1
# Copyright (C) 2006-2011 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
"""Tests for Knit data structure"""
19
from cStringIO import StringIO
32
from bzrlib.errors import (
36
from bzrlib.index import *
37
from bzrlib.knit import (
42
_VFContentMapGenerator,
48
from bzrlib.patiencediff import PatienceSequenceMatcher
49
from bzrlib.repofmt import (
53
from bzrlib.tests import (
55
TestCaseWithMemoryTransport,
56
TestCaseWithTransport,
59
from bzrlib.versionedfile import (
62
network_bytes_to_kind_and_offset,
63
RecordingVersionedFilesDecorator,
65
from bzrlib.tests import (
70
compiled_knit_feature = features.ModuleAvailableFeature(
71
'bzrlib._knit_load_data_pyx')
74
class KnitContentTestsMixin(object):
76
def test_constructor(self):
77
content = self._make_content([])
80
content = self._make_content([])
81
self.assertEqual(content.text(), [])
83
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
84
self.assertEqual(content.text(), ["text1", "text2"])
87
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
89
self.assertIsInstance(copy, content.__class__)
90
self.assertEqual(copy.annotate(), content.annotate())
92
def assertDerivedBlocksEqual(self, source, target, noeol=False):
93
"""Assert that the derived matching blocks match real output"""
94
source_lines = source.splitlines(True)
95
target_lines = target.splitlines(True)
97
if noeol and not line.endswith('\n'):
101
source_content = self._make_content([(None, nl(l)) for l in source_lines])
102
target_content = self._make_content([(None, nl(l)) for l in target_lines])
103
line_delta = source_content.line_delta(target_content)
104
delta_blocks = list(KnitContent.get_line_delta_blocks(line_delta,
105
source_lines, target_lines))
106
matcher = PatienceSequenceMatcher(None, source_lines, target_lines)
107
matcher_blocks = list(matcher.get_matching_blocks())
108
self.assertEqual(matcher_blocks, delta_blocks)
110
def test_get_line_delta_blocks(self):
111
self.assertDerivedBlocksEqual('a\nb\nc\n', 'q\nc\n')
112
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1)
113
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1A)
114
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1B)
115
self.assertDerivedBlocksEqual(TEXT_1B, TEXT_1A)
116
self.assertDerivedBlocksEqual(TEXT_1A, TEXT_1B)
117
self.assertDerivedBlocksEqual(TEXT_1A, '')
118
self.assertDerivedBlocksEqual('', TEXT_1A)
119
self.assertDerivedBlocksEqual('', '')
120
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd')
122
def test_get_line_delta_blocks_noeol(self):
123
"""Handle historical knit deltas safely
125
Some existing knit deltas don't consider the last line to differ
126
when the only difference whether it has a final newline.
128
New knit deltas appear to always consider the last line to differ
131
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd\n', noeol=True)
132
self.assertDerivedBlocksEqual('a\nb\nc\nd\n', 'a\nb\nc', noeol=True)
133
self.assertDerivedBlocksEqual('a\nb\nc\n', 'a\nb\nc', noeol=True)
134
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\n', noeol=True)
146
Banana cup cake recipe
156
Banana cup cake recipe
158
- bananas (do not use plantains!!!)
165
Banana cup cake recipe
182
class TestPlainKnitContent(TestCase, KnitContentTestsMixin):
184
def _make_content(self, lines):
185
annotated_content = AnnotatedKnitContent(lines)
186
return PlainKnitContent(annotated_content.text(), 'bogus')
188
def test_annotate(self):
189
content = self._make_content([])
190
self.assertEqual(content.annotate(), [])
192
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
193
self.assertEqual(content.annotate(),
194
[("bogus", "text1"), ("bogus", "text2")])
196
def test_line_delta(self):
197
content1 = self._make_content([("", "a"), ("", "b")])
198
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
199
self.assertEqual(content1.line_delta(content2),
200
[(1, 2, 2, ["a", "c"])])
202
def test_line_delta_iter(self):
203
content1 = self._make_content([("", "a"), ("", "b")])
204
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
205
it = content1.line_delta_iter(content2)
206
self.assertEqual(it.next(), (1, 2, 2, ["a", "c"]))
207
self.assertRaises(StopIteration, it.next)
210
class TestAnnotatedKnitContent(TestCase, KnitContentTestsMixin):
212
def _make_content(self, lines):
213
return AnnotatedKnitContent(lines)
215
def test_annotate(self):
216
content = self._make_content([])
217
self.assertEqual(content.annotate(), [])
219
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
220
self.assertEqual(content.annotate(),
221
[("origin1", "text1"), ("origin2", "text2")])
223
def test_line_delta(self):
224
content1 = self._make_content([("", "a"), ("", "b")])
225
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
226
self.assertEqual(content1.line_delta(content2),
227
[(1, 2, 2, [("", "a"), ("", "c")])])
229
def test_line_delta_iter(self):
230
content1 = self._make_content([("", "a"), ("", "b")])
231
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
232
it = content1.line_delta_iter(content2)
233
self.assertEqual(it.next(), (1, 2, 2, [("", "a"), ("", "c")]))
234
self.assertRaises(StopIteration, it.next)
237
class MockTransport(object):
239
def __init__(self, file_lines=None):
240
self.file_lines = file_lines
242
# We have no base directory for the MockTransport
245
def get(self, filename):
246
if self.file_lines is None:
247
raise NoSuchFile(filename)
249
return StringIO("\n".join(self.file_lines))
251
def readv(self, relpath, offsets):
252
fp = self.get(relpath)
253
for offset, size in offsets:
255
yield offset, fp.read(size)
257
def __getattr__(self, name):
258
def queue_call(*args, **kwargs):
259
self.calls.append((name, args, kwargs))
263
class MockReadvFailingTransport(MockTransport):
264
"""Fail in the middle of a readv() result.
266
This Transport will successfully yield the first two requested hunks, but
267
raise NoSuchFile for the rest.
270
def readv(self, relpath, offsets):
272
for result in MockTransport.readv(self, relpath, offsets):
274
# we use 2 because the first offset is the pack header, the second
275
# is the first actual content requset
277
raise errors.NoSuchFile(relpath)
281
class KnitRecordAccessTestsMixin(object):
282
"""Tests for getting and putting knit records."""
284
def test_add_raw_records(self):
285
"""Add_raw_records adds records retrievable later."""
286
access = self.get_access()
287
memos = access.add_raw_records([('key', 10)], '1234567890')
288
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
290
def test_add_several_raw_records(self):
291
"""add_raw_records with many records and read some back."""
292
access = self.get_access()
293
memos = access.add_raw_records([('key', 10), ('key2', 2), ('key3', 5)],
295
self.assertEqual(['1234567890', '12', '34567'],
296
list(access.get_raw_records(memos)))
297
self.assertEqual(['1234567890'],
298
list(access.get_raw_records(memos[0:1])))
299
self.assertEqual(['12'],
300
list(access.get_raw_records(memos[1:2])))
301
self.assertEqual(['34567'],
302
list(access.get_raw_records(memos[2:3])))
303
self.assertEqual(['1234567890', '34567'],
304
list(access.get_raw_records(memos[0:1] + memos[2:3])))
307
class TestKnitKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
308
"""Tests for the .kndx implementation."""
310
def get_access(self):
311
"""Get a .knit style access instance."""
312
mapper = ConstantMapper("foo")
313
access = _KnitKeyAccess(self.get_transport(), mapper)
317
class _TestException(Exception):
318
"""Just an exception for local tests to use."""
321
class TestPackKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
322
"""Tests for the pack based access."""
324
def get_access(self):
325
return self._get_access()[0]
327
def _get_access(self, packname='packfile', index='FOO'):
328
transport = self.get_transport()
329
def write_data(bytes):
330
transport.append_bytes(packname, bytes)
331
writer = pack.ContainerWriter(write_data)
333
access = pack_repo._DirectPackAccess({})
334
access.set_writer(writer, index, (transport, packname))
335
return access, writer
337
def make_pack_file(self):
338
"""Create a pack file with 2 records."""
339
access, writer = self._get_access(packname='packname', index='foo')
341
memos.extend(access.add_raw_records([('key1', 10)], '1234567890'))
342
memos.extend(access.add_raw_records([('key2', 5)], '12345'))
346
def test_pack_collection_pack_retries(self):
347
"""An explicit pack of a pack collection succeeds even when a
348
concurrent pack happens.
350
builder = self.make_branch_builder('.')
351
builder.start_series()
352
builder.build_snapshot('rev-1', None, [
353
('add', ('', 'root-id', 'directory', None)),
354
('add', ('file', 'file-id', 'file', 'content\nrev 1\n')),
356
builder.build_snapshot('rev-2', ['rev-1'], [
357
('modify', ('file-id', 'content\nrev 2\n')),
359
builder.build_snapshot('rev-3', ['rev-2'], [
360
('modify', ('file-id', 'content\nrev 3\n')),
362
self.addCleanup(builder.finish_series)
363
b = builder.get_branch()
364
self.addCleanup(b.lock_write().unlock)
366
collection = repo._pack_collection
367
# Concurrently repack the repo.
368
reopened_repo = repo.bzrdir.open_repository()
373
def make_vf_for_retrying(self):
374
"""Create 3 packs and a reload function.
376
Originally, 2 pack files will have the data, but one will be missing.
377
And then the third will be used in place of the first two if reload()
380
:return: (versioned_file, reload_counter)
381
versioned_file a KnitVersionedFiles using the packs for access
383
builder = self.make_branch_builder('.', format="1.9")
384
builder.start_series()
385
builder.build_snapshot('rev-1', None, [
386
('add', ('', 'root-id', 'directory', None)),
387
('add', ('file', 'file-id', 'file', 'content\nrev 1\n')),
389
builder.build_snapshot('rev-2', ['rev-1'], [
390
('modify', ('file-id', 'content\nrev 2\n')),
392
builder.build_snapshot('rev-3', ['rev-2'], [
393
('modify', ('file-id', 'content\nrev 3\n')),
395
builder.finish_series()
396
b = builder.get_branch()
398
self.addCleanup(b.unlock)
399
# Pack these three revisions into another pack file, but don't remove
402
collection = repo._pack_collection
403
collection.ensure_loaded()
404
orig_packs = collection.packs
405
packer = knitpack_repo.KnitPacker(collection, orig_packs, '.testpack')
406
new_pack = packer.pack()
407
# forget about the new pack
411
# Set up a reload() function that switches to using the new pack file
412
new_index = new_pack.revision_index
413
access_tuple = new_pack.access_tuple()
414
reload_counter = [0, 0, 0]
416
reload_counter[0] += 1
417
if reload_counter[1] > 0:
418
# We already reloaded, nothing more to do
419
reload_counter[2] += 1
421
reload_counter[1] += 1
422
vf._index._graph_index._indices[:] = [new_index]
423
vf._access._indices.clear()
424
vf._access._indices[new_index] = access_tuple
426
# Delete one of the pack files so the data will need to be reloaded. We
427
# will delete the file with 'rev-2' in it
428
trans, name = orig_packs[1].access_tuple()
430
# We don't have the index trigger reloading because we want to test
431
# that we reload when the .pack disappears
432
vf._access._reload_func = reload
433
return vf, reload_counter
435
def make_reload_func(self, return_val=True):
438
reload_called[0] += 1
440
return reload_called, reload
442
def make_retry_exception(self):
443
# We raise a real exception so that sys.exc_info() is properly
446
raise _TestException('foobar')
447
except _TestException, e:
448
retry_exc = errors.RetryWithNewPacks(None, reload_occurred=False,
449
exc_info=sys.exc_info())
450
# GZ 2010-08-10: Cycle with exc_info affects 3 tests
453
def test_read_from_several_packs(self):
454
access, writer = self._get_access()
456
memos.extend(access.add_raw_records([('key', 10)], '1234567890'))
458
access, writer = self._get_access('pack2', 'FOOBAR')
459
memos.extend(access.add_raw_records([('key', 5)], '12345'))
461
access, writer = self._get_access('pack3', 'BAZ')
462
memos.extend(access.add_raw_records([('key', 5)], 'alpha'))
464
transport = self.get_transport()
465
access = pack_repo._DirectPackAccess({"FOO":(transport, 'packfile'),
466
"FOOBAR":(transport, 'pack2'),
467
"BAZ":(transport, 'pack3')})
468
self.assertEqual(['1234567890', '12345', 'alpha'],
469
list(access.get_raw_records(memos)))
470
self.assertEqual(['1234567890'],
471
list(access.get_raw_records(memos[0:1])))
472
self.assertEqual(['12345'],
473
list(access.get_raw_records(memos[1:2])))
474
self.assertEqual(['alpha'],
475
list(access.get_raw_records(memos[2:3])))
476
self.assertEqual(['1234567890', 'alpha'],
477
list(access.get_raw_records(memos[0:1] + memos[2:3])))
479
def test_set_writer(self):
480
"""The writer should be settable post construction."""
481
access = pack_repo._DirectPackAccess({})
482
transport = self.get_transport()
483
packname = 'packfile'
485
def write_data(bytes):
486
transport.append_bytes(packname, bytes)
487
writer = pack.ContainerWriter(write_data)
489
access.set_writer(writer, index, (transport, packname))
490
memos = access.add_raw_records([('key', 10)], '1234567890')
492
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
494
def test_missing_index_raises_retry(self):
495
memos = self.make_pack_file()
496
transport = self.get_transport()
497
reload_called, reload_func = self.make_reload_func()
498
# Note that the index key has changed from 'foo' to 'bar'
499
access = pack_repo._DirectPackAccess({'bar':(transport, 'packname')},
500
reload_func=reload_func)
501
e = self.assertListRaises(errors.RetryWithNewPacks,
502
access.get_raw_records, memos)
503
# Because a key was passed in which does not match our index list, we
504
# assume that the listing was already reloaded
505
self.assertTrue(e.reload_occurred)
506
self.assertIsInstance(e.exc_info, tuple)
507
self.assertIs(e.exc_info[0], KeyError)
508
self.assertIsInstance(e.exc_info[1], KeyError)
510
def test_missing_index_raises_key_error_with_no_reload(self):
511
memos = self.make_pack_file()
512
transport = self.get_transport()
513
# Note that the index key has changed from 'foo' to 'bar'
514
access = pack_repo._DirectPackAccess({'bar':(transport, 'packname')})
515
e = self.assertListRaises(KeyError, access.get_raw_records, memos)
517
def test_missing_file_raises_retry(self):
518
memos = self.make_pack_file()
519
transport = self.get_transport()
520
reload_called, reload_func = self.make_reload_func()
521
# Note that the 'filename' has been changed to 'different-packname'
522
access = pack_repo._DirectPackAccess(
523
{'foo':(transport, 'different-packname')},
524
reload_func=reload_func)
525
e = self.assertListRaises(errors.RetryWithNewPacks,
526
access.get_raw_records, memos)
527
# The file has gone missing, so we assume we need to reload
528
self.assertFalse(e.reload_occurred)
529
self.assertIsInstance(e.exc_info, tuple)
530
self.assertIs(e.exc_info[0], errors.NoSuchFile)
531
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
532
self.assertEqual('different-packname', e.exc_info[1].path)
534
def test_missing_file_raises_no_such_file_with_no_reload(self):
535
memos = self.make_pack_file()
536
transport = self.get_transport()
537
# Note that the 'filename' has been changed to 'different-packname'
538
access = pack_repo._DirectPackAccess(
539
{'foo': (transport, 'different-packname')})
540
e = self.assertListRaises(errors.NoSuchFile,
541
access.get_raw_records, memos)
543
def test_failing_readv_raises_retry(self):
544
memos = self.make_pack_file()
545
transport = self.get_transport()
546
failing_transport = MockReadvFailingTransport(
547
[transport.get_bytes('packname')])
548
reload_called, reload_func = self.make_reload_func()
549
access = pack_repo._DirectPackAccess(
550
{'foo': (failing_transport, 'packname')},
551
reload_func=reload_func)
552
# Asking for a single record will not trigger the Mock failure
553
self.assertEqual(['1234567890'],
554
list(access.get_raw_records(memos[:1])))
555
self.assertEqual(['12345'],
556
list(access.get_raw_records(memos[1:2])))
557
# A multiple offset readv() will fail mid-way through
558
e = self.assertListRaises(errors.RetryWithNewPacks,
559
access.get_raw_records, memos)
560
# The file has gone missing, so we assume we need to reload
561
self.assertFalse(e.reload_occurred)
562
self.assertIsInstance(e.exc_info, tuple)
563
self.assertIs(e.exc_info[0], errors.NoSuchFile)
564
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
565
self.assertEqual('packname', e.exc_info[1].path)
567
def test_failing_readv_raises_no_such_file_with_no_reload(self):
568
memos = self.make_pack_file()
569
transport = self.get_transport()
570
failing_transport = MockReadvFailingTransport(
571
[transport.get_bytes('packname')])
572
reload_called, reload_func = self.make_reload_func()
573
access = pack_repo._DirectPackAccess(
574
{'foo':(failing_transport, 'packname')})
575
# Asking for a single record will not trigger the Mock failure
576
self.assertEqual(['1234567890'],
577
list(access.get_raw_records(memos[:1])))
578
self.assertEqual(['12345'],
579
list(access.get_raw_records(memos[1:2])))
580
# A multiple offset readv() will fail mid-way through
581
e = self.assertListRaises(errors.NoSuchFile,
582
access.get_raw_records, memos)
584
def test_reload_or_raise_no_reload(self):
585
access = pack_repo._DirectPackAccess({}, reload_func=None)
586
retry_exc = self.make_retry_exception()
587
# Without a reload_func, we will just re-raise the original exception
588
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
590
def test_reload_or_raise_reload_changed(self):
591
reload_called, reload_func = self.make_reload_func(return_val=True)
592
access = pack_repo._DirectPackAccess({}, reload_func=reload_func)
593
retry_exc = self.make_retry_exception()
594
access.reload_or_raise(retry_exc)
595
self.assertEqual([1], reload_called)
596
retry_exc.reload_occurred=True
597
access.reload_or_raise(retry_exc)
598
self.assertEqual([2], reload_called)
600
def test_reload_or_raise_reload_no_change(self):
601
reload_called, reload_func = self.make_reload_func(return_val=False)
602
access = pack_repo._DirectPackAccess({}, reload_func=reload_func)
603
retry_exc = self.make_retry_exception()
604
# If reload_occurred is False, then we consider it an error to have
605
# reload_func() return False (no changes).
606
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
607
self.assertEqual([1], reload_called)
608
retry_exc.reload_occurred=True
609
# If reload_occurred is True, then we assume nothing changed because
610
# it had changed earlier, but didn't change again
611
access.reload_or_raise(retry_exc)
612
self.assertEqual([2], reload_called)
614
def test_annotate_retries(self):
615
vf, reload_counter = self.make_vf_for_retrying()
616
# It is a little bit bogus to annotate the Revision VF, but it works,
617
# as we have ancestry stored there
619
reload_lines = vf.annotate(key)
620
self.assertEqual([1, 1, 0], reload_counter)
621
plain_lines = vf.annotate(key)
622
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
623
if reload_lines != plain_lines:
624
self.fail('Annotation was not identical with reloading.')
625
# Now delete the packs-in-use, which should trigger another reload, but
626
# this time we just raise an exception because we can't recover
627
for trans, name in vf._access._indices.itervalues():
629
self.assertRaises(errors.NoSuchFile, vf.annotate, key)
630
self.assertEqual([2, 1, 1], reload_counter)
632
def test__get_record_map_retries(self):
633
vf, reload_counter = self.make_vf_for_retrying()
634
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
635
records = vf._get_record_map(keys)
636
self.assertEqual(keys, sorted(records.keys()))
637
self.assertEqual([1, 1, 0], reload_counter)
638
# Now delete the packs-in-use, which should trigger another reload, but
639
# this time we just raise an exception because we can't recover
640
for trans, name in vf._access._indices.itervalues():
642
self.assertRaises(errors.NoSuchFile, vf._get_record_map, keys)
643
self.assertEqual([2, 1, 1], reload_counter)
645
def test_get_record_stream_retries(self):
646
vf, reload_counter = self.make_vf_for_retrying()
647
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
648
record_stream = vf.get_record_stream(keys, 'topological', False)
649
record = record_stream.next()
650
self.assertEqual(('rev-1',), record.key)
651
self.assertEqual([0, 0, 0], reload_counter)
652
record = record_stream.next()
653
self.assertEqual(('rev-2',), record.key)
654
self.assertEqual([1, 1, 0], reload_counter)
655
record = record_stream.next()
656
self.assertEqual(('rev-3',), record.key)
657
self.assertEqual([1, 1, 0], reload_counter)
658
# Now delete all pack files, and see that we raise the right error
659
for trans, name in vf._access._indices.itervalues():
661
self.assertListRaises(errors.NoSuchFile,
662
vf.get_record_stream, keys, 'topological', False)
664
def test_iter_lines_added_or_present_in_keys_retries(self):
665
vf, reload_counter = self.make_vf_for_retrying()
666
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
667
# Unfortunately, iter_lines_added_or_present_in_keys iterates the
668
# result in random order (determined by the iteration order from a
669
# set()), so we don't have any solid way to trigger whether data is
670
# read before or after. However we tried to delete the middle node to
671
# exercise the code well.
672
# What we care about is that all lines are always yielded, but not
675
reload_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
676
self.assertEqual([1, 1, 0], reload_counter)
677
# Now do it again, to make sure the result is equivalent
678
plain_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
679
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
680
self.assertEqual(plain_lines, reload_lines)
681
self.assertEqual(21, len(plain_lines))
682
# Now delete all pack files, and see that we raise the right error
683
for trans, name in vf._access._indices.itervalues():
685
self.assertListRaises(errors.NoSuchFile,
686
vf.iter_lines_added_or_present_in_keys, keys)
687
self.assertEqual([2, 1, 1], reload_counter)
689
def test_get_record_stream_yields_disk_sorted_order(self):
690
# if we get 'unordered' pick a semi-optimal order for reading. The
691
# order should be grouped by pack file, and then by position in file
692
repo = self.make_repository('test', format='pack-0.92')
694
self.addCleanup(repo.unlock)
695
repo.start_write_group()
697
vf.add_lines(('f-id', 'rev-5'), [('f-id', 'rev-4')], ['lines\n'])
698
vf.add_lines(('f-id', 'rev-1'), [], ['lines\n'])
699
vf.add_lines(('f-id', 'rev-2'), [('f-id', 'rev-1')], ['lines\n'])
700
repo.commit_write_group()
701
# We inserted them as rev-5, rev-1, rev-2, we should get them back in
703
stream = vf.get_record_stream([('f-id', 'rev-1'), ('f-id', 'rev-5'),
704
('f-id', 'rev-2')], 'unordered', False)
705
keys = [r.key for r in stream]
706
self.assertEqual([('f-id', 'rev-5'), ('f-id', 'rev-1'),
707
('f-id', 'rev-2')], keys)
708
repo.start_write_group()
709
vf.add_lines(('f-id', 'rev-4'), [('f-id', 'rev-3')], ['lines\n'])
710
vf.add_lines(('f-id', 'rev-3'), [('f-id', 'rev-2')], ['lines\n'])
711
vf.add_lines(('f-id', 'rev-6'), [('f-id', 'rev-5')], ['lines\n'])
712
repo.commit_write_group()
713
# Request in random order, to make sure the output order isn't based on
715
request_keys = set(('f-id', 'rev-%d' % i) for i in range(1, 7))
716
stream = vf.get_record_stream(request_keys, 'unordered', False)
717
keys = [r.key for r in stream]
718
# We want to get the keys back in disk order, but it doesn't matter
719
# which pack we read from first. So this can come back in 2 orders
720
alt1 = [('f-id', 'rev-%d' % i) for i in [4, 3, 6, 5, 1, 2]]
721
alt2 = [('f-id', 'rev-%d' % i) for i in [5, 1, 2, 4, 3, 6]]
722
if keys != alt1 and keys != alt2:
723
self.fail('Returned key order did not match either expected order.'
724
' expected %s or %s, not %s'
725
% (alt1, alt2, keys))
728
class LowLevelKnitDataTests(TestCase):
730
def create_gz_content(self, text):
732
gz_file = gzip.GzipFile(mode='wb', fileobj=sio)
735
return sio.getvalue()
737
def make_multiple_records(self):
738
"""Create the content for multiple records."""
739
sha1sum = osutils.sha_string('foo\nbar\n')
741
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
746
record_1 = (0, len(gz_txt), sha1sum)
747
total_txt.append(gz_txt)
748
sha1sum = osutils.sha_string('baz\n')
749
gz_txt = self.create_gz_content('version rev-id-2 1 %s\n'
753
record_2 = (record_1[1], len(gz_txt), sha1sum)
754
total_txt.append(gz_txt)
755
return total_txt, record_1, record_2
757
def test_valid_knit_data(self):
758
sha1sum = osutils.sha_string('foo\nbar\n')
759
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
764
transport = MockTransport([gz_txt])
765
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
766
knit = KnitVersionedFiles(None, access)
767
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
769
contents = list(knit._read_records_iter(records))
770
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'],
771
'4e48e2c9a3d2ca8a708cb0cc545700544efb5021')], contents)
773
raw_contents = list(knit._read_records_iter_raw(records))
774
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
776
def test_multiple_records_valid(self):
777
total_txt, record_1, record_2 = self.make_multiple_records()
778
transport = MockTransport([''.join(total_txt)])
779
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
780
knit = KnitVersionedFiles(None, access)
781
records = [(('rev-id-1',), (('rev-id-1',), record_1[0], record_1[1])),
782
(('rev-id-2',), (('rev-id-2',), record_2[0], record_2[1]))]
784
contents = list(knit._read_records_iter(records))
785
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'], record_1[2]),
786
(('rev-id-2',), ['baz\n'], record_2[2])],
789
raw_contents = list(knit._read_records_iter_raw(records))
790
self.assertEqual([(('rev-id-1',), total_txt[0], record_1[2]),
791
(('rev-id-2',), total_txt[1], record_2[2])],
794
def test_not_enough_lines(self):
795
sha1sum = osutils.sha_string('foo\n')
796
# record says 2 lines data says 1
797
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
801
transport = MockTransport([gz_txt])
802
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
803
knit = KnitVersionedFiles(None, access)
804
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
805
self.assertRaises(errors.KnitCorrupt, list,
806
knit._read_records_iter(records))
808
# read_records_iter_raw won't detect that sort of mismatch/corruption
809
raw_contents = list(knit._read_records_iter_raw(records))
810
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
812
def test_too_many_lines(self):
813
sha1sum = osutils.sha_string('foo\nbar\n')
814
# record says 1 lines data says 2
815
gz_txt = self.create_gz_content('version rev-id-1 1 %s\n'
820
transport = MockTransport([gz_txt])
821
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
822
knit = KnitVersionedFiles(None, access)
823
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
824
self.assertRaises(errors.KnitCorrupt, list,
825
knit._read_records_iter(records))
827
# read_records_iter_raw won't detect that sort of mismatch/corruption
828
raw_contents = list(knit._read_records_iter_raw(records))
829
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
831
def test_mismatched_version_id(self):
832
sha1sum = osutils.sha_string('foo\nbar\n')
833
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
838
transport = MockTransport([gz_txt])
839
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
840
knit = KnitVersionedFiles(None, access)
841
# We are asking for rev-id-2, but the data is rev-id-1
842
records = [(('rev-id-2',), (('rev-id-2',), 0, len(gz_txt)))]
843
self.assertRaises(errors.KnitCorrupt, list,
844
knit._read_records_iter(records))
846
# read_records_iter_raw detects mismatches in the header
847
self.assertRaises(errors.KnitCorrupt, list,
848
knit._read_records_iter_raw(records))
850
def test_uncompressed_data(self):
851
sha1sum = osutils.sha_string('foo\nbar\n')
852
txt = ('version rev-id-1 2 %s\n'
857
transport = MockTransport([txt])
858
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
859
knit = KnitVersionedFiles(None, access)
860
records = [(('rev-id-1',), (('rev-id-1',), 0, len(txt)))]
862
# We don't have valid gzip data ==> corrupt
863
self.assertRaises(errors.KnitCorrupt, list,
864
knit._read_records_iter(records))
866
# read_records_iter_raw will notice the bad data
867
self.assertRaises(errors.KnitCorrupt, list,
868
knit._read_records_iter_raw(records))
870
def test_corrupted_data(self):
871
sha1sum = osutils.sha_string('foo\nbar\n')
872
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
877
# Change 2 bytes in the middle to \xff
878
gz_txt = gz_txt[:10] + '\xff\xff' + gz_txt[12:]
879
transport = MockTransport([gz_txt])
880
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
881
knit = KnitVersionedFiles(None, access)
882
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
883
self.assertRaises(errors.KnitCorrupt, list,
884
knit._read_records_iter(records))
885
# read_records_iter_raw will barf on bad gz data
886
self.assertRaises(errors.KnitCorrupt, list,
887
knit._read_records_iter_raw(records))
890
class LowLevelKnitIndexTests(TestCase):
892
def get_knit_index(self, transport, name, mode):
893
mapper = ConstantMapper(name)
894
from bzrlib._knit_load_data_py import _load_data_py
895
self.overrideAttr(knit, '_load_data', _load_data_py)
896
allow_writes = lambda: 'w' in mode
897
return _KndxIndex(transport, mapper, lambda:None, allow_writes, lambda:True)
899
def test_create_file(self):
900
transport = MockTransport()
901
index = self.get_knit_index(transport, "filename", "w")
903
call = transport.calls.pop(0)
904
# call[1][1] is a StringIO - we can't test it by simple equality.
905
self.assertEqual('put_file_non_atomic', call[0])
906
self.assertEqual('filename.kndx', call[1][0])
907
# With no history, _KndxIndex writes a new index:
908
self.assertEqual(_KndxIndex.HEADER,
909
call[1][1].getvalue())
910
self.assertEqual({'create_parent_dir': True}, call[2])
912
def test_read_utf8_version_id(self):
913
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
914
utf8_revision_id = unicode_revision_id.encode('utf-8')
915
transport = MockTransport([
917
'%s option 0 1 :' % (utf8_revision_id,)
919
index = self.get_knit_index(transport, "filename", "r")
920
# _KndxIndex is a private class, and deals in utf8 revision_ids, not
921
# Unicode revision_ids.
922
self.assertEqual({(utf8_revision_id,):()},
923
index.get_parent_map(index.keys()))
924
self.assertFalse((unicode_revision_id,) in index.keys())
926
def test_read_utf8_parents(self):
927
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
928
utf8_revision_id = unicode_revision_id.encode('utf-8')
929
transport = MockTransport([
931
"version option 0 1 .%s :" % (utf8_revision_id,)
933
index = self.get_knit_index(transport, "filename", "r")
934
self.assertEqual({("version",):((utf8_revision_id,),)},
935
index.get_parent_map(index.keys()))
937
def test_read_ignore_corrupted_lines(self):
938
transport = MockTransport([
941
"corrupted options 0 1 .b .c ",
942
"version options 0 1 :"
944
index = self.get_knit_index(transport, "filename", "r")
945
self.assertEqual(1, len(index.keys()))
946
self.assertEqual(set([("version",)]), index.keys())
948
def test_read_corrupted_header(self):
949
transport = MockTransport(['not a bzr knit index header\n'])
950
index = self.get_knit_index(transport, "filename", "r")
951
self.assertRaises(KnitHeaderError, index.keys)
953
def test_read_duplicate_entries(self):
954
transport = MockTransport([
956
"parent options 0 1 :",
957
"version options1 0 1 0 :",
958
"version options2 1 2 .other :",
959
"version options3 3 4 0 .other :"
961
index = self.get_knit_index(transport, "filename", "r")
962
self.assertEqual(2, len(index.keys()))
963
# check that the index used is the first one written. (Specific
964
# to KnitIndex style indices.
965
self.assertEqual("1", index._dictionary_compress([("version",)]))
966
self.assertEqual((("version",), 3, 4), index.get_position(("version",)))
967
self.assertEqual(["options3"], index.get_options(("version",)))
968
self.assertEqual({("version",):(("parent",), ("other",))},
969
index.get_parent_map([("version",)]))
971
def test_read_compressed_parents(self):
972
transport = MockTransport([
976
"c option 0 1 1 0 :",
978
index = self.get_knit_index(transport, "filename", "r")
979
self.assertEqual({("b",):(("a",),), ("c",):(("b",), ("a",))},
980
index.get_parent_map([("b",), ("c",)]))
982
def test_write_utf8_version_id(self):
983
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
984
utf8_revision_id = unicode_revision_id.encode('utf-8')
985
transport = MockTransport([
988
index = self.get_knit_index(transport, "filename", "r")
990
((utf8_revision_id,), ["option"], ((utf8_revision_id,), 0, 1), [])])
991
call = transport.calls.pop(0)
992
# call[1][1] is a StringIO - we can't test it by simple equality.
993
self.assertEqual('put_file_non_atomic', call[0])
994
self.assertEqual('filename.kndx', call[1][0])
995
# With no history, _KndxIndex writes a new index:
996
self.assertEqual(_KndxIndex.HEADER +
997
"\n%s option 0 1 :" % (utf8_revision_id,),
998
call[1][1].getvalue())
999
self.assertEqual({'create_parent_dir': True}, call[2])
1001
def test_write_utf8_parents(self):
1002
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
1003
utf8_revision_id = unicode_revision_id.encode('utf-8')
1004
transport = MockTransport([
1007
index = self.get_knit_index(transport, "filename", "r")
1009
(("version",), ["option"], (("version",), 0, 1), [(utf8_revision_id,)])])
1010
call = transport.calls.pop(0)
1011
# call[1][1] is a StringIO - we can't test it by simple equality.
1012
self.assertEqual('put_file_non_atomic', call[0])
1013
self.assertEqual('filename.kndx', call[1][0])
1014
# With no history, _KndxIndex writes a new index:
1015
self.assertEqual(_KndxIndex.HEADER +
1016
"\nversion option 0 1 .%s :" % (utf8_revision_id,),
1017
call[1][1].getvalue())
1018
self.assertEqual({'create_parent_dir': True}, call[2])
1020
def test_keys(self):
1021
transport = MockTransport([
1024
index = self.get_knit_index(transport, "filename", "r")
1026
self.assertEqual(set(), index.keys())
1028
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1029
self.assertEqual(set([("a",)]), index.keys())
1031
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1032
self.assertEqual(set([("a",)]), index.keys())
1034
index.add_records([(("b",), ["option"], (("b",), 0, 1), [])])
1035
self.assertEqual(set([("a",), ("b",)]), index.keys())
1037
def add_a_b(self, index, random_id=None):
1039
if random_id is not None:
1040
kwargs["random_id"] = random_id
1042
(("a",), ["option"], (("a",), 0, 1), [("b",)]),
1043
(("a",), ["opt"], (("a",), 1, 2), [("c",)]),
1044
(("b",), ["option"], (("b",), 2, 3), [("a",)])
1047
def assertIndexIsAB(self, index):
1052
index.get_parent_map(index.keys()))
1053
self.assertEqual((("a",), 1, 2), index.get_position(("a",)))
1054
self.assertEqual((("b",), 2, 3), index.get_position(("b",)))
1055
self.assertEqual(["opt"], index.get_options(("a",)))
1057
def test_add_versions(self):
1058
transport = MockTransport([
1061
index = self.get_knit_index(transport, "filename", "r")
1064
call = transport.calls.pop(0)
1065
# call[1][1] is a StringIO - we can't test it by simple equality.
1066
self.assertEqual('put_file_non_atomic', call[0])
1067
self.assertEqual('filename.kndx', call[1][0])
1068
# With no history, _KndxIndex writes a new index:
1071
"\na option 0 1 .b :"
1073
"\nb option 2 3 0 :",
1074
call[1][1].getvalue())
1075
self.assertEqual({'create_parent_dir': True}, call[2])
1076
self.assertIndexIsAB(index)
1078
def test_add_versions_random_id_is_accepted(self):
1079
transport = MockTransport([
1082
index = self.get_knit_index(transport, "filename", "r")
1083
self.add_a_b(index, random_id=True)
1085
def test_delay_create_and_add_versions(self):
1086
transport = MockTransport()
1088
index = self.get_knit_index(transport, "filename", "w")
1090
self.assertEqual([], transport.calls)
1093
#[ {"dir_mode": 0777, "create_parent_dir": True, "mode": "wb"},
1095
# Two calls: one during which we load the existing index (and when its
1096
# missing create it), then a second where we write the contents out.
1097
self.assertEqual(2, len(transport.calls))
1098
call = transport.calls.pop(0)
1099
self.assertEqual('put_file_non_atomic', call[0])
1100
self.assertEqual('filename.kndx', call[1][0])
1101
# With no history, _KndxIndex writes a new index:
1102
self.assertEqual(_KndxIndex.HEADER, call[1][1].getvalue())
1103
self.assertEqual({'create_parent_dir': True}, call[2])
1104
call = transport.calls.pop(0)
1105
# call[1][1] is a StringIO - we can't test it by simple equality.
1106
self.assertEqual('put_file_non_atomic', call[0])
1107
self.assertEqual('filename.kndx', call[1][0])
1108
# With no history, _KndxIndex writes a new index:
1111
"\na option 0 1 .b :"
1113
"\nb option 2 3 0 :",
1114
call[1][1].getvalue())
1115
self.assertEqual({'create_parent_dir': True}, call[2])
1117
def assertTotalBuildSize(self, size, keys, positions):
1118
self.assertEqual(size,
1119
knit._get_total_build_size(None, keys, positions))
1121
def test__get_total_build_size(self):
1123
('a',): (('fulltext', False), (('a',), 0, 100), None),
1124
('b',): (('line-delta', False), (('b',), 100, 21), ('a',)),
1125
('c',): (('line-delta', False), (('c',), 121, 35), ('b',)),
1126
('d',): (('line-delta', False), (('d',), 156, 12), ('b',)),
1128
self.assertTotalBuildSize(100, [('a',)], positions)
1129
self.assertTotalBuildSize(121, [('b',)], positions)
1130
# c needs both a & b
1131
self.assertTotalBuildSize(156, [('c',)], positions)
1132
# we shouldn't count 'b' twice
1133
self.assertTotalBuildSize(156, [('b',), ('c',)], positions)
1134
self.assertTotalBuildSize(133, [('d',)], positions)
1135
self.assertTotalBuildSize(168, [('c',), ('d',)], positions)
1137
def test_get_position(self):
1138
transport = MockTransport([
1143
index = self.get_knit_index(transport, "filename", "r")
1145
self.assertEqual((("a",), 0, 1), index.get_position(("a",)))
1146
self.assertEqual((("b",), 1, 2), index.get_position(("b",)))
1148
def test_get_method(self):
1149
transport = MockTransport([
1151
"a fulltext,unknown 0 1 :",
1152
"b unknown,line-delta 1 2 :",
1155
index = self.get_knit_index(transport, "filename", "r")
1157
self.assertEqual("fulltext", index.get_method("a"))
1158
self.assertEqual("line-delta", index.get_method("b"))
1159
self.assertRaises(errors.KnitIndexUnknownMethod, index.get_method, "c")
1161
def test_get_options(self):
1162
transport = MockTransport([
1167
index = self.get_knit_index(transport, "filename", "r")
1169
self.assertEqual(["opt1"], index.get_options("a"))
1170
self.assertEqual(["opt2", "opt3"], index.get_options("b"))
1172
def test_get_parent_map(self):
1173
transport = MockTransport([
1176
"b option 1 2 0 .c :",
1177
"c option 1 2 1 0 .e :"
1179
index = self.get_knit_index(transport, "filename", "r")
1183
("b",):(("a",), ("c",)),
1184
("c",):(("b",), ("a",), ("e",)),
1185
}, index.get_parent_map(index.keys()))
1187
def test_impossible_parent(self):
1188
"""Test we get KnitCorrupt if the parent couldn't possibly exist."""
1189
transport = MockTransport([
1192
"b option 0 1 4 :" # We don't have a 4th record
1194
index = self.get_knit_index(transport, 'filename', 'r')
1196
self.assertRaises(errors.KnitCorrupt, index.keys)
1197
except TypeError, e:
1198
if (str(e) == ('exceptions must be strings, classes, or instances,'
1199
' not exceptions.IndexError')):
1200
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1201
' raising new style exceptions with python'
1206
def test_corrupted_parent(self):
1207
transport = MockTransport([
1211
"c option 0 1 1v :", # Can't have a parent of '1v'
1213
index = self.get_knit_index(transport, 'filename', 'r')
1215
self.assertRaises(errors.KnitCorrupt, index.keys)
1216
except TypeError, e:
1217
if (str(e) == ('exceptions must be strings, classes, or instances,'
1218
' not exceptions.ValueError')):
1219
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1220
' raising new style exceptions with python'
1225
def test_corrupted_parent_in_list(self):
1226
transport = MockTransport([
1230
"c option 0 1 1 v :", # Can't have a parent of 'v'
1232
index = self.get_knit_index(transport, 'filename', 'r')
1234
self.assertRaises(errors.KnitCorrupt, index.keys)
1235
except TypeError, e:
1236
if (str(e) == ('exceptions must be strings, classes, or instances,'
1237
' not exceptions.ValueError')):
1238
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1239
' raising new style exceptions with python'
1244
def test_invalid_position(self):
1245
transport = MockTransport([
1249
index = self.get_knit_index(transport, 'filename', 'r')
1251
self.assertRaises(errors.KnitCorrupt, index.keys)
1252
except TypeError, e:
1253
if (str(e) == ('exceptions must be strings, classes, or instances,'
1254
' not exceptions.ValueError')):
1255
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1256
' raising new style exceptions with python'
1261
def test_invalid_size(self):
1262
transport = MockTransport([
1266
index = self.get_knit_index(transport, 'filename', 'r')
1268
self.assertRaises(errors.KnitCorrupt, index.keys)
1269
except TypeError, e:
1270
if (str(e) == ('exceptions must be strings, classes, or instances,'
1271
' not exceptions.ValueError')):
1272
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1273
' raising new style exceptions with python'
1278
def test_scan_unvalidated_index_not_implemented(self):
1279
transport = MockTransport()
1280
index = self.get_knit_index(transport, 'filename', 'r')
1282
NotImplementedError, index.scan_unvalidated_index,
1283
'dummy graph_index')
1285
NotImplementedError, index.get_missing_compression_parents)
1287
def test_short_line(self):
1288
transport = MockTransport([
1291
"b option 10 10 0", # This line isn't terminated, ignored
1293
index = self.get_knit_index(transport, "filename", "r")
1294
self.assertEqual(set([('a',)]), index.keys())
1296
def test_skip_incomplete_record(self):
1297
# A line with bogus data should just be skipped
1298
transport = MockTransport([
1301
"b option 10 10 0", # This line isn't terminated, ignored
1302
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1304
index = self.get_knit_index(transport, "filename", "r")
1305
self.assertEqual(set([('a',), ('c',)]), index.keys())
1307
def test_trailing_characters(self):
1308
# A line with bogus data should just be skipped
1309
transport = MockTransport([
1312
"b option 10 10 0 :a", # This line has extra trailing characters
1313
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1315
index = self.get_knit_index(transport, "filename", "r")
1316
self.assertEqual(set([('a',), ('c',)]), index.keys())
1319
class LowLevelKnitIndexTests_c(LowLevelKnitIndexTests):
1321
_test_needs_features = [compiled_knit_feature]
1323
def get_knit_index(self, transport, name, mode):
1324
mapper = ConstantMapper(name)
1325
from bzrlib._knit_load_data_pyx import _load_data_c
1326
self.overrideAttr(knit, '_load_data', _load_data_c)
1327
allow_writes = lambda: mode == 'w'
1328
return _KndxIndex(transport, mapper, lambda:None,
1329
allow_writes, lambda:True)
1332
class Test_KnitAnnotator(TestCaseWithMemoryTransport):
1334
def make_annotator(self):
1335
factory = knit.make_pack_factory(True, True, 1)
1336
vf = factory(self.get_transport())
1337
return knit._KnitAnnotator(vf)
1339
def test__expand_fulltext(self):
1340
ann = self.make_annotator()
1341
rev_key = ('rev-id',)
1342
ann._num_compression_children[rev_key] = 1
1343
res = ann._expand_record(rev_key, (('parent-id',),), None,
1344
['line1\n', 'line2\n'], ('fulltext', True))
1345
# The content object and text lines should be cached appropriately
1346
self.assertEqual(['line1\n', 'line2'], res)
1347
content_obj = ann._content_objects[rev_key]
1348
self.assertEqual(['line1\n', 'line2\n'], content_obj._lines)
1349
self.assertEqual(res, content_obj.text())
1350
self.assertEqual(res, ann._text_cache[rev_key])
1352
def test__expand_delta_comp_parent_not_available(self):
1353
# Parent isn't available yet, so we return nothing, but queue up this
1354
# node for later processing
1355
ann = self.make_annotator()
1356
rev_key = ('rev-id',)
1357
parent_key = ('parent-id',)
1358
record = ['0,1,1\n', 'new-line\n']
1359
details = ('line-delta', False)
1360
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1362
self.assertEqual(None, res)
1363
self.assertTrue(parent_key in ann._pending_deltas)
1364
pending = ann._pending_deltas[parent_key]
1365
self.assertEqual(1, len(pending))
1366
self.assertEqual((rev_key, (parent_key,), record, details), pending[0])
1368
def test__expand_record_tracks_num_children(self):
1369
ann = self.make_annotator()
1370
rev_key = ('rev-id',)
1371
rev2_key = ('rev2-id',)
1372
parent_key = ('parent-id',)
1373
record = ['0,1,1\n', 'new-line\n']
1374
details = ('line-delta', False)
1375
ann._num_compression_children[parent_key] = 2
1376
ann._expand_record(parent_key, (), None, ['line1\n', 'line2\n'],
1377
('fulltext', False))
1378
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1380
self.assertEqual({parent_key: 1}, ann._num_compression_children)
1381
# Expanding the second child should remove the content object, and the
1382
# num_compression_children entry
1383
res = ann._expand_record(rev2_key, (parent_key,), parent_key,
1385
self.assertFalse(parent_key in ann._content_objects)
1386
self.assertEqual({}, ann._num_compression_children)
1387
# We should not cache the content_objects for rev2 and rev, because
1388
# they do not have compression children of their own.
1389
self.assertEqual({}, ann._content_objects)
1391
def test__expand_delta_records_blocks(self):
1392
ann = self.make_annotator()
1393
rev_key = ('rev-id',)
1394
parent_key = ('parent-id',)
1395
record = ['0,1,1\n', 'new-line\n']
1396
details = ('line-delta', True)
1397
ann._num_compression_children[parent_key] = 2
1398
ann._expand_record(parent_key, (), None,
1399
['line1\n', 'line2\n', 'line3\n'],
1400
('fulltext', False))
1401
ann._expand_record(rev_key, (parent_key,), parent_key, record, details)
1402
self.assertEqual({(rev_key, parent_key): [(1, 1, 1), (3, 3, 0)]},
1403
ann._matching_blocks)
1404
rev2_key = ('rev2-id',)
1405
record = ['0,1,1\n', 'new-line\n']
1406
details = ('line-delta', False)
1407
ann._expand_record(rev2_key, (parent_key,), parent_key, record, details)
1408
self.assertEqual([(1, 1, 2), (3, 3, 0)],
1409
ann._matching_blocks[(rev2_key, parent_key)])
1411
def test__get_parent_ann_uses_matching_blocks(self):
1412
ann = self.make_annotator()
1413
rev_key = ('rev-id',)
1414
parent_key = ('parent-id',)
1415
parent_ann = [(parent_key,)]*3
1416
block_key = (rev_key, parent_key)
1417
ann._annotations_cache[parent_key] = parent_ann
1418
ann._matching_blocks[block_key] = [(0, 1, 1), (3, 3, 0)]
1419
# We should not try to access any parent_lines content, because we know
1420
# we already have the matching blocks
1421
par_ann, blocks = ann._get_parent_annotations_and_matches(rev_key,
1422
['1\n', '2\n', '3\n'], parent_key)
1423
self.assertEqual(parent_ann, par_ann)
1424
self.assertEqual([(0, 1, 1), (3, 3, 0)], blocks)
1425
self.assertEqual({}, ann._matching_blocks)
1427
def test__process_pending(self):
1428
ann = self.make_annotator()
1429
rev_key = ('rev-id',)
1432
record = ['0,1,1\n', 'new-line\n']
1433
details = ('line-delta', False)
1434
p1_record = ['line1\n', 'line2\n']
1435
ann._num_compression_children[p1_key] = 1
1436
res = ann._expand_record(rev_key, (p1_key,p2_key), p1_key,
1438
self.assertEqual(None, res)
1439
# self.assertTrue(p1_key in ann._pending_deltas)
1440
self.assertEqual({}, ann._pending_annotation)
1441
# Now insert p1, and we should be able to expand the delta
1442
res = ann._expand_record(p1_key, (), None, p1_record,
1443
('fulltext', False))
1444
self.assertEqual(p1_record, res)
1445
ann._annotations_cache[p1_key] = [(p1_key,)]*2
1446
res = ann._process_pending(p1_key)
1447
self.assertEqual([], res)
1448
self.assertFalse(p1_key in ann._pending_deltas)
1449
self.assertTrue(p2_key in ann._pending_annotation)
1450
self.assertEqual({p2_key: [(rev_key, (p1_key, p2_key))]},
1451
ann._pending_annotation)
1452
# Now fill in parent 2, and pending annotation should be satisfied
1453
res = ann._expand_record(p2_key, (), None, [], ('fulltext', False))
1454
ann._annotations_cache[p2_key] = []
1455
res = ann._process_pending(p2_key)
1456
self.assertEqual([rev_key], res)
1457
self.assertEqual({}, ann._pending_annotation)
1458
self.assertEqual({}, ann._pending_deltas)
1460
def test_record_delta_removes_basis(self):
1461
ann = self.make_annotator()
1462
ann._expand_record(('parent-id',), (), None,
1463
['line1\n', 'line2\n'], ('fulltext', False))
1464
ann._num_compression_children['parent-id'] = 2
1466
def test_annotate_special_text(self):
1467
ann = self.make_annotator()
1469
rev1_key = ('rev-1',)
1470
rev2_key = ('rev-2',)
1471
rev3_key = ('rev-3',)
1472
spec_key = ('special:',)
1473
vf.add_lines(rev1_key, [], ['initial content\n'])
1474
vf.add_lines(rev2_key, [rev1_key], ['initial content\n',
1477
vf.add_lines(rev3_key, [rev1_key], ['initial content\n',
1480
spec_text = ('initial content\n'
1484
ann.add_special_text(spec_key, [rev2_key, rev3_key], spec_text)
1485
anns, lines = ann.annotate(spec_key)
1486
self.assertEqual([(rev1_key,),
1487
(rev2_key, rev3_key),
1491
self.assertEqualDiff(spec_text, ''.join(lines))
1494
class KnitTests(TestCaseWithTransport):
1495
"""Class containing knit test helper routines."""
1497
def make_test_knit(self, annotate=False, name='test'):
1498
mapper = ConstantMapper(name)
1499
return make_file_factory(annotate, mapper)(self.get_transport())
1502
class TestBadShaError(KnitTests):
1503
"""Tests for handling of sha errors."""
1505
def test_sha_exception_has_text(self):
1506
# having the failed text included in the error allows for recovery.
1507
source = self.make_test_knit()
1508
target = self.make_test_knit(name="target")
1509
if not source._max_delta_chain:
1510
raise TestNotApplicable(
1511
"cannot get delta-caused sha failures without deltas.")
1514
broken = ('broken',)
1515
source.add_lines(basis, (), ['foo\n'])
1516
source.add_lines(broken, (basis,), ['foo\n', 'bar\n'])
1517
# Seed target with a bad basis text
1518
target.add_lines(basis, (), ['gam\n'])
1519
target.insert_record_stream(
1520
source.get_record_stream([broken], 'unordered', False))
1521
err = self.assertRaises(errors.KnitCorrupt,
1522
target.get_record_stream([broken], 'unordered', True
1523
).next().get_bytes_as, 'chunked')
1524
self.assertEqual(['gam\n', 'bar\n'], err.content)
1525
# Test for formatting with live data
1526
self.assertStartsWith(str(err), "Knit ")
1529
class TestKnitIndex(KnitTests):
1531
def test_add_versions_dictionary_compresses(self):
1532
"""Adding versions to the index should update the lookup dict"""
1533
knit = self.make_test_knit()
1535
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1536
self.check_file_contents('test.kndx',
1537
'# bzr knit index 8\n'
1539
'a-1 fulltext 0 0 :'
1542
(('a-2',), ['fulltext'], (('a-2',), 0, 0), [('a-1',)]),
1543
(('a-3',), ['fulltext'], (('a-3',), 0, 0), [('a-2',)]),
1545
self.check_file_contents('test.kndx',
1546
'# bzr knit index 8\n'
1548
'a-1 fulltext 0 0 :\n'
1549
'a-2 fulltext 0 0 0 :\n'
1550
'a-3 fulltext 0 0 1 :'
1552
self.assertEqual(set([('a-3',), ('a-1',), ('a-2',)]), idx.keys())
1554
('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False)),
1555
('a-2',): ((('a-2',), 0, 0), None, (('a-1',),), ('fulltext', False)),
1556
('a-3',): ((('a-3',), 0, 0), None, (('a-2',),), ('fulltext', False)),
1557
}, idx.get_build_details(idx.keys()))
1558
self.assertEqual({('a-1',):(),
1559
('a-2',):(('a-1',),),
1560
('a-3',):(('a-2',),),},
1561
idx.get_parent_map(idx.keys()))
1563
def test_add_versions_fails_clean(self):
1564
"""If add_versions fails in the middle, it restores a pristine state.
1566
Any modifications that are made to the index are reset if all versions
1569
# This cheats a little bit by passing in a generator which will
1570
# raise an exception before the processing finishes
1571
# Other possibilities would be to have an version with the wrong number
1572
# of entries, or to make the backing transport unable to write any
1575
knit = self.make_test_knit()
1577
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1579
class StopEarly(Exception):
1582
def generate_failure():
1583
"""Add some entries and then raise an exception"""
1584
yield (('a-2',), ['fulltext'], (None, 0, 0), ('a-1',))
1585
yield (('a-3',), ['fulltext'], (None, 0, 0), ('a-2',))
1588
# Assert the pre-condition
1590
self.assertEqual(set([('a-1',)]), set(idx.keys()))
1592
{('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False))},
1593
idx.get_build_details([('a-1',)]))
1594
self.assertEqual({('a-1',):()}, idx.get_parent_map(idx.keys()))
1597
self.assertRaises(StopEarly, idx.add_records, generate_failure())
1598
# And it shouldn't be modified
1601
def test_knit_index_ignores_empty_files(self):
1602
# There was a race condition in older bzr, where a ^C at the right time
1603
# could leave an empty .kndx file, which bzr would later claim was a
1604
# corrupted file since the header was not present. In reality, the file
1605
# just wasn't created, so it should be ignored.
1606
t = transport.get_transport_from_path('.')
1607
t.put_bytes('test.kndx', '')
1609
knit = self.make_test_knit()
1611
def test_knit_index_checks_header(self):
1612
t = transport.get_transport_from_path('.')
1613
t.put_bytes('test.kndx', '# not really a knit header\n\n')
1614
k = self.make_test_knit()
1615
self.assertRaises(KnitHeaderError, k.keys)
1618
class TestGraphIndexKnit(KnitTests):
1619
"""Tests for knits using a GraphIndex rather than a KnitIndex."""
1621
def make_g_index(self, name, ref_lists=0, nodes=[]):
1622
builder = GraphIndexBuilder(ref_lists)
1623
for node, references, value in nodes:
1624
builder.add_node(node, references, value)
1625
stream = builder.finish()
1626
trans = self.get_transport()
1627
size = trans.put_file(name, stream)
1628
return GraphIndex(trans, name, size)
1630
def two_graph_index(self, deltas=False, catch_adds=False):
1631
"""Build a two-graph index.
1633
:param deltas: If true, use underlying indices with two node-ref
1634
lists and 'parent' set to a delta-compressed against tail.
1636
# build a complex graph across several indices.
1638
# delta compression inn the index
1639
index1 = self.make_g_index('1', 2, [
1640
(('tip', ), 'N0 100', ([('parent', )], [], )),
1641
(('tail', ), '', ([], []))])
1642
index2 = self.make_g_index('2', 2, [
1643
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], [('tail', )])),
1644
(('separate', ), '', ([], []))])
1646
# just blob location and graph in the index.
1647
index1 = self.make_g_index('1', 1, [
1648
(('tip', ), 'N0 100', ([('parent', )], )),
1649
(('tail', ), '', ([], ))])
1650
index2 = self.make_g_index('2', 1, [
1651
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], )),
1652
(('separate', ), '', ([], ))])
1653
combined_index = CombinedGraphIndex([index1, index2])
1655
self.combined_index = combined_index
1656
self.caught_entries = []
1657
add_callback = self.catch_add
1660
return _KnitGraphIndex(combined_index, lambda:True, deltas=deltas,
1661
add_callback=add_callback)
1663
def test_keys(self):
1664
index = self.two_graph_index()
1665
self.assertEqual(set([('tail',), ('tip',), ('parent',), ('separate',)]),
1668
def test_get_position(self):
1669
index = self.two_graph_index()
1670
self.assertEqual((index._graph_index._indices[0], 0, 100), index.get_position(('tip',)))
1671
self.assertEqual((index._graph_index._indices[1], 100, 78), index.get_position(('parent',)))
1673
def test_get_method_deltas(self):
1674
index = self.two_graph_index(deltas=True)
1675
self.assertEqual('fulltext', index.get_method(('tip',)))
1676
self.assertEqual('line-delta', index.get_method(('parent',)))
1678
def test_get_method_no_deltas(self):
1679
# check that the parent-history lookup is ignored with deltas=False.
1680
index = self.two_graph_index(deltas=False)
1681
self.assertEqual('fulltext', index.get_method(('tip',)))
1682
self.assertEqual('fulltext', index.get_method(('parent',)))
1684
def test_get_options_deltas(self):
1685
index = self.two_graph_index(deltas=True)
1686
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1687
self.assertEqual(['line-delta'], index.get_options(('parent',)))
1689
def test_get_options_no_deltas(self):
1690
# check that the parent-history lookup is ignored with deltas=False.
1691
index = self.two_graph_index(deltas=False)
1692
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1693
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1695
def test_get_parent_map(self):
1696
index = self.two_graph_index()
1697
self.assertEqual({('parent',):(('tail',), ('ghost',))},
1698
index.get_parent_map([('parent',), ('ghost',)]))
1700
def catch_add(self, entries):
1701
self.caught_entries.append(entries)
1703
def test_add_no_callback_errors(self):
1704
index = self.two_graph_index()
1705
self.assertRaises(errors.ReadOnlyError, index.add_records,
1706
[(('new',), 'fulltext,no-eol', (None, 50, 60), ['separate'])])
1708
def test_add_version_smoke(self):
1709
index = self.two_graph_index(catch_adds=True)
1710
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60),
1712
self.assertEqual([[(('new', ), 'N50 60', ((('separate',),),))]],
1713
self.caught_entries)
1715
def test_add_version_delta_not_delta_index(self):
1716
index = self.two_graph_index(catch_adds=True)
1717
self.assertRaises(errors.KnitCorrupt, index.add_records,
1718
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1719
self.assertEqual([], self.caught_entries)
1721
def test_add_version_same_dup(self):
1722
index = self.two_graph_index(catch_adds=True)
1723
# options can be spelt two different ways
1724
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
1725
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
1726
# position/length are ignored (because each pack could have fulltext or
1727
# delta, and be at a different position.
1728
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1730
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1732
# but neither should have added data:
1733
self.assertEqual([[], [], [], []], self.caught_entries)
1735
def test_add_version_different_dup(self):
1736
index = self.two_graph_index(deltas=True, catch_adds=True)
1738
self.assertRaises(errors.KnitCorrupt, index.add_records,
1739
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1740
self.assertRaises(errors.KnitCorrupt, index.add_records,
1741
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1743
self.assertRaises(errors.KnitCorrupt, index.add_records,
1744
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1745
self.assertEqual([], self.caught_entries)
1747
def test_add_versions_nodeltas(self):
1748
index = self.two_graph_index(catch_adds=True)
1750
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1751
(('new2',), 'fulltext', (None, 0, 6), [('new',)]),
1753
self.assertEqual([(('new', ), 'N50 60', ((('separate',),),)),
1754
(('new2', ), ' 0 6', ((('new',),),))],
1755
sorted(self.caught_entries[0]))
1756
self.assertEqual(1, len(self.caught_entries))
1758
def test_add_versions_deltas(self):
1759
index = self.two_graph_index(deltas=True, catch_adds=True)
1761
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1762
(('new2',), 'line-delta', (None, 0, 6), [('new',)]),
1764
self.assertEqual([(('new', ), 'N50 60', ((('separate',),), ())),
1765
(('new2', ), ' 0 6', ((('new',),), (('new',),), ))],
1766
sorted(self.caught_entries[0]))
1767
self.assertEqual(1, len(self.caught_entries))
1769
def test_add_versions_delta_not_delta_index(self):
1770
index = self.two_graph_index(catch_adds=True)
1771
self.assertRaises(errors.KnitCorrupt, index.add_records,
1772
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1773
self.assertEqual([], self.caught_entries)
1775
def test_add_versions_random_id_accepted(self):
1776
index = self.two_graph_index(catch_adds=True)
1777
index.add_records([], random_id=True)
1779
def test_add_versions_same_dup(self):
1780
index = self.two_graph_index(catch_adds=True)
1781
# options can be spelt two different ways
1782
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100),
1784
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100),
1786
# position/length are ignored (because each pack could have fulltext or
1787
# delta, and be at a different position.
1788
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1790
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1792
# but neither should have added data.
1793
self.assertEqual([[], [], [], []], self.caught_entries)
1795
def test_add_versions_different_dup(self):
1796
index = self.two_graph_index(deltas=True, catch_adds=True)
1798
self.assertRaises(errors.KnitCorrupt, index.add_records,
1799
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1800
self.assertRaises(errors.KnitCorrupt, index.add_records,
1801
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1803
self.assertRaises(errors.KnitCorrupt, index.add_records,
1804
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1805
# change options in the second record
1806
self.assertRaises(errors.KnitCorrupt, index.add_records,
1807
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)]),
1808
(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1809
self.assertEqual([], self.caught_entries)
1811
def make_g_index_missing_compression_parent(self):
1812
graph_index = self.make_g_index('missing_comp', 2,
1813
[(('tip', ), ' 100 78',
1814
([('missing-parent', ), ('ghost', )], [('missing-parent', )]))])
1817
def make_g_index_missing_parent(self):
1818
graph_index = self.make_g_index('missing_parent', 2,
1819
[(('parent', ), ' 100 78', ([], [])),
1820
(('tip', ), ' 100 78',
1821
([('parent', ), ('missing-parent', )], [('parent', )])),
1825
def make_g_index_no_external_refs(self):
1826
graph_index = self.make_g_index('no_external_refs', 2,
1827
[(('rev', ), ' 100 78',
1828
([('parent', ), ('ghost', )], []))])
1831
def test_add_good_unvalidated_index(self):
1832
unvalidated = self.make_g_index_no_external_refs()
1833
combined = CombinedGraphIndex([unvalidated])
1834
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1835
index.scan_unvalidated_index(unvalidated)
1836
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1838
def test_add_missing_compression_parent_unvalidated_index(self):
1839
unvalidated = self.make_g_index_missing_compression_parent()
1840
combined = CombinedGraphIndex([unvalidated])
1841
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1842
index.scan_unvalidated_index(unvalidated)
1843
# This also checks that its only the compression parent that is
1844
# examined, otherwise 'ghost' would also be reported as a missing
1847
frozenset([('missing-parent',)]),
1848
index.get_missing_compression_parents())
1850
def test_add_missing_noncompression_parent_unvalidated_index(self):
1851
unvalidated = self.make_g_index_missing_parent()
1852
combined = CombinedGraphIndex([unvalidated])
1853
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1854
track_external_parent_refs=True)
1855
index.scan_unvalidated_index(unvalidated)
1857
frozenset([('missing-parent',)]), index.get_missing_parents())
1859
def test_track_external_parent_refs(self):
1860
g_index = self.make_g_index('empty', 2, [])
1861
combined = CombinedGraphIndex([g_index])
1862
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1863
add_callback=self.catch_add, track_external_parent_refs=True)
1864
self.caught_entries = []
1866
(('new-key',), 'fulltext,no-eol', (None, 50, 60),
1867
[('parent-1',), ('parent-2',)])])
1869
frozenset([('parent-1',), ('parent-2',)]),
1870
index.get_missing_parents())
1872
def test_add_unvalidated_index_with_present_external_references(self):
1873
index = self.two_graph_index(deltas=True)
1874
# Ugly hack to get at one of the underlying GraphIndex objects that
1875
# two_graph_index built.
1876
unvalidated = index._graph_index._indices[1]
1877
# 'parent' is an external ref of _indices[1] (unvalidated), but is
1878
# present in _indices[0].
1879
index.scan_unvalidated_index(unvalidated)
1880
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1882
def make_new_missing_parent_g_index(self, name):
1883
missing_parent = name + '-missing-parent'
1884
graph_index = self.make_g_index(name, 2,
1885
[((name + 'tip', ), ' 100 78',
1886
([(missing_parent, ), ('ghost', )], [(missing_parent, )]))])
1889
def test_add_mulitiple_unvalidated_indices_with_missing_parents(self):
1890
g_index_1 = self.make_new_missing_parent_g_index('one')
1891
g_index_2 = self.make_new_missing_parent_g_index('two')
1892
combined = CombinedGraphIndex([g_index_1, g_index_2])
1893
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1894
index.scan_unvalidated_index(g_index_1)
1895
index.scan_unvalidated_index(g_index_2)
1897
frozenset([('one-missing-parent',), ('two-missing-parent',)]),
1898
index.get_missing_compression_parents())
1900
def test_add_mulitiple_unvalidated_indices_with_mutual_dependencies(self):
1901
graph_index_a = self.make_g_index('one', 2,
1902
[(('parent-one', ), ' 100 78', ([('non-compression-parent',)], [])),
1903
(('child-of-two', ), ' 100 78',
1904
([('parent-two',)], [('parent-two',)]))])
1905
graph_index_b = self.make_g_index('two', 2,
1906
[(('parent-two', ), ' 100 78', ([('non-compression-parent',)], [])),
1907
(('child-of-one', ), ' 100 78',
1908
([('parent-one',)], [('parent-one',)]))])
1909
combined = CombinedGraphIndex([graph_index_a, graph_index_b])
1910
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1911
index.scan_unvalidated_index(graph_index_a)
1912
index.scan_unvalidated_index(graph_index_b)
1914
frozenset([]), index.get_missing_compression_parents())
1917
class TestNoParentsGraphIndexKnit(KnitTests):
1918
"""Tests for knits using _KnitGraphIndex with no parents."""
1920
def make_g_index(self, name, ref_lists=0, nodes=[]):
1921
builder = GraphIndexBuilder(ref_lists)
1922
for node, references in nodes:
1923
builder.add_node(node, references)
1924
stream = builder.finish()
1925
trans = self.get_transport()
1926
size = trans.put_file(name, stream)
1927
return GraphIndex(trans, name, size)
1929
def test_add_good_unvalidated_index(self):
1930
unvalidated = self.make_g_index('unvalidated')
1931
combined = CombinedGraphIndex([unvalidated])
1932
index = _KnitGraphIndex(combined, lambda: True, parents=False)
1933
index.scan_unvalidated_index(unvalidated)
1934
self.assertEqual(frozenset(),
1935
index.get_missing_compression_parents())
1937
def test_parents_deltas_incompatible(self):
1938
index = CombinedGraphIndex([])
1939
self.assertRaises(errors.KnitError, _KnitGraphIndex, lambda:True,
1940
index, deltas=True, parents=False)
1942
def two_graph_index(self, catch_adds=False):
1943
"""Build a two-graph index.
1945
:param deltas: If true, use underlying indices with two node-ref
1946
lists and 'parent' set to a delta-compressed against tail.
1948
# put several versions in the index.
1949
index1 = self.make_g_index('1', 0, [
1950
(('tip', ), 'N0 100'),
1952
index2 = self.make_g_index('2', 0, [
1953
(('parent', ), ' 100 78'),
1954
(('separate', ), '')])
1955
combined_index = CombinedGraphIndex([index1, index2])
1957
self.combined_index = combined_index
1958
self.caught_entries = []
1959
add_callback = self.catch_add
1962
return _KnitGraphIndex(combined_index, lambda:True, parents=False,
1963
add_callback=add_callback)
1965
def test_keys(self):
1966
index = self.two_graph_index()
1967
self.assertEqual(set([('tail',), ('tip',), ('parent',), ('separate',)]),
1970
def test_get_position(self):
1971
index = self.two_graph_index()
1972
self.assertEqual((index._graph_index._indices[0], 0, 100),
1973
index.get_position(('tip',)))
1974
self.assertEqual((index._graph_index._indices[1], 100, 78),
1975
index.get_position(('parent',)))
1977
def test_get_method(self):
1978
index = self.two_graph_index()
1979
self.assertEqual('fulltext', index.get_method(('tip',)))
1980
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1982
def test_get_options(self):
1983
index = self.two_graph_index()
1984
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1985
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1987
def test_get_parent_map(self):
1988
index = self.two_graph_index()
1989
self.assertEqual({('parent',):None},
1990
index.get_parent_map([('parent',), ('ghost',)]))
1992
def catch_add(self, entries):
1993
self.caught_entries.append(entries)
1995
def test_add_no_callback_errors(self):
1996
index = self.two_graph_index()
1997
self.assertRaises(errors.ReadOnlyError, index.add_records,
1998
[(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)])])
2000
def test_add_version_smoke(self):
2001
index = self.two_graph_index(catch_adds=True)
2002
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60), [])])
2003
self.assertEqual([[(('new', ), 'N50 60')]],
2004
self.caught_entries)
2006
def test_add_version_delta_not_delta_index(self):
2007
index = self.two_graph_index(catch_adds=True)
2008
self.assertRaises(errors.KnitCorrupt, index.add_records,
2009
[(('new',), 'no-eol,line-delta', (None, 0, 100), [])])
2010
self.assertEqual([], self.caught_entries)
2012
def test_add_version_same_dup(self):
2013
index = self.two_graph_index(catch_adds=True)
2014
# options can be spelt two different ways
2015
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
2016
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
2017
# position/length are ignored (because each pack could have fulltext or
2018
# delta, and be at a different position.
2019
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
2020
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
2021
# but neither should have added data.
2022
self.assertEqual([[], [], [], []], self.caught_entries)
2024
def test_add_version_different_dup(self):
2025
index = self.two_graph_index(catch_adds=True)
2027
self.assertRaises(errors.KnitCorrupt, index.add_records,
2028
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2029
self.assertRaises(errors.KnitCorrupt, index.add_records,
2030
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
2031
self.assertRaises(errors.KnitCorrupt, index.add_records,
2032
[(('tip',), 'fulltext', (None, 0, 100), [])])
2034
self.assertRaises(errors.KnitCorrupt, index.add_records,
2035
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
2036
self.assertEqual([], self.caught_entries)
2038
def test_add_versions(self):
2039
index = self.two_graph_index(catch_adds=True)
2041
(('new',), 'fulltext,no-eol', (None, 50, 60), []),
2042
(('new2',), 'fulltext', (None, 0, 6), []),
2044
self.assertEqual([(('new', ), 'N50 60'), (('new2', ), ' 0 6')],
2045
sorted(self.caught_entries[0]))
2046
self.assertEqual(1, len(self.caught_entries))
2048
def test_add_versions_delta_not_delta_index(self):
2049
index = self.two_graph_index(catch_adds=True)
2050
self.assertRaises(errors.KnitCorrupt, index.add_records,
2051
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
2052
self.assertEqual([], self.caught_entries)
2054
def test_add_versions_parents_not_parents_index(self):
2055
index = self.two_graph_index(catch_adds=True)
2056
self.assertRaises(errors.KnitCorrupt, index.add_records,
2057
[(('new',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
2058
self.assertEqual([], self.caught_entries)
2060
def test_add_versions_random_id_accepted(self):
2061
index = self.two_graph_index(catch_adds=True)
2062
index.add_records([], random_id=True)
2064
def test_add_versions_same_dup(self):
2065
index = self.two_graph_index(catch_adds=True)
2066
# options can be spelt two different ways
2067
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
2068
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
2069
# position/length are ignored (because each pack could have fulltext or
2070
# delta, and be at a different position.
2071
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
2072
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
2073
# but neither should have added data.
2074
self.assertEqual([[], [], [], []], self.caught_entries)
2076
def test_add_versions_different_dup(self):
2077
index = self.two_graph_index(catch_adds=True)
2079
self.assertRaises(errors.KnitCorrupt, index.add_records,
2080
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2081
self.assertRaises(errors.KnitCorrupt, index.add_records,
2082
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
2083
self.assertRaises(errors.KnitCorrupt, index.add_records,
2084
[(('tip',), 'fulltext', (None, 0, 100), [])])
2086
self.assertRaises(errors.KnitCorrupt, index.add_records,
2087
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
2088
# change options in the second record
2089
self.assertRaises(errors.KnitCorrupt, index.add_records,
2090
[(('tip',), 'fulltext,no-eol', (None, 0, 100), []),
2091
(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2092
self.assertEqual([], self.caught_entries)
2095
class TestKnitVersionedFiles(KnitTests):
2097
def assertGroupKeysForIo(self, exp_groups, keys, non_local_keys,
2098
positions, _min_buffer_size=None):
2099
kvf = self.make_test_knit()
2100
if _min_buffer_size is None:
2101
_min_buffer_size = knit._STREAM_MIN_BUFFER_SIZE
2102
self.assertEqual(exp_groups, kvf._group_keys_for_io(keys,
2103
non_local_keys, positions,
2104
_min_buffer_size=_min_buffer_size))
2106
def assertSplitByPrefix(self, expected_map, expected_prefix_order,
2108
split, prefix_order = KnitVersionedFiles._split_by_prefix(keys)
2109
self.assertEqual(expected_map, split)
2110
self.assertEqual(expected_prefix_order, prefix_order)
2112
def test__group_keys_for_io(self):
2113
ft_detail = ('fulltext', False)
2114
ld_detail = ('line-delta', False)
2122
f_a: (ft_detail, (f_a, 0, 100), None),
2123
f_b: (ld_detail, (f_b, 100, 21), f_a),
2124
f_c: (ld_detail, (f_c, 180, 15), f_b),
2125
g_a: (ft_detail, (g_a, 121, 35), None),
2126
g_b: (ld_detail, (g_b, 156, 12), g_a),
2127
g_c: (ld_detail, (g_c, 195, 13), g_a),
2129
self.assertGroupKeysForIo([([f_a], set())],
2130
[f_a], [], positions)
2131
self.assertGroupKeysForIo([([f_a], set([f_a]))],
2132
[f_a], [f_a], positions)
2133
self.assertGroupKeysForIo([([f_a, f_b], set([]))],
2134
[f_a, f_b], [], positions)
2135
self.assertGroupKeysForIo([([f_a, f_b], set([f_b]))],
2136
[f_a, f_b], [f_b], positions)
2137
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2138
[f_a, g_a, f_b, g_b], [], positions)
2139
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2140
[f_a, g_a, f_b, g_b], [], positions,
2141
_min_buffer_size=150)
2142
self.assertGroupKeysForIo([([f_a, f_b], set()), ([g_a, g_b], set())],
2143
[f_a, g_a, f_b, g_b], [], positions,
2144
_min_buffer_size=100)
2145
self.assertGroupKeysForIo([([f_c], set()), ([g_b], set())],
2146
[f_c, g_b], [], positions,
2147
_min_buffer_size=125)
2148
self.assertGroupKeysForIo([([g_b, f_c], set())],
2149
[g_b, f_c], [], positions,
2150
_min_buffer_size=125)
2152
def test__split_by_prefix(self):
2153
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2154
'g': [('g', 'b'), ('g', 'a')],
2156
[('f', 'a'), ('g', 'b'),
2157
('g', 'a'), ('f', 'b')])
2159
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2160
'g': [('g', 'b'), ('g', 'a')],
2162
[('f', 'a'), ('f', 'b'),
2163
('g', 'b'), ('g', 'a')])
2165
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2166
'g': [('g', 'b'), ('g', 'a')],
2168
[('f', 'a'), ('f', 'b'),
2169
('g', 'b'), ('g', 'a')])
2171
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2172
'g': [('g', 'b'), ('g', 'a')],
2173
'': [('a',), ('b',)]
2175
[('f', 'a'), ('g', 'b'),
2177
('g', 'a'), ('f', 'b')])
2180
class TestStacking(KnitTests):
2182
def get_basis_and_test_knit(self):
2183
basis = self.make_test_knit(name='basis')
2184
basis = RecordingVersionedFilesDecorator(basis)
2185
test = self.make_test_knit(name='test')
2186
test.add_fallback_versioned_files(basis)
2189
def test_add_fallback_versioned_files(self):
2190
basis = self.make_test_knit(name='basis')
2191
test = self.make_test_knit(name='test')
2192
# It must not error; other tests test that the fallback is referred to
2193
# when accessing data.
2194
test.add_fallback_versioned_files(basis)
2196
def test_add_lines(self):
2197
# lines added to the test are not added to the basis
2198
basis, test = self.get_basis_and_test_knit()
2200
key_basis = ('bar',)
2201
key_cross_border = ('quux',)
2202
key_delta = ('zaphod',)
2203
test.add_lines(key, (), ['foo\n'])
2204
self.assertEqual({}, basis.get_parent_map([key]))
2205
# lines added to the test that reference across the stack do a
2207
basis.add_lines(key_basis, (), ['foo\n'])
2209
test.add_lines(key_cross_border, (key_basis,), ['foo\n'])
2210
self.assertEqual('fulltext', test._index.get_method(key_cross_border))
2211
# we don't even need to look at the basis to see that this should be
2212
# stored as a fulltext
2213
self.assertEqual([], basis.calls)
2214
# Subsequent adds do delta.
2216
test.add_lines(key_delta, (key_cross_border,), ['foo\n'])
2217
self.assertEqual('line-delta', test._index.get_method(key_delta))
2218
self.assertEqual([], basis.calls)
2220
def test_annotate(self):
2221
# annotations from the test knit are answered without asking the basis
2222
basis, test = self.get_basis_and_test_knit()
2224
key_basis = ('bar',)
2225
key_missing = ('missing',)
2226
test.add_lines(key, (), ['foo\n'])
2227
details = test.annotate(key)
2228
self.assertEqual([(key, 'foo\n')], details)
2229
self.assertEqual([], basis.calls)
2230
# But texts that are not in the test knit are looked for in the basis
2232
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2234
details = test.annotate(key_basis)
2235
self.assertEqual([(key_basis, 'foo\n'), (key_basis, 'bar\n')], details)
2236
# Not optimised to date:
2237
# self.assertEqual([("annotate", key_basis)], basis.calls)
2238
self.assertEqual([('get_parent_map', set([key_basis])),
2239
('get_parent_map', set([key_basis])),
2240
('get_record_stream', [key_basis], 'topological', True)],
2243
def test_check(self):
2244
# At the moment checking a stacked knit does implicitly check the
2246
basis, test = self.get_basis_and_test_knit()
2249
def test_get_parent_map(self):
2250
# parents in the test knit are answered without asking the basis
2251
basis, test = self.get_basis_and_test_knit()
2253
key_basis = ('bar',)
2254
key_missing = ('missing',)
2255
test.add_lines(key, (), [])
2256
parent_map = test.get_parent_map([key])
2257
self.assertEqual({key: ()}, parent_map)
2258
self.assertEqual([], basis.calls)
2259
# But parents that are not in the test knit are looked for in the basis
2260
basis.add_lines(key_basis, (), [])
2262
parent_map = test.get_parent_map([key, key_basis, key_missing])
2263
self.assertEqual({key: (),
2264
key_basis: ()}, parent_map)
2265
self.assertEqual([("get_parent_map", set([key_basis, key_missing]))],
2268
def test_get_record_stream_unordered_fulltexts(self):
2269
# records from the test knit are answered without asking the basis:
2270
basis, test = self.get_basis_and_test_knit()
2272
key_basis = ('bar',)
2273
key_missing = ('missing',)
2274
test.add_lines(key, (), ['foo\n'])
2275
records = list(test.get_record_stream([key], 'unordered', True))
2276
self.assertEqual(1, len(records))
2277
self.assertEqual([], basis.calls)
2278
# Missing (from test knit) objects are retrieved from the basis:
2279
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2281
records = list(test.get_record_stream([key_basis, key_missing],
2283
self.assertEqual(2, len(records))
2284
calls = list(basis.calls)
2285
for record in records:
2286
self.assertSubset([record.key], (key_basis, key_missing))
2287
if record.key == key_missing:
2288
self.assertIsInstance(record, AbsentContentFactory)
2290
reference = list(basis.get_record_stream([key_basis],
2291
'unordered', True))[0]
2292
self.assertEqual(reference.key, record.key)
2293
self.assertEqual(reference.sha1, record.sha1)
2294
self.assertEqual(reference.storage_kind, record.storage_kind)
2295
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2296
record.get_bytes_as(record.storage_kind))
2297
self.assertEqual(reference.get_bytes_as('fulltext'),
2298
record.get_bytes_as('fulltext'))
2299
# It's not strictly minimal, but it seems reasonable for now for it to
2300
# ask which fallbacks have which parents.
2302
("get_parent_map", set([key_basis, key_missing])),
2303
("get_record_stream", [key_basis], 'unordered', True)],
2306
def test_get_record_stream_ordered_fulltexts(self):
2307
# ordering is preserved down into the fallback store.
2308
basis, test = self.get_basis_and_test_knit()
2310
key_basis = ('bar',)
2311
key_basis_2 = ('quux',)
2312
key_missing = ('missing',)
2313
test.add_lines(key, (key_basis,), ['foo\n'])
2314
# Missing (from test knit) objects are retrieved from the basis:
2315
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2316
basis.add_lines(key_basis_2, (), ['quux\n'])
2318
# ask for in non-topological order
2319
records = list(test.get_record_stream(
2320
[key, key_basis, key_missing, key_basis_2], 'topological', True))
2321
self.assertEqual(4, len(records))
2323
for record in records:
2324
self.assertSubset([record.key],
2325
(key_basis, key_missing, key_basis_2, key))
2326
if record.key == key_missing:
2327
self.assertIsInstance(record, AbsentContentFactory)
2329
results.append((record.key, record.sha1, record.storage_kind,
2330
record.get_bytes_as('fulltext')))
2331
calls = list(basis.calls)
2332
order = [record[0] for record in results]
2333
self.assertEqual([key_basis_2, key_basis, key], order)
2334
for result in results:
2335
if result[0] == key:
2339
record = source.get_record_stream([result[0]], 'unordered',
2341
self.assertEqual(record.key, result[0])
2342
self.assertEqual(record.sha1, result[1])
2343
# We used to check that the storage kind matched, but actually it
2344
# depends on whether it was sourced from the basis, or in a single
2345
# group, because asking for full texts returns proxy objects to a
2346
# _ContentMapGenerator object; so checking the kind is unneeded.
2347
self.assertEqual(record.get_bytes_as('fulltext'), result[3])
2348
# It's not strictly minimal, but it seems reasonable for now for it to
2349
# ask which fallbacks have which parents.
2351
("get_parent_map", set([key_basis, key_basis_2, key_missing])),
2352
# topological is requested from the fallback, because that is what
2353
# was requested at the top level.
2354
("get_record_stream", [key_basis_2, key_basis], 'topological', True)],
2357
def test_get_record_stream_unordered_deltas(self):
2358
# records from the test knit are answered without asking the basis:
2359
basis, test = self.get_basis_and_test_knit()
2361
key_basis = ('bar',)
2362
key_missing = ('missing',)
2363
test.add_lines(key, (), ['foo\n'])
2364
records = list(test.get_record_stream([key], 'unordered', False))
2365
self.assertEqual(1, len(records))
2366
self.assertEqual([], basis.calls)
2367
# Missing (from test knit) objects are retrieved from the basis:
2368
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2370
records = list(test.get_record_stream([key_basis, key_missing],
2371
'unordered', False))
2372
self.assertEqual(2, len(records))
2373
calls = list(basis.calls)
2374
for record in records:
2375
self.assertSubset([record.key], (key_basis, key_missing))
2376
if record.key == key_missing:
2377
self.assertIsInstance(record, AbsentContentFactory)
2379
reference = list(basis.get_record_stream([key_basis],
2380
'unordered', False))[0]
2381
self.assertEqual(reference.key, record.key)
2382
self.assertEqual(reference.sha1, record.sha1)
2383
self.assertEqual(reference.storage_kind, record.storage_kind)
2384
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2385
record.get_bytes_as(record.storage_kind))
2386
# It's not strictly minimal, but it seems reasonable for now for it to
2387
# ask which fallbacks have which parents.
2389
("get_parent_map", set([key_basis, key_missing])),
2390
("get_record_stream", [key_basis], 'unordered', False)],
2393
def test_get_record_stream_ordered_deltas(self):
2394
# ordering is preserved down into the fallback store.
2395
basis, test = self.get_basis_and_test_knit()
2397
key_basis = ('bar',)
2398
key_basis_2 = ('quux',)
2399
key_missing = ('missing',)
2400
test.add_lines(key, (key_basis,), ['foo\n'])
2401
# Missing (from test knit) objects are retrieved from the basis:
2402
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2403
basis.add_lines(key_basis_2, (), ['quux\n'])
2405
# ask for in non-topological order
2406
records = list(test.get_record_stream(
2407
[key, key_basis, key_missing, key_basis_2], 'topological', False))
2408
self.assertEqual(4, len(records))
2410
for record in records:
2411
self.assertSubset([record.key],
2412
(key_basis, key_missing, key_basis_2, key))
2413
if record.key == key_missing:
2414
self.assertIsInstance(record, AbsentContentFactory)
2416
results.append((record.key, record.sha1, record.storage_kind,
2417
record.get_bytes_as(record.storage_kind)))
2418
calls = list(basis.calls)
2419
order = [record[0] for record in results]
2420
self.assertEqual([key_basis_2, key_basis, key], order)
2421
for result in results:
2422
if result[0] == key:
2426
record = source.get_record_stream([result[0]], 'unordered',
2428
self.assertEqual(record.key, result[0])
2429
self.assertEqual(record.sha1, result[1])
2430
self.assertEqual(record.storage_kind, result[2])
2431
self.assertEqual(record.get_bytes_as(record.storage_kind), result[3])
2432
# It's not strictly minimal, but it seems reasonable for now for it to
2433
# ask which fallbacks have which parents.
2435
("get_parent_map", set([key_basis, key_basis_2, key_missing])),
2436
("get_record_stream", [key_basis_2, key_basis], 'topological', False)],
2439
def test_get_sha1s(self):
2440
# sha1's in the test knit are answered without asking the basis
2441
basis, test = self.get_basis_and_test_knit()
2443
key_basis = ('bar',)
2444
key_missing = ('missing',)
2445
test.add_lines(key, (), ['foo\n'])
2446
key_sha1sum = osutils.sha_string('foo\n')
2447
sha1s = test.get_sha1s([key])
2448
self.assertEqual({key: key_sha1sum}, sha1s)
2449
self.assertEqual([], basis.calls)
2450
# But texts that are not in the test knit are looked for in the basis
2451
# directly (rather than via text reconstruction) so that remote servers
2452
# etc don't have to answer with full content.
2453
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2454
basis_sha1sum = osutils.sha_string('foo\nbar\n')
2456
sha1s = test.get_sha1s([key, key_missing, key_basis])
2457
self.assertEqual({key: key_sha1sum,
2458
key_basis: basis_sha1sum}, sha1s)
2459
self.assertEqual([("get_sha1s", set([key_basis, key_missing]))],
2462
def test_insert_record_stream(self):
2463
# records are inserted as normal; insert_record_stream builds on
2464
# add_lines, so a smoke test should be all that's needed:
2466
key_basis = ('bar',)
2467
key_delta = ('zaphod',)
2468
basis, test = self.get_basis_and_test_knit()
2469
source = self.make_test_knit(name='source')
2470
basis.add_lines(key_basis, (), ['foo\n'])
2472
source.add_lines(key_basis, (), ['foo\n'])
2473
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2474
stream = source.get_record_stream([key_delta], 'unordered', False)
2475
test.insert_record_stream(stream)
2476
# XXX: this does somewhat too many calls in making sure of whether it
2477
# has to recreate the full text.
2478
self.assertEqual([("get_parent_map", set([key_basis])),
2479
('get_parent_map', set([key_basis])),
2480
('get_record_stream', [key_basis], 'unordered', True)],
2482
self.assertEqual({key_delta:(key_basis,)},
2483
test.get_parent_map([key_delta]))
2484
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2485
'unordered', True).next().get_bytes_as('fulltext'))
2487
def test_iter_lines_added_or_present_in_keys(self):
2488
# Lines from the basis are returned, and lines for a given key are only
2492
# all sources are asked for keys:
2493
basis, test = self.get_basis_and_test_knit()
2494
basis.add_lines(key1, (), ["foo"])
2496
lines = list(test.iter_lines_added_or_present_in_keys([key1]))
2497
self.assertEqual([("foo\n", key1)], lines)
2498
self.assertEqual([("iter_lines_added_or_present_in_keys", set([key1]))],
2500
# keys in both are not duplicated:
2501
test.add_lines(key2, (), ["bar\n"])
2502
basis.add_lines(key2, (), ["bar\n"])
2504
lines = list(test.iter_lines_added_or_present_in_keys([key2]))
2505
self.assertEqual([("bar\n", key2)], lines)
2506
self.assertEqual([], basis.calls)
2508
def test_keys(self):
2511
# all sources are asked for keys:
2512
basis, test = self.get_basis_and_test_knit()
2514
self.assertEqual(set(), set(keys))
2515
self.assertEqual([("keys",)], basis.calls)
2516
# keys from a basis are returned:
2517
basis.add_lines(key1, (), [])
2520
self.assertEqual(set([key1]), set(keys))
2521
self.assertEqual([("keys",)], basis.calls)
2522
# keys in both are not duplicated:
2523
test.add_lines(key2, (), [])
2524
basis.add_lines(key2, (), [])
2527
self.assertEqual(2, len(keys))
2528
self.assertEqual(set([key1, key2]), set(keys))
2529
self.assertEqual([("keys",)], basis.calls)
2531
def test_add_mpdiffs(self):
2532
# records are inserted as normal; add_mpdiff builds on
2533
# add_lines, so a smoke test should be all that's needed:
2535
key_basis = ('bar',)
2536
key_delta = ('zaphod',)
2537
basis, test = self.get_basis_and_test_knit()
2538
source = self.make_test_knit(name='source')
2539
basis.add_lines(key_basis, (), ['foo\n'])
2541
source.add_lines(key_basis, (), ['foo\n'])
2542
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2543
diffs = source.make_mpdiffs([key_delta])
2544
test.add_mpdiffs([(key_delta, (key_basis,),
2545
source.get_sha1s([key_delta])[key_delta], diffs[0])])
2546
self.assertEqual([("get_parent_map", set([key_basis])),
2547
('get_record_stream', [key_basis], 'unordered', True),],
2549
self.assertEqual({key_delta:(key_basis,)},
2550
test.get_parent_map([key_delta]))
2551
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2552
'unordered', True).next().get_bytes_as('fulltext'))
2554
def test_make_mpdiffs(self):
2555
# Generating an mpdiff across a stacking boundary should detect parent
2559
key_right = ('zaphod',)
2560
basis, test = self.get_basis_and_test_knit()
2561
basis.add_lines(key_left, (), ['bar\n'])
2562
basis.add_lines(key_right, (), ['zaphod\n'])
2564
test.add_lines(key, (key_left, key_right),
2565
['bar\n', 'foo\n', 'zaphod\n'])
2566
diffs = test.make_mpdiffs([key])
2568
multiparent.MultiParent([multiparent.ParentText(0, 0, 0, 1),
2569
multiparent.NewText(['foo\n']),
2570
multiparent.ParentText(1, 0, 2, 1)])],
2572
self.assertEqual(3, len(basis.calls))
2574
("get_parent_map", set([key_left, key_right])),
2575
("get_parent_map", set([key_left, key_right])),
2578
last_call = basis.calls[-1]
2579
self.assertEqual('get_record_stream', last_call[0])
2580
self.assertEqual(set([key_left, key_right]), set(last_call[1]))
2581
self.assertEqual('topological', last_call[2])
2582
self.assertEqual(True, last_call[3])
2585
class TestNetworkBehaviour(KnitTests):
2586
"""Tests for getting data out of/into knits over the network."""
2588
def test_include_delta_closure_generates_a_knit_delta_closure(self):
2589
vf = self.make_test_knit(name='test')
2590
# put in three texts, giving ft, delta, delta
2591
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2592
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2593
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2594
# But heuristics could interfere, so check what happened:
2595
self.assertEqual(['knit-ft-gz', 'knit-delta-gz', 'knit-delta-gz'],
2596
[record.storage_kind for record in
2597
vf.get_record_stream([('base',), ('d1',), ('d2',)],
2598
'topological', False)])
2599
# generate a stream of just the deltas include_delta_closure=True,
2600
# serialise to the network, and check that we get a delta closure on the wire.
2601
stream = vf.get_record_stream([('d1',), ('d2',)], 'topological', True)
2602
netb = [record.get_bytes_as(record.storage_kind) for record in stream]
2603
# The first bytes should be a memo from _ContentMapGenerator, and the
2604
# second bytes should be empty (because its a API proxy not something
2605
# for wire serialisation.
2606
self.assertEqual('', netb[1])
2608
kind, line_end = network_bytes_to_kind_and_offset(bytes)
2609
self.assertEqual('knit-delta-closure', kind)
2612
class TestContentMapGenerator(KnitTests):
2613
"""Tests for ContentMapGenerator"""
2615
def test_get_record_stream_gives_records(self):
2616
vf = self.make_test_knit(name='test')
2617
# put in three texts, giving ft, delta, delta
2618
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2619
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2620
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2621
keys = [('d1',), ('d2',)]
2622
generator = _VFContentMapGenerator(vf, keys,
2623
global_map=vf.get_parent_map(keys))
2624
for record in generator.get_record_stream():
2625
if record.key == ('d1',):
2626
self.assertEqual('d1\n', record.get_bytes_as('fulltext'))
2628
self.assertEqual('d2\n', record.get_bytes_as('fulltext'))
2630
def test_get_record_stream_kinds_are_raw(self):
2631
vf = self.make_test_knit(name='test')
2632
# put in three texts, giving ft, delta, delta
2633
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2634
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2635
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2636
keys = [('base',), ('d1',), ('d2',)]
2637
generator = _VFContentMapGenerator(vf, keys,
2638
global_map=vf.get_parent_map(keys))
2639
kinds = {('base',): 'knit-delta-closure',
2640
('d1',): 'knit-delta-closure-ref',
2641
('d2',): 'knit-delta-closure-ref',
2643
for record in generator.get_record_stream():
2644
self.assertEqual(kinds[record.key], record.storage_kind)
added
removed
bzrlib/tests/test_knit.py
