, 'filename', 'r')
1193
self.assertRaises(errors.KnitCorrupt, index.keys)
1194
except TypeError, e:
1195
if (str(e) == ('exceptions must be strings, classes, or instances,'
1196
' not exceptions.ValueError')
1197
and sys.version_info[0:2] >= (2,5)):
1198
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1199
' raising new style exceptions with python'
1204
def test_invalid_position(self):
1205
transport = MockTransport([
1209
index = self.get_knit_index(transport, 'filename', 'r')
1211
self.assertRaises(errors.KnitCorrupt, index.keys)
1212
except TypeError, e:
1213
if (str(e) == ('exceptions must be strings, classes, or instances,'
1214
' not exceptions.ValueError')
1215
and sys.version_info[0:2] >= (2,5)):
1216
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1217
' raising new style exceptions with python'
1222
def test_invalid_size(self):
1223
transport = MockTransport([
1227
index = self.get_knit_index(transport, 'filename', 'r')
1229
self.assertRaises(errors.KnitCorrupt, index.keys)
1230
except TypeError, e:
1231
if (str(e) == ('exceptions must be strings, classes, or instances,'
1232
' not exceptions.ValueError')
1233
and sys.version_info[0:2] >= (2,5)):
1234
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1235
' raising new style exceptions with python'
1240
def test_short_line(self):
1241
transport = MockTransport([
1244
"b option 10 10 0", # This line isn't terminated, ignored
1246
index = self.get_knit_index(transport, "filename", "r")
1247
self.assertEqual(set([('a',)]), index.keys())
1249
def test_skip_incomplete_record(self):
1250
# A line with bogus data should just be skipped
1251
transport = MockTransport([
1254
"b option 10 10 0", # This line isn't terminated, ignored
1255
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1257
index = self.get_knit_index(transport, "filename", "r")
1258
self.assertEqual(set([('a',), ('c',)]), index.keys())
1260
def test_trailing_characters(self):
1261
# A line with bogus data should just be skipped
1262
transport = MockTransport([
1265
"b option 10 10 0 :a", # This line has extra trailing characters
1266
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1268
index = self.get_knit_index(transport, "filename", "r")
1269
self.assertEqual(set([('a',), ('c',)]), index.keys())
1272
class LowLevelKnitIndexTests_c(LowLevelKnitIndexTests):
1274
_test_needs_features = [CompiledKnitFeature]
1276
def get_knit_index(self, transport, name, mode):
1277
mapper = ConstantMapper(name)
1278
orig = knit._load_data
1280
knit._load_data = orig
1281
self.addCleanup(reset)
1282
from bzrlib._knit_load_data_c import _load_data_c
1283
knit._load_data = _load_data_c
1284
allow_writes = lambda: mode == 'w'
1285
return _KndxIndex(transport, mapper, lambda:None, allow_writes, lambda:True)
1288
class KnitTests(TestCaseWithTransport):
1289
"""Class containing knit test helper routines."""
1291
def make_test_knit(self, annotate=False, name='test'):
1292
mapper = ConstantMapper(name)
1293
return make_file_factory(annotate, mapper)(self.get_transport())
1296
class TestBadShaError(KnitTests):
1297
"""Tests for handling of sha errors."""
1299
def test_exception_has_text(self):
1300
# having the failed text included in the error allows for recovery.
1301
source = self.make_test_knit()
1302
target = self.make_test_knit(name="target")
1303
if not source._max_delta_chain:
1304
raise TestNotApplicable(
1305
"cannot get delta-caused sha failures without deltas.")
1308
broken = ('broken',)
1309
source.add_lines(basis, (), ['foo\n'])
1310
source.add_lines(broken, (basis,), ['foo\n', 'bar\n'])
1311
# Seed target with a bad basis text
1312
target.add_lines(basis, (), ['gam\n'])
1313
target.insert_record_stream(
1314
source.get_record_stream([broken], 'unordered', False))
1315
err = self.assertRaises(errors.KnitCorrupt,
1316
target.get_record_stream([broken], 'unordered', True).next)
1317
self.assertEqual(['gam\n', 'bar\n'], err.content)
1318
# Test for formatting with live data
1319
self.assertStartsWith(str(err), "Knit ")
1322
class TestKnitIndex(KnitTests):
1324
def test_add_versions_dictionary_compresses(self):
1325
"""Adding versions to the index should update the lookup dict"""
1326
knit = self.make_test_knit()
1328
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1329
self.check_file_contents('test.kndx',
1330
'# bzr knit index 8\n'
1332
'a-1 fulltext 0 0 :'
1335
(('a-2',), ['fulltext'], (('a-2',), 0, 0), [('a-1',)]),
1336
(('a-3',), ['fulltext'], (('a-3',), 0, 0), [('a-2',)]),
1338
self.check_file_contents('test.kndx',
1339
'# bzr knit index 8\n'
1341
'a-1 fulltext 0 0 :\n'
1342
'a-2 fulltext 0 0 0 :\n'
1343
'a-3 fulltext 0 0 1 :'
1345
self.assertEqual(set([('a-3',), ('a-1',), ('a-2',)]), idx.keys())
1347
('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False)),
1348
('a-2',): ((('a-2',), 0, 0), None, (('a-1',),), ('fulltext', False)),
1349
('a-3',): ((('a-3',), 0, 0), None, (('a-2',),), ('fulltext', False)),
1350
}, idx.get_build_details(idx.keys()))
1351
self.assertEqual({('a-1',):(),
1352
('a-2',):(('a-1',),),
1353
('a-3',):(('a-2',),),},
1354
idx.get_parent_map(idx.keys()))
1356
def test_add_versions_fails_clean(self):
1357
"""If add_versions fails in the middle, it restores a pristine state.
1359
Any modifications that are made to the index are reset if all versions
1362
# This cheats a little bit by passing in a generator which will
1363
# raise an exception before the processing finishes
1364
# Other possibilities would be to have an version with the wrong number
1365
# of entries, or to make the backing transport unable to write any
1368
knit = self.make_test_knit()
1370
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1372
class StopEarly(Exception):
1375
def generate_failure():
1376
"""Add some entries and then raise an exception"""
1377
yield (('a-2',), ['fulltext'], (None, 0, 0), ('a-1',))
1378
yield (('a-3',), ['fulltext'], (None, 0, 0), ('a-2',))
1381
# Assert the pre-condition
1383
self.assertEqual(set([('a-1',)]), set(idx.keys()))
1385
{('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False))},
1386
idx.get_build_details([('a-1',)]))
1387
self.assertEqual({('a-1',):()}, idx.get_parent_map(idx.keys()))
1390
self.assertRaises(StopEarly, idx.add_records, generate_failure())
1391
# And it shouldn't be modified
1394
def test_knit_index_ignores_empty_files(self):
1395
# There was a race condition in older bzr, where a ^C at the right time
1396
# could leave an empty .kndx file, which bzr would later claim was a
1397
# corrupted file since the header was not present. In reality, the file
1398
# just wasn't created, so it should be ignored.
1399
t = get_transport('.')
1400
t.put_bytes('test.kndx', '')
1402
knit = self.make_test_knit()
1404
def test_knit_index_checks_header(self):
1405
t = get_transport('.')
1406
t.put_bytes('test.kndx', '# not really a knit header\n\n')
1407
k = self.make_test_knit()
1408
self.assertRaises(KnitHeaderError, k.keys)
1411
class TestGraphIndexKnit(KnitTests):
1412
"""Tests for knits using a GraphIndex rather than a KnitIndex."""
1414
def make_g_index(self, name, ref_lists=0, nodes=[]):
1415
builder = GraphIndexBuilder(ref_lists)
1416
for node, references, value in nodes:
1417
builder.add_node(node, references, value)
1418
stream = builder.finish()
1419
trans = self.get_transport()
1420
size = trans.put_file(name, stream)
1421
return GraphIndex(trans, name, size)
1423
def two_graph_index(self, deltas=False, catch_adds=False):
1424
"""Build a two-graph index.
1426
:param deltas: If true, use underlying indices with two node-ref
1427
lists and 'parent' set to a delta-compressed against tail.
1429
# build a complex graph across several indices.
1431
# delta compression inn the index
1432
index1 = self.make_g_index('1', 2, [
1433
(('tip', ), 'N0 100', ([('parent', )], [], )),
1434
(('tail', ), '', ([], []))])
1435
index2 = self.make_g_index('2', 2, [
1436
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], [('tail', )])),
1437
(('separate', ), '', ([], []))])
1439
# just blob location and graph in the index.
1440
index1 = self.make_g_index('1', 1, [
1441
(('tip', ), 'N0 100', ([('parent', )], )),
1442
(('tail', ), '', ([], ))])
1443
index2 = self.make_g_index('2', 1, [
1444
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], )),
1445
(('separate', ), '', ([], ))])
1446
combined_index = CombinedGraphIndex([index1, index2])
1448
self.combined_index = combined_index
1449
self.caught_entries = []
1450
add_callback = self.catch_add
1453
return _KnitGraphIndex(combined_index, lambda:True, deltas=deltas,
1454
add_callback=add_callback)
1456
def test_keys(self):
1457
index = self.two_graph_index()
1458
self.assertEqual(set([('tail',), ('tip',), ('parent',), ('separate',)]),
1461
def test_get_position(self):
1462
index = self.two_graph_index()
1463
self.assertEqual((index._graph_index._indices[0], 0, 100), index.get_position(('tip',)))
1464
self.assertEqual((index._graph_index._indices[1], 100, 78), index.get_position(('parent',)))
1466
def test_get_method_deltas(self):
1467
index = self.two_graph_index(deltas=True)
1468
self.assertEqual('fulltext', index.get_method(('tip',)))
1469
self.assertEqual('line-delta', index.get_method(('parent',)))
1471
def test_get_method_no_deltas(self):
1472
# check that the parent-history lookup is ignored with deltas=False.
1473
index = self.two_graph_index(deltas=False)
1474
self.assertEqual('fulltext', index.get_method(('tip',)))
1475
self.assertEqual('fulltext', index.get_method(('parent',)))
1477
def test_get_options_deltas(self):
1478
index = self.two_graph_index(deltas=True)
1479
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1480
self.assertEqual(['line-delta'], index.get_options(('parent',)))
1482
def test_get_options_no_deltas(self):
1483
# check that the parent-history lookup is ignored with deltas=False.
1484
index = self.two_graph_index(deltas=False)
1485
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1486
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1488
def test_get_parent_map(self):
1489
index = self.two_graph_index()
1490
self.assertEqual({('parent',):(('tail',), ('ghost',))},
1491
index.get_parent_map([('parent',), ('ghost',)]))
1493
def catch_add(self, entries):
1494
self.caught_entries.append(entries)
1496
def test_add_no_callback_errors(self):
1497
index = self.two_graph_index()
1498
self.assertRaises(errors.ReadOnlyError, index.add_records,
1499
[(('new',), 'fulltext,no-eol', (None, 50, 60), ['separate'])])
1501
def test_add_version_smoke(self):
1502
index = self.two_graph_index(catch_adds=True)
1503
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60),
1505
self.assertEqual([[(('new', ), 'N50 60', ((('separate',),),))]],
1506
self.caught_entries)
1508
def test_add_version_delta_not_delta_index(self):
1509
index = self.two_graph_index(catch_adds=True)
1510
self.assertRaises(errors.KnitCorrupt, index.add_records,
1511
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1512
self.assertEqual([], self.caught_entries)
1514
def test_add_version_same_dup(self):
1515
index = self.two_graph_index(catch_adds=True)
1516
# options can be spelt two different ways
1517
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
1518
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
1519
# position/length are ignored (because each pack could have fulltext or
1520
# delta, and be at a different position.
1521
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1523
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1525
# but neither should have added data:
1526
self.assertEqual([[], [], [], []], self.caught_entries)
1528
def test_add_version_different_dup(self):
1529
index = self.two_graph_index(deltas=True, catch_adds=True)
1531
self.assertRaises(errors.KnitCorrupt, index.add_records,
1532
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1533
self.assertRaises(errors.KnitCorrupt, index.add_records,
1534
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1536
self.assertRaises(errors.KnitCorrupt, index.add_records,
1537
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1538
self.assertEqual([], self.caught_entries)
1540
def test_add_versions_nodeltas(self):
1541
index = self.two_graph_index(catch_adds=True)
1543
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1544
(('new2',), 'fulltext', (None, 0, 6), [('new',)]),
1546
self.assertEqual([(('new', ), 'N50 60', ((('separate',),),)),
1547
(('new2', ), ' 0 6', ((('new',),),))],
1548
sorted(self.caught_entries[0]))
1549
self.assertEqual(1, len(self.caught_entries))
1551
def test_add_versions_deltas(self):
1552
index = self.two_graph_index(deltas=True, catch_adds=True)
1554
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1555
(('new2',), 'line-delta', (None, 0, 6), [('new',)]),
1557
self.assertEqual([(('new', ), 'N50 60', ((('separate',),), ())),
1558
(('new2', ), ' 0 6', ((('new',),), (('new',),), ))],
1559
sorted(self.caught_entries[0]))
1560
self.assertEqual(1, len(self.caught_entries))
1562
def test_add_versions_delta_not_delta_index(self):
1563
index = self.two_graph_index(catch_adds=True)
1564
self.assertRaises(errors.KnitCorrupt, index.add_records,
1565
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1566
self.assertEqual([], self.caught_entries)
1568
def test_add_versions_random_id_accepted(self):
1569
index = self.two_graph_index(catch_adds=True)
1570
index.add_records([], random_id=True)
1572
def test_add_versions_same_dup(self):
1573
index = self.two_graph_index(catch_adds=True)
1574
# options can be spelt two different ways
1575
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100),
1577
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100),
1579
# position/length are ignored (because each pack could have fulltext or
1580
# delta, and be at a different position.
1581
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1583
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1585
# but neither should have added data.
1586
self.assertEqual([[], [], [], []], self.caught_entries)
1588
def test_add_versions_different_dup(self):
1589
index = self.two_graph_index(deltas=True, catch_adds=True)
1591
self.assertRaises(errors.KnitCorrupt, index.add_records,
1592
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1593
self.assertRaises(errors.KnitCorrupt, index.add_records,
1594
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1596
self.assertRaises(errors.KnitCorrupt, index.add_records,
1597
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1598
# change options in the second record
1599
self.assertRaises(errors.KnitCorrupt, index.add_records,
1600
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)]),
1601
(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1602
self.assertEqual([], self.caught_entries)
1605
class TestNoParentsGraphIndexKnit(KnitTests):
1606
"""Tests for knits using _KnitGraphIndex with no parents."""
1608
def make_g_index(self, name, ref_lists=0, nodes=[]):
1609
builder = GraphIndexBuilder(ref_lists)
1610
for node, references in nodes:
1611
builder.add_node(node, references)
1612
stream = builder.finish()
1613
trans = self.get_transport()
1614
size = trans.put_file(name, stream)
1615
return GraphIndex(trans, name, size)
1617
def test_parents_deltas_incompatible(self):
1618
index = CombinedGraphIndex([])
1619
self.assertRaises(errors.KnitError, _KnitGraphIndex, lambda:True,
1620
index, deltas=True, parents=False)
1622
def two_graph_index(self, catch_adds=False):
1623
"""Build a two-graph index.
1625
:param deltas: If true, use underlying indices with two node-ref
1626
lists and 'parent' set to a delta-compressed against tail.
1628
# put several versions in the index.
1629
index1 = self.make_g_index('1', 0, [
1630
(('tip', ), 'N0 100'),
1632
index2 = self.make_g_index('2', 0, [
1633
(('parent', ), ' 100 78'),
1634
(('separate', ), '')])
1635
combined_index = CombinedGraphIndex([index1, index2])
1637
self.combined_index = combined_index
1638
self.caught_entries = []
1639
add_callback = self.catch_add
1642
return _KnitGraphIndex(combined_index, lambda:True, parents=False,
1643
add_callback=add_callback)
1645
def test_keys(self):
1646
index = self.two_graph_index()
1647
self.assertEqual(set([('tail',), ('tip',), ('parent',), ('separate',)]),
1650
def test_get_position(self):
1651
index = self.two_graph_index()
1652
self.assertEqual((index._graph_index._indices[0], 0, 100),
1653
index.get_position(('tip',)))
1654
self.assertEqual((index._graph_index._indices[1], 100, 78),
1655
index.get_position(('parent',)))
1657
def test_get_method(self):
1658
index = self.two_graph_index()
1659
self.assertEqual('fulltext', index.get_method(('tip',)))
1660
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1662
def test_get_options(self):
1663
index = self.two_graph_index()
1664
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1665
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1667
def test_get_parent_map(self):
1668
index = self.two_graph_index()
1669
self.assertEqual({('parent',):None},
1670
index.get_parent_map([('parent',), ('ghost',)]))
1672
def catch_add(self, entries):
1673
self.caught_entries.append(entries)
1675
def test_add_no_callback_errors(self):
1676
index = self.two_graph_index()
1677
self.assertRaises(errors.ReadOnlyError, index.add_records,
1678
[(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)])])
1680
def test_add_version_smoke(self):
1681
index = self.two_graph_index(catch_adds=True)
1682
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60), [])])
1683
self.assertEqual([[(('new', ), 'N50 60')]],
1684
self.caught_entries)
1686
def test_add_version_delta_not_delta_index(self):
1687
index = self.two_graph_index(catch_adds=True)
1688
self.assertRaises(errors.KnitCorrupt, index.add_records,
1689
[(('new',), 'no-eol,line-delta', (None, 0, 100), [])])
1690
self.assertEqual([], self.caught_entries)
1692
def test_add_version_same_dup(self):
1693
index = self.two_graph_index(catch_adds=True)
1694
# options can be spelt two different ways
1695
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1696
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
1697
# position/length are ignored (because each pack could have fulltext or
1698
# delta, and be at a different position.
1699
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
1700
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
1701
# but neither should have added data.
1702
self.assertEqual([[], [], [], []], self.caught_entries)
1704
def test_add_version_different_dup(self):
1705
index = self.two_graph_index(catch_adds=True)
1707
self.assertRaises(errors.KnitCorrupt, index.add_records,
1708
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
1709
self.assertRaises(errors.KnitCorrupt, index.add_records,
1710
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
1711
self.assertRaises(errors.KnitCorrupt, index.add_records,
1712
[(('tip',), 'fulltext', (None, 0, 100), [])])
1714
self.assertRaises(errors.KnitCorrupt, index.add_records,
1715
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
1716
self.assertEqual([], self.caught_entries)
1718
def test_add_versions(self):
1719
index = self.two_graph_index(catch_adds=True)
1721
(('new',), 'fulltext,no-eol', (None, 50, 60), []),
1722
(('new2',), 'fulltext', (None, 0, 6), []),
1724
self.assertEqual([(('new', ), 'N50 60'), (('new2', ), ' 0 6')],
1725
sorted(self.caught_entries[0]))
1726
self.assertEqual(1, len(self.caught_entries))
1728
def test_add_versions_delta_not_delta_index(self):
1729
index = self.two_graph_index(catch_adds=True)
1730
self.assertRaises(errors.KnitCorrupt, index.add_records,
1731
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1732
self.assertEqual([], self.caught_entries)
1734
def test_add_versions_parents_not_parents_index(self):
1735
index = self.two_graph_index(catch_adds=True)
1736
self.assertRaises(errors.KnitCorrupt, index.add_records,
1737
[(('new',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
1738
self.assertEqual([], self.caught_entries)
1740
def test_add_versions_random_id_accepted(self):
1741
index = self.two_graph_index(catch_adds=True)
1742
index.add_records([], random_id=True)
1744
def test_add_versions_same_dup(self):
1745
index = self.two_graph_index(catch_adds=True)
1746
# options can be spelt two different ways
1747
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1748
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
1749
# position/length are ignored (because each pack could have fulltext or
1750
# delta, and be at a different position.
1751
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
1752
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
1753
# but neither should have added data.
1754
self.assertEqual([[], [], [], []], self.caught_entries)
1756
def test_add_versions_different_dup(self):
1757
index = self.two_graph_index(catch_adds=True)
1759
self.assertRaises(errors.KnitCorrupt, index.add_records,
1760
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
1761
self.assertRaises(errors.KnitCorrupt, index.add_records,
1762
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
1763
self.assertRaises(errors.KnitCorrupt, index.add_records,
1764
[(('tip',), 'fulltext', (None, 0, 100), [])])
1766
self.assertRaises(errors.KnitCorrupt, index.add_records,
1767
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
1768
# change options in the second record
1769
self.assertRaises(errors.KnitCorrupt, index.add_records,
1770
[(('tip',), 'fulltext,no-eol', (None, 0, 100), []),
1771
(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
1772
self.assertEqual([], self.caught_entries)
1775
class TestStacking(KnitTests):
1777
def get_basis_and_test_knit(self):
1778
basis = self.make_test_knit(name='basis')
1779
basis = RecordingVersionedFilesDecorator(basis)
1780
test = self.make_test_knit(name='test')
1781
test.add_fallback_versioned_files(basis)
1784
def test_add_fallback_versioned_files(self):
1785
basis = self.make_test_knit(name='basis')
1786
test = self.make_test_knit(name='test')
1787
# It must not error; other tests test that the fallback is referred to
1788
# when accessing data.
1789
test.add_fallback_versioned_files(basis)
1791
def test_add_lines(self):
1792
# lines added to the test are not added to the basis
1793
basis, test = self.get_basis_and_test_knit()
1795
key_basis = ('bar',)
1796
key_cross_border = ('quux',)
1797
key_delta = ('zaphod',)
1798
test.add_lines(key, (), ['foo\n'])
1799
self.assertEqual({}, basis.get_parent_map([key]))
1800
# lines added to the test that reference across the stack do a
1802
basis.add_lines(key_basis, (), ['foo\n'])
1804
test.add_lines(key_cross_border, (key_basis,), ['foo\n'])
1805
self.assertEqual('fulltext', test._index.get_method(key_cross_border))
1806
# we don't even need to look at the basis to see that this should be
1807
# stored as a fulltext
1808
self.assertEqual([], basis.calls)
1809
# Subsequent adds do delta.
1811
test.add_lines(key_delta, (key_cross_border,), ['foo\n'])
1812
self.assertEqual('line-delta', test._index.get_method(key_delta))
1813
self.assertEqual([], basis.calls)
1815
def test_annotate(self):
1816
# annotations from the test knit are answered without asking the basis
1817
basis, test = self.get_basis_and_test_knit()
1819
key_basis = ('bar',)
1820
key_missing = ('missing',)
1821
test.add_lines(key, (), ['foo\n'])
1822
details = test.annotate(key)
1823
self.assertEqual([(key, 'foo\n')], details)
1824
self.assertEqual([], basis.calls)
1825
# But texts that are not in the test knit are looked for in the basis
1827
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
1829
details = test.annotate(key_basis)
1830
self.assertEqual([(key_basis, 'foo\n'), (key_basis, 'bar\n')], details)
1831
# Not optimised to date:
1832
# self.assertEqual([("annotate", key_basis)], basis.calls)
1833
self.assertEqual([('get_parent_map', set([key_basis])),
1834
('get_parent_map', set([key_basis])),
1835
('get_parent_map', set([key_basis])),
1836
('get_record_stream', [key_basis], 'unordered', True)],
1839
def test_check(self):
1840
# At the moment checking a stacked knit does implicitly check the
1842
basis, test = self.get_basis_and_test_knit()
1845
def test_get_parent_map(self):
1846
# parents in the test knit are answered without asking the basis
1847
basis, test = self.get_basis_and_test_knit()
1849
key_basis = ('bar',)
1850
key_missing = ('missing',)
1851
test.add_lines(key, (), [])
1852
parent_map = test.get_parent_map([key])
1853
self.assertEqual({key: ()}, parent_map)
1854
self.assertEqual([], basis.calls)
1855
# But parents that are not in the test knit are looked for in the basis
1856
basis.add_lines(key_basis, (), [])
1858
parent_map = test.get_parent_map([key, key_basis, key_missing])
1859
self.assertEqual({key: (),
1860
key_basis: ()}, parent_map)
1861
self.assertEqual([("get_parent_map", set([key_basis, key_missing]))],
1864
def test_get_record_stream_unordered_fulltexts(self):
1865
# records from the test knit are answered without asking the basis:
1866
basis, test = self.get_basis_and_test_knit()
1868
key_basis = ('bar',)
1869
key_missing = ('missing',)
1870
test.add_lines(key, (), ['foo\n'])
1871
records = list(test.get_record_stream([key], 'unordered', True))
1872
self.assertEqual(1, len(records))
1873
self.assertEqual([], basis.calls)
1874
# Missing (from test knit) objects are retrieved from the basis:
1875
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
1877
records = list(test.get_record_stream([key_basis, key_missing],
1879
self.assertEqual(2, len(records))
1880
calls = list(basis.calls)
1881
for record in records:
1882
self.assertSubset([record.key], (key_basis, key_missing))
1883
if record.key == key_missing:
1884
self.assertIsInstance(record, AbsentContentFactory)
1886
reference = list(basis.get_record_stream([key_basis],
1887
'unordered', True))[0]
1888
self.assertEqual(reference.key, record.key)
1889
self.assertEqual(reference.sha1, record.sha1)
1890
self.assertEqual(reference.storage_kind, record.storage_kind)
1891
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
1892
record.get_bytes_as(record.storage_kind))
1893
self.assertEqual(reference.get_bytes_as('fulltext'),
1894
record.get_bytes_as('fulltext'))
1895
# It's not strictly minimal, but it seems reasonable for now for it to
1896
# ask which fallbacks have which parents.
1898
("get_parent_map", set([key_basis, key_missing])),
1899
("get_record_stream", [key_basis], 'unordered', True)],
1902
def test_get_record_stream_ordered_fulltexts(self):
1903
# ordering is preserved down into the fallback store.
1904
basis, test = self.get_basis_and_test_knit()
1906
key_basis = ('bar',)
1907
key_basis_2 = ('quux',)
1908
key_missing = ('missing',)
1909
test.add_lines(key, (key_basis,), ['foo\n'])
1910
# Missing (from test knit) objects are retrieved from the basis:
1911
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
1912
basis.add_lines(key_basis_2, (), ['quux\n'])
1914
# ask for in non-topological order
1915
records = list(test.get_record_stream(
1916
[key, key_basis, key_missing, key_basis_2], 'topological', True))
1917
self.assertEqual(4, len(records))
1919
for record in records:
1920
self.assertSubset([record.key],
1921
(key_basis, key_missing, key_basis_2, key))
1922
if record.key == key_missing:
1923
self.assertIsInstance(record, AbsentContentFactory)
1925
results.append((record.key, record.sha1, record.storage_kind,
1926
record.get_bytes_as('fulltext')))
1927
calls = list(basis.calls)
1928
order = [record[0] for record in results]
1929
self.assertEqual([key_basis_2, key_basis, key], order)
1930
for result in results:
1931
if result[0] == key:
1935
record = source.get_record_stream([result[0]], 'unordered',
1937
self.assertEqual(record.key, result[0])
1938
self.assertEqual(record.sha1, result[1])
1939
self.assertEqual(record.storage_kind, result[2])
1940
self.assertEqual(record.get_bytes_as('fulltext'), result[3])
1941
# It's not strictly minimal, but it seems reasonable for now for it to
1942
# ask which fallbacks have which parents.
1944
("get_parent_map", set([key_basis, key_basis_2, key_missing])),
1945
# unordered is asked for by the underlying worker as it still
1946
# buffers everything while answering - which is a problem!
1947
("get_record_stream", [key_basis_2, key_basis], 'unordered', True)],
1950
def test_get_record_stream_unordered_deltas(self):
1951
# records from the test knit are answered without asking the basis:
1952
basis, test = self.get_basis_and_test_knit()
1954
key_basis = ('bar',)
1955
key_missing = ('missing',)
1956
test.add_lines(key, (), ['foo\n'])
1957
records = list(test.get_record_stream([key], 'unordered', False))
1958
self.assertEqual(1, len(records))
1959
self.assertEqual([], basis.calls)
1960
# Missing (from test knit) objects are retrieved from the basis:
1961
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
1963
records = list(test.get_record_stream([key_basis, key_missing],
1964
'unordered', False))
1965
self.assertEqual(2, len(records))
1966
calls = list(basis.calls)
1967
for record in records:
1968
self.assertSubset([record.key], (key_basis, key_missing))
1969
if record.key == key_missing:
1970
self.assertIsInstance(record, AbsentContentFactory)
1972
reference = list(basis.get_record_stream([key_basis],
1973
'unordered', False))[0]
1974
self.assertEqual(reference.key, record.key)
1975
self.assertEqual(reference.sha1, record.sha1)
1976
self.assertEqual(reference.storage_kind, record.storage_kind)
1977
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
1978
record.get_bytes_as(record.storage_kind))
1979
# It's not strictly minimal, but it seems reasonable for now for it to
1980
# ask which fallbacks have which parents.
1982
("get_parent_map", set([key_basis, key_missing])),
1983
("get_record_stream", [key_basis], 'unordered', False)],
1986
def test_get_record_stream_ordered_deltas(self):
1987
# ordering is preserved down into the fallback store.
1988
basis, test = self.get_basis_and_test_knit()
1990
key_basis = ('bar',)
1991
key_basis_2 = ('quux',)
1992
key_missing = ('missing',)
1993
test.add_lines(key, (key_basis,), ['foo\n'])
1994
# Missing (from test knit) objects are retrieved from the basis:
1995
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
1996
basis.add_lines(key_basis_2, (), ['quux\n'])
1998
# ask for in non-topological order
1999
records = list(test.get_record_stream(
2000
[key, key_basis, key_missing, key_basis_2], 'topological', False))
2001
self.assertEqual(4, len(records))
2003
for record in records:
2004
self.assertSubset([record.key],
2005
(key_basis, key_missing, key_basis_2, key))
2006
if record.key == key_missing:
2007
self.assertIsInstance(record, AbsentContentFactory)
2009
results.append((record.key, record.sha1, record.storage_kind,
2010
record.get_bytes_as(record.storage_kind)))
2011
calls = list(basis.calls)
2012
order = [record[0] for record in results]
2013
self.assertEqual([key_basis_2, key_basis, key], order)
2014
for result in results:
2015
if result[0] == key:
2019
record = source.get_record_stream([result[0]], 'unordered',
2021
self.assertEqual(record.key, result[0])
2022
self.assertEqual(record.sha1, result[1])
2023
self.assertEqual(record.storage_kind, result[2])
2024
self.assertEqual(record.get_bytes_as(record.storage_kind), result[3])
2025
# It's not strictly minimal, but it seems reasonable for now for it to
2026
# ask which fallbacks have which parents.
2028
("get_parent_map", set([key_basis, key_basis_2, key_missing])),
2029
("get_record_stream", [key_basis_2, key_basis], 'topological', False)],
2032
def test_get_sha1s(self):
2033
# sha1's in the test knit are answered without asking the basis
2034
basis, test = self.get_basis_and_test_knit()
2036
key_basis = ('bar',)
2037
key_missing = ('missing',)
2038
test.add_lines(key, (), ['foo\n'])
2039
key_sha1sum = osutils.sha('foo\n').hexdigest()
2040
sha1s = test.get_sha1s([key])
2041
self.assertEqual({key: key_sha1sum}, sha1s)
2042
self.assertEqual([], basis.calls)
2043
# But texts that are not in the test knit are looked for in the basis
2044
# directly (rather than via text reconstruction) so that remote servers
2045
# etc don't have to answer with full content.
2046
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2047
basis_sha1sum = osutils.sha('foo\nbar\n').hexdigest()
2049
sha1s = test.get_sha1s([key, key_missing, key_basis])
2050
self.assertEqual({key: key_sha1sum,
2051
key_basis: basis_sha1sum}, sha1s)
2052
self.assertEqual([("get_sha1s", set([key_basis, key_missing]))],
2055
def test_insert_record_stream(self):
2056
# records are inserted as normal; insert_record_stream builds on
2057
# add_lines, so a smoke test should be all that's needed:
2059
key_basis = ('bar',)
2060
key_delta = ('zaphod',)
2061
basis, test = self.get_basis_and_test_knit()
2062
source = self.make_test_knit(name='source')
2063
basis.add_lines(key_basis, (), ['foo\n'])
2065
source.add_lines(key_basis, (), ['foo\n'])
2066
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2067
stream = source.get_record_stream([key_delta], 'unordered', False)
2068
test.insert_record_stream(stream)
2069
# XXX: this does somewhat too many calls in making sure of whether it
2070
# has to recreate the full text.
2071
self.assertEqual([("get_parent_map", set([key_basis])),
2072
('get_parent_map', set([key_basis])),
2073
('get_record_stream', [key_basis], 'unordered', True)],
2075
self.assertEqual({key_delta:(key_basis,)},
2076
test.get_parent_map([key_delta]))
2077
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2078
'unordered', True).next().get_bytes_as('fulltext'))
2080
def test_iter_lines_added_or_present_in_keys(self):
2081
# Lines from the basis are returned, and lines for a given key are only
2085
# all sources are asked for keys:
2086
basis, test = self.get_basis_and_test_knit()
2087
basis.add_lines(key1, (), ["foo"])
2089
lines = list(test.iter_lines_added_or_present_in_keys([key1]))
2090
self.assertEqual([("foo\n", key1)], lines)
2091
self.assertEqual([("iter_lines_added_or_present_in_keys", set([key1]))],
2093
# keys in both are not duplicated:
2094
test.add_lines(key2, (), ["bar\n"])
2095
basis.add_lines(key2, (), ["bar\n"])
2097
lines = list(test.iter_lines_added_or_present_in_keys([key2]))
2098
self.assertEqual([("bar\n", key2)], lines)
2099
self.assertEqual([], basis.calls)
2101
def test_keys(self):
2104
# all sources are asked for keys:
2105
basis, test = self.get_basis_and_test_knit()
2107
self.assertEqual(set(), set(keys))
2108
self.assertEqual([("keys",)], basis.calls)
2109
# keys from a basis are returned:
2110
basis.add_lines(key1, (), [])
2113
self.assertEqual(set([key1]), set(keys))
2114
self.assertEqual([("keys",)], basis.calls)
2115
# keys in both are not duplicated:
2116
test.add_lines(key2, (), [])
2117
basis.add_lines(key2, (), [])
2120
self.assertEqual(2, len(keys))
2121
self.assertEqual(set([key1, key2]), set(keys))
2122
self.assertEqual([("keys",)], basis.calls)
2124
def test_add_mpdiffs(self):
2125
# records are inserted as normal; add_mpdiff builds on
2126
# add_lines, so a smoke test should be all that's needed:
2128
key_basis = ('bar',)
2129
key_delta = ('zaphod',)
2130
basis, test = self.get_basis_and_test_knit()
2131
source = self.make_test_knit(name='source')
2132
basis.add_lines(key_basis, (), ['foo\n'])
2134
source.add_lines(key_basis, (), ['foo\n'])
2135
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2136
diffs = source.make_mpdiffs([key_delta])
2137
test.add_mpdiffs([(key_delta, (key_basis,),
2138
source.get_sha1s([key_delta])[key_delta], diffs[0])])
2139
self.assertEqual([("get_parent_map", set([key_basis])),
2140
('get_record_stream', [key_basis], 'unordered', True),],
2142
self.assertEqual({key_delta:(key_basis,)},
2143
test.get_parent_map([key_delta]))
2144
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2145
'unordered', True).next().get_bytes_as('fulltext'))
2147
def test_make_mpdiffs(self):
2148
# Generating an mpdiff across a stacking boundary should detect parent
2152
key_right = ('zaphod',)
2153
basis, test = self.get_basis_and_test_knit()
2154
basis.add_lines(key_left, (), ['bar\n'])
2155
basis.add_lines(key_right, (), ['zaphod\n'])
2157
test.add_lines(key, (key_left, key_right),
2158
['bar\n', 'foo\n', 'zaphod\n'])
2159
diffs = test.make_mpdiffs([key])
2161
multiparent.MultiParent([multiparent.ParentText(0, 0, 0, 1),
2162
multiparent.NewText(['foo\n']),
2163
multiparent.ParentText(1, 0, 2, 1)])],
2165
self.assertEqual(3, len(basis.calls))
2167
("get_parent_map", set([key_left, key_right])),
2168
("get_parent_map", set([key_left, key_right])),
2171
last_call = basis.calls[-1]
2172
self.assertEqual('get_record_stream', last_call[0])
2173
self.assertEqual(set([key_left, key_right]), set(last_call[1]))
2174
self.assertEqual('unordered', last_call[2])
2175
self.assertEqual(True, last_call[3])
added
removed
bzrlib/tests/test_knit.py
