1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
"""Tests for Knit data structure"""
19
from cStringIO import StringIO
33
from bzrlib.errors import (
34
RevisionAlreadyPresent,
39
from bzrlib.index import *
40
from bzrlib.knit import (
46
_VFContentMapGenerator,
53
from bzrlib.repofmt import pack_repo
54
from bzrlib.tests import (
58
TestCaseWithMemoryTransport,
59
TestCaseWithTransport,
62
from bzrlib.transport import get_transport
63
from bzrlib.transport.memory import MemoryTransport
64
from bzrlib.tuned_gzip import GzipFile
65
from bzrlib.versionedfile import (
68
network_bytes_to_kind_and_offset,
69
RecordingVersionedFilesDecorator,
73
compiled_knit_feature = tests.ModuleAvailableFeature(
74
'bzrlib._knit_load_data_pyx')
77
class KnitContentTestsMixin(object):
79
def test_constructor(self):
80
content = self._make_content([])
83
content = self._make_content([])
84
self.assertEqual(content.text(), [])
86
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
87
self.assertEqual(content.text(), ["text1", "text2"])
90
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
92
self.assertIsInstance(copy, content.__class__)
93
self.assertEqual(copy.annotate(), content.annotate())
95
def assertDerivedBlocksEqual(self, source, target, noeol=False):
96
"""Assert that the derived matching blocks match real output"""
97
source_lines = source.splitlines(True)
98
target_lines = target.splitlines(True)
100
if noeol and not line.endswith('\n'):
104
source_content = self._make_content([(None, nl(l)) for l in source_lines])
105
target_content = self._make_content([(None, nl(l)) for l in target_lines])
106
line_delta = source_content.line_delta(target_content)
107
delta_blocks = list(KnitContent.get_line_delta_blocks(line_delta,
108
source_lines, target_lines))
109
matcher = KnitSequenceMatcher(None, source_lines, target_lines)
110
matcher_blocks = list(list(matcher.get_matching_blocks()))
111
self.assertEqual(matcher_blocks, delta_blocks)
113
def test_get_line_delta_blocks(self):
114
self.assertDerivedBlocksEqual('a\nb\nc\n', 'q\nc\n')
115
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1)
116
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1A)
117
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1B)
118
self.assertDerivedBlocksEqual(TEXT_1B, TEXT_1A)
119
self.assertDerivedBlocksEqual(TEXT_1A, TEXT_1B)
120
self.assertDerivedBlocksEqual(TEXT_1A, '')
121
self.assertDerivedBlocksEqual('', TEXT_1A)
122
self.assertDerivedBlocksEqual('', '')
123
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd')
125
def test_get_line_delta_blocks_noeol(self):
126
"""Handle historical knit deltas safely
128
Some existing knit deltas don't consider the last line to differ
129
when the only difference whether it has a final newline.
131
New knit deltas appear to always consider the last line to differ
134
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd\n', noeol=True)
135
self.assertDerivedBlocksEqual('a\nb\nc\nd\n', 'a\nb\nc', noeol=True)
136
self.assertDerivedBlocksEqual('a\nb\nc\n', 'a\nb\nc', noeol=True)
137
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\n', noeol=True)
149
Banana cup cake recipe
159
Banana cup cake recipe
161
- bananas (do not use plantains!!!)
168
Banana cup cake recipe
185
class TestPlainKnitContent(TestCase, KnitContentTestsMixin):
187
def _make_content(self, lines):
188
annotated_content = AnnotatedKnitContent(lines)
189
return PlainKnitContent(annotated_content.text(), 'bogus')
191
def test_annotate(self):
192
content = self._make_content([])
193
self.assertEqual(content.annotate(), [])
195
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
196
self.assertEqual(content.annotate(),
197
[("bogus", "text1"), ("bogus", "text2")])
199
def test_line_delta(self):
200
content1 = self._make_content([("", "a"), ("", "b")])
201
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
202
self.assertEqual(content1.line_delta(content2),
203
[(1, 2, 2, ["a", "c"])])
205
def test_line_delta_iter(self):
206
content1 = self._make_content([("", "a"), ("", "b")])
207
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
208
it = content1.line_delta_iter(content2)
209
self.assertEqual(it.next(), (1, 2, 2, ["a", "c"]))
210
self.assertRaises(StopIteration, it.next)
213
class TestAnnotatedKnitContent(TestCase, KnitContentTestsMixin):
215
def _make_content(self, lines):
216
return AnnotatedKnitContent(lines)
218
def test_annotate(self):
219
content = self._make_content([])
220
self.assertEqual(content.annotate(), [])
222
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
223
self.assertEqual(content.annotate(),
224
[("origin1", "text1"), ("origin2", "text2")])
226
def test_line_delta(self):
227
content1 = self._make_content([("", "a"), ("", "b")])
228
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
229
self.assertEqual(content1.line_delta(content2),
230
[(1, 2, 2, [("", "a"), ("", "c")])])
232
def test_line_delta_iter(self):
233
content1 = self._make_content([("", "a"), ("", "b")])
234
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
235
it = content1.line_delta_iter(content2)
236
self.assertEqual(it.next(), (1, 2, 2, [("", "a"), ("", "c")]))
237
self.assertRaises(StopIteration, it.next)
240
class MockTransport(object):
242
def __init__(self, file_lines=None):
243
self.file_lines = file_lines
245
# We have no base directory for the MockTransport
248
def get(self, filename):
249
if self.file_lines is None:
250
raise NoSuchFile(filename)
252
return StringIO("\n".join(self.file_lines))
254
def readv(self, relpath, offsets):
255
fp = self.get(relpath)
256
for offset, size in offsets:
258
yield offset, fp.read(size)
260
def __getattr__(self, name):
261
def queue_call(*args, **kwargs):
262
self.calls.append((name, args, kwargs))
266
class MockReadvFailingTransport(MockTransport):
267
"""Fail in the middle of a readv() result.
269
This Transport will successfully yield the first two requested hunks, but
270
raise NoSuchFile for the rest.
273
def readv(self, relpath, offsets):
275
for result in MockTransport.readv(self, relpath, offsets):
277
# we use 2 because the first offset is the pack header, the second
278
# is the first actual content requset
280
raise errors.NoSuchFile(relpath)
284
class KnitRecordAccessTestsMixin(object):
285
"""Tests for getting and putting knit records."""
287
def test_add_raw_records(self):
288
"""Add_raw_records adds records retrievable later."""
289
access = self.get_access()
290
memos = access.add_raw_records([('key', 10)], '1234567890')
291
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
293
def test_add_several_raw_records(self):
294
"""add_raw_records with many records and read some back."""
295
access = self.get_access()
296
memos = access.add_raw_records([('key', 10), ('key2', 2), ('key3', 5)],
298
self.assertEqual(['1234567890', '12', '34567'],
299
list(access.get_raw_records(memos)))
300
self.assertEqual(['1234567890'],
301
list(access.get_raw_records(memos[0:1])))
302
self.assertEqual(['12'],
303
list(access.get_raw_records(memos[1:2])))
304
self.assertEqual(['34567'],
305
list(access.get_raw_records(memos[2:3])))
306
self.assertEqual(['1234567890', '34567'],
307
list(access.get_raw_records(memos[0:1] + memos[2:3])))
310
class TestKnitKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
311
"""Tests for the .kndx implementation."""
313
def get_access(self):
314
"""Get a .knit style access instance."""
315
mapper = ConstantMapper("foo")
316
access = _KnitKeyAccess(self.get_transport(), mapper)
320
class _TestException(Exception):
321
"""Just an exception for local tests to use."""
324
class TestPackKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
325
"""Tests for the pack based access."""
327
def get_access(self):
328
return self._get_access()[0]
330
def _get_access(self, packname='packfile', index='FOO'):
331
transport = self.get_transport()
332
def write_data(bytes):
333
transport.append_bytes(packname, bytes)
334
writer = pack.ContainerWriter(write_data)
336
access = _DirectPackAccess({})
337
access.set_writer(writer, index, (transport, packname))
338
return access, writer
340
def make_pack_file(self):
341
"""Create a pack file with 2 records."""
342
access, writer = self._get_access(packname='packname', index='foo')
344
memos.extend(access.add_raw_records([('key1', 10)], '1234567890'))
345
memos.extend(access.add_raw_records([('key2', 5)], '12345'))
349
def make_vf_for_retrying(self):
350
"""Create 3 packs and a reload function.
352
Originally, 2 pack files will have the data, but one will be missing.
353
And then the third will be used in place of the first two if reload()
356
:return: (versioned_file, reload_counter)
357
versioned_file a KnitVersionedFiles using the packs for access
359
builder = self.make_branch_builder('.', format="1.9")
360
builder.start_series()
361
builder.build_snapshot('rev-1', None, [
362
('add', ('', 'root-id', 'directory', None)),
363
('add', ('file', 'file-id', 'file', 'content\nrev 1\n')),
365
builder.build_snapshot('rev-2', ['rev-1'], [
366
('modify', ('file-id', 'content\nrev 2\n')),
368
builder.build_snapshot('rev-3', ['rev-2'], [
369
('modify', ('file-id', 'content\nrev 3\n')),
371
builder.finish_series()
372
b = builder.get_branch()
374
self.addCleanup(b.unlock)
375
# Pack these three revisions into another pack file, but don't remove
378
collection = repo._pack_collection
379
collection.ensure_loaded()
380
orig_packs = collection.packs
381
packer = pack_repo.Packer(collection, orig_packs, '.testpack')
382
new_pack = packer.pack()
383
# forget about the new pack
387
# Set up a reload() function that switches to using the new pack file
388
new_index = new_pack.revision_index
389
access_tuple = new_pack.access_tuple()
390
reload_counter = [0, 0, 0]
392
reload_counter[0] += 1
393
if reload_counter[1] > 0:
394
# We already reloaded, nothing more to do
395
reload_counter[2] += 1
397
reload_counter[1] += 1
398
vf._index._graph_index._indices[:] = [new_index]
399
vf._access._indices.clear()
400
vf._access._indices[new_index] = access_tuple
402
# Delete one of the pack files so the data will need to be reloaded. We
403
# will delete the file with 'rev-2' in it
404
trans, name = orig_packs[1].access_tuple()
406
# We don't have the index trigger reloading because we want to test
407
# that we reload when the .pack disappears
408
vf._access._reload_func = reload
409
return vf, reload_counter
411
def make_reload_func(self, return_val=True):
414
reload_called[0] += 1
416
return reload_called, reload
418
def make_retry_exception(self):
419
# We raise a real exception so that sys.exc_info() is properly
422
raise _TestException('foobar')
423
except _TestException, e:
424
retry_exc = errors.RetryWithNewPacks(None, reload_occurred=False,
425
exc_info=sys.exc_info())
428
def test_read_from_several_packs(self):
429
access, writer = self._get_access()
431
memos.extend(access.add_raw_records([('key', 10)], '1234567890'))
433
access, writer = self._get_access('pack2', 'FOOBAR')
434
memos.extend(access.add_raw_records([('key', 5)], '12345'))
436
access, writer = self._get_access('pack3', 'BAZ')
437
memos.extend(access.add_raw_records([('key', 5)], 'alpha'))
439
transport = self.get_transport()
440
access = _DirectPackAccess({"FOO":(transport, 'packfile'),
441
"FOOBAR":(transport, 'pack2'),
442
"BAZ":(transport, 'pack3')})
443
self.assertEqual(['1234567890', '12345', 'alpha'],
444
list(access.get_raw_records(memos)))
445
self.assertEqual(['1234567890'],
446
list(access.get_raw_records(memos[0:1])))
447
self.assertEqual(['12345'],
448
list(access.get_raw_records(memos[1:2])))
449
self.assertEqual(['alpha'],
450
list(access.get_raw_records(memos[2:3])))
451
self.assertEqual(['1234567890', 'alpha'],
452
list(access.get_raw_records(memos[0:1] + memos[2:3])))
454
def test_set_writer(self):
455
"""The writer should be settable post construction."""
456
access = _DirectPackAccess({})
457
transport = self.get_transport()
458
packname = 'packfile'
460
def write_data(bytes):
461
transport.append_bytes(packname, bytes)
462
writer = pack.ContainerWriter(write_data)
464
access.set_writer(writer, index, (transport, packname))
465
memos = access.add_raw_records([('key', 10)], '1234567890')
467
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
469
def test_missing_index_raises_retry(self):
470
memos = self.make_pack_file()
471
transport = self.get_transport()
472
reload_called, reload_func = self.make_reload_func()
473
# Note that the index key has changed from 'foo' to 'bar'
474
access = _DirectPackAccess({'bar':(transport, 'packname')},
475
reload_func=reload_func)
476
e = self.assertListRaises(errors.RetryWithNewPacks,
477
access.get_raw_records, memos)
478
# Because a key was passed in which does not match our index list, we
479
# assume that the listing was already reloaded
480
self.assertTrue(e.reload_occurred)
481
self.assertIsInstance(e.exc_info, tuple)
482
self.assertIs(e.exc_info[0], KeyError)
483
self.assertIsInstance(e.exc_info[1], KeyError)
485
def test_missing_index_raises_key_error_with_no_reload(self):
486
memos = self.make_pack_file()
487
transport = self.get_transport()
488
# Note that the index key has changed from 'foo' to 'bar'
489
access = _DirectPackAccess({'bar':(transport, 'packname')})
490
e = self.assertListRaises(KeyError, access.get_raw_records, memos)
492
def test_missing_file_raises_retry(self):
493
memos = self.make_pack_file()
494
transport = self.get_transport()
495
reload_called, reload_func = self.make_reload_func()
496
# Note that the 'filename' has been changed to 'different-packname'
497
access = _DirectPackAccess({'foo':(transport, 'different-packname')},
498
reload_func=reload_func)
499
e = self.assertListRaises(errors.RetryWithNewPacks,
500
access.get_raw_records, memos)
501
# The file has gone missing, so we assume we need to reload
502
self.assertFalse(e.reload_occurred)
503
self.assertIsInstance(e.exc_info, tuple)
504
self.assertIs(e.exc_info[0], errors.NoSuchFile)
505
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
506
self.assertEqual('different-packname', e.exc_info[1].path)
508
def test_missing_file_raises_no_such_file_with_no_reload(self):
509
memos = self.make_pack_file()
510
transport = self.get_transport()
511
# Note that the 'filename' has been changed to 'different-packname'
512
access = _DirectPackAccess({'foo':(transport, 'different-packname')})
513
e = self.assertListRaises(errors.NoSuchFile,
514
access.get_raw_records, memos)
516
def test_failing_readv_raises_retry(self):
517
memos = self.make_pack_file()
518
transport = self.get_transport()
519
failing_transport = MockReadvFailingTransport(
520
[transport.get_bytes('packname')])
521
reload_called, reload_func = self.make_reload_func()
522
access = _DirectPackAccess({'foo':(failing_transport, 'packname')},
523
reload_func=reload_func)
524
# Asking for a single record will not trigger the Mock failure
525
self.assertEqual(['1234567890'],
526
list(access.get_raw_records(memos[:1])))
527
self.assertEqual(['12345'],
528
list(access.get_raw_records(memos[1:2])))
529
# A multiple offset readv() will fail mid-way through
530
e = self.assertListRaises(errors.RetryWithNewPacks,
531
access.get_raw_records, memos)
532
# The file has gone missing, so we assume we need to reload
533
self.assertFalse(e.reload_occurred)
534
self.assertIsInstance(e.exc_info, tuple)
535
self.assertIs(e.exc_info[0], errors.NoSuchFile)
536
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
537
self.assertEqual('packname', e.exc_info[1].path)
539
def test_failing_readv_raises_no_such_file_with_no_reload(self):
540
memos = self.make_pack_file()
541
transport = self.get_transport()
542
failing_transport = MockReadvFailingTransport(
543
[transport.get_bytes('packname')])
544
reload_called, reload_func = self.make_reload_func()
545
access = _DirectPackAccess({'foo':(failing_transport, 'packname')})
546
# Asking for a single record will not trigger the Mock failure
547
self.assertEqual(['1234567890'],
548
list(access.get_raw_records(memos[:1])))
549
self.assertEqual(['12345'],
550
list(access.get_raw_records(memos[1:2])))
551
# A multiple offset readv() will fail mid-way through
552
e = self.assertListRaises(errors.NoSuchFile,
553
access.get_raw_records, memos)
555
def test_reload_or_raise_no_reload(self):
556
access = _DirectPackAccess({}, reload_func=None)
557
retry_exc = self.make_retry_exception()
558
# Without a reload_func, we will just re-raise the original exception
559
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
561
def test_reload_or_raise_reload_changed(self):
562
reload_called, reload_func = self.make_reload_func(return_val=True)
563
access = _DirectPackAccess({}, reload_func=reload_func)
564
retry_exc = self.make_retry_exception()
565
access.reload_or_raise(retry_exc)
566
self.assertEqual([1], reload_called)
567
retry_exc.reload_occurred=True
568
access.reload_or_raise(retry_exc)
569
self.assertEqual([2], reload_called)
571
def test_reload_or_raise_reload_no_change(self):
572
reload_called, reload_func = self.make_reload_func(return_val=False)
573
access = _DirectPackAccess({}, reload_func=reload_func)
574
retry_exc = self.make_retry_exception()
575
# If reload_occurred is False, then we consider it an error to have
576
# reload_func() return False (no changes).
577
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
578
self.assertEqual([1], reload_called)
579
retry_exc.reload_occurred=True
580
# If reload_occurred is True, then we assume nothing changed because
581
# it had changed earlier, but didn't change again
582
access.reload_or_raise(retry_exc)
583
self.assertEqual([2], reload_called)
585
def test_annotate_retries(self):
586
vf, reload_counter = self.make_vf_for_retrying()
587
# It is a little bit bogus to annotate the Revision VF, but it works,
588
# as we have ancestry stored there
590
reload_lines = vf.annotate(key)
591
self.assertEqual([1, 1, 0], reload_counter)
592
plain_lines = vf.annotate(key)
593
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
594
if reload_lines != plain_lines:
595
self.fail('Annotation was not identical with reloading.')
596
# Now delete the packs-in-use, which should trigger another reload, but
597
# this time we just raise an exception because we can't recover
598
for trans, name in vf._access._indices.itervalues():
600
self.assertRaises(errors.NoSuchFile, vf.annotate, key)
601
self.assertEqual([2, 1, 1], reload_counter)
603
def test__get_record_map_retries(self):
604
vf, reload_counter = self.make_vf_for_retrying()
605
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
606
records = vf._get_record_map(keys)
607
self.assertEqual(keys, sorted(records.keys()))
608
self.assertEqual([1, 1, 0], reload_counter)
609
# Now delete the packs-in-use, which should trigger another reload, but
610
# this time we just raise an exception because we can't recover
611
for trans, name in vf._access._indices.itervalues():
613
self.assertRaises(errors.NoSuchFile, vf._get_record_map, keys)
614
self.assertEqual([2, 1, 1], reload_counter)
616
def test_get_record_stream_retries(self):
617
vf, reload_counter = self.make_vf_for_retrying()
618
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
619
record_stream = vf.get_record_stream(keys, 'topological', False)
620
record = record_stream.next()
621
self.assertEqual(('rev-1',), record.key)
622
self.assertEqual([0, 0, 0], reload_counter)
623
record = record_stream.next()
624
self.assertEqual(('rev-2',), record.key)
625
self.assertEqual([1, 1, 0], reload_counter)
626
record = record_stream.next()
627
self.assertEqual(('rev-3',), record.key)
628
self.assertEqual([1, 1, 0], reload_counter)
629
# Now delete all pack files, and see that we raise the right error
630
for trans, name in vf._access._indices.itervalues():
632
self.assertListRaises(errors.NoSuchFile,
633
vf.get_record_stream, keys, 'topological', False)
635
def test_iter_lines_added_or_present_in_keys_retries(self):
636
vf, reload_counter = self.make_vf_for_retrying()
637
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
638
# Unfortunately, iter_lines_added_or_present_in_keys iterates the
639
# result in random order (determined by the iteration order from a
640
# set()), so we don't have any solid way to trigger whether data is
641
# read before or after. However we tried to delete the middle node to
642
# exercise the code well.
643
# What we care about is that all lines are always yielded, but not
646
reload_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
647
self.assertEqual([1, 1, 0], reload_counter)
648
# Now do it again, to make sure the result is equivalent
649
plain_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
650
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
651
self.assertEqual(plain_lines, reload_lines)
652
self.assertEqual(21, len(plain_lines))
653
# Now delete all pack files, and see that we raise the right error
654
for trans, name in vf._access._indices.itervalues():
656
self.assertListRaises(errors.NoSuchFile,
657
vf.iter_lines_added_or_present_in_keys, keys)
658
self.assertEqual([2, 1, 1], reload_counter)
660
def test_get_record_stream_yields_disk_sorted_order(self):
661
# if we get 'unordered' pick a semi-optimal order for reading. The
662
# order should be grouped by pack file, and then by position in file
663
repo = self.make_repository('test', format='pack-0.92')
665
self.addCleanup(repo.unlock)
666
repo.start_write_group()
668
vf.add_lines(('f-id', 'rev-5'), [('f-id', 'rev-4')], ['lines\n'])
669
vf.add_lines(('f-id', 'rev-1'), [], ['lines\n'])
670
vf.add_lines(('f-id', 'rev-2'), [('f-id', 'rev-1')], ['lines\n'])
671
repo.commit_write_group()
672
# We inserted them as rev-5, rev-1, rev-2, we should get them back in
674
stream = vf.get_record_stream([('f-id', 'rev-1'), ('f-id', 'rev-5'),
675
('f-id', 'rev-2')], 'unordered', False)
676
keys = [r.key for r in stream]
677
self.assertEqual([('f-id', 'rev-5'), ('f-id', 'rev-1'),
678
('f-id', 'rev-2')], keys)
679
repo.start_write_group()
680
vf.add_lines(('f-id', 'rev-4'), [('f-id', 'rev-3')], ['lines\n'])
681
vf.add_lines(('f-id', 'rev-3'), [('f-id', 'rev-2')], ['lines\n'])
682
vf.add_lines(('f-id', 'rev-6'), [('f-id', 'rev-5')], ['lines\n'])
683
repo.commit_write_group()
684
# Request in random order, to make sure the output order isn't based on
686
request_keys = set(('f-id', 'rev-%d' % i) for i in range(1, 7))
687
stream = vf.get_record_stream(request_keys, 'unordered', False)
688
keys = [r.key for r in stream]
689
# We want to get the keys back in disk order, but it doesn't matter
690
# which pack we read from first. So this can come back in 2 orders
691
alt1 = [('f-id', 'rev-%d' % i) for i in [4, 3, 6, 5, 1, 2]]
692
alt2 = [('f-id', 'rev-%d' % i) for i in [5, 1, 2, 4, 3, 6]]
693
if keys != alt1 and keys != alt2:
694
self.fail('Returned key order did not match either expected order.'
695
' expected %s or %s, not %s'
696
% (alt1, alt2, keys))
699
class LowLevelKnitDataTests(TestCase):
701
def create_gz_content(self, text):
703
gz_file = gzip.GzipFile(mode='wb', fileobj=sio)
706
return sio.getvalue()
708
def make_multiple_records(self):
709
"""Create the content for multiple records."""
710
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
712
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
717
record_1 = (0, len(gz_txt), sha1sum)
718
total_txt.append(gz_txt)
719
sha1sum = osutils.sha('baz\n').hexdigest()
720
gz_txt = self.create_gz_content('version rev-id-2 1 %s\n'
724
record_2 = (record_1[1], len(gz_txt), sha1sum)
725
total_txt.append(gz_txt)
726
return total_txt, record_1, record_2
728
def test_valid_knit_data(self):
729
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
730
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
735
transport = MockTransport([gz_txt])
736
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
737
knit = KnitVersionedFiles(None, access)
738
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
740
contents = list(knit._read_records_iter(records))
741
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'],
742
'4e48e2c9a3d2ca8a708cb0cc545700544efb5021')], contents)
744
raw_contents = list(knit._read_records_iter_raw(records))
745
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
747
def test_multiple_records_valid(self):
748
total_txt, record_1, record_2 = self.make_multiple_records()
749
transport = MockTransport([''.join(total_txt)])
750
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
751
knit = KnitVersionedFiles(None, access)
752
records = [(('rev-id-1',), (('rev-id-1',), record_1[0], record_1[1])),
753
(('rev-id-2',), (('rev-id-2',), record_2[0], record_2[1]))]
755
contents = list(knit._read_records_iter(records))
756
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'], record_1[2]),
757
(('rev-id-2',), ['baz\n'], record_2[2])],
760
raw_contents = list(knit._read_records_iter_raw(records))
761
self.assertEqual([(('rev-id-1',), total_txt[0], record_1[2]),
762
(('rev-id-2',), total_txt[1], record_2[2])],
765
def test_not_enough_lines(self):
766
sha1sum = osutils.sha('foo\n').hexdigest()
767
# record says 2 lines data says 1
768
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
772
transport = MockTransport([gz_txt])
773
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
774
knit = KnitVersionedFiles(None, access)
775
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
776
self.assertRaises(errors.KnitCorrupt, list,
777
knit._read_records_iter(records))
779
# read_records_iter_raw won't detect that sort of mismatch/corruption
780
raw_contents = list(knit._read_records_iter_raw(records))
781
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
783
def test_too_many_lines(self):
784
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
785
# record says 1 lines data says 2
786
gz_txt = self.create_gz_content('version rev-id-1 1 %s\n'
791
transport = MockTransport([gz_txt])
792
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
793
knit = KnitVersionedFiles(None, access)
794
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
795
self.assertRaises(errors.KnitCorrupt, list,
796
knit._read_records_iter(records))
798
# read_records_iter_raw won't detect that sort of mismatch/corruption
799
raw_contents = list(knit._read_records_iter_raw(records))
800
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
802
def test_mismatched_version_id(self):
803
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
804
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
809
transport = MockTransport([gz_txt])
810
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
811
knit = KnitVersionedFiles(None, access)
812
# We are asking for rev-id-2, but the data is rev-id-1
813
records = [(('rev-id-2',), (('rev-id-2',), 0, len(gz_txt)))]
814
self.assertRaises(errors.KnitCorrupt, list,
815
knit._read_records_iter(records))
817
# read_records_iter_raw detects mismatches in the header
818
self.assertRaises(errors.KnitCorrupt, list,
819
knit._read_records_iter_raw(records))
821
def test_uncompressed_data(self):
822
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
823
txt = ('version rev-id-1 2 %s\n'
828
transport = MockTransport([txt])
829
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
830
knit = KnitVersionedFiles(None, access)
831
records = [(('rev-id-1',), (('rev-id-1',), 0, len(txt)))]
833
# We don't have valid gzip data ==> corrupt
834
self.assertRaises(errors.KnitCorrupt, list,
835
knit._read_records_iter(records))
837
# read_records_iter_raw will notice the bad data
838
self.assertRaises(errors.KnitCorrupt, list,
839
knit._read_records_iter_raw(records))
841
def test_corrupted_data(self):
842
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
843
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
848
# Change 2 bytes in the middle to \xff
849
gz_txt = gz_txt[:10] + '\xff\xff' + gz_txt[12:]
850
transport = MockTransport([gz_txt])
851
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
852
knit = KnitVersionedFiles(None, access)
853
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
854
self.assertRaises(errors.KnitCorrupt, list,
855
knit._read_records_iter(records))
856
# read_records_iter_raw will barf on bad gz data
857
self.assertRaises(errors.KnitCorrupt, list,
858
knit._read_records_iter_raw(records))
861
class LowLevelKnitIndexTests(TestCase):
863
def get_knit_index(self, transport, name, mode):
864
mapper = ConstantMapper(name)
865
orig = knit._load_data
867
knit._load_data = orig
868
self.addCleanup(reset)
869
from bzrlib._knit_load_data_py import _load_data_py
870
knit._load_data = _load_data_py
871
allow_writes = lambda: 'w' in mode
872
return _KndxIndex(transport, mapper, lambda:None, allow_writes, lambda:True)
874
def test_create_file(self):
875
transport = MockTransport()
876
index = self.get_knit_index(transport, "filename", "w")
878
call = transport.calls.pop(0)
879
# call[1][1] is a StringIO - we can't test it by simple equality.
880
self.assertEqual('put_file_non_atomic', call[0])
881
self.assertEqual('filename.kndx', call[1][0])
882
# With no history, _KndxIndex writes a new index:
883
self.assertEqual(_KndxIndex.HEADER,
884
call[1][1].getvalue())
885
self.assertEqual({'create_parent_dir': True}, call[2])
887
def test_read_utf8_version_id(self):
888
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
889
utf8_revision_id = unicode_revision_id.encode('utf-8')
890
transport = MockTransport([
892
'%s option 0 1 :' % (utf8_revision_id,)
894
index = self.get_knit_index(transport, "filename", "r")
895
# _KndxIndex is a private class, and deals in utf8 revision_ids, not
896
# Unicode revision_ids.
897
self.assertEqual({(utf8_revision_id,):()},
898
index.get_parent_map(index.keys()))
899
self.assertFalse((unicode_revision_id,) in index.keys())
901
def test_read_utf8_parents(self):
902
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
903
utf8_revision_id = unicode_revision_id.encode('utf-8')
904
transport = MockTransport([
906
"version option 0 1 .%s :" % (utf8_revision_id,)
908
index = self.get_knit_index(transport, "filename", "r")
909
self.assertEqual({("version",):((utf8_revision_id,),)},
910
index.get_parent_map(index.keys()))
912
def test_read_ignore_corrupted_lines(self):
913
transport = MockTransport([
916
"corrupted options 0 1 .b .c ",
917
"version options 0 1 :"
919
index = self.get_knit_index(transport, "filename", "r")
920
self.assertEqual(1, len(index.keys()))
921
self.assertEqual(set([("version",)]), index.keys())
923
def test_read_corrupted_header(self):
924
transport = MockTransport(['not a bzr knit index header\n'])
925
index = self.get_knit_index(transport, "filename", "r")
926
self.assertRaises(KnitHeaderError, index.keys)
928
def test_read_duplicate_entries(self):
929
transport = MockTransport([
931
"parent options 0 1 :",
932
"version options1 0 1 0 :",
933
"version options2 1 2 .other :",
934
"version options3 3 4 0 .other :"
936
index = self.get_knit_index(transport, "filename", "r")
937
self.assertEqual(2, len(index.keys()))
938
# check that the index used is the first one written. (Specific
939
# to KnitIndex style indices.
940
self.assertEqual("1", index._dictionary_compress([("version",)]))
941
self.assertEqual((("version",), 3, 4), index.get_position(("version",)))
942
self.assertEqual(["options3"], index.get_options(("version",)))
943
self.assertEqual({("version",):(("parent",), ("other",))},
944
index.get_parent_map([("version",)]))
946
def test_read_compressed_parents(self):
947
transport = MockTransport([
951
"c option 0 1 1 0 :",
953
index = self.get_knit_index(transport, "filename", "r")
954
self.assertEqual({("b",):(("a",),), ("c",):(("b",), ("a",))},
955
index.get_parent_map([("b",), ("c",)]))
957
def test_write_utf8_version_id(self):
958
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
959
utf8_revision_id = unicode_revision_id.encode('utf-8')
960
transport = MockTransport([
963
index = self.get_knit_index(transport, "filename", "r")
965
((utf8_revision_id,), ["option"], ((utf8_revision_id,), 0, 1), [])])
966
call = transport.calls.pop(0)
967
# call[1][1] is a StringIO - we can't test it by simple equality.
968
self.assertEqual('put_file_non_atomic', call[0])
969
self.assertEqual('filename.kndx', call[1][0])
970
# With no history, _KndxIndex writes a new index:
971
self.assertEqual(_KndxIndex.HEADER +
972
"\n%s option 0 1 :" % (utf8_revision_id,),
973
call[1][1].getvalue())
974
self.assertEqual({'create_parent_dir': True}, call[2])
976
def test_write_utf8_parents(self):
977
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
978
utf8_revision_id = unicode_revision_id.encode('utf-8')
979
transport = MockTransport([
982
index = self.get_knit_index(transport, "filename", "r")
984
(("version",), ["option"], (("version",), 0, 1), [(utf8_revision_id,)])])
985
call = transport.calls.pop(0)
986
# call[1][1] is a StringIO - we can't test it by simple equality.
987
self.assertEqual('put_file_non_atomic', call[0])
988
self.assertEqual('filename.kndx', call[1][0])
989
# With no history, _KndxIndex writes a new index:
990
self.assertEqual(_KndxIndex.HEADER +
991
"\nversion option 0 1 .%s :" % (utf8_revision_id,),
992
call[1][1].getvalue())
993
self.assertEqual({'create_parent_dir': True}, call[2])
996
transport = MockTransport([
999
index = self.get_knit_index(transport, "filename", "r")
1001
self.assertEqual(set(), index.keys())
1003
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1004
self.assertEqual(set([("a",)]), index.keys())
1006
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1007
self.assertEqual(set([("a",)]), index.keys())
1009
index.add_records([(("b",), ["option"], (("b",), 0, 1), [])])
1010
self.assertEqual(set([("a",), ("b",)]), index.keys())
1012
def add_a_b(self, index, random_id=None):
1014
if random_id is not None:
1015
kwargs["random_id"] = random_id
1017
(("a",), ["option"], (("a",), 0, 1), [("b",)]),
1018
(("a",), ["opt"], (("a",), 1, 2), [("c",)]),
1019
(("b",), ["option"], (("b",), 2, 3), [("a",)])
1022
def assertIndexIsAB(self, index):
1027
index.get_parent_map(index.keys()))
1028
self.assertEqual((("a",), 1, 2), index.get_position(("a",)))
1029
self.assertEqual((("b",), 2, 3), index.get_position(("b",)))
1030
self.assertEqual(["opt"], index.get_options(("a",)))
1032
def test_add_versions(self):
1033
transport = MockTransport([
1036
index = self.get_knit_index(transport, "filename", "r")
1039
call = transport.calls.pop(0)
1040
# call[1][1] is a StringIO - we can't test it by simple equality.
1041
self.assertEqual('put_file_non_atomic', call[0])
1042
self.assertEqual('filename.kndx', call[1][0])
1043
# With no history, _KndxIndex writes a new index:
1046
"\na option 0 1 .b :"
1048
"\nb option 2 3 0 :",
1049
call[1][1].getvalue())
1050
self.assertEqual({'create_parent_dir': True}, call[2])
1051
self.assertIndexIsAB(index)
1053
def test_add_versions_random_id_is_accepted(self):
1054
transport = MockTransport([
1057
index = self.get_knit_index(transport, "filename", "r")
1058
self.add_a_b(index, random_id=True)
1060
def test_delay_create_and_add_versions(self):
1061
transport = MockTransport()
1063
index = self.get_knit_index(transport, "filename", "w")
1065
self.assertEqual([], transport.calls)
1068
#[ {"dir_mode": 0777, "create_parent_dir": True, "mode": "wb"},
1070
# Two calls: one during which we load the existing index (and when its
1071
# missing create it), then a second where we write the contents out.
1072
self.assertEqual(2, len(transport.calls))
1073
call = transport.calls.pop(0)
1074
self.assertEqual('put_file_non_atomic', call[0])
1075
self.assertEqual('filename.kndx', call[1][0])
1076
# With no history, _KndxIndex writes a new index:
1077
self.assertEqual(_KndxIndex.HEADER, call[1][1].getvalue())
1078
self.assertEqual({'create_parent_dir': True}, call[2])
1079
call = transport.calls.pop(0)
1080
# call[1][1] is a StringIO - we can't test it by simple equality.
1081
self.assertEqual('put_file_non_atomic', call[0])
1082
self.assertEqual('filename.kndx', call[1][0])
1083
# With no history, _KndxIndex writes a new index:
1086
"\na option 0 1 .b :"
1088
"\nb option 2 3 0 :",
1089
call[1][1].getvalue())
1090
self.assertEqual({'create_parent_dir': True}, call[2])
1092
def assertTotalBuildSize(self, size, keys, positions):
1093
self.assertEqual(size,
1094
knit._get_total_build_size(None, keys, positions))
1096
def test__get_total_build_size(self):
1098
('a',): (('fulltext', False), (('a',), 0, 100), None),
1099
('b',): (('line-delta', False), (('b',), 100, 21), ('a',)),
1100
('c',): (('line-delta', False), (('c',), 121, 35), ('b',)),
1101
('d',): (('line-delta', False), (('d',), 156, 12), ('b',)),
1103
self.assertTotalBuildSize(100, [('a',)], positions)
1104
self.assertTotalBuildSize(121, [('b',)], positions)
1105
# c needs both a & b
1106
self.assertTotalBuildSize(156, [('c',)], positions)
1107
# we shouldn't count 'b' twice
1108
self.assertTotalBuildSize(156, [('b',), ('c',)], positions)
1109
self.assertTotalBuildSize(133, [('d',)], positions)
1110
self.assertTotalBuildSize(168, [('c',), ('d',)], positions)
1112
def test_get_position(self):
1113
transport = MockTransport([
1118
index = self.get_knit_index(transport, "filename", "r")
1120
self.assertEqual((("a",), 0, 1), index.get_position(("a",)))
1121
self.assertEqual((("b",), 1, 2), index.get_position(("b",)))
1123
def test_get_method(self):
1124
transport = MockTransport([
1126
"a fulltext,unknown 0 1 :",
1127
"b unknown,line-delta 1 2 :",
1130
index = self.get_knit_index(transport, "filename", "r")
1132
self.assertEqual("fulltext", index.get_method("a"))
1133
self.assertEqual("line-delta", index.get_method("b"))
1134
self.assertRaises(errors.KnitIndexUnknownMethod, index.get_method, "c")
1136
def test_get_options(self):
1137
transport = MockTransport([
1142
index = self.get_knit_index(transport, "filename", "r")
1144
self.assertEqual(["opt1"], index.get_options("a"))
1145
self.assertEqual(["opt2", "opt3"], index.get_options("b"))
1147
def test_get_parent_map(self):
1148
transport = MockTransport([
1151
"b option 1 2 0 .c :",
1152
"c option 1 2 1 0 .e :"
1154
index = self.get_knit_index(transport, "filename", "r")
1158
("b",):(("a",), ("c",)),
1159
("c",):(("b",), ("a",), ("e",)),
1160
}, index.get_parent_map(index.keys()))
1162
def test_impossible_parent(self):
1163
"""Test we get KnitCorrupt if the parent couldn't possibly exist."""
1164
transport = MockTransport([
1167
"b option 0 1 4 :" # We don't have a 4th record
1169
index = self.get_knit_index(transport, 'filename', 'r')
1171
self.assertRaises(errors.KnitCorrupt, index.keys)
1172
except TypeError, e:
1173
if (str(e) == ('exceptions must be strings, classes, or instances,'
1174
' not exceptions.IndexError')
1175
and sys.version_info[0:2] >= (2,5)):
1176
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1177
' raising new style exceptions with python'
1182
def test_corrupted_parent(self):
1183
transport = MockTransport([
1187
"c option 0 1 1v :", # Can't have a parent of '1v'
1189
index = self.get_knit_index(transport, 'filename', 'r')
1191
self.assertRaises(errors.KnitCorrupt, index.keys)
1192
except TypeError, e:
1193
if (str(e) == ('exceptions must be strings, classes, or instances,'
1194
' not exceptions.ValueError')
1195
and sys.version_info[0:2] >= (2,5)):
1196
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1197
' raising new style exceptions with python'
1202
def test_corrupted_parent_in_list(self):
1203
transport = MockTransport([
1207
"c option 0 1 1 v :", # Can't have a parent of 'v'
1209
index = self.get_knit_index(transport, 'filename', 'r')
1211
self.assertRaises(errors.KnitCorrupt, index.keys)
1212
except TypeError, e:
1213
if (str(e) == ('exceptions must be strings, classes, or instances,'
1214
' not exceptions.ValueError')
1215
and sys.version_info[0:2] >= (2,5)):
1216
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1217
' raising new style exceptions with python'
1222
def test_invalid_position(self):
1223
transport = MockTransport([
1227
index = self.get_knit_index(transport, 'filename', 'r')
1229
self.assertRaises(errors.KnitCorrupt, index.keys)
1230
except TypeError, e:
1231
if (str(e) == ('exceptions must be strings, classes, or instances,'
1232
' not exceptions.ValueError')
1233
and sys.version_info[0:2] >= (2,5)):
1234
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1235
' raising new style exceptions with python'
1240
def test_invalid_size(self):
1241
transport = MockTransport([
1245
index = self.get_knit_index(transport, 'filename', 'r')
1247
self.assertRaises(errors.KnitCorrupt, index.keys)
1248
except TypeError, e:
1249
if (str(e) == ('exceptions must be strings, classes, or instances,'
1250
' not exceptions.ValueError')
1251
and sys.version_info[0:2] >= (2,5)):
1252
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1253
' raising new style exceptions with python'
1258
def test_scan_unvalidated_index_not_implemented(self):
1259
transport = MockTransport()
1260
index = self.get_knit_index(transport, 'filename', 'r')
1262
NotImplementedError, index.scan_unvalidated_index,
1263
'dummy graph_index')
1265
NotImplementedError, index.get_missing_compression_parents)
1267
def test_short_line(self):
1268
transport = MockTransport([
1271
"b option 10 10 0", # This line isn't terminated, ignored
1273
index = self.get_knit_index(transport, "filename", "r")
1274
self.assertEqual(set([('a',)]), index.keys())
1276
def test_skip_incomplete_record(self):
1277
# A line with bogus data should just be skipped
1278
transport = MockTransport([
1281
"b option 10 10 0", # This line isn't terminated, ignored
1282
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1284
index = self.get_knit_index(transport, "filename", "r")
1285
self.assertEqual(set([('a',), ('c',)]), index.keys())
1287
def test_trailing_characters(self):
1288
# A line with bogus data should just be skipped
1289
transport = MockTransport([
1292
"b option 10 10 0 :a", # This line has extra trailing characters
1293
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1295
index = self.get_knit_index(transport, "filename", "r")
1296
self.assertEqual(set([('a',), ('c',)]), index.keys())
1299
class LowLevelKnitIndexTests_c(LowLevelKnitIndexTests):
1301
_test_needs_features = [compiled_knit_feature]
1303
def get_knit_index(self, transport, name, mode):
1304
mapper = ConstantMapper(name)
1305
orig = knit._load_data
1307
knit._load_data = orig
1308
self.addCleanup(reset)
1309
from bzrlib._knit_load_data_pyx import _load_data_c
1310
knit._load_data = _load_data_c
1311
allow_writes = lambda: mode == 'w'
1312
return _KndxIndex(transport, mapper, lambda:None, allow_writes, lambda:True)
1315
class Test_KnitAnnotator(TestCaseWithMemoryTransport):
1317
def make_annotator(self):
1318
factory = knit.make_pack_factory(True, True, 1)
1319
vf = factory(self.get_transport())
1320
return knit._KnitAnnotator(vf)
1322
def test__expand_fulltext(self):
1323
ann = self.make_annotator()
1324
rev_key = ('rev-id',)
1325
ann._num_compression_children[rev_key] = 1
1326
res = ann._expand_record(rev_key, (('parent-id',),), None,
1327
['line1\n', 'line2\n'], ('fulltext', True))
1328
# The content object and text lines should be cached appropriately
1329
self.assertEqual(['line1\n', 'line2'], res)
1330
content_obj = ann._content_objects[rev_key]
1331
self.assertEqual(['line1\n', 'line2\n'], content_obj._lines)
1332
self.assertEqual(res, content_obj.text())
1333
self.assertEqual(res, ann._text_cache[rev_key])
1335
def test__expand_delta_comp_parent_not_available(self):
1336
# Parent isn't available yet, so we return nothing, but queue up this
1337
# node for later processing
1338
ann = self.make_annotator()
1339
rev_key = ('rev-id',)
1340
parent_key = ('parent-id',)
1341
record = ['0,1,1\n', 'new-line\n']
1342
details = ('line-delta', False)
1343
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1345
self.assertEqual(None, res)
1346
self.assertTrue(parent_key in ann._pending_deltas)
1347
pending = ann._pending_deltas[parent_key]
1348
self.assertEqual(1, len(pending))
1349
self.assertEqual((rev_key, (parent_key,), record, details), pending[0])
1351
def test__expand_record_tracks_num_children(self):
1352
ann = self.make_annotator()
1353
rev_key = ('rev-id',)
1354
rev2_key = ('rev2-id',)
1355
parent_key = ('parent-id',)
1356
record = ['0,1,1\n', 'new-line\n']
1357
details = ('line-delta', False)
1358
ann._num_compression_children[parent_key] = 2
1359
ann._expand_record(parent_key, (), None, ['line1\n', 'line2\n'],
1360
('fulltext', False))
1361
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1363
self.assertEqual({parent_key: 1}, ann._num_compression_children)
1364
# Expanding the second child should remove the content object, and the
1365
# num_compression_children entry
1366
res = ann._expand_record(rev2_key, (parent_key,), parent_key,
1368
self.assertFalse(parent_key in ann._content_objects)
1369
self.assertEqual({}, ann._num_compression_children)
1370
# We should not cache the content_objects for rev2 and rev, because
1371
# they do not have compression children of their own.
1372
self.assertEqual({}, ann._content_objects)
1374
def test__expand_delta_records_blocks(self):
1375
ann = self.make_annotator()
1376
rev_key = ('rev-id',)
1377
parent_key = ('parent-id',)
1378
record = ['0,1,1\n', 'new-line\n']
1379
details = ('line-delta', True)
1380
ann._num_compression_children[parent_key] = 2
1381
ann._expand_record(parent_key, (), None,
1382
['line1\n', 'line2\n', 'line3\n'],
1383
('fulltext', False))
1384
ann._expand_record(rev_key, (parent_key,), parent_key, record, details)
1385
self.assertEqual({(rev_key, parent_key): [(1, 1, 1), (3, 3, 0)]},
1386
ann._matching_blocks)
1387
rev2_key = ('rev2-id',)
1388
record = ['0,1,1\n', 'new-line\n']
1389
details = ('line-delta', False)
1390
ann._expand_record(rev2_key, (parent_key,), parent_key, record, details)
1391
self.assertEqual([(1, 1, 2), (3, 3, 0)],
1392
ann._matching_blocks[(rev2_key, parent_key)])
1394
def test__get_parent_ann_uses_matching_blocks(self):
1395
ann = self.make_annotator()
1396
rev_key = ('rev-id',)
1397
parent_key = ('parent-id',)
1398
parent_ann = [(parent_key,)]*3
1399
block_key = (rev_key, parent_key)
1400
ann._annotations_cache[parent_key] = parent_ann
1401
ann._matching_blocks[block_key] = [(0, 1, 1), (3, 3, 0)]
1402
# We should not try to access any parent_lines content, because we know
1403
# we already have the matching blocks
1404
par_ann, blocks = ann._get_parent_annotations_and_matches(rev_key,
1405
['1\n', '2\n', '3\n'], parent_key)
1406
self.assertEqual(parent_ann, par_ann)
1407
self.assertEqual([(0, 1, 1), (3, 3, 0)], blocks)
1408
self.assertEqual({}, ann._matching_blocks)
1410
def test__process_pending(self):
1411
ann = self.make_annotator()
1412
rev_key = ('rev-id',)
1415
record = ['0,1,1\n', 'new-line\n']
1416
details = ('line-delta', False)
1417
p1_record = ['line1\n', 'line2\n']
1418
ann._num_compression_children[p1_key] = 1
1419
res = ann._expand_record(rev_key, (p1_key,p2_key), p1_key,
1421
self.assertEqual(None, res)
1422
# self.assertTrue(p1_key in ann._pending_deltas)
1423
self.assertEqual({}, ann._pending_annotation)
1424
# Now insert p1, and we should be able to expand the delta
1425
res = ann._expand_record(p1_key, (), None, p1_record,
1426
('fulltext', False))
1427
self.assertEqual(p1_record, res)
1428
ann._annotations_cache[p1_key] = [(p1_key,)]*2
1429
res = ann._process_pending(p1_key)
1430
self.assertEqual([], res)
1431
self.assertFalse(p1_key in ann._pending_deltas)
1432
self.assertTrue(p2_key in ann._pending_annotation)
1433
self.assertEqual({p2_key: [(rev_key, (p1_key, p2_key))]},
1434
ann._pending_annotation)
1435
# Now fill in parent 2, and pending annotation should be satisfied
1436
res = ann._expand_record(p2_key, (), None, [], ('fulltext', False))
1437
ann._annotations_cache[p2_key] = []
1438
res = ann._process_pending(p2_key)
1439
self.assertEqual([rev_key], res)
1440
self.assertEqual({}, ann._pending_annotation)
1441
self.assertEqual({}, ann._pending_deltas)
1443
def test_record_delta_removes_basis(self):
1444
ann = self.make_annotator()
1445
ann._expand_record(('parent-id',), (), None,
1446
['line1\n', 'line2\n'], ('fulltext', False))
1447
ann._num_compression_children['parent-id'] = 2
1449
def test_annotate_special_text(self):
1450
ann = self.make_annotator()
1452
rev1_key = ('rev-1',)
1453
rev2_key = ('rev-2',)
1454
rev3_key = ('rev-3',)
1455
spec_key = ('special:',)
1456
vf.add_lines(rev1_key, [], ['initial content\n'])
1457
vf.add_lines(rev2_key, [rev1_key], ['initial content\n',
1460
vf.add_lines(rev3_key, [rev1_key], ['initial content\n',
1463
spec_text = ('initial content\n'
1467
ann.add_special_text(spec_key, [rev2_key, rev3_key], spec_text)
1468
anns, lines = ann.annotate(spec_key)
1469
self.assertEqual([(rev1_key,),
1470
(rev2_key, rev3_key),
1474
self.assertEqualDiff(spec_text, ''.join(lines))
1477
class KnitTests(TestCaseWithTransport):
1478
"""Class containing knit test helper routines."""
1480
def make_test_knit(self, annotate=False, name='test'):
1481
mapper = ConstantMapper(name)
1482
return make_file_factory(annotate, mapper)(self.get_transport())
1485
class TestBadShaError(KnitTests):
1486
"""Tests for handling of sha errors."""
1488
def test_sha_exception_has_text(self):
1489
# having the failed text included in the error allows for recovery.
1490
source = self.make_test_knit()
1491
target = self.make_test_knit(name="target")
1492
if not source._max_delta_chain:
1493
raise TestNotApplicable(
1494
"cannot get delta-caused sha failures without deltas.")
1497
broken = ('broken',)
1498
source.add_lines(basis, (), ['foo\n'])
1499
source.add_lines(broken, (basis,), ['foo\n', 'bar\n'])
1500
# Seed target with a bad basis text
1501
target.add_lines(basis, (), ['gam\n'])
1502
target.insert_record_stream(
1503
source.get_record_stream([broken], 'unordered', False))
1504
err = self.assertRaises(errors.KnitCorrupt,
1505
target.get_record_stream([broken], 'unordered', True
1506
).next().get_bytes_as, 'chunked')
1507
self.assertEqual(['gam\n', 'bar\n'], err.content)
1508
# Test for formatting with live data
1509
self.assertStartsWith(str(err), "Knit ")
1512
class TestKnitIndex(KnitTests):
1514
def test_add_versions_dictionary_compresses(self):
1515
"""Adding versions to the index should update the lookup dict"""
1516
knit = self.make_test_knit()
1518
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1519
self.check_file_contents('test.kndx',
1520
'# bzr knit index 8\n'
1522
'a-1 fulltext 0 0 :'
1525
(('a-2',), ['fulltext'], (('a-2',), 0, 0), [('a-1',)]),
1526
(('a-3',), ['fulltext'], (('a-3',), 0, 0), [('a-2',)]),
1528
self.check_file_contents('test.kndx',
1529
'# bzr knit index 8\n'
1531
'a-1 fulltext 0 0 :\n'
1532
'a-2 fulltext 0 0 0 :\n'
1533
'a-3 fulltext 0 0 1 :'
1535
self.assertEqual(set([('a-3',), ('a-1',), ('a-2',)]), idx.keys())
1537
('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False)),
1538
('a-2',): ((('a-2',), 0, 0), None, (('a-1',),), ('fulltext', False)),
1539
('a-3',): ((('a-3',), 0, 0), None, (('a-2',),), ('fulltext', False)),
1540
}, idx.get_build_details(idx.keys()))
1541
self.assertEqual({('a-1',):(),
1542
('a-2',):(('a-1',),),
1543
('a-3',):(('a-2',),),},
1544
idx.get_parent_map(idx.keys()))
1546
def test_add_versions_fails_clean(self):
1547
"""If add_versions fails in the middle, it restores a pristine state.
1549
Any modifications that are made to the index are reset if all versions
1552
# This cheats a little bit by passing in a generator which will
1553
# raise an exception before the processing finishes
1554
# Other possibilities would be to have an version with the wrong number
1555
# of entries, or to make the backing transport unable to write any
1558
knit = self.make_test_knit()
1560
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1562
class StopEarly(Exception):
1565
def generate_failure():
1566
"""Add some entries and then raise an exception"""
1567
yield (('a-2',), ['fulltext'], (None, 0, 0), ('a-1',))
1568
yield (('a-3',), ['fulltext'], (None, 0, 0), ('a-2',))
1571
# Assert the pre-condition
1573
self.assertEqual(set([('a-1',)]), set(idx.keys()))
1575
{('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False))},
1576
idx.get_build_details([('a-1',)]))
1577
self.assertEqual({('a-1',):()}, idx.get_parent_map(idx.keys()))
1580
self.assertRaises(StopEarly, idx.add_records, generate_failure())
1581
# And it shouldn't be modified
1584
def test_knit_index_ignores_empty_files(self):
1585
# There was a race condition in older bzr, where a ^C at the right time
1586
# could leave an empty .kndx file, which bzr would later claim was a
1587
# corrupted file since the header was not present. In reality, the file
1588
# just wasn't created, so it should be ignored.
1589
t = get_transport('.')
1590
t.put_bytes('test.kndx', '')
1592
knit = self.make_test_knit()
1594
def test_knit_index_checks_header(self):
1595
t = get_transport('.')
1596
t.put_bytes('test.kndx', '# not really a knit header\n\n')
1597
k = self.make_test_knit()
1598
self.assertRaises(KnitHeaderError, k.keys)
1601
class TestGraphIndexKnit(KnitTests):
1602
"""Tests for knits using a GraphIndex rather than a KnitIndex."""
1604
def make_g_index(self, name, ref_lists=0, nodes=[]):
1605
builder = GraphIndexBuilder(ref_lists)
1606
for node, references, value in nodes:
1607
builder.add_node(node, references, value)
1608
stream = builder.finish()
1609
trans = self.get_transport()
1610
size = trans.put_file(name, stream)
1611
return GraphIndex(trans, name, size)
1613
def two_graph_index(self, deltas=False, catch_adds=False):
1614
"""Build a two-graph index.
1616
:param deltas: If true, use underlying indices with two node-ref
1617
lists and 'parent' set to a delta-compressed against tail.
1619
# build a complex graph across several indices.
1621
# delta compression inn the index
1622
index1 = self.make_g_index('1', 2, [
1623
(('tip', ), 'N0 100', ([('parent', )], [], )),
1624
(('tail', ), '', ([], []))])
1625
index2 = self.make_g_index('2', 2, [
1626
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], [('tail', )])),
1627
(('separate', ), '', ([], []))])
1629
# just blob location and graph in the index.
1630
index1 = self.make_g_index('1', 1, [
1631
(('tip', ), 'N0 100', ([('parent', )], )),
1632
(('tail', ), '', ([], ))])
1633
index2 = self.make_g_index('2', 1, [
1634
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], )),
1635
(('separate', ), '', ([], ))])
1636
combined_index = CombinedGraphIndex([index1, index2])
1638
self.combined_index = combined_index
1639
self.caught_entries = []
1640
add_callback = self.catch_add
1643
return _KnitGraphIndex(combined_index, lambda:True, deltas=deltas,
1644
add_callback=add_callback)
1646
def test_keys(self):
1647
index = self.two_graph_index()
1648
self.assertEqual(set([('tail',), ('tip',), ('parent',), ('separate',)]),
1651
def test_get_position(self):
1652
index = self.two_graph_index()
1653
self.assertEqual((index._graph_index._indices[0], 0, 100), index.get_position(('tip',)))
1654
self.assertEqual((index._graph_index._indices[1], 100, 78), index.get_position(('parent',)))
1656
def test_get_method_deltas(self):
1657
index = self.two_graph_index(deltas=True)
1658
self.assertEqual('fulltext', index.get_method(('tip',)))
1659
self.assertEqual('line-delta', index.get_method(('parent',)))
1661
def test_get_method_no_deltas(self):
1662
# check that the parent-history lookup is ignored with deltas=False.
1663
index = self.two_graph_index(deltas=False)
1664
self.assertEqual('fulltext', index.get_method(('tip',)))
1665
self.assertEqual('fulltext', index.get_method(('parent',)))
1667
def test_get_options_deltas(self):
1668
index = self.two_graph_index(deltas=True)
1669
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1670
self.assertEqual(['line-delta'], index.get_options(('parent',)))
1672
def test_get_options_no_deltas(self):
1673
# check that the parent-history lookup is ignored with deltas=False.
1674
index = self.two_graph_index(deltas=False)
1675
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1676
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1678
def test_get_parent_map(self):
1679
index = self.two_graph_index()
1680
self.assertEqual({('parent',):(('tail',), ('ghost',))},
1681
index.get_parent_map([('parent',), ('ghost',)]))
1683
def catch_add(self, entries):
1684
self.caught_entries.append(entries)
1686
def test_add_no_callback_errors(self):
1687
index = self.two_graph_index()
1688
self.assertRaises(errors.ReadOnlyError, index.add_records,
1689
[(('new',), 'fulltext,no-eol', (None, 50, 60), ['separate'])])
1691
def test_add_version_smoke(self):
1692
index = self.two_graph_index(catch_adds=True)
1693
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60),
1695
self.assertEqual([[(('new', ), 'N50 60', ((('separate',),),))]],
1696
self.caught_entries)
1698
def test_add_version_delta_not_delta_index(self):
1699
index = self.two_graph_index(catch_adds=True)
1700
self.assertRaises(errors.KnitCorrupt, index.add_records,
1701
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1702
self.assertEqual([], self.caught_entries)
1704
def test_add_version_same_dup(self):
1705
index = self.two_graph_index(catch_adds=True)
1706
# options can be spelt two different ways
1707
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
1708
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
1709
# position/length are ignored (because each pack could have fulltext or
1710
# delta, and be at a different position.
1711
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1713
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1715
# but neither should have added data:
1716
self.assertEqual([[], [], [], []], self.caught_entries)
1718
def test_add_version_different_dup(self):
1719
index = self.two_graph_index(deltas=True, catch_adds=True)
1721
self.assertRaises(errors.KnitCorrupt, index.add_records,
1722
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1723
self.assertRaises(errors.KnitCorrupt, index.add_records,
1724
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1726
self.assertRaises(errors.KnitCorrupt, index.add_records,
1727
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1728
self.assertEqual([], self.caught_entries)
1730
def test_add_versions_nodeltas(self):
1731
index = self.two_graph_index(catch_adds=True)
1733
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1734
(('new2',), 'fulltext', (None, 0, 6), [('new',)]),
1736
self.assertEqual([(('new', ), 'N50 60', ((('separate',),),)),
1737
(('new2', ), ' 0 6', ((('new',),),))],
1738
sorted(self.caught_entries[0]))
1739
self.assertEqual(1, len(self.caught_entries))
1741
def test_add_versions_deltas(self):
1742
index = self.two_graph_index(deltas=True, catch_adds=True)
1744
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1745
(('new2',), 'line-delta', (None, 0, 6), [('new',)]),
1747
self.assertEqual([(('new', ), 'N50 60', ((('separate',),), ())),
1748
(('new2', ), ' 0 6', ((('new',),), (('new',),), ))],
1749
sorted(self.caught_entries[0]))
1750
self.assertEqual(1, len(self.caught_entries))
1752
def test_add_versions_delta_not_delta_index(self):
1753
index = self.two_graph_index(catch_adds=True)
1754
self.assertRaises(errors.KnitCorrupt, index.add_records,
1755
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1756
self.assertEqual([], self.caught_entries)
1758
def test_add_versions_random_id_accepted(self):
1759
index = self.two_graph_index(catch_adds=True)
1760
index.add_records([], random_id=True)
1762
def test_add_versions_same_dup(self):
1763
index = self.two_graph_index(catch_adds=True)
1764
# options can be spelt two different ways
1765
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100),
1767
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100),
1769
# position/length are ignored (because each pack could have fulltext or
1770
# delta, and be at a different position.
1771
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1773
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1775
# but neither should have added data.
1776
self.assertEqual([[], [], [], []], self.caught_entries)
1778
def test_add_versions_different_dup(self):
1779
index = self.two_graph_index(deltas=True, catch_adds=True)
1781
self.assertRaises(errors.KnitCorrupt, index.add_records,
1782
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1783
self.assertRaises(errors.KnitCorrupt, index.add_records,
1784
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1786
self.assertRaises(errors.KnitCorrupt, index.add_records,
1787
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1788
# change options in the second record
1789
self.assertRaises(errors.KnitCorrupt, index.add_records,
1790
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)]),
1791
(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1792
self.assertEqual([], self.caught_entries)
1794
def make_g_index_missing_compression_parent(self):
1795
graph_index = self.make_g_index('missing_comp', 2,
1796
[(('tip', ), ' 100 78',
1797
([('missing-parent', ), ('ghost', )], [('missing-parent', )]))])
1800
def make_g_index_missing_parent(self):
1801
graph_index = self.make_g_index('missing_parent', 2,
1802
[(('parent', ), ' 100 78', ([], [])),
1803
(('tip', ), ' 100 78',
1804
([('parent', ), ('missing-parent', )], [('parent', )])),
1808
def make_g_index_no_external_refs(self):
1809
graph_index = self.make_g_index('no_external_refs', 2,
1810
[(('rev', ), ' 100 78',
1811
([('parent', ), ('ghost', )], []))])
1814
def test_add_good_unvalidated_index(self):
1815
unvalidated = self.make_g_index_no_external_refs()
1816
combined = CombinedGraphIndex([unvalidated])
1817
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1818
index.scan_unvalidated_index(unvalidated)
1819
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1821
def test_add_missing_compression_parent_unvalidated_index(self):
1822
unvalidated = self.make_g_index_missing_compression_parent()
1823
combined = CombinedGraphIndex([unvalidated])
1824
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1825
index.scan_unvalidated_index(unvalidated)
1826
# This also checks that its only the compression parent that is
1827
# examined, otherwise 'ghost' would also be reported as a missing
1830
frozenset([('missing-parent',)]),
1831
index.get_missing_compression_parents())
1833
def test_add_missing_noncompression_parent_unvalidated_index(self):
1834
unvalidated = self.make_g_index_missing_parent()
1835
combined = CombinedGraphIndex([unvalidated])
1836
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1837
track_external_parent_refs=True)
1838
index.scan_unvalidated_index(unvalidated)
1840
frozenset([('missing-parent',)]), index.get_missing_parents())
1842
def test_track_external_parent_refs(self):
1843
g_index = self.make_g_index('empty', 2, [])
1844
combined = CombinedGraphIndex([g_index])
1845
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1846
add_callback=self.catch_add, track_external_parent_refs=True)
1847
self.caught_entries = []
1849
(('new-key',), 'fulltext,no-eol', (None, 50, 60),
1850
[('parent-1',), ('parent-2',)])])
1852
frozenset([('parent-1',), ('parent-2',)]),
1853
index.get_missing_parents())
1855
def test_add_unvalidated_index_with_present_external_references(self):
1856
index = self.two_graph_index(deltas=True)
1857
# Ugly hack to get at one of the underlying GraphIndex objects that
1858
# two_graph_index built.
1859
unvalidated = index._graph_index._indices[1]
1860
# 'parent' is an external ref of _indices[1] (unvalidated), but is
1861
# present in _indices[0].
1862
index.scan_unvalidated_index(unvalidated)
1863
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1865
def make_new_missing_parent_g_index(self, name):
1866
missing_parent = name + '-missing-parent'
1867
graph_index = self.make_g_index(name, 2,
1868
[((name + 'tip', ), ' 100 78',
1869
([(missing_parent, ), ('ghost', )], [(missing_parent, )]))])
1872
def test_add_mulitiple_unvalidated_indices_with_missing_parents(self):
1873
g_index_1 = self.make_new_missing_parent_g_index('one')
1874
g_index_2 = self.make_new_missing_parent_g_index('two')
1875
combined = CombinedGraphIndex([g_index_1, g_index_2])
1876
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1877
index.scan_unvalidated_index(g_index_1)
1878
index.scan_unvalidated_index(g_index_2)
1880
frozenset([('one-missing-parent',), ('two-missing-parent',)]),
1881
index.get_missing_compression_parents())
1883
def test_add_mulitiple_unvalidated_indices_with_mutual_dependencies(self):
1884
graph_index_a = self.make_g_index('one', 2,
1885
[(('parent-one', ), ' 100 78', ([('non-compression-parent',)], [])),
1886
(('child-of-two', ), ' 100 78',
1887
([('parent-two',)], [('parent-two',)]))])
1888
graph_index_b = self.make_g_index('two', 2,
1889
[(('parent-two', ), ' 100 78', ([('non-compression-parent',)], [])),
1890
(('child-of-one', ), ' 100 78',
1891
([('parent-one',)], [('parent-one',)]))])
1892
combined = CombinedGraphIndex([graph_index_a, graph_index_b])
1893
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1894
index.scan_unvalidated_index(graph_index_a)
1895
index.scan_unvalidated_index(graph_index_b)
1897
frozenset([]), index.get_missing_compression_parents())
1900
class TestNoParentsGraphIndexKnit(KnitTests):
1901
"""Tests for knits using _KnitGraphIndex with no parents."""
1903
def make_g_index(self, name, ref_lists=0, nodes=[]):
1904
builder = GraphIndexBuilder(ref_lists)
1905
for node, references in nodes:
1906
builder.add_node(node, references)
1907
stream = builder.finish()
1908
trans = self.get_transport()
1909
size = trans.put_file(name, stream)
1910
return GraphIndex(trans, name, size)
1912
def test_add_good_unvalidated_index(self):
1913
unvalidated = self.make_g_index('unvalidated')
1914
combined = CombinedGraphIndex([unvalidated])
1915
index = _KnitGraphIndex(combined, lambda: True, parents=False)
1916
index.scan_unvalidated_index(unvalidated)
1917
self.assertEqual(frozenset(),
1918
index.get_missing_compression_parents())
1920
def test_parents_deltas_incompatible(self):
1921
index = CombinedGraphIndex([])
1922
self.assertRaises(errors.KnitError, _KnitGraphIndex, lambda:True,
1923
index, deltas=True, parents=False)
1925
def two_graph_index(self, catch_adds=False):
1926
"""Build a two-graph index.
1928
:param deltas: If true, use underlying indices with two node-ref
1929
lists and 'parent' set to a delta-compressed against tail.
1931
# put several versions in the index.
1932
index1 = self.make_g_index('1', 0, [
1933
(('tip', ), 'N0 100'),
1935
index2 = self.make_g_index('2', 0, [
1936
(('parent', ), ' 100 78'),
1937
(('separate', ), '')])
1938
combined_index = CombinedGraphIndex([index1, index2])
1940
self.combined_index = combined_index
1941
self.caught_entries = []
1942
add_callback = self.catch_add
1945
return _KnitGraphIndex(combined_index, lambda:True, parents=False,
1946
add_callback=add_callback)
1948
def test_keys(self):
1949
index = self.two_graph_index()
1950
self.assertEqual(set([('tail',), ('tip',), ('parent',), ('separate',)]),
1953
def test_get_position(self):
1954
index = self.two_graph_index()
1955
self.assertEqual((index._graph_index._indices[0], 0, 100),
1956
index.get_position(('tip',)))
1957
self.assertEqual((index._graph_index._indices[1], 100, 78),
1958
index.get_position(('parent',)))
1960
def test_get_method(self):
1961
index = self.two_graph_index()
1962
self.assertEqual('fulltext', index.get_method(('tip',)))
1963
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1965
def test_get_options(self):
1966
index = self.two_graph_index()
1967
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1968
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1970
def test_get_parent_map(self):
1971
index = self.two_graph_index()
1972
self.assertEqual({('parent',):None},
1973
index.get_parent_map([('parent',), ('ghost',)]))
1975
def catch_add(self, entries):
1976
self.caught_entries.append(entries)
1978
def test_add_no_callback_errors(self):
1979
index = self.two_graph_index()
1980
self.assertRaises(errors.ReadOnlyError, index.add_records,
1981
[(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)])])
1983
def test_add_version_smoke(self):
1984
index = self.two_graph_index(catch_adds=True)
1985
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60), [])])
1986
self.assertEqual([[(('new', ), 'N50 60')]],
1987
self.caught_entries)
1989
def test_add_version_delta_not_delta_index(self):
1990
index = self.two_graph_index(catch_adds=True)
1991
self.assertRaises(errors.KnitCorrupt, index.add_records,
1992
[(('new',), 'no-eol,line-delta', (None, 0, 100), [])])
1993
self.assertEqual([], self.caught_entries)
1995
def test_add_version_same_dup(self):
1996
index = self.two_graph_index(catch_adds=True)
1997
# options can be spelt two different ways
1998
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1999
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
2000
# position/length are ignored (because each pack could have fulltext or
2001
# delta, and be at a different position.
2002
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
2003
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
2004
# but neither should have added data.
2005
self.assertEqual([[], [], [], []], self.caught_entries)
2007
def test_add_version_different_dup(self):
2008
index = self.two_graph_index(catch_adds=True)
2010
self.assertRaises(errors.KnitCorrupt, index.add_records,
2011
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2012
self.assertRaises(errors.KnitCorrupt, index.add_records,
2013
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
2014
self.assertRaises(errors.KnitCorrupt, index.add_records,
2015
[(('tip',), 'fulltext', (None, 0, 100), [])])
2017
self.assertRaises(errors.KnitCorrupt, index.add_records,
2018
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
2019
self.assertEqual([], self.caught_entries)
2021
def test_add_versions(self):
2022
index = self.two_graph_index(catch_adds=True)
2024
(('new',), 'fulltext,no-eol', (None, 50, 60), []),
2025
(('new2',), 'fulltext', (None, 0, 6), []),
2027
self.assertEqual([(('new', ), 'N50 60'), (('new2', ), ' 0 6')],
2028
sorted(self.caught_entries[0]))
2029
self.assertEqual(1, len(self.caught_entries))
2031
def test_add_versions_delta_not_delta_index(self):
2032
index = self.two_graph_index(catch_adds=True)
2033
self.assertRaises(errors.KnitCorrupt, index.add_records,
2034
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
2035
self.assertEqual([], self.caught_entries)
2037
def test_add_versions_parents_not_parents_index(self):
2038
index = self.two_graph_index(catch_adds=True)
2039
self.assertRaises(errors.KnitCorrupt, index.add_records,
2040
[(('new',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
2041
self.assertEqual([], self.caught_entries)
2043
def test_add_versions_random_id_accepted(self):
2044
index = self.two_graph_index(catch_adds=True)
2045
index.add_records([], random_id=True)
2047
def test_add_versions_same_dup(self):
2048
index = self.two_graph_index(catch_adds=True)
2049
# options can be spelt two different ways
2050
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
2051
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
2052
# position/length are ignored (because each pack could have fulltext or
2053
# delta, and be at a different position.
2054
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
2055
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
2056
# but neither should have added data.
2057
self.assertEqual([[], [], [], []], self.caught_entries)
2059
def test_add_versions_different_dup(self):
2060
index = self.two_graph_index(catch_adds=True)
2062
self.assertRaises(errors.KnitCorrupt, index.add_records,
2063
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2064
self.assertRaises(errors.KnitCorrupt, index.add_records,
2065
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
2066
self.assertRaises(errors.KnitCorrupt, index.add_records,
2067
[(('tip',), 'fulltext', (None, 0, 100), [])])
2069
self.assertRaises(errors.KnitCorrupt, index.add_records,
2070
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
2071
# change options in the second record
2072
self.assertRaises(errors.KnitCorrupt, index.add_records,
2073
[(('tip',), 'fulltext,no-eol', (None, 0, 100), []),
2074
(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2075
self.assertEqual([], self.caught_entries)
2078
class TestKnitVersionedFiles(KnitTests):
2080
def assertGroupKeysForIo(self, exp_groups, keys, non_local_keys,
2081
positions, _min_buffer_size=None):
2082
kvf = self.make_test_knit()
2083
if _min_buffer_size is None:
2084
_min_buffer_size = knit._STREAM_MIN_BUFFER_SIZE
2085
self.assertEqual(exp_groups, kvf._group_keys_for_io(keys,
2086
non_local_keys, positions,
2087
_min_buffer_size=_min_buffer_size))
2089
def assertSplitByPrefix(self, expected_map, expected_prefix_order,
2091
split, prefix_order = KnitVersionedFiles._split_by_prefix(keys)
2092
self.assertEqual(expected_map, split)
2093
self.assertEqual(expected_prefix_order, prefix_order)
2095
def test__group_keys_for_io(self):
2096
ft_detail = ('fulltext', False)
2097
ld_detail = ('line-delta', False)
2105
f_a: (ft_detail, (f_a, 0, 100), None),
2106
f_b: (ld_detail, (f_b, 100, 21), f_a),
2107
f_c: (ld_detail, (f_c, 180, 15), f_b),
2108
g_a: (ft_detail, (g_a, 121, 35), None),
2109
g_b: (ld_detail, (g_b, 156, 12), g_a),
2110
g_c: (ld_detail, (g_c, 195, 13), g_a),
2112
self.assertGroupKeysForIo([([f_a], set())],
2113
[f_a], [], positions)
2114
self.assertGroupKeysForIo([([f_a], set([f_a]))],
2115
[f_a], [f_a], positions)
2116
self.assertGroupKeysForIo([([f_a, f_b], set([]))],
2117
[f_a, f_b], [], positions)
2118
self.assertGroupKeysForIo([([f_a, f_b], set([f_b]))],
2119
[f_a, f_b], [f_b], positions)
2120
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2121
[f_a, g_a, f_b, g_b], [], positions)
2122
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2123
[f_a, g_a, f_b, g_b], [], positions,
2124
_min_buffer_size=150)
2125
self.assertGroupKeysForIo([([f_a, f_b], set()), ([g_a, g_b], set())],
2126
[f_a, g_a, f_b, g_b], [], positions,
2127
_min_buffer_size=100)
2128
self.assertGroupKeysForIo([([f_c], set()), ([g_b], set())],
2129
[f_c, g_b], [], positions,
2130
_min_buffer_size=125)
2131
self.assertGroupKeysForIo([([g_b, f_c], set())],
2132
[g_b, f_c], [], positions,
2133
_min_buffer_size=125)
2135
def test__split_by_prefix(self):
2136
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2137
'g': [('g', 'b'), ('g', 'a')],
2139
[('f', 'a'), ('g', 'b'),
2140
('g', 'a'), ('f', 'b')])
2142
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2143
'g': [('g', 'b'), ('g', 'a')],
2145
[('f', 'a'), ('f', 'b'),
2146
('g', 'b'), ('g', 'a')])
2148
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2149
'g': [('g', 'b'), ('g', 'a')],
2151
[('f', 'a'), ('f', 'b'),
2152
('g', 'b'), ('g', 'a')])
2154
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2155
'g': [('g', 'b'), ('g', 'a')],
2156
'': [('a',), ('b',)]
2158
[('f', 'a'), ('g', 'b'),
2160
('g', 'a'), ('f', 'b')])
2163
class TestStacking(KnitTests):
2165
def get_basis_and_test_knit(self):
2166
basis = self.make_test_knit(name='basis')
2167
basis = RecordingVersionedFilesDecorator(basis)
2168
test = self.make_test_knit(name='test')
2169
test.add_fallback_versioned_files(basis)
2172
def test_add_fallback_versioned_files(self):
2173
basis = self.make_test_knit(name='basis')
2174
test = self.make_test_knit(name='test')
2175
# It must not error; other tests test that the fallback is referred to
2176
# when accessing data.
2177
test.add_fallback_versioned_files(basis)
2179
def test_add_lines(self):
2180
# lines added to the test are not added to the basis
2181
basis, test = self.get_basis_and_test_knit()
2183
key_basis = ('bar',)
2184
key_cross_border = ('quux',)
2185
key_delta = ('zaphod',)
2186
test.add_lines(key, (), ['foo\n'])
2187
self.assertEqual({}, basis.get_parent_map([key]))
2188
# lines added to the test that reference across the stack do a
2190
basis.add_lines(key_basis, (), ['foo\n'])
2192
test.add_lines(key_cross_border, (key_basis,), ['foo\n'])
2193
self.assertEqual('fulltext', test._index.get_method(key_cross_border))
2194
# we don't even need to look at the basis to see that this should be
2195
# stored as a fulltext
2196
self.assertEqual([], basis.calls)
2197
# Subsequent adds do delta.
2199
test.add_lines(key_delta, (key_cross_border,), ['foo\n'])
2200
self.assertEqual('line-delta', test._index.get_method(key_delta))
2201
self.assertEqual([], basis.calls)
2203
def test_annotate(self):
2204
# annotations from the test knit are answered without asking the basis
2205
basis, test = self.get_basis_and_test_knit()
2207
key_basis = ('bar',)
2208
key_missing = ('missing',)
2209
test.add_lines(key, (), ['foo\n'])
2210
details = test.annotate(key)
2211
self.assertEqual([(key, 'foo\n')], details)
2212
self.assertEqual([], basis.calls)
2213
# But texts that are not in the test knit are looked for in the basis
2215
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2217
details = test.annotate(key_basis)
2218
self.assertEqual([(key_basis, 'foo\n'), (key_basis, 'bar\n')], details)
2219
# Not optimised to date:
2220
# self.assertEqual([("annotate", key_basis)], basis.calls)
2221
self.assertEqual([('get_parent_map', set([key_basis])),
2222
('get_parent_map', set([key_basis])),
2223
('get_record_stream', [key_basis], 'topological', True)],
2226
def test_check(self):
2227
# At the moment checking a stacked knit does implicitly check the
2229
basis, test = self.get_basis_and_test_knit()
2232
def test_get_parent_map(self):
2233
# parents in the test knit are answered without asking the basis
2234
basis, test = self.get_basis_and_test_knit()
2236
key_basis = ('bar',)
2237
key_missing = ('missing',)
2238
test.add_lines(key, (), [])
2239
parent_map = test.get_parent_map([key])
2240
self.assertEqual({key: ()}, parent_map)
2241
self.assertEqual([], basis.calls)
2242
# But parents that are not in the test knit are looked for in the basis
2243
basis.add_lines(key_basis, (), [])
2245
parent_map = test.get_parent_map([key, key_basis, key_missing])
2246
self.assertEqual({key: (),
2247
key_basis: ()}, parent_map)
2248
self.assertEqual([("get_parent_map", set([key_basis, key_missing]))],
2251
def test_get_record_stream_unordered_fulltexts(self):
2252
# records from the test knit are answered without asking the basis:
2253
basis, test = self.get_basis_and_test_knit()
2255
key_basis = ('bar',)
2256
key_missing = ('missing',)
2257
test.add_lines(key, (), ['foo\n'])
2258
records = list(test.get_record_stream([key], 'unordered', True))
2259
self.assertEqual(1, len(records))
2260
self.assertEqual([], basis.calls)
2261
# Missing (from test knit) objects are retrieved from the basis:
2262
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2264
records = list(test.get_record_stream([key_basis, key_missing],
2266
self.assertEqual(2, len(records))
2267
calls = list(basis.calls)
2268
for record in records:
2269
self.assertSubset([record.key], (key_basis, key_missing))
2270
if record.key == key_missing:
2271
self.assertIsInstance(record, AbsentContentFactory)
2273
reference = list(basis.get_record_stream([key_basis],
2274
'unordered', True))[0]
2275
self.assertEqual(reference.key, record.key)
2276
self.assertEqual(reference.sha1, record.sha1)
2277
self.assertEqual(reference.storage_kind, record.storage_kind)
2278
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2279
record.get_bytes_as(record.storage_kind))
2280
self.assertEqual(reference.get_bytes_as('fulltext'),
2281
record.get_bytes_as('fulltext'))
2282
# It's not strictly minimal, but it seems reasonable for now for it to
2283
# ask which fallbacks have which parents.
2285
("get_parent_map", set([key_basis, key_missing])),
2286
("get_record_stream", [key_basis], 'unordered', True)],
2289
def test_get_record_stream_ordered_fulltexts(self):
2290
# ordering is preserved down into the fallback store.
2291
basis, test = self.get_basis_and_test_knit()
2293
key_basis = ('bar',)
2294
key_basis_2 = ('quux',)
2295
key_missing = ('missing',)
2296
test.add_lines(key, (key_basis,), ['foo\n'])
2297
# Missing (from test knit) objects are retrieved from the basis:
2298
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2299
basis.add_lines(key_basis_2, (), ['quux\n'])
2301
# ask for in non-topological order
2302
records = list(test.get_record_stream(
2303
[key, key_basis, key_missing, key_basis_2], 'topological', True))
2304
self.assertEqual(4, len(records))
2306
for record in records:
2307
self.assertSubset([record.key],
2308
(key_basis, key_missing, key_basis_2, key))
2309
if record.key == key_missing:
2310
self.assertIsInstance(record, AbsentContentFactory)
2312
results.append((record.key, record.sha1, record.storage_kind,
2313
record.get_bytes_as('fulltext')))
2314
calls = list(basis.calls)
2315
order = [record[0] for record in results]
2316
self.assertEqual([key_basis_2, key_basis, key], order)
2317
for result in results:
2318
if result[0] == key:
2322
record = source.get_record_stream([result[0]], 'unordered',
2324
self.assertEqual(record.key, result[0])
2325
self.assertEqual(record.sha1, result[1])
2326
# We used to check that the storage kind matched, but actually it
2327
# depends on whether it was sourced from the basis, or in a single
2328
# group, because asking for full texts returns proxy objects to a
2329
# _ContentMapGenerator object; so checking the kind is unneeded.
2330
self.assertEqual(record.get_bytes_as('fulltext'), result[3])
2331
# It's not strictly minimal, but it seems reasonable for now for it to
2332
# ask which fallbacks have which parents.
2334
("get_parent_map", set([key_basis, key_basis_2, key_missing])),
2335
# topological is requested from the fallback, because that is what
2336
# was requested at the top level.
2337
("get_record_stream", [key_basis_2, key_basis], 'topological', True)],
2340
def test_get_record_stream_unordered_deltas(self):
2341
# records from the test knit are answered without asking the basis:
2342
basis, test = self.get_basis_and_test_knit()
2344
key_basis = ('bar',)
2345
key_missing = ('missing',)
2346
test.add_lines(key, (), ['foo\n'])
2347
records = list(test.get_record_stream([key], 'unordered', False))
2348
self.assertEqual(1, len(records))
2349
self.assertEqual([], basis.calls)
2350
# Missing (from test knit) objects are retrieved from the basis:
2351
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2353
records = list(test.get_record_stream([key_basis, key_missing],
2354
'unordered', False))
2355
self.assertEqual(2, len(records))
2356
calls = list(basis.calls)
2357
for record in records:
2358
self.assertSubset([record.key], (key_basis, key_missing))
2359
if record.key == key_missing:
2360
self.assertIsInstance(record, AbsentContentFactory)
2362
reference = list(basis.get_record_stream([key_basis],
2363
'unordered', False))[0]
2364
self.assertEqual(reference.key, record.key)
2365
self.assertEqual(reference.sha1, record.sha1)
2366
self.assertEqual(reference.storage_kind, record.storage_kind)
2367
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2368
record.get_bytes_as(record.storage_kind))
2369
# It's not strictly minimal, but it seems reasonable for now for it to
2370
# ask which fallbacks have which parents.
2372
("get_parent_map", set([key_basis, key_missing])),
2373
("get_record_stream", [key_basis], 'unordered', False)],
2376
def test_get_record_stream_ordered_deltas(self):
2377
# ordering is preserved down into the fallback store.
2378
basis, test = self.get_basis_and_test_knit()
2380
key_basis = ('bar',)
2381
key_basis_2 = ('quux',)
2382
key_missing = ('missing',)
2383
test.add_lines(key, (key_basis,), ['foo\n'])
2384
# Missing (from test knit) objects are retrieved from the basis:
2385
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2386
basis.add_lines(key_basis_2, (), ['quux\n'])
2388
# ask for in non-topological order
2389
records = list(test.get_record_stream(
2390
[key, key_basis, key_missing, key_basis_2], 'topological', False))
2391
self.assertEqual(4, len(records))
2393
for record in records:
2394
self.assertSubset([record.key],
2395
(key_basis, key_missing, key_basis_2, key))
2396
if record.key == key_missing:
2397
self.assertIsInstance(record, AbsentContentFactory)
2399
results.append((record.key, record.sha1, record.storage_kind,
2400
record.get_bytes_as(record.storage_kind)))
2401
calls = list(basis.calls)
2402
order = [record[0] for record in results]
2403
self.assertEqual([key_basis_2, key_basis, key], order)
2404
for result in results:
2405
if result[0] == key:
2409
record = source.get_record_stream([result[0]], 'unordered',
2411
self.assertEqual(record.key, result[0])
2412
self.assertEqual(record.sha1, result[1])
2413
self.assertEqual(record.storage_kind, result[2])
2414
self.assertEqual(record.get_bytes_as(record.storage_kind), result[3])
2415
# It's not strictly minimal, but it seems reasonable for now for it to
2416
# ask which fallbacks have which parents.
2418
("get_parent_map", set([key_basis, key_basis_2, key_missing])),
2419
("get_record_stream", [key_basis_2, key_basis], 'topological', False)],
2422
def test_get_sha1s(self):
2423
# sha1's in the test knit are answered without asking the basis
2424
basis, test = self.get_basis_and_test_knit()
2426
key_basis = ('bar',)
2427
key_missing = ('missing',)
2428
test.add_lines(key, (), ['foo\n'])
2429
key_sha1sum = osutils.sha('foo\n').hexdigest()
2430
sha1s = test.get_sha1s([key])
2431
self.assertEqual({key: key_sha1sum}, sha1s)
2432
self.assertEqual([], basis.calls)
2433
# But texts that are not in the test knit are looked for in the basis
2434
# directly (rather than via text reconstruction) so that remote servers
2435
# etc don't have to answer with full content.
2436
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2437
basis_sha1sum = osutils.sha('foo\nbar\n').hexdigest()
2439
sha1s = test.get_sha1s([key, key_missing, key_basis])
2440
self.assertEqual({key: key_sha1sum,
2441
key_basis: basis_sha1sum}, sha1s)
2442
self.assertEqual([("get_sha1s", set([key_basis, key_missing]))],
2445
def test_insert_record_stream(self):
2446
# records are inserted as normal; insert_record_stream builds on
2447
# add_lines, so a smoke test should be all that's needed:
2449
key_basis = ('bar',)
2450
key_delta = ('zaphod',)
2451
basis, test = self.get_basis_and_test_knit()
2452
source = self.make_test_knit(name='source')
2453
basis.add_lines(key_basis, (), ['foo\n'])
2455
source.add_lines(key_basis, (), ['foo\n'])
2456
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2457
stream = source.get_record_stream([key_delta], 'unordered', False)
2458
test.insert_record_stream(stream)
2459
# XXX: this does somewhat too many calls in making sure of whether it
2460
# has to recreate the full text.
2461
self.assertEqual([("get_parent_map", set([key_basis])),
2462
('get_parent_map', set([key_basis])),
2463
('get_record_stream', [key_basis], 'unordered', True)],
2465
self.assertEqual({key_delta:(key_basis,)},
2466
test.get_parent_map([key_delta]))
2467
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2468
'unordered', True).next().get_bytes_as('fulltext'))
2470
def test_iter_lines_added_or_present_in_keys(self):
2471
# Lines from the basis are returned, and lines for a given key are only
2475
# all sources are asked for keys:
2476
basis, test = self.get_basis_and_test_knit()
2477
basis.add_lines(key1, (), ["foo"])
2479
lines = list(test.iter_lines_added_or_present_in_keys([key1]))
2480
self.assertEqual([("foo\n", key1)], lines)
2481
self.assertEqual([("iter_lines_added_or_present_in_keys", set([key1]))],
2483
# keys in both are not duplicated:
2484
test.add_lines(key2, (), ["bar\n"])
2485
basis.add_lines(key2, (), ["bar\n"])
2487
lines = list(test.iter_lines_added_or_present_in_keys([key2]))
2488
self.assertEqual([("bar\n", key2)], lines)
2489
self.assertEqual([], basis.calls)
2491
def test_keys(self):
2494
# all sources are asked for keys:
2495
basis, test = self.get_basis_and_test_knit()
2497
self.assertEqual(set(), set(keys))
2498
self.assertEqual([("keys",)], basis.calls)
2499
# keys from a basis are returned:
2500
basis.add_lines(key1, (), [])
2503
self.assertEqual(set([key1]), set(keys))
2504
self.assertEqual([("keys",)], basis.calls)
2505
# keys in both are not duplicated:
2506
test.add_lines(key2, (), [])
2507
basis.add_lines(key2, (), [])
2510
self.assertEqual(2, len(keys))
2511
self.assertEqual(set([key1, key2]), set(keys))
2512
self.assertEqual([("keys",)], basis.calls)
2514
def test_add_mpdiffs(self):
2515
# records are inserted as normal; add_mpdiff builds on
2516
# add_lines, so a smoke test should be all that's needed:
2518
key_basis = ('bar',)
2519
key_delta = ('zaphod',)
2520
basis, test = self.get_basis_and_test_knit()
2521
source = self.make_test_knit(name='source')
2522
basis.add_lines(key_basis, (), ['foo\n'])
2524
source.add_lines(key_basis, (), ['foo\n'])
2525
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2526
diffs = source.make_mpdiffs([key_delta])
2527
test.add_mpdiffs([(key_delta, (key_basis,),
2528
source.get_sha1s([key_delta])[key_delta], diffs[0])])
2529
self.assertEqual([("get_parent_map", set([key_basis])),
2530
('get_record_stream', [key_basis], 'unordered', True),],
2532
self.assertEqual({key_delta:(key_basis,)},
2533
test.get_parent_map([key_delta]))
2534
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2535
'unordered', True).next().get_bytes_as('fulltext'))
2537
def test_make_mpdiffs(self):
2538
# Generating an mpdiff across a stacking boundary should detect parent
2542
key_right = ('zaphod',)
2543
basis, test = self.get_basis_and_test_knit()
2544
basis.add_lines(key_left, (), ['bar\n'])
2545
basis.add_lines(key_right, (), ['zaphod\n'])
2547
test.add_lines(key, (key_left, key_right),
2548
['bar\n', 'foo\n', 'zaphod\n'])
2549
diffs = test.make_mpdiffs([key])
2551
multiparent.MultiParent([multiparent.ParentText(0, 0, 0, 1),
2552
multiparent.NewText(['foo\n']),
2553
multiparent.ParentText(1, 0, 2, 1)])],
2555
self.assertEqual(3, len(basis.calls))
2557
("get_parent_map", set([key_left, key_right])),
2558
("get_parent_map", set([key_left, key_right])),
2561
last_call = basis.calls[-1]
2562
self.assertEqual('get_record_stream', last_call[0])
2563
self.assertEqual(set([key_left, key_right]), set(last_call[1]))
2564
self.assertEqual('topological', last_call[2])
2565
self.assertEqual(True, last_call[3])
2568
class TestNetworkBehaviour(KnitTests):
2569
"""Tests for getting data out of/into knits over the network."""
2571
def test_include_delta_closure_generates_a_knit_delta_closure(self):
2572
vf = self.make_test_knit(name='test')
2573
# put in three texts, giving ft, delta, delta
2574
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2575
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2576
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2577
# But heuristics could interfere, so check what happened:
2578
self.assertEqual(['knit-ft-gz', 'knit-delta-gz', 'knit-delta-gz'],
2579
[record.storage_kind for record in
2580
vf.get_record_stream([('base',), ('d1',), ('d2',)],
2581
'topological', False)])
2582
# generate a stream of just the deltas include_delta_closure=True,
2583
# serialise to the network, and check that we get a delta closure on the wire.
2584
stream = vf.get_record_stream([('d1',), ('d2',)], 'topological', True)
2585
netb = [record.get_bytes_as(record.storage_kind) for record in stream]
2586
# The first bytes should be a memo from _ContentMapGenerator, and the
2587
# second bytes should be empty (because its a API proxy not something
2588
# for wire serialisation.
2589
self.assertEqual('', netb[1])
2591
kind, line_end = network_bytes_to_kind_and_offset(bytes)
2592
self.assertEqual('knit-delta-closure', kind)
2595
class TestContentMapGenerator(KnitTests):
2596
"""Tests for ContentMapGenerator"""
2598
def test_get_record_stream_gives_records(self):
2599
vf = self.make_test_knit(name='test')
2600
# put in three texts, giving ft, delta, delta
2601
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2602
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2603
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2604
keys = [('d1',), ('d2',)]
2605
generator = _VFContentMapGenerator(vf, keys,
2606
global_map=vf.get_parent_map(keys))
2607
for record in generator.get_record_stream():
2608
if record.key == ('d1',):
2609
self.assertEqual('d1\n', record.get_bytes_as('fulltext'))
2611
self.assertEqual('d2\n', record.get_bytes_as('fulltext'))
2613
def test_get_record_stream_kinds_are_raw(self):
2614
vf = self.make_test_knit(name='test')
2615
# put in three texts, giving ft, delta, delta
2616
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2617
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2618
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2619
keys = [('base',), ('d1',), ('d2',)]
2620
generator = _VFContentMapGenerator(vf, keys,
2621
global_map=vf.get_parent_map(keys))
2622
kinds = {('base',): 'knit-delta-closure',
2623
('d1',): 'knit-delta-closure-ref',
2624
('d2',): 'knit-delta-closure-ref',
2626
for record in generator.get_record_stream():
2627
self.assertEqual(kinds[record.key], record.storage_kind)
added
removed
bzrlib/tests/test_knit.py
