1
# Copyright (C) 2006-2011 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
"""Tests for Knit data structure"""
19
from cStringIO import StringIO
32
from bzrlib.errors import (
36
from bzrlib.index import *
37
from bzrlib.knit import (
42
_VFContentMapGenerator,
49
from bzrlib.patiencediff import PatienceSequenceMatcher
50
from bzrlib.repofmt import pack_repo
51
from bzrlib.tests import (
53
TestCaseWithMemoryTransport,
54
TestCaseWithTransport,
57
from bzrlib.versionedfile import (
60
network_bytes_to_kind_and_offset,
61
RecordingVersionedFilesDecorator,
65
compiled_knit_feature = tests.ModuleAvailableFeature(
66
'bzrlib._knit_load_data_pyx')
69
class KnitContentTestsMixin(object):
71
def test_constructor(self):
72
content = self._make_content([])
75
content = self._make_content([])
76
self.assertEqual(content.text(), [])
78
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
79
self.assertEqual(content.text(), ["text1", "text2"])
82
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
84
self.assertIsInstance(copy, content.__class__)
85
self.assertEqual(copy.annotate(), content.annotate())
87
def assertDerivedBlocksEqual(self, source, target, noeol=False):
88
"""Assert that the derived matching blocks match real output"""
89
source_lines = source.splitlines(True)
90
target_lines = target.splitlines(True)
92
if noeol and not line.endswith('\n'):
96
source_content = self._make_content([(None, nl(l)) for l in source_lines])
97
target_content = self._make_content([(None, nl(l)) for l in target_lines])
98
line_delta = source_content.line_delta(target_content)
99
delta_blocks = list(KnitContent.get_line_delta_blocks(line_delta,
100
source_lines, target_lines))
101
matcher = PatienceSequenceMatcher(None, source_lines, target_lines)
102
matcher_blocks = list(matcher.get_matching_blocks())
103
self.assertEqual(matcher_blocks, delta_blocks)
105
def test_get_line_delta_blocks(self):
106
self.assertDerivedBlocksEqual('a\nb\nc\n', 'q\nc\n')
107
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1)
108
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1A)
109
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1B)
110
self.assertDerivedBlocksEqual(TEXT_1B, TEXT_1A)
111
self.assertDerivedBlocksEqual(TEXT_1A, TEXT_1B)
112
self.assertDerivedBlocksEqual(TEXT_1A, '')
113
self.assertDerivedBlocksEqual('', TEXT_1A)
114
self.assertDerivedBlocksEqual('', '')
115
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd')
117
def test_get_line_delta_blocks_noeol(self):
118
"""Handle historical knit deltas safely
120
Some existing knit deltas don't consider the last line to differ
121
when the only difference whether it has a final newline.
123
New knit deltas appear to always consider the last line to differ
126
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd\n', noeol=True)
127
self.assertDerivedBlocksEqual('a\nb\nc\nd\n', 'a\nb\nc', noeol=True)
128
self.assertDerivedBlocksEqual('a\nb\nc\n', 'a\nb\nc', noeol=True)
129
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\n', noeol=True)
141
Banana cup cake recipe
151
Banana cup cake recipe
153
- bananas (do not use plantains!!!)
160
Banana cup cake recipe
177
class TestPlainKnitContent(TestCase, KnitContentTestsMixin):
179
def _make_content(self, lines):
180
annotated_content = AnnotatedKnitContent(lines)
181
return PlainKnitContent(annotated_content.text(), 'bogus')
183
def test_annotate(self):
184
content = self._make_content([])
185
self.assertEqual(content.annotate(), [])
187
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
188
self.assertEqual(content.annotate(),
189
[("bogus", "text1"), ("bogus", "text2")])
191
def test_line_delta(self):
192
content1 = self._make_content([("", "a"), ("", "b")])
193
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
194
self.assertEqual(content1.line_delta(content2),
195
[(1, 2, 2, ["a", "c"])])
197
def test_line_delta_iter(self):
198
content1 = self._make_content([("", "a"), ("", "b")])
199
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
200
it = content1.line_delta_iter(content2)
201
self.assertEqual(it.next(), (1, 2, 2, ["a", "c"]))
202
self.assertRaises(StopIteration, it.next)
205
class TestAnnotatedKnitContent(TestCase, KnitContentTestsMixin):
207
def _make_content(self, lines):
208
return AnnotatedKnitContent(lines)
210
def test_annotate(self):
211
content = self._make_content([])
212
self.assertEqual(content.annotate(), [])
214
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
215
self.assertEqual(content.annotate(),
216
[("origin1", "text1"), ("origin2", "text2")])
218
def test_line_delta(self):
219
content1 = self._make_content([("", "a"), ("", "b")])
220
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
221
self.assertEqual(content1.line_delta(content2),
222
[(1, 2, 2, [("", "a"), ("", "c")])])
224
def test_line_delta_iter(self):
225
content1 = self._make_content([("", "a"), ("", "b")])
226
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
227
it = content1.line_delta_iter(content2)
228
self.assertEqual(it.next(), (1, 2, 2, [("", "a"), ("", "c")]))
229
self.assertRaises(StopIteration, it.next)
232
class MockTransport(object):
234
def __init__(self, file_lines=None):
235
self.file_lines = file_lines
237
# We have no base directory for the MockTransport
240
def get(self, filename):
241
if self.file_lines is None:
242
raise NoSuchFile(filename)
244
return StringIO("\n".join(self.file_lines))
246
def readv(self, relpath, offsets):
247
fp = self.get(relpath)
248
for offset, size in offsets:
250
yield offset, fp.read(size)
252
def __getattr__(self, name):
253
def queue_call(*args, **kwargs):
254
self.calls.append((name, args, kwargs))
258
class MockReadvFailingTransport(MockTransport):
259
"""Fail in the middle of a readv() result.
261
This Transport will successfully yield the first two requested hunks, but
262
raise NoSuchFile for the rest.
265
def readv(self, relpath, offsets):
267
for result in MockTransport.readv(self, relpath, offsets):
269
# we use 2 because the first offset is the pack header, the second
270
# is the first actual content requset
272
raise errors.NoSuchFile(relpath)
276
class KnitRecordAccessTestsMixin(object):
277
"""Tests for getting and putting knit records."""
279
def test_add_raw_records(self):
280
"""Add_raw_records adds records retrievable later."""
281
access = self.get_access()
282
memos = access.add_raw_records([('key', 10)], '1234567890')
283
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
285
def test_add_several_raw_records(self):
286
"""add_raw_records with many records and read some back."""
287
access = self.get_access()
288
memos = access.add_raw_records([('key', 10), ('key2', 2), ('key3', 5)],
290
self.assertEqual(['1234567890', '12', '34567'],
291
list(access.get_raw_records(memos)))
292
self.assertEqual(['1234567890'],
293
list(access.get_raw_records(memos[0:1])))
294
self.assertEqual(['12'],
295
list(access.get_raw_records(memos[1:2])))
296
self.assertEqual(['34567'],
297
list(access.get_raw_records(memos[2:3])))
298
self.assertEqual(['1234567890', '34567'],
299
list(access.get_raw_records(memos[0:1] + memos[2:3])))
302
class TestKnitKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
303
"""Tests for the .kndx implementation."""
305
def get_access(self):
306
"""Get a .knit style access instance."""
307
mapper = ConstantMapper("foo")
308
access = _KnitKeyAccess(self.get_transport(), mapper)
312
class _TestException(Exception):
313
"""Just an exception for local tests to use."""
316
class TestPackKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
317
"""Tests for the pack based access."""
319
def get_access(self):
320
return self._get_access()[0]
322
def _get_access(self, packname='packfile', index='FOO'):
323
transport = self.get_transport()
324
def write_data(bytes):
325
transport.append_bytes(packname, bytes)
326
writer = pack.ContainerWriter(write_data)
328
access = pack_repo._DirectPackAccess({})
329
access.set_writer(writer, index, (transport, packname))
330
return access, writer
332
def make_pack_file(self):
333
"""Create a pack file with 2 records."""
334
access, writer = self._get_access(packname='packname', index='foo')
336
memos.extend(access.add_raw_records([('key1', 10)], '1234567890'))
337
memos.extend(access.add_raw_records([('key2', 5)], '12345'))
341
def test_pack_collection_pack_retries(self):
342
"""An explicit pack of a pack collection succeeds even when a
343
concurrent pack happens.
345
builder = self.make_branch_builder('.')
346
builder.start_series()
347
builder.build_snapshot('rev-1', None, [
348
('add', ('', 'root-id', 'directory', None)),
349
('add', ('file', 'file-id', 'file', 'content\nrev 1\n')),
351
builder.build_snapshot('rev-2', ['rev-1'], [
352
('modify', ('file-id', 'content\nrev 2\n')),
354
builder.build_snapshot('rev-3', ['rev-2'], [
355
('modify', ('file-id', 'content\nrev 3\n')),
357
self.addCleanup(builder.finish_series)
358
b = builder.get_branch()
359
self.addCleanup(b.lock_write().unlock)
361
collection = repo._pack_collection
362
# Concurrently repack the repo.
363
reopened_repo = repo.bzrdir.open_repository()
368
def make_vf_for_retrying(self):
369
"""Create 3 packs and a reload function.
371
Originally, 2 pack files will have the data, but one will be missing.
372
And then the third will be used in place of the first two if reload()
375
:return: (versioned_file, reload_counter)
376
versioned_file a KnitVersionedFiles using the packs for access
378
builder = self.make_branch_builder('.', format="1.9")
379
builder.start_series()
380
builder.build_snapshot('rev-1', None, [
381
('add', ('', 'root-id', 'directory', None)),
382
('add', ('file', 'file-id', 'file', 'content\nrev 1\n')),
384
builder.build_snapshot('rev-2', ['rev-1'], [
385
('modify', ('file-id', 'content\nrev 2\n')),
387
builder.build_snapshot('rev-3', ['rev-2'], [
388
('modify', ('file-id', 'content\nrev 3\n')),
390
builder.finish_series()
391
b = builder.get_branch()
393
self.addCleanup(b.unlock)
394
# Pack these three revisions into another pack file, but don't remove
397
collection = repo._pack_collection
398
collection.ensure_loaded()
399
orig_packs = collection.packs
400
packer = pack_repo.Packer(collection, orig_packs, '.testpack')
401
new_pack = packer.pack()
402
# forget about the new pack
406
# Set up a reload() function that switches to using the new pack file
407
new_index = new_pack.revision_index
408
access_tuple = new_pack.access_tuple()
409
reload_counter = [0, 0, 0]
411
reload_counter[0] += 1
412
if reload_counter[1] > 0:
413
# We already reloaded, nothing more to do
414
reload_counter[2] += 1
416
reload_counter[1] += 1
417
vf._index._graph_index._indices[:] = [new_index]
418
vf._access._indices.clear()
419
vf._access._indices[new_index] = access_tuple
421
# Delete one of the pack files so the data will need to be reloaded. We
422
# will delete the file with 'rev-2' in it
423
trans, name = orig_packs[1].access_tuple()
425
# We don't have the index trigger reloading because we want to test
426
# that we reload when the .pack disappears
427
vf._access._reload_func = reload
428
return vf, reload_counter
430
def make_reload_func(self, return_val=True):
433
reload_called[0] += 1
435
return reload_called, reload
437
def make_retry_exception(self):
438
# We raise a real exception so that sys.exc_info() is properly
441
raise _TestException('foobar')
442
except _TestException, e:
443
retry_exc = errors.RetryWithNewPacks(None, reload_occurred=False,
444
exc_info=sys.exc_info())
447
def test_read_from_several_packs(self):
448
access, writer = self._get_access()
450
memos.extend(access.add_raw_records([('key', 10)], '1234567890'))
452
access, writer = self._get_access('pack2', 'FOOBAR')
453
memos.extend(access.add_raw_records([('key', 5)], '12345'))
455
access, writer = self._get_access('pack3', 'BAZ')
456
memos.extend(access.add_raw_records([('key', 5)], 'alpha'))
458
transport = self.get_transport()
459
access = pack_repo._DirectPackAccess({"FOO":(transport, 'packfile'),
460
"FOOBAR":(transport, 'pack2'),
461
"BAZ":(transport, 'pack3')})
462
self.assertEqual(['1234567890', '12345', 'alpha'],
463
list(access.get_raw_records(memos)))
464
self.assertEqual(['1234567890'],
465
list(access.get_raw_records(memos[0:1])))
466
self.assertEqual(['12345'],
467
list(access.get_raw_records(memos[1:2])))
468
self.assertEqual(['alpha'],
469
list(access.get_raw_records(memos[2:3])))
470
self.assertEqual(['1234567890', 'alpha'],
471
list(access.get_raw_records(memos[0:1] + memos[2:3])))
473
def test_set_writer(self):
474
"""The writer should be settable post construction."""
475
access = pack_repo._DirectPackAccess({})
476
transport = self.get_transport()
477
packname = 'packfile'
479
def write_data(bytes):
480
transport.append_bytes(packname, bytes)
481
writer = pack.ContainerWriter(write_data)
483
access.set_writer(writer, index, (transport, packname))
484
memos = access.add_raw_records([('key', 10)], '1234567890')
486
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
488
def test_missing_index_raises_retry(self):
489
memos = self.make_pack_file()
490
transport = self.get_transport()
491
reload_called, reload_func = self.make_reload_func()
492
# Note that the index key has changed from 'foo' to 'bar'
493
access = pack_repo._DirectPackAccess({'bar':(transport, 'packname')},
494
reload_func=reload_func)
495
e = self.assertListRaises(errors.RetryWithNewPacks,
496
access.get_raw_records, memos)
497
# Because a key was passed in which does not match our index list, we
498
# assume that the listing was already reloaded
499
self.assertTrue(e.reload_occurred)
500
self.assertIsInstance(e.exc_info, tuple)
501
self.assertIs(e.exc_info[0], KeyError)
502
self.assertIsInstance(e.exc_info[1], KeyError)
504
def test_missing_index_raises_key_error_with_no_reload(self):
505
memos = self.make_pack_file()
506
transport = self.get_transport()
507
# Note that the index key has changed from 'foo' to 'bar'
508
access = pack_repo._DirectPackAccess({'bar':(transport, 'packname')})
509
e = self.assertListRaises(KeyError, access.get_raw_records, memos)
511
def test_missing_file_raises_retry(self):
512
memos = self.make_pack_file()
513
transport = self.get_transport()
514
reload_called, reload_func = self.make_reload_func()
515
# Note that the 'filename' has been changed to 'different-packname'
516
access = pack_repo._DirectPackAccess(
517
{'foo':(transport, 'different-packname')},
518
reload_func=reload_func)
519
e = self.assertListRaises(errors.RetryWithNewPacks,
520
access.get_raw_records, memos)
521
# The file has gone missing, so we assume we need to reload
522
self.assertFalse(e.reload_occurred)
523
self.assertIsInstance(e.exc_info, tuple)
524
self.assertIs(e.exc_info[0], errors.NoSuchFile)
525
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
526
self.assertEqual('different-packname', e.exc_info[1].path)
528
def test_missing_file_raises_no_such_file_with_no_reload(self):
529
memos = self.make_pack_file()
530
transport = self.get_transport()
531
# Note that the 'filename' has been changed to 'different-packname'
532
access = pack_repo._DirectPackAccess(
533
{'foo': (transport, 'different-packname')})
534
e = self.assertListRaises(errors.NoSuchFile,
535
access.get_raw_records, memos)
537
def test_failing_readv_raises_retry(self):
538
memos = self.make_pack_file()
539
transport = self.get_transport()
540
failing_transport = MockReadvFailingTransport(
541
[transport.get_bytes('packname')])
542
reload_called, reload_func = self.make_reload_func()
543
access = pack_repo._DirectPackAccess(
544
{'foo': (failing_transport, 'packname')},
545
reload_func=reload_func)
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.RetryWithNewPacks,
553
access.get_raw_records, memos)
554
# The file has gone missing, so we assume we need to reload
555
self.assertFalse(e.reload_occurred)
556
self.assertIsInstance(e.exc_info, tuple)
557
self.assertIs(e.exc_info[0], errors.NoSuchFile)
558
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
559
self.assertEqual('packname', e.exc_info[1].path)
561
def test_failing_readv_raises_no_such_file_with_no_reload(self):
562
memos = self.make_pack_file()
563
transport = self.get_transport()
564
failing_transport = MockReadvFailingTransport(
565
[transport.get_bytes('packname')])
566
reload_called, reload_func = self.make_reload_func()
567
access = pack_repo._DirectPackAccess(
568
{'foo':(failing_transport, 'packname')})
569
# Asking for a single record will not trigger the Mock failure
570
self.assertEqual(['1234567890'],
571
list(access.get_raw_records(memos[:1])))
572
self.assertEqual(['12345'],
573
list(access.get_raw_records(memos[1:2])))
574
# A multiple offset readv() will fail mid-way through
575
e = self.assertListRaises(errors.NoSuchFile,
576
access.get_raw_records, memos)
578
def test_reload_or_raise_no_reload(self):
579
access = pack_repo._DirectPackAccess({}, reload_func=None)
580
retry_exc = self.make_retry_exception()
581
# Without a reload_func, we will just re-raise the original exception
582
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
584
def test_reload_or_raise_reload_changed(self):
585
reload_called, reload_func = self.make_reload_func(return_val=True)
586
access = pack_repo._DirectPackAccess({}, reload_func=reload_func)
587
retry_exc = self.make_retry_exception()
588
access.reload_or_raise(retry_exc)
589
self.assertEqual([1], reload_called)
590
retry_exc.reload_occurred=True
591
access.reload_or_raise(retry_exc)
592
self.assertEqual([2], reload_called)
594
def test_reload_or_raise_reload_no_change(self):
595
reload_called, reload_func = self.make_reload_func(return_val=False)
596
access = pack_repo._DirectPackAccess({}, reload_func=reload_func)
597
retry_exc = self.make_retry_exception()
598
# If reload_occurred is False, then we consider it an error to have
599
# reload_func() return False (no changes).
600
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
601
self.assertEqual([1], reload_called)
602
retry_exc.reload_occurred=True
603
# If reload_occurred is True, then we assume nothing changed because
604
# it had changed earlier, but didn't change again
605
access.reload_or_raise(retry_exc)
606
self.assertEqual([2], reload_called)
608
def test_annotate_retries(self):
609
vf, reload_counter = self.make_vf_for_retrying()
610
# It is a little bit bogus to annotate the Revision VF, but it works,
611
# as we have ancestry stored there
613
reload_lines = vf.annotate(key)
614
self.assertEqual([1, 1, 0], reload_counter)
615
plain_lines = vf.annotate(key)
616
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
617
if reload_lines != plain_lines:
618
self.fail('Annotation was not identical with reloading.')
619
# Now delete the packs-in-use, which should trigger another reload, but
620
# this time we just raise an exception because we can't recover
621
for trans, name in vf._access._indices.itervalues():
623
self.assertRaises(errors.NoSuchFile, vf.annotate, key)
624
self.assertEqual([2, 1, 1], reload_counter)
626
def test__get_record_map_retries(self):
627
vf, reload_counter = self.make_vf_for_retrying()
628
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
629
records = vf._get_record_map(keys)
630
self.assertEqual(keys, sorted(records.keys()))
631
self.assertEqual([1, 1, 0], reload_counter)
632
# Now delete the packs-in-use, which should trigger another reload, but
633
# this time we just raise an exception because we can't recover
634
for trans, name in vf._access._indices.itervalues():
636
self.assertRaises(errors.NoSuchFile, vf._get_record_map, keys)
637
self.assertEqual([2, 1, 1], reload_counter)
639
def test_get_record_stream_retries(self):
640
vf, reload_counter = self.make_vf_for_retrying()
641
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
642
record_stream = vf.get_record_stream(keys, 'topological', False)
643
record = record_stream.next()
644
self.assertEqual(('rev-1',), record.key)
645
self.assertEqual([0, 0, 0], reload_counter)
646
record = record_stream.next()
647
self.assertEqual(('rev-2',), record.key)
648
self.assertEqual([1, 1, 0], reload_counter)
649
record = record_stream.next()
650
self.assertEqual(('rev-3',), record.key)
651
self.assertEqual([1, 1, 0], reload_counter)
652
# Now delete all pack files, and see that we raise the right error
653
for trans, name in vf._access._indices.itervalues():
655
self.assertListRaises(errors.NoSuchFile,
656
vf.get_record_stream, keys, 'topological', False)
658
def test_iter_lines_added_or_present_in_keys_retries(self):
659
vf, reload_counter = self.make_vf_for_retrying()
660
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
661
# Unfortunately, iter_lines_added_or_present_in_keys iterates the
662
# result in random order (determined by the iteration order from a
663
# set()), so we don't have any solid way to trigger whether data is
664
# read before or after. However we tried to delete the middle node to
665
# exercise the code well.
666
# What we care about is that all lines are always yielded, but not
669
reload_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
670
self.assertEqual([1, 1, 0], reload_counter)
671
# Now do it again, to make sure the result is equivalent
672
plain_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
673
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
674
self.assertEqual(plain_lines, reload_lines)
675
self.assertEqual(21, len(plain_lines))
676
# Now delete all pack files, and see that we raise the right error
677
for trans, name in vf._access._indices.itervalues():
679
self.assertListRaises(errors.NoSuchFile,
680
vf.iter_lines_added_or_present_in_keys, keys)
681
self.assertEqual([2, 1, 1], reload_counter)
683
def test_get_record_stream_yields_disk_sorted_order(self):
684
# if we get 'unordered' pick a semi-optimal order for reading. The
685
# order should be grouped by pack file, and then by position in file
686
repo = self.make_repository('test', format='pack-0.92')
688
self.addCleanup(repo.unlock)
689
repo.start_write_group()
691
vf.add_lines(('f-id', 'rev-5'), [('f-id', 'rev-4')], ['lines\n'])
692
vf.add_lines(('f-id', 'rev-1'), [], ['lines\n'])
693
vf.add_lines(('f-id', 'rev-2'), [('f-id', 'rev-1')], ['lines\n'])
694
repo.commit_write_group()
695
# We inserted them as rev-5, rev-1, rev-2, we should get them back in
697
stream = vf.get_record_stream([('f-id', 'rev-1'), ('f-id', 'rev-5'),
698
('f-id', 'rev-2')], 'unordered', False)
699
keys = [r.key for r in stream]
700
self.assertEqual([('f-id', 'rev-5'), ('f-id', 'rev-1'),
701
('f-id', 'rev-2')], keys)
702
repo.start_write_group()
703
vf.add_lines(('f-id', 'rev-4'), [('f-id', 'rev-3')], ['lines\n'])
704
vf.add_lines(('f-id', 'rev-3'), [('f-id', 'rev-2')], ['lines\n'])
705
vf.add_lines(('f-id', 'rev-6'), [('f-id', 'rev-5')], ['lines\n'])
706
repo.commit_write_group()
707
# Request in random order, to make sure the output order isn't based on
709
request_keys = set(('f-id', 'rev-%d' % i) for i in range(1, 7))
710
stream = vf.get_record_stream(request_keys, 'unordered', False)
711
keys = [r.key for r in stream]
712
# We want to get the keys back in disk order, but it doesn't matter
713
# which pack we read from first. So this can come back in 2 orders
714
alt1 = [('f-id', 'rev-%d' % i) for i in [4, 3, 6, 5, 1, 2]]
715
alt2 = [('f-id', 'rev-%d' % i) for i in [5, 1, 2, 4, 3, 6]]
716
if keys != alt1 and keys != alt2:
717
self.fail('Returned key order did not match either expected order.'
718
' expected %s or %s, not %s'
719
% (alt1, alt2, keys))
722
class LowLevelKnitDataTests(TestCase):
724
def create_gz_content(self, text):
726
gz_file = gzip.GzipFile(mode='wb', fileobj=sio)
729
return sio.getvalue()
731
def make_multiple_records(self):
732
"""Create the content for multiple records."""
733
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
735
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
740
record_1 = (0, len(gz_txt), sha1sum)
741
total_txt.append(gz_txt)
742
sha1sum = osutils.sha('baz\n').hexdigest()
743
gz_txt = self.create_gz_content('version rev-id-2 1 %s\n'
747
record_2 = (record_1[1], len(gz_txt), sha1sum)
748
total_txt.append(gz_txt)
749
return total_txt, record_1, record_2
751
def test_valid_knit_data(self):
752
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
753
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
758
transport = MockTransport([gz_txt])
759
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
760
knit = KnitVersionedFiles(None, access)
761
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
763
contents = list(knit._read_records_iter(records))
764
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'],
765
'4e48e2c9a3d2ca8a708cb0cc545700544efb5021')], contents)
767
raw_contents = list(knit._read_records_iter_raw(records))
768
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
770
def test_multiple_records_valid(self):
771
total_txt, record_1, record_2 = self.make_multiple_records()
772
transport = MockTransport([''.join(total_txt)])
773
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
774
knit = KnitVersionedFiles(None, access)
775
records = [(('rev-id-1',), (('rev-id-1',), record_1[0], record_1[1])),
776
(('rev-id-2',), (('rev-id-2',), record_2[0], record_2[1]))]
778
contents = list(knit._read_records_iter(records))
779
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'], record_1[2]),
780
(('rev-id-2',), ['baz\n'], record_2[2])],
783
raw_contents = list(knit._read_records_iter_raw(records))
784
self.assertEqual([(('rev-id-1',), total_txt[0], record_1[2]),
785
(('rev-id-2',), total_txt[1], record_2[2])],
788
def test_not_enough_lines(self):
789
sha1sum = osutils.sha('foo\n').hexdigest()
790
# record says 2 lines data says 1
791
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
795
transport = MockTransport([gz_txt])
796
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
797
knit = KnitVersionedFiles(None, access)
798
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
799
self.assertRaises(errors.KnitCorrupt, list,
800
knit._read_records_iter(records))
802
# read_records_iter_raw won't detect that sort of mismatch/corruption
803
raw_contents = list(knit._read_records_iter_raw(records))
804
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
806
def test_too_many_lines(self):
807
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
808
# record says 1 lines data says 2
809
gz_txt = self.create_gz_content('version rev-id-1 1 %s\n'
814
transport = MockTransport([gz_txt])
815
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
816
knit = KnitVersionedFiles(None, access)
817
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
818
self.assertRaises(errors.KnitCorrupt, list,
819
knit._read_records_iter(records))
821
# read_records_iter_raw won't detect that sort of mismatch/corruption
822
raw_contents = list(knit._read_records_iter_raw(records))
823
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
825
def test_mismatched_version_id(self):
826
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
827
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
832
transport = MockTransport([gz_txt])
833
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
834
knit = KnitVersionedFiles(None, access)
835
# We are asking for rev-id-2, but the data is rev-id-1
836
records = [(('rev-id-2',), (('rev-id-2',), 0, len(gz_txt)))]
837
self.assertRaises(errors.KnitCorrupt, list,
838
knit._read_records_iter(records))
840
# read_records_iter_raw detects mismatches in the header
841
self.assertRaises(errors.KnitCorrupt, list,
842
knit._read_records_iter_raw(records))
844
def test_uncompressed_data(self):
845
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
846
txt = ('version rev-id-1 2 %s\n'
851
transport = MockTransport([txt])
852
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
853
knit = KnitVersionedFiles(None, access)
854
records = [(('rev-id-1',), (('rev-id-1',), 0, len(txt)))]
856
# We don't have valid gzip data ==> corrupt
857
self.assertRaises(errors.KnitCorrupt, list,
858
knit._read_records_iter(records))
860
# read_records_iter_raw will notice the bad data
861
self.assertRaises(errors.KnitCorrupt, list,
862
knit._read_records_iter_raw(records))
864
def test_corrupted_data(self):
865
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
866
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
871
# Change 2 bytes in the middle to \xff
872
gz_txt = gz_txt[:10] + '\xff\xff' + gz_txt[12:]
873
transport = MockTransport([gz_txt])
874
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
875
knit = KnitVersionedFiles(None, access)
876
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
877
self.assertRaises(errors.KnitCorrupt, list,
878
knit._read_records_iter(records))
879
# read_records_iter_raw will barf on bad gz data
880
self.assertRaises(errors.KnitCorrupt, list,
881
knit._read_records_iter_raw(records))
884
class LowLevelKnitIndexTests(TestCase):
886
def get_knit_index(self, transport, name, mode):
887
mapper = ConstantMapper(name)
888
from bzrlib._knit_load_data_py import _load_data_py
889
self.overrideAttr(knit, '_load_data', _load_data_py)
890
allow_writes = lambda: 'w' in mode
891
return _KndxIndex(transport, mapper, lambda:None, allow_writes, lambda:True)
893
def test_create_file(self):
894
transport = MockTransport()
895
index = self.get_knit_index(transport, "filename", "w")
897
call = transport.calls.pop(0)
898
# call[1][1] is a StringIO - we can't test it by simple equality.
899
self.assertEqual('put_file_non_atomic', call[0])
900
self.assertEqual('filename.kndx', call[1][0])
901
# With no history, _KndxIndex writes a new index:
902
self.assertEqual(_KndxIndex.HEADER,
903
call[1][1].getvalue())
904
self.assertEqual({'create_parent_dir': True}, call[2])
906
def test_read_utf8_version_id(self):
907
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
908
utf8_revision_id = unicode_revision_id.encode('utf-8')
909
transport = MockTransport([
911
'%s option 0 1 :' % (utf8_revision_id,)
913
index = self.get_knit_index(transport, "filename", "r")
914
# _KndxIndex is a private class, and deals in utf8 revision_ids, not
915
# Unicode revision_ids.
916
self.assertEqual({(utf8_revision_id,):()},
917
index.get_parent_map(index.keys()))
918
self.assertFalse((unicode_revision_id,) in index.keys())
920
def test_read_utf8_parents(self):
921
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
922
utf8_revision_id = unicode_revision_id.encode('utf-8')
923
transport = MockTransport([
925
"version option 0 1 .%s :" % (utf8_revision_id,)
927
index = self.get_knit_index(transport, "filename", "r")
928
self.assertEqual({("version",):((utf8_revision_id,),)},
929
index.get_parent_map(index.keys()))
931
def test_read_ignore_corrupted_lines(self):
932
transport = MockTransport([
935
"corrupted options 0 1 .b .c ",
936
"version options 0 1 :"
938
index = self.get_knit_index(transport, "filename", "r")
939
self.assertEqual(1, len(index.keys()))
940
self.assertEqual(set([("version",)]), index.keys())
942
def test_read_corrupted_header(self):
943
transport = MockTransport(['not a bzr knit index header\n'])
944
index = self.get_knit_index(transport, "filename", "r")
945
self.assertRaises(KnitHeaderError, index.keys)
947
def test_read_duplicate_entries(self):
948
transport = MockTransport([
950
"parent options 0 1 :",
951
"version options1 0 1 0 :",
952
"version options2 1 2 .other :",
953
"version options3 3 4 0 .other :"
955
index = self.get_knit_index(transport, "filename", "r")
956
self.assertEqual(2, len(index.keys()))
957
# check that the index used is the first one written. (Specific
958
# to KnitIndex style indices.
959
self.assertEqual("1", index._dictionary_compress([("version",)]))
960
self.assertEqual((("version",), 3, 4), index.get_position(("version",)))
961
self.assertEqual(["options3"], index.get_options(("version",)))
962
self.assertEqual({("version",):(("parent",), ("other",))},
963
index.get_parent_map([("version",)]))
965
def test_read_compressed_parents(self):
966
transport = MockTransport([
970
"c option 0 1 1 0 :",
972
index = self.get_knit_index(transport, "filename", "r")
973
self.assertEqual({("b",):(("a",),), ("c",):(("b",), ("a",))},
974
index.get_parent_map([("b",), ("c",)]))
976
def test_write_utf8_version_id(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
((utf8_revision_id,), ["option"], ((utf8_revision_id,), 0, 1), [])])
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
"\n%s option 0 1 :" % (utf8_revision_id,),
992
call[1][1].getvalue())
993
self.assertEqual({'create_parent_dir': True}, call[2])
995
def test_write_utf8_parents(self):
996
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
997
utf8_revision_id = unicode_revision_id.encode('utf-8')
998
transport = MockTransport([
1001
index = self.get_knit_index(transport, "filename", "r")
1003
(("version",), ["option"], (("version",), 0, 1), [(utf8_revision_id,)])])
1004
call = transport.calls.pop(0)
1005
# call[1][1] is a StringIO - we can't test it by simple equality.
1006
self.assertEqual('put_file_non_atomic', call[0])
1007
self.assertEqual('filename.kndx', call[1][0])
1008
# With no history, _KndxIndex writes a new index:
1009
self.assertEqual(_KndxIndex.HEADER +
1010
"\nversion option 0 1 .%s :" % (utf8_revision_id,),
1011
call[1][1].getvalue())
1012
self.assertEqual({'create_parent_dir': True}, call[2])
1014
def test_keys(self):
1015
transport = MockTransport([
1018
index = self.get_knit_index(transport, "filename", "r")
1020
self.assertEqual(set(), index.keys())
1022
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1023
self.assertEqual(set([("a",)]), index.keys())
1025
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1026
self.assertEqual(set([("a",)]), index.keys())
1028
index.add_records([(("b",), ["option"], (("b",), 0, 1), [])])
1029
self.assertEqual(set([("a",), ("b",)]), index.keys())
1031
def add_a_b(self, index, random_id=None):
1033
if random_id is not None:
1034
kwargs["random_id"] = random_id
1036
(("a",), ["option"], (("a",), 0, 1), [("b",)]),
1037
(("a",), ["opt"], (("a",), 1, 2), [("c",)]),
1038
(("b",), ["option"], (("b",), 2, 3), [("a",)])
1041
def assertIndexIsAB(self, index):
1046
index.get_parent_map(index.keys()))
1047
self.assertEqual((("a",), 1, 2), index.get_position(("a",)))
1048
self.assertEqual((("b",), 2, 3), index.get_position(("b",)))
1049
self.assertEqual(["opt"], index.get_options(("a",)))
1051
def test_add_versions(self):
1052
transport = MockTransport([
1055
index = self.get_knit_index(transport, "filename", "r")
1058
call = transport.calls.pop(0)
1059
# call[1][1] is a StringIO - we can't test it by simple equality.
1060
self.assertEqual('put_file_non_atomic', call[0])
1061
self.assertEqual('filename.kndx', call[1][0])
1062
# With no history, _KndxIndex writes a new index:
1065
"\na option 0 1 .b :"
1067
"\nb option 2 3 0 :",
1068
call[1][1].getvalue())
1069
self.assertEqual({'create_parent_dir': True}, call[2])
1070
self.assertIndexIsAB(index)
1072
def test_add_versions_random_id_is_accepted(self):
1073
transport = MockTransport([
1076
index = self.get_knit_index(transport, "filename", "r")
1077
self.add_a_b(index, random_id=True)
1079
def test_delay_create_and_add_versions(self):
1080
transport = MockTransport()
1082
index = self.get_knit_index(transport, "filename", "w")
1084
self.assertEqual([], transport.calls)
1087
#[ {"dir_mode": 0777, "create_parent_dir": True, "mode": "wb"},
1089
# Two calls: one during which we load the existing index (and when its
1090
# missing create it), then a second where we write the contents out.
1091
self.assertEqual(2, len(transport.calls))
1092
call = transport.calls.pop(0)
1093
self.assertEqual('put_file_non_atomic', call[0])
1094
self.assertEqual('filename.kndx', call[1][0])
1095
# With no history, _KndxIndex writes a new index:
1096
self.assertEqual(_KndxIndex.HEADER, call[1][1].getvalue())
1097
self.assertEqual({'create_parent_dir': True}, call[2])
1098
call = transport.calls.pop(0)
1099
# call[1][1] is a StringIO - we can't test it by simple equality.
1100
self.assertEqual('put_file_non_atomic', call[0])
1101
self.assertEqual('filename.kndx', call[1][0])
1102
# With no history, _KndxIndex writes a new index:
1105
"\na option 0 1 .b :"
1107
"\nb option 2 3 0 :",
1108
call[1][1].getvalue())
1109
self.assertEqual({'create_parent_dir': True}, call[2])
1111
def assertTotalBuildSize(self, size, keys, positions):
1112
self.assertEqual(size,
1113
knit._get_total_build_size(None, keys, positions))
1115
def test__get_total_build_size(self):
1117
('a',): (('fulltext', False), (('a',), 0, 100), None),
1118
('b',): (('line-delta', False), (('b',), 100, 21), ('a',)),
1119
('c',): (('line-delta', False), (('c',), 121, 35), ('b',)),
1120
('d',): (('line-delta', False), (('d',), 156, 12), ('b',)),
1122
self.assertTotalBuildSize(100, [('a',)], positions)
1123
self.assertTotalBuildSize(121, [('b',)], positions)
1124
# c needs both a & b
1125
self.assertTotalBuildSize(156, [('c',)], positions)
1126
# we shouldn't count 'b' twice
1127
self.assertTotalBuildSize(156, [('b',), ('c',)], positions)
1128
self.assertTotalBuildSize(133, [('d',)], positions)
1129
self.assertTotalBuildSize(168, [('c',), ('d',)], positions)
1131
def test_get_position(self):
1132
transport = MockTransport([
1137
index = self.get_knit_index(transport, "filename", "r")
1139
self.assertEqual((("a",), 0, 1), index.get_position(("a",)))
1140
self.assertEqual((("b",), 1, 2), index.get_position(("b",)))
1142
def test_get_method(self):
1143
transport = MockTransport([
1145
"a fulltext,unknown 0 1 :",
1146
"b unknown,line-delta 1 2 :",
1149
index = self.get_knit_index(transport, "filename", "r")
1151
self.assertEqual("fulltext", index.get_method("a"))
1152
self.assertEqual("line-delta", index.get_method("b"))
1153
self.assertRaises(errors.KnitIndexUnknownMethod, index.get_method, "c")
1155
def test_get_options(self):
1156
transport = MockTransport([
1161
index = self.get_knit_index(transport, "filename", "r")
1163
self.assertEqual(["opt1"], index.get_options("a"))
1164
self.assertEqual(["opt2", "opt3"], index.get_options("b"))
1166
def test_get_parent_map(self):
1167
transport = MockTransport([
1170
"b option 1 2 0 .c :",
1171
"c option 1 2 1 0 .e :"
1173
index = self.get_knit_index(transport, "filename", "r")
1177
("b",):(("a",), ("c",)),
1178
("c",):(("b",), ("a",), ("e",)),
1179
}, index.get_parent_map(index.keys()))
1181
def test_impossible_parent(self):
1182
"""Test we get KnitCorrupt if the parent couldn't possibly exist."""
1183
transport = MockTransport([
1186
"b option 0 1 4 :" # We don't have a 4th record
1188
index = self.get_knit_index(transport, 'filename', 'r')
1190
self.assertRaises(errors.KnitCorrupt, index.keys)
1191
except TypeError, e:
1192
if (str(e) == ('exceptions must be strings, classes, or instances,'
1193
' not exceptions.IndexError')
1194
and sys.version_info[0:2] >= (2,5)):
1195
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1196
' raising new style exceptions with python'
1201
def test_corrupted_parent(self):
1202
transport = MockTransport([
1206
"c option 0 1 1v :", # Can't have a parent of '1v'
1208
index = self.get_knit_index(transport, 'filename', 'r')
1210
self.assertRaises(errors.KnitCorrupt, index.keys)
1211
except TypeError, e:
1212
if (str(e) == ('exceptions must be strings, classes, or instances,'
1213
' not exceptions.ValueError')
1214
and sys.version_info[0:2] >= (2,5)):
1215
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1216
' raising new style exceptions with python'
1221
def test_corrupted_parent_in_list(self):
1222
transport = MockTransport([
1226
"c option 0 1 1 v :", # Can't have a parent of 'v'
1228
index = self.get_knit_index(transport, 'filename', 'r')
1230
self.assertRaises(errors.KnitCorrupt, index.keys)
1231
except TypeError, e:
1232
if (str(e) == ('exceptions must be strings, classes, or instances,'
1233
' not exceptions.ValueError')
1234
and sys.version_info[0:2] >= (2,5)):
1235
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1236
' raising new style exceptions with python'
1241
def test_invalid_position(self):
1242
transport = MockTransport([
1246
index = self.get_knit_index(transport, 'filename', 'r')
1248
self.assertRaises(errors.KnitCorrupt, index.keys)
1249
except TypeError, e:
1250
if (str(e) == ('exceptions must be strings, classes, or instances,'
1251
' not exceptions.ValueError')
1252
and sys.version_info[0:2] >= (2,5)):
1253
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1254
' raising new style exceptions with python'
1259
def test_invalid_size(self):
1260
transport = MockTransport([
1264
index = self.get_knit_index(transport, 'filename', 'r')
1266
self.assertRaises(errors.KnitCorrupt, index.keys)
1267
except TypeError, e:
1268
if (str(e) == ('exceptions must be strings, classes, or instances,'
1269
' not exceptions.ValueError')
1270
and sys.version_info[0:2] >= (2,5)):
1271
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1272
' raising new style exceptions with python'
1277
def test_scan_unvalidated_index_not_implemented(self):
1278
transport = MockTransport()
1279
index = self.get_knit_index(transport, 'filename', 'r')
1281
NotImplementedError, index.scan_unvalidated_index,
1282
'dummy graph_index')
1284
NotImplementedError, index.get_missing_compression_parents)
1286
def test_short_line(self):
1287
transport = MockTransport([
1290
"b option 10 10 0", # This line isn't terminated, ignored
1292
index = self.get_knit_index(transport, "filename", "r")
1293
self.assertEqual(set([('a',)]), index.keys())
1295
def test_skip_incomplete_record(self):
1296
# A line with bogus data should just be skipped
1297
transport = MockTransport([
1300
"b option 10 10 0", # This line isn't terminated, ignored
1301
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1303
index = self.get_knit_index(transport, "filename", "r")
1304
self.assertEqual(set([('a',), ('c',)]), index.keys())
1306
def test_trailing_characters(self):
1307
# A line with bogus data should just be skipped
1308
transport = MockTransport([
1311
"b option 10 10 0 :a", # This line has extra trailing characters
1312
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1314
index = self.get_knit_index(transport, "filename", "r")
1315
self.assertEqual(set([('a',), ('c',)]), index.keys())
1318
class LowLevelKnitIndexTests_c(LowLevelKnitIndexTests):
1320
_test_needs_features = [compiled_knit_feature]
1322
def get_knit_index(self, transport, name, mode):
1323
mapper = ConstantMapper(name)
1324
from bzrlib._knit_load_data_pyx import _load_data_c
1325
self.overrideAttr(knit, '_load_data', _load_data_c)
1326
allow_writes = lambda: mode == 'w'
1327
return _KndxIndex(transport, mapper, lambda:None,
1328
allow_writes, lambda:True)
1331
class Test_KnitAnnotator(TestCaseWithMemoryTransport):
1333
def make_annotator(self):
1334
factory = knit.make_pack_factory(True, True, 1)
1335
vf = factory(self.get_transport())
1336
return knit._KnitAnnotator(vf)
1338
def test__expand_fulltext(self):
1339
ann = self.make_annotator()
1340
rev_key = ('rev-id',)
1341
ann._num_compression_children[rev_key] = 1
1342
res = ann._expand_record(rev_key, (('parent-id',),), None,
1343
['line1\n', 'line2\n'], ('fulltext', True))
1344
# The content object and text lines should be cached appropriately
1345
self.assertEqual(['line1\n', 'line2'], res)
1346
content_obj = ann._content_objects[rev_key]
1347
self.assertEqual(['line1\n', 'line2\n'], content_obj._lines)
1348
self.assertEqual(res, content_obj.text())
1349
self.assertEqual(res, ann._text_cache[rev_key])
1351
def test__expand_delta_comp_parent_not_available(self):
1352
# Parent isn't available yet, so we return nothing, but queue up this
1353
# node for later processing
1354
ann = self.make_annotator()
1355
rev_key = ('rev-id',)
1356
parent_key = ('parent-id',)
1357
record = ['0,1,1\n', 'new-line\n']
1358
details = ('line-delta', False)
1359
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1361
self.assertEqual(None, res)
1362
self.assertTrue(parent_key in ann._pending_deltas)
1363
pending = ann._pending_deltas[parent_key]
1364
self.assertEqual(1, len(pending))
1365
self.assertEqual((rev_key, (parent_key,), record, details), pending[0])
1367
def test__expand_record_tracks_num_children(self):
1368
ann = self.make_annotator()
1369
rev_key = ('rev-id',)
1370
rev2_key = ('rev2-id',)
1371
parent_key = ('parent-id',)
1372
record = ['0,1,1\n', 'new-line\n']
1373
details = ('line-delta', False)
1374
ann._num_compression_children[parent_key] = 2
1375
ann._expand_record(parent_key, (), None, ['line1\n', 'line2\n'],
1376
('fulltext', False))
1377
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1379
self.assertEqual({parent_key: 1}, ann._num_compression_children)
1380
# Expanding the second child should remove the content object, and the
1381
# num_compression_children entry
1382
res = ann._expand_record(rev2_key, (parent_key,), parent_key,
1384
self.assertFalse(parent_key in ann._content_objects)
1385
self.assertEqual({}, ann._num_compression_children)
1386
# We should not cache the content_objects for rev2 and rev, because
1387
# they do not have compression children of their own.
1388
self.assertEqual({}, ann._content_objects)
1390
def test__expand_delta_records_blocks(self):
1391
ann = self.make_annotator()
1392
rev_key = ('rev-id',)
1393
parent_key = ('parent-id',)
1394
record = ['0,1,1\n', 'new-line\n']
1395
details = ('line-delta', True)
1396
ann._num_compression_children[parent_key] = 2
1397
ann._expand_record(parent_key, (), None,
1398
['line1\n', 'line2\n', 'line3\n'],
1399
('fulltext', False))
1400
ann._expand_record(rev_key, (parent_key,), parent_key, record, details)
1401
self.assertEqual({(rev_key, parent_key): [(1, 1, 1), (3, 3, 0)]},
1402
ann._matching_blocks)
1403
rev2_key = ('rev2-id',)
1404
record = ['0,1,1\n', 'new-line\n']
1405
details = ('line-delta', False)
1406
ann._expand_record(rev2_key, (parent_key,), parent_key, record, details)
1407
self.assertEqual([(1, 1, 2), (3, 3, 0)],
1408
ann._matching_blocks[(rev2_key, parent_key)])
1410
def test__get_parent_ann_uses_matching_blocks(self):
1411
ann = self.make_annotator()
1412
rev_key = ('rev-id',)
1413
parent_key = ('parent-id',)
1414
parent_ann = [(parent_key,)]*3
1415
block_key = (rev_key, parent_key)
1416
ann._annotations_cache[parent_key] = parent_ann
1417
ann._matching_blocks[block_key] = [(0, 1, 1), (3, 3, 0)]
1418
# We should not try to access any parent_lines content, because we know
1419
# we already have the matching blocks
1420
par_ann, blocks = ann._get_parent_annotations_and_matches(rev_key,
1421
['1\n', '2\n', '3\n'], parent_key)
1422
self.assertEqual(parent_ann, par_ann)
1423
self.assertEqual([(0, 1, 1), (3, 3, 0)], blocks)
1424
self.assertEqual({}, ann._matching_blocks)
1426
def test__process_pending(self):
1427
ann = self.make_annotator()
1428
rev_key = ('rev-id',)
1431
record = ['0,1,1\n', 'new-line\n']
1432
details = ('line-delta', False)
1433
p1_record = ['line1\n', 'line2\n']
1434
ann._num_compression_children[p1_key] = 1
1435
res = ann._expand_record(rev_key, (p1_key,p2_key), p1_key,
1437
self.assertEqual(None, res)
1438
# self.assertTrue(p1_key in ann._pending_deltas)
1439
self.assertEqual({}, ann._pending_annotation)
1440
# Now insert p1, and we should be able to expand the delta
1441
res = ann._expand_record(p1_key, (), None, p1_record,
1442
('fulltext', False))
1443
self.assertEqual(p1_record, res)
1444
ann._annotations_cache[p1_key] = [(p1_key,)]*2
1445
res = ann._process_pending(p1_key)
1446
self.assertEqual([], res)
1447
self.assertFalse(p1_key in ann._pending_deltas)
1448
self.assertTrue(p2_key in ann._pending_annotation)
1449
self.assertEqual({p2_key: [(rev_key, (p1_key, p2_key))]},
1450
ann._pending_annotation)
1451
# Now fill in parent 2, and pending annotation should be satisfied
1452
res = ann._expand_record(p2_key, (), None, [], ('fulltext', False))
1453
ann._annotations_cache[p2_key] = []
1454
res = ann._process_pending(p2_key)
1455
self.assertEqual([rev_key], res)
1456
self.assertEqual({}, ann._pending_annotation)
1457
self.assertEqual({}, ann._pending_deltas)
1459
def test_record_delta_removes_basis(self):
1460
ann = self.make_annotator()
1461
ann._expand_record(('parent-id',), (), None,
1462
['line1\n', 'line2\n'], ('fulltext', False))
1463
ann._num_compression_children['parent-id'] = 2
1465
def test_annotate_special_text(self):
1466
ann = self.make_annotator()
1468
rev1_key = ('rev-1',)
1469
rev2_key = ('rev-2',)
1470
rev3_key = ('rev-3',)
1471
spec_key = ('special:',)
1472
vf.add_lines(rev1_key, [], ['initial content\n'])
1473
vf.add_lines(rev2_key, [rev1_key], ['initial content\n',
1476
vf.add_lines(rev3_key, [rev1_key], ['initial content\n',
1479
spec_text = ('initial content\n'
1483
ann.add_special_text(spec_key, [rev2_key, rev3_key], spec_text)
1484
anns, lines = ann.annotate(spec_key)
1485
self.assertEqual([(rev1_key,),
1486
(rev2_key, rev3_key),
1490
self.assertEqualDiff(spec_text, ''.join(lines))
1493
class KnitTests(TestCaseWithTransport):
1494
"""Class containing knit test helper routines."""
1496
def make_test_knit(self, annotate=False, name='test'):
1497
mapper = ConstantMapper(name)
1498
return make_file_factory(annotate, mapper)(self.get_transport())
1501
class TestBadShaError(KnitTests):
1502
"""Tests for handling of sha errors."""
1504
def test_sha_exception_has_text(self):
1505
# having the failed text included in the error allows for recovery.
1506
source = self.make_test_knit()
1507
target = self.make_test_knit(name="target")
1508
if not source._max_delta_chain:
1509
raise TestNotApplicable(
1510
"cannot get delta-caused sha failures without deltas.")
1513
broken = ('broken',)
1514
source.add_lines(basis, (), ['foo\n'])
1515
source.add_lines(broken, (basis,), ['foo\n', 'bar\n'])
1516
# Seed target with a bad basis text
1517
target.add_lines(basis, (), ['gam\n'])
1518
target.insert_record_stream(
1519
source.get_record_stream([broken], 'unordered', False))
1520
err = self.assertRaises(errors.KnitCorrupt,
1521
target.get_record_stream([broken], 'unordered', True
1522
).next().get_bytes_as, 'chunked')
1523
self.assertEqual(['gam\n', 'bar\n'], err.content)
1524
# Test for formatting with live data
1525
self.assertStartsWith(str(err), "Knit ")
1528
class TestKnitIndex(KnitTests):
1530
def test_add_versions_dictionary_compresses(self):
1531
"""Adding versions to the index should update the lookup dict"""
1532
knit = self.make_test_knit()
1534
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1535
self.check_file_contents('test.kndx',
1536
'# bzr knit index 8\n'
1538
'a-1 fulltext 0 0 :'
1541
(('a-2',), ['fulltext'], (('a-2',), 0, 0), [('a-1',)]),
1542
(('a-3',), ['fulltext'], (('a-3',), 0, 0), [('a-2',)]),
1544
self.check_file_contents('test.kndx',
1545
'# bzr knit index 8\n'
1547
'a-1 fulltext 0 0 :\n'
1548
'a-2 fulltext 0 0 0 :\n'
1549
'a-3 fulltext 0 0 1 :'
1551
self.assertEqual(set([('a-3',), ('a-1',), ('a-2',)]), idx.keys())
1553
('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False)),
1554
('a-2',): ((('a-2',), 0, 0), None, (('a-1',),), ('fulltext', False)),
1555
('a-3',): ((('a-3',), 0, 0), None, (('a-2',),), ('fulltext', False)),
1556
}, idx.get_build_details(idx.keys()))
1557
self.assertEqual({('a-1',):(),
1558
('a-2',):(('a-1',),),
1559
('a-3',):(('a-2',),),},
1560
idx.get_parent_map(idx.keys()))
1562
def test_add_versions_fails_clean(self):
1563
"""If add_versions fails in the middle, it restores a pristine state.
1565
Any modifications that are made to the index are reset if all versions
1568
# This cheats a little bit by passing in a generator which will
1569
# raise an exception before the processing finishes
1570
# Other possibilities would be to have an version with the wrong number
1571
# of entries, or to make the backing transport unable to write any
1574
knit = self.make_test_knit()
1576
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1578
class StopEarly(Exception):
1581
def generate_failure():
1582
"""Add some entries and then raise an exception"""
1583
yield (('a-2',), ['fulltext'], (None, 0, 0), ('a-1',))
1584
yield (('a-3',), ['fulltext'], (None, 0, 0), ('a-2',))
1587
# Assert the pre-condition
1589
self.assertEqual(set([('a-1',)]), set(idx.keys()))
1591
{('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False))},
1592
idx.get_build_details([('a-1',)]))
1593
self.assertEqual({('a-1',):()}, idx.get_parent_map(idx.keys()))
1596
self.assertRaises(StopEarly, idx.add_records, generate_failure())
1597
# And it shouldn't be modified
1600
def test_knit_index_ignores_empty_files(self):
1601
# There was a race condition in older bzr, where a ^C at the right time
1602
# could leave an empty .kndx file, which bzr would later claim was a
1603
# corrupted file since the header was not present. In reality, the file
1604
# just wasn't created, so it should be ignored.
1605
t = transport.get_transport('.')
1606
t.put_bytes('test.kndx', '')
1608
knit = self.make_test_knit()
1610
def test_knit_index_checks_header(self):
1611
t = transport.get_transport('.')
1612
t.put_bytes('test.kndx', '# not really a knit header\n\n')
1613
k = self.make_test_knit()
1614
self.assertRaises(KnitHeaderError, k.keys)
1617
class TestGraphIndexKnit(KnitTests):
1618
"""Tests for knits using a GraphIndex rather than a KnitIndex."""
1620
def make_g_index(self, name, ref_lists=0, nodes=[]):
1621
builder = GraphIndexBuilder(ref_lists)
1622
for node, references, value in nodes:
1623
builder.add_node(node, references, value)
1624
stream = builder.finish()
1625
trans = self.get_transport()
1626
size = trans.put_file(name, stream)
1627
return GraphIndex(trans, name, size)
1629
def two_graph_index(self, deltas=False, catch_adds=False):
1630
"""Build a two-graph index.
1632
:param deltas: If true, use underlying indices with two node-ref
1633
lists and 'parent' set to a delta-compressed against tail.
1635
# build a complex graph across several indices.
1637
# delta compression inn the index
1638
index1 = self.make_g_index('1', 2, [
1639
(('tip', ), 'N0 100', ([('parent', )], [], )),
1640
(('tail', ), '', ([], []))])
1641
index2 = self.make_g_index('2', 2, [
1642
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], [('tail', )])),
1643
(('separate', ), '', ([], []))])
1645
# just blob location and graph in the index.
1646
index1 = self.make_g_index('1', 1, [
1647
(('tip', ), 'N0 100', ([('parent', )], )),
1648
(('tail', ), '', ([], ))])
1649
index2 = self.make_g_index('2', 1, [
1650
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], )),
1651
(('separate', ), '', ([], ))])
1652
combined_index = CombinedGraphIndex([index1, index2])
1654
self.combined_index = combined_index
1655
self.caught_entries = []
1656
add_callback = self.catch_add
1659
return _KnitGraphIndex(combined_index, lambda:True, deltas=deltas,
1660
add_callback=add_callback)
1662
def test_keys(self):
1663
index = self.two_graph_index()
1664
self.assertEqual(set([('tail',), ('tip',), ('parent',), ('separate',)]),
1667
def test_get_position(self):
1668
index = self.two_graph_index()
1669
self.assertEqual((index._graph_index._indices[0], 0, 100), index.get_position(('tip',)))
1670
self.assertEqual((index._graph_index._indices[1], 100, 78), index.get_position(('parent',)))
1672
def test_get_method_deltas(self):
1673
index = self.two_graph_index(deltas=True)
1674
self.assertEqual('fulltext', index.get_method(('tip',)))
1675
self.assertEqual('line-delta', index.get_method(('parent',)))
1677
def test_get_method_no_deltas(self):
1678
# check that the parent-history lookup is ignored with deltas=False.
1679
index = self.two_graph_index(deltas=False)
1680
self.assertEqual('fulltext', index.get_method(('tip',)))
1681
self.assertEqual('fulltext', index.get_method(('parent',)))
1683
def test_get_options_deltas(self):
1684
index = self.two_graph_index(deltas=True)
1685
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1686
self.assertEqual(['line-delta'], index.get_options(('parent',)))
1688
def test_get_options_no_deltas(self):
1689
# check that the parent-history lookup is ignored with deltas=False.
1690
index = self.two_graph_index(deltas=False)
1691
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1692
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1694
def test_get_parent_map(self):
1695
index = self.two_graph_index()
1696
self.assertEqual({('parent',):(('tail',), ('ghost',))},
1697
index.get_parent_map([('parent',), ('ghost',)]))
1699
def catch_add(self, entries):
1700
self.caught_entries.append(entries)
1702
def test_add_no_callback_errors(self):
1703
index = self.two_graph_index()
1704
self.assertRaises(errors.ReadOnlyError, index.add_records,
1705
[(('new',), 'fulltext,no-eol', (None, 50, 60), ['separate'])])
1707
def test_add_version_smoke(self):
1708
index = self.two_graph_index(catch_adds=True)
1709
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60),
1711
self.assertEqual([[(('new', ), 'N50 60', ((('separate',),),))]],
1712
self.caught_entries)
1714
def test_add_version_delta_not_delta_index(self):
1715
index = self.two_graph_index(catch_adds=True)
1716
self.assertRaises(errors.KnitCorrupt, index.add_records,
1717
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1718
self.assertEqual([], self.caught_entries)
1720
def test_add_version_same_dup(self):
1721
index = self.two_graph_index(catch_adds=True)
1722
# options can be spelt two different ways
1723
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
1724
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
1725
# position/length are ignored (because each pack could have fulltext or
1726
# delta, and be at a different position.
1727
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1729
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1731
# but neither should have added data:
1732
self.assertEqual([[], [], [], []], self.caught_entries)
1734
def test_add_version_different_dup(self):
1735
index = self.two_graph_index(deltas=True, catch_adds=True)
1737
self.assertRaises(errors.KnitCorrupt, index.add_records,
1738
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1739
self.assertRaises(errors.KnitCorrupt, index.add_records,
1740
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1742
self.assertRaises(errors.KnitCorrupt, index.add_records,
1743
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1744
self.assertEqual([], self.caught_entries)
1746
def test_add_versions_nodeltas(self):
1747
index = self.two_graph_index(catch_adds=True)
1749
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1750
(('new2',), 'fulltext', (None, 0, 6), [('new',)]),
1752
self.assertEqual([(('new', ), 'N50 60', ((('separate',),),)),
1753
(('new2', ), ' 0 6', ((('new',),),))],
1754
sorted(self.caught_entries[0]))
1755
self.assertEqual(1, len(self.caught_entries))
1757
def test_add_versions_deltas(self):
1758
index = self.two_graph_index(deltas=True, catch_adds=True)
1760
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1761
(('new2',), 'line-delta', (None, 0, 6), [('new',)]),
1763
self.assertEqual([(('new', ), 'N50 60', ((('separate',),), ())),
1764
(('new2', ), ' 0 6', ((('new',),), (('new',),), ))],
1765
sorted(self.caught_entries[0]))
1766
self.assertEqual(1, len(self.caught_entries))
1768
def test_add_versions_delta_not_delta_index(self):
1769
index = self.two_graph_index(catch_adds=True)
1770
self.assertRaises(errors.KnitCorrupt, index.add_records,
1771
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1772
self.assertEqual([], self.caught_entries)
1774
def test_add_versions_random_id_accepted(self):
1775
index = self.two_graph_index(catch_adds=True)
1776
index.add_records([], random_id=True)
1778
def test_add_versions_same_dup(self):
1779
index = self.two_graph_index(catch_adds=True)
1780
# options can be spelt two different ways
1781
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100),
1783
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100),
1785
# position/length are ignored (because each pack could have fulltext or
1786
# delta, and be at a different position.
1787
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1789
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1791
# but neither should have added data.
1792
self.assertEqual([[], [], [], []], self.caught_entries)
1794
def test_add_versions_different_dup(self):
1795
index = self.two_graph_index(deltas=True, catch_adds=True)
1797
self.assertRaises(errors.KnitCorrupt, index.add_records,
1798
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1799
self.assertRaises(errors.KnitCorrupt, index.add_records,
1800
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1802
self.assertRaises(errors.KnitCorrupt, index.add_records,
1803
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1804
# change options in the second record
1805
self.assertRaises(errors.KnitCorrupt, index.add_records,
1806
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)]),
1807
(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1808
self.assertEqual([], self.caught_entries)
1810
def make_g_index_missing_compression_parent(self):
1811
graph_index = self.make_g_index('missing_comp', 2,
1812
[(('tip', ), ' 100 78',
1813
([('missing-parent', ), ('ghost', )], [('missing-parent', )]))])
1816
def make_g_index_missing_parent(self):
1817
graph_index = self.make_g_index('missing_parent', 2,
1818
[(('parent', ), ' 100 78', ([], [])),
1819
(('tip', ), ' 100 78',
1820
([('parent', ), ('missing-parent', )], [('parent', )])),
1824
def make_g_index_no_external_refs(self):
1825
graph_index = self.make_g_index('no_external_refs', 2,
1826
[(('rev', ), ' 100 78',
1827
([('parent', ), ('ghost', )], []))])
1830
def test_add_good_unvalidated_index(self):
1831
unvalidated = self.make_g_index_no_external_refs()
1832
combined = CombinedGraphIndex([unvalidated])
1833
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1834
index.scan_unvalidated_index(unvalidated)
1835
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1837
def test_add_missing_compression_parent_unvalidated_index(self):
1838
unvalidated = self.make_g_index_missing_compression_parent()
1839
combined = CombinedGraphIndex([unvalidated])
1840
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1841
index.scan_unvalidated_index(unvalidated)
1842
# This also checks that its only the compression parent that is
1843
# examined, otherwise 'ghost' would also be reported as a missing
1846
frozenset([('missing-parent',)]),
1847
index.get_missing_compression_parents())
1849
def test_add_missing_noncompression_parent_unvalidated_index(self):
1850
unvalidated = self.make_g_index_missing_parent()
1851
combined = CombinedGraphIndex([unvalidated])
1852
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1853
track_external_parent_refs=True)
1854
index.scan_unvalidated_index(unvalidated)
1856
frozenset([('missing-parent',)]), index.get_missing_parents())
1858
def test_track_external_parent_refs(self):
1859
g_index = self.make_g_index('empty', 2, [])
1860
combined = CombinedGraphIndex([g_index])
1861
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1862
add_callback=self.catch_add, track_external_parent_refs=True)
1863
self.caught_entries = []
1865
(('new-key',), 'fulltext,no-eol', (None, 50, 60),
1866
[('parent-1',), ('parent-2',)])])
1868
frozenset([('parent-1',), ('parent-2',)]),
1869
index.get_missing_parents())
1871
def test_add_unvalidated_index_with_present_external_references(self):
1872
index = self.two_graph_index(deltas=True)
1873
# Ugly hack to get at one of the underlying GraphIndex objects that
1874
# two_graph_index built.
1875
unvalidated = index._graph_index._indices[1]
1876
# 'parent' is an external ref of _indices[1] (unvalidated), but is
1877
# present in _indices[0].
1878
index.scan_unvalidated_index(unvalidated)
1879
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1881
def make_new_missing_parent_g_index(self, name):
1882
missing_parent = name + '-missing-parent'
1883
graph_index = self.make_g_index(name, 2,
1884
[((name + 'tip', ), ' 100 78',
1885
([(missing_parent, ), ('ghost', )], [(missing_parent, )]))])
1888
def test_add_mulitiple_unvalidated_indices_with_missing_parents(self):
1889
g_index_1 = self.make_new_missing_parent_g_index('one')
1890
g_index_2 = self.make_new_missing_parent_g_index('two')
1891
combined = CombinedGraphIndex([g_index_1, g_index_2])
1892
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1893
index.scan_unvalidated_index(g_index_1)
1894
index.scan_unvalidated_index(g_index_2)
1896
frozenset([('one-missing-parent',), ('two-missing-parent',)]),
1897
index.get_missing_compression_parents())
1899
def test_add_mulitiple_unvalidated_indices_with_mutual_dependencies(self):
1900
graph_index_a = self.make_g_index('one', 2,
1901
[(('parent-one', ), ' 100 78', ([('non-compression-parent',)], [])),
1902
(('child-of-two', ), ' 100 78',
1903
([('parent-two',)], [('parent-two',)]))])
1904
graph_index_b = self.make_g_index('two', 2,
1905
[(('parent-two', ), ' 100 78', ([('non-compression-parent',)], [])),
1906
(('child-of-one', ), ' 100 78',
1907
([('parent-one',)], [('parent-one',)]))])
1908
combined = CombinedGraphIndex([graph_index_a, graph_index_b])
1909
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1910
index.scan_unvalidated_index(graph_index_a)
1911
index.scan_unvalidated_index(graph_index_b)
1913
frozenset([]), index.get_missing_compression_parents())
1916
class TestNoParentsGraphIndexKnit(KnitTests):
1917
"""Tests for knits using _KnitGraphIndex with no parents."""
1919
def make_g_index(self, name, ref_lists=0, nodes=[]):
1920
builder = GraphIndexBuilder(ref_lists)
1921
for node, references in nodes:
1922
builder.add_node(node, references)
1923
stream = builder.finish()
1924
trans = self.get_transport()
1925
size = trans.put_file(name, stream)
1926
return GraphIndex(trans, name, size)
1928
def test_add_good_unvalidated_index(self):
1929
unvalidated = self.make_g_index('unvalidated')
1930
combined = CombinedGraphIndex([unvalidated])
1931
index = _KnitGraphIndex(combined, lambda: True, parents=False)
1932
index.scan_unvalidated_index(unvalidated)
1933
self.assertEqual(frozenset(),
1934
index.get_missing_compression_parents())
1936
def test_parents_deltas_incompatible(self):
1937
index = CombinedGraphIndex([])
1938
self.assertRaises(errors.KnitError, _KnitGraphIndex, lambda:True,
1939
index, deltas=True, parents=False)
1941
def two_graph_index(self, catch_adds=False):
1942
"""Build a two-graph index.
1944
:param deltas: If true, use underlying indices with two node-ref
1945
lists and 'parent' set to a delta-compressed against tail.
1947
# put several versions in the index.
1948
index1 = self.make_g_index('1', 0, [
1949
(('tip', ), 'N0 100'),
1951
index2 = self.make_g_index('2', 0, [
1952
(('parent', ), ' 100 78'),
1953
(('separate', ), '')])
1954
combined_index = CombinedGraphIndex([index1, index2])
1956
self.combined_index = combined_index
1957
self.caught_entries = []
1958
add_callback = self.catch_add
1961
return _KnitGraphIndex(combined_index, lambda:True, parents=False,
1962
add_callback=add_callback)
1964
def test_keys(self):
1965
index = self.two_graph_index()
1966
self.assertEqual(set([('tail',), ('tip',), ('parent',), ('separate',)]),
1969
def test_get_position(self):
1970
index = self.two_graph_index()
1971
self.assertEqual((index._graph_index._indices[0], 0, 100),
1972
index.get_position(('tip',)))
1973
self.assertEqual((index._graph_index._indices[1], 100, 78),
1974
index.get_position(('parent',)))
1976
def test_get_method(self):
1977
index = self.two_graph_index()
1978
self.assertEqual('fulltext', index.get_method(('tip',)))
1979
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1981
def test_get_options(self):
1982
index = self.two_graph_index()
1983
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1984
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1986
def test_get_parent_map(self):
1987
index = self.two_graph_index()
1988
self.assertEqual({('parent',):None},
1989
index.get_parent_map([('parent',), ('ghost',)]))
1991
def catch_add(self, entries):
1992
self.caught_entries.append(entries)
1994
def test_add_no_callback_errors(self):
1995
index = self.two_graph_index()
1996
self.assertRaises(errors.ReadOnlyError, index.add_records,
1997
[(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)])])
1999
def test_add_version_smoke(self):
2000
index = self.two_graph_index(catch_adds=True)
2001
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60), [])])
2002
self.assertEqual([[(('new', ), 'N50 60')]],
2003
self.caught_entries)
2005
def test_add_version_delta_not_delta_index(self):
2006
index = self.two_graph_index(catch_adds=True)
2007
self.assertRaises(errors.KnitCorrupt, index.add_records,
2008
[(('new',), 'no-eol,line-delta', (None, 0, 100), [])])
2009
self.assertEqual([], self.caught_entries)
2011
def test_add_version_same_dup(self):
2012
index = self.two_graph_index(catch_adds=True)
2013
# options can be spelt two different ways
2014
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
2015
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
2016
# position/length are ignored (because each pack could have fulltext or
2017
# delta, and be at a different position.
2018
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
2019
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
2020
# but neither should have added data.
2021
self.assertEqual([[], [], [], []], self.caught_entries)
2023
def test_add_version_different_dup(self):
2024
index = self.two_graph_index(catch_adds=True)
2026
self.assertRaises(errors.KnitCorrupt, index.add_records,
2027
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2028
self.assertRaises(errors.KnitCorrupt, index.add_records,
2029
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
2030
self.assertRaises(errors.KnitCorrupt, index.add_records,
2031
[(('tip',), 'fulltext', (None, 0, 100), [])])
2033
self.assertRaises(errors.KnitCorrupt, index.add_records,
2034
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
2035
self.assertEqual([], self.caught_entries)
2037
def test_add_versions(self):
2038
index = self.two_graph_index(catch_adds=True)
2040
(('new',), 'fulltext,no-eol', (None, 50, 60), []),
2041
(('new2',), 'fulltext', (None, 0, 6), []),
2043
self.assertEqual([(('new', ), 'N50 60'), (('new2', ), ' 0 6')],
2044
sorted(self.caught_entries[0]))
2045
self.assertEqual(1, len(self.caught_entries))
2047
def test_add_versions_delta_not_delta_index(self):
2048
index = self.two_graph_index(catch_adds=True)
2049
self.assertRaises(errors.KnitCorrupt, index.add_records,
2050
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
2051
self.assertEqual([], self.caught_entries)
2053
def test_add_versions_parents_not_parents_index(self):
2054
index = self.two_graph_index(catch_adds=True)
2055
self.assertRaises(errors.KnitCorrupt, index.add_records,
2056
[(('new',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
2057
self.assertEqual([], self.caught_entries)
2059
def test_add_versions_random_id_accepted(self):
2060
index = self.two_graph_index(catch_adds=True)
2061
index.add_records([], random_id=True)
2063
def test_add_versions_same_dup(self):
2064
index = self.two_graph_index(catch_adds=True)
2065
# options can be spelt two different ways
2066
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
2067
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
2068
# position/length are ignored (because each pack could have fulltext or
2069
# delta, and be at a different position.
2070
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
2071
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
2072
# but neither should have added data.
2073
self.assertEqual([[], [], [], []], self.caught_entries)
2075
def test_add_versions_different_dup(self):
2076
index = self.two_graph_index(catch_adds=True)
2078
self.assertRaises(errors.KnitCorrupt, index.add_records,
2079
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2080
self.assertRaises(errors.KnitCorrupt, index.add_records,
2081
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
2082
self.assertRaises(errors.KnitCorrupt, index.add_records,
2083
[(('tip',), 'fulltext', (None, 0, 100), [])])
2085
self.assertRaises(errors.KnitCorrupt, index.add_records,
2086
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
2087
# change options in the second record
2088
self.assertRaises(errors.KnitCorrupt, index.add_records,
2089
[(('tip',), 'fulltext,no-eol', (None, 0, 100), []),
2090
(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2091
self.assertEqual([], self.caught_entries)
2094
class TestKnitVersionedFiles(KnitTests):
2096
def assertGroupKeysForIo(self, exp_groups, keys, non_local_keys,
2097
positions, _min_buffer_size=None):
2098
kvf = self.make_test_knit()
2099
if _min_buffer_size is None:
2100
_min_buffer_size = knit._STREAM_MIN_BUFFER_SIZE
2101
self.assertEqual(exp_groups, kvf._group_keys_for_io(keys,
2102
non_local_keys, positions,
2103
_min_buffer_size=_min_buffer_size))
2105
def assertSplitByPrefix(self, expected_map, expected_prefix_order,
2107
split, prefix_order = KnitVersionedFiles._split_by_prefix(keys)
2108
self.assertEqual(expected_map, split)
2109
self.assertEqual(expected_prefix_order, prefix_order)
2111
def test__group_keys_for_io(self):
2112
ft_detail = ('fulltext', False)
2113
ld_detail = ('line-delta', False)
2121
f_a: (ft_detail, (f_a, 0, 100), None),
2122
f_b: (ld_detail, (f_b, 100, 21), f_a),
2123
f_c: (ld_detail, (f_c, 180, 15), f_b),
2124
g_a: (ft_detail, (g_a, 121, 35), None),
2125
g_b: (ld_detail, (g_b, 156, 12), g_a),
2126
g_c: (ld_detail, (g_c, 195, 13), g_a),
2128
self.assertGroupKeysForIo([([f_a], set())],
2129
[f_a], [], positions)
2130
self.assertGroupKeysForIo([([f_a], set([f_a]))],
2131
[f_a], [f_a], positions)
2132
self.assertGroupKeysForIo([([f_a, f_b], set([]))],
2133
[f_a, f_b], [], positions)
2134
self.assertGroupKeysForIo([([f_a, f_b], set([f_b]))],
2135
[f_a, f_b], [f_b], positions)
2136
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2137
[f_a, g_a, f_b, g_b], [], positions)
2138
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2139
[f_a, g_a, f_b, g_b], [], positions,
2140
_min_buffer_size=150)
2141
self.assertGroupKeysForIo([([f_a, f_b], set()), ([g_a, g_b], set())],
2142
[f_a, g_a, f_b, g_b], [], positions,
2143
_min_buffer_size=100)
2144
self.assertGroupKeysForIo([([f_c], set()), ([g_b], set())],
2145
[f_c, g_b], [], positions,
2146
_min_buffer_size=125)
2147
self.assertGroupKeysForIo([([g_b, f_c], set())],
2148
[g_b, f_c], [], positions,
2149
_min_buffer_size=125)
2151
def test__split_by_prefix(self):
2152
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2153
'g': [('g', 'b'), ('g', 'a')],
2155
[('f', 'a'), ('g', 'b'),
2156
('g', 'a'), ('f', 'b')])
2158
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2159
'g': [('g', 'b'), ('g', 'a')],
2161
[('f', 'a'), ('f', 'b'),
2162
('g', 'b'), ('g', 'a')])
2164
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2165
'g': [('g', 'b'), ('g', 'a')],
2167
[('f', 'a'), ('f', 'b'),
2168
('g', 'b'), ('g', 'a')])
2170
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2171
'g': [('g', 'b'), ('g', 'a')],
2172
'': [('a',), ('b',)]
2174
[('f', 'a'), ('g', 'b'),
2176
('g', 'a'), ('f', 'b')])
2179
class TestStacking(KnitTests):
2181
def get_basis_and_test_knit(self):
2182
basis = self.make_test_knit(name='basis')
2183
basis = RecordingVersionedFilesDecorator(basis)
2184
test = self.make_test_knit(name='test')
2185
test.add_fallback_versioned_files(basis)
2188
def test_add_fallback_versioned_files(self):
2189
basis = self.make_test_knit(name='basis')
2190
test = self.make_test_knit(name='test')
2191
# It must not error; other tests test that the fallback is referred to
2192
# when accessing data.
2193
test.add_fallback_versioned_files(basis)
2195
def test_add_lines(self):
2196
# lines added to the test are not added to the basis
2197
basis, test = self.get_basis_and_test_knit()
2199
key_basis = ('bar',)
2200
key_cross_border = ('quux',)
2201
key_delta = ('zaphod',)
2202
test.add_lines(key, (), ['foo\n'])
2203
self.assertEqual({}, basis.get_parent_map([key]))
2204
# lines added to the test that reference across the stack do a
2206
basis.add_lines(key_basis, (), ['foo\n'])
2208
test.add_lines(key_cross_border, (key_basis,), ['foo\n'])
2209
self.assertEqual('fulltext', test._index.get_method(key_cross_border))
2210
# we don't even need to look at the basis to see that this should be
2211
# stored as a fulltext
2212
self.assertEqual([], basis.calls)
2213
# Subsequent adds do delta.
2215
test.add_lines(key_delta, (key_cross_border,), ['foo\n'])
2216
self.assertEqual('line-delta', test._index.get_method(key_delta))
2217
self.assertEqual([], basis.calls)
2219
def test_annotate(self):
2220
# annotations from the test knit are answered without asking the basis
2221
basis, test = self.get_basis_and_test_knit()
2223
key_basis = ('bar',)
2224
key_missing = ('missing',)
2225
test.add_lines(key, (), ['foo\n'])
2226
details = test.annotate(key)
2227
self.assertEqual([(key, 'foo\n')], details)
2228
self.assertEqual([], basis.calls)
2229
# But texts that are not in the test knit are looked for in the basis
2231
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2233
details = test.annotate(key_basis)
2234
self.assertEqual([(key_basis, 'foo\n'), (key_basis, 'bar\n')], details)
2235
# Not optimised to date:
2236
# self.assertEqual([("annotate", key_basis)], basis.calls)
2237
self.assertEqual([('get_parent_map', set([key_basis])),
2238
('get_parent_map', set([key_basis])),
2239
('get_record_stream', [key_basis], 'topological', True)],
2242
def test_check(self):
2243
# At the moment checking a stacked knit does implicitly check the
2245
basis, test = self.get_basis_and_test_knit()
2248
def test_get_parent_map(self):
2249
# parents in the test knit are answered without asking the basis
2250
basis, test = self.get_basis_and_test_knit()
2252
key_basis = ('bar',)
2253
key_missing = ('missing',)
2254
test.add_lines(key, (), [])
2255
parent_map = test.get_parent_map([key])
2256
self.assertEqual({key: ()}, parent_map)
2257
self.assertEqual([], basis.calls)
2258
# But parents that are not in the test knit are looked for in the basis
2259
basis.add_lines(key_basis, (), [])
2261
parent_map = test.get_parent_map([key, key_basis, key_missing])
2262
self.assertEqual({key: (),
2263
key_basis: ()}, parent_map)
2264
self.assertEqual([("get_parent_map", set([key_basis, key_missing]))],
2267
def test_get_record_stream_unordered_fulltexts(self):
2268
# records from the test knit are answered without asking the basis:
2269
basis, test = self.get_basis_and_test_knit()
2271
key_basis = ('bar',)
2272
key_missing = ('missing',)
2273
test.add_lines(key, (), ['foo\n'])
2274
records = list(test.get_record_stream([key], 'unordered', True))
2275
self.assertEqual(1, len(records))
2276
self.assertEqual([], basis.calls)
2277
# Missing (from test knit) objects are retrieved from the basis:
2278
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2280
records = list(test.get_record_stream([key_basis, key_missing],
2282
self.assertEqual(2, len(records))
2283
calls = list(basis.calls)
2284
for record in records:
2285
self.assertSubset([record.key], (key_basis, key_missing))
2286
if record.key == key_missing:
2287
self.assertIsInstance(record, AbsentContentFactory)
2289
reference = list(basis.get_record_stream([key_basis],
2290
'unordered', True))[0]
2291
self.assertEqual(reference.key, record.key)
2292
self.assertEqual(reference.sha1, record.sha1)
2293
self.assertEqual(reference.storage_kind, record.storage_kind)
2294
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2295
record.get_bytes_as(record.storage_kind))
2296
self.assertEqual(reference.get_bytes_as('fulltext'),
2297
record.get_bytes_as('fulltext'))
2298
# It's not strictly minimal, but it seems reasonable for now for it to
2299
# ask which fallbacks have which parents.
2301
("get_parent_map", set([key_basis, key_missing])),
2302
("get_record_stream", [key_basis], 'unordered', True)],
2305
def test_get_record_stream_ordered_fulltexts(self):
2306
# ordering is preserved down into the fallback store.
2307
basis, test = self.get_basis_and_test_knit()
2309
key_basis = ('bar',)
2310
key_basis_2 = ('quux',)
2311
key_missing = ('missing',)
2312
test.add_lines(key, (key_basis,), ['foo\n'])
2313
# Missing (from test knit) objects are retrieved from the basis:
2314
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2315
basis.add_lines(key_basis_2, (), ['quux\n'])
2317
# ask for in non-topological order
2318
records = list(test.get_record_stream(
2319
[key, key_basis, key_missing, key_basis_2], 'topological', True))
2320
self.assertEqual(4, len(records))
2322
for record in records:
2323
self.assertSubset([record.key],
2324
(key_basis, key_missing, key_basis_2, key))
2325
if record.key == key_missing:
2326
self.assertIsInstance(record, AbsentContentFactory)
2328
results.append((record.key, record.sha1, record.storage_kind,
2329
record.get_bytes_as('fulltext')))
2330
calls = list(basis.calls)
2331
order = [record[0] for record in results]
2332
self.assertEqual([key_basis_2, key_basis, key], order)
2333
for result in results:
2334
if result[0] == key:
2338
record = source.get_record_stream([result[0]], 'unordered',
2340
self.assertEqual(record.key, result[0])
2341
self.assertEqual(record.sha1, result[1])
2342
# We used to check that the storage kind matched, but actually it
2343
# depends on whether it was sourced from the basis, or in a single
2344
# group, because asking for full texts returns proxy objects to a
2345
# _ContentMapGenerator object; so checking the kind is unneeded.
2346
self.assertEqual(record.get_bytes_as('fulltext'), result[3])
2347
# It's not strictly minimal, but it seems reasonable for now for it to
2348
# ask which fallbacks have which parents.
2350
("get_parent_map", set([key_basis, key_basis_2, key_missing])),
2351
# topological is requested from the fallback, because that is what
2352
# was requested at the top level.
2353
("get_record_stream", [key_basis_2, key_basis], 'topological', True)],
2356
def test_get_record_stream_unordered_deltas(self):
2357
# records from the test knit are answered without asking the basis:
2358
basis, test = self.get_basis_and_test_knit()
2360
key_basis = ('bar',)
2361
key_missing = ('missing',)
2362
test.add_lines(key, (), ['foo\n'])
2363
records = list(test.get_record_stream([key], 'unordered', False))
2364
self.assertEqual(1, len(records))
2365
self.assertEqual([], basis.calls)
2366
# Missing (from test knit) objects are retrieved from the basis:
2367
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2369
records = list(test.get_record_stream([key_basis, key_missing],
2370
'unordered', False))
2371
self.assertEqual(2, len(records))
2372
calls = list(basis.calls)
2373
for record in records:
2374
self.assertSubset([record.key], (key_basis, key_missing))
2375
if record.key == key_missing:
2376
self.assertIsInstance(record, AbsentContentFactory)
2378
reference = list(basis.get_record_stream([key_basis],
2379
'unordered', False))[0]
2380
self.assertEqual(reference.key, record.key)
2381
self.assertEqual(reference.sha1, record.sha1)
2382
self.assertEqual(reference.storage_kind, record.storage_kind)
2383
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2384
record.get_bytes_as(record.storage_kind))
2385
# It's not strictly minimal, but it seems reasonable for now for it to
2386
# ask which fallbacks have which parents.
2388
("get_parent_map", set([key_basis, key_missing])),
2389
("get_record_stream", [key_basis], 'unordered', False)],
2392
def test_get_record_stream_ordered_deltas(self):
2393
# ordering is preserved down into the fallback store.
2394
basis, test = self.get_basis_and_test_knit()
2396
key_basis = ('bar',)
2397
key_basis_2 = ('quux',)
2398
key_missing = ('missing',)
2399
test.add_lines(key, (key_basis,), ['foo\n'])
2400
# Missing (from test knit) objects are retrieved from the basis:
2401
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2402
basis.add_lines(key_basis_2, (), ['quux\n'])
2404
# ask for in non-topological order
2405
records = list(test.get_record_stream(
2406
[key, key_basis, key_missing, key_basis_2], 'topological', False))
2407
self.assertEqual(4, len(records))
2409
for record in records:
2410
self.assertSubset([record.key],
2411
(key_basis, key_missing, key_basis_2, key))
2412
if record.key == key_missing:
2413
self.assertIsInstance(record, AbsentContentFactory)
2415
results.append((record.key, record.sha1, record.storage_kind,
2416
record.get_bytes_as(record.storage_kind)))
2417
calls = list(basis.calls)
2418
order = [record[0] for record in results]
2419
self.assertEqual([key_basis_2, key_basis, key], order)
2420
for result in results:
2421
if result[0] == key:
2425
record = source.get_record_stream([result[0]], 'unordered',
2427
self.assertEqual(record.key, result[0])
2428
self.assertEqual(record.sha1, result[1])
2429
self.assertEqual(record.storage_kind, result[2])
2430
self.assertEqual(record.get_bytes_as(record.storage_kind), result[3])
2431
# It's not strictly minimal, but it seems reasonable for now for it to
2432
# ask which fallbacks have which parents.
2434
("get_parent_map", set([key_basis, key_basis_2, key_missing])),
2435
("get_record_stream", [key_basis_2, key_basis], 'topological', False)],
2438
def test_get_sha1s(self):
2439
# sha1's in the test knit are answered without asking the basis
2440
basis, test = self.get_basis_and_test_knit()
2442
key_basis = ('bar',)
2443
key_missing = ('missing',)
2444
test.add_lines(key, (), ['foo\n'])
2445
key_sha1sum = osutils.sha('foo\n').hexdigest()
2446
sha1s = test.get_sha1s([key])
2447
self.assertEqual({key: key_sha1sum}, sha1s)
2448
self.assertEqual([], basis.calls)
2449
# But texts that are not in the test knit are looked for in the basis
2450
# directly (rather than via text reconstruction) so that remote servers
2451
# etc don't have to answer with full content.
2452
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2453
basis_sha1sum = osutils.sha('foo\nbar\n').hexdigest()
2455
sha1s = test.get_sha1s([key, key_missing, key_basis])
2456
self.assertEqual({key: key_sha1sum,
2457
key_basis: basis_sha1sum}, sha1s)
2458
self.assertEqual([("get_sha1s", set([key_basis, key_missing]))],
2461
def test_insert_record_stream(self):
2462
# records are inserted as normal; insert_record_stream builds on
2463
# add_lines, so a smoke test should be all that's needed:
2465
key_basis = ('bar',)
2466
key_delta = ('zaphod',)
2467
basis, test = self.get_basis_and_test_knit()
2468
source = self.make_test_knit(name='source')
2469
basis.add_lines(key_basis, (), ['foo\n'])
2471
source.add_lines(key_basis, (), ['foo\n'])
2472
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2473
stream = source.get_record_stream([key_delta], 'unordered', False)
2474
test.insert_record_stream(stream)
2475
# XXX: this does somewhat too many calls in making sure of whether it
2476
# has to recreate the full text.
2477
self.assertEqual([("get_parent_map", set([key_basis])),
2478
('get_parent_map', set([key_basis])),
2479
('get_record_stream', [key_basis], 'unordered', True)],
2481
self.assertEqual({key_delta:(key_basis,)},
2482
test.get_parent_map([key_delta]))
2483
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2484
'unordered', True).next().get_bytes_as('fulltext'))
2486
def test_iter_lines_added_or_present_in_keys(self):
2487
# Lines from the basis are returned, and lines for a given key are only
2491
# all sources are asked for keys:
2492
basis, test = self.get_basis_and_test_knit()
2493
basis.add_lines(key1, (), ["foo"])
2495
lines = list(test.iter_lines_added_or_present_in_keys([key1]))
2496
self.assertEqual([("foo\n", key1)], lines)
2497
self.assertEqual([("iter_lines_added_or_present_in_keys", set([key1]))],
2499
# keys in both are not duplicated:
2500
test.add_lines(key2, (), ["bar\n"])
2501
basis.add_lines(key2, (), ["bar\n"])
2503
lines = list(test.iter_lines_added_or_present_in_keys([key2]))
2504
self.assertEqual([("bar\n", key2)], lines)
2505
self.assertEqual([], basis.calls)
2507
def test_keys(self):
2510
# all sources are asked for keys:
2511
basis, test = self.get_basis_and_test_knit()
2513
self.assertEqual(set(), set(keys))
2514
self.assertEqual([("keys",)], basis.calls)
2515
# keys from a basis are returned:
2516
basis.add_lines(key1, (), [])
2519
self.assertEqual(set([key1]), set(keys))
2520
self.assertEqual([("keys",)], basis.calls)
2521
# keys in both are not duplicated:
2522
test.add_lines(key2, (), [])
2523
basis.add_lines(key2, (), [])
2526
self.assertEqual(2, len(keys))
2527
self.assertEqual(set([key1, key2]), set(keys))
2528
self.assertEqual([("keys",)], basis.calls)
2530
def test_add_mpdiffs(self):
2531
# records are inserted as normal; add_mpdiff builds on
2532
# add_lines, so a smoke test should be all that's needed:
2534
key_basis = ('bar',)
2535
key_delta = ('zaphod',)
2536
basis, test = self.get_basis_and_test_knit()
2537
source = self.make_test_knit(name='source')
2538
basis.add_lines(key_basis, (), ['foo\n'])
2540
source.add_lines(key_basis, (), ['foo\n'])
2541
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2542
diffs = source.make_mpdiffs([key_delta])
2543
test.add_mpdiffs([(key_delta, (key_basis,),
2544
source.get_sha1s([key_delta])[key_delta], diffs[0])])
2545
self.assertEqual([("get_parent_map", set([key_basis])),
2546
('get_record_stream', [key_basis], 'unordered', True),],
2548
self.assertEqual({key_delta:(key_basis,)},
2549
test.get_parent_map([key_delta]))
2550
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2551
'unordered', True).next().get_bytes_as('fulltext'))
2553
def test_make_mpdiffs(self):
2554
# Generating an mpdiff across a stacking boundary should detect parent
2558
key_right = ('zaphod',)
2559
basis, test = self.get_basis_and_test_knit()
2560
basis.add_lines(key_left, (), ['bar\n'])
2561
basis.add_lines(key_right, (), ['zaphod\n'])
2563
test.add_lines(key, (key_left, key_right),
2564
['bar\n', 'foo\n', 'zaphod\n'])
2565
diffs = test.make_mpdiffs([key])
2567
multiparent.MultiParent([multiparent.ParentText(0, 0, 0, 1),
2568
multiparent.NewText(['foo\n']),
2569
multiparent.ParentText(1, 0, 2, 1)])],
2571
self.assertEqual(3, len(basis.calls))
2573
("get_parent_map", set([key_left, key_right])),
2574
("get_parent_map", set([key_left, key_right])),
2577
last_call = basis.calls[-1]
2578
self.assertEqual('get_record_stream', last_call[0])
2579
self.assertEqual(set([key_left, key_right]), set(last_call[1]))
2580
self.assertEqual('topological', last_call[2])
2581
self.assertEqual(True, last_call[3])
2584
class TestNetworkBehaviour(KnitTests):
2585
"""Tests for getting data out of/into knits over the network."""
2587
def test_include_delta_closure_generates_a_knit_delta_closure(self):
2588
vf = self.make_test_knit(name='test')
2589
# put in three texts, giving ft, delta, delta
2590
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2591
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2592
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2593
# But heuristics could interfere, so check what happened:
2594
self.assertEqual(['knit-ft-gz', 'knit-delta-gz', 'knit-delta-gz'],
2595
[record.storage_kind for record in
2596
vf.get_record_stream([('base',), ('d1',), ('d2',)],
2597
'topological', False)])
2598
# generate a stream of just the deltas include_delta_closure=True,
2599
# serialise to the network, and check that we get a delta closure on the wire.
2600
stream = vf.get_record_stream([('d1',), ('d2',)], 'topological', True)
2601
netb = [record.get_bytes_as(record.storage_kind) for record in stream]
2602
# The first bytes should be a memo from _ContentMapGenerator, and the
2603
# second bytes should be empty (because its a API proxy not something
2604
# for wire serialisation.
2605
self.assertEqual('', netb[1])
2607
kind, line_end = network_bytes_to_kind_and_offset(bytes)
2608
self.assertEqual('knit-delta-closure', kind)
2611
class TestContentMapGenerator(KnitTests):
2612
"""Tests for ContentMapGenerator"""
2614
def test_get_record_stream_gives_records(self):
2615
vf = self.make_test_knit(name='test')
2616
# put in three texts, giving ft, delta, delta
2617
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2618
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2619
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2620
keys = [('d1',), ('d2',)]
2621
generator = _VFContentMapGenerator(vf, keys,
2622
global_map=vf.get_parent_map(keys))
2623
for record in generator.get_record_stream():
2624
if record.key == ('d1',):
2625
self.assertEqual('d1\n', record.get_bytes_as('fulltext'))
2627
self.assertEqual('d2\n', record.get_bytes_as('fulltext'))
2629
def test_get_record_stream_kinds_are_raw(self):
2630
vf = self.make_test_knit(name='test')
2631
# put in three texts, giving ft, delta, delta
2632
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2633
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2634
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2635
keys = [('base',), ('d1',), ('d2',)]
2636
generator = _VFContentMapGenerator(vf, keys,
2637
global_map=vf.get_parent_map(keys))
2638
kinds = {('base',): 'knit-delta-closure',
2639
('d1',): 'knit-delta-closure-ref',
2640
('d2',): 'knit-delta-closure-ref',
2642
for record in generator.get_record_stream():
2643
self.assertEqual(kinds[record.key], record.storage_kind)
added
removed
bzrlib/tests/test_knit.py
