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,
67
compiled_knit_feature = tests.ModuleAvailableFeature(
68
'bzrlib._knit_load_data_pyx')
71
class KnitContentTestsMixin(object):
73
def test_constructor(self):
74
content = self._make_content([])
77
content = self._make_content([])
78
self.assertEqual(content.text(), [])
80
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
81
self.assertEqual(content.text(), ["text1", "text2"])
84
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
86
self.assertIsInstance(copy, content.__class__)
87
self.assertEqual(copy.annotate(), content.annotate())
89
def assertDerivedBlocksEqual(self, source, target, noeol=False):
90
"""Assert that the derived matching blocks match real output"""
91
source_lines = source.splitlines(True)
92
target_lines = target.splitlines(True)
94
if noeol and not line.endswith('\n'):
98
source_content = self._make_content([(None, nl(l)) for l in source_lines])
99
target_content = self._make_content([(None, nl(l)) for l in target_lines])
100
line_delta = source_content.line_delta(target_content)
101
delta_blocks = list(KnitContent.get_line_delta_blocks(line_delta,
102
source_lines, target_lines))
103
matcher = PatienceSequenceMatcher(None, source_lines, target_lines)
104
matcher_blocks = list(matcher.get_matching_blocks())
105
self.assertEqual(matcher_blocks, delta_blocks)
107
def test_get_line_delta_blocks(self):
108
self.assertDerivedBlocksEqual('a\nb\nc\n', 'q\nc\n')
109
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1)
110
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1A)
111
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1B)
112
self.assertDerivedBlocksEqual(TEXT_1B, TEXT_1A)
113
self.assertDerivedBlocksEqual(TEXT_1A, TEXT_1B)
114
self.assertDerivedBlocksEqual(TEXT_1A, '')
115
self.assertDerivedBlocksEqual('', TEXT_1A)
116
self.assertDerivedBlocksEqual('', '')
117
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd')
119
def test_get_line_delta_blocks_noeol(self):
120
"""Handle historical knit deltas safely
122
Some existing knit deltas don't consider the last line to differ
123
when the only difference whether it has a final newline.
125
New knit deltas appear to always consider the last line to differ
128
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd\n', noeol=True)
129
self.assertDerivedBlocksEqual('a\nb\nc\nd\n', 'a\nb\nc', noeol=True)
130
self.assertDerivedBlocksEqual('a\nb\nc\n', 'a\nb\nc', noeol=True)
131
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\n', noeol=True)
143
Banana cup cake recipe
153
Banana cup cake recipe
155
- bananas (do not use plantains!!!)
162
Banana cup cake recipe
179
class TestPlainKnitContent(TestCase, KnitContentTestsMixin):
181
def _make_content(self, lines):
182
annotated_content = AnnotatedKnitContent(lines)
183
return PlainKnitContent(annotated_content.text(), 'bogus')
185
def test_annotate(self):
186
content = self._make_content([])
187
self.assertEqual(content.annotate(), [])
189
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
190
self.assertEqual(content.annotate(),
191
[("bogus", "text1"), ("bogus", "text2")])
193
def test_line_delta(self):
194
content1 = self._make_content([("", "a"), ("", "b")])
195
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
196
self.assertEqual(content1.line_delta(content2),
197
[(1, 2, 2, ["a", "c"])])
199
def test_line_delta_iter(self):
200
content1 = self._make_content([("", "a"), ("", "b")])
201
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
202
it = content1.line_delta_iter(content2)
203
self.assertEqual(it.next(), (1, 2, 2, ["a", "c"]))
204
self.assertRaises(StopIteration, it.next)
207
class TestAnnotatedKnitContent(TestCase, KnitContentTestsMixin):
209
def _make_content(self, lines):
210
return AnnotatedKnitContent(lines)
212
def test_annotate(self):
213
content = self._make_content([])
214
self.assertEqual(content.annotate(), [])
216
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
217
self.assertEqual(content.annotate(),
218
[("origin1", "text1"), ("origin2", "text2")])
220
def test_line_delta(self):
221
content1 = self._make_content([("", "a"), ("", "b")])
222
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
223
self.assertEqual(content1.line_delta(content2),
224
[(1, 2, 2, [("", "a"), ("", "c")])])
226
def test_line_delta_iter(self):
227
content1 = self._make_content([("", "a"), ("", "b")])
228
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
229
it = content1.line_delta_iter(content2)
230
self.assertEqual(it.next(), (1, 2, 2, [("", "a"), ("", "c")]))
231
self.assertRaises(StopIteration, it.next)
234
class MockTransport(object):
236
def __init__(self, file_lines=None):
237
self.file_lines = file_lines
239
# We have no base directory for the MockTransport
242
def get(self, filename):
243
if self.file_lines is None:
244
raise NoSuchFile(filename)
246
return StringIO("\n".join(self.file_lines))
248
def readv(self, relpath, offsets):
249
fp = self.get(relpath)
250
for offset, size in offsets:
252
yield offset, fp.read(size)
254
def __getattr__(self, name):
255
def queue_call(*args, **kwargs):
256
self.calls.append((name, args, kwargs))
260
class MockReadvFailingTransport(MockTransport):
261
"""Fail in the middle of a readv() result.
263
This Transport will successfully yield the first two requested hunks, but
264
raise NoSuchFile for the rest.
267
def readv(self, relpath, offsets):
269
for result in MockTransport.readv(self, relpath, offsets):
271
# we use 2 because the first offset is the pack header, the second
272
# is the first actual content requset
274
raise errors.NoSuchFile(relpath)
278
class KnitRecordAccessTestsMixin(object):
279
"""Tests for getting and putting knit records."""
281
def test_add_raw_records(self):
282
"""Add_raw_records adds records retrievable later."""
283
access = self.get_access()
284
memos = access.add_raw_records([('key', 10)], '1234567890')
285
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
287
def test_add_several_raw_records(self):
288
"""add_raw_records with many records and read some back."""
289
access = self.get_access()
290
memos = access.add_raw_records([('key', 10), ('key2', 2), ('key3', 5)],
292
self.assertEqual(['1234567890', '12', '34567'],
293
list(access.get_raw_records(memos)))
294
self.assertEqual(['1234567890'],
295
list(access.get_raw_records(memos[0:1])))
296
self.assertEqual(['12'],
297
list(access.get_raw_records(memos[1:2])))
298
self.assertEqual(['34567'],
299
list(access.get_raw_records(memos[2:3])))
300
self.assertEqual(['1234567890', '34567'],
301
list(access.get_raw_records(memos[0:1] + memos[2:3])))
304
class TestKnitKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
305
"""Tests for the .kndx implementation."""
307
def get_access(self):
308
"""Get a .knit style access instance."""
309
mapper = ConstantMapper("foo")
310
access = _KnitKeyAccess(self.get_transport(), mapper)
314
class _TestException(Exception):
315
"""Just an exception for local tests to use."""
318
class TestPackKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
319
"""Tests for the pack based access."""
321
def get_access(self):
322
return self._get_access()[0]
324
def _get_access(self, packname='packfile', index='FOO'):
325
transport = self.get_transport()
326
def write_data(bytes):
327
transport.append_bytes(packname, bytes)
328
writer = pack.ContainerWriter(write_data)
330
access = pack_repo._DirectPackAccess({})
331
access.set_writer(writer, index, (transport, packname))
332
return access, writer
334
def make_pack_file(self):
335
"""Create a pack file with 2 records."""
336
access, writer = self._get_access(packname='packname', index='foo')
338
memos.extend(access.add_raw_records([('key1', 10)], '1234567890'))
339
memos.extend(access.add_raw_records([('key2', 5)], '12345'))
343
def test_pack_collection_pack_retries(self):
344
"""An explicit pack of a pack collection succeeds even when a
345
concurrent pack happens.
347
builder = self.make_branch_builder('.')
348
builder.start_series()
349
builder.build_snapshot('rev-1', None, [
350
('add', ('', 'root-id', 'directory', None)),
351
('add', ('file', 'file-id', 'file', 'content\nrev 1\n')),
353
builder.build_snapshot('rev-2', ['rev-1'], [
354
('modify', ('file-id', 'content\nrev 2\n')),
356
builder.build_snapshot('rev-3', ['rev-2'], [
357
('modify', ('file-id', 'content\nrev 3\n')),
359
self.addCleanup(builder.finish_series)
360
b = builder.get_branch()
361
self.addCleanup(b.lock_write().unlock)
363
collection = repo._pack_collection
364
# Concurrently repack the repo.
365
reopened_repo = repo.bzrdir.open_repository()
370
def make_vf_for_retrying(self):
371
"""Create 3 packs and a reload function.
373
Originally, 2 pack files will have the data, but one will be missing.
374
And then the third will be used in place of the first two if reload()
377
:return: (versioned_file, reload_counter)
378
versioned_file a KnitVersionedFiles using the packs for access
380
builder = self.make_branch_builder('.', format="1.9")
381
builder.start_series()
382
builder.build_snapshot('rev-1', None, [
383
('add', ('', 'root-id', 'directory', None)),
384
('add', ('file', 'file-id', 'file', 'content\nrev 1\n')),
386
builder.build_snapshot('rev-2', ['rev-1'], [
387
('modify', ('file-id', 'content\nrev 2\n')),
389
builder.build_snapshot('rev-3', ['rev-2'], [
390
('modify', ('file-id', 'content\nrev 3\n')),
392
builder.finish_series()
393
b = builder.get_branch()
395
self.addCleanup(b.unlock)
396
# Pack these three revisions into another pack file, but don't remove
399
collection = repo._pack_collection
400
collection.ensure_loaded()
401
orig_packs = collection.packs
402
packer = knitpack_repo.KnitPacker(collection, orig_packs, '.testpack')
403
new_pack = packer.pack()
404
# forget about the new pack
408
# Set up a reload() function that switches to using the new pack file
409
new_index = new_pack.revision_index
410
access_tuple = new_pack.access_tuple()
411
reload_counter = [0, 0, 0]
413
reload_counter[0] += 1
414
if reload_counter[1] > 0:
415
# We already reloaded, nothing more to do
416
reload_counter[2] += 1
418
reload_counter[1] += 1
419
vf._index._graph_index._indices[:] = [new_index]
420
vf._access._indices.clear()
421
vf._access._indices[new_index] = access_tuple
423
# Delete one of the pack files so the data will need to be reloaded. We
424
# will delete the file with 'rev-2' in it
425
trans, name = orig_packs[1].access_tuple()
427
# We don't have the index trigger reloading because we want to test
428
# that we reload when the .pack disappears
429
vf._access._reload_func = reload
430
return vf, reload_counter
432
def make_reload_func(self, return_val=True):
435
reload_called[0] += 1
437
return reload_called, reload
439
def make_retry_exception(self):
440
# We raise a real exception so that sys.exc_info() is properly
443
raise _TestException('foobar')
444
except _TestException, e:
445
retry_exc = errors.RetryWithNewPacks(None, reload_occurred=False,
446
exc_info=sys.exc_info())
447
# GZ 2010-08-10: Cycle with exc_info affects 3 tests
450
def test_read_from_several_packs(self):
451
access, writer = self._get_access()
453
memos.extend(access.add_raw_records([('key', 10)], '1234567890'))
455
access, writer = self._get_access('pack2', 'FOOBAR')
456
memos.extend(access.add_raw_records([('key', 5)], '12345'))
458
access, writer = self._get_access('pack3', 'BAZ')
459
memos.extend(access.add_raw_records([('key', 5)], 'alpha'))
461
transport = self.get_transport()
462
access = pack_repo._DirectPackAccess({"FOO":(transport, 'packfile'),
463
"FOOBAR":(transport, 'pack2'),
464
"BAZ":(transport, 'pack3')})
465
self.assertEqual(['1234567890', '12345', 'alpha'],
466
list(access.get_raw_records(memos)))
467
self.assertEqual(['1234567890'],
468
list(access.get_raw_records(memos[0:1])))
469
self.assertEqual(['12345'],
470
list(access.get_raw_records(memos[1:2])))
471
self.assertEqual(['alpha'],
472
list(access.get_raw_records(memos[2:3])))
473
self.assertEqual(['1234567890', 'alpha'],
474
list(access.get_raw_records(memos[0:1] + memos[2:3])))
476
def test_set_writer(self):
477
"""The writer should be settable post construction."""
478
access = pack_repo._DirectPackAccess({})
479
transport = self.get_transport()
480
packname = 'packfile'
482
def write_data(bytes):
483
transport.append_bytes(packname, bytes)
484
writer = pack.ContainerWriter(write_data)
486
access.set_writer(writer, index, (transport, packname))
487
memos = access.add_raw_records([('key', 10)], '1234567890')
489
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
491
def test_missing_index_raises_retry(self):
492
memos = self.make_pack_file()
493
transport = self.get_transport()
494
reload_called, reload_func = self.make_reload_func()
495
# Note that the index key has changed from 'foo' to 'bar'
496
access = pack_repo._DirectPackAccess({'bar':(transport, 'packname')},
497
reload_func=reload_func)
498
e = self.assertListRaises(errors.RetryWithNewPacks,
499
access.get_raw_records, memos)
500
# Because a key was passed in which does not match our index list, we
501
# assume that the listing was already reloaded
502
self.assertTrue(e.reload_occurred)
503
self.assertIsInstance(e.exc_info, tuple)
504
self.assertIs(e.exc_info[0], KeyError)
505
self.assertIsInstance(e.exc_info[1], KeyError)
507
def test_missing_index_raises_key_error_with_no_reload(self):
508
memos = self.make_pack_file()
509
transport = self.get_transport()
510
# Note that the index key has changed from 'foo' to 'bar'
511
access = pack_repo._DirectPackAccess({'bar':(transport, 'packname')})
512
e = self.assertListRaises(KeyError, access.get_raw_records, memos)
514
def test_missing_file_raises_retry(self):
515
memos = self.make_pack_file()
516
transport = self.get_transport()
517
reload_called, reload_func = self.make_reload_func()
518
# Note that the 'filename' has been changed to 'different-packname'
519
access = pack_repo._DirectPackAccess(
520
{'foo':(transport, 'different-packname')},
521
reload_func=reload_func)
522
e = self.assertListRaises(errors.RetryWithNewPacks,
523
access.get_raw_records, memos)
524
# The file has gone missing, so we assume we need to reload
525
self.assertFalse(e.reload_occurred)
526
self.assertIsInstance(e.exc_info, tuple)
527
self.assertIs(e.exc_info[0], errors.NoSuchFile)
528
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
529
self.assertEqual('different-packname', e.exc_info[1].path)
531
def test_missing_file_raises_no_such_file_with_no_reload(self):
532
memos = self.make_pack_file()
533
transport = self.get_transport()
534
# Note that the 'filename' has been changed to 'different-packname'
535
access = pack_repo._DirectPackAccess(
536
{'foo': (transport, 'different-packname')})
537
e = self.assertListRaises(errors.NoSuchFile,
538
access.get_raw_records, memos)
540
def test_failing_readv_raises_retry(self):
541
memos = self.make_pack_file()
542
transport = self.get_transport()
543
failing_transport = MockReadvFailingTransport(
544
[transport.get_bytes('packname')])
545
reload_called, reload_func = self.make_reload_func()
546
access = pack_repo._DirectPackAccess(
547
{'foo': (failing_transport, 'packname')},
548
reload_func=reload_func)
549
# Asking for a single record will not trigger the Mock failure
550
self.assertEqual(['1234567890'],
551
list(access.get_raw_records(memos[:1])))
552
self.assertEqual(['12345'],
553
list(access.get_raw_records(memos[1:2])))
554
# A multiple offset readv() will fail mid-way through
555
e = self.assertListRaises(errors.RetryWithNewPacks,
556
access.get_raw_records, memos)
557
# The file has gone missing, so we assume we need to reload
558
self.assertFalse(e.reload_occurred)
559
self.assertIsInstance(e.exc_info, tuple)
560
self.assertIs(e.exc_info[0], errors.NoSuchFile)
561
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
562
self.assertEqual('packname', e.exc_info[1].path)
564
def test_failing_readv_raises_no_such_file_with_no_reload(self):
565
memos = self.make_pack_file()
566
transport = self.get_transport()
567
failing_transport = MockReadvFailingTransport(
568
[transport.get_bytes('packname')])
569
reload_called, reload_func = self.make_reload_func()
570
access = pack_repo._DirectPackAccess(
571
{'foo':(failing_transport, 'packname')})
572
# Asking for a single record will not trigger the Mock failure
573
self.assertEqual(['1234567890'],
574
list(access.get_raw_records(memos[:1])))
575
self.assertEqual(['12345'],
576
list(access.get_raw_records(memos[1:2])))
577
# A multiple offset readv() will fail mid-way through
578
e = self.assertListRaises(errors.NoSuchFile,
579
access.get_raw_records, memos)
581
def test_reload_or_raise_no_reload(self):
582
access = pack_repo._DirectPackAccess({}, reload_func=None)
583
retry_exc = self.make_retry_exception()
584
# Without a reload_func, we will just re-raise the original exception
585
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
587
def test_reload_or_raise_reload_changed(self):
588
reload_called, reload_func = self.make_reload_func(return_val=True)
589
access = pack_repo._DirectPackAccess({}, reload_func=reload_func)
590
retry_exc = self.make_retry_exception()
591
access.reload_or_raise(retry_exc)
592
self.assertEqual([1], reload_called)
593
retry_exc.reload_occurred=True
594
access.reload_or_raise(retry_exc)
595
self.assertEqual([2], reload_called)
597
def test_reload_or_raise_reload_no_change(self):
598
reload_called, reload_func = self.make_reload_func(return_val=False)
599
access = pack_repo._DirectPackAccess({}, reload_func=reload_func)
600
retry_exc = self.make_retry_exception()
601
# If reload_occurred is False, then we consider it an error to have
602
# reload_func() return False (no changes).
603
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
604
self.assertEqual([1], reload_called)
605
retry_exc.reload_occurred=True
606
# If reload_occurred is True, then we assume nothing changed because
607
# it had changed earlier, but didn't change again
608
access.reload_or_raise(retry_exc)
609
self.assertEqual([2], reload_called)
611
def test_annotate_retries(self):
612
vf, reload_counter = self.make_vf_for_retrying()
613
# It is a little bit bogus to annotate the Revision VF, but it works,
614
# as we have ancestry stored there
616
reload_lines = vf.annotate(key)
617
self.assertEqual([1, 1, 0], reload_counter)
618
plain_lines = vf.annotate(key)
619
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
620
if reload_lines != plain_lines:
621
self.fail('Annotation was not identical with reloading.')
622
# Now delete the packs-in-use, which should trigger another reload, but
623
# this time we just raise an exception because we can't recover
624
for trans, name in vf._access._indices.itervalues():
626
self.assertRaises(errors.NoSuchFile, vf.annotate, key)
627
self.assertEqual([2, 1, 1], reload_counter)
629
def test__get_record_map_retries(self):
630
vf, reload_counter = self.make_vf_for_retrying()
631
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
632
records = vf._get_record_map(keys)
633
self.assertEqual(keys, sorted(records.keys()))
634
self.assertEqual([1, 1, 0], reload_counter)
635
# Now delete the packs-in-use, which should trigger another reload, but
636
# this time we just raise an exception because we can't recover
637
for trans, name in vf._access._indices.itervalues():
639
self.assertRaises(errors.NoSuchFile, vf._get_record_map, keys)
640
self.assertEqual([2, 1, 1], reload_counter)
642
def test_get_record_stream_retries(self):
643
vf, reload_counter = self.make_vf_for_retrying()
644
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
645
record_stream = vf.get_record_stream(keys, 'topological', False)
646
record = record_stream.next()
647
self.assertEqual(('rev-1',), record.key)
648
self.assertEqual([0, 0, 0], reload_counter)
649
record = record_stream.next()
650
self.assertEqual(('rev-2',), record.key)
651
self.assertEqual([1, 1, 0], reload_counter)
652
record = record_stream.next()
653
self.assertEqual(('rev-3',), record.key)
654
self.assertEqual([1, 1, 0], reload_counter)
655
# Now delete all pack files, and see that we raise the right error
656
for trans, name in vf._access._indices.itervalues():
658
self.assertListRaises(errors.NoSuchFile,
659
vf.get_record_stream, keys, 'topological', False)
661
def test_iter_lines_added_or_present_in_keys_retries(self):
662
vf, reload_counter = self.make_vf_for_retrying()
663
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
664
# Unfortunately, iter_lines_added_or_present_in_keys iterates the
665
# result in random order (determined by the iteration order from a
666
# set()), so we don't have any solid way to trigger whether data is
667
# read before or after. However we tried to delete the middle node to
668
# exercise the code well.
669
# What we care about is that all lines are always yielded, but not
672
reload_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
673
self.assertEqual([1, 1, 0], reload_counter)
674
# Now do it again, to make sure the result is equivalent
675
plain_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
676
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
677
self.assertEqual(plain_lines, reload_lines)
678
self.assertEqual(21, len(plain_lines))
679
# Now delete all pack files, and see that we raise the right error
680
for trans, name in vf._access._indices.itervalues():
682
self.assertListRaises(errors.NoSuchFile,
683
vf.iter_lines_added_or_present_in_keys, keys)
684
self.assertEqual([2, 1, 1], reload_counter)
686
def test_get_record_stream_yields_disk_sorted_order(self):
687
# if we get 'unordered' pick a semi-optimal order for reading. The
688
# order should be grouped by pack file, and then by position in file
689
repo = self.make_repository('test', format='pack-0.92')
691
self.addCleanup(repo.unlock)
692
repo.start_write_group()
694
vf.add_lines(('f-id', 'rev-5'), [('f-id', 'rev-4')], ['lines\n'])
695
vf.add_lines(('f-id', 'rev-1'), [], ['lines\n'])
696
vf.add_lines(('f-id', 'rev-2'), [('f-id', 'rev-1')], ['lines\n'])
697
repo.commit_write_group()
698
# We inserted them as rev-5, rev-1, rev-2, we should get them back in
700
stream = vf.get_record_stream([('f-id', 'rev-1'), ('f-id', 'rev-5'),
701
('f-id', 'rev-2')], 'unordered', False)
702
keys = [r.key for r in stream]
703
self.assertEqual([('f-id', 'rev-5'), ('f-id', 'rev-1'),
704
('f-id', 'rev-2')], keys)
705
repo.start_write_group()
706
vf.add_lines(('f-id', 'rev-4'), [('f-id', 'rev-3')], ['lines\n'])
707
vf.add_lines(('f-id', 'rev-3'), [('f-id', 'rev-2')], ['lines\n'])
708
vf.add_lines(('f-id', 'rev-6'), [('f-id', 'rev-5')], ['lines\n'])
709
repo.commit_write_group()
710
# Request in random order, to make sure the output order isn't based on
712
request_keys = set(('f-id', 'rev-%d' % i) for i in range(1, 7))
713
stream = vf.get_record_stream(request_keys, 'unordered', False)
714
keys = [r.key for r in stream]
715
# We want to get the keys back in disk order, but it doesn't matter
716
# which pack we read from first. So this can come back in 2 orders
717
alt1 = [('f-id', 'rev-%d' % i) for i in [4, 3, 6, 5, 1, 2]]
718
alt2 = [('f-id', 'rev-%d' % i) for i in [5, 1, 2, 4, 3, 6]]
719
if keys != alt1 and keys != alt2:
720
self.fail('Returned key order did not match either expected order.'
721
' expected %s or %s, not %s'
722
% (alt1, alt2, keys))
725
class LowLevelKnitDataTests(TestCase):
727
def create_gz_content(self, text):
729
gz_file = gzip.GzipFile(mode='wb', fileobj=sio)
732
return sio.getvalue()
734
def make_multiple_records(self):
735
"""Create the content for multiple records."""
736
sha1sum = osutils.sha_string('foo\nbar\n')
738
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
743
record_1 = (0, len(gz_txt), sha1sum)
744
total_txt.append(gz_txt)
745
sha1sum = osutils.sha_string('baz\n')
746
gz_txt = self.create_gz_content('version rev-id-2 1 %s\n'
750
record_2 = (record_1[1], len(gz_txt), sha1sum)
751
total_txt.append(gz_txt)
752
return total_txt, record_1, record_2
754
def test_valid_knit_data(self):
755
sha1sum = osutils.sha_string('foo\nbar\n')
756
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
761
transport = MockTransport([gz_txt])
762
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
763
knit = KnitVersionedFiles(None, access)
764
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
766
contents = list(knit._read_records_iter(records))
767
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'],
768
'4e48e2c9a3d2ca8a708cb0cc545700544efb5021')], contents)
770
raw_contents = list(knit._read_records_iter_raw(records))
771
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
773
def test_multiple_records_valid(self):
774
total_txt, record_1, record_2 = self.make_multiple_records()
775
transport = MockTransport([''.join(total_txt)])
776
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
777
knit = KnitVersionedFiles(None, access)
778
records = [(('rev-id-1',), (('rev-id-1',), record_1[0], record_1[1])),
779
(('rev-id-2',), (('rev-id-2',), record_2[0], record_2[1]))]
781
contents = list(knit._read_records_iter(records))
782
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'], record_1[2]),
783
(('rev-id-2',), ['baz\n'], record_2[2])],
786
raw_contents = list(knit._read_records_iter_raw(records))
787
self.assertEqual([(('rev-id-1',), total_txt[0], record_1[2]),
788
(('rev-id-2',), total_txt[1], record_2[2])],
791
def test_not_enough_lines(self):
792
sha1sum = osutils.sha_string('foo\n')
793
# record says 2 lines data says 1
794
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
798
transport = MockTransport([gz_txt])
799
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
800
knit = KnitVersionedFiles(None, access)
801
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
802
self.assertRaises(errors.KnitCorrupt, list,
803
knit._read_records_iter(records))
805
# read_records_iter_raw won't detect that sort of mismatch/corruption
806
raw_contents = list(knit._read_records_iter_raw(records))
807
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
809
def test_too_many_lines(self):
810
sha1sum = osutils.sha_string('foo\nbar\n')
811
# record says 1 lines data says 2
812
gz_txt = self.create_gz_content('version rev-id-1 1 %s\n'
817
transport = MockTransport([gz_txt])
818
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
819
knit = KnitVersionedFiles(None, access)
820
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
821
self.assertRaises(errors.KnitCorrupt, list,
822
knit._read_records_iter(records))
824
# read_records_iter_raw won't detect that sort of mismatch/corruption
825
raw_contents = list(knit._read_records_iter_raw(records))
826
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
828
def test_mismatched_version_id(self):
829
sha1sum = osutils.sha_string('foo\nbar\n')
830
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
835
transport = MockTransport([gz_txt])
836
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
837
knit = KnitVersionedFiles(None, access)
838
# We are asking for rev-id-2, but the data is rev-id-1
839
records = [(('rev-id-2',), (('rev-id-2',), 0, len(gz_txt)))]
840
self.assertRaises(errors.KnitCorrupt, list,
841
knit._read_records_iter(records))
843
# read_records_iter_raw detects mismatches in the header
844
self.assertRaises(errors.KnitCorrupt, list,
845
knit._read_records_iter_raw(records))
847
def test_uncompressed_data(self):
848
sha1sum = osutils.sha_string('foo\nbar\n')
849
txt = ('version rev-id-1 2 %s\n'
854
transport = MockTransport([txt])
855
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
856
knit = KnitVersionedFiles(None, access)
857
records = [(('rev-id-1',), (('rev-id-1',), 0, len(txt)))]
859
# We don't have valid gzip data ==> corrupt
860
self.assertRaises(errors.KnitCorrupt, list,
861
knit._read_records_iter(records))
863
# read_records_iter_raw will notice the bad data
864
self.assertRaises(errors.KnitCorrupt, list,
865
knit._read_records_iter_raw(records))
867
def test_corrupted_data(self):
868
sha1sum = osutils.sha_string('foo\nbar\n')
869
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
874
# Change 2 bytes in the middle to \xff
875
gz_txt = gz_txt[:10] + '\xff\xff' + gz_txt[12:]
876
transport = MockTransport([gz_txt])
877
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
878
knit = KnitVersionedFiles(None, access)
879
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
880
self.assertRaises(errors.KnitCorrupt, list,
881
knit._read_records_iter(records))
882
# read_records_iter_raw will barf on bad gz data
883
self.assertRaises(errors.KnitCorrupt, list,
884
knit._read_records_iter_raw(records))
887
class LowLevelKnitIndexTests(TestCase):
889
def get_knit_index(self, transport, name, mode):
890
mapper = ConstantMapper(name)
891
from bzrlib._knit_load_data_py import _load_data_py
892
self.overrideAttr(knit, '_load_data', _load_data_py)
893
allow_writes = lambda: 'w' in mode
894
return _KndxIndex(transport, mapper, lambda:None, allow_writes, lambda:True)
896
def test_create_file(self):
897
transport = MockTransport()
898
index = self.get_knit_index(transport, "filename", "w")
900
call = transport.calls.pop(0)
901
# call[1][1] is a StringIO - we can't test it by simple equality.
902
self.assertEqual('put_file_non_atomic', call[0])
903
self.assertEqual('filename.kndx', call[1][0])
904
# With no history, _KndxIndex writes a new index:
905
self.assertEqual(_KndxIndex.HEADER,
906
call[1][1].getvalue())
907
self.assertEqual({'create_parent_dir': True}, call[2])
909
def test_read_utf8_version_id(self):
910
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
911
utf8_revision_id = unicode_revision_id.encode('utf-8')
912
transport = MockTransport([
914
'%s option 0 1 :' % (utf8_revision_id,)
916
index = self.get_knit_index(transport, "filename", "r")
917
# _KndxIndex is a private class, and deals in utf8 revision_ids, not
918
# Unicode revision_ids.
919
self.assertEqual({(utf8_revision_id,):()},
920
index.get_parent_map(index.keys()))
921
self.assertFalse((unicode_revision_id,) in index.keys())
923
def test_read_utf8_parents(self):
924
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
925
utf8_revision_id = unicode_revision_id.encode('utf-8')
926
transport = MockTransport([
928
"version option 0 1 .%s :" % (utf8_revision_id,)
930
index = self.get_knit_index(transport, "filename", "r")
931
self.assertEqual({("version",):((utf8_revision_id,),)},
932
index.get_parent_map(index.keys()))
934
def test_read_ignore_corrupted_lines(self):
935
transport = MockTransport([
938
"corrupted options 0 1 .b .c ",
939
"version options 0 1 :"
941
index = self.get_knit_index(transport, "filename", "r")
942
self.assertEqual(1, len(index.keys()))
943
self.assertEqual(set([("version",)]), index.keys())
945
def test_read_corrupted_header(self):
946
transport = MockTransport(['not a bzr knit index header\n'])
947
index = self.get_knit_index(transport, "filename", "r")
948
self.assertRaises(KnitHeaderError, index.keys)
950
def test_read_duplicate_entries(self):
951
transport = MockTransport([
953
"parent options 0 1 :",
954
"version options1 0 1 0 :",
955
"version options2 1 2 .other :",
956
"version options3 3 4 0 .other :"
958
index = self.get_knit_index(transport, "filename", "r")
959
self.assertEqual(2, len(index.keys()))
960
# check that the index used is the first one written. (Specific
961
# to KnitIndex style indices.
962
self.assertEqual("1", index._dictionary_compress([("version",)]))
963
self.assertEqual((("version",), 3, 4), index.get_position(("version",)))
964
self.assertEqual(["options3"], index.get_options(("version",)))
965
self.assertEqual({("version",):(("parent",), ("other",))},
966
index.get_parent_map([("version",)]))
968
def test_read_compressed_parents(self):
969
transport = MockTransport([
973
"c option 0 1 1 0 :",
975
index = self.get_knit_index(transport, "filename", "r")
976
self.assertEqual({("b",):(("a",),), ("c",):(("b",), ("a",))},
977
index.get_parent_map([("b",), ("c",)]))
979
def test_write_utf8_version_id(self):
980
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
981
utf8_revision_id = unicode_revision_id.encode('utf-8')
982
transport = MockTransport([
985
index = self.get_knit_index(transport, "filename", "r")
987
((utf8_revision_id,), ["option"], ((utf8_revision_id,), 0, 1), [])])
988
call = transport.calls.pop(0)
989
# call[1][1] is a StringIO - we can't test it by simple equality.
990
self.assertEqual('put_file_non_atomic', call[0])
991
self.assertEqual('filename.kndx', call[1][0])
992
# With no history, _KndxIndex writes a new index:
993
self.assertEqual(_KndxIndex.HEADER +
994
"\n%s option 0 1 :" % (utf8_revision_id,),
995
call[1][1].getvalue())
996
self.assertEqual({'create_parent_dir': True}, call[2])
998
def test_write_utf8_parents(self):
999
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
1000
utf8_revision_id = unicode_revision_id.encode('utf-8')
1001
transport = MockTransport([
1004
index = self.get_knit_index(transport, "filename", "r")
1006
(("version",), ["option"], (("version",), 0, 1), [(utf8_revision_id,)])])
1007
call = transport.calls.pop(0)
1008
# call[1][1] is a StringIO - we can't test it by simple equality.
1009
self.assertEqual('put_file_non_atomic', call[0])
1010
self.assertEqual('filename.kndx', call[1][0])
1011
# With no history, _KndxIndex writes a new index:
1012
self.assertEqual(_KndxIndex.HEADER +
1013
"\nversion option 0 1 .%s :" % (utf8_revision_id,),
1014
call[1][1].getvalue())
1015
self.assertEqual({'create_parent_dir': True}, call[2])
1017
def test_keys(self):
1018
transport = MockTransport([
1021
index = self.get_knit_index(transport, "filename", "r")
1023
self.assertEqual(set(), index.keys())
1025
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1026
self.assertEqual(set([("a",)]), index.keys())
1028
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1029
self.assertEqual(set([("a",)]), index.keys())
1031
index.add_records([(("b",), ["option"], (("b",), 0, 1), [])])
1032
self.assertEqual(set([("a",), ("b",)]), index.keys())
1034
def add_a_b(self, index, random_id=None):
1036
if random_id is not None:
1037
kwargs["random_id"] = random_id
1039
(("a",), ["option"], (("a",), 0, 1), [("b",)]),
1040
(("a",), ["opt"], (("a",), 1, 2), [("c",)]),
1041
(("b",), ["option"], (("b",), 2, 3), [("a",)])
1044
def assertIndexIsAB(self, index):
1049
index.get_parent_map(index.keys()))
1050
self.assertEqual((("a",), 1, 2), index.get_position(("a",)))
1051
self.assertEqual((("b",), 2, 3), index.get_position(("b",)))
1052
self.assertEqual(["opt"], index.get_options(("a",)))
1054
def test_add_versions(self):
1055
transport = MockTransport([
1058
index = self.get_knit_index(transport, "filename", "r")
1061
call = transport.calls.pop(0)
1062
# call[1][1] is a StringIO - we can't test it by simple equality.
1063
self.assertEqual('put_file_non_atomic', call[0])
1064
self.assertEqual('filename.kndx', call[1][0])
1065
# With no history, _KndxIndex writes a new index:
1068
"\na option 0 1 .b :"
1070
"\nb option 2 3 0 :",
1071
call[1][1].getvalue())
1072
self.assertEqual({'create_parent_dir': True}, call[2])
1073
self.assertIndexIsAB(index)
1075
def test_add_versions_random_id_is_accepted(self):
1076
transport = MockTransport([
1079
index = self.get_knit_index(transport, "filename", "r")
1080
self.add_a_b(index, random_id=True)
1082
def test_delay_create_and_add_versions(self):
1083
transport = MockTransport()
1085
index = self.get_knit_index(transport, "filename", "w")
1087
self.assertEqual([], transport.calls)
1090
#[ {"dir_mode": 0777, "create_parent_dir": True, "mode": "wb"},
1092
# Two calls: one during which we load the existing index (and when its
1093
# missing create it), then a second where we write the contents out.
1094
self.assertEqual(2, len(transport.calls))
1095
call = transport.calls.pop(0)
1096
self.assertEqual('put_file_non_atomic', call[0])
1097
self.assertEqual('filename.kndx', call[1][0])
1098
# With no history, _KndxIndex writes a new index:
1099
self.assertEqual(_KndxIndex.HEADER, call[1][1].getvalue())
1100
self.assertEqual({'create_parent_dir': True}, call[2])
1101
call = transport.calls.pop(0)
1102
# call[1][1] is a StringIO - we can't test it by simple equality.
1103
self.assertEqual('put_file_non_atomic', call[0])
1104
self.assertEqual('filename.kndx', call[1][0])
1105
# With no history, _KndxIndex writes a new index:
1108
"\na option 0 1 .b :"
1110
"\nb option 2 3 0 :",
1111
call[1][1].getvalue())
1112
self.assertEqual({'create_parent_dir': True}, call[2])
1114
def assertTotalBuildSize(self, size, keys, positions):
1115
self.assertEqual(size,
1116
knit._get_total_build_size(None, keys, positions))
1118
def test__get_total_build_size(self):
1120
('a',): (('fulltext', False), (('a',), 0, 100), None),
1121
('b',): (('line-delta', False), (('b',), 100, 21), ('a',)),
1122
('c',): (('line-delta', False), (('c',), 121, 35), ('b',)),
1123
('d',): (('line-delta', False), (('d',), 156, 12), ('b',)),
1125
self.assertTotalBuildSize(100, [('a',)], positions)
1126
self.assertTotalBuildSize(121, [('b',)], positions)
1127
# c needs both a & b
1128
self.assertTotalBuildSize(156, [('c',)], positions)
1129
# we shouldn't count 'b' twice
1130
self.assertTotalBuildSize(156, [('b',), ('c',)], positions)
1131
self.assertTotalBuildSize(133, [('d',)], positions)
1132
self.assertTotalBuildSize(168, [('c',), ('d',)], positions)
1134
def test_get_position(self):
1135
transport = MockTransport([
1140
index = self.get_knit_index(transport, "filename", "r")
1142
self.assertEqual((("a",), 0, 1), index.get_position(("a",)))
1143
self.assertEqual((("b",), 1, 2), index.get_position(("b",)))
1145
def test_get_method(self):
1146
transport = MockTransport([
1148
"a fulltext,unknown 0 1 :",
1149
"b unknown,line-delta 1 2 :",
1152
index = self.get_knit_index(transport, "filename", "r")
1154
self.assertEqual("fulltext", index.get_method("a"))
1155
self.assertEqual("line-delta", index.get_method("b"))
1156
self.assertRaises(errors.KnitIndexUnknownMethod, index.get_method, "c")
1158
def test_get_options(self):
1159
transport = MockTransport([
1164
index = self.get_knit_index(transport, "filename", "r")
1166
self.assertEqual(["opt1"], index.get_options("a"))
1167
self.assertEqual(["opt2", "opt3"], index.get_options("b"))
1169
def test_get_parent_map(self):
1170
transport = MockTransport([
1173
"b option 1 2 0 .c :",
1174
"c option 1 2 1 0 .e :"
1176
index = self.get_knit_index(transport, "filename", "r")
1180
("b",):(("a",), ("c",)),
1181
("c",):(("b",), ("a",), ("e",)),
1182
}, index.get_parent_map(index.keys()))
1184
def test_impossible_parent(self):
1185
"""Test we get KnitCorrupt if the parent couldn't possibly exist."""
1186
transport = MockTransport([
1189
"b option 0 1 4 :" # We don't have a 4th record
1191
index = self.get_knit_index(transport, 'filename', 'r')
1193
self.assertRaises(errors.KnitCorrupt, index.keys)
1194
except TypeError, e:
1195
if (str(e) == ('exceptions must be strings, classes, or instances,'
1196
' not exceptions.IndexError')):
1197
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1198
' raising new style exceptions with python'
1203
def test_corrupted_parent(self):
1204
transport = MockTransport([
1208
"c option 0 1 1v :", # Can't have a parent of '1v'
1210
index = self.get_knit_index(transport, 'filename', 'r')
1212
self.assertRaises(errors.KnitCorrupt, index.keys)
1213
except TypeError, e:
1214
if (str(e) == ('exceptions must be strings, classes, or instances,'
1215
' not exceptions.ValueError')):
1216
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1217
' raising new style exceptions with python'
1222
def test_corrupted_parent_in_list(self):
1223
transport = MockTransport([
1227
"c option 0 1 1 v :", # Can't have a parent of 'v'
1229
index = self.get_knit_index(transport, 'filename', 'r')
1231
self.assertRaises(errors.KnitCorrupt, index.keys)
1232
except TypeError, e:
1233
if (str(e) == ('exceptions must be strings, classes, or instances,'
1234
' not exceptions.ValueError')):
1235
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1236
' raising new style exceptions with python'
1241
def test_invalid_position(self):
1242
transport = MockTransport([
1246
index = self.get_knit_index(transport, 'filename', 'r')
1248
self.assertRaises(errors.KnitCorrupt, index.keys)
1249
except TypeError, e:
1250
if (str(e) == ('exceptions must be strings, classes, or instances,'
1251
' not exceptions.ValueError')):
1252
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1253
' raising new style exceptions with python'
1258
def test_invalid_size(self):
1259
transport = MockTransport([
1263
index = self.get_knit_index(transport, 'filename', 'r')
1265
self.assertRaises(errors.KnitCorrupt, index.keys)
1266
except TypeError, e:
1267
if (str(e) == ('exceptions must be strings, classes, or instances,'
1268
' not exceptions.ValueError')):
1269
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1270
' raising new style exceptions with python'
1275
def test_scan_unvalidated_index_not_implemented(self):
1276
transport = MockTransport()
1277
index = self.get_knit_index(transport, 'filename', 'r')
1279
NotImplementedError, index.scan_unvalidated_index,
1280
'dummy graph_index')
1282
NotImplementedError, index.get_missing_compression_parents)
1284
def test_short_line(self):
1285
transport = MockTransport([
1288
"b option 10 10 0", # This line isn't terminated, ignored
1290
index = self.get_knit_index(transport, "filename", "r")
1291
self.assertEqual(set([('a',)]), index.keys())
1293
def test_skip_incomplete_record(self):
1294
# A line with bogus data should just be skipped
1295
transport = MockTransport([
1298
"b option 10 10 0", # This line isn't terminated, ignored
1299
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1301
index = self.get_knit_index(transport, "filename", "r")
1302
self.assertEqual(set([('a',), ('c',)]), index.keys())
1304
def test_trailing_characters(self):
1305
# A line with bogus data should just be skipped
1306
transport = MockTransport([
1309
"b option 10 10 0 :a", # This line has extra trailing characters
1310
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1312
index = self.get_knit_index(transport, "filename", "r")
1313
self.assertEqual(set([('a',), ('c',)]), index.keys())
1316
class LowLevelKnitIndexTests_c(LowLevelKnitIndexTests):
1318
_test_needs_features = [compiled_knit_feature]
1320
def get_knit_index(self, transport, name, mode):
1321
mapper = ConstantMapper(name)
1322
from bzrlib._knit_load_data_pyx import _load_data_c
1323
self.overrideAttr(knit, '_load_data', _load_data_c)
1324
allow_writes = lambda: mode == 'w'
1325
return _KndxIndex(transport, mapper, lambda:None,
1326
allow_writes, lambda:True)
1329
class Test_KnitAnnotator(TestCaseWithMemoryTransport):
1331
def make_annotator(self):
1332
factory = knit.make_pack_factory(True, True, 1)
1333
vf = factory(self.get_transport())
1334
return knit._KnitAnnotator(vf)
1336
def test__expand_fulltext(self):
1337
ann = self.make_annotator()
1338
rev_key = ('rev-id',)
1339
ann._num_compression_children[rev_key] = 1
1340
res = ann._expand_record(rev_key, (('parent-id',),), None,
1341
['line1\n', 'line2\n'], ('fulltext', True))
1342
# The content object and text lines should be cached appropriately
1343
self.assertEqual(['line1\n', 'line2'], res)
1344
content_obj = ann._content_objects[rev_key]
1345
self.assertEqual(['line1\n', 'line2\n'], content_obj._lines)
1346
self.assertEqual(res, content_obj.text())
1347
self.assertEqual(res, ann._text_cache[rev_key])
1349
def test__expand_delta_comp_parent_not_available(self):
1350
# Parent isn't available yet, so we return nothing, but queue up this
1351
# node for later processing
1352
ann = self.make_annotator()
1353
rev_key = ('rev-id',)
1354
parent_key = ('parent-id',)
1355
record = ['0,1,1\n', 'new-line\n']
1356
details = ('line-delta', False)
1357
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1359
self.assertEqual(None, res)
1360
self.assertTrue(parent_key in ann._pending_deltas)
1361
pending = ann._pending_deltas[parent_key]
1362
self.assertEqual(1, len(pending))
1363
self.assertEqual((rev_key, (parent_key,), record, details), pending[0])
1365
def test__expand_record_tracks_num_children(self):
1366
ann = self.make_annotator()
1367
rev_key = ('rev-id',)
1368
rev2_key = ('rev2-id',)
1369
parent_key = ('parent-id',)
1370
record = ['0,1,1\n', 'new-line\n']
1371
details = ('line-delta', False)
1372
ann._num_compression_children[parent_key] = 2
1373
ann._expand_record(parent_key, (), None, ['line1\n', 'line2\n'],
1374
('fulltext', False))
1375
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1377
self.assertEqual({parent_key: 1}, ann._num_compression_children)
1378
# Expanding the second child should remove the content object, and the
1379
# num_compression_children entry
1380
res = ann._expand_record(rev2_key, (parent_key,), parent_key,
1382
self.assertFalse(parent_key in ann._content_objects)
1383
self.assertEqual({}, ann._num_compression_children)
1384
# We should not cache the content_objects for rev2 and rev, because
1385
# they do not have compression children of their own.
1386
self.assertEqual({}, ann._content_objects)
1388
def test__expand_delta_records_blocks(self):
1389
ann = self.make_annotator()
1390
rev_key = ('rev-id',)
1391
parent_key = ('parent-id',)
1392
record = ['0,1,1\n', 'new-line\n']
1393
details = ('line-delta', True)
1394
ann._num_compression_children[parent_key] = 2
1395
ann._expand_record(parent_key, (), None,
1396
['line1\n', 'line2\n', 'line3\n'],
1397
('fulltext', False))
1398
ann._expand_record(rev_key, (parent_key,), parent_key, record, details)
1399
self.assertEqual({(rev_key, parent_key): [(1, 1, 1), (3, 3, 0)]},
1400
ann._matching_blocks)
1401
rev2_key = ('rev2-id',)
1402
record = ['0,1,1\n', 'new-line\n']
1403
details = ('line-delta', False)
1404
ann._expand_record(rev2_key, (parent_key,), parent_key, record, details)
1405
self.assertEqual([(1, 1, 2), (3, 3, 0)],
1406
ann._matching_blocks[(rev2_key, parent_key)])
1408
def test__get_parent_ann_uses_matching_blocks(self):
1409
ann = self.make_annotator()
1410
rev_key = ('rev-id',)
1411
parent_key = ('parent-id',)
1412
parent_ann = [(parent_key,)]*3
1413
block_key = (rev_key, parent_key)
1414
ann._annotations_cache[parent_key] = parent_ann
1415
ann._matching_blocks[block_key] = [(0, 1, 1), (3, 3, 0)]
1416
# We should not try to access any parent_lines content, because we know
1417
# we already have the matching blocks
1418
par_ann, blocks = ann._get_parent_annotations_and_matches(rev_key,
1419
['1\n', '2\n', '3\n'], parent_key)
1420
self.assertEqual(parent_ann, par_ann)
1421
self.assertEqual([(0, 1, 1), (3, 3, 0)], blocks)
1422
self.assertEqual({}, ann._matching_blocks)
1424
def test__process_pending(self):
1425
ann = self.make_annotator()
1426
rev_key = ('rev-id',)
1429
record = ['0,1,1\n', 'new-line\n']
1430
details = ('line-delta', False)
1431
p1_record = ['line1\n', 'line2\n']
1432
ann._num_compression_children[p1_key] = 1
1433
res = ann._expand_record(rev_key, (p1_key,p2_key), p1_key,
1435
self.assertEqual(None, res)
1436
# self.assertTrue(p1_key in ann._pending_deltas)
1437
self.assertEqual({}, ann._pending_annotation)
1438
# Now insert p1, and we should be able to expand the delta
1439
res = ann._expand_record(p1_key, (), None, p1_record,
1440
('fulltext', False))
1441
self.assertEqual(p1_record, res)
1442
ann._annotations_cache[p1_key] = [(p1_key,)]*2
1443
res = ann._process_pending(p1_key)
1444
self.assertEqual([], res)
1445
self.assertFalse(p1_key in ann._pending_deltas)
1446
self.assertTrue(p2_key in ann._pending_annotation)
1447
self.assertEqual({p2_key: [(rev_key, (p1_key, p2_key))]},
1448
ann._pending_annotation)
1449
# Now fill in parent 2, and pending annotation should be satisfied
1450
res = ann._expand_record(p2_key, (), None, [], ('fulltext', False))
1451
ann._annotations_cache[p2_key] = []
1452
res = ann._process_pending(p2_key)
1453
self.assertEqual([rev_key], res)
1454
self.assertEqual({}, ann._pending_annotation)
1455
self.assertEqual({}, ann._pending_deltas)
1457
def test_record_delta_removes_basis(self):
1458
ann = self.make_annotator()
1459
ann._expand_record(('parent-id',), (), None,
1460
['line1\n', 'line2\n'], ('fulltext', False))
1461
ann._num_compression_children['parent-id'] = 2
1463
def test_annotate_special_text(self):
1464
ann = self.make_annotator()
1466
rev1_key = ('rev-1',)
1467
rev2_key = ('rev-2',)
1468
rev3_key = ('rev-3',)
1469
spec_key = ('special:',)
1470
vf.add_lines(rev1_key, [], ['initial content\n'])
1471
vf.add_lines(rev2_key, [rev1_key], ['initial content\n',
1474
vf.add_lines(rev3_key, [rev1_key], ['initial content\n',
1477
spec_text = ('initial content\n'
1481
ann.add_special_text(spec_key, [rev2_key, rev3_key], spec_text)
1482
anns, lines = ann.annotate(spec_key)
1483
self.assertEqual([(rev1_key,),
1484
(rev2_key, rev3_key),
1488
self.assertEqualDiff(spec_text, ''.join(lines))
1491
class KnitTests(TestCaseWithTransport):
1492
"""Class containing knit test helper routines."""
1494
def make_test_knit(self, annotate=False, name='test'):
1495
mapper = ConstantMapper(name)
1496
return make_file_factory(annotate, mapper)(self.get_transport())
1499
class TestBadShaError(KnitTests):
1500
"""Tests for handling of sha errors."""
1502
def test_sha_exception_has_text(self):
1503
# having the failed text included in the error allows for recovery.
1504
source = self.make_test_knit()
1505
target = self.make_test_knit(name="target")
1506
if not source._max_delta_chain:
1507
raise TestNotApplicable(
1508
"cannot get delta-caused sha failures without deltas.")
1511
broken = ('broken',)
1512
source.add_lines(basis, (), ['foo\n'])
1513
source.add_lines(broken, (basis,), ['foo\n', 'bar\n'])
1514
# Seed target with a bad basis text
1515
target.add_lines(basis, (), ['gam\n'])
1516
target.insert_record_stream(
1517
source.get_record_stream([broken], 'unordered', False))
1518
err = self.assertRaises(errors.KnitCorrupt,
1519
target.get_record_stream([broken], 'unordered', True
1520
).next().get_bytes_as, 'chunked')
1521
self.assertEqual(['gam\n', 'bar\n'], err.content)
1522
# Test for formatting with live data
1523
self.assertStartsWith(str(err), "Knit ")
1526
class TestKnitIndex(KnitTests):
1528
def test_add_versions_dictionary_compresses(self):
1529
"""Adding versions to the index should update the lookup dict"""
1530
knit = self.make_test_knit()
1532
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1533
self.check_file_contents('test.kndx',
1534
'# bzr knit index 8\n'
1536
'a-1 fulltext 0 0 :'
1539
(('a-2',), ['fulltext'], (('a-2',), 0, 0), [('a-1',)]),
1540
(('a-3',), ['fulltext'], (('a-3',), 0, 0), [('a-2',)]),
1542
self.check_file_contents('test.kndx',
1543
'# bzr knit index 8\n'
1545
'a-1 fulltext 0 0 :\n'
1546
'a-2 fulltext 0 0 0 :\n'
1547
'a-3 fulltext 0 0 1 :'
1549
self.assertEqual(set([('a-3',), ('a-1',), ('a-2',)]), idx.keys())
1551
('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False)),
1552
('a-2',): ((('a-2',), 0, 0), None, (('a-1',),), ('fulltext', False)),
1553
('a-3',): ((('a-3',), 0, 0), None, (('a-2',),), ('fulltext', False)),
1554
}, idx.get_build_details(idx.keys()))
1555
self.assertEqual({('a-1',):(),
1556
('a-2',):(('a-1',),),
1557
('a-3',):(('a-2',),),},
1558
idx.get_parent_map(idx.keys()))
1560
def test_add_versions_fails_clean(self):
1561
"""If add_versions fails in the middle, it restores a pristine state.
1563
Any modifications that are made to the index are reset if all versions
1566
# This cheats a little bit by passing in a generator which will
1567
# raise an exception before the processing finishes
1568
# Other possibilities would be to have an version with the wrong number
1569
# of entries, or to make the backing transport unable to write any
1572
knit = self.make_test_knit()
1574
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1576
class StopEarly(Exception):
1579
def generate_failure():
1580
"""Add some entries and then raise an exception"""
1581
yield (('a-2',), ['fulltext'], (None, 0, 0), ('a-1',))
1582
yield (('a-3',), ['fulltext'], (None, 0, 0), ('a-2',))
1585
# Assert the pre-condition
1587
self.assertEqual(set([('a-1',)]), set(idx.keys()))
1589
{('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False))},
1590
idx.get_build_details([('a-1',)]))
1591
self.assertEqual({('a-1',):()}, idx.get_parent_map(idx.keys()))
1594
self.assertRaises(StopEarly, idx.add_records, generate_failure())
1595
# And it shouldn't be modified
1598
def test_knit_index_ignores_empty_files(self):
1599
# There was a race condition in older bzr, where a ^C at the right time
1600
# could leave an empty .kndx file, which bzr would later claim was a
1601
# corrupted file since the header was not present. In reality, the file
1602
# just wasn't created, so it should be ignored.
1603
t = transport.get_transport('.')
1604
t.put_bytes('test.kndx', '')
1606
knit = self.make_test_knit()
1608
def test_knit_index_checks_header(self):
1609
t = transport.get_transport('.')
1610
t.put_bytes('test.kndx', '# not really a knit header\n\n')
1611
k = self.make_test_knit()
1612
self.assertRaises(KnitHeaderError, k.keys)
1615
class TestGraphIndexKnit(KnitTests):
1616
"""Tests for knits using a GraphIndex rather than a KnitIndex."""
1618
def make_g_index(self, name, ref_lists=0, nodes=[]):
1619
builder = GraphIndexBuilder(ref_lists)
1620
for node, references, value in nodes:
1621
builder.add_node(node, references, value)
1622
stream = builder.finish()
1623
trans = self.get_transport()
1624
size = trans.put_file(name, stream)
1625
return GraphIndex(trans, name, size)
1627
def two_graph_index(self, deltas=False, catch_adds=False):
1628
"""Build a two-graph index.
1630
:param deltas: If true, use underlying indices with two node-ref
1631
lists and 'parent' set to a delta-compressed against tail.
1633
# build a complex graph across several indices.
1635
# delta compression inn the index
1636
index1 = self.make_g_index('1', 2, [
1637
(('tip', ), 'N0 100', ([('parent', )], [], )),
1638
(('tail', ), '', ([], []))])
1639
index2 = self.make_g_index('2', 2, [
1640
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], [('tail', )])),
1641
(('separate', ), '', ([], []))])
1643
# just blob location and graph in the index.
1644
index1 = self.make_g_index('1', 1, [
1645
(('tip', ), 'N0 100', ([('parent', )], )),
1646
(('tail', ), '', ([], ))])
1647
index2 = self.make_g_index('2', 1, [
1648
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], )),
1649
(('separate', ), '', ([], ))])
1650
combined_index = CombinedGraphIndex([index1, index2])
1652
self.combined_index = combined_index
1653
self.caught_entries = []
1654
add_callback = self.catch_add
1657
return _KnitGraphIndex(combined_index, lambda:True, deltas=deltas,
1658
add_callback=add_callback)
1660
def test_keys(self):
1661
index = self.two_graph_index()
1662
self.assertEqual(set([('tail',), ('tip',), ('parent',), ('separate',)]),
1665
def test_get_position(self):
1666
index = self.two_graph_index()
1667
self.assertEqual((index._graph_index._indices[0], 0, 100), index.get_position(('tip',)))
1668
self.assertEqual((index._graph_index._indices[1], 100, 78), index.get_position(('parent',)))
1670
def test_get_method_deltas(self):
1671
index = self.two_graph_index(deltas=True)
1672
self.assertEqual('fulltext', index.get_method(('tip',)))
1673
self.assertEqual('line-delta', index.get_method(('parent',)))
1675
def test_get_method_no_deltas(self):
1676
# check that the parent-history lookup is ignored with deltas=False.
1677
index = self.two_graph_index(deltas=False)
1678
self.assertEqual('fulltext', index.get_method(('tip',)))
1679
self.assertEqual('fulltext', index.get_method(('parent',)))
1681
def test_get_options_deltas(self):
1682
index = self.two_graph_index(deltas=True)
1683
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1684
self.assertEqual(['line-delta'], index.get_options(('parent',)))
1686
def test_get_options_no_deltas(self):
1687
# check that the parent-history lookup is ignored with deltas=False.
1688
index = self.two_graph_index(deltas=False)
1689
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1690
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1692
def test_get_parent_map(self):
1693
index = self.two_graph_index()
1694
self.assertEqual({('parent',):(('tail',), ('ghost',))},
1695
index.get_parent_map([('parent',), ('ghost',)]))
1697
def catch_add(self, entries):
1698
self.caught_entries.append(entries)
1700
def test_add_no_callback_errors(self):
1701
index = self.two_graph_index()
1702
self.assertRaises(errors.ReadOnlyError, index.add_records,
1703
[(('new',), 'fulltext,no-eol', (None, 50, 60), ['separate'])])
1705
def test_add_version_smoke(self):
1706
index = self.two_graph_index(catch_adds=True)
1707
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60),
1709
self.assertEqual([[(('new', ), 'N50 60', ((('separate',),),))]],
1710
self.caught_entries)
1712
def test_add_version_delta_not_delta_index(self):
1713
index = self.two_graph_index(catch_adds=True)
1714
self.assertRaises(errors.KnitCorrupt, index.add_records,
1715
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1716
self.assertEqual([], self.caught_entries)
1718
def test_add_version_same_dup(self):
1719
index = self.two_graph_index(catch_adds=True)
1720
# options can be spelt two different ways
1721
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
1722
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
1723
# position/length are ignored (because each pack could have fulltext or
1724
# delta, and be at a different position.
1725
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1727
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1729
# but neither should have added data:
1730
self.assertEqual([[], [], [], []], self.caught_entries)
1732
def test_add_version_different_dup(self):
1733
index = self.two_graph_index(deltas=True, catch_adds=True)
1735
self.assertRaises(errors.KnitCorrupt, index.add_records,
1736
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1737
self.assertRaises(errors.KnitCorrupt, index.add_records,
1738
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1740
self.assertRaises(errors.KnitCorrupt, index.add_records,
1741
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1742
self.assertEqual([], self.caught_entries)
1744
def test_add_versions_nodeltas(self):
1745
index = self.two_graph_index(catch_adds=True)
1747
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1748
(('new2',), 'fulltext', (None, 0, 6), [('new',)]),
1750
self.assertEqual([(('new', ), 'N50 60', ((('separate',),),)),
1751
(('new2', ), ' 0 6', ((('new',),),))],
1752
sorted(self.caught_entries[0]))
1753
self.assertEqual(1, len(self.caught_entries))
1755
def test_add_versions_deltas(self):
1756
index = self.two_graph_index(deltas=True, catch_adds=True)
1758
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1759
(('new2',), 'line-delta', (None, 0, 6), [('new',)]),
1761
self.assertEqual([(('new', ), 'N50 60', ((('separate',),), ())),
1762
(('new2', ), ' 0 6', ((('new',),), (('new',),), ))],
1763
sorted(self.caught_entries[0]))
1764
self.assertEqual(1, len(self.caught_entries))
1766
def test_add_versions_delta_not_delta_index(self):
1767
index = self.two_graph_index(catch_adds=True)
1768
self.assertRaises(errors.KnitCorrupt, index.add_records,
1769
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1770
self.assertEqual([], self.caught_entries)
1772
def test_add_versions_random_id_accepted(self):
1773
index = self.two_graph_index(catch_adds=True)
1774
index.add_records([], random_id=True)
1776
def test_add_versions_same_dup(self):
1777
index = self.two_graph_index(catch_adds=True)
1778
# options can be spelt two different ways
1779
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100),
1781
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100),
1783
# position/length are ignored (because each pack could have fulltext or
1784
# delta, and be at a different position.
1785
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1787
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1789
# but neither should have added data.
1790
self.assertEqual([[], [], [], []], self.caught_entries)
1792
def test_add_versions_different_dup(self):
1793
index = self.two_graph_index(deltas=True, catch_adds=True)
1795
self.assertRaises(errors.KnitCorrupt, index.add_records,
1796
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1797
self.assertRaises(errors.KnitCorrupt, index.add_records,
1798
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1800
self.assertRaises(errors.KnitCorrupt, index.add_records,
1801
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1802
# change options in the second record
1803
self.assertRaises(errors.KnitCorrupt, index.add_records,
1804
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)]),
1805
(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1806
self.assertEqual([], self.caught_entries)
1808
def make_g_index_missing_compression_parent(self):
1809
graph_index = self.make_g_index('missing_comp', 2,
1810
[(('tip', ), ' 100 78',
1811
([('missing-parent', ), ('ghost', )], [('missing-parent', )]))])
1814
def make_g_index_missing_parent(self):
1815
graph_index = self.make_g_index('missing_parent', 2,
1816
[(('parent', ), ' 100 78', ([], [])),
1817
(('tip', ), ' 100 78',
1818
([('parent', ), ('missing-parent', )], [('parent', )])),
1822
def make_g_index_no_external_refs(self):
1823
graph_index = self.make_g_index('no_external_refs', 2,
1824
[(('rev', ), ' 100 78',
1825
([('parent', ), ('ghost', )], []))])
1828
def test_add_good_unvalidated_index(self):
1829
unvalidated = self.make_g_index_no_external_refs()
1830
combined = CombinedGraphIndex([unvalidated])
1831
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1832
index.scan_unvalidated_index(unvalidated)
1833
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1835
def test_add_missing_compression_parent_unvalidated_index(self):
1836
unvalidated = self.make_g_index_missing_compression_parent()
1837
combined = CombinedGraphIndex([unvalidated])
1838
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1839
index.scan_unvalidated_index(unvalidated)
1840
# This also checks that its only the compression parent that is
1841
# examined, otherwise 'ghost' would also be reported as a missing
1844
frozenset([('missing-parent',)]),
1845
index.get_missing_compression_parents())
1847
def test_add_missing_noncompression_parent_unvalidated_index(self):
1848
unvalidated = self.make_g_index_missing_parent()
1849
combined = CombinedGraphIndex([unvalidated])
1850
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1851
track_external_parent_refs=True)
1852
index.scan_unvalidated_index(unvalidated)
1854
frozenset([('missing-parent',)]), index.get_missing_parents())
1856
def test_track_external_parent_refs(self):
1857
g_index = self.make_g_index('empty', 2, [])
1858
combined = CombinedGraphIndex([g_index])
1859
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1860
add_callback=self.catch_add, track_external_parent_refs=True)
1861
self.caught_entries = []
1863
(('new-key',), 'fulltext,no-eol', (None, 50, 60),
1864
[('parent-1',), ('parent-2',)])])
1866
frozenset([('parent-1',), ('parent-2',)]),
1867
index.get_missing_parents())
1869
def test_add_unvalidated_index_with_present_external_references(self):
1870
index = self.two_graph_index(deltas=True)
1871
# Ugly hack to get at one of the underlying GraphIndex objects that
1872
# two_graph_index built.
1873
unvalidated = index._graph_index._indices[1]
1874
# 'parent' is an external ref of _indices[1] (unvalidated), but is
1875
# present in _indices[0].
1876
index.scan_unvalidated_index(unvalidated)
1877
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1879
def make_new_missing_parent_g_index(self, name):
1880
missing_parent = name + '-missing-parent'
1881
graph_index = self.make_g_index(name, 2,
1882
[((name + 'tip', ), ' 100 78',
1883
([(missing_parent, ), ('ghost', )], [(missing_parent, )]))])
1886
def test_add_mulitiple_unvalidated_indices_with_missing_parents(self):
1887
g_index_1 = self.make_new_missing_parent_g_index('one')
1888
g_index_2 = self.make_new_missing_parent_g_index('two')
1889
combined = CombinedGraphIndex([g_index_1, g_index_2])
1890
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1891
index.scan_unvalidated_index(g_index_1)
1892
index.scan_unvalidated_index(g_index_2)
1894
frozenset([('one-missing-parent',), ('two-missing-parent',)]),
1895
index.get_missing_compression_parents())
1897
def test_add_mulitiple_unvalidated_indices_with_mutual_dependencies(self):
1898
graph_index_a = self.make_g_index('one', 2,
1899
[(('parent-one', ), ' 100 78', ([('non-compression-parent',)], [])),
1900
(('child-of-two', ), ' 100 78',
1901
([('parent-two',)], [('parent-two',)]))])
1902
graph_index_b = self.make_g_index('two', 2,
1903
[(('parent-two', ), ' 100 78', ([('non-compression-parent',)], [])),
1904
(('child-of-one', ), ' 100 78',
1905
([('parent-one',)], [('parent-one',)]))])
1906
combined = CombinedGraphIndex([graph_index_a, graph_index_b])
1907
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1908
index.scan_unvalidated_index(graph_index_a)
1909
index.scan_unvalidated_index(graph_index_b)
1911
frozenset([]), index.get_missing_compression_parents())
1914
class TestNoParentsGraphIndexKnit(KnitTests):
1915
"""Tests for knits using _KnitGraphIndex with no parents."""
1917
def make_g_index(self, name, ref_lists=0, nodes=[]):
1918
builder = GraphIndexBuilder(ref_lists)
1919
for node, references in nodes:
1920
builder.add_node(node, references)
1921
stream = builder.finish()
1922
trans = self.get_transport()
1923
size = trans.put_file(name, stream)
1924
return GraphIndex(trans, name, size)
1926
def test_add_good_unvalidated_index(self):
1927
unvalidated = self.make_g_index('unvalidated')
1928
combined = CombinedGraphIndex([unvalidated])
1929
index = _KnitGraphIndex(combined, lambda: True, parents=False)
1930
index.scan_unvalidated_index(unvalidated)
1931
self.assertEqual(frozenset(),
1932
index.get_missing_compression_parents())
1934
def test_parents_deltas_incompatible(self):
1935
index = CombinedGraphIndex([])
1936
self.assertRaises(errors.KnitError, _KnitGraphIndex, lambda:True,
1937
index, deltas=True, parents=False)
1939
def two_graph_index(self, catch_adds=False):
1940
"""Build a two-graph index.
1942
:param deltas: If true, use underlying indices with two node-ref
1943
lists and 'parent' set to a delta-compressed against tail.
1945
# put several versions in the index.
1946
index1 = self.make_g_index('1', 0, [
1947
(('tip', ), 'N0 100'),
1949
index2 = self.make_g_index('2', 0, [
1950
(('parent', ), ' 100 78'),
1951
(('separate', ), '')])
1952
combined_index = CombinedGraphIndex([index1, index2])
1954
self.combined_index = combined_index
1955
self.caught_entries = []
1956
add_callback = self.catch_add
1959
return _KnitGraphIndex(combined_index, lambda:True, parents=False,
1960
add_callback=add_callback)
1962
def test_keys(self):
1963
index = self.two_graph_index()
1964
self.assertEqual(set([('tail',), ('tip',), ('parent',), ('separate',)]),
1967
def test_get_position(self):
1968
index = self.two_graph_index()
1969
self.assertEqual((index._graph_index._indices[0], 0, 100),
1970
index.get_position(('tip',)))
1971
self.assertEqual((index._graph_index._indices[1], 100, 78),
1972
index.get_position(('parent',)))
1974
def test_get_method(self):
1975
index = self.two_graph_index()
1976
self.assertEqual('fulltext', index.get_method(('tip',)))
1977
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1979
def test_get_options(self):
1980
index = self.two_graph_index()
1981
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1982
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1984
def test_get_parent_map(self):
1985
index = self.two_graph_index()
1986
self.assertEqual({('parent',):None},
1987
index.get_parent_map([('parent',), ('ghost',)]))
1989
def catch_add(self, entries):
1990
self.caught_entries.append(entries)
1992
def test_add_no_callback_errors(self):
1993
index = self.two_graph_index()
1994
self.assertRaises(errors.ReadOnlyError, index.add_records,
1995
[(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)])])
1997
def test_add_version_smoke(self):
1998
index = self.two_graph_index(catch_adds=True)
1999
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60), [])])
2000
self.assertEqual([[(('new', ), 'N50 60')]],
2001
self.caught_entries)
2003
def test_add_version_delta_not_delta_index(self):
2004
index = self.two_graph_index(catch_adds=True)
2005
self.assertRaises(errors.KnitCorrupt, index.add_records,
2006
[(('new',), 'no-eol,line-delta', (None, 0, 100), [])])
2007
self.assertEqual([], self.caught_entries)
2009
def test_add_version_same_dup(self):
2010
index = self.two_graph_index(catch_adds=True)
2011
# options can be spelt two different ways
2012
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
2013
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
2014
# position/length are ignored (because each pack could have fulltext or
2015
# delta, and be at a different position.
2016
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
2017
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
2018
# but neither should have added data.
2019
self.assertEqual([[], [], [], []], self.caught_entries)
2021
def test_add_version_different_dup(self):
2022
index = self.two_graph_index(catch_adds=True)
2024
self.assertRaises(errors.KnitCorrupt, index.add_records,
2025
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2026
self.assertRaises(errors.KnitCorrupt, index.add_records,
2027
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
2028
self.assertRaises(errors.KnitCorrupt, index.add_records,
2029
[(('tip',), 'fulltext', (None, 0, 100), [])])
2031
self.assertRaises(errors.KnitCorrupt, index.add_records,
2032
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
2033
self.assertEqual([], self.caught_entries)
2035
def test_add_versions(self):
2036
index = self.two_graph_index(catch_adds=True)
2038
(('new',), 'fulltext,no-eol', (None, 50, 60), []),
2039
(('new2',), 'fulltext', (None, 0, 6), []),
2041
self.assertEqual([(('new', ), 'N50 60'), (('new2', ), ' 0 6')],
2042
sorted(self.caught_entries[0]))
2043
self.assertEqual(1, len(self.caught_entries))
2045
def test_add_versions_delta_not_delta_index(self):
2046
index = self.two_graph_index(catch_adds=True)
2047
self.assertRaises(errors.KnitCorrupt, index.add_records,
2048
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
2049
self.assertEqual([], self.caught_entries)
2051
def test_add_versions_parents_not_parents_index(self):
2052
index = self.two_graph_index(catch_adds=True)
2053
self.assertRaises(errors.KnitCorrupt, index.add_records,
2054
[(('new',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
2055
self.assertEqual([], self.caught_entries)
2057
def test_add_versions_random_id_accepted(self):
2058
index = self.two_graph_index(catch_adds=True)
2059
index.add_records([], random_id=True)
2061
def test_add_versions_same_dup(self):
2062
index = self.two_graph_index(catch_adds=True)
2063
# options can be spelt two different ways
2064
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
2065
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
2066
# position/length are ignored (because each pack could have fulltext or
2067
# delta, and be at a different position.
2068
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
2069
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
2070
# but neither should have added data.
2071
self.assertEqual([[], [], [], []], self.caught_entries)
2073
def test_add_versions_different_dup(self):
2074
index = self.two_graph_index(catch_adds=True)
2076
self.assertRaises(errors.KnitCorrupt, index.add_records,
2077
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2078
self.assertRaises(errors.KnitCorrupt, index.add_records,
2079
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
2080
self.assertRaises(errors.KnitCorrupt, index.add_records,
2081
[(('tip',), 'fulltext', (None, 0, 100), [])])
2083
self.assertRaises(errors.KnitCorrupt, index.add_records,
2084
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
2085
# change options in the second record
2086
self.assertRaises(errors.KnitCorrupt, index.add_records,
2087
[(('tip',), 'fulltext,no-eol', (None, 0, 100), []),
2088
(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2089
self.assertEqual([], self.caught_entries)
2092
class TestKnitVersionedFiles(KnitTests):
2094
def assertGroupKeysForIo(self, exp_groups, keys, non_local_keys,
2095
positions, _min_buffer_size=None):
2096
kvf = self.make_test_knit()
2097
if _min_buffer_size is None:
2098
_min_buffer_size = knit._STREAM_MIN_BUFFER_SIZE
2099
self.assertEqual(exp_groups, kvf._group_keys_for_io(keys,
2100
non_local_keys, positions,
2101
_min_buffer_size=_min_buffer_size))
2103
def assertSplitByPrefix(self, expected_map, expected_prefix_order,
2105
split, prefix_order = KnitVersionedFiles._split_by_prefix(keys)
2106
self.assertEqual(expected_map, split)
2107
self.assertEqual(expected_prefix_order, prefix_order)
2109
def test__group_keys_for_io(self):
2110
ft_detail = ('fulltext', False)
2111
ld_detail = ('line-delta', False)
2119
f_a: (ft_detail, (f_a, 0, 100), None),
2120
f_b: (ld_detail, (f_b, 100, 21), f_a),
2121
f_c: (ld_detail, (f_c, 180, 15), f_b),
2122
g_a: (ft_detail, (g_a, 121, 35), None),
2123
g_b: (ld_detail, (g_b, 156, 12), g_a),
2124
g_c: (ld_detail, (g_c, 195, 13), g_a),
2126
self.assertGroupKeysForIo([([f_a], set())],
2127
[f_a], [], positions)
2128
self.assertGroupKeysForIo([([f_a], set([f_a]))],
2129
[f_a], [f_a], positions)
2130
self.assertGroupKeysForIo([([f_a, f_b], set([]))],
2131
[f_a, f_b], [], positions)
2132
self.assertGroupKeysForIo([([f_a, f_b], set([f_b]))],
2133
[f_a, f_b], [f_b], positions)
2134
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2135
[f_a, g_a, f_b, g_b], [], positions)
2136
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2137
[f_a, g_a, f_b, g_b], [], positions,
2138
_min_buffer_size=150)
2139
self.assertGroupKeysForIo([([f_a, f_b], set()), ([g_a, g_b], set())],
2140
[f_a, g_a, f_b, g_b], [], positions,
2141
_min_buffer_size=100)
2142
self.assertGroupKeysForIo([([f_c], set()), ([g_b], set())],
2143
[f_c, g_b], [], positions,
2144
_min_buffer_size=125)
2145
self.assertGroupKeysForIo([([g_b, f_c], set())],
2146
[g_b, f_c], [], positions,
2147
_min_buffer_size=125)
2149
def test__split_by_prefix(self):
2150
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2151
'g': [('g', 'b'), ('g', 'a')],
2153
[('f', 'a'), ('g', 'b'),
2154
('g', 'a'), ('f', 'b')])
2156
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2157
'g': [('g', 'b'), ('g', 'a')],
2159
[('f', 'a'), ('f', 'b'),
2160
('g', 'b'), ('g', 'a')])
2162
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2163
'g': [('g', 'b'), ('g', 'a')],
2165
[('f', 'a'), ('f', 'b'),
2166
('g', 'b'), ('g', 'a')])
2168
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2169
'g': [('g', 'b'), ('g', 'a')],
2170
'': [('a',), ('b',)]
2172
[('f', 'a'), ('g', 'b'),
2174
('g', 'a'), ('f', 'b')])
2177
class TestStacking(KnitTests):
2179
def get_basis_and_test_knit(self):
2180
basis = self.make_test_knit(name='basis')
2181
basis = RecordingVersionedFilesDecorator(basis)
2182
test = self.make_test_knit(name='test')
2183
test.add_fallback_versioned_files(basis)
2186
def test_add_fallback_versioned_files(self):
2187
basis = self.make_test_knit(name='basis')
2188
test = self.make_test_knit(name='test')
2189
# It must not error; other tests test that the fallback is referred to
2190
# when accessing data.
2191
test.add_fallback_versioned_files(basis)
2193
def test_add_lines(self):
2194
# lines added to the test are not added to the basis
2195
basis, test = self.get_basis_and_test_knit()
2197
key_basis = ('bar',)
2198
key_cross_border = ('quux',)
2199
key_delta = ('zaphod',)
2200
test.add_lines(key, (), ['foo\n'])
2201
self.assertEqual({}, basis.get_parent_map([key]))
2202
# lines added to the test that reference across the stack do a
2204
basis.add_lines(key_basis, (), ['foo\n'])
2206
test.add_lines(key_cross_border, (key_basis,), ['foo\n'])
2207
self.assertEqual('fulltext', test._index.get_method(key_cross_border))
2208
# we don't even need to look at the basis to see that this should be
2209
# stored as a fulltext
2210
self.assertEqual([], basis.calls)
2211
# Subsequent adds do delta.
2213
test.add_lines(key_delta, (key_cross_border,), ['foo\n'])
2214
self.assertEqual('line-delta', test._index.get_method(key_delta))
2215
self.assertEqual([], basis.calls)
2217
def test_annotate(self):
2218
# annotations from the test knit are answered without asking the basis
2219
basis, test = self.get_basis_and_test_knit()
2221
key_basis = ('bar',)
2222
key_missing = ('missing',)
2223
test.add_lines(key, (), ['foo\n'])
2224
details = test.annotate(key)
2225
self.assertEqual([(key, 'foo\n')], details)
2226
self.assertEqual([], basis.calls)
2227
# But texts that are not in the test knit are looked for in the basis
2229
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2231
details = test.annotate(key_basis)
2232
self.assertEqual([(key_basis, 'foo\n'), (key_basis, 'bar\n')], details)
2233
# Not optimised to date:
2234
# self.assertEqual([("annotate", key_basis)], basis.calls)
2235
self.assertEqual([('get_parent_map', set([key_basis])),
2236
('get_parent_map', set([key_basis])),
2237
('get_record_stream', [key_basis], 'topological', True)],
2240
def test_check(self):
2241
# At the moment checking a stacked knit does implicitly check the
2243
basis, test = self.get_basis_and_test_knit()
2246
def test_get_parent_map(self):
2247
# parents in the test knit are answered without asking the basis
2248
basis, test = self.get_basis_and_test_knit()
2250
key_basis = ('bar',)
2251
key_missing = ('missing',)
2252
test.add_lines(key, (), [])
2253
parent_map = test.get_parent_map([key])
2254
self.assertEqual({key: ()}, parent_map)
2255
self.assertEqual([], basis.calls)
2256
# But parents that are not in the test knit are looked for in the basis
2257
basis.add_lines(key_basis, (), [])
2259
parent_map = test.get_parent_map([key, key_basis, key_missing])
2260
self.assertEqual({key: (),
2261
key_basis: ()}, parent_map)
2262
self.assertEqual([("get_parent_map", set([key_basis, key_missing]))],
2265
def test_get_record_stream_unordered_fulltexts(self):
2266
# records from the test knit are answered without asking the basis:
2267
basis, test = self.get_basis_and_test_knit()
2269
key_basis = ('bar',)
2270
key_missing = ('missing',)
2271
test.add_lines(key, (), ['foo\n'])
2272
records = list(test.get_record_stream([key], 'unordered', True))
2273
self.assertEqual(1, len(records))
2274
self.assertEqual([], basis.calls)
2275
# Missing (from test knit) objects are retrieved from the basis:
2276
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2278
records = list(test.get_record_stream([key_basis, key_missing],
2280
self.assertEqual(2, len(records))
2281
calls = list(basis.calls)
2282
for record in records:
2283
self.assertSubset([record.key], (key_basis, key_missing))
2284
if record.key == key_missing:
2285
self.assertIsInstance(record, AbsentContentFactory)
2287
reference = list(basis.get_record_stream([key_basis],
2288
'unordered', True))[0]
2289
self.assertEqual(reference.key, record.key)
2290
self.assertEqual(reference.sha1, record.sha1)
2291
self.assertEqual(reference.storage_kind, record.storage_kind)
2292
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2293
record.get_bytes_as(record.storage_kind))
2294
self.assertEqual(reference.get_bytes_as('fulltext'),
2295
record.get_bytes_as('fulltext'))
2296
# It's not strictly minimal, but it seems reasonable for now for it to
2297
# ask which fallbacks have which parents.
2299
("get_parent_map", set([key_basis, key_missing])),
2300
("get_record_stream", [key_basis], 'unordered', True)],
2303
def test_get_record_stream_ordered_fulltexts(self):
2304
# ordering is preserved down into the fallback store.
2305
basis, test = self.get_basis_and_test_knit()
2307
key_basis = ('bar',)
2308
key_basis_2 = ('quux',)
2309
key_missing = ('missing',)
2310
test.add_lines(key, (key_basis,), ['foo\n'])
2311
# Missing (from test knit) objects are retrieved from the basis:
2312
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2313
basis.add_lines(key_basis_2, (), ['quux\n'])
2315
# ask for in non-topological order
2316
records = list(test.get_record_stream(
2317
[key, key_basis, key_missing, key_basis_2], 'topological', True))
2318
self.assertEqual(4, len(records))
2320
for record in records:
2321
self.assertSubset([record.key],
2322
(key_basis, key_missing, key_basis_2, key))
2323
if record.key == key_missing:
2324
self.assertIsInstance(record, AbsentContentFactory)
2326
results.append((record.key, record.sha1, record.storage_kind,
2327
record.get_bytes_as('fulltext')))
2328
calls = list(basis.calls)
2329
order = [record[0] for record in results]
2330
self.assertEqual([key_basis_2, key_basis, key], order)
2331
for result in results:
2332
if result[0] == key:
2336
record = source.get_record_stream([result[0]], 'unordered',
2338
self.assertEqual(record.key, result[0])
2339
self.assertEqual(record.sha1, result[1])
2340
# We used to check that the storage kind matched, but actually it
2341
# depends on whether it was sourced from the basis, or in a single
2342
# group, because asking for full texts returns proxy objects to a
2343
# _ContentMapGenerator object; so checking the kind is unneeded.
2344
self.assertEqual(record.get_bytes_as('fulltext'), result[3])
2345
# It's not strictly minimal, but it seems reasonable for now for it to
2346
# ask which fallbacks have which parents.
2348
("get_parent_map", set([key_basis, key_basis_2, key_missing])),
2349
# topological is requested from the fallback, because that is what
2350
# was requested at the top level.
2351
("get_record_stream", [key_basis_2, key_basis], 'topological', True)],
2354
def test_get_record_stream_unordered_deltas(self):
2355
# records from the test knit are answered without asking the basis:
2356
basis, test = self.get_basis_and_test_knit()
2358
key_basis = ('bar',)
2359
key_missing = ('missing',)
2360
test.add_lines(key, (), ['foo\n'])
2361
records = list(test.get_record_stream([key], 'unordered', False))
2362
self.assertEqual(1, len(records))
2363
self.assertEqual([], basis.calls)
2364
# Missing (from test knit) objects are retrieved from the basis:
2365
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2367
records = list(test.get_record_stream([key_basis, key_missing],
2368
'unordered', False))
2369
self.assertEqual(2, len(records))
2370
calls = list(basis.calls)
2371
for record in records:
2372
self.assertSubset([record.key], (key_basis, key_missing))
2373
if record.key == key_missing:
2374
self.assertIsInstance(record, AbsentContentFactory)
2376
reference = list(basis.get_record_stream([key_basis],
2377
'unordered', False))[0]
2378
self.assertEqual(reference.key, record.key)
2379
self.assertEqual(reference.sha1, record.sha1)
2380
self.assertEqual(reference.storage_kind, record.storage_kind)
2381
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2382
record.get_bytes_as(record.storage_kind))
2383
# It's not strictly minimal, but it seems reasonable for now for it to
2384
# ask which fallbacks have which parents.
2386
("get_parent_map", set([key_basis, key_missing])),
2387
("get_record_stream", [key_basis], 'unordered', False)],
2390
def test_get_record_stream_ordered_deltas(self):
2391
# ordering is preserved down into the fallback store.
2392
basis, test = self.get_basis_and_test_knit()
2394
key_basis = ('bar',)
2395
key_basis_2 = ('quux',)
2396
key_missing = ('missing',)
2397
test.add_lines(key, (key_basis,), ['foo\n'])
2398
# Missing (from test knit) objects are retrieved from the basis:
2399
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2400
basis.add_lines(key_basis_2, (), ['quux\n'])
2402
# ask for in non-topological order
2403
records = list(test.get_record_stream(
2404
[key, key_basis, key_missing, key_basis_2], 'topological', False))
2405
self.assertEqual(4, len(records))
2407
for record in records:
2408
self.assertSubset([record.key],
2409
(key_basis, key_missing, key_basis_2, key))
2410
if record.key == key_missing:
2411
self.assertIsInstance(record, AbsentContentFactory)
2413
results.append((record.key, record.sha1, record.storage_kind,
2414
record.get_bytes_as(record.storage_kind)))
2415
calls = list(basis.calls)
2416
order = [record[0] for record in results]
2417
self.assertEqual([key_basis_2, key_basis, key], order)
2418
for result in results:
2419
if result[0] == key:
2423
record = source.get_record_stream([result[0]], 'unordered',
2425
self.assertEqual(record.key, result[0])
2426
self.assertEqual(record.sha1, result[1])
2427
self.assertEqual(record.storage_kind, result[2])
2428
self.assertEqual(record.get_bytes_as(record.storage_kind), result[3])
2429
# It's not strictly minimal, but it seems reasonable for now for it to
2430
# ask which fallbacks have which parents.
2432
("get_parent_map", set([key_basis, key_basis_2, key_missing])),
2433
("get_record_stream", [key_basis_2, key_basis], 'topological', False)],
2436
def test_get_sha1s(self):
2437
# sha1's in the test knit are answered without asking the basis
2438
basis, test = self.get_basis_and_test_knit()
2440
key_basis = ('bar',)
2441
key_missing = ('missing',)
2442
test.add_lines(key, (), ['foo\n'])
2443
key_sha1sum = osutils.sha_string('foo\n')
2444
sha1s = test.get_sha1s([key])
2445
self.assertEqual({key: key_sha1sum}, sha1s)
2446
self.assertEqual([], basis.calls)
2447
# But texts that are not in the test knit are looked for in the basis
2448
# directly (rather than via text reconstruction) so that remote servers
2449
# etc don't have to answer with full content.
2450
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2451
basis_sha1sum = osutils.sha_string('foo\nbar\n')
2453
sha1s = test.get_sha1s([key, key_missing, key_basis])
2454
self.assertEqual({key: key_sha1sum,
2455
key_basis: basis_sha1sum}, sha1s)
2456
self.assertEqual([("get_sha1s", set([key_basis, key_missing]))],
2459
def test_insert_record_stream(self):
2460
# records are inserted as normal; insert_record_stream builds on
2461
# add_lines, so a smoke test should be all that's needed:
2463
key_basis = ('bar',)
2464
key_delta = ('zaphod',)
2465
basis, test = self.get_basis_and_test_knit()
2466
source = self.make_test_knit(name='source')
2467
basis.add_lines(key_basis, (), ['foo\n'])
2469
source.add_lines(key_basis, (), ['foo\n'])
2470
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2471
stream = source.get_record_stream([key_delta], 'unordered', False)
2472
test.insert_record_stream(stream)
2473
# XXX: this does somewhat too many calls in making sure of whether it
2474
# has to recreate the full text.
2475
self.assertEqual([("get_parent_map", set([key_basis])),
2476
('get_parent_map', set([key_basis])),
2477
('get_record_stream', [key_basis], 'unordered', True)],
2479
self.assertEqual({key_delta:(key_basis,)},
2480
test.get_parent_map([key_delta]))
2481
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2482
'unordered', True).next().get_bytes_as('fulltext'))
2484
def test_iter_lines_added_or_present_in_keys(self):
2485
# Lines from the basis are returned, and lines for a given key are only
2489
# all sources are asked for keys:
2490
basis, test = self.get_basis_and_test_knit()
2491
basis.add_lines(key1, (), ["foo"])
2493
lines = list(test.iter_lines_added_or_present_in_keys([key1]))
2494
self.assertEqual([("foo\n", key1)], lines)
2495
self.assertEqual([("iter_lines_added_or_present_in_keys", set([key1]))],
2497
# keys in both are not duplicated:
2498
test.add_lines(key2, (), ["bar\n"])
2499
basis.add_lines(key2, (), ["bar\n"])
2501
lines = list(test.iter_lines_added_or_present_in_keys([key2]))
2502
self.assertEqual([("bar\n", key2)], lines)
2503
self.assertEqual([], basis.calls)
2505
def test_keys(self):
2508
# all sources are asked for keys:
2509
basis, test = self.get_basis_and_test_knit()
2511
self.assertEqual(set(), set(keys))
2512
self.assertEqual([("keys",)], basis.calls)
2513
# keys from a basis are returned:
2514
basis.add_lines(key1, (), [])
2517
self.assertEqual(set([key1]), set(keys))
2518
self.assertEqual([("keys",)], basis.calls)
2519
# keys in both are not duplicated:
2520
test.add_lines(key2, (), [])
2521
basis.add_lines(key2, (), [])
2524
self.assertEqual(2, len(keys))
2525
self.assertEqual(set([key1, key2]), set(keys))
2526
self.assertEqual([("keys",)], basis.calls)
2528
def test_add_mpdiffs(self):
2529
# records are inserted as normal; add_mpdiff builds on
2530
# add_lines, so a smoke test should be all that's needed:
2532
key_basis = ('bar',)
2533
key_delta = ('zaphod',)
2534
basis, test = self.get_basis_and_test_knit()
2535
source = self.make_test_knit(name='source')
2536
basis.add_lines(key_basis, (), ['foo\n'])
2538
source.add_lines(key_basis, (), ['foo\n'])
2539
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2540
diffs = source.make_mpdiffs([key_delta])
2541
test.add_mpdiffs([(key_delta, (key_basis,),
2542
source.get_sha1s([key_delta])[key_delta], diffs[0])])
2543
self.assertEqual([("get_parent_map", set([key_basis])),
2544
('get_record_stream', [key_basis], 'unordered', True),],
2546
self.assertEqual({key_delta:(key_basis,)},
2547
test.get_parent_map([key_delta]))
2548
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2549
'unordered', True).next().get_bytes_as('fulltext'))
2551
def test_make_mpdiffs(self):
2552
# Generating an mpdiff across a stacking boundary should detect parent
2556
key_right = ('zaphod',)
2557
basis, test = self.get_basis_and_test_knit()
2558
basis.add_lines(key_left, (), ['bar\n'])
2559
basis.add_lines(key_right, (), ['zaphod\n'])
2561
test.add_lines(key, (key_left, key_right),
2562
['bar\n', 'foo\n', 'zaphod\n'])
2563
diffs = test.make_mpdiffs([key])
2565
multiparent.MultiParent([multiparent.ParentText(0, 0, 0, 1),
2566
multiparent.NewText(['foo\n']),
2567
multiparent.ParentText(1, 0, 2, 1)])],
2569
self.assertEqual(3, len(basis.calls))
2571
("get_parent_map", set([key_left, key_right])),
2572
("get_parent_map", set([key_left, key_right])),
2575
last_call = basis.calls[-1]
2576
self.assertEqual('get_record_stream', last_call[0])
2577
self.assertEqual(set([key_left, key_right]), set(last_call[1]))
2578
self.assertEqual('topological', last_call[2])
2579
self.assertEqual(True, last_call[3])
2582
class TestNetworkBehaviour(KnitTests):
2583
"""Tests for getting data out of/into knits over the network."""
2585
def test_include_delta_closure_generates_a_knit_delta_closure(self):
2586
vf = self.make_test_knit(name='test')
2587
# put in three texts, giving ft, delta, delta
2588
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2589
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2590
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2591
# But heuristics could interfere, so check what happened:
2592
self.assertEqual(['knit-ft-gz', 'knit-delta-gz', 'knit-delta-gz'],
2593
[record.storage_kind for record in
2594
vf.get_record_stream([('base',), ('d1',), ('d2',)],
2595
'topological', False)])
2596
# generate a stream of just the deltas include_delta_closure=True,
2597
# serialise to the network, and check that we get a delta closure on the wire.
2598
stream = vf.get_record_stream([('d1',), ('d2',)], 'topological', True)
2599
netb = [record.get_bytes_as(record.storage_kind) for record in stream]
2600
# The first bytes should be a memo from _ContentMapGenerator, and the
2601
# second bytes should be empty (because its a API proxy not something
2602
# for wire serialisation.
2603
self.assertEqual('', netb[1])
2605
kind, line_end = network_bytes_to_kind_and_offset(bytes)
2606
self.assertEqual('knit-delta-closure', kind)
2609
class TestContentMapGenerator(KnitTests):
2610
"""Tests for ContentMapGenerator"""
2612
def test_get_record_stream_gives_records(self):
2613
vf = self.make_test_knit(name='test')
2614
# put in three texts, giving ft, delta, delta
2615
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2616
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2617
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2618
keys = [('d1',), ('d2',)]
2619
generator = _VFContentMapGenerator(vf, keys,
2620
global_map=vf.get_parent_map(keys))
2621
for record in generator.get_record_stream():
2622
if record.key == ('d1',):
2623
self.assertEqual('d1\n', record.get_bytes_as('fulltext'))
2625
self.assertEqual('d2\n', record.get_bytes_as('fulltext'))
2627
def test_get_record_stream_kinds_are_raw(self):
2628
vf = self.make_test_knit(name='test')
2629
# put in three texts, giving ft, delta, delta
2630
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2631
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2632
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2633
keys = [('base',), ('d1',), ('d2',)]
2634
generator = _VFContentMapGenerator(vf, keys,
2635
global_map=vf.get_parent_map(keys))
2636
kinds = {('base',): 'knit-delta-closure',
2637
('d1',): 'knit-delta-closure-ref',
2638
('d2',): 'knit-delta-closure-ref',
2640
for record in generator.get_record_stream():
2641
self.assertEqual(kinds[record.key], record.storage_kind)
added
removed
bzrlib/tests/test_knit.py
