1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
"""Tests for Knit data structure"""
19
from cStringIO import StringIO
32
from bzrlib.errors import (
33
RevisionAlreadyPresent,
38
from bzrlib.index import *
39
from bzrlib.knit import (
45
_VFContentMapGenerator,
52
from bzrlib.repofmt import pack_repo
53
from bzrlib.tests import (
57
TestCaseWithMemoryTransport,
58
TestCaseWithTransport,
61
from bzrlib.transport import get_transport
62
from bzrlib.transport.memory import MemoryTransport
63
from bzrlib.tuned_gzip import GzipFile
64
from bzrlib.versionedfile import (
67
network_bytes_to_kind_and_offset,
68
RecordingVersionedFilesDecorator,
72
compiled_knit_feature = tests.ModuleAvailableFeature(
73
'bzrlib._knit_load_data_pyx')
76
class KnitContentTestsMixin(object):
78
def test_constructor(self):
79
content = self._make_content([])
82
content = self._make_content([])
83
self.assertEqual(content.text(), [])
85
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
86
self.assertEqual(content.text(), ["text1", "text2"])
89
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
91
self.assertIsInstance(copy, content.__class__)
92
self.assertEqual(copy.annotate(), content.annotate())
94
def assertDerivedBlocksEqual(self, source, target, noeol=False):
95
"""Assert that the derived matching blocks match real output"""
96
source_lines = source.splitlines(True)
97
target_lines = target.splitlines(True)
99
if noeol and not line.endswith('\n'):
103
source_content = self._make_content([(None, nl(l)) for l in source_lines])
104
target_content = self._make_content([(None, nl(l)) for l in target_lines])
105
line_delta = source_content.line_delta(target_content)
106
delta_blocks = list(KnitContent.get_line_delta_blocks(line_delta,
107
source_lines, target_lines))
108
matcher = KnitSequenceMatcher(None, source_lines, target_lines)
109
matcher_blocks = list(list(matcher.get_matching_blocks()))
110
self.assertEqual(matcher_blocks, delta_blocks)
112
def test_get_line_delta_blocks(self):
113
self.assertDerivedBlocksEqual('a\nb\nc\n', 'q\nc\n')
114
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1)
115
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1A)
116
self.assertDerivedBlocksEqual(TEXT_1, TEXT_1B)
117
self.assertDerivedBlocksEqual(TEXT_1B, TEXT_1A)
118
self.assertDerivedBlocksEqual(TEXT_1A, TEXT_1B)
119
self.assertDerivedBlocksEqual(TEXT_1A, '')
120
self.assertDerivedBlocksEqual('', TEXT_1A)
121
self.assertDerivedBlocksEqual('', '')
122
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd')
124
def test_get_line_delta_blocks_noeol(self):
125
"""Handle historical knit deltas safely
127
Some existing knit deltas don't consider the last line to differ
128
when the only difference whether it has a final newline.
130
New knit deltas appear to always consider the last line to differ
133
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\nd\n', noeol=True)
134
self.assertDerivedBlocksEqual('a\nb\nc\nd\n', 'a\nb\nc', noeol=True)
135
self.assertDerivedBlocksEqual('a\nb\nc\n', 'a\nb\nc', noeol=True)
136
self.assertDerivedBlocksEqual('a\nb\nc', 'a\nb\nc\n', noeol=True)
148
Banana cup cake recipe
158
Banana cup cake recipe
160
- bananas (do not use plantains!!!)
167
Banana cup cake recipe
184
class TestPlainKnitContent(TestCase, KnitContentTestsMixin):
186
def _make_content(self, lines):
187
annotated_content = AnnotatedKnitContent(lines)
188
return PlainKnitContent(annotated_content.text(), 'bogus')
190
def test_annotate(self):
191
content = self._make_content([])
192
self.assertEqual(content.annotate(), [])
194
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
195
self.assertEqual(content.annotate(),
196
[("bogus", "text1"), ("bogus", "text2")])
198
def test_line_delta(self):
199
content1 = self._make_content([("", "a"), ("", "b")])
200
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
201
self.assertEqual(content1.line_delta(content2),
202
[(1, 2, 2, ["a", "c"])])
204
def test_line_delta_iter(self):
205
content1 = self._make_content([("", "a"), ("", "b")])
206
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
207
it = content1.line_delta_iter(content2)
208
self.assertEqual(it.next(), (1, 2, 2, ["a", "c"]))
209
self.assertRaises(StopIteration, it.next)
212
class TestAnnotatedKnitContent(TestCase, KnitContentTestsMixin):
214
def _make_content(self, lines):
215
return AnnotatedKnitContent(lines)
217
def test_annotate(self):
218
content = self._make_content([])
219
self.assertEqual(content.annotate(), [])
221
content = self._make_content([("origin1", "text1"), ("origin2", "text2")])
222
self.assertEqual(content.annotate(),
223
[("origin1", "text1"), ("origin2", "text2")])
225
def test_line_delta(self):
226
content1 = self._make_content([("", "a"), ("", "b")])
227
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
228
self.assertEqual(content1.line_delta(content2),
229
[(1, 2, 2, [("", "a"), ("", "c")])])
231
def test_line_delta_iter(self):
232
content1 = self._make_content([("", "a"), ("", "b")])
233
content2 = self._make_content([("", "a"), ("", "a"), ("", "c")])
234
it = content1.line_delta_iter(content2)
235
self.assertEqual(it.next(), (1, 2, 2, [("", "a"), ("", "c")]))
236
self.assertRaises(StopIteration, it.next)
239
class MockTransport(object):
241
def __init__(self, file_lines=None):
242
self.file_lines = file_lines
244
# We have no base directory for the MockTransport
247
def get(self, filename):
248
if self.file_lines is None:
249
raise NoSuchFile(filename)
251
return StringIO("\n".join(self.file_lines))
253
def readv(self, relpath, offsets):
254
fp = self.get(relpath)
255
for offset, size in offsets:
257
yield offset, fp.read(size)
259
def __getattr__(self, name):
260
def queue_call(*args, **kwargs):
261
self.calls.append((name, args, kwargs))
265
class MockReadvFailingTransport(MockTransport):
266
"""Fail in the middle of a readv() result.
268
This Transport will successfully yield the first two requested hunks, but
269
raise NoSuchFile for the rest.
272
def readv(self, relpath, offsets):
274
for result in MockTransport.readv(self, relpath, offsets):
276
# we use 2 because the first offset is the pack header, the second
277
# is the first actual content requset
279
raise errors.NoSuchFile(relpath)
283
class KnitRecordAccessTestsMixin(object):
284
"""Tests for getting and putting knit records."""
286
def test_add_raw_records(self):
287
"""Add_raw_records adds records retrievable later."""
288
access = self.get_access()
289
memos = access.add_raw_records([('key', 10)], '1234567890')
290
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
292
def test_add_several_raw_records(self):
293
"""add_raw_records with many records and read some back."""
294
access = self.get_access()
295
memos = access.add_raw_records([('key', 10), ('key2', 2), ('key3', 5)],
297
self.assertEqual(['1234567890', '12', '34567'],
298
list(access.get_raw_records(memos)))
299
self.assertEqual(['1234567890'],
300
list(access.get_raw_records(memos[0:1])))
301
self.assertEqual(['12'],
302
list(access.get_raw_records(memos[1:2])))
303
self.assertEqual(['34567'],
304
list(access.get_raw_records(memos[2:3])))
305
self.assertEqual(['1234567890', '34567'],
306
list(access.get_raw_records(memos[0:1] + memos[2:3])))
309
class TestKnitKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
310
"""Tests for the .kndx implementation."""
312
def get_access(self):
313
"""Get a .knit style access instance."""
314
mapper = ConstantMapper("foo")
315
access = _KnitKeyAccess(self.get_transport(), mapper)
319
class _TestException(Exception):
320
"""Just an exception for local tests to use."""
323
class TestPackKnitAccess(TestCaseWithMemoryTransport, KnitRecordAccessTestsMixin):
324
"""Tests for the pack based access."""
326
def get_access(self):
327
return self._get_access()[0]
329
def _get_access(self, packname='packfile', index='FOO'):
330
transport = self.get_transport()
331
def write_data(bytes):
332
transport.append_bytes(packname, bytes)
333
writer = pack.ContainerWriter(write_data)
335
access = _DirectPackAccess({})
336
access.set_writer(writer, index, (transport, packname))
337
return access, writer
339
def make_pack_file(self):
340
"""Create a pack file with 2 records."""
341
access, writer = self._get_access(packname='packname', index='foo')
343
memos.extend(access.add_raw_records([('key1', 10)], '1234567890'))
344
memos.extend(access.add_raw_records([('key2', 5)], '12345'))
348
def make_vf_for_retrying(self):
349
"""Create 3 packs and a reload function.
351
Originally, 2 pack files will have the data, but one will be missing.
352
And then the third will be used in place of the first two if reload()
355
:return: (versioned_file, reload_counter)
356
versioned_file a KnitVersionedFiles using the packs for access
358
builder = self.make_branch_builder('.', format="1.9")
359
builder.start_series()
360
builder.build_snapshot('rev-1', None, [
361
('add', ('', 'root-id', 'directory', None)),
362
('add', ('file', 'file-id', 'file', 'content\nrev 1\n')),
364
builder.build_snapshot('rev-2', ['rev-1'], [
365
('modify', ('file-id', 'content\nrev 2\n')),
367
builder.build_snapshot('rev-3', ['rev-2'], [
368
('modify', ('file-id', 'content\nrev 3\n')),
370
builder.finish_series()
371
b = builder.get_branch()
373
self.addCleanup(b.unlock)
374
# Pack these three revisions into another pack file, but don't remove
377
collection = repo._pack_collection
378
collection.ensure_loaded()
379
orig_packs = collection.packs
380
packer = pack_repo.Packer(collection, orig_packs, '.testpack')
381
new_pack = packer.pack()
382
# forget about the new pack
386
# Set up a reload() function that switches to using the new pack file
387
new_index = new_pack.revision_index
388
access_tuple = new_pack.access_tuple()
389
reload_counter = [0, 0, 0]
391
reload_counter[0] += 1
392
if reload_counter[1] > 0:
393
# We already reloaded, nothing more to do
394
reload_counter[2] += 1
396
reload_counter[1] += 1
397
vf._index._graph_index._indices[:] = [new_index]
398
vf._access._indices.clear()
399
vf._access._indices[new_index] = access_tuple
401
# Delete one of the pack files so the data will need to be reloaded. We
402
# will delete the file with 'rev-2' in it
403
trans, name = orig_packs[1].access_tuple()
405
# We don't have the index trigger reloading because we want to test
406
# that we reload when the .pack disappears
407
vf._access._reload_func = reload
408
return vf, reload_counter
410
def make_reload_func(self, return_val=True):
413
reload_called[0] += 1
415
return reload_called, reload
417
def make_retry_exception(self):
418
# We raise a real exception so that sys.exc_info() is properly
421
raise _TestException('foobar')
422
except _TestException, e:
423
retry_exc = errors.RetryWithNewPacks(None, reload_occurred=False,
424
exc_info=sys.exc_info())
427
def test_read_from_several_packs(self):
428
access, writer = self._get_access()
430
memos.extend(access.add_raw_records([('key', 10)], '1234567890'))
432
access, writer = self._get_access('pack2', 'FOOBAR')
433
memos.extend(access.add_raw_records([('key', 5)], '12345'))
435
access, writer = self._get_access('pack3', 'BAZ')
436
memos.extend(access.add_raw_records([('key', 5)], 'alpha'))
438
transport = self.get_transport()
439
access = _DirectPackAccess({"FOO":(transport, 'packfile'),
440
"FOOBAR":(transport, 'pack2'),
441
"BAZ":(transport, 'pack3')})
442
self.assertEqual(['1234567890', '12345', 'alpha'],
443
list(access.get_raw_records(memos)))
444
self.assertEqual(['1234567890'],
445
list(access.get_raw_records(memos[0:1])))
446
self.assertEqual(['12345'],
447
list(access.get_raw_records(memos[1:2])))
448
self.assertEqual(['alpha'],
449
list(access.get_raw_records(memos[2:3])))
450
self.assertEqual(['1234567890', 'alpha'],
451
list(access.get_raw_records(memos[0:1] + memos[2:3])))
453
def test_set_writer(self):
454
"""The writer should be settable post construction."""
455
access = _DirectPackAccess({})
456
transport = self.get_transport()
457
packname = 'packfile'
459
def write_data(bytes):
460
transport.append_bytes(packname, bytes)
461
writer = pack.ContainerWriter(write_data)
463
access.set_writer(writer, index, (transport, packname))
464
memos = access.add_raw_records([('key', 10)], '1234567890')
466
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
468
def test_missing_index_raises_retry(self):
469
memos = self.make_pack_file()
470
transport = self.get_transport()
471
reload_called, reload_func = self.make_reload_func()
472
# Note that the index key has changed from 'foo' to 'bar'
473
access = _DirectPackAccess({'bar':(transport, 'packname')},
474
reload_func=reload_func)
475
e = self.assertListRaises(errors.RetryWithNewPacks,
476
access.get_raw_records, memos)
477
# Because a key was passed in which does not match our index list, we
478
# assume that the listing was already reloaded
479
self.assertTrue(e.reload_occurred)
480
self.assertIsInstance(e.exc_info, tuple)
481
self.assertIs(e.exc_info[0], KeyError)
482
self.assertIsInstance(e.exc_info[1], KeyError)
484
def test_missing_index_raises_key_error_with_no_reload(self):
485
memos = self.make_pack_file()
486
transport = self.get_transport()
487
# Note that the index key has changed from 'foo' to 'bar'
488
access = _DirectPackAccess({'bar':(transport, 'packname')})
489
e = self.assertListRaises(KeyError, access.get_raw_records, memos)
491
def test_missing_file_raises_retry(self):
492
memos = self.make_pack_file()
493
transport = self.get_transport()
494
reload_called, reload_func = self.make_reload_func()
495
# Note that the 'filename' has been changed to 'different-packname'
496
access = _DirectPackAccess({'foo':(transport, 'different-packname')},
497
reload_func=reload_func)
498
e = self.assertListRaises(errors.RetryWithNewPacks,
499
access.get_raw_records, memos)
500
# The file has gone missing, so we assume we need to reload
501
self.assertFalse(e.reload_occurred)
502
self.assertIsInstance(e.exc_info, tuple)
503
self.assertIs(e.exc_info[0], errors.NoSuchFile)
504
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
505
self.assertEqual('different-packname', e.exc_info[1].path)
507
def test_missing_file_raises_no_such_file_with_no_reload(self):
508
memos = self.make_pack_file()
509
transport = self.get_transport()
510
# Note that the 'filename' has been changed to 'different-packname'
511
access = _DirectPackAccess({'foo':(transport, 'different-packname')})
512
e = self.assertListRaises(errors.NoSuchFile,
513
access.get_raw_records, memos)
515
def test_failing_readv_raises_retry(self):
516
memos = self.make_pack_file()
517
transport = self.get_transport()
518
failing_transport = MockReadvFailingTransport(
519
[transport.get_bytes('packname')])
520
reload_called, reload_func = self.make_reload_func()
521
access = _DirectPackAccess({'foo':(failing_transport, 'packname')},
522
reload_func=reload_func)
523
# Asking for a single record will not trigger the Mock failure
524
self.assertEqual(['1234567890'],
525
list(access.get_raw_records(memos[:1])))
526
self.assertEqual(['12345'],
527
list(access.get_raw_records(memos[1:2])))
528
# A multiple offset readv() will fail mid-way through
529
e = self.assertListRaises(errors.RetryWithNewPacks,
530
access.get_raw_records, memos)
531
# The file has gone missing, so we assume we need to reload
532
self.assertFalse(e.reload_occurred)
533
self.assertIsInstance(e.exc_info, tuple)
534
self.assertIs(e.exc_info[0], errors.NoSuchFile)
535
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
536
self.assertEqual('packname', e.exc_info[1].path)
538
def test_failing_readv_raises_no_such_file_with_no_reload(self):
539
memos = self.make_pack_file()
540
transport = self.get_transport()
541
failing_transport = MockReadvFailingTransport(
542
[transport.get_bytes('packname')])
543
reload_called, reload_func = self.make_reload_func()
544
access = _DirectPackAccess({'foo':(failing_transport, 'packname')})
545
# Asking for a single record will not trigger the Mock failure
546
self.assertEqual(['1234567890'],
547
list(access.get_raw_records(memos[:1])))
548
self.assertEqual(['12345'],
549
list(access.get_raw_records(memos[1:2])))
550
# A multiple offset readv() will fail mid-way through
551
e = self.assertListRaises(errors.NoSuchFile,
552
access.get_raw_records, memos)
554
def test_reload_or_raise_no_reload(self):
555
access = _DirectPackAccess({}, reload_func=None)
556
retry_exc = self.make_retry_exception()
557
# Without a reload_func, we will just re-raise the original exception
558
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
560
def test_reload_or_raise_reload_changed(self):
561
reload_called, reload_func = self.make_reload_func(return_val=True)
562
access = _DirectPackAccess({}, reload_func=reload_func)
563
retry_exc = self.make_retry_exception()
564
access.reload_or_raise(retry_exc)
565
self.assertEqual([1], reload_called)
566
retry_exc.reload_occurred=True
567
access.reload_or_raise(retry_exc)
568
self.assertEqual([2], reload_called)
570
def test_reload_or_raise_reload_no_change(self):
571
reload_called, reload_func = self.make_reload_func(return_val=False)
572
access = _DirectPackAccess({}, reload_func=reload_func)
573
retry_exc = self.make_retry_exception()
574
# If reload_occurred is False, then we consider it an error to have
575
# reload_func() return False (no changes).
576
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
577
self.assertEqual([1], reload_called)
578
retry_exc.reload_occurred=True
579
# If reload_occurred is True, then we assume nothing changed because
580
# it had changed earlier, but didn't change again
581
access.reload_or_raise(retry_exc)
582
self.assertEqual([2], reload_called)
584
def test_annotate_retries(self):
585
vf, reload_counter = self.make_vf_for_retrying()
586
# It is a little bit bogus to annotate the Revision VF, but it works,
587
# as we have ancestry stored there
589
reload_lines = vf.annotate(key)
590
self.assertEqual([1, 1, 0], reload_counter)
591
plain_lines = vf.annotate(key)
592
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
593
if reload_lines != plain_lines:
594
self.fail('Annotation was not identical with reloading.')
595
# Now delete the packs-in-use, which should trigger another reload, but
596
# this time we just raise an exception because we can't recover
597
for trans, name in vf._access._indices.itervalues():
599
self.assertRaises(errors.NoSuchFile, vf.annotate, key)
600
self.assertEqual([2, 1, 1], reload_counter)
602
def test__get_record_map_retries(self):
603
vf, reload_counter = self.make_vf_for_retrying()
604
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
605
records = vf._get_record_map(keys)
606
self.assertEqual(keys, sorted(records.keys()))
607
self.assertEqual([1, 1, 0], reload_counter)
608
# Now delete the packs-in-use, which should trigger another reload, but
609
# this time we just raise an exception because we can't recover
610
for trans, name in vf._access._indices.itervalues():
612
self.assertRaises(errors.NoSuchFile, vf._get_record_map, keys)
613
self.assertEqual([2, 1, 1], reload_counter)
615
def test_get_record_stream_retries(self):
616
vf, reload_counter = self.make_vf_for_retrying()
617
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
618
record_stream = vf.get_record_stream(keys, 'topological', False)
619
record = record_stream.next()
620
self.assertEqual(('rev-1',), record.key)
621
self.assertEqual([0, 0, 0], reload_counter)
622
record = record_stream.next()
623
self.assertEqual(('rev-2',), record.key)
624
self.assertEqual([1, 1, 0], reload_counter)
625
record = record_stream.next()
626
self.assertEqual(('rev-3',), record.key)
627
self.assertEqual([1, 1, 0], reload_counter)
628
# Now delete all pack files, and see that we raise the right error
629
for trans, name in vf._access._indices.itervalues():
631
self.assertListRaises(errors.NoSuchFile,
632
vf.get_record_stream, keys, 'topological', False)
634
def test_iter_lines_added_or_present_in_keys_retries(self):
635
vf, reload_counter = self.make_vf_for_retrying()
636
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
637
# Unfortunately, iter_lines_added_or_present_in_keys iterates the
638
# result in random order (determined by the iteration order from a
639
# set()), so we don't have any solid way to trigger whether data is
640
# read before or after. However we tried to delete the middle node to
641
# exercise the code well.
642
# What we care about is that all lines are always yielded, but not
645
reload_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
646
self.assertEqual([1, 1, 0], reload_counter)
647
# Now do it again, to make sure the result is equivalent
648
plain_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
649
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
650
self.assertEqual(plain_lines, reload_lines)
651
self.assertEqual(21, len(plain_lines))
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.iter_lines_added_or_present_in_keys, keys)
657
self.assertEqual([2, 1, 1], reload_counter)
659
def test_get_record_stream_yields_disk_sorted_order(self):
660
# if we get 'unordered' pick a semi-optimal order for reading. The
661
# order should be grouped by pack file, and then by position in file
662
repo = self.make_repository('test', format='pack-0.92')
664
self.addCleanup(repo.unlock)
665
repo.start_write_group()
667
vf.add_lines(('f-id', 'rev-5'), [('f-id', 'rev-4')], ['lines\n'])
668
vf.add_lines(('f-id', 'rev-1'), [], ['lines\n'])
669
vf.add_lines(('f-id', 'rev-2'), [('f-id', 'rev-1')], ['lines\n'])
670
repo.commit_write_group()
671
# We inserted them as rev-5, rev-1, rev-2, we should get them back in
673
stream = vf.get_record_stream([('f-id', 'rev-1'), ('f-id', 'rev-5'),
674
('f-id', 'rev-2')], 'unordered', False)
675
keys = [r.key for r in stream]
676
self.assertEqual([('f-id', 'rev-5'), ('f-id', 'rev-1'),
677
('f-id', 'rev-2')], keys)
678
repo.start_write_group()
679
vf.add_lines(('f-id', 'rev-4'), [('f-id', 'rev-3')], ['lines\n'])
680
vf.add_lines(('f-id', 'rev-3'), [('f-id', 'rev-2')], ['lines\n'])
681
vf.add_lines(('f-id', 'rev-6'), [('f-id', 'rev-5')], ['lines\n'])
682
repo.commit_write_group()
683
# Request in random order, to make sure the output order isn't based on
685
request_keys = set(('f-id', 'rev-%d' % i) for i in range(1, 7))
686
stream = vf.get_record_stream(request_keys, 'unordered', False)
687
keys = [r.key for r in stream]
688
# We want to get the keys back in disk order, but it doesn't matter
689
# which pack we read from first. So this can come back in 2 orders
690
alt1 = [('f-id', 'rev-%d' % i) for i in [4, 3, 6, 5, 1, 2]]
691
alt2 = [('f-id', 'rev-%d' % i) for i in [5, 1, 2, 4, 3, 6]]
692
if keys != alt1 and keys != alt2:
693
self.fail('Returned key order did not match either expected order.'
694
' expected %s or %s, not %s'
695
% (alt1, alt2, keys))
698
class LowLevelKnitDataTests(TestCase):
700
def create_gz_content(self, text):
702
gz_file = gzip.GzipFile(mode='wb', fileobj=sio)
705
return sio.getvalue()
707
def make_multiple_records(self):
708
"""Create the content for multiple records."""
709
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
711
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
716
record_1 = (0, len(gz_txt), sha1sum)
717
total_txt.append(gz_txt)
718
sha1sum = osutils.sha('baz\n').hexdigest()
719
gz_txt = self.create_gz_content('version rev-id-2 1 %s\n'
723
record_2 = (record_1[1], len(gz_txt), sha1sum)
724
total_txt.append(gz_txt)
725
return total_txt, record_1, record_2
727
def test_valid_knit_data(self):
728
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
729
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
734
transport = MockTransport([gz_txt])
735
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
736
knit = KnitVersionedFiles(None, access)
737
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
739
contents = list(knit._read_records_iter(records))
740
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'],
741
'4e48e2c9a3d2ca8a708cb0cc545700544efb5021')], contents)
743
raw_contents = list(knit._read_records_iter_raw(records))
744
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
746
def test_multiple_records_valid(self):
747
total_txt, record_1, record_2 = self.make_multiple_records()
748
transport = MockTransport([''.join(total_txt)])
749
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
750
knit = KnitVersionedFiles(None, access)
751
records = [(('rev-id-1',), (('rev-id-1',), record_1[0], record_1[1])),
752
(('rev-id-2',), (('rev-id-2',), record_2[0], record_2[1]))]
754
contents = list(knit._read_records_iter(records))
755
self.assertEqual([(('rev-id-1',), ['foo\n', 'bar\n'], record_1[2]),
756
(('rev-id-2',), ['baz\n'], record_2[2])],
759
raw_contents = list(knit._read_records_iter_raw(records))
760
self.assertEqual([(('rev-id-1',), total_txt[0], record_1[2]),
761
(('rev-id-2',), total_txt[1], record_2[2])],
764
def test_not_enough_lines(self):
765
sha1sum = osutils.sha('foo\n').hexdigest()
766
# record says 2 lines data says 1
767
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
771
transport = MockTransport([gz_txt])
772
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
773
knit = KnitVersionedFiles(None, access)
774
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
775
self.assertRaises(errors.KnitCorrupt, list,
776
knit._read_records_iter(records))
778
# read_records_iter_raw won't detect that sort of mismatch/corruption
779
raw_contents = list(knit._read_records_iter_raw(records))
780
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
782
def test_too_many_lines(self):
783
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
784
# record says 1 lines data says 2
785
gz_txt = self.create_gz_content('version rev-id-1 1 %s\n'
790
transport = MockTransport([gz_txt])
791
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
792
knit = KnitVersionedFiles(None, access)
793
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
794
self.assertRaises(errors.KnitCorrupt, list,
795
knit._read_records_iter(records))
797
# read_records_iter_raw won't detect that sort of mismatch/corruption
798
raw_contents = list(knit._read_records_iter_raw(records))
799
self.assertEqual([(('rev-id-1',), gz_txt, sha1sum)], raw_contents)
801
def test_mismatched_version_id(self):
802
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
803
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
808
transport = MockTransport([gz_txt])
809
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
810
knit = KnitVersionedFiles(None, access)
811
# We are asking for rev-id-2, but the data is rev-id-1
812
records = [(('rev-id-2',), (('rev-id-2',), 0, len(gz_txt)))]
813
self.assertRaises(errors.KnitCorrupt, list,
814
knit._read_records_iter(records))
816
# read_records_iter_raw detects mismatches in the header
817
self.assertRaises(errors.KnitCorrupt, list,
818
knit._read_records_iter_raw(records))
820
def test_uncompressed_data(self):
821
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
822
txt = ('version rev-id-1 2 %s\n'
827
transport = MockTransport([txt])
828
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
829
knit = KnitVersionedFiles(None, access)
830
records = [(('rev-id-1',), (('rev-id-1',), 0, len(txt)))]
832
# We don't have valid gzip data ==> corrupt
833
self.assertRaises(errors.KnitCorrupt, list,
834
knit._read_records_iter(records))
836
# read_records_iter_raw will notice the bad data
837
self.assertRaises(errors.KnitCorrupt, list,
838
knit._read_records_iter_raw(records))
840
def test_corrupted_data(self):
841
sha1sum = osutils.sha('foo\nbar\n').hexdigest()
842
gz_txt = self.create_gz_content('version rev-id-1 2 %s\n'
847
# Change 2 bytes in the middle to \xff
848
gz_txt = gz_txt[:10] + '\xff\xff' + gz_txt[12:]
849
transport = MockTransport([gz_txt])
850
access = _KnitKeyAccess(transport, ConstantMapper('filename'))
851
knit = KnitVersionedFiles(None, access)
852
records = [(('rev-id-1',), (('rev-id-1',), 0, len(gz_txt)))]
853
self.assertRaises(errors.KnitCorrupt, list,
854
knit._read_records_iter(records))
855
# read_records_iter_raw will barf on bad gz data
856
self.assertRaises(errors.KnitCorrupt, list,
857
knit._read_records_iter_raw(records))
860
class LowLevelKnitIndexTests(TestCase):
862
def get_knit_index(self, transport, name, mode):
863
mapper = ConstantMapper(name)
864
orig = knit._load_data
866
knit._load_data = orig
867
self.addCleanup(reset)
868
from bzrlib._knit_load_data_py import _load_data_py
869
knit._load_data = _load_data_py
870
allow_writes = lambda: 'w' in mode
871
return _KndxIndex(transport, mapper, lambda:None, allow_writes, lambda:True)
873
def test_create_file(self):
874
transport = MockTransport()
875
index = self.get_knit_index(transport, "filename", "w")
877
call = transport.calls.pop(0)
878
# call[1][1] is a StringIO - we can't test it by simple equality.
879
self.assertEqual('put_file_non_atomic', call[0])
880
self.assertEqual('filename.kndx', call[1][0])
881
# With no history, _KndxIndex writes a new index:
882
self.assertEqual(_KndxIndex.HEADER,
883
call[1][1].getvalue())
884
self.assertEqual({'create_parent_dir': True}, call[2])
886
def test_read_utf8_version_id(self):
887
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
888
utf8_revision_id = unicode_revision_id.encode('utf-8')
889
transport = MockTransport([
891
'%s option 0 1 :' % (utf8_revision_id,)
893
index = self.get_knit_index(transport, "filename", "r")
894
# _KndxIndex is a private class, and deals in utf8 revision_ids, not
895
# Unicode revision_ids.
896
self.assertEqual({(utf8_revision_id,):()},
897
index.get_parent_map(index.keys()))
898
self.assertFalse((unicode_revision_id,) in index.keys())
900
def test_read_utf8_parents(self):
901
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
902
utf8_revision_id = unicode_revision_id.encode('utf-8')
903
transport = MockTransport([
905
"version option 0 1 .%s :" % (utf8_revision_id,)
907
index = self.get_knit_index(transport, "filename", "r")
908
self.assertEqual({("version",):((utf8_revision_id,),)},
909
index.get_parent_map(index.keys()))
911
def test_read_ignore_corrupted_lines(self):
912
transport = MockTransport([
915
"corrupted options 0 1 .b .c ",
916
"version options 0 1 :"
918
index = self.get_knit_index(transport, "filename", "r")
919
self.assertEqual(1, len(index.keys()))
920
self.assertEqual(set([("version",)]), index.keys())
922
def test_read_corrupted_header(self):
923
transport = MockTransport(['not a bzr knit index header\n'])
924
index = self.get_knit_index(transport, "filename", "r")
925
self.assertRaises(KnitHeaderError, index.keys)
927
def test_read_duplicate_entries(self):
928
transport = MockTransport([
930
"parent options 0 1 :",
931
"version options1 0 1 0 :",
932
"version options2 1 2 .other :",
933
"version options3 3 4 0 .other :"
935
index = self.get_knit_index(transport, "filename", "r")
936
self.assertEqual(2, len(index.keys()))
937
# check that the index used is the first one written. (Specific
938
# to KnitIndex style indices.
939
self.assertEqual("1", index._dictionary_compress([("version",)]))
940
self.assertEqual((("version",), 3, 4), index.get_position(("version",)))
941
self.assertEqual(["options3"], index.get_options(("version",)))
942
self.assertEqual({("version",):(("parent",), ("other",))},
943
index.get_parent_map([("version",)]))
945
def test_read_compressed_parents(self):
946
transport = MockTransport([
950
"c option 0 1 1 0 :",
952
index = self.get_knit_index(transport, "filename", "r")
953
self.assertEqual({("b",):(("a",),), ("c",):(("b",), ("a",))},
954
index.get_parent_map([("b",), ("c",)]))
956
def test_write_utf8_version_id(self):
957
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
958
utf8_revision_id = unicode_revision_id.encode('utf-8')
959
transport = MockTransport([
962
index = self.get_knit_index(transport, "filename", "r")
964
((utf8_revision_id,), ["option"], ((utf8_revision_id,), 0, 1), [])])
965
call = transport.calls.pop(0)
966
# call[1][1] is a StringIO - we can't test it by simple equality.
967
self.assertEqual('put_file_non_atomic', call[0])
968
self.assertEqual('filename.kndx', call[1][0])
969
# With no history, _KndxIndex writes a new index:
970
self.assertEqual(_KndxIndex.HEADER +
971
"\n%s option 0 1 :" % (utf8_revision_id,),
972
call[1][1].getvalue())
973
self.assertEqual({'create_parent_dir': True}, call[2])
975
def test_write_utf8_parents(self):
976
unicode_revision_id = u"version-\N{CYRILLIC CAPITAL LETTER A}"
977
utf8_revision_id = unicode_revision_id.encode('utf-8')
978
transport = MockTransport([
981
index = self.get_knit_index(transport, "filename", "r")
983
(("version",), ["option"], (("version",), 0, 1), [(utf8_revision_id,)])])
984
call = transport.calls.pop(0)
985
# call[1][1] is a StringIO - we can't test it by simple equality.
986
self.assertEqual('put_file_non_atomic', call[0])
987
self.assertEqual('filename.kndx', call[1][0])
988
# With no history, _KndxIndex writes a new index:
989
self.assertEqual(_KndxIndex.HEADER +
990
"\nversion option 0 1 .%s :" % (utf8_revision_id,),
991
call[1][1].getvalue())
992
self.assertEqual({'create_parent_dir': True}, call[2])
995
transport = MockTransport([
998
index = self.get_knit_index(transport, "filename", "r")
1000
self.assertEqual(set(), index.keys())
1002
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1003
self.assertEqual(set([("a",)]), index.keys())
1005
index.add_records([(("a",), ["option"], (("a",), 0, 1), [])])
1006
self.assertEqual(set([("a",)]), index.keys())
1008
index.add_records([(("b",), ["option"], (("b",), 0, 1), [])])
1009
self.assertEqual(set([("a",), ("b",)]), index.keys())
1011
def add_a_b(self, index, random_id=None):
1013
if random_id is not None:
1014
kwargs["random_id"] = random_id
1016
(("a",), ["option"], (("a",), 0, 1), [("b",)]),
1017
(("a",), ["opt"], (("a",), 1, 2), [("c",)]),
1018
(("b",), ["option"], (("b",), 2, 3), [("a",)])
1021
def assertIndexIsAB(self, index):
1026
index.get_parent_map(index.keys()))
1027
self.assertEqual((("a",), 1, 2), index.get_position(("a",)))
1028
self.assertEqual((("b",), 2, 3), index.get_position(("b",)))
1029
self.assertEqual(["opt"], index.get_options(("a",)))
1031
def test_add_versions(self):
1032
transport = MockTransport([
1035
index = self.get_knit_index(transport, "filename", "r")
1038
call = transport.calls.pop(0)
1039
# call[1][1] is a StringIO - we can't test it by simple equality.
1040
self.assertEqual('put_file_non_atomic', call[0])
1041
self.assertEqual('filename.kndx', call[1][0])
1042
# With no history, _KndxIndex writes a new index:
1045
"\na option 0 1 .b :"
1047
"\nb option 2 3 0 :",
1048
call[1][1].getvalue())
1049
self.assertEqual({'create_parent_dir': True}, call[2])
1050
self.assertIndexIsAB(index)
1052
def test_add_versions_random_id_is_accepted(self):
1053
transport = MockTransport([
1056
index = self.get_knit_index(transport, "filename", "r")
1057
self.add_a_b(index, random_id=True)
1059
def test_delay_create_and_add_versions(self):
1060
transport = MockTransport()
1062
index = self.get_knit_index(transport, "filename", "w")
1064
self.assertEqual([], transport.calls)
1067
#[ {"dir_mode": 0777, "create_parent_dir": True, "mode": "wb"},
1069
# Two calls: one during which we load the existing index (and when its
1070
# missing create it), then a second where we write the contents out.
1071
self.assertEqual(2, len(transport.calls))
1072
call = transport.calls.pop(0)
1073
self.assertEqual('put_file_non_atomic', call[0])
1074
self.assertEqual('filename.kndx', call[1][0])
1075
# With no history, _KndxIndex writes a new index:
1076
self.assertEqual(_KndxIndex.HEADER, call[1][1].getvalue())
1077
self.assertEqual({'create_parent_dir': True}, call[2])
1078
call = transport.calls.pop(0)
1079
# call[1][1] is a StringIO - we can't test it by simple equality.
1080
self.assertEqual('put_file_non_atomic', call[0])
1081
self.assertEqual('filename.kndx', call[1][0])
1082
# With no history, _KndxIndex writes a new index:
1085
"\na option 0 1 .b :"
1087
"\nb option 2 3 0 :",
1088
call[1][1].getvalue())
1089
self.assertEqual({'create_parent_dir': True}, call[2])
1091
def assertTotalBuildSize(self, size, keys, positions):
1092
self.assertEqual(size,
1093
knit._get_total_build_size(None, keys, positions))
1095
def test__get_total_build_size(self):
1097
('a',): (('fulltext', False), (('a',), 0, 100), None),
1098
('b',): (('line-delta', False), (('b',), 100, 21), ('a',)),
1099
('c',): (('line-delta', False), (('c',), 121, 35), ('b',)),
1100
('d',): (('line-delta', False), (('d',), 156, 12), ('b',)),
1102
self.assertTotalBuildSize(100, [('a',)], positions)
1103
self.assertTotalBuildSize(121, [('b',)], positions)
1104
# c needs both a & b
1105
self.assertTotalBuildSize(156, [('c',)], positions)
1106
# we shouldn't count 'b' twice
1107
self.assertTotalBuildSize(156, [('b',), ('c',)], positions)
1108
self.assertTotalBuildSize(133, [('d',)], positions)
1109
self.assertTotalBuildSize(168, [('c',), ('d',)], positions)
1111
def test_get_position(self):
1112
transport = MockTransport([
1117
index = self.get_knit_index(transport, "filename", "r")
1119
self.assertEqual((("a",), 0, 1), index.get_position(("a",)))
1120
self.assertEqual((("b",), 1, 2), index.get_position(("b",)))
1122
def test_get_method(self):
1123
transport = MockTransport([
1125
"a fulltext,unknown 0 1 :",
1126
"b unknown,line-delta 1 2 :",
1129
index = self.get_knit_index(transport, "filename", "r")
1131
self.assertEqual("fulltext", index.get_method("a"))
1132
self.assertEqual("line-delta", index.get_method("b"))
1133
self.assertRaises(errors.KnitIndexUnknownMethod, index.get_method, "c")
1135
def test_get_options(self):
1136
transport = MockTransport([
1141
index = self.get_knit_index(transport, "filename", "r")
1143
self.assertEqual(["opt1"], index.get_options("a"))
1144
self.assertEqual(["opt2", "opt3"], index.get_options("b"))
1146
def test_get_parent_map(self):
1147
transport = MockTransport([
1150
"b option 1 2 0 .c :",
1151
"c option 1 2 1 0 .e :"
1153
index = self.get_knit_index(transport, "filename", "r")
1157
("b",):(("a",), ("c",)),
1158
("c",):(("b",), ("a",), ("e",)),
1159
}, index.get_parent_map(index.keys()))
1161
def test_impossible_parent(self):
1162
"""Test we get KnitCorrupt if the parent couldn't possibly exist."""
1163
transport = MockTransport([
1166
"b option 0 1 4 :" # We don't have a 4th record
1168
index = self.get_knit_index(transport, 'filename', 'r')
1170
self.assertRaises(errors.KnitCorrupt, index.keys)
1171
except TypeError, e:
1172
if (str(e) == ('exceptions must be strings, classes, or instances,'
1173
' not exceptions.IndexError')
1174
and sys.version_info[0:2] >= (2,5)):
1175
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1176
' raising new style exceptions with python'
1181
def test_corrupted_parent(self):
1182
transport = MockTransport([
1186
"c option 0 1 1v :", # Can't have a parent of '1v'
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.ValueError')
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_in_list(self):
1202
transport = MockTransport([
1206
"c option 0 1 1 v :", # Can't have a parent of 'v'
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_invalid_position(self):
1222
transport = MockTransport([
1226
index = self.get_knit_index(transport, 'filename', 'r')
1228
self.assertRaises(errors.KnitCorrupt, index.keys)
1229
except TypeError, e:
1230
if (str(e) == ('exceptions must be strings, classes, or instances,'
1231
' not exceptions.ValueError')
1232
and sys.version_info[0:2] >= (2,5)):
1233
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1234
' raising new style exceptions with python'
1239
def test_invalid_size(self):
1240
transport = MockTransport([
1244
index = self.get_knit_index(transport, 'filename', 'r')
1246
self.assertRaises(errors.KnitCorrupt, index.keys)
1247
except TypeError, e:
1248
if (str(e) == ('exceptions must be strings, classes, or instances,'
1249
' not exceptions.ValueError')
1250
and sys.version_info[0:2] >= (2,5)):
1251
self.knownFailure('Pyrex <0.9.5 fails with TypeError when'
1252
' raising new style exceptions with python'
1257
def test_scan_unvalidated_index_not_implemented(self):
1258
transport = MockTransport()
1259
index = self.get_knit_index(transport, 'filename', 'r')
1261
NotImplementedError, index.scan_unvalidated_index,
1262
'dummy graph_index')
1264
NotImplementedError, index.get_missing_compression_parents)
1266
def test_short_line(self):
1267
transport = MockTransport([
1270
"b option 10 10 0", # This line isn't terminated, ignored
1272
index = self.get_knit_index(transport, "filename", "r")
1273
self.assertEqual(set([('a',)]), index.keys())
1275
def test_skip_incomplete_record(self):
1276
# A line with bogus data should just be skipped
1277
transport = MockTransport([
1280
"b option 10 10 0", # This line isn't terminated, ignored
1281
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1283
index = self.get_knit_index(transport, "filename", "r")
1284
self.assertEqual(set([('a',), ('c',)]), index.keys())
1286
def test_trailing_characters(self):
1287
# A line with bogus data should just be skipped
1288
transport = MockTransport([
1291
"b option 10 10 0 :a", # This line has extra trailing characters
1292
"c option 20 10 0 :", # Properly terminated, and starts with '\n'
1294
index = self.get_knit_index(transport, "filename", "r")
1295
self.assertEqual(set([('a',), ('c',)]), index.keys())
1298
class LowLevelKnitIndexTests_c(LowLevelKnitIndexTests):
1300
_test_needs_features = [compiled_knit_feature]
1302
def get_knit_index(self, transport, name, mode):
1303
mapper = ConstantMapper(name)
1304
orig = knit._load_data
1306
knit._load_data = orig
1307
self.addCleanup(reset)
1308
from bzrlib._knit_load_data_pyx import _load_data_c
1309
knit._load_data = _load_data_c
1310
allow_writes = lambda: mode == 'w'
1311
return _KndxIndex(transport, mapper, lambda:None, allow_writes, lambda:True)
1314
class Test_KnitAnnotator(TestCaseWithMemoryTransport):
1316
def make_annotator(self):
1317
factory = knit.make_pack_factory(True, True, 1)
1318
vf = factory(self.get_transport())
1319
return knit._KnitAnnotator(vf)
1321
def test__expand_fulltext(self):
1322
ann = self.make_annotator()
1323
rev_key = ('rev-id',)
1324
ann._num_compression_children[rev_key] = 1
1325
res = ann._expand_record(rev_key, (('parent-id',),), None,
1326
['line1\n', 'line2\n'], ('fulltext', True))
1327
# The content object and text lines should be cached appropriately
1328
self.assertEqual(['line1\n', 'line2'], res)
1329
content_obj = ann._content_objects[rev_key]
1330
self.assertEqual(['line1\n', 'line2\n'], content_obj._lines)
1331
self.assertEqual(res, content_obj.text())
1332
self.assertEqual(res, ann._text_cache[rev_key])
1334
def test__expand_delta_comp_parent_not_available(self):
1335
# Parent isn't available yet, so we return nothing, but queue up this
1336
# node for later processing
1337
ann = self.make_annotator()
1338
rev_key = ('rev-id',)
1339
parent_key = ('parent-id',)
1340
record = ['0,1,1\n', 'new-line\n']
1341
details = ('line-delta', False)
1342
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1344
self.assertEqual(None, res)
1345
self.assertTrue(parent_key in ann._pending_deltas)
1346
pending = ann._pending_deltas[parent_key]
1347
self.assertEqual(1, len(pending))
1348
self.assertEqual((rev_key, (parent_key,), record, details), pending[0])
1350
def test__expand_record_tracks_num_children(self):
1351
ann = self.make_annotator()
1352
rev_key = ('rev-id',)
1353
rev2_key = ('rev2-id',)
1354
parent_key = ('parent-id',)
1355
record = ['0,1,1\n', 'new-line\n']
1356
details = ('line-delta', False)
1357
ann._num_compression_children[parent_key] = 2
1358
ann._expand_record(parent_key, (), None, ['line1\n', 'line2\n'],
1359
('fulltext', False))
1360
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1362
self.assertEqual({parent_key: 1}, ann._num_compression_children)
1363
# Expanding the second child should remove the content object, and the
1364
# num_compression_children entry
1365
res = ann._expand_record(rev2_key, (parent_key,), parent_key,
1367
self.assertFalse(parent_key in ann._content_objects)
1368
self.assertEqual({}, ann._num_compression_children)
1369
# We should not cache the content_objects for rev2 and rev, because
1370
# they do not have compression children of their own.
1371
self.assertEqual({}, ann._content_objects)
1373
def test__expand_delta_records_blocks(self):
1374
ann = self.make_annotator()
1375
rev_key = ('rev-id',)
1376
parent_key = ('parent-id',)
1377
record = ['0,1,1\n', 'new-line\n']
1378
details = ('line-delta', True)
1379
ann._num_compression_children[parent_key] = 2
1380
ann._expand_record(parent_key, (), None,
1381
['line1\n', 'line2\n', 'line3\n'],
1382
('fulltext', False))
1383
ann._expand_record(rev_key, (parent_key,), parent_key, record, details)
1384
self.assertEqual({(rev_key, parent_key): [(1, 1, 1), (3, 3, 0)]},
1385
ann._matching_blocks)
1386
rev2_key = ('rev2-id',)
1387
record = ['0,1,1\n', 'new-line\n']
1388
details = ('line-delta', False)
1389
ann._expand_record(rev2_key, (parent_key,), parent_key, record, details)
1390
self.assertEqual([(1, 1, 2), (3, 3, 0)],
1391
ann._matching_blocks[(rev2_key, parent_key)])
1393
def test__get_parent_ann_uses_matching_blocks(self):
1394
ann = self.make_annotator()
1395
rev_key = ('rev-id',)
1396
parent_key = ('parent-id',)
1397
parent_ann = [(parent_key,)]*3
1398
block_key = (rev_key, parent_key)
1399
ann._annotations_cache[parent_key] = parent_ann
1400
ann._matching_blocks[block_key] = [(0, 1, 1), (3, 3, 0)]
1401
# We should not try to access any parent_lines content, because we know
1402
# we already have the matching blocks
1403
par_ann, blocks = ann._get_parent_annotations_and_matches(rev_key,
1404
['1\n', '2\n', '3\n'], parent_key)
1405
self.assertEqual(parent_ann, par_ann)
1406
self.assertEqual([(0, 1, 1), (3, 3, 0)], blocks)
1407
self.assertEqual({}, ann._matching_blocks)
1409
def test__process_pending(self):
1410
ann = self.make_annotator()
1411
rev_key = ('rev-id',)
1414
record = ['0,1,1\n', 'new-line\n']
1415
details = ('line-delta', False)
1416
p1_record = ['line1\n', 'line2\n']
1417
ann._num_compression_children[p1_key] = 1
1418
res = ann._expand_record(rev_key, (p1_key,p2_key), p1_key,
1420
self.assertEqual(None, res)
1421
# self.assertTrue(p1_key in ann._pending_deltas)
1422
self.assertEqual({}, ann._pending_annotation)
1423
# Now insert p1, and we should be able to expand the delta
1424
res = ann._expand_record(p1_key, (), None, p1_record,
1425
('fulltext', False))
1426
self.assertEqual(p1_record, res)
1427
ann._annotations_cache[p1_key] = [(p1_key,)]*2
1428
res = ann._process_pending(p1_key)
1429
self.assertEqual([], res)
1430
self.assertFalse(p1_key in ann._pending_deltas)
1431
self.assertTrue(p2_key in ann._pending_annotation)
1432
self.assertEqual({p2_key: [(rev_key, (p1_key, p2_key))]},
1433
ann._pending_annotation)
1434
# Now fill in parent 2, and pending annotation should be satisfied
1435
res = ann._expand_record(p2_key, (), None, [], ('fulltext', False))
1436
ann._annotations_cache[p2_key] = []
1437
res = ann._process_pending(p2_key)
1438
self.assertEqual([rev_key], res)
1439
self.assertEqual({}, ann._pending_annotation)
1440
self.assertEqual({}, ann._pending_deltas)
1442
def test_record_delta_removes_basis(self):
1443
ann = self.make_annotator()
1444
ann._expand_record(('parent-id',), (), None,
1445
['line1\n', 'line2\n'], ('fulltext', False))
1446
ann._num_compression_children['parent-id'] = 2
1448
def test_annotate_special_text(self):
1449
ann = self.make_annotator()
1451
rev1_key = ('rev-1',)
1452
rev2_key = ('rev-2',)
1453
rev3_key = ('rev-3',)
1454
spec_key = ('special:',)
1455
vf.add_lines(rev1_key, [], ['initial content\n'])
1456
vf.add_lines(rev2_key, [rev1_key], ['initial content\n',
1459
vf.add_lines(rev3_key, [rev1_key], ['initial content\n',
1462
spec_text = ('initial content\n'
1466
ann.add_special_text(spec_key, [rev2_key, rev3_key], spec_text)
1467
anns, lines = ann.annotate(spec_key)
1468
self.assertEqual([(rev1_key,),
1469
(rev2_key, rev3_key),
1473
self.assertEqualDiff(spec_text, ''.join(lines))
1476
class KnitTests(TestCaseWithTransport):
1477
"""Class containing knit test helper routines."""
1479
def make_test_knit(self, annotate=False, name='test'):
1480
mapper = ConstantMapper(name)
1481
return make_file_factory(annotate, mapper)(self.get_transport())
1484
class TestBadShaError(KnitTests):
1485
"""Tests for handling of sha errors."""
1487
def test_sha_exception_has_text(self):
1488
# having the failed text included in the error allows for recovery.
1489
source = self.make_test_knit()
1490
target = self.make_test_knit(name="target")
1491
if not source._max_delta_chain:
1492
raise TestNotApplicable(
1493
"cannot get delta-caused sha failures without deltas.")
1496
broken = ('broken',)
1497
source.add_lines(basis, (), ['foo\n'])
1498
source.add_lines(broken, (basis,), ['foo\n', 'bar\n'])
1499
# Seed target with a bad basis text
1500
target.add_lines(basis, (), ['gam\n'])
1501
target.insert_record_stream(
1502
source.get_record_stream([broken], 'unordered', False))
1503
err = self.assertRaises(errors.KnitCorrupt,
1504
target.get_record_stream([broken], 'unordered', True
1505
).next().get_bytes_as, 'chunked')
1506
self.assertEqual(['gam\n', 'bar\n'], err.content)
1507
# Test for formatting with live data
1508
self.assertStartsWith(str(err), "Knit ")
1511
class TestKnitIndex(KnitTests):
1513
def test_add_versions_dictionary_compresses(self):
1514
"""Adding versions to the index should update the lookup dict"""
1515
knit = self.make_test_knit()
1517
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1518
self.check_file_contents('test.kndx',
1519
'# bzr knit index 8\n'
1521
'a-1 fulltext 0 0 :'
1524
(('a-2',), ['fulltext'], (('a-2',), 0, 0), [('a-1',)]),
1525
(('a-3',), ['fulltext'], (('a-3',), 0, 0), [('a-2',)]),
1527
self.check_file_contents('test.kndx',
1528
'# bzr knit index 8\n'
1530
'a-1 fulltext 0 0 :\n'
1531
'a-2 fulltext 0 0 0 :\n'
1532
'a-3 fulltext 0 0 1 :'
1534
self.assertEqual(set([('a-3',), ('a-1',), ('a-2',)]), idx.keys())
1536
('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False)),
1537
('a-2',): ((('a-2',), 0, 0), None, (('a-1',),), ('fulltext', False)),
1538
('a-3',): ((('a-3',), 0, 0), None, (('a-2',),), ('fulltext', False)),
1539
}, idx.get_build_details(idx.keys()))
1540
self.assertEqual({('a-1',):(),
1541
('a-2',):(('a-1',),),
1542
('a-3',):(('a-2',),),},
1543
idx.get_parent_map(idx.keys()))
1545
def test_add_versions_fails_clean(self):
1546
"""If add_versions fails in the middle, it restores a pristine state.
1548
Any modifications that are made to the index are reset if all versions
1551
# This cheats a little bit by passing in a generator which will
1552
# raise an exception before the processing finishes
1553
# Other possibilities would be to have an version with the wrong number
1554
# of entries, or to make the backing transport unable to write any
1557
knit = self.make_test_knit()
1559
idx.add_records([(('a-1',), ['fulltext'], (('a-1',), 0, 0), [])])
1561
class StopEarly(Exception):
1564
def generate_failure():
1565
"""Add some entries and then raise an exception"""
1566
yield (('a-2',), ['fulltext'], (None, 0, 0), ('a-1',))
1567
yield (('a-3',), ['fulltext'], (None, 0, 0), ('a-2',))
1570
# Assert the pre-condition
1572
self.assertEqual(set([('a-1',)]), set(idx.keys()))
1574
{('a-1',): ((('a-1',), 0, 0), None, (), ('fulltext', False))},
1575
idx.get_build_details([('a-1',)]))
1576
self.assertEqual({('a-1',):()}, idx.get_parent_map(idx.keys()))
1579
self.assertRaises(StopEarly, idx.add_records, generate_failure())
1580
# And it shouldn't be modified
1583
def test_knit_index_ignores_empty_files(self):
1584
# There was a race condition in older bzr, where a ^C at the right time
1585
# could leave an empty .kndx file, which bzr would later claim was a
1586
# corrupted file since the header was not present. In reality, the file
1587
# just wasn't created, so it should be ignored.
1588
t = get_transport('.')
1589
t.put_bytes('test.kndx', '')
1591
knit = self.make_test_knit()
1593
def test_knit_index_checks_header(self):
1594
t = get_transport('.')
1595
t.put_bytes('test.kndx', '# not really a knit header\n\n')
1596
k = self.make_test_knit()
1597
self.assertRaises(KnitHeaderError, k.keys)
1600
class TestGraphIndexKnit(KnitTests):
1601
"""Tests for knits using a GraphIndex rather than a KnitIndex."""
1603
def make_g_index(self, name, ref_lists=0, nodes=[]):
1604
builder = GraphIndexBuilder(ref_lists)
1605
for node, references, value in nodes:
1606
builder.add_node(node, references, value)
1607
stream = builder.finish()
1608
trans = self.get_transport()
1609
size = trans.put_file(name, stream)
1610
return GraphIndex(trans, name, size)
1612
def two_graph_index(self, deltas=False, catch_adds=False):
1613
"""Build a two-graph index.
1615
:param deltas: If true, use underlying indices with two node-ref
1616
lists and 'parent' set to a delta-compressed against tail.
1618
# build a complex graph across several indices.
1620
# delta compression inn the index
1621
index1 = self.make_g_index('1', 2, [
1622
(('tip', ), 'N0 100', ([('parent', )], [], )),
1623
(('tail', ), '', ([], []))])
1624
index2 = self.make_g_index('2', 2, [
1625
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], [('tail', )])),
1626
(('separate', ), '', ([], []))])
1628
# just blob location and graph in the index.
1629
index1 = self.make_g_index('1', 1, [
1630
(('tip', ), 'N0 100', ([('parent', )], )),
1631
(('tail', ), '', ([], ))])
1632
index2 = self.make_g_index('2', 1, [
1633
(('parent', ), ' 100 78', ([('tail', ), ('ghost', )], )),
1634
(('separate', ), '', ([], ))])
1635
combined_index = CombinedGraphIndex([index1, index2])
1637
self.combined_index = combined_index
1638
self.caught_entries = []
1639
add_callback = self.catch_add
1642
return _KnitGraphIndex(combined_index, lambda:True, deltas=deltas,
1643
add_callback=add_callback)
1645
def test_keys(self):
1646
index = self.two_graph_index()
1647
self.assertEqual(set([('tail',), ('tip',), ('parent',), ('separate',)]),
1650
def test_get_position(self):
1651
index = self.two_graph_index()
1652
self.assertEqual((index._graph_index._indices[0], 0, 100), index.get_position(('tip',)))
1653
self.assertEqual((index._graph_index._indices[1], 100, 78), index.get_position(('parent',)))
1655
def test_get_method_deltas(self):
1656
index = self.two_graph_index(deltas=True)
1657
self.assertEqual('fulltext', index.get_method(('tip',)))
1658
self.assertEqual('line-delta', index.get_method(('parent',)))
1660
def test_get_method_no_deltas(self):
1661
# check that the parent-history lookup is ignored with deltas=False.
1662
index = self.two_graph_index(deltas=False)
1663
self.assertEqual('fulltext', index.get_method(('tip',)))
1664
self.assertEqual('fulltext', index.get_method(('parent',)))
1666
def test_get_options_deltas(self):
1667
index = self.two_graph_index(deltas=True)
1668
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1669
self.assertEqual(['line-delta'], index.get_options(('parent',)))
1671
def test_get_options_no_deltas(self):
1672
# check that the parent-history lookup is ignored with deltas=False.
1673
index = self.two_graph_index(deltas=False)
1674
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1675
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1677
def test_get_parent_map(self):
1678
index = self.two_graph_index()
1679
self.assertEqual({('parent',):(('tail',), ('ghost',))},
1680
index.get_parent_map([('parent',), ('ghost',)]))
1682
def catch_add(self, entries):
1683
self.caught_entries.append(entries)
1685
def test_add_no_callback_errors(self):
1686
index = self.two_graph_index()
1687
self.assertRaises(errors.ReadOnlyError, index.add_records,
1688
[(('new',), 'fulltext,no-eol', (None, 50, 60), ['separate'])])
1690
def test_add_version_smoke(self):
1691
index = self.two_graph_index(catch_adds=True)
1692
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60),
1694
self.assertEqual([[(('new', ), 'N50 60', ((('separate',),),))]],
1695
self.caught_entries)
1697
def test_add_version_delta_not_delta_index(self):
1698
index = self.two_graph_index(catch_adds=True)
1699
self.assertRaises(errors.KnitCorrupt, index.add_records,
1700
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1701
self.assertEqual([], self.caught_entries)
1703
def test_add_version_same_dup(self):
1704
index = self.two_graph_index(catch_adds=True)
1705
# options can be spelt two different ways
1706
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
1707
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
1708
# position/length are ignored (because each pack could have fulltext or
1709
# delta, and be at a different position.
1710
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1712
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1714
# but neither should have added data:
1715
self.assertEqual([[], [], [], []], self.caught_entries)
1717
def test_add_version_different_dup(self):
1718
index = self.two_graph_index(deltas=True, catch_adds=True)
1720
self.assertRaises(errors.KnitCorrupt, index.add_records,
1721
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1722
self.assertRaises(errors.KnitCorrupt, index.add_records,
1723
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1725
self.assertRaises(errors.KnitCorrupt, index.add_records,
1726
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1727
self.assertEqual([], self.caught_entries)
1729
def test_add_versions_nodeltas(self):
1730
index = self.two_graph_index(catch_adds=True)
1732
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1733
(('new2',), 'fulltext', (None, 0, 6), [('new',)]),
1735
self.assertEqual([(('new', ), 'N50 60', ((('separate',),),)),
1736
(('new2', ), ' 0 6', ((('new',),),))],
1737
sorted(self.caught_entries[0]))
1738
self.assertEqual(1, len(self.caught_entries))
1740
def test_add_versions_deltas(self):
1741
index = self.two_graph_index(deltas=True, catch_adds=True)
1743
(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)]),
1744
(('new2',), 'line-delta', (None, 0, 6), [('new',)]),
1746
self.assertEqual([(('new', ), 'N50 60', ((('separate',),), ())),
1747
(('new2', ), ' 0 6', ((('new',),), (('new',),), ))],
1748
sorted(self.caught_entries[0]))
1749
self.assertEqual(1, len(self.caught_entries))
1751
def test_add_versions_delta_not_delta_index(self):
1752
index = self.two_graph_index(catch_adds=True)
1753
self.assertRaises(errors.KnitCorrupt, index.add_records,
1754
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1755
self.assertEqual([], self.caught_entries)
1757
def test_add_versions_random_id_accepted(self):
1758
index = self.two_graph_index(catch_adds=True)
1759
index.add_records([], random_id=True)
1761
def test_add_versions_same_dup(self):
1762
index = self.two_graph_index(catch_adds=True)
1763
# options can be spelt two different ways
1764
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100),
1766
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100),
1768
# position/length are ignored (because each pack could have fulltext or
1769
# delta, and be at a different position.
1770
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100),
1772
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000),
1774
# but neither should have added data.
1775
self.assertEqual([[], [], [], []], self.caught_entries)
1777
def test_add_versions_different_dup(self):
1778
index = self.two_graph_index(deltas=True, catch_adds=True)
1780
self.assertRaises(errors.KnitCorrupt, index.add_records,
1781
[(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1782
self.assertRaises(errors.KnitCorrupt, index.add_records,
1783
[(('tip',), 'fulltext', (None, 0, 100), [('parent',)])])
1785
self.assertRaises(errors.KnitCorrupt, index.add_records,
1786
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1787
# change options in the second record
1788
self.assertRaises(errors.KnitCorrupt, index.add_records,
1789
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)]),
1790
(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1791
self.assertEqual([], self.caught_entries)
1793
def make_g_index_missing_compression_parent(self):
1794
graph_index = self.make_g_index('missing_comp', 2,
1795
[(('tip', ), ' 100 78',
1796
([('missing-parent', ), ('ghost', )], [('missing-parent', )]))])
1799
def make_g_index_missing_parent(self):
1800
graph_index = self.make_g_index('missing_parent', 2,
1801
[(('parent', ), ' 100 78', ([], [])),
1802
(('tip', ), ' 100 78',
1803
([('parent', ), ('missing-parent', )], [('parent', )])),
1807
def make_g_index_no_external_refs(self):
1808
graph_index = self.make_g_index('no_external_refs', 2,
1809
[(('rev', ), ' 100 78',
1810
([('parent', ), ('ghost', )], []))])
1813
def test_add_good_unvalidated_index(self):
1814
unvalidated = self.make_g_index_no_external_refs()
1815
combined = CombinedGraphIndex([unvalidated])
1816
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1817
index.scan_unvalidated_index(unvalidated)
1818
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1820
def test_add_missing_compression_parent_unvalidated_index(self):
1821
unvalidated = self.make_g_index_missing_compression_parent()
1822
combined = CombinedGraphIndex([unvalidated])
1823
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1824
index.scan_unvalidated_index(unvalidated)
1825
# This also checks that its only the compression parent that is
1826
# examined, otherwise 'ghost' would also be reported as a missing
1829
frozenset([('missing-parent',)]),
1830
index.get_missing_compression_parents())
1832
def test_add_missing_noncompression_parent_unvalidated_index(self):
1833
unvalidated = self.make_g_index_missing_parent()
1834
combined = CombinedGraphIndex([unvalidated])
1835
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1836
track_external_parent_refs=True)
1837
index.scan_unvalidated_index(unvalidated)
1839
frozenset([('missing-parent',)]), index.get_missing_parents())
1841
def test_track_external_parent_refs(self):
1842
g_index = self.make_g_index('empty', 2, [])
1843
combined = CombinedGraphIndex([g_index])
1844
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1845
add_callback=self.catch_add, track_external_parent_refs=True)
1846
self.caught_entries = []
1848
(('new-key',), 'fulltext,no-eol', (None, 50, 60),
1849
[('parent-1',), ('parent-2',)])])
1851
frozenset([('parent-1',), ('parent-2',)]),
1852
index.get_missing_parents())
1854
def test_add_unvalidated_index_with_present_external_references(self):
1855
index = self.two_graph_index(deltas=True)
1856
# Ugly hack to get at one of the underlying GraphIndex objects that
1857
# two_graph_index built.
1858
unvalidated = index._graph_index._indices[1]
1859
# 'parent' is an external ref of _indices[1] (unvalidated), but is
1860
# present in _indices[0].
1861
index.scan_unvalidated_index(unvalidated)
1862
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1864
def make_new_missing_parent_g_index(self, name):
1865
missing_parent = name + '-missing-parent'
1866
graph_index = self.make_g_index(name, 2,
1867
[((name + 'tip', ), ' 100 78',
1868
([(missing_parent, ), ('ghost', )], [(missing_parent, )]))])
1871
def test_add_mulitiple_unvalidated_indices_with_missing_parents(self):
1872
g_index_1 = self.make_new_missing_parent_g_index('one')
1873
g_index_2 = self.make_new_missing_parent_g_index('two')
1874
combined = CombinedGraphIndex([g_index_1, g_index_2])
1875
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1876
index.scan_unvalidated_index(g_index_1)
1877
index.scan_unvalidated_index(g_index_2)
1879
frozenset([('one-missing-parent',), ('two-missing-parent',)]),
1880
index.get_missing_compression_parents())
1882
def test_add_mulitiple_unvalidated_indices_with_mutual_dependencies(self):
1883
graph_index_a = self.make_g_index('one', 2,
1884
[(('parent-one', ), ' 100 78', ([('non-compression-parent',)], [])),
1885
(('child-of-two', ), ' 100 78',
1886
([('parent-two',)], [('parent-two',)]))])
1887
graph_index_b = self.make_g_index('two', 2,
1888
[(('parent-two', ), ' 100 78', ([('non-compression-parent',)], [])),
1889
(('child-of-one', ), ' 100 78',
1890
([('parent-one',)], [('parent-one',)]))])
1891
combined = CombinedGraphIndex([graph_index_a, graph_index_b])
1892
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1893
index.scan_unvalidated_index(graph_index_a)
1894
index.scan_unvalidated_index(graph_index_b)
1896
frozenset([]), index.get_missing_compression_parents())
1899
class TestNoParentsGraphIndexKnit(KnitTests):
1900
"""Tests for knits using _KnitGraphIndex with no parents."""
1902
def make_g_index(self, name, ref_lists=0, nodes=[]):
1903
builder = GraphIndexBuilder(ref_lists)
1904
for node, references in nodes:
1905
builder.add_node(node, references)
1906
stream = builder.finish()
1907
trans = self.get_transport()
1908
size = trans.put_file(name, stream)
1909
return GraphIndex(trans, name, size)
1911
def test_add_good_unvalidated_index(self):
1912
unvalidated = self.make_g_index('unvalidated')
1913
combined = CombinedGraphIndex([unvalidated])
1914
index = _KnitGraphIndex(combined, lambda: True, parents=False)
1915
index.scan_unvalidated_index(unvalidated)
1916
self.assertEqual(frozenset(),
1917
index.get_missing_compression_parents())
1919
def test_parents_deltas_incompatible(self):
1920
index = CombinedGraphIndex([])
1921
self.assertRaises(errors.KnitError, _KnitGraphIndex, lambda:True,
1922
index, deltas=True, parents=False)
1924
def two_graph_index(self, catch_adds=False):
1925
"""Build a two-graph index.
1927
:param deltas: If true, use underlying indices with two node-ref
1928
lists and 'parent' set to a delta-compressed against tail.
1930
# put several versions in the index.
1931
index1 = self.make_g_index('1', 0, [
1932
(('tip', ), 'N0 100'),
1934
index2 = self.make_g_index('2', 0, [
1935
(('parent', ), ' 100 78'),
1936
(('separate', ), '')])
1937
combined_index = CombinedGraphIndex([index1, index2])
1939
self.combined_index = combined_index
1940
self.caught_entries = []
1941
add_callback = self.catch_add
1944
return _KnitGraphIndex(combined_index, lambda:True, parents=False,
1945
add_callback=add_callback)
1947
def test_keys(self):
1948
index = self.two_graph_index()
1949
self.assertEqual(set([('tail',), ('tip',), ('parent',), ('separate',)]),
1952
def test_get_position(self):
1953
index = self.two_graph_index()
1954
self.assertEqual((index._graph_index._indices[0], 0, 100),
1955
index.get_position(('tip',)))
1956
self.assertEqual((index._graph_index._indices[1], 100, 78),
1957
index.get_position(('parent',)))
1959
def test_get_method(self):
1960
index = self.two_graph_index()
1961
self.assertEqual('fulltext', index.get_method(('tip',)))
1962
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1964
def test_get_options(self):
1965
index = self.two_graph_index()
1966
self.assertEqual(['fulltext', 'no-eol'], index.get_options(('tip',)))
1967
self.assertEqual(['fulltext'], index.get_options(('parent',)))
1969
def test_get_parent_map(self):
1970
index = self.two_graph_index()
1971
self.assertEqual({('parent',):None},
1972
index.get_parent_map([('parent',), ('ghost',)]))
1974
def catch_add(self, entries):
1975
self.caught_entries.append(entries)
1977
def test_add_no_callback_errors(self):
1978
index = self.two_graph_index()
1979
self.assertRaises(errors.ReadOnlyError, index.add_records,
1980
[(('new',), 'fulltext,no-eol', (None, 50, 60), [('separate',)])])
1982
def test_add_version_smoke(self):
1983
index = self.two_graph_index(catch_adds=True)
1984
index.add_records([(('new',), 'fulltext,no-eol', (None, 50, 60), [])])
1985
self.assertEqual([[(('new', ), 'N50 60')]],
1986
self.caught_entries)
1988
def test_add_version_delta_not_delta_index(self):
1989
index = self.two_graph_index(catch_adds=True)
1990
self.assertRaises(errors.KnitCorrupt, index.add_records,
1991
[(('new',), 'no-eol,line-delta', (None, 0, 100), [])])
1992
self.assertEqual([], self.caught_entries)
1994
def test_add_version_same_dup(self):
1995
index = self.two_graph_index(catch_adds=True)
1996
# options can be spelt two different ways
1997
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
1998
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
1999
# position/length are ignored (because each pack could have fulltext or
2000
# delta, and be at a different position.
2001
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
2002
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
2003
# but neither should have added data.
2004
self.assertEqual([[], [], [], []], self.caught_entries)
2006
def test_add_version_different_dup(self):
2007
index = self.two_graph_index(catch_adds=True)
2009
self.assertRaises(errors.KnitCorrupt, index.add_records,
2010
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2011
self.assertRaises(errors.KnitCorrupt, index.add_records,
2012
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
2013
self.assertRaises(errors.KnitCorrupt, index.add_records,
2014
[(('tip',), 'fulltext', (None, 0, 100), [])])
2016
self.assertRaises(errors.KnitCorrupt, index.add_records,
2017
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
2018
self.assertEqual([], self.caught_entries)
2020
def test_add_versions(self):
2021
index = self.two_graph_index(catch_adds=True)
2023
(('new',), 'fulltext,no-eol', (None, 50, 60), []),
2024
(('new2',), 'fulltext', (None, 0, 6), []),
2026
self.assertEqual([(('new', ), 'N50 60'), (('new2', ), ' 0 6')],
2027
sorted(self.caught_entries[0]))
2028
self.assertEqual(1, len(self.caught_entries))
2030
def test_add_versions_delta_not_delta_index(self):
2031
index = self.two_graph_index(catch_adds=True)
2032
self.assertRaises(errors.KnitCorrupt, index.add_records,
2033
[(('new',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
2034
self.assertEqual([], self.caught_entries)
2036
def test_add_versions_parents_not_parents_index(self):
2037
index = self.two_graph_index(catch_adds=True)
2038
self.assertRaises(errors.KnitCorrupt, index.add_records,
2039
[(('new',), 'no-eol,fulltext', (None, 0, 100), [('parent',)])])
2040
self.assertEqual([], self.caught_entries)
2042
def test_add_versions_random_id_accepted(self):
2043
index = self.two_graph_index(catch_adds=True)
2044
index.add_records([], random_id=True)
2046
def test_add_versions_same_dup(self):
2047
index = self.two_graph_index(catch_adds=True)
2048
# options can be spelt two different ways
2049
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 100), [])])
2050
index.add_records([(('tip',), 'no-eol,fulltext', (None, 0, 100), [])])
2051
# position/length are ignored (because each pack could have fulltext or
2052
# delta, and be at a different position.
2053
index.add_records([(('tip',), 'fulltext,no-eol', (None, 50, 100), [])])
2054
index.add_records([(('tip',), 'fulltext,no-eol', (None, 0, 1000), [])])
2055
# but neither should have added data.
2056
self.assertEqual([[], [], [], []], self.caught_entries)
2058
def test_add_versions_different_dup(self):
2059
index = self.two_graph_index(catch_adds=True)
2061
self.assertRaises(errors.KnitCorrupt, index.add_records,
2062
[(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2063
self.assertRaises(errors.KnitCorrupt, index.add_records,
2064
[(('tip',), 'line-delta,no-eol', (None, 0, 100), [])])
2065
self.assertRaises(errors.KnitCorrupt, index.add_records,
2066
[(('tip',), 'fulltext', (None, 0, 100), [])])
2068
self.assertRaises(errors.KnitCorrupt, index.add_records,
2069
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)])])
2070
# change options in the second record
2071
self.assertRaises(errors.KnitCorrupt, index.add_records,
2072
[(('tip',), 'fulltext,no-eol', (None, 0, 100), []),
2073
(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2074
self.assertEqual([], self.caught_entries)
2077
class TestKnitVersionedFiles(KnitTests):
2079
def assertGroupKeysForIo(self, exp_groups, keys, non_local_keys,
2080
positions, _min_buffer_size=None):
2081
kvf = self.make_test_knit()
2082
if _min_buffer_size is None:
2083
_min_buffer_size = knit._STREAM_MIN_BUFFER_SIZE
2084
self.assertEqual(exp_groups, kvf._group_keys_for_io(keys,
2085
non_local_keys, positions,
2086
_min_buffer_size=_min_buffer_size))
2088
def assertSplitByPrefix(self, expected_map, expected_prefix_order,
2090
split, prefix_order = KnitVersionedFiles._split_by_prefix(keys)
2091
self.assertEqual(expected_map, split)
2092
self.assertEqual(expected_prefix_order, prefix_order)
2094
def test__group_keys_for_io(self):
2095
ft_detail = ('fulltext', False)
2096
ld_detail = ('line-delta', False)
2104
f_a: (ft_detail, (f_a, 0, 100), None),
2105
f_b: (ld_detail, (f_b, 100, 21), f_a),
2106
f_c: (ld_detail, (f_c, 180, 15), f_b),
2107
g_a: (ft_detail, (g_a, 121, 35), None),
2108
g_b: (ld_detail, (g_b, 156, 12), g_a),
2109
g_c: (ld_detail, (g_c, 195, 13), g_a),
2111
self.assertGroupKeysForIo([([f_a], set())],
2112
[f_a], [], positions)
2113
self.assertGroupKeysForIo([([f_a], set([f_a]))],
2114
[f_a], [f_a], positions)
2115
self.assertGroupKeysForIo([([f_a, f_b], set([]))],
2116
[f_a, f_b], [], positions)
2117
self.assertGroupKeysForIo([([f_a, f_b], set([f_b]))],
2118
[f_a, f_b], [f_b], positions)
2119
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2120
[f_a, g_a, f_b, g_b], [], positions)
2121
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2122
[f_a, g_a, f_b, g_b], [], positions,
2123
_min_buffer_size=150)
2124
self.assertGroupKeysForIo([([f_a, f_b], set()), ([g_a, g_b], set())],
2125
[f_a, g_a, f_b, g_b], [], positions,
2126
_min_buffer_size=100)
2127
self.assertGroupKeysForIo([([f_c], set()), ([g_b], set())],
2128
[f_c, g_b], [], positions,
2129
_min_buffer_size=125)
2130
self.assertGroupKeysForIo([([g_b, f_c], set())],
2131
[g_b, f_c], [], positions,
2132
_min_buffer_size=125)
2134
def test__split_by_prefix(self):
2135
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2136
'g': [('g', 'b'), ('g', 'a')],
2138
[('f', 'a'), ('g', 'b'),
2139
('g', 'a'), ('f', 'b')])
2141
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2142
'g': [('g', 'b'), ('g', 'a')],
2144
[('f', 'a'), ('f', 'b'),
2145
('g', 'b'), ('g', 'a')])
2147
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2148
'g': [('g', 'b'), ('g', 'a')],
2150
[('f', 'a'), ('f', 'b'),
2151
('g', 'b'), ('g', 'a')])
2153
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2154
'g': [('g', 'b'), ('g', 'a')],
2155
'': [('a',), ('b',)]
2157
[('f', 'a'), ('g', 'b'),
2159
('g', 'a'), ('f', 'b')])
2162
class TestStacking(KnitTests):
2164
def get_basis_and_test_knit(self):
2165
basis = self.make_test_knit(name='basis')
2166
basis = RecordingVersionedFilesDecorator(basis)
2167
test = self.make_test_knit(name='test')
2168
test.add_fallback_versioned_files(basis)
2171
def test_add_fallback_versioned_files(self):
2172
basis = self.make_test_knit(name='basis')
2173
test = self.make_test_knit(name='test')
2174
# It must not error; other tests test that the fallback is referred to
2175
# when accessing data.
2176
test.add_fallback_versioned_files(basis)
2178
def test_add_lines(self):
2179
# lines added to the test are not added to the basis
2180
basis, test = self.get_basis_and_test_knit()
2182
key_basis = ('bar',)
2183
key_cross_border = ('quux',)
2184
key_delta = ('zaphod',)
2185
test.add_lines(key, (), ['foo\n'])
2186
self.assertEqual({}, basis.get_parent_map([key]))
2187
# lines added to the test that reference across the stack do a
2189
basis.add_lines(key_basis, (), ['foo\n'])
2191
test.add_lines(key_cross_border, (key_basis,), ['foo\n'])
2192
self.assertEqual('fulltext', test._index.get_method(key_cross_border))
2193
# we don't even need to look at the basis to see that this should be
2194
# stored as a fulltext
2195
self.assertEqual([], basis.calls)
2196
# Subsequent adds do delta.
2198
test.add_lines(key_delta, (key_cross_border,), ['foo\n'])
2199
self.assertEqual('line-delta', test._index.get_method(key_delta))
2200
self.assertEqual([], basis.calls)
2202
def test_annotate(self):
2203
# annotations from the test knit are answered without asking the basis
2204
basis, test = self.get_basis_and_test_knit()
2206
key_basis = ('bar',)
2207
key_missing = ('missing',)
2208
test.add_lines(key, (), ['foo\n'])
2209
details = test.annotate(key)
2210
self.assertEqual([(key, 'foo\n')], details)
2211
self.assertEqual([], basis.calls)
2212
# But texts that are not in the test knit are looked for in the basis
2214
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2216
details = test.annotate(key_basis)
2217
self.assertEqual([(key_basis, 'foo\n'), (key_basis, 'bar\n')], details)
2218
# Not optimised to date:
2219
# self.assertEqual([("annotate", key_basis)], basis.calls)
2220
self.assertEqual([('get_parent_map', set([key_basis])),
2221
('get_parent_map', set([key_basis])),
2222
('get_record_stream', [key_basis], 'topological', True)],
2225
def test_check(self):
2226
# At the moment checking a stacked knit does implicitly check the
2228
basis, test = self.get_basis_and_test_knit()
2231
def test_get_parent_map(self):
2232
# parents in the test knit are answered without asking the basis
2233
basis, test = self.get_basis_and_test_knit()
2235
key_basis = ('bar',)
2236
key_missing = ('missing',)
2237
test.add_lines(key, (), [])
2238
parent_map = test.get_parent_map([key])
2239
self.assertEqual({key: ()}, parent_map)
2240
self.assertEqual([], basis.calls)
2241
# But parents that are not in the test knit are looked for in the basis
2242
basis.add_lines(key_basis, (), [])
2244
parent_map = test.get_parent_map([key, key_basis, key_missing])
2245
self.assertEqual({key: (),
2246
key_basis: ()}, parent_map)
2247
self.assertEqual([("get_parent_map", set([key_basis, key_missing]))],
2250
def test_get_record_stream_unordered_fulltexts(self):
2251
# records from the test knit are answered without asking the basis:
2252
basis, test = self.get_basis_and_test_knit()
2254
key_basis = ('bar',)
2255
key_missing = ('missing',)
2256
test.add_lines(key, (), ['foo\n'])
2257
records = list(test.get_record_stream([key], 'unordered', True))
2258
self.assertEqual(1, len(records))
2259
self.assertEqual([], basis.calls)
2260
# Missing (from test knit) objects are retrieved from the basis:
2261
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2263
records = list(test.get_record_stream([key_basis, key_missing],
2265
self.assertEqual(2, len(records))
2266
calls = list(basis.calls)
2267
for record in records:
2268
self.assertSubset([record.key], (key_basis, key_missing))
2269
if record.key == key_missing:
2270
self.assertIsInstance(record, AbsentContentFactory)
2272
reference = list(basis.get_record_stream([key_basis],
2273
'unordered', True))[0]
2274
self.assertEqual(reference.key, record.key)
2275
self.assertEqual(reference.sha1, record.sha1)
2276
self.assertEqual(reference.storage_kind, record.storage_kind)
2277
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2278
record.get_bytes_as(record.storage_kind))
2279
self.assertEqual(reference.get_bytes_as('fulltext'),
2280
record.get_bytes_as('fulltext'))
2281
# It's not strictly minimal, but it seems reasonable for now for it to
2282
# ask which fallbacks have which parents.
2284
("get_parent_map", set([key_basis, key_missing])),
2285
("get_record_stream", [key_basis], 'unordered', True)],
2288
def test_get_record_stream_ordered_fulltexts(self):
2289
# ordering is preserved down into the fallback store.
2290
basis, test = self.get_basis_and_test_knit()
2292
key_basis = ('bar',)
2293
key_basis_2 = ('quux',)
2294
key_missing = ('missing',)
2295
test.add_lines(key, (key_basis,), ['foo\n'])
2296
# Missing (from test knit) objects are retrieved from the basis:
2297
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2298
basis.add_lines(key_basis_2, (), ['quux\n'])
2300
# ask for in non-topological order
2301
records = list(test.get_record_stream(
2302
[key, key_basis, key_missing, key_basis_2], 'topological', True))
2303
self.assertEqual(4, len(records))
2305
for record in records:
2306
self.assertSubset([record.key],
2307
(key_basis, key_missing, key_basis_2, key))
2308
if record.key == key_missing:
2309
self.assertIsInstance(record, AbsentContentFactory)
2311
results.append((record.key, record.sha1, record.storage_kind,
2312
record.get_bytes_as('fulltext')))
2313
calls = list(basis.calls)
2314
order = [record[0] for record in results]
2315
self.assertEqual([key_basis_2, key_basis, key], order)
2316
for result in results:
2317
if result[0] == key:
2321
record = source.get_record_stream([result[0]], 'unordered',
2323
self.assertEqual(record.key, result[0])
2324
self.assertEqual(record.sha1, result[1])
2325
# We used to check that the storage kind matched, but actually it
2326
# depends on whether it was sourced from the basis, or in a single
2327
# group, because asking for full texts returns proxy objects to a
2328
# _ContentMapGenerator object; so checking the kind is unneeded.
2329
self.assertEqual(record.get_bytes_as('fulltext'), result[3])
2330
# It's not strictly minimal, but it seems reasonable for now for it to
2331
# ask which fallbacks have which parents.
2333
("get_parent_map", set([key_basis, key_basis_2, key_missing])),
2334
# topological is requested from the fallback, because that is what
2335
# was requested at the top level.
2336
("get_record_stream", [key_basis_2, key_basis], 'topological', True)],
2339
def test_get_record_stream_unordered_deltas(self):
2340
# records from the test knit are answered without asking the basis:
2341
basis, test = self.get_basis_and_test_knit()
2343
key_basis = ('bar',)
2344
key_missing = ('missing',)
2345
test.add_lines(key, (), ['foo\n'])
2346
records = list(test.get_record_stream([key], 'unordered', False))
2347
self.assertEqual(1, len(records))
2348
self.assertEqual([], basis.calls)
2349
# Missing (from test knit) objects are retrieved from the basis:
2350
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2352
records = list(test.get_record_stream([key_basis, key_missing],
2353
'unordered', False))
2354
self.assertEqual(2, len(records))
2355
calls = list(basis.calls)
2356
for record in records:
2357
self.assertSubset([record.key], (key_basis, key_missing))
2358
if record.key == key_missing:
2359
self.assertIsInstance(record, AbsentContentFactory)
2361
reference = list(basis.get_record_stream([key_basis],
2362
'unordered', False))[0]
2363
self.assertEqual(reference.key, record.key)
2364
self.assertEqual(reference.sha1, record.sha1)
2365
self.assertEqual(reference.storage_kind, record.storage_kind)
2366
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2367
record.get_bytes_as(record.storage_kind))
2368
# It's not strictly minimal, but it seems reasonable for now for it to
2369
# ask which fallbacks have which parents.
2371
("get_parent_map", set([key_basis, key_missing])),
2372
("get_record_stream", [key_basis], 'unordered', False)],
2375
def test_get_record_stream_ordered_deltas(self):
2376
# ordering is preserved down into the fallback store.
2377
basis, test = self.get_basis_and_test_knit()
2379
key_basis = ('bar',)
2380
key_basis_2 = ('quux',)
2381
key_missing = ('missing',)
2382
test.add_lines(key, (key_basis,), ['foo\n'])
2383
# Missing (from test knit) objects are retrieved from the basis:
2384
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2385
basis.add_lines(key_basis_2, (), ['quux\n'])
2387
# ask for in non-topological order
2388
records = list(test.get_record_stream(
2389
[key, key_basis, key_missing, key_basis_2], 'topological', False))
2390
self.assertEqual(4, len(records))
2392
for record in records:
2393
self.assertSubset([record.key],
2394
(key_basis, key_missing, key_basis_2, key))
2395
if record.key == key_missing:
2396
self.assertIsInstance(record, AbsentContentFactory)
2398
results.append((record.key, record.sha1, record.storage_kind,
2399
record.get_bytes_as(record.storage_kind)))
2400
calls = list(basis.calls)
2401
order = [record[0] for record in results]
2402
self.assertEqual([key_basis_2, key_basis, key], order)
2403
for result in results:
2404
if result[0] == key:
2408
record = source.get_record_stream([result[0]], 'unordered',
2410
self.assertEqual(record.key, result[0])
2411
self.assertEqual(record.sha1, result[1])
2412
self.assertEqual(record.storage_kind, result[2])
2413
self.assertEqual(record.get_bytes_as(record.storage_kind), result[3])
2414
# It's not strictly minimal, but it seems reasonable for now for it to
2415
# ask which fallbacks have which parents.
2417
("get_parent_map", set([key_basis, key_basis_2, key_missing])),
2418
("get_record_stream", [key_basis_2, key_basis], 'topological', False)],
2421
def test_get_sha1s(self):
2422
# sha1's in the test knit are answered without asking the basis
2423
basis, test = self.get_basis_and_test_knit()
2425
key_basis = ('bar',)
2426
key_missing = ('missing',)
2427
test.add_lines(key, (), ['foo\n'])
2428
key_sha1sum = osutils.sha('foo\n').hexdigest()
2429
sha1s = test.get_sha1s([key])
2430
self.assertEqual({key: key_sha1sum}, sha1s)
2431
self.assertEqual([], basis.calls)
2432
# But texts that are not in the test knit are looked for in the basis
2433
# directly (rather than via text reconstruction) so that remote servers
2434
# etc don't have to answer with full content.
2435
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2436
basis_sha1sum = osutils.sha('foo\nbar\n').hexdigest()
2438
sha1s = test.get_sha1s([key, key_missing, key_basis])
2439
self.assertEqual({key: key_sha1sum,
2440
key_basis: basis_sha1sum}, sha1s)
2441
self.assertEqual([("get_sha1s", set([key_basis, key_missing]))],
2444
def test_insert_record_stream(self):
2445
# records are inserted as normal; insert_record_stream builds on
2446
# add_lines, so a smoke test should be all that's needed:
2448
key_basis = ('bar',)
2449
key_delta = ('zaphod',)
2450
basis, test = self.get_basis_and_test_knit()
2451
source = self.make_test_knit(name='source')
2452
basis.add_lines(key_basis, (), ['foo\n'])
2454
source.add_lines(key_basis, (), ['foo\n'])
2455
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2456
stream = source.get_record_stream([key_delta], 'unordered', False)
2457
test.insert_record_stream(stream)
2458
# XXX: this does somewhat too many calls in making sure of whether it
2459
# has to recreate the full text.
2460
self.assertEqual([("get_parent_map", set([key_basis])),
2461
('get_parent_map', set([key_basis])),
2462
('get_record_stream', [key_basis], 'unordered', True)],
2464
self.assertEqual({key_delta:(key_basis,)},
2465
test.get_parent_map([key_delta]))
2466
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2467
'unordered', True).next().get_bytes_as('fulltext'))
2469
def test_iter_lines_added_or_present_in_keys(self):
2470
# Lines from the basis are returned, and lines for a given key are only
2474
# all sources are asked for keys:
2475
basis, test = self.get_basis_and_test_knit()
2476
basis.add_lines(key1, (), ["foo"])
2478
lines = list(test.iter_lines_added_or_present_in_keys([key1]))
2479
self.assertEqual([("foo\n", key1)], lines)
2480
self.assertEqual([("iter_lines_added_or_present_in_keys", set([key1]))],
2482
# keys in both are not duplicated:
2483
test.add_lines(key2, (), ["bar\n"])
2484
basis.add_lines(key2, (), ["bar\n"])
2486
lines = list(test.iter_lines_added_or_present_in_keys([key2]))
2487
self.assertEqual([("bar\n", key2)], lines)
2488
self.assertEqual([], basis.calls)
2490
def test_keys(self):
2493
# all sources are asked for keys:
2494
basis, test = self.get_basis_and_test_knit()
2496
self.assertEqual(set(), set(keys))
2497
self.assertEqual([("keys",)], basis.calls)
2498
# keys from a basis are returned:
2499
basis.add_lines(key1, (), [])
2502
self.assertEqual(set([key1]), set(keys))
2503
self.assertEqual([("keys",)], basis.calls)
2504
# keys in both are not duplicated:
2505
test.add_lines(key2, (), [])
2506
basis.add_lines(key2, (), [])
2509
self.assertEqual(2, len(keys))
2510
self.assertEqual(set([key1, key2]), set(keys))
2511
self.assertEqual([("keys",)], basis.calls)
2513
def test_add_mpdiffs(self):
2514
# records are inserted as normal; add_mpdiff builds on
2515
# add_lines, so a smoke test should be all that's needed:
2517
key_basis = ('bar',)
2518
key_delta = ('zaphod',)
2519
basis, test = self.get_basis_and_test_knit()
2520
source = self.make_test_knit(name='source')
2521
basis.add_lines(key_basis, (), ['foo\n'])
2523
source.add_lines(key_basis, (), ['foo\n'])
2524
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2525
diffs = source.make_mpdiffs([key_delta])
2526
test.add_mpdiffs([(key_delta, (key_basis,),
2527
source.get_sha1s([key_delta])[key_delta], diffs[0])])
2528
self.assertEqual([("get_parent_map", set([key_basis])),
2529
('get_record_stream', [key_basis], 'unordered', True),],
2531
self.assertEqual({key_delta:(key_basis,)},
2532
test.get_parent_map([key_delta]))
2533
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2534
'unordered', True).next().get_bytes_as('fulltext'))
2536
def test_make_mpdiffs(self):
2537
# Generating an mpdiff across a stacking boundary should detect parent
2541
key_right = ('zaphod',)
2542
basis, test = self.get_basis_and_test_knit()
2543
basis.add_lines(key_left, (), ['bar\n'])
2544
basis.add_lines(key_right, (), ['zaphod\n'])
2546
test.add_lines(key, (key_left, key_right),
2547
['bar\n', 'foo\n', 'zaphod\n'])
2548
diffs = test.make_mpdiffs([key])
2550
multiparent.MultiParent([multiparent.ParentText(0, 0, 0, 1),
2551
multiparent.NewText(['foo\n']),
2552
multiparent.ParentText(1, 0, 2, 1)])],
2554
self.assertEqual(3, len(basis.calls))
2556
("get_parent_map", set([key_left, key_right])),
2557
("get_parent_map", set([key_left, key_right])),
2560
last_call = basis.calls[-1]
2561
self.assertEqual('get_record_stream', last_call[0])
2562
self.assertEqual(set([key_left, key_right]), set(last_call[1]))
2563
self.assertEqual('topological', last_call[2])
2564
self.assertEqual(True, last_call[3])
2567
class TestNetworkBehaviour(KnitTests):
2568
"""Tests for getting data out of/into knits over the network."""
2570
def test_include_delta_closure_generates_a_knit_delta_closure(self):
2571
vf = self.make_test_knit(name='test')
2572
# put in three texts, giving ft, delta, delta
2573
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2574
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2575
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2576
# But heuristics could interfere, so check what happened:
2577
self.assertEqual(['knit-ft-gz', 'knit-delta-gz', 'knit-delta-gz'],
2578
[record.storage_kind for record in
2579
vf.get_record_stream([('base',), ('d1',), ('d2',)],
2580
'topological', False)])
2581
# generate a stream of just the deltas include_delta_closure=True,
2582
# serialise to the network, and check that we get a delta closure on the wire.
2583
stream = vf.get_record_stream([('d1',), ('d2',)], 'topological', True)
2584
netb = [record.get_bytes_as(record.storage_kind) for record in stream]
2585
# The first bytes should be a memo from _ContentMapGenerator, and the
2586
# second bytes should be empty (because its a API proxy not something
2587
# for wire serialisation.
2588
self.assertEqual('', netb[1])
2590
kind, line_end = network_bytes_to_kind_and_offset(bytes)
2591
self.assertEqual('knit-delta-closure', kind)
2594
class TestContentMapGenerator(KnitTests):
2595
"""Tests for ContentMapGenerator"""
2597
def test_get_record_stream_gives_records(self):
2598
vf = self.make_test_knit(name='test')
2599
# put in three texts, giving ft, delta, delta
2600
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2601
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2602
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2603
keys = [('d1',), ('d2',)]
2604
generator = _VFContentMapGenerator(vf, keys,
2605
global_map=vf.get_parent_map(keys))
2606
for record in generator.get_record_stream():
2607
if record.key == ('d1',):
2608
self.assertEqual('d1\n', record.get_bytes_as('fulltext'))
2610
self.assertEqual('d2\n', record.get_bytes_as('fulltext'))
2612
def test_get_record_stream_kinds_are_raw(self):
2613
vf = self.make_test_knit(name='test')
2614
# put in three texts, giving ft, delta, delta
2615
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2616
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2617
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2618
keys = [('base',), ('d1',), ('d2',)]
2619
generator = _VFContentMapGenerator(vf, keys,
2620
global_map=vf.get_parent_map(keys))
2621
kinds = {('base',): 'knit-delta-closure',
2622
('d1',): 'knit-delta-closure-ref',
2623
('d2',): 'knit-delta-closure-ref',
2625
for record in generator.get_record_stream():
2626
self.assertEqual(kinds[record.key], record.storage_kind)
added
removed
bzrlib/tests/test_knit.py
