1
# Copyright (C) 2006-2010 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"""
19
from cStringIO import StringIO
33
from bzrlib.errors import (
34
RevisionAlreadyPresent,
39
from bzrlib.index import *
40
from bzrlib.knit import (
46
_VFContentMapGenerator,
53
from bzrlib.repofmt import pack_repo
54
from bzrlib.tests import (
58
TestCaseWithMemoryTransport,
59
TestCaseWithTransport,
62
from bzrlib.transport import get_transport
63
from bzrlib.transport.memory import MemoryTransport
64
from bzrlib.tuned_gzip import GzipFile
65
from bzrlib.versionedfile import (
68
network_bytes_to_kind_and_offset,
69
RecordingVersionedFilesDecorator,
73
compiled_knit_feature = tests.ModuleAvailableFeature(
74
'bzrlib._knit_load_data_pyx')
77
class KnitContentTestsMixin(object):
79
def test_constructor(self):
80
content = self._make_content([])
83
content = self._make_content([])
84
self.assertEqual(content.text(), [])
86
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
87
self.assertEqual(content.text(), ["text1", "text2"])
90
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
92
self.assertIsInstance(copy, content.__class__)
93
self.assertEqual(copy.annotate(), content.annotate())
95
def assertDerivedBlocksEqual(self, source, target, noeol=False):
96
"""Assert that the derived matching blocks match real output"""
97
source_lines = source.splitlines(True)
98
target_lines = target.splitlines(True)
100
if noeol and not line.endswith('\n'):
104
source_content = self._make_content([(None, nl(l)) for l in source_lines])
105
target_content = self._make_content([(None, nl(l)) for l in target_lines])
106
line_delta = source_content.line_delta(target_content)
107
delta_blocks = list(KnitContent.get_line_delta_blocks(line_delta,
108
source_lines, target_lines))
109
matcher = KnitSequenceMatcher(None, source_lines, target_lines)
110
matcher_blocks = list(list(matcher.get_matching_blocks()))
111
self.assertEqual(matcher_blocks, delta_blocks)
113
def test_get_line_delta_blocks(self):
114
self.assertDerivedBlocksEqual('a\nb\nc\n', 'q\nc\n')
115
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1)
116
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1A)
117
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1B)
118
self.assertDerivedBlocksEqual(TEXT_1B, TEXT_1A)
119
self.assertDerivedBlocksEqual(TEXT_1A, TEXT_1B)
120
self.assertDerivedBlocksEqual(TEXT_1A, '')
121
self.assertDerivedBlocksEqual('', TEXT_1A)
122
self.assertDerivedBlocksEqual('', '')
123
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd')
125
def test_get_line_delta_blocks_noeol(self):
126
"""Handle historical knit deltas safely
128
Some existing knit deltas don't consider the last line to differ
129
when the only difference whether it has a final newline.
131
New knit deltas appear to always consider the last line to differ
134
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd\n', noeol=True)
135
self.assertDerivedBlocksEqual('a\nb\nc\nd\n', 'a\nb\nc', noeol=True)
136
self.assertDerivedBlocksEqual('a\nb\nc\n', 'a\nb\nc', noeol=True)
137
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\n', noeol=True)
149
Banana cup cake recipe
159
Banana cup cake recipe
161
- bananas (do not use plantains!!!)
168
Banana cup cake recipe
185
class TestPlainKnitContent(TestCase, KnitContentTestsMixin):
187
def _make_content(self, lines):
188
annotated_content = AnnotatedKnitContent(lines)
189
return PlainKnitContent(annotated_content.text(), 'bogus')
191
def test_annotate(self):
192
content = self._make_content([])
193
self.assertEqual(content.annotate(), [])
195
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
196
self.assertEqual(content.annotate(),
197
[("bogus", "text1"), ("bogus", "text2")])
199
def test_line_delta(self):
200
content1 = self._make_content([("", "a"), ("", "b")])
201
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
202
self.assertEqual(content1.line_delta(content2),
203
[(1, 2, 2, ["a", "c"])])
205
def test_line_delta_iter(self):
206
content1 = self._make_content([("", "a"), ("", "b")])
207
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
208
it = content1.line_delta_iter(content2)
209
self.assertEqual(it.next(), (1, 2, 2, ["a", "c"]))
210
self.assertRaises(StopIteration, it.next)
213
class TestAnnotatedKnitContent(TestCase, KnitContentTestsMixin):
215
def _make_content(self, lines):
216
return AnnotatedKnitContent(lines)
218
def test_annotate(self):
219
content = self._make_content([])
220
self.assertEqual(content.annotate(), [])
222
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
223
self.assertEqual(content.annotate(),
224
[("origin1", "text1"), ("origin2", "text2")])
226
def test_line_delta(self):
227
content1 = self._make_content([("", "a"), ("", "b")])
228
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
229
self.assertEqual(content1.line_delta(content2),
230
[(1, 2, 2, [("", "a"), ("", "c")])])
232
def test_line_delta_iter(self):
233
content1 = self._make_content([("", "a"), ("", "b")])
234
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
235
it = content1.line_delta_iter(content2)
236
self.assertEqual(it.next(), (1, 2, 2, [("", "a"), ("", "c")]))
237
self.assertRaises(StopIteration, it.next)
240
class MockTransport(object):
242
def __init__(self, file_lines=None):
243
self.file_lines = file_lines
245
# We have no base directory for the MockTransport
248
def get(self, filename):
249
if self.file_lines is None:
250
raise NoSuchFile(filename)
252
return StringIO("\n".join(self.file_lines))
254
def readv(self, relpath, offsets):
255
fp = self.get(relpath)
256
for offset, size in offsets:
258
yield offset, fp.read(size)
260
def __getattr__(self, name):
261
def queue_call(*args, **kwargs):
262
self.calls.append((name, args, kwargs))
266
class MockReadvFailingTransport(MockTransport):
267
"""Fail in the middle of a readv() result.
269
This Transport will successfully yield the first two requested hunks, but
270
raise NoSuchFile for the rest.
273
def readv(self, relpath, offsets):
275
for result in MockTransport.readv(self, relpath, offsets):
277
# we use 2 because the first offset is the pack header, the second
278
# is the first actual content requset
280
raise errors.NoSuchFile(relpath)
284
class KnitRecordAccessTestsMixin(object):
285
"""Tests for getting and putting knit records."""
287
def test_add_raw_records(self):
288
"""Add_raw_records adds records retrievable later."""
289
access = self.get_access()
290
memos = access.add_raw_records([('key', 10)], '1234567890')
291
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
293
def test_add_several_raw_records(self):
294
"""add_raw_records with many records and read some back."""
295
access = self.get_access()
296
memos = access.add_raw_records([('key', 10), ('key2', 2), ('key3', 5)],
298
self.assertEqual(['1234567890', '12', '34567'],
299
list(access.get_raw_records(memos)))
300
self.assertEqual(['1234567890'],
301
list(access.get_raw_records(memos[0:1])))
302
self.assertEqual(['12'],
303
list(access.get_raw_records(memos[1:2])))
304
self.assertEqual(['34567'],
305
list(access.get_raw_records(memos[2:3])))
306
self.assertEqual(['1234567890', '34567'],
307
list(access.get_raw_records(memos[0:1] + memos[2:3])))
310
class TestKnitKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
311
"""Tests for the .kndx implementation."""
313
def get_access(self):
314
"""Get a .knit style access instance."""
315
mapper = ConstantMapper("foo")
316
access = _KnitKeyAccess(self.get_transport(), mapper)
320
class _TestException(Exception):
321
"""Just an exception for local tests to use."""
324
class TestPackKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
325
"""Tests for the pack based access."""
327
def get_access(self):
328
return self._get_access()[0]
330
def _get_access(self, packname='packfile', index='FOO'):
331
transport = self.get_transport()
332
def write_data(bytes):
333
transport.append_bytes(packname, bytes)
334
writer = pack.ContainerWriter(write_data)
336
access = _DirectPackAccess({})
337
access.set_writer(writer, index, (transport, packname))
338
return access, writer
340
def make_pack_file(self):
341
"""Create a pack file with 2 records."""
342
access, writer = self._get_access(packname='packname', index='foo')
344
memos.extend(access.add_raw_records([('key1', 10)], '1234567890'))
345
memos.extend(access.add_raw_records([('key2', 5)], '12345'))
349
def make_vf_for_retrying(self):
350
"""Create 3 packs and a reload function.
352
Originally, 2 pack files will have the data, but one will be missing.
353
And then the third will be used in place of the first two if reload()
356
:return: (versioned_file, reload_counter)
357
versioned_file a KnitVersionedFiles using the packs for access
359
builder = self.make_branch_builder('.', format="1.9")
360
builder.start_series()
361
builder.build_snapshot('rev-1', None, [
362
('add', ('', 'root-id', 'directory', None)),
363
('add', ('file', 'file-id', 'file', 'content\nrev 1\n')),
365
builder.build_snapshot('rev-2', ['rev-1'], [
366
('modify', ('file-id', 'content\nrev 2\n')),
368
builder.build_snapshot('rev-3', ['rev-2'], [
369
('modify', ('file-id', 'content\nrev 3\n')),
371
builder.finish_series()
372
b = builder.get_branch()
374
self.addCleanup(b.unlock)
375
# Pack these three revisions into another pack file, but don't remove
378
collection = repo._pack_collection
379
collection.ensure_loaded()
380
orig_packs = collection.packs
381
packer = pack_repo.Packer(collection, orig_packs, '.testpack')
382
new_pack = packer.pack()
383
# forget about the new pack
387
# Set up a reload() function that switches to using the new pack file
388
new_index = new_pack.revision_index
389
access_tuple = new_pack.access_tuple()
390
reload_counter = [0, 0, 0]
392
reload_counter[0] += 1
393
if reload_counter[1] > 0:
394
# We already reloaded, nothing more to do
395
reload_counter[2] += 1
397
reload_counter[1] += 1
398
vf._index._graph_index._indices[:] = [new_index]
399
vf._access._indices.clear()
400
vf._access._indices[new_index] = access_tuple
402
# Delete one of the pack files so the data will need to be reloaded. We
403
# will delete the file with 'rev-2' in it
404
trans, name = orig_packs[1].access_tuple()
406
# We don't have the index trigger reloading because we want to test
407
# that we reload when the .pack disappears
408
vf._access._reload_func = reload
409
return vf, reload_counter
411
def make_reload_func(self, return_val=True):
414
reload_called[0] += 1
416
return reload_called, reload
418
def make_retry_exception(self):
419
# We raise a real exception so that sys.exc_info() is properly
422
raise _TestException('foobar')
423
except _TestException, e:
424
retry_exc = errors.RetryWithNewPacks(None, reload_occurred=False,
425
exc_info=sys.exc_info())
428
def test_read_from_several_packs(self):
429
access, writer = self._get_access()
431
memos.extend(access.add_raw_records([('key', 10)], '1234567890'))
433
access, writer = self._get_access('pack2', 'FOOBAR')
434
memos.extend(access.add_raw_records([('key', 5)], '12345'))
436
access, writer = self._get_access('pack3', 'BAZ')
437
memos.extend(access.add_raw_records([('key', 5)], 'alpha'))
439
transport = self.get_transport()
440
access = _DirectPackAccess({"FOO":(transport, 'packfile'),
441
"FOOBAR":(transport, 'pack2'),
442
"BAZ":(transport, 'pack3')})
443
self.assertEqual(['1234567890', '12345', 'alpha'],
444
list(access.get_raw_records(memos)))
445
self.assertEqual(['1234567890'],
446
list(access.get_raw_records(memos[0:1])))
447
self.assertEqual(['12345'],
448
list(access.get_raw_records(memos[1:2])))
449
self.assertEqual(['alpha'],
450
list(access.get_raw_records(memos[2:3])))
451
self.assertEqual(['1234567890', 'alpha'],
452
list(access.get_raw_records(memos[0:1] + memos[2:3])))
454
def test_set_writer(self):
455
"""The writer should be settable post construction."""
456
access = _DirectPackAccess({})
457
transport = self.get_transport()
458
packname = 'packfile'
460
def write_data(bytes):
461
transport.append_bytes(packname, bytes)
462
writer = pack.ContainerWriter(write_data)
464
access.set_writer(writer, index, (transport, packname))
465
memos = access.add_raw_records([('key', 10)], '1234567890')
467
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
469
def test_missing_index_raises_retry(self):
470
memos = self.make_pack_file()
471
transport = self.get_transport()
472
reload_called, reload_func = self.make_reload_func()
473
# Note that the index key has changed from 'foo' to 'bar'
474
access = _DirectPackAccess({'bar':(transport, 'packname')},
475
reload_func=reload_func)
476
e = self.assertListRaises(errors.RetryWithNewPacks,
477
access.get_raw_records, memos)
478
# Because a key was passed in which does not match our index list, we
479
# assume that the listing was already reloaded
480
self.assertTrue(e.reload_occurred)
481
self.assertIsInstance(e.exc_info, tuple)
482
self.assertIs(e.exc_info[0], KeyError)
483
self.assertIsInstance(e.exc_info[1], KeyError)
485
def test_missing_index_raises_key_error_with_no_reload(self):
486
memos = self.make_pack_file()
487
transport = self.get_transport()
488
# Note that the index key has changed from 'foo' to 'bar'
489
access = _DirectPackAccess({'bar':(transport, 'packname')})
490
e = self.assertListRaises(KeyError, access.get_raw_records, memos)
492
def test_missing_file_raises_retry(self):
493
memos = self.make_pack_file()
494
transport = self.get_transport()
495
reload_called, reload_func = self.make_reload_func()
496
# Note that the 'filename' has been changed to 'different-packname'
497
access = _DirectPackAccess({'foo':(transport, 'different-packname')},
498
reload_func=reload_func)
499
e = self.assertListRaises(errors.RetryWithNewPacks,
500
access.get_raw_records, memos)
501
# The file has gone missing, so we assume we need to reload
502
self.assertFalse(e.reload_occurred)
503
self.assertIsInstance(e.exc_info, tuple)
504
self.assertIs(e.exc_info[0], errors.NoSuchFile)
505
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
506
self.assertEqual('different-packname', e.exc_info[1].path)
508
def test_missing_file_raises_no_such_file_with_no_reload(self):
509
memos = self.make_pack_file()
510
transport = self.get_transport()
511
# Note that the 'filename' has been changed to 'different-packname'
512
access = _DirectPackAccess({'foo':(transport, 'different-packname')})
513
e = self.assertListRaises(errors.NoSuchFile,
514
access.get_raw_records, memos)
516
def test_failing_readv_raises_retry(self):
517
memos = self.make_pack_file()
518
transport = self.get_transport()
519
failing_transport = MockReadvFailingTransport(
520
[transport.get_bytes('packname')])
521
reload_called, reload_func = self.make_reload_func()
522
access = _DirectPackAccess({'foo':(failing_transport, 'packname')},
523
reload_func=reload_func)
524
# Asking for a single record will not trigger the Mock failure
525
self.assertEqual(['1234567890'],
526
list(access.get_raw_records(memos[:1])))
527
self.assertEqual(['12345'],
528
list(access.get_raw_records(memos[1:2])))
529
# A multiple offset readv() will fail mid-way through
530
e = self.assertListRaises(errors.RetryWithNewPacks,
531
access.get_raw_records, memos)
532
# The file has gone missing, so we assume we need to reload
533
self.assertFalse(e.reload_occurred)
534
self.assertIsInstance(e.exc_info, tuple)
535
self.assertIs(e.exc_info[0], errors.NoSuchFile)
536
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
537
self.assertEqual('packname', e.exc_info[1].path)
539
def test_failing_readv_raises_no_such_file_with_no_reload(self):
540
memos = self.make_pack_file()
541
transport = self.get_transport()
542
failing_transport = MockReadvFailingTransport(
543
[transport.get_bytes('packname')])
544
reload_called, reload_func = self.make_reload_func()
545
access = _DirectPackAccess({'foo':(failing_transport, 'packname')})
546
# Asking for a single record will not trigger the Mock failure
547
self.assertEqual(['1234567890'],
548
list(access.get_raw_records(memos[:1])))
549
self.assertEqual(['12345'],
550
list(access.get_raw_records(memos[1:2])))
551
# A multiple offset readv() will fail mid-way through
552
e = self.assertListRaises(errors.NoSuchFile,
553
access.get_raw_records, memos)
555
def test_reload_or_raise_no_reload(self):
556
access = _DirectPackAccess({}, reload_func=None)
557
retry_exc = self.make_retry_exception()
558
# Without a reload_func, we will just re-raise the original exception
559
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
561
def test_reload_or_raise_reload_changed(self):
562
reload_called, reload_func = self.make_reload_func(return_val=True)
563
access = _DirectPackAccess({}, reload_func=reload_func)
564
retry_exc = self.make_retry_exception()
565
access.reload_or_raise(retry_exc)
566
self.assertEqual([1], reload_called)
567
retry_exc.reload_occurred=True
568
access.reload_or_raise(retry_exc)
569
self.assertEqual([2], reload_called)
571
def test_reload_or_raise_reload_no_change(self):
572
reload_called, reload_func = self.make_reload_func(return_val=False)
573
access = _DirectPackAccess({}, reload_func=reload_func)
574
retry_exc = self.make_retry_exception()
575
# If reload_occurred is False, then we consider it an error to have
576
# reload_func() return False (no changes).
577
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
578
self.assertEqual([1], reload_called)
579
retry_exc.reload_occurred=True
580
# If reload_occurred is True, then we assume nothing changed because
581
# it had changed earlier, but didn't change again
582
access.reload_or_raise(retry_exc)
583
self.assertEqual([2], reload_called)
585
def test_annotate_retries(self):
586
vf, reload_counter = self.make_vf_for_retrying()
587
# It is a little bit bogus to annotate the Revision VF, but it works,
588
# as we have ancestry stored there
590
reload_lines = vf.annotate(key)
591
self.assertEqual([1, 1, 0], reload_counter)
592
plain_lines = vf.annotate(key)
593
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
594
if reload_lines != plain_lines:
595
self.fail('Annotation was not identical with reloading.')
596
# Now delete the packs-in-use, which should trigger another reload, but
597
# this time we just raise an exception because we can't recover
598
for trans, name in vf._access._indices.itervalues():
600
self.assertRaises(errors.NoSuchFile, vf.annotate, key)
601
self.assertEqual([2, 1, 1], reload_counter)
603
def test__get_record_map_retries(self):
604
vf, reload_counter = self.make_vf_for_retrying()
605
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
606
records = vf._get_record_map(keys)
607
self.assertEqual(keys, sorted(records.keys()))
608
self.assertEqual([1, 1, 0], reload_counter)
609
# Now delete the packs-in-use, which should trigger another reload, but
610
# this time we just raise an exception because we can't recover
611
for trans, name in vf._access._indices.itervalues():
613
self.assertRaises(errors.NoSuchFile, vf._get_record_map, keys)
614
self.assertEqual([2, 1, 1], reload_counter)
616
def test_get_record_stream_retries(self):
617
vf, reload_counter = self.make_vf_for_retrying()
618
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
619
record_stream = vf.get_record_stream(keys, 'topological', False)
620
record = record_stream.next()
621
self.assertEqual(('rev-1',), record.key)
622
self.assertEqual([0, 0, 0], reload_counter)
623
record = record_stream.next()
624
self.assertEqual(('rev-2',), record.key)
625
self.assertEqual([1, 1, 0], reload_counter)
626
record = record_stream.next()
627
self.assertEqual(('rev-3',), record.key)
628
self.assertEqual([1, 1, 0], reload_counter)
629
# Now delete all pack files, and see that we raise the right error
630
for trans, name in vf._access._indices.itervalues():
632
self.assertListRaises(errors.NoSuchFile,
633
vf.get_record_stream, keys, 'topological', False)
635
def test_iter_lines_added_or_present_in_keys_retries(self):
636
vf, reload_counter = self.make_vf_for_retrying()
637
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
638
# Unfortunately, iter_lines_added_or_present_in_keys iterates the
639
# result in random order (determined by the iteration order from a
640
# set()), so we don't have any solid way to trigger whether data is
641
# read before or after. However we tried to delete the middle node to
642
# exercise the code well.
643
# What we care about is that all lines are always yielded, but not
646
reload_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
647
self.assertEqual([1, 1, 0], reload_counter)
648
# Now do it again, to make sure the result is equivalent
649
plain_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
650
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
651
self.assertEqual(plain_lines, reload_lines)
652
self.assertEqual(21, len(plain_lines))
653
# Now delete all pack files, and see that we raise the right error
654
for trans, name in vf._access._indices.itervalues():
656
self.assertListRaises(errors.NoSuchFile,
657
vf.iter_lines_added_or_present_in_keys, keys)
658
self.assertEqual([2, 1, 1], reload_counter)
660
def test_get_record_stream_yields_disk_sorted_order(self):
661
# if we get 'unordered' pick a semi-optimal order for reading. The
662
# order should be grouped by pack file, and then by position in file
663
repo = self.make_repository('test', format='pack-0.92')
665
self.addCleanup(repo.unlock)
666
repo.start_write_group()
668
vf.add_lines(('f-id', 'rev-5'), [('f-id', 'rev-4')], ['lines\n'])
669
vf.add_lines(('f-id', 'rev-1'), [], ['lines\n'])
670
vf.add_lines(('f-id', 'rev-2'), [('f-id', 'rev-1')], ['lines\n'])
671
repo.commit_write_group()
672
# We inserted them as rev-5, rev-1, rev-2, we should get them back in
674
stream = vf.get_record_stream([('f-id', 'rev-1'), ('f-id', 'rev-5'),
675
('f-id', 'rev-2')], 'unordered', False)
676
keys = [r.key for r in stream]
677
self.assertEqual([('f-id', 'rev-5'), ('f-id', 'rev-1'),
678
('f-id', 'rev-2')], keys)
679
repo.start_write_group()
680
vf.add_lines(('f-id', 'rev-4'), [('f-id', 'rev-3')], ['lines\n'])
681
vf.add_lines(('f-id', 'rev-3'), [('f-id', 'rev-2')], ['lines\n'])
682
vf.add_lines(('f-id', 'rev-6'), [('f-id', 'rev-5')], ['lines\n'])
683
repo.commit_write_group()
684
# Request in random order, to make sure the output order isn't based on
686
request_keys = set(('f-id', 'rev-%d' % i) for i in range(1, 7))
687
stream = vf.get_record_stream(request_keys, 'unordered', False)
688
keys = [r.key for r in stream]
689
# We want to get the keys back in disk order, but it doesn't matter
690
# which pack we read from first. So this can come back in 2 orders
691
alt1 = [('f-id', 'rev-%d' % i) for i in [4, 3, 6, 5, 1, 2]]
692
alt2 = [('f-id', 'rev-%d' % i) for i in [5, 1, 2, 4, 3, 6]]
693
if keys != alt1 and keys != alt2:
694
self.fail('Returned key order did not match either expected order.'
695
' expected %s or %s, not %s'
696
% (alt1, alt2, keys))
699
class LowLevelKnitDataTests(TestCase):
701
def create_gz_content(self, text):
703
gz_file = gzip.GzipFile(mode='wb', fileobj=sio)
706
return sio.getvalue()
708
def make_multiple_records(self):
709
"""Create the content for multiple records."""
710
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
712
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
717
record_1 = (0, len(gz_txt), sha1sum)
718
total_txt.append(gz_txt)
719
sha1sum = osutils.sha('baz\n').hexdigest()
720
gz_txt = self.create_gz_content('version rev-id-2 1 %s\n'
724
record_2 = (record_1[1], len(gz_txt), sha1sum)
725
total_txt.append(gz_txt)
726
return total_txt, record_1, record_2
728
def test_valid_knit_data(self):
729
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
730
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
735
transport = MockTransport([gz_txt])
736
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
737
knit = KnitVersionedFiles(None, access)
738
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
740
contents = list(knit._read_records_iter(records))
741
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'],
742
'4e48e2c9a3d2ca8a708cb0cc545700544efb5021')], contents)
744
raw_contents = list(knit._read_records_iter_raw(records))
745
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
747
def test_multiple_records_valid(self):
748
total_txt, record_1, record_2 = self.make_multiple_records()
749
transport = MockTransport([''.join(total_txt)])
750
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
751
knit = KnitVersionedFiles(None, access)
752
records = [(('rev-id-1',), (('rev-id-1',), record_1[0], record_1[1])),
753
(('rev-id-2',), (('rev-id-2',), record_2[0], record_2[1]))]
755
contents = list(knit._read_records_iter(records))
756
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'], record_1[2]),
757
(('rev-id-2',), ['baz\n'], record_2[2])],
760
raw_contents = list(knit._read_records_iter_raw(records))
761
self.assertEqual([(('rev-id-1',), total_txt[0], record_1[2]),
762
(('rev-id-2',), total_txt[1], record_2[2])],
765
def test_not_enough_lines(self):
766
sha1sum = osutils.sha('foo\n').hexdigest()
767
# record says 2 lines data says 1
768
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
772
transport = MockTransport([gz_txt])
773
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
774
knit = KnitVersionedFiles(None, access)
775
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
776
self.assertRaises(errors.KnitCorrupt, list,
777
knit._read_records_iter(records))
779
# read_records_iter_raw won't detect that sort of mismatch/corruption
780
raw_contents = list(knit._read_records_iter_raw(records))
781
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
783
def test_too_many_lines(self):
784
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
785
# record says 1 lines data says 2
786
gz_txt = self.create_gz_content('version rev-id-1 1 %s\n'
791
transport = MockTransport([gz_txt])
792
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
793
knit = KnitVersionedFiles(None, access)
794
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
795
self.assertRaises(errors.KnitCorrupt, list,
796
knit._read_records_iter(records))
798
# read_records_iter_raw won't detect that sort of mismatch/corruption
799
raw_contents = list(knit._read_records_iter_raw(records))
800
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
802
def test_mismatched_version_id(self):
803
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
804
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
809
transport = MockTransport([gz_txt])
810
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
811
knit = KnitVersionedFiles(None, access)
812
# We are asking for rev-id-2, but the data is rev-id-1
813
records = [(('rev-id-2',), (('rev-id-2',), 0, len(gz_txt)))]
814
self.assertRaises(errors.KnitCorrupt, list,
815
knit._read_records_iter(records))
817
# read_records_iter_raw detects mismatches in the header
818
self.assertRaises(errors.KnitCorrupt, list,
819
knit._read_records_iter_raw(records))
821
def test_uncompressed_data(self):
822
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
823
txt = ('version rev-id-1 2 %s\n'
828
transport = MockTransport([txt])
829
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
830
knit = KnitVersionedFiles(None, access)
831
records = [(('rev-id-1',), (('rev-id-1',), 0, len(txt)))]
833
# We don't have valid gzip data ==> corrupt
834
self.assertRaises(errors.KnitCorrupt, list,
835
knit._read_records_iter(records))
837
# read_records_iter_raw will notice the bad data
838
self.assertRaises(errors.KnitCorrupt, list,
839
knit._read_records_iter_raw(records))
841
def test_corrupted_data(self):
842
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
843
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
848
# Change 2 bytes in the middle to \xff
849
gz_txt = gz_txt[:10] + '\xff\xff' + gz_txt[12:]
850
transport = MockTransport([gz_txt])
851
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
852
knit = KnitVersionedFiles(None, access)
853
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
854
self.assertRaises(errors.KnitCorrupt, list,
855
knit._read_records_iter(records))
856
# read_records_iter_raw will barf on bad gz data
857
self.assertRaises(errors.KnitCorrupt, list,
858
knit._read_records_iter_raw(records))
861
class LowLevelKnitIndexTests(TestCase):
863
def get_knit_index(self, transport, name, mode):
864
mapper = ConstantMapper(name)
865
from bzrlib._knit_load_data_py import _load_data_py
866
self.overrideAttr(knit, '_load_data', _load_data_py)
867
allow_writes = lambda: 'w' in mode
868
return _KndxIndex(transport, mapper, lambda:None, allow_writes, lambda:True)
870
def test_create_file(self):
871
transport = MockTransport()
872
index = self.get_knit_index(transport, "filename", "w")
874
call = transport.calls.pop(0)
875
# call[1][1] is a StringIO - we can't test it by simple equality.
876
self.assertEqual('put_file_non_atomic', call[0])
877
self.assertEqual('filename.kndx', call[1][0])
878
# With no history, _KndxIndex writes a new index:
879
self.assertEqual(_KndxIndex.HEADER,
880
call[1][1].getvalue())
881
self.assertEqual({'create_parent_dir': True}, call[2])
883
def test_read_utf8_version_id(self):
884
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
885
utf8_revision_id = unicode_revision_id.encode('utf-8')
886
transport = MockTransport([
888
'%s option 0 1 :' % (utf8_revision_id,)
890
index = self.get_knit_index(transport, "filename", "r")
891
# _KndxIndex is a private class, and deals in utf8 revision_ids, not
892
# Unicode revision_ids.
893
self.assertEqual({(utf8_revision_id,):()},
894
index.get_parent_map(index.keys()))
895
self.assertFalse((unicode_revision_id,) in index.keys())
897
def test_read_utf8_parents(self):
898
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
899
utf8_revision_id = unicode_revision_id.encode('utf-8')
900
transport = MockTransport([
902
"version option 0 1 .%s :" % (utf8_revision_id,)
904
index = self.get_knit_index(transport, "filename", "r")
905
self.assertEqual({("version",):((utf8_revision_id,),)},
906
index.get_parent_map(index.keys()))
908
def test_read_ignore_corrupted_lines(self):
909
transport = MockTransport([
912
"corrupted options 0 1 .b .c ",
913
"version options 0 1 :"
915
index = self.get_knit_index(transport, "filename", "r")
916
self.assertEqual(1, len(index.keys()))
917
self.assertEqual(set([("version",)]), index.keys())
919
def test_read_corrupted_header(self):
920
transport = MockTransport(['not a bzr knit index header\n'])
921
index = self.get_knit_index(transport, "filename", "r")
922
self.assertRaises(KnitHeaderError, index.keys)
924
def test_read_duplicate_entries(self):
925
transport = MockTransport([
927
"parent options 0 1 :",
928
"version options1 0 1 0 :",
929
"version options2 1 2 .other :",
930
"version options3 3 4 0 .other :"
932
index = self.get_knit_index(transport, "filename", "r")
933
self.assertEqual(2, len(index.keys()))
934
# check that the index used is the first one written. (Specific
935
# to KnitIndex style indices.
936
self.assertEqual("1", index._dictionary_compress([("version",)]))
937
self.assertEqual((("version",), 3, 4), index.get_position(("version",)))
938
self.assertEqual(["options3"], index.get_options(("version",)))
939
self.assertEqual({("version",):(("parent",), ("other",))},
940
index.get_parent_map([("version",)]))
942
def test_read_compressed_parents(self):
943
transport = MockTransport([
947
"c option 0 1 1 0 :",
949
index = self.get_knit_index(transport, "filename", "r")
950
self.assertEqual({("b",):(("a",),), ("c",):(("b",), ("a",))},
951
index.get_parent_map([("b",), ("c",)]))
953
def test_write_utf8_version_id(self):
954
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
955
utf8_revision_id = unicode_revision_id.encode('utf-8')
956
transport = MockTransport([
959
index = self.get_knit_index(transport, "filename", "r")
961
((utf8_revision_id,), ["option"], ((utf8_revision_id,), 0, 1), [])])
962
call = transport.calls.pop(0)
963
# call[1][1] is a StringIO - we can't test it by simple equality.
964
self.assertEqual('put_file_non_atomic', call[0])
965
self.assertEqual('filename.kndx', call[1][0])
966
# With no history, _KndxIndex writes a new index:
967
self.assertEqual(_KndxIndex.HEADER +
968
"\n%s option 0 1 :" % (utf8_revision_id,),
969
call[1][1].getvalue())
970
self.assertEqual({'create_parent_dir': True}, call[2])
972
def test_write_utf8_parents(self):
973
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
974
utf8_revision_id = unicode_revision_id.encode('utf-8')
975
transport = MockTransport([
978
index = self.get_knit_index(transport, "filename", "r")
980
(("version",), ["option"], (("version",), 0, 1), [(utf8_revision_id,)])])
981
call = transport.calls.pop(0)
982
# call[1][1] is a StringIO - we can't test it by simple equality.
983
self.assertEqual('put_file_non_atomic', call[0])
984
self.assertEqual('filename.kndx', call[1][0])
985
# With no history, _KndxIndex writes a new index:
986
self.assertEqual(_KndxIndex.HEADER +
987
"\nversion option 0 1 .%s :" % (utf8_revision_id,),
988
call[1][1].getvalue())
989
self.assertEqual({'create_parent_dir': True}, call[2])
992
transport = MockTransport([
995
index = self.get_knit_index(transport, "filename", "r")
997
self.assertEqual(set(), index.keys())
999
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1000
self.assertEqual(set([("a",)]), index.keys())
1002
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1003
self.assertEqual(set([("a",)]), index.keys())
1005
index.add_records([(("b",), ["option"], (("b",), 0, 1), [])])
1006
self.assertEqual(set([("a",), ("b",)]), index.keys())
1008
def add_a_b(self, index, random_id=None):
1010
if random_id is not None:
1011
kwargs["random_id"] = random_id
1013
(("a",), ["option"], (("a",), 0, 1), [("b",)]),
1014
(("a",), ["opt"], (("a",), 1, 2), [("c",)]),
1015
(("b",), ["option"], (("b",), 2, 3), [("a",)])
1018
def assertIndexIsAB(self, index):
1023
index.get_parent_map(index.keys()))
1024
self.assertEqual((("a",), 1, 2), index.get_position(("a",)))
1025
self.assertEqual((("b",), 2, 3), index.get_position(("b",)))
1026
self.assertEqual(["opt"], index.get_options(("a",)))
1028
def test_add_versions(self):
1029
transport = MockTransport([
1032
index = self.get_knit_index(transport, "filename", "r")
1035
call = transport.calls.pop(0)
1036
# call[1][1] is a StringIO - we can't test it by simple equality.
1037
self.assertEqual('put_file_non_atomic', call[0])
1038
self.assertEqual('filename.kndx', call[1][0])
1039
# With no history, _KndxIndex writes a new index:
1042
"\na option 0 1 .b :"
1044
"\nb option 2 3 0 :",
1045
call[1][1].getvalue())
1046
self.assertEqual({'create_parent_dir': True}, call[2])
1047
self.assertIndexIsAB(index)
1049
def test_add_versions_random_id_is_accepted(self):
1050
transport = MockTransport([
1053
index = self.get_knit_index(transport, "filename", "r")
1054
self.add_a_b(index, random_id=True)
1056
def test_delay_create_and_add_versions(self):
1057
transport = MockTransport()
1059
index = self.get_knit_index(transport, "filename", "w")
1061
self.assertEqual([], transport.calls)
1064
#[ {"dir_mode": 0777, "create_parent_dir": True, "mode": "wb"},
1066
# Two calls: one during which we load the existing index (and when its
1067
# missing create it), then a second where we write the contents out.
1068
self.assertEqual(2, len(transport.calls))
1069
call = transport.calls.pop(0)
1070
self.assertEqual('put_file_non_atomic', call[0])
1071
self.assertEqual('filename.kndx', call[1][0])
1072
# With no history, _KndxIndex writes a new index:
1073
self.assertEqual(_KndxIndex.HEADER, call[1][1].getvalue())
1074
self.assertEqual({'create_parent_dir': True}, call[2])
1075
call = transport.calls.pop(0)
1076
# call[1][1] is a StringIO - we can't test it by simple equality.
1077
self.assertEqual('put_file_non_atomic', call[0])
1078
self.assertEqual('filename.kndx', call[1][0])
1079
# With no history, _KndxIndex writes a new index:
1082
"\na option 0 1 .b :"
1084
"\nb option 2 3 0 :",
1085
call[1][1].getvalue())
1086
self.assertEqual({'create_parent_dir': True}, call[2])
1088
def assertTotalBuildSize(self, size, keys, positions):
1089
self.assertEqual(size,
1090
knit._get_total_build_size(None, keys, positions))
1092
def test__get_total_build_size(self):
1094
('a',): (('fulltext', False), (('a',), 0, 100), None),
1095
('b',): (('line-delta', False), (('b',), 100, 21), ('a',)),
1096
('c',): (('line-delta', False), (('c',), 121, 35), ('b',)),
1097
('d',): (('line-delta', False), (('d',), 156, 12), ('b',)),
1099
self.assertTotalBuildSize(100, [('a',)], positions)
1100
self.assertTotalBuildSize(121, [('b',)], positions)
1101
# c needs both a & b
1102
self.assertTotalBuildSize(156, [('c',)], positions)
1103
# we shouldn't count 'b' twice
1104
self.assertTotalBuildSize(156, [('b',), ('c',)], positions)
1105
self.assertTotalBuildSize(133, [('d',)], positions)
1106
self.assertTotalBuildSize(168, [('c',), ('d',)], positions)
1108
def test_get_position(self):
1109
transport = MockTransport([
1114
index = self.get_knit_index(transport, "filename", "r")
1116
self.assertEqual((("a",), 0, 1), index.get_position(("a",)))
1117
self.assertEqual((("b",), 1, 2), index.get_position(("b",)))
1119
def test_get_method(self):
1120
transport = MockTransport([
1122
"a fulltext,unknown 0 1 :",
1123
"b unknown,line-delta 1 2 :",
1126
index = self.get_knit_index(transport, "filename", "r")
1128
self.assertEqual("fulltext", index.get_method("a"))
1129
self.assertEqual("line-delta", index.get_method("b"))
1130
self.assertRaises(errors.KnitIndexUnknownMethod, index.get_method, "c")
1132
def test_get_options(self):
1133
transport = MockTransport([
1138
index = self.get_knit_index(transport, "filename", "r")
1140
self.assertEqual(["opt1"], index.get_options("a"))
1141
self.assertEqual(["opt2", "opt3"], index.get_options("b"))
1143
def test_get_parent_map(self):
1144
transport = MockTransport([
1147
"b option 1 2 0 .c :",
1148
"c option 1 2 1 0 .e :"
1150
index = self.get_knit_index(transport, "filename", "r")
1154
("b",):(("a",), ("c",)),
1155
("c",):(("b",), ("a",), ("e",)),
1156
}, index.get_parent_map(index.keys()))
1158
def test_impossible_parent(self):
1159
"""Test we get KnitCorrupt if the parent couldn't possibly exist."""
1160
transport = MockTransport([
1163
"b option 0 1 4 :" # We don't have a 4th record
1165
index = self.get_knit_index(transport, 'filename', 'r')
1167
self.assertRaises(errors.KnitCorrupt, index.keys)
1168
except TypeError, e:
1169
if (str(e) == ('exceptions must be strings, classes, or instances,'
1170
' not exceptions.IndexError')
1171
and sys.version_info[0:2] >= (2,5)):
1172
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1173
' raising new style exceptions with python'
1178
def test_corrupted_parent(self):
1179
transport = MockTransport([
1183
"c option 0 1 1v :", # Can't have a parent of '1v'
1185
index = self.get_knit_index(transport, 'filename', 'r')
1187
self.assertRaises(errors.KnitCorrupt, index.keys)
1188
except TypeError, e:
1189
if (str(e) == ('exceptions must be strings, classes, or instances,'
1190
' not exceptions.ValueError')
1191
and sys.version_info[0:2] >= (2,5)):
1192
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1193
' raising new style exceptions with python'
1198
def test_corrupted_parent_in_list(self):
1199
transport = MockTransport([
1203
"c option 0 1 1 v :", # Can't have a parent of 'v'
1205
index = self.get_knit_index(transport, 'filename', 'r')
1207
self.assertRaises(errors.KnitCorrupt, index.keys)
1208
except TypeError, e:
1209
if (str(e) == ('exceptions must be strings, classes, or instances,'
1210
' not exceptions.ValueError')
1211
and sys.version_info[0:2] >= (2,5)):
1212
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1213
' raising new style exceptions with python'
1218
def test_invalid_position(self):
1219
transport = MockTransport([
1223
index = self.get_knit_index(transport, 'filename', 'r')
1225
self.assertRaises(errors.KnitCorrupt, index.keys)
1226
except TypeError, e:
1227
if (str(e) == ('exceptions must be strings, classes, or instances,'
1228
' not exceptions.ValueError')
1229
and sys.version_info[0:2] >= (2,5)):
1230
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1231
' raising new style exceptions with python'
1236
def test_invalid_size(self):
1237
transport = MockTransport([
1241
index = self.get_knit_index(transport, 'filename', 'r')
1243
self.assertRaises(errors.KnitCorrupt, index.keys)
1244
except TypeError, e:
1245
if (str(e) == ('exceptions must be strings, classes, or instances,'
1246
' not exceptions.ValueError')
1247
and sys.version_info[0:2] >= (2,5)):
1248
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1249
' raising new style exceptions with python'
1254
def test_scan_unvalidated_index_not_implemented(self):
1255
transport = MockTransport()
1256
index = self.get_knit_index(transport, 'filename', 'r')
1258
NotImplementedError, index.scan_unvalidated_index,
1259
'dummy graph_index')
1261
NotImplementedError, index.get_missing_compression_parents)
1263
def test_short_line(self):
1264
transport = MockTransport([
1267
"b option 10 10 0", # This line isn't terminated, ignored
1269
index = self.get_knit_index(transport, "filename", "r")
1270
self.assertEqual(set([('a',)]), index.keys())
1272
def test_skip_incomplete_record(self):
1273
# A line with bogus data should just be skipped
1274
transport = MockTransport([
1277
"b option 10 10 0", # This line isn't terminated, ignored
1278
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1280
index = self.get_knit_index(transport, "filename", "r")
1281
self.assertEqual(set([('a',), ('c',)]), index.keys())
1283
def test_trailing_characters(self):
1284
# A line with bogus data should just be skipped
1285
transport = MockTransport([
1288
"b option 10 10 0 :a", # This line has extra trailing characters
1289
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1291
index = self.get_knit_index(transport, "filename", "r")
1292
self.assertEqual(set([('a',), ('c',)]), index.keys())
1295
class LowLevelKnitIndexTests_c(LowLevelKnitIndexTests):
1297
_test_needs_features = [compiled_knit_feature]
1299
def get_knit_index(self, transport, name, mode):
1300
mapper = ConstantMapper(name)
1301
from bzrlib._knit_load_data_pyx import _load_data_c
1302
self.overrideAttr(knit, '_load_data', _load_data_c)
1303
allow_writes = lambda: mode == 'w'
1304
return _KndxIndex(transport, mapper, lambda:None,
1305
allow_writes, lambda:True)
1308
class Test_KnitAnnotator(TestCaseWithMemoryTransport):
1310
def make_annotator(self):
1311
factory = knit.make_pack_factory(True, True, 1)
1312
vf = factory(self.get_transport())
1313
return knit._KnitAnnotator(vf)
1315
def test__expand_fulltext(self):
1316
ann = self.make_annotator()
1317
rev_key = ('rev-id',)
1318
ann._num_compression_children[rev_key] = 1
1319
res = ann._expand_record(rev_key, (('parent-id',),), None,
1320
['line1\n', 'line2\n'], ('fulltext', True))
1321
# The content object and text lines should be cached appropriately
1322
self.assertEqual(['line1\n', 'line2'], res)
1323
content_obj = ann._content_objects[rev_key]
1324
self.assertEqual(['line1\n', 'line2\n'], content_obj._lines)
1325
self.assertEqual(res, content_obj.text())
1326
self.assertEqual(res, ann._text_cache[rev_key])
1328
def test__expand_delta_comp_parent_not_available(self):
1329
# Parent isn't available yet, so we return nothing, but queue up this
1330
# node for later processing
1331
ann = self.make_annotator()
1332
rev_key = ('rev-id',)
1333
parent_key = ('parent-id',)
1334
record = ['0,1,1\n', 'new-line\n']
1335
details = ('line-delta', False)
1336
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1338
self.assertEqual(None, res)
1339
self.assertTrue(parent_key in ann._pending_deltas)
1340
pending = ann._pending_deltas[parent_key]
1341
self.assertEqual(1, len(pending))
1342
self.assertEqual((rev_key, (parent_key,), record, details), pending[0])
1344
def test__expand_record_tracks_num_children(self):
1345
ann = self.make_annotator()
1346
rev_key = ('rev-id',)
1347
rev2_key = ('rev2-id',)
1348
parent_key = ('parent-id',)
1349
record = ['0,1,1\n', 'new-line\n']
1350
details = ('line-delta', False)
1351
ann._num_compression_children[parent_key] = 2
1352
ann._expand_record(parent_key, (), None, ['line1\n', 'line2\n'],
1353
('fulltext', False))
1354
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1356
self.assertEqual({parent_key: 1}, ann._num_compression_children)
1357
# Expanding the second child should remove the content object, and the
1358
# num_compression_children entry
1359
res = ann._expand_record(rev2_key, (parent_key,), parent_key,
1361
self.assertFalse(parent_key in ann._content_objects)
1362
self.assertEqual({}, ann._num_compression_children)
1363
# We should not cache the content_objects for rev2 and rev, because
1364
# they do not have compression children of their own.
1365
self.assertEqual({}, ann._content_objects)
1367
def test__expand_delta_records_blocks(self):
1368
ann = self.make_annotator()
1369
rev_key = ('rev-id',)
1370
parent_key = ('parent-id',)
1371
record = ['0,1,1\n', 'new-line\n']
1372
details = ('line-delta', True)
1373
ann._num_compression_children[parent_key] = 2
1374
ann._expand_record(parent_key, (), None,
1375
['line1\n', 'line2\n', 'line3\n'],
1376
('fulltext', False))
1377
ann._expand_record(rev_key, (parent_key,), parent_key, record, details)
1378
self.assertEqual({(rev_key, parent_key): [(1, 1, 1), (3, 3, 0)]},
1379
ann._matching_blocks)
1380
rev2_key = ('rev2-id',)
1381
record = ['0,1,1\n', 'new-line\n']
1382
details = ('line-delta', False)
1383
ann._expand_record(rev2_key, (parent_key,), parent_key, record, details)
1384
self.assertEqual([(1, 1, 2), (3, 3, 0)],
1385
ann._matching_blocks[(rev2_key, parent_key)])
1387
def test__get_parent_ann_uses_matching_blocks(self):
1388
ann = self.make_annotator()
1389
rev_key = ('rev-id',)
1390
parent_key = ('parent-id',)
1391
parent_ann = [(parent_key,)]*3
1392
block_key = (rev_key, parent_key)
1393
ann._annotations_cache[parent_key] = parent_ann
1394
ann._matching_blocks[block_key] = [(0, 1, 1), (3, 3, 0)]
1395
# We should not try to access any parent_lines content, because we know
1396
# we already have the matching blocks
1397
par_ann, blocks = ann._get_parent_annotations_and_matches(rev_key,
1398
['1\n', '2\n', '3\n'], parent_key)
1399
self.assertEqual(parent_ann, par_ann)
1400
self.assertEqual([(0, 1, 1), (3, 3, 0)], blocks)
1401
self.assertEqual({}, ann._matching_blocks)
1403
def test__process_pending(self):
1404
ann = self.make_annotator()
1405
rev_key = ('rev-id',)
1408
record = ['0,1,1\n', 'new-line\n']
1409
details = ('line-delta', False)
1410
p1_record = ['line1\n', 'line2\n']
1411
ann._num_compression_children[p1_key] = 1
1412
res = ann._expand_record(rev_key, (p1_key,p2_key), p1_key,
1414
self.assertEqual(None, res)
1415
# self.assertTrue(p1_key in ann._pending_deltas)
1416
self.assertEqual({}, ann._pending_annotation)
1417
# Now insert p1, and we should be able to expand the delta
1418
res = ann._expand_record(p1_key, (), None, p1_record,
1419
('fulltext', False))
1420
self.assertEqual(p1_record, res)
1421
ann._annotations_cache[p1_key] = [(p1_key,)]*2
1422
res = ann._process_pending(p1_key)
1423
self.assertEqual([], res)
1424
self.assertFalse(p1_key in ann._pending_deltas)
1425
self.assertTrue(p2_key in ann._pending_annotation)
1426
self.assertEqual({p2_key: [(rev_key, (p1_key, p2_key))]},
1427
ann._pending_annotation)
1428
# Now fill in parent 2, and pending annotation should be satisfied
1429
res = ann._expand_record(p2_key, (), None, [], ('fulltext', False))
1430
ann._annotations_cache[p2_key] = []
1431
res = ann._process_pending(p2_key)
1432
self.assertEqual([rev_key], res)
1433
self.assertEqual({}, ann._pending_annotation)
1434
self.assertEqual({}, ann._pending_deltas)
1436
def test_record_delta_removes_basis(self):
1437
ann = self.make_annotator()
1438
ann._expand_record(('parent-id',), (), None,
1439
['line1\n', 'line2\n'], ('fulltext', False))
1440
ann._num_compression_children['parent-id'] = 2
1442
def test_annotate_special_text(self):
1443
ann = self.make_annotator()
1445
rev1_key = ('rev-1',)
1446
rev2_key = ('rev-2',)
1447
rev3_key = ('rev-3',)
1448
spec_key = ('special:',)
1449
vf.add_lines(rev1_key, [], ['initial content\n'])
1450
vf.add_lines(rev2_key, [rev1_key], ['initial content\n',
1453
vf.add_lines(rev3_key, [rev1_key], ['initial content\n',
1456
spec_text = ('initial content\n'
1460
ann.add_special_text(spec_key, [rev2_key, rev3_key], spec_text)
1461
anns, lines = ann.annotate(spec_key)
1462
self.assertEqual([(rev1_key,),
1463
(rev2_key, rev3_key),
1467
self.assertEqualDiff(spec_text, ''.join(lines))
1470
class KnitTests(TestCaseWithTransport):
1471
"""Class containing knit test helper routines."""
1473
def make_test_knit(self, annotate=False, name='test'):
1474
mapper = ConstantMapper(name)
1475
return make_file_factory(annotate, mapper)(self.get_transport())
1478
class TestBadShaError(KnitTests):
1479
"""Tests for handling of sha errors."""
1481
def test_sha_exception_has_text(self):
1482
# having the failed text included in the error allows for recovery.
1483
source = self.make_test_knit()
1484
target = self.make_test_knit(name="target")
1485
if not source._max_delta_chain:
1486
raise TestNotApplicable(
1487
"cannot get delta-caused sha failures without deltas.")
1490
broken = ('broken',)
1491
source.add_lines(basis, (), ['foo\n'])
1492
source.add_lines(broken, (basis,), ['foo\n', 'bar\n'])
1493
# Seed target with a bad basis text
1494
target.add_lines(basis, (), ['gam\n'])
1495
target.insert_record_stream(
1496
source.get_record_stream([broken], 'unordered', False))
1497
err = self.assertRaises(errors.KnitCorrupt,
1498
target.get_record_stream([broken], 'unordered', True
1499
).next().get_bytes_as, 'chunked')
1500
self.assertEqual(['gam\n', 'bar\n'], err.content)
1501
# Test for formatting with live data
1502
self.assertStartsWith(str(err), "Knit ")
1505
class TestKnitIndex(KnitTests):
1507
def test_add_versions_dictionary_compresses(self):
1508
"""Adding versions to the index should update the lookup dict"""
1509
knit = self.make_test_knit()
1511
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1512
self.check_file_contents('test.kndx',
1513
'# bzr knit index 8\n'
1515
'a-1 fulltext 0 0 :'
1518
(('a-2',), ['fulltext'], (('a-2',), 0, 0), [('a-1',)]),
1519
(('a-3',), ['fulltext'], (('a-3',), 0, 0), [('a-2',)]),
1521
self.check_file_contents('test.kndx',
1522
'# bzr knit index 8\n'
1524
'a-1 fulltext 0 0 :\n'
1525
'a-2 fulltext 0 0 0 :\n'
1526
'a-3 fulltext 0 0 1 :'
1528
self.assertEqual(set([('a-3',), ('a-1',), ('a-2',)]), idx.keys())
1530
('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False)),
1531
('a-2',): ((('a-2',), 0, 0), None, (('a-1',),), ('fulltext', False)),
1532
('a-3',): ((('a-3',), 0, 0), None, (('a-2',),), ('fulltext', False)),
1533
}, idx.get_build_details(idx.keys()))
1534
self.assertEqual({('a-1',):(),
1535
('a-2',):(('a-1',),),
1536
('a-3',):(('a-2',),),},
1537
idx.get_parent_map(idx.keys()))
1539
def test_add_versions_fails_clean(self):
1540
"""If add_versions fails in the middle, it restores a pristine state.
1542
Any modifications that are made to the index are reset if all versions
1545
# This cheats a little bit by passing in a generator which will
1546
# raise an exception before the processing finishes
1547
# Other possibilities would be to have an version with the wrong number
1548
# of entries, or to make the backing transport unable to write any
1551
knit = self.make_test_knit()
1553
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1555
class StopEarly(Exception):
1558
def generate_failure():
1559
"""Add some entries and then raise an exception"""
1560
yield (('a-2',), ['fulltext'], (None, 0, 0), ('a-1',))
1561
yield (('a-3',), ['fulltext'], (None, 0, 0), ('a-2',))
1564
# Assert the pre-condition
1566
self.assertEqual(set([('a-1',)]), set(idx.keys()))
1568
{('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False))},
1569
idx.get_build_details([('a-1',)]))
1570
self.assertEqual({('a-1',):()}, idx.get_parent_map(idx.keys()))
1573
self.assertRaises(StopEarly, idx.add_records, generate_failure())
1574
# And it shouldn't be modified
1577
def test_knit_index_ignores_empty_files(self):
1578
# There was a race condition in older bzr, where a ^C at the right time
1579
# could leave an empty .kndx file, which bzr would later claim was a
1580
# corrupted file since the header was not present. In reality, the file
1581
# just wasn't created, so it should be ignored.
1582
t = get_transport('.')
1583
t.put_bytes('test.kndx', '')
1585
knit = self.make_test_knit()
1587
def test_knit_index_checks_header(self):
1588
t = get_transport('.')
1589
t.put_bytes('test.kndx', '# not really a knit header\n\n')
1590
k = self.make_test_knit()
1591
self.assertRaises(KnitHeaderError, k.keys)
1594
class TestGraphIndexKnit(KnitTests):
1595
"""Tests for knits using a GraphIndex rather than a KnitIndex."""
1597
def make_g_index(self, name, ref_lists=0, nodes=[]):
1598
builder = GraphIndexBuilder(ref_lists)
1599
for node, references, value in nodes:
1600
builder.add_node(node, references, value)
1601
stream = builder.finish()
1602
trans = self.get_transport()
1603
size = trans.put_file(name, stream)
1604
return GraphIndex(trans, name, size)
1606
def two_graph_index(self, deltas=False, catch_adds=False):
1607
"""Build a two-graph index.
1609
:param deltas: If true, use underlying indices with two node-ref
1610
lists and 'parent' set to a delta-compressed against tail.
1612
# build a complex graph across several indices.
1614
# delta compression inn the index
1615
index1 = self.make_g_index('1', 2, [
1616
(('tip', ), 'N0 100', ([('parent', )], [], )),
1617
(('tail', ), '', ([], []))])
1618
index2 = self.make_g_index('2', 2, [
1619
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], [('tail', )])),
1620
(('separate', ), '', ([], []))])
1622
# just blob location and graph in the index.
1623
index1 = self.make_g_index('1', 1, [
1624
(('tip', ), 'N0 100', ([('parent', )], )),
1625
(('tail', ), '', ([], ))])
1626
index2 = self.make_g_index('2', 1, [
1627
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], )),
1628
(('separate', ), '', ([], ))])
1629
combined_index = CombinedGraphIndex([index1, index2])
1631
self.combined_index = combined_index
1632
self.caught_entries = []
1633
add_callback = self.catch_add
1636
return _KnitGraphIndex(combined_index, lambda:True, deltas=deltas,
1637
add_callback=add_callback)
1639
def test_keys(self):
1640
index = self.two_graph_index()
1641
self.assertEqual(set([('tail',), ('tip',), ('parent',), ('separate',)]),
1644
def test_get_position(self):
1645
index = self.two_graph_index()
1646
self.assertEqual((index._graph_index._indices[0], 0, 100), index.get_position(('tip',)))
1647
self.assertEqual((index._graph_index._indices[1], 100, 78), index.get_position(('parent',)))
1649
def test_get_method_deltas(self):
1650
index = self.two_graph_index(deltas=True)
1651
self.assertEqual('fulltext', index.get_method(('tip',)))
1652
self.assertEqual('line-delta', index.get_method(('parent',)))
1654
def test_get_method_no_deltas(self):
1655
# check that the parent-history lookup is ignored with deltas=False.
1656
index = self.two_graph_index(deltas=False)
1657
self.assertEqual('fulltext', index.get_method(('tip',)))
1658
self.assertEqual('fulltext', index.get_method(('parent',)))
1660
def test_get_options_deltas(self):
1661
index = self.two_graph_index(deltas=True)
1662
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1663
self.assertEqual(['line-delta'], index.get_options(('parent',)))
1665
def test_get_options_no_deltas(self):
1666
# check that the parent-history lookup is ignored with deltas=False.
1667
index = self.two_graph_index(deltas=False)
1668
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1669
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1671
def test_get_parent_map(self):
1672
index = self.two_graph_index()
1673
self.assertEqual({('parent',):(('tail',), ('ghost',))},
1674
index.get_parent_map([('parent',), ('ghost',)]))
1676
def catch_add(self, entries):
1677
self.caught_entries.append(entries)
1679
def test_add_no_callback_errors(self):
1680
index = self.two_graph_index()
1681
self.assertRaises(errors.ReadOnlyError, index.add_records,
1682
[(('new',), 'fulltext,no-eol', (None, 50, 60), ['separate'])])
1684
def test_add_version_smoke(self):
1685
index = self.two_graph_index(catch_adds=True)
1686
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60),
1688
self.assertEqual([[(('new', ), 'N50 60', ((('separate',),),))]],
1689
self.caught_entries)
1691
def test_add_version_delta_not_delta_index(self):
1692
index = self.two_graph_index(catch_adds=True)
1693
self.assertRaises(errors.KnitCorrupt, index.add_records,
1694
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1695
self.assertEqual([], self.caught_entries)
1697
def test_add_version_same_dup(self):
1698
index = self.two_graph_index(catch_adds=True)
1699
# options can be spelt two different ways
1700
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
1701
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
1702
# position/length are ignored (because each pack could have fulltext or
1703
# delta, and be at a different position.
1704
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1706
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1708
# but neither should have added data:
1709
self.assertEqual([[], [], [], []], self.caught_entries)
1711
def test_add_version_different_dup(self):
1712
index = self.two_graph_index(deltas=True, catch_adds=True)
1714
self.assertRaises(errors.KnitCorrupt, index.add_records,
1715
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1716
self.assertRaises(errors.KnitCorrupt, index.add_records,
1717
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1719
self.assertRaises(errors.KnitCorrupt, index.add_records,
1720
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1721
self.assertEqual([], self.caught_entries)
1723
def test_add_versions_nodeltas(self):
1724
index = self.two_graph_index(catch_adds=True)
1726
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1727
(('new2',), 'fulltext', (None, 0, 6), [('new',)]),
1729
self.assertEqual([(('new', ), 'N50 60', ((('separate',),),)),
1730
(('new2', ), ' 0 6', ((('new',),),))],
1731
sorted(self.caught_entries[0]))
1732
self.assertEqual(1, len(self.caught_entries))
1734
def test_add_versions_deltas(self):
1735
index = self.two_graph_index(deltas=True, catch_adds=True)
1737
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1738
(('new2',), 'line-delta', (None, 0, 6), [('new',)]),
1740
self.assertEqual([(('new', ), 'N50 60', ((('separate',),), ())),
1741
(('new2', ), ' 0 6', ((('new',),), (('new',),), ))],
1742
sorted(self.caught_entries[0]))
1743
self.assertEqual(1, len(self.caught_entries))
1745
def test_add_versions_delta_not_delta_index(self):
1746
index = self.two_graph_index(catch_adds=True)
1747
self.assertRaises(errors.KnitCorrupt, index.add_records,
1748
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1749
self.assertEqual([], self.caught_entries)
1751
def test_add_versions_random_id_accepted(self):
1752
index = self.two_graph_index(catch_adds=True)
1753
index.add_records([], random_id=True)
1755
def test_add_versions_same_dup(self):
1756
index = self.two_graph_index(catch_adds=True)
1757
# options can be spelt two different ways
1758
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100),
1760
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100),
1762
# position/length are ignored (because each pack could have fulltext or
1763
# delta, and be at a different position.
1764
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1766
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1768
# but neither should have added data.
1769
self.assertEqual([[], [], [], []], self.caught_entries)
1771
def test_add_versions_different_dup(self):
1772
index = self.two_graph_index(deltas=True, catch_adds=True)
1774
self.assertRaises(errors.KnitCorrupt, index.add_records,
1775
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1776
self.assertRaises(errors.KnitCorrupt, index.add_records,
1777
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1779
self.assertRaises(errors.KnitCorrupt, index.add_records,
1780
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1781
# change options in the second record
1782
self.assertRaises(errors.KnitCorrupt, index.add_records,
1783
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)]),
1784
(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1785
self.assertEqual([], self.caught_entries)
1787
def make_g_index_missing_compression_parent(self):
1788
graph_index = self.make_g_index('missing_comp', 2,
1789
[(('tip', ), ' 100 78',
1790
([('missing-parent', ), ('ghost', )], [('missing-parent', )]))])
1793
def make_g_index_missing_parent(self):
1794
graph_index = self.make_g_index('missing_parent', 2,
1795
[(('parent', ), ' 100 78', ([], [])),
1796
(('tip', ), ' 100 78',
1797
([('parent', ), ('missing-parent', )], [('parent', )])),
1801
def make_g_index_no_external_refs(self):
1802
graph_index = self.make_g_index('no_external_refs', 2,
1803
[(('rev', ), ' 100 78',
1804
([('parent', ), ('ghost', )], []))])
1807
def test_add_good_unvalidated_index(self):
1808
unvalidated = self.make_g_index_no_external_refs()
1809
combined = CombinedGraphIndex([unvalidated])
1810
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1811
index.scan_unvalidated_index(unvalidated)
1812
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1814
def test_add_missing_compression_parent_unvalidated_index(self):
1815
unvalidated = self.make_g_index_missing_compression_parent()
1816
combined = CombinedGraphIndex([unvalidated])
1817
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1818
index.scan_unvalidated_index(unvalidated)
1819
# This also checks that its only the compression parent that is
1820
# examined, otherwise 'ghost' would also be reported as a missing
1823
frozenset([('missing-parent',)]),
1824
index.get_missing_compression_parents())
1826
def test_add_missing_noncompression_parent_unvalidated_index(self):
1827
unvalidated = self.make_g_index_missing_parent()
1828
combined = CombinedGraphIndex([unvalidated])
1829
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1830
track_external_parent_refs=True)
1831
index.scan_unvalidated_index(unvalidated)
1833
frozenset([('missing-parent',)]), index.get_missing_parents())
1835
def test_track_external_parent_refs(self):
1836
g_index = self.make_g_index('empty', 2, [])
1837
combined = CombinedGraphIndex([g_index])
1838
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1839
add_callback=self.catch_add, track_external_parent_refs=True)
1840
self.caught_entries = []
1842
(('new-key',), 'fulltext,no-eol', (None, 50, 60),
1843
[('parent-1',), ('parent-2',)])])
1845
frozenset([('parent-1',), ('parent-2',)]),
1846
index.get_missing_parents())
1848
def test_add_unvalidated_index_with_present_external_references(self):
1849
index = self.two_graph_index(deltas=True)
1850
# Ugly hack to get at one of the underlying GraphIndex objects that
1851
# two_graph_index built.
1852
unvalidated = index._graph_index._indices[1]
1853
# 'parent' is an external ref of _indices[1] (unvalidated), but is
1854
# present in _indices[0].
1855
index.scan_unvalidated_index(unvalidated)
1856
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1858
def make_new_missing_parent_g_index(self, name):
1859
missing_parent = name + '-missing-parent'
1860
graph_index = self.make_g_index(name, 2,
1861
[((name + 'tip', ), ' 100 78',
1862
([(missing_parent, ), ('ghost', )], [(missing_parent, )]))])
1865
def test_add_mulitiple_unvalidated_indices_with_missing_parents(self):
1866
g_index_1 = self.make_new_missing_parent_g_index('one')
1867
g_index_2 = self.make_new_missing_parent_g_index('two')
1868
combined = CombinedGraphIndex([g_index_1, g_index_2])
1869
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1870
index.scan_unvalidated_index(g_index_1)
1871
index.scan_unvalidated_index(g_index_2)
1873
frozenset([('one-missing-parent',), ('two-missing-parent',)]),
1874
index.get_missing_compression_parents())
1876
def test_add_mulitiple_unvalidated_indices_with_mutual_dependencies(self):
1877
graph_index_a = self.make_g_index('one', 2,
1878
[(('parent-one', ), ' 100 78', ([('non-compression-parent',)], [])),
1879
(('child-of-two', ), ' 100 78',
1880
([('parent-two',)], [('parent-two',)]))])
1881
graph_index_b = self.make_g_index('two', 2,
1882
[(('parent-two', ), ' 100 78', ([('non-compression-parent',)], [])),
1883
(('child-of-one', ), ' 100 78',
1884
([('parent-one',)], [('parent-one',)]))])
1885
combined = CombinedGraphIndex([graph_index_a, graph_index_b])
1886
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1887
index.scan_unvalidated_index(graph_index_a)
1888
index.scan_unvalidated_index(graph_index_b)
1890
frozenset([]), index.get_missing_compression_parents())
1893
class TestNoParentsGraphIndexKnit(KnitTests):
1894
"""Tests for knits using _KnitGraphIndex with no parents."""
1896
def make_g_index(self, name, ref_lists=0, nodes=[]):
1897
builder = GraphIndexBuilder(ref_lists)
1898
for node, references in nodes:
1899
builder.add_node(node, references)
1900
stream = builder.finish()
1901
trans = self.get_transport()
1902
size = trans.put_file(name, stream)
1903
return GraphIndex(trans, name, size)
1905
def test_add_good_unvalidated_index(self):
1906
unvalidated = self.make_g_index('unvalidated')
1907
combined = CombinedGraphIndex([unvalidated])
1908
index = _KnitGraphIndex(combined, lambda: True, parents=False)
1909
index.scan_unvalidated_index(unvalidated)
1910
self.assertEqual(frozenset(),
1911
index.get_missing_compression_parents())
1913
def test_parents_deltas_incompatible(self):
1914
index = CombinedGraphIndex([])
1915
self.assertRaises(errors.KnitError, _KnitGraphIndex, lambda:True,
1916
index, deltas=True, parents=False)
1918
def two_graph_index(self, catch_adds=False):
1919
"""Build a two-graph index.
1921
:param deltas: If true, use underlying indices with two node-ref
1922
lists and 'parent' set to a delta-compressed against tail.
1924
# put several versions in the index.
1925
index1 = self.make_g_index('1', 0, [
1926
(('tip', ), 'N0 100'),
1928
index2 = self.make_g_index('2', 0, [
1929
(('parent', ), ' 100 78'),
1930
(('separate', ), '')])
1931
combined_index = CombinedGraphIndex([index1, index2])
1933
self.combined_index = combined_index
1934
self.caught_entries = []
1935
add_callback = self.catch_add
1938
return _KnitGraphIndex(combined_index, lambda:True, parents=False,
1939
add_callback=add_callback)
1941
def test_keys(self):
1942
index = self.two_graph_index()
1943
self.assertEqual(set([('tail',), ('tip',), ('parent',), ('separate',)]),
1946
def test_get_position(self):
1947
index = self.two_graph_index()
1948
self.assertEqual((index._graph_index._indices[0], 0, 100),
1949
index.get_position(('tip',)))
1950
self.assertEqual((index._graph_index._indices[1], 100, 78),
1951
index.get_position(('parent',)))
1953
def test_get_method(self):
1954
index = self.two_graph_index()
1955
self.assertEqual('fulltext', index.get_method(('tip',)))
1956
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1958
def test_get_options(self):
1959
index = self.two_graph_index()
1960
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1961
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1963
def test_get_parent_map(self):
1964
index = self.two_graph_index()
1965
self.assertEqual({('parent',):None},
1966
index.get_parent_map([('parent',), ('ghost',)]))
1968
def catch_add(self, entries):
1969
self.caught_entries.append(entries)
1971
def test_add_no_callback_errors(self):
1972
index = self.two_graph_index()
1973
self.assertRaises(errors.ReadOnlyError, index.add_records,
1974
[(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)])])
1976
def test_add_version_smoke(self):
1977
index = self.two_graph_index(catch_adds=True)
1978
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60), [])])
1979
self.assertEqual([[(('new', ), 'N50 60')]],
1980
self.caught_entries)
1982
def test_add_version_delta_not_delta_index(self):
1983
index = self.two_graph_index(catch_adds=True)
1984
self.assertRaises(errors.KnitCorrupt, index.add_records,
1985
[(('new',), 'no-eol,line-delta', (None, 0, 100), [])])
1986
self.assertEqual([], self.caught_entries)
1988
def test_add_version_same_dup(self):
1989
index = self.two_graph_index(catch_adds=True)
1990
# options can be spelt two different ways
1991
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1992
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
1993
# position/length are ignored (because each pack could have fulltext or
1994
# delta, and be at a different position.
1995
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
1996
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
1997
# but neither should have added data.
1998
self.assertEqual([[], [], [], []], self.caught_entries)
2000
def test_add_version_different_dup(self):
2001
index = self.two_graph_index(catch_adds=True)
2003
self.assertRaises(errors.KnitCorrupt, index.add_records,
2004
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2005
self.assertRaises(errors.KnitCorrupt, index.add_records,
2006
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
2007
self.assertRaises(errors.KnitCorrupt, index.add_records,
2008
[(('tip',), 'fulltext', (None, 0, 100), [])])
2010
self.assertRaises(errors.KnitCorrupt, index.add_records,
2011
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
2012
self.assertEqual([], self.caught_entries)
2014
def test_add_versions(self):
2015
index = self.two_graph_index(catch_adds=True)
2017
(('new',), 'fulltext,no-eol', (None, 50, 60), []),
2018
(('new2',), 'fulltext', (None, 0, 6), []),
2020
self.assertEqual([(('new', ), 'N50 60'), (('new2', ), ' 0 6')],
2021
sorted(self.caught_entries[0]))
2022
self.assertEqual(1, len(self.caught_entries))
2024
def test_add_versions_delta_not_delta_index(self):
2025
index = self.two_graph_index(catch_adds=True)
2026
self.assertRaises(errors.KnitCorrupt, index.add_records,
2027
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
2028
self.assertEqual([], self.caught_entries)
2030
def test_add_versions_parents_not_parents_index(self):
2031
index = self.two_graph_index(catch_adds=True)
2032
self.assertRaises(errors.KnitCorrupt, index.add_records,
2033
[(('new',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
2034
self.assertEqual([], self.caught_entries)
2036
def test_add_versions_random_id_accepted(self):
2037
index = self.two_graph_index(catch_adds=True)
2038
index.add_records([], random_id=True)
2040
def test_add_versions_same_dup(self):
2041
index = self.two_graph_index(catch_adds=True)
2042
# options can be spelt two different ways
2043
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
2044
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
2045
# position/length are ignored (because each pack could have fulltext or
2046
# delta, and be at a different position.
2047
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
2048
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
2049
# but neither should have added data.
2050
self.assertEqual([[], [], [], []], self.caught_entries)
2052
def test_add_versions_different_dup(self):
2053
index = self.two_graph_index(catch_adds=True)
2055
self.assertRaises(errors.KnitCorrupt, index.add_records,
2056
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2057
self.assertRaises(errors.KnitCorrupt, index.add_records,
2058
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
2059
self.assertRaises(errors.KnitCorrupt, index.add_records,
2060
[(('tip',), 'fulltext', (None, 0, 100), [])])
2062
self.assertRaises(errors.KnitCorrupt, index.add_records,
2063
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
2064
# change options in the second record
2065
self.assertRaises(errors.KnitCorrupt, index.add_records,
2066
[(('tip',), 'fulltext,no-eol', (None, 0, 100), []),
2067
(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2068
self.assertEqual([], self.caught_entries)
2071
class TestKnitVersionedFiles(KnitTests):
2073
def assertGroupKeysForIo(self, exp_groups, keys, non_local_keys,
2074
positions, _min_buffer_size=None):
2075
kvf = self.make_test_knit()
2076
if _min_buffer_size is None:
2077
_min_buffer_size = knit._STREAM_MIN_BUFFER_SIZE
2078
self.assertEqual(exp_groups, kvf._group_keys_for_io(keys,
2079
non_local_keys, positions,
2080
_min_buffer_size=_min_buffer_size))
2082
def assertSplitByPrefix(self, expected_map, expected_prefix_order,
2084
split, prefix_order = KnitVersionedFiles._split_by_prefix(keys)
2085
self.assertEqual(expected_map, split)
2086
self.assertEqual(expected_prefix_order, prefix_order)
2088
def test__group_keys_for_io(self):
2089
ft_detail = ('fulltext', False)
2090
ld_detail = ('line-delta', False)
2098
f_a: (ft_detail, (f_a, 0, 100), None),
2099
f_b: (ld_detail, (f_b, 100, 21), f_a),
2100
f_c: (ld_detail, (f_c, 180, 15), f_b),
2101
g_a: (ft_detail, (g_a, 121, 35), None),
2102
g_b: (ld_detail, (g_b, 156, 12), g_a),
2103
g_c: (ld_detail, (g_c, 195, 13), g_a),
2105
self.assertGroupKeysForIo([([f_a], set())],
2106
[f_a], [], positions)
2107
self.assertGroupKeysForIo([([f_a], set([f_a]))],
2108
[f_a], [f_a], positions)
2109
self.assertGroupKeysForIo([([f_a, f_b], set([]))],
2110
[f_a, f_b], [], positions)
2111
self.assertGroupKeysForIo([([f_a, f_b], set([f_b]))],
2112
[f_a, f_b], [f_b], positions)
2113
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2114
[f_a, g_a, f_b, g_b], [], positions)
2115
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2116
[f_a, g_a, f_b, g_b], [], positions,
2117
_min_buffer_size=150)
2118
self.assertGroupKeysForIo([([f_a, f_b], set()), ([g_a, g_b], set())],
2119
[f_a, g_a, f_b, g_b], [], positions,
2120
_min_buffer_size=100)
2121
self.assertGroupKeysForIo([([f_c], set()), ([g_b], set())],
2122
[f_c, g_b], [], positions,
2123
_min_buffer_size=125)
2124
self.assertGroupKeysForIo([([g_b, f_c], set())],
2125
[g_b, f_c], [], positions,
2126
_min_buffer_size=125)
2128
def test__split_by_prefix(self):
2129
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2130
'g': [('g', 'b'), ('g', 'a')],
2132
[('f', 'a'), ('g', 'b'),
2133
('g', 'a'), ('f', 'b')])
2135
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2136
'g': [('g', 'b'), ('g', 'a')],
2138
[('f', 'a'), ('f', 'b'),
2139
('g', 'b'), ('g', 'a')])
2141
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2142
'g': [('g', 'b'), ('g', 'a')],
2144
[('f', 'a'), ('f', 'b'),
2145
('g', 'b'), ('g', 'a')])
2147
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2148
'g': [('g', 'b'), ('g', 'a')],
2149
'': [('a',), ('b',)]
2151
[('f', 'a'), ('g', 'b'),
2153
('g', 'a'), ('f', 'b')])
2156
class TestStacking(KnitTests):
2158
def get_basis_and_test_knit(self):
2159
basis = self.make_test_knit(name='basis')
2160
basis = RecordingVersionedFilesDecorator(basis)
2161
test = self.make_test_knit(name='test')
2162
test.add_fallback_versioned_files(basis)
2165
def test_add_fallback_versioned_files(self):
2166
basis = self.make_test_knit(name='basis')
2167
test = self.make_test_knit(name='test')
2168
# It must not error; other tests test that the fallback is referred to
2169
# when accessing data.
2170
test.add_fallback_versioned_files(basis)
2172
def test_add_lines(self):
2173
# lines added to the test are not added to the basis
2174
basis, test = self.get_basis_and_test_knit()
2176
key_basis = ('bar',)
2177
key_cross_border = ('quux',)
2178
key_delta = ('zaphod',)
2179
test.add_lines(key, (), ['foo\n'])
2180
self.assertEqual({}, basis.get_parent_map([key]))
2181
# lines added to the test that reference across the stack do a
2183
basis.add_lines(key_basis, (), ['foo\n'])
2185
test.add_lines(key_cross_border, (key_basis,), ['foo\n'])
2186
self.assertEqual('fulltext', test._index.get_method(key_cross_border))
2187
# we don't even need to look at the basis to see that this should be
2188
# stored as a fulltext
2189
self.assertEqual([], basis.calls)
2190
# Subsequent adds do delta.
2192
test.add_lines(key_delta, (key_cross_border,), ['foo\n'])
2193
self.assertEqual('line-delta', test._index.get_method(key_delta))
2194
self.assertEqual([], basis.calls)
2196
def test_annotate(self):
2197
# annotations from the test knit are answered without asking the basis
2198
basis, test = self.get_basis_and_test_knit()
2200
key_basis = ('bar',)
2201
key_missing = ('missing',)
2202
test.add_lines(key, (), ['foo\n'])
2203
details = test.annotate(key)
2204
self.assertEqual([(key, 'foo\n')], details)
2205
self.assertEqual([], basis.calls)
2206
# But texts that are not in the test knit are looked for in the basis
2208
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2210
details = test.annotate(key_basis)
2211
self.assertEqual([(key_basis, 'foo\n'), (key_basis, 'bar\n')], details)
2212
# Not optimised to date:
2213
# self.assertEqual([("annotate", key_basis)], basis.calls)
2214
self.assertEqual([('get_parent_map', set([key_basis])),
2215
('get_parent_map', set([key_basis])),
2216
('get_record_stream', [key_basis], 'topological', True)],
2219
def test_check(self):
2220
# At the moment checking a stacked knit does implicitly check the
2222
basis, test = self.get_basis_and_test_knit()
2225
def test_get_parent_map(self):
2226
# parents in the test knit are answered without asking the basis
2227
basis, test = self.get_basis_and_test_knit()
2229
key_basis = ('bar',)
2230
key_missing = ('missing',)
2231
test.add_lines(key, (), [])
2232
parent_map = test.get_parent_map([key])
2233
self.assertEqual({key: ()}, parent_map)
2234
self.assertEqual([], basis.calls)
2235
# But parents that are not in the test knit are looked for in the basis
2236
basis.add_lines(key_basis, (), [])
2238
parent_map = test.get_parent_map([key, key_basis, key_missing])
2239
self.assertEqual({key: (),
2240
key_basis: ()}, parent_map)
2241
self.assertEqual([("get_parent_map", set([key_basis, key_missing]))],
2244
def test_get_record_stream_unordered_fulltexts(self):
2245
# records from the test knit are answered without asking the basis:
2246
basis, test = self.get_basis_and_test_knit()
2248
key_basis = ('bar',)
2249
key_missing = ('missing',)
2250
test.add_lines(key, (), ['foo\n'])
2251
records = list(test.get_record_stream([key], 'unordered', True))
2252
self.assertEqual(1, len(records))
2253
self.assertEqual([], basis.calls)
2254
# Missing (from test knit) objects are retrieved from the basis:
2255
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2257
records = list(test.get_record_stream([key_basis, key_missing],
2259
self.assertEqual(2, len(records))
2260
calls = list(basis.calls)
2261
for record in records:
2262
self.assertSubset([record.key], (key_basis, key_missing))
2263
if record.key == key_missing:
2264
self.assertIsInstance(record, AbsentContentFactory)
2266
reference = list(basis.get_record_stream([key_basis],
2267
'unordered', True))[0]
2268
self.assertEqual(reference.key, record.key)
2269
self.assertEqual(reference.sha1, record.sha1)
2270
self.assertEqual(reference.storage_kind, record.storage_kind)
2271
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2272
record.get_bytes_as(record.storage_kind))
2273
self.assertEqual(reference.get_bytes_as('fulltext'),
2274
record.get_bytes_as('fulltext'))
2275
# It's not strictly minimal, but it seems reasonable for now for it to
2276
# ask which fallbacks have which parents.
2278
("get_parent_map", set([key_basis, key_missing])),
2279
("get_record_stream", [key_basis], 'unordered', True)],
2282
def test_get_record_stream_ordered_fulltexts(self):
2283
# ordering is preserved down into the fallback store.
2284
basis, test = self.get_basis_and_test_knit()
2286
key_basis = ('bar',)
2287
key_basis_2 = ('quux',)
2288
key_missing = ('missing',)
2289
test.add_lines(key, (key_basis,), ['foo\n'])
2290
# Missing (from test knit) objects are retrieved from the basis:
2291
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2292
basis.add_lines(key_basis_2, (), ['quux\n'])
2294
# ask for in non-topological order
2295
records = list(test.get_record_stream(
2296
[key, key_basis, key_missing, key_basis_2], 'topological', True))
2297
self.assertEqual(4, len(records))
2299
for record in records:
2300
self.assertSubset([record.key],
2301
(key_basis, key_missing, key_basis_2, key))
2302
if record.key == key_missing:
2303
self.assertIsInstance(record, AbsentContentFactory)
2305
results.append((record.key, record.sha1, record.storage_kind,
2306
record.get_bytes_as('fulltext')))
2307
calls = list(basis.calls)
2308
order = [record[0] for record in results]
2309
self.assertEqual([key_basis_2, key_basis, key], order)
2310
for result in results:
2311
if result[0] == key:
2315
record = source.get_record_stream([result[0]], 'unordered',
2317
self.assertEqual(record.key, result[0])
2318
self.assertEqual(record.sha1, result[1])
2319
# We used to check that the storage kind matched, but actually it
2320
# depends on whether it was sourced from the basis, or in a single
2321
# group, because asking for full texts returns proxy objects to a
2322
# _ContentMapGenerator object; so checking the kind is unneeded.
2323
self.assertEqual(record.get_bytes_as('fulltext'), result[3])
2324
# It's not strictly minimal, but it seems reasonable for now for it to
2325
# ask which fallbacks have which parents.
2327
("get_parent_map", set([key_basis, key_basis_2, key_missing])),
2328
# topological is requested from the fallback, because that is what
2329
# was requested at the top level.
2330
("get_record_stream", [key_basis_2, key_basis], 'topological', True)],
2333
def test_get_record_stream_unordered_deltas(self):
2334
# records from the test knit are answered without asking the basis:
2335
basis, test = self.get_basis_and_test_knit()
2337
key_basis = ('bar',)
2338
key_missing = ('missing',)
2339
test.add_lines(key, (), ['foo\n'])
2340
records = list(test.get_record_stream([key], 'unordered', False))
2341
self.assertEqual(1, len(records))
2342
self.assertEqual([], basis.calls)
2343
# Missing (from test knit) objects are retrieved from the basis:
2344
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2346
records = list(test.get_record_stream([key_basis, key_missing],
2347
'unordered', False))
2348
self.assertEqual(2, len(records))
2349
calls = list(basis.calls)
2350
for record in records:
2351
self.assertSubset([record.key], (key_basis, key_missing))
2352
if record.key == key_missing:
2353
self.assertIsInstance(record, AbsentContentFactory)
2355
reference = list(basis.get_record_stream([key_basis],
2356
'unordered', False))[0]
2357
self.assertEqual(reference.key, record.key)
2358
self.assertEqual(reference.sha1, record.sha1)
2359
self.assertEqual(reference.storage_kind, record.storage_kind)
2360
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2361
record.get_bytes_as(record.storage_kind))
2362
# It's not strictly minimal, but it seems reasonable for now for it to
2363
# ask which fallbacks have which parents.
2365
("get_parent_map", set([key_basis, key_missing])),
2366
("get_record_stream", [key_basis], 'unordered', False)],
2369
def test_get_record_stream_ordered_deltas(self):
2370
# ordering is preserved down into the fallback store.
2371
basis, test = self.get_basis_and_test_knit()
2373
key_basis = ('bar',)
2374
key_basis_2 = ('quux',)
2375
key_missing = ('missing',)
2376
test.add_lines(key, (key_basis,), ['foo\n'])
2377
# Missing (from test knit) objects are retrieved from the basis:
2378
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2379
basis.add_lines(key_basis_2, (), ['quux\n'])
2381
# ask for in non-topological order
2382
records = list(test.get_record_stream(
2383
[key, key_basis, key_missing, key_basis_2], 'topological', False))
2384
self.assertEqual(4, len(records))
2386
for record in records:
2387
self.assertSubset([record.key],
2388
(key_basis, key_missing, key_basis_2, key))
2389
if record.key == key_missing:
2390
self.assertIsInstance(record, AbsentContentFactory)
2392
results.append((record.key, record.sha1, record.storage_kind,
2393
record.get_bytes_as(record.storage_kind)))
2394
calls = list(basis.calls)
2395
order = [record[0] for record in results]
2396
self.assertEqual([key_basis_2, key_basis, key], order)
2397
for result in results:
2398
if result[0] == key:
2402
record = source.get_record_stream([result[0]], 'unordered',
2404
self.assertEqual(record.key, result[0])
2405
self.assertEqual(record.sha1, result[1])
2406
self.assertEqual(record.storage_kind, result[2])
2407
self.assertEqual(record.get_bytes_as(record.storage_kind), result[3])
2408
# It's not strictly minimal, but it seems reasonable for now for it to
2409
# ask which fallbacks have which parents.
2411
("get_parent_map", set([key_basis, key_basis_2, key_missing])),
2412
("get_record_stream", [key_basis_2, key_basis], 'topological', False)],
2415
def test_get_sha1s(self):
2416
# sha1's in the test knit are answered without asking the basis
2417
basis, test = self.get_basis_and_test_knit()
2419
key_basis = ('bar',)
2420
key_missing = ('missing',)
2421
test.add_lines(key, (), ['foo\n'])
2422
key_sha1sum = osutils.sha('foo\n').hexdigest()
2423
sha1s = test.get_sha1s([key])
2424
self.assertEqual({key: key_sha1sum}, sha1s)
2425
self.assertEqual([], basis.calls)
2426
# But texts that are not in the test knit are looked for in the basis
2427
# directly (rather than via text reconstruction) so that remote servers
2428
# etc don't have to answer with full content.
2429
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2430
basis_sha1sum = osutils.sha('foo\nbar\n').hexdigest()
2432
sha1s = test.get_sha1s([key, key_missing, key_basis])
2433
self.assertEqual({key: key_sha1sum,
2434
key_basis: basis_sha1sum}, sha1s)
2435
self.assertEqual([("get_sha1s", set([key_basis, key_missing]))],
2438
def test_insert_record_stream(self):
2439
# records are inserted as normal; insert_record_stream builds on
2440
# add_lines, so a smoke test should be all that's needed:
2442
key_basis = ('bar',)
2443
key_delta = ('zaphod',)
2444
basis, test = self.get_basis_and_test_knit()
2445
source = self.make_test_knit(name='source')
2446
basis.add_lines(key_basis, (), ['foo\n'])
2448
source.add_lines(key_basis, (), ['foo\n'])
2449
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2450
stream = source.get_record_stream([key_delta], 'unordered', False)
2451
test.insert_record_stream(stream)
2452
# XXX: this does somewhat too many calls in making sure of whether it
2453
# has to recreate the full text.
2454
self.assertEqual([("get_parent_map", set([key_basis])),
2455
('get_parent_map', set([key_basis])),
2456
('get_record_stream', [key_basis], 'unordered', True)],
2458
self.assertEqual({key_delta:(key_basis,)},
2459
test.get_parent_map([key_delta]))
2460
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2461
'unordered', True).next().get_bytes_as('fulltext'))
2463
def test_iter_lines_added_or_present_in_keys(self):
2464
# Lines from the basis are returned, and lines for a given key are only
2468
# all sources are asked for keys:
2469
basis, test = self.get_basis_and_test_knit()
2470
basis.add_lines(key1, (), ["foo"])
2472
lines = list(test.iter_lines_added_or_present_in_keys([key1]))
2473
self.assertEqual([("foo\n", key1)], lines)
2474
self.assertEqual([("iter_lines_added_or_present_in_keys", set([key1]))],
2476
# keys in both are not duplicated:
2477
test.add_lines(key2, (), ["bar\n"])
2478
basis.add_lines(key2, (), ["bar\n"])
2480
lines = list(test.iter_lines_added_or_present_in_keys([key2]))
2481
self.assertEqual([("bar\n", key2)], lines)
2482
self.assertEqual([], basis.calls)
2484
def test_keys(self):
2487
# all sources are asked for keys:
2488
basis, test = self.get_basis_and_test_knit()
2490
self.assertEqual(set(), set(keys))
2491
self.assertEqual([("keys",)], basis.calls)
2492
# keys from a basis are returned:
2493
basis.add_lines(key1, (), [])
2496
self.assertEqual(set([key1]), set(keys))
2497
self.assertEqual([("keys",)], basis.calls)
2498
# keys in both are not duplicated:
2499
test.add_lines(key2, (), [])
2500
basis.add_lines(key2, (), [])
2503
self.assertEqual(2, len(keys))
2504
self.assertEqual(set([key1, key2]), set(keys))
2505
self.assertEqual([("keys",)], basis.calls)
2507
def test_add_mpdiffs(self):
2508
# records are inserted as normal; add_mpdiff builds on
2509
# add_lines, so a smoke test should be all that's needed:
2511
key_basis = ('bar',)
2512
key_delta = ('zaphod',)
2513
basis, test = self.get_basis_and_test_knit()
2514
source = self.make_test_knit(name='source')
2515
basis.add_lines(key_basis, (), ['foo\n'])
2517
source.add_lines(key_basis, (), ['foo\n'])
2518
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2519
diffs = source.make_mpdiffs([key_delta])
2520
test.add_mpdiffs([(key_delta, (key_basis,),
2521
source.get_sha1s([key_delta])[key_delta], diffs[0])])
2522
self.assertEqual([("get_parent_map", set([key_basis])),
2523
('get_record_stream', [key_basis], 'unordered', True),],
2525
self.assertEqual({key_delta:(key_basis,)},
2526
test.get_parent_map([key_delta]))
2527
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2528
'unordered', True).next().get_bytes_as('fulltext'))
2530
def test_make_mpdiffs(self):
2531
# Generating an mpdiff across a stacking boundary should detect parent
2535
key_right = ('zaphod',)
2536
basis, test = self.get_basis_and_test_knit()
2537
basis.add_lines(key_left, (), ['bar\n'])
2538
basis.add_lines(key_right, (), ['zaphod\n'])
2540
test.add_lines(key, (key_left, key_right),
2541
['bar\n', 'foo\n', 'zaphod\n'])
2542
diffs = test.make_mpdiffs([key])
2544
multiparent.MultiParent([multiparent.ParentText(0, 0, 0, 1),
2545
multiparent.NewText(['foo\n']),
2546
multiparent.ParentText(1, 0, 2, 1)])],
2548
self.assertEqual(3, len(basis.calls))
2550
("get_parent_map", set([key_left, key_right])),
2551
("get_parent_map", set([key_left, key_right])),
2554
last_call = basis.calls[-1]
2555
self.assertEqual('get_record_stream', last_call[0])
2556
self.assertEqual(set([key_left, key_right]), set(last_call[1]))
2557
self.assertEqual('topological', last_call[2])
2558
self.assertEqual(True, last_call[3])
2561
class TestNetworkBehaviour(KnitTests):
2562
"""Tests for getting data out of/into knits over the network."""
2564
def test_include_delta_closure_generates_a_knit_delta_closure(self):
2565
vf = self.make_test_knit(name='test')
2566
# put in three texts, giving ft, delta, delta
2567
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2568
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2569
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2570
# But heuristics could interfere, so check what happened:
2571
self.assertEqual(['knit-ft-gz', 'knit-delta-gz', 'knit-delta-gz'],
2572
[record.storage_kind for record in
2573
vf.get_record_stream([('base',), ('d1',), ('d2',)],
2574
'topological', False)])
2575
# generate a stream of just the deltas include_delta_closure=True,
2576
# serialise to the network, and check that we get a delta closure on the wire.
2577
stream = vf.get_record_stream([('d1',), ('d2',)], 'topological', True)
2578
netb = [record.get_bytes_as(record.storage_kind) for record in stream]
2579
# The first bytes should be a memo from _ContentMapGenerator, and the
2580
# second bytes should be empty (because its a API proxy not something
2581
# for wire serialisation.
2582
self.assertEqual('', netb[1])
2584
kind, line_end = network_bytes_to_kind_and_offset(bytes)
2585
self.assertEqual('knit-delta-closure', kind)
2588
class TestContentMapGenerator(KnitTests):
2589
"""Tests for ContentMapGenerator"""
2591
def test_get_record_stream_gives_records(self):
2592
vf = self.make_test_knit(name='test')
2593
# put in three texts, giving ft, delta, delta
2594
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2595
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2596
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2597
keys = [('d1',), ('d2',)]
2598
generator = _VFContentMapGenerator(vf, keys,
2599
global_map=vf.get_parent_map(keys))
2600
for record in generator.get_record_stream():
2601
if record.key == ('d1',):
2602
self.assertEqual('d1\n', record.get_bytes_as('fulltext'))
2604
self.assertEqual('d2\n', record.get_bytes_as('fulltext'))
2606
def test_get_record_stream_kinds_are_raw(self):
2607
vf = self.make_test_knit(name='test')
2608
# put in three texts, giving ft, delta, delta
2609
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2610
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2611
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2612
keys = [('base',), ('d1',), ('d2',)]
2613
generator = _VFContentMapGenerator(vf, keys,
2614
global_map=vf.get_parent_map(keys))
2615
kinds = {('base',): 'knit-delta-closure',
2616
('d1',): 'knit-delta-closure-ref',
2617
('d2',): 'knit-delta-closure-ref',
2619
for record in generator.get_record_stream():
2620
self.assertEqual(kinds[record.key], record.storage_kind)
added
removed
bzrlib/tests/test_knit.py
