323
347
access.set_writer(writer, index, (transport, packname))
324
348
return access, writer
350
def make_pack_file(self):
351
"""Create a pack file with 2 records."""
352
access, writer = self._get_access(packname='packname', index='foo')
354
memos.extend(access.add_raw_records([('key1', 10)], '1234567890'))
355
memos.extend(access.add_raw_records([('key2', 5)], '12345'))
359
def make_vf_for_retrying(self):
360
"""Create 3 packs and a reload function.
362
Originally, 2 pack files will have the data, but one will be missing.
363
And then the third will be used in place of the first two if reload()
366
:return: (versioned_file, reload_counter)
367
versioned_file a KnitVersionedFiles using the packs for access
369
builder = self.make_branch_builder('.', format="1.9")
370
builder.start_series()
371
builder.build_snapshot('rev-1', None, [
372
('add', ('', 'root-id', 'directory', None)),
373
('add', ('file', 'file-id', 'file', 'content\nrev 1\n')),
375
builder.build_snapshot('rev-2', ['rev-1'], [
376
('modify', ('file-id', 'content\nrev 2\n')),
378
builder.build_snapshot('rev-3', ['rev-2'], [
379
('modify', ('file-id', 'content\nrev 3\n')),
381
builder.finish_series()
382
b = builder.get_branch()
384
self.addCleanup(b.unlock)
385
# Pack these three revisions into another pack file, but don't remove
388
collection = repo._pack_collection
389
collection.ensure_loaded()
390
orig_packs = collection.packs
391
packer = pack_repo.Packer(collection, orig_packs, '.testpack')
392
new_pack = packer.pack()
393
# forget about the new pack
397
# Set up a reload() function that switches to using the new pack file
398
new_index = new_pack.revision_index
399
access_tuple = new_pack.access_tuple()
400
reload_counter = [0, 0, 0]
402
reload_counter[0] += 1
403
if reload_counter[1] > 0:
404
# We already reloaded, nothing more to do
405
reload_counter[2] += 1
407
reload_counter[1] += 1
408
vf._index._graph_index._indices[:] = [new_index]
409
vf._access._indices.clear()
410
vf._access._indices[new_index] = access_tuple
412
# Delete one of the pack files so the data will need to be reloaded. We
413
# will delete the file with 'rev-2' in it
414
trans, name = orig_packs[1].access_tuple()
416
# We don't have the index trigger reloading because we want to test
417
# that we reload when the .pack disappears
418
vf._access._reload_func = reload
419
return vf, reload_counter
421
def make_reload_func(self, return_val=True):
424
reload_called[0] += 1
426
return reload_called, reload
428
def make_retry_exception(self):
429
# We raise a real exception so that sys.exc_info() is properly
432
raise _TestException('foobar')
433
except _TestException, e:
434
retry_exc = errors.RetryWithNewPacks(None, reload_occurred=False,
435
exc_info=sys.exc_info())
326
438
def test_read_from_several_packs(self):
327
439
access, writer = self._get_access()
365
477
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
479
def test_missing_index_raises_retry(self):
480
memos = self.make_pack_file()
481
transport = self.get_transport()
482
reload_called, reload_func = self.make_reload_func()
483
# Note that the index key has changed from 'foo' to 'bar'
484
access = _DirectPackAccess({'bar':(transport, 'packname')},
485
reload_func=reload_func)
486
e = self.assertListRaises(errors.RetryWithNewPacks,
487
access.get_raw_records, memos)
488
# Because a key was passed in which does not match our index list, we
489
# assume that the listing was already reloaded
490
self.assertTrue(e.reload_occurred)
491
self.assertIsInstance(e.exc_info, tuple)
492
self.assertIs(e.exc_info[0], KeyError)
493
self.assertIsInstance(e.exc_info[1], KeyError)
495
def test_missing_index_raises_key_error_with_no_reload(self):
496
memos = self.make_pack_file()
497
transport = self.get_transport()
498
# Note that the index key has changed from 'foo' to 'bar'
499
access = _DirectPackAccess({'bar':(transport, 'packname')})
500
e = self.assertListRaises(KeyError, access.get_raw_records, memos)
502
def test_missing_file_raises_retry(self):
503
memos = self.make_pack_file()
504
transport = self.get_transport()
505
reload_called, reload_func = self.make_reload_func()
506
# Note that the 'filename' has been changed to 'different-packname'
507
access = _DirectPackAccess({'foo':(transport, 'different-packname')},
508
reload_func=reload_func)
509
e = self.assertListRaises(errors.RetryWithNewPacks,
510
access.get_raw_records, memos)
511
# The file has gone missing, so we assume we need to reload
512
self.assertFalse(e.reload_occurred)
513
self.assertIsInstance(e.exc_info, tuple)
514
self.assertIs(e.exc_info[0], errors.NoSuchFile)
515
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
516
self.assertEqual('different-packname', e.exc_info[1].path)
518
def test_missing_file_raises_no_such_file_with_no_reload(self):
519
memos = self.make_pack_file()
520
transport = self.get_transport()
521
# Note that the 'filename' has been changed to 'different-packname'
522
access = _DirectPackAccess({'foo':(transport, 'different-packname')})
523
e = self.assertListRaises(errors.NoSuchFile,
524
access.get_raw_records, memos)
526
def test_failing_readv_raises_retry(self):
527
memos = self.make_pack_file()
528
transport = self.get_transport()
529
failing_transport = MockReadvFailingTransport(
530
[transport.get_bytes('packname')])
531
reload_called, reload_func = self.make_reload_func()
532
access = _DirectPackAccess({'foo':(failing_transport, 'packname')},
533
reload_func=reload_func)
534
# Asking for a single record will not trigger the Mock failure
535
self.assertEqual(['1234567890'],
536
list(access.get_raw_records(memos[:1])))
537
self.assertEqual(['12345'],
538
list(access.get_raw_records(memos[1:2])))
539
# A multiple offset readv() will fail mid-way through
540
e = self.assertListRaises(errors.RetryWithNewPacks,
541
access.get_raw_records, memos)
542
# The file has gone missing, so we assume we need to reload
543
self.assertFalse(e.reload_occurred)
544
self.assertIsInstance(e.exc_info, tuple)
545
self.assertIs(e.exc_info[0], errors.NoSuchFile)
546
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
547
self.assertEqual('packname', e.exc_info[1].path)
549
def test_failing_readv_raises_no_such_file_with_no_reload(self):
550
memos = self.make_pack_file()
551
transport = self.get_transport()
552
failing_transport = MockReadvFailingTransport(
553
[transport.get_bytes('packname')])
554
reload_called, reload_func = self.make_reload_func()
555
access = _DirectPackAccess({'foo':(failing_transport, 'packname')})
556
# Asking for a single record will not trigger the Mock failure
557
self.assertEqual(['1234567890'],
558
list(access.get_raw_records(memos[:1])))
559
self.assertEqual(['12345'],
560
list(access.get_raw_records(memos[1:2])))
561
# A multiple offset readv() will fail mid-way through
562
e = self.assertListRaises(errors.NoSuchFile,
563
access.get_raw_records, memos)
565
def test_reload_or_raise_no_reload(self):
566
access = _DirectPackAccess({}, reload_func=None)
567
retry_exc = self.make_retry_exception()
568
# Without a reload_func, we will just re-raise the original exception
569
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
571
def test_reload_or_raise_reload_changed(self):
572
reload_called, reload_func = self.make_reload_func(return_val=True)
573
access = _DirectPackAccess({}, reload_func=reload_func)
574
retry_exc = self.make_retry_exception()
575
access.reload_or_raise(retry_exc)
576
self.assertEqual([1], reload_called)
577
retry_exc.reload_occurred=True
578
access.reload_or_raise(retry_exc)
579
self.assertEqual([2], reload_called)
581
def test_reload_or_raise_reload_no_change(self):
582
reload_called, reload_func = self.make_reload_func(return_val=False)
583
access = _DirectPackAccess({}, reload_func=reload_func)
584
retry_exc = self.make_retry_exception()
585
# If reload_occurred is False, then we consider it an error to have
586
# reload_func() return False (no changes).
587
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
588
self.assertEqual([1], reload_called)
589
retry_exc.reload_occurred=True
590
# If reload_occurred is True, then we assume nothing changed because
591
# it had changed earlier, but didn't change again
592
access.reload_or_raise(retry_exc)
593
self.assertEqual([2], reload_called)
595
def test_annotate_retries(self):
596
vf, reload_counter = self.make_vf_for_retrying()
597
# It is a little bit bogus to annotate the Revision VF, but it works,
598
# as we have ancestry stored there
600
reload_lines = vf.annotate(key)
601
self.assertEqual([1, 1, 0], reload_counter)
602
plain_lines = vf.annotate(key)
603
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
604
if reload_lines != plain_lines:
605
self.fail('Annotation was not identical with reloading.')
606
# Now delete the packs-in-use, which should trigger another reload, but
607
# this time we just raise an exception because we can't recover
608
for trans, name in vf._access._indices.itervalues():
610
self.assertRaises(errors.NoSuchFile, vf.annotate, key)
611
self.assertEqual([2, 1, 1], reload_counter)
613
def test__get_record_map_retries(self):
614
vf, reload_counter = self.make_vf_for_retrying()
615
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
616
records = vf._get_record_map(keys)
617
self.assertEqual(keys, sorted(records.keys()))
618
self.assertEqual([1, 1, 0], reload_counter)
619
# Now delete the packs-in-use, which should trigger another reload, but
620
# this time we just raise an exception because we can't recover
621
for trans, name in vf._access._indices.itervalues():
623
self.assertRaises(errors.NoSuchFile, vf._get_record_map, keys)
624
self.assertEqual([2, 1, 1], reload_counter)
626
def test_get_record_stream_retries(self):
627
vf, reload_counter = self.make_vf_for_retrying()
628
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
629
record_stream = vf.get_record_stream(keys, 'topological', False)
630
record = record_stream.next()
631
self.assertEqual(('rev-1',), record.key)
632
self.assertEqual([0, 0, 0], reload_counter)
633
record = record_stream.next()
634
self.assertEqual(('rev-2',), record.key)
635
self.assertEqual([1, 1, 0], reload_counter)
636
record = record_stream.next()
637
self.assertEqual(('rev-3',), record.key)
638
self.assertEqual([1, 1, 0], reload_counter)
639
# Now delete all pack files, and see that we raise the right error
640
for trans, name in vf._access._indices.itervalues():
642
self.assertListRaises(errors.NoSuchFile,
643
vf.get_record_stream, keys, 'topological', False)
645
def test_iter_lines_added_or_present_in_keys_retries(self):
646
vf, reload_counter = self.make_vf_for_retrying()
647
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
648
# Unfortunately, iter_lines_added_or_present_in_keys iterates the
649
# result in random order (determined by the iteration order from a
650
# set()), so we don't have any solid way to trigger whether data is
651
# read before or after. However we tried to delete the middle node to
652
# exercise the code well.
653
# What we care about is that all lines are always yielded, but not
656
reload_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
657
self.assertEqual([1, 1, 0], reload_counter)
658
# Now do it again, to make sure the result is equivalent
659
plain_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
660
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
661
self.assertEqual(plain_lines, reload_lines)
662
self.assertEqual(21, len(plain_lines))
663
# Now delete all pack files, and see that we raise the right error
664
for trans, name in vf._access._indices.itervalues():
666
self.assertListRaises(errors.NoSuchFile,
667
vf.iter_lines_added_or_present_in_keys, keys)
668
self.assertEqual([2, 1, 1], reload_counter)
670
def test_get_record_stream_yields_disk_sorted_order(self):
671
# if we get 'unordered' pick a semi-optimal order for reading. The
672
# order should be grouped by pack file, and then by position in file
673
repo = self.make_repository('test', format='pack-0.92')
675
self.addCleanup(repo.unlock)
676
repo.start_write_group()
678
vf.add_lines(('f-id', 'rev-5'), [('f-id', 'rev-4')], ['lines\n'])
679
vf.add_lines(('f-id', 'rev-1'), [], ['lines\n'])
680
vf.add_lines(('f-id', 'rev-2'), [('f-id', 'rev-1')], ['lines\n'])
681
repo.commit_write_group()
682
# We inserted them as rev-5, rev-1, rev-2, we should get them back in
684
stream = vf.get_record_stream([('f-id', 'rev-1'), ('f-id', 'rev-5'),
685
('f-id', 'rev-2')], 'unordered', False)
686
keys = [r.key for r in stream]
687
self.assertEqual([('f-id', 'rev-5'), ('f-id', 'rev-1'),
688
('f-id', 'rev-2')], keys)
689
repo.start_write_group()
690
vf.add_lines(('f-id', 'rev-4'), [('f-id', 'rev-3')], ['lines\n'])
691
vf.add_lines(('f-id', 'rev-3'), [('f-id', 'rev-2')], ['lines\n'])
692
vf.add_lines(('f-id', 'rev-6'), [('f-id', 'rev-5')], ['lines\n'])
693
repo.commit_write_group()
694
# Request in random order, to make sure the output order isn't based on
696
request_keys = set(('f-id', 'rev-%d' % i) for i in range(1, 7))
697
stream = vf.get_record_stream(request_keys, 'unordered', False)
698
keys = [r.key for r in stream]
699
# We want to get the keys back in disk order, but it doesn't matter
700
# which pack we read from first. So this can come back in 2 orders
701
alt1 = [('f-id', 'rev-%d' % i) for i in [4, 3, 6, 5, 1, 2]]
702
alt2 = [('f-id', 'rev-%d' % i) for i in [5, 1, 2, 4, 3, 6]]
703
if keys != alt1 and keys != alt2:
704
self.fail('Returned key order did not match either expected order.'
705
' expected %s or %s, not %s'
706
% (alt1, alt2, keys))
368
709
class LowLevelKnitDataTests(TestCase):
909
1317
knit._load_data = orig
910
1318
self.addCleanup(reset)
911
from bzrlib._knit_load_data_c import _load_data_c
1319
from bzrlib._knit_load_data_pyx import _load_data_c
912
1320
knit._load_data = _load_data_c
913
1321
allow_writes = lambda: mode == 'w'
914
1322
return _KndxIndex(transport, mapper, lambda:None, allow_writes, lambda:True)
1325
class Test_KnitAnnotator(TestCaseWithMemoryTransport):
1327
def make_annotator(self):
1328
factory = knit.make_pack_factory(True, True, 1)
1329
vf = factory(self.get_transport())
1330
return knit._KnitAnnotator(vf)
1332
def test__expand_fulltext(self):
1333
ann = self.make_annotator()
1334
rev_key = ('rev-id',)
1335
ann._num_compression_children[rev_key] = 1
1336
res = ann._expand_record(rev_key, (('parent-id',),), None,
1337
['line1\n', 'line2\n'], ('fulltext', True))
1338
# The content object and text lines should be cached appropriately
1339
self.assertEqual(['line1\n', 'line2'], res)
1340
content_obj = ann._content_objects[rev_key]
1341
self.assertEqual(['line1\n', 'line2\n'], content_obj._lines)
1342
self.assertEqual(res, content_obj.text())
1343
self.assertEqual(res, ann._text_cache[rev_key])
1345
def test__expand_delta_comp_parent_not_available(self):
1346
# Parent isn't available yet, so we return nothing, but queue up this
1347
# node for later processing
1348
ann = self.make_annotator()
1349
rev_key = ('rev-id',)
1350
parent_key = ('parent-id',)
1351
record = ['0,1,1\n', 'new-line\n']
1352
details = ('line-delta', False)
1353
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1355
self.assertEqual(None, res)
1356
self.assertTrue(parent_key in ann._pending_deltas)
1357
pending = ann._pending_deltas[parent_key]
1358
self.assertEqual(1, len(pending))
1359
self.assertEqual((rev_key, (parent_key,), record, details), pending[0])
1361
def test__expand_record_tracks_num_children(self):
1362
ann = self.make_annotator()
1363
rev_key = ('rev-id',)
1364
rev2_key = ('rev2-id',)
1365
parent_key = ('parent-id',)
1366
record = ['0,1,1\n', 'new-line\n']
1367
details = ('line-delta', False)
1368
ann._num_compression_children[parent_key] = 2
1369
ann._expand_record(parent_key, (), None, ['line1\n', 'line2\n'],
1370
('fulltext', False))
1371
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1373
self.assertEqual({parent_key: 1}, ann._num_compression_children)
1374
# Expanding the second child should remove the content object, and the
1375
# num_compression_children entry
1376
res = ann._expand_record(rev2_key, (parent_key,), parent_key,
1378
self.assertFalse(parent_key in ann._content_objects)
1379
self.assertEqual({}, ann._num_compression_children)
1380
# We should not cache the content_objects for rev2 and rev, because
1381
# they do not have compression children of their own.
1382
self.assertEqual({}, ann._content_objects)
1384
def test__expand_delta_records_blocks(self):
1385
ann = self.make_annotator()
1386
rev_key = ('rev-id',)
1387
parent_key = ('parent-id',)
1388
record = ['0,1,1\n', 'new-line\n']
1389
details = ('line-delta', True)
1390
ann._num_compression_children[parent_key] = 2
1391
ann._expand_record(parent_key, (), None,
1392
['line1\n', 'line2\n', 'line3\n'],
1393
('fulltext', False))
1394
ann._expand_record(rev_key, (parent_key,), parent_key, record, details)
1395
self.assertEqual({(rev_key, parent_key): [(1, 1, 1), (3, 3, 0)]},
1396
ann._matching_blocks)
1397
rev2_key = ('rev2-id',)
1398
record = ['0,1,1\n', 'new-line\n']
1399
details = ('line-delta', False)
1400
ann._expand_record(rev2_key, (parent_key,), parent_key, record, details)
1401
self.assertEqual([(1, 1, 2), (3, 3, 0)],
1402
ann._matching_blocks[(rev2_key, parent_key)])
1404
def test__get_parent_ann_uses_matching_blocks(self):
1405
ann = self.make_annotator()
1406
rev_key = ('rev-id',)
1407
parent_key = ('parent-id',)
1408
parent_ann = [(parent_key,)]*3
1409
block_key = (rev_key, parent_key)
1410
ann._annotations_cache[parent_key] = parent_ann
1411
ann._matching_blocks[block_key] = [(0, 1, 1), (3, 3, 0)]
1412
# We should not try to access any parent_lines content, because we know
1413
# we already have the matching blocks
1414
par_ann, blocks = ann._get_parent_annotations_and_matches(rev_key,
1415
['1\n', '2\n', '3\n'], parent_key)
1416
self.assertEqual(parent_ann, par_ann)
1417
self.assertEqual([(0, 1, 1), (3, 3, 0)], blocks)
1418
self.assertEqual({}, ann._matching_blocks)
1420
def test__process_pending(self):
1421
ann = self.make_annotator()
1422
rev_key = ('rev-id',)
1425
record = ['0,1,1\n', 'new-line\n']
1426
details = ('line-delta', False)
1427
p1_record = ['line1\n', 'line2\n']
1428
ann._num_compression_children[p1_key] = 1
1429
res = ann._expand_record(rev_key, (p1_key,p2_key), p1_key,
1431
self.assertEqual(None, res)
1432
# self.assertTrue(p1_key in ann._pending_deltas)
1433
self.assertEqual({}, ann._pending_annotation)
1434
# Now insert p1, and we should be able to expand the delta
1435
res = ann._expand_record(p1_key, (), None, p1_record,
1436
('fulltext', False))
1437
self.assertEqual(p1_record, res)
1438
ann._annotations_cache[p1_key] = [(p1_key,)]*2
1439
res = ann._process_pending(p1_key)
1440
self.assertEqual([], res)
1441
self.assertFalse(p1_key in ann._pending_deltas)
1442
self.assertTrue(p2_key in ann._pending_annotation)
1443
self.assertEqual({p2_key: [(rev_key, (p1_key, p2_key))]},
1444
ann._pending_annotation)
1445
# Now fill in parent 2, and pending annotation should be satisfied
1446
res = ann._expand_record(p2_key, (), None, [], ('fulltext', False))
1447
ann._annotations_cache[p2_key] = []
1448
res = ann._process_pending(p2_key)
1449
self.assertEqual([rev_key], res)
1450
self.assertEqual({}, ann._pending_annotation)
1451
self.assertEqual({}, ann._pending_deltas)
1453
def test_record_delta_removes_basis(self):
1454
ann = self.make_annotator()
1455
ann._expand_record(('parent-id',), (), None,
1456
['line1\n', 'line2\n'], ('fulltext', False))
1457
ann._num_compression_children['parent-id'] = 2
1459
def test_annotate_special_text(self):
1460
ann = self.make_annotator()
1462
rev1_key = ('rev-1',)
1463
rev2_key = ('rev-2',)
1464
rev3_key = ('rev-3',)
1465
spec_key = ('special:',)
1466
vf.add_lines(rev1_key, [], ['initial content\n'])
1467
vf.add_lines(rev2_key, [rev1_key], ['initial content\n',
1470
vf.add_lines(rev3_key, [rev1_key], ['initial content\n',
1473
spec_text = ('initial content\n'
1477
ann.add_special_text(spec_key, [rev2_key, rev3_key], spec_text)
1478
anns, lines = ann.annotate(spec_key)
1479
self.assertEqual([(rev1_key,),
1480
(rev2_key, rev3_key),
1484
self.assertEqualDiff(spec_text, ''.join(lines))
917
1487
class KnitTests(TestCaseWithTransport):
918
1488
"""Class containing knit test helper routines."""
1205
1798
# change options in the second record
1206
1799
self.assertRaises(errors.KnitCorrupt, index.add_records,
1207
1800
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)]),
1208
(('tip',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1801
(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1209
1802
self.assertEqual([], self.caught_entries)
1804
def make_g_index_missing_compression_parent(self):
1805
graph_index = self.make_g_index('missing_comp', 2,
1806
[(('tip', ), ' 100 78',
1807
([('missing-parent', ), ('ghost', )], [('missing-parent', )]))])
1810
def make_g_index_missing_parent(self):
1811
graph_index = self.make_g_index('missing_parent', 2,
1812
[(('parent', ), ' 100 78', ([], [])),
1813
(('tip', ), ' 100 78',
1814
([('parent', ), ('missing-parent', )], [('parent', )])),
1818
def make_g_index_no_external_refs(self):
1819
graph_index = self.make_g_index('no_external_refs', 2,
1820
[(('rev', ), ' 100 78',
1821
([('parent', ), ('ghost', )], []))])
1824
def test_add_good_unvalidated_index(self):
1825
unvalidated = self.make_g_index_no_external_refs()
1826
combined = CombinedGraphIndex([unvalidated])
1827
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1828
index.scan_unvalidated_index(unvalidated)
1829
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1831
def test_add_missing_compression_parent_unvalidated_index(self):
1832
unvalidated = self.make_g_index_missing_compression_parent()
1833
combined = CombinedGraphIndex([unvalidated])
1834
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1835
index.scan_unvalidated_index(unvalidated)
1836
# This also checks that its only the compression parent that is
1837
# examined, otherwise 'ghost' would also be reported as a missing
1840
frozenset([('missing-parent',)]),
1841
index.get_missing_compression_parents())
1843
def test_add_missing_noncompression_parent_unvalidated_index(self):
1844
unvalidated = self.make_g_index_missing_parent()
1845
combined = CombinedGraphIndex([unvalidated])
1846
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1847
track_external_parent_refs=True)
1848
index.scan_unvalidated_index(unvalidated)
1850
frozenset([('missing-parent',)]), index.get_missing_parents())
1852
def test_track_external_parent_refs(self):
1853
g_index = self.make_g_index('empty', 2, [])
1854
combined = CombinedGraphIndex([g_index])
1855
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1856
add_callback=self.catch_add, track_external_parent_refs=True)
1857
self.caught_entries = []
1859
(('new-key',), 'fulltext,no-eol', (None, 50, 60),
1860
[('parent-1',), ('parent-2',)])])
1862
frozenset([('parent-1',), ('parent-2',)]),
1863
index.get_missing_parents())
1865
def test_add_unvalidated_index_with_present_external_references(self):
1866
index = self.two_graph_index(deltas=True)
1867
# Ugly hack to get at one of the underlying GraphIndex objects that
1868
# two_graph_index built.
1869
unvalidated = index._graph_index._indices[1]
1870
# 'parent' is an external ref of _indices[1] (unvalidated), but is
1871
# present in _indices[0].
1872
index.scan_unvalidated_index(unvalidated)
1873
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1875
def make_new_missing_parent_g_index(self, name):
1876
missing_parent = name + '-missing-parent'
1877
graph_index = self.make_g_index(name, 2,
1878
[((name + 'tip', ), ' 100 78',
1879
([(missing_parent, ), ('ghost', )], [(missing_parent, )]))])
1882
def test_add_mulitiple_unvalidated_indices_with_missing_parents(self):
1883
g_index_1 = self.make_new_missing_parent_g_index('one')
1884
g_index_2 = self.make_new_missing_parent_g_index('two')
1885
combined = CombinedGraphIndex([g_index_1, g_index_2])
1886
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1887
index.scan_unvalidated_index(g_index_1)
1888
index.scan_unvalidated_index(g_index_2)
1890
frozenset([('one-missing-parent',), ('two-missing-parent',)]),
1891
index.get_missing_compression_parents())
1893
def test_add_mulitiple_unvalidated_indices_with_mutual_dependencies(self):
1894
graph_index_a = self.make_g_index('one', 2,
1895
[(('parent-one', ), ' 100 78', ([('non-compression-parent',)], [])),
1896
(('child-of-two', ), ' 100 78',
1897
([('parent-two',)], [('parent-two',)]))])
1898
graph_index_b = self.make_g_index('two', 2,
1899
[(('parent-two', ), ' 100 78', ([('non-compression-parent',)], [])),
1900
(('child-of-one', ), ' 100 78',
1901
([('parent-one',)], [('parent-one',)]))])
1902
combined = CombinedGraphIndex([graph_index_a, graph_index_b])
1903
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1904
index.scan_unvalidated_index(graph_index_a)
1905
index.scan_unvalidated_index(graph_index_b)
1907
frozenset([]), index.get_missing_compression_parents())
1212
1910
class TestNoParentsGraphIndexKnit(KnitTests):
1213
1911
"""Tests for knits using _KnitGraphIndex with no parents."""
1379
2085
self.assertEqual([], self.caught_entries)
2088
class TestKnitVersionedFiles(KnitTests):
2090
def assertGroupKeysForIo(self, exp_groups, keys, non_local_keys,
2091
positions, _min_buffer_size=None):
2092
kvf = self.make_test_knit()
2093
if _min_buffer_size is None:
2094
_min_buffer_size = knit._STREAM_MIN_BUFFER_SIZE
2095
self.assertEqual(exp_groups, kvf._group_keys_for_io(keys,
2096
non_local_keys, positions,
2097
_min_buffer_size=_min_buffer_size))
2099
def assertSplitByPrefix(self, expected_map, expected_prefix_order,
2101
split, prefix_order = KnitVersionedFiles._split_by_prefix(keys)
2102
self.assertEqual(expected_map, split)
2103
self.assertEqual(expected_prefix_order, prefix_order)
2105
def test__group_keys_for_io(self):
2106
ft_detail = ('fulltext', False)
2107
ld_detail = ('line-delta', False)
2115
f_a: (ft_detail, (f_a, 0, 100), None),
2116
f_b: (ld_detail, (f_b, 100, 21), f_a),
2117
f_c: (ld_detail, (f_c, 180, 15), f_b),
2118
g_a: (ft_detail, (g_a, 121, 35), None),
2119
g_b: (ld_detail, (g_b, 156, 12), g_a),
2120
g_c: (ld_detail, (g_c, 195, 13), g_a),
2122
self.assertGroupKeysForIo([([f_a], set())],
2123
[f_a], [], positions)
2124
self.assertGroupKeysForIo([([f_a], set([f_a]))],
2125
[f_a], [f_a], positions)
2126
self.assertGroupKeysForIo([([f_a, f_b], set([]))],
2127
[f_a, f_b], [], positions)
2128
self.assertGroupKeysForIo([([f_a, f_b], set([f_b]))],
2129
[f_a, f_b], [f_b], positions)
2130
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2131
[f_a, g_a, f_b, g_b], [], positions)
2132
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2133
[f_a, g_a, f_b, g_b], [], positions,
2134
_min_buffer_size=150)
2135
self.assertGroupKeysForIo([([f_a, f_b], set()), ([g_a, g_b], set())],
2136
[f_a, g_a, f_b, g_b], [], positions,
2137
_min_buffer_size=100)
2138
self.assertGroupKeysForIo([([f_c], set()), ([g_b], set())],
2139
[f_c, g_b], [], positions,
2140
_min_buffer_size=125)
2141
self.assertGroupKeysForIo([([g_b, f_c], set())],
2142
[g_b, f_c], [], positions,
2143
_min_buffer_size=125)
2145
def test__split_by_prefix(self):
2146
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2147
'g': [('g', 'b'), ('g', 'a')],
2149
[('f', 'a'), ('g', 'b'),
2150
('g', 'a'), ('f', 'b')])
2152
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2153
'g': [('g', 'b'), ('g', 'a')],
2155
[('f', 'a'), ('f', 'b'),
2156
('g', 'b'), ('g', 'a')])
2158
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2159
'g': [('g', 'b'), ('g', 'a')],
2161
[('f', 'a'), ('f', 'b'),
2162
('g', 'b'), ('g', 'a')])
2164
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2165
'g': [('g', 'b'), ('g', 'a')],
2166
'': [('a',), ('b',)]
2168
[('f', 'a'), ('g', 'b'),
2170
('g', 'a'), ('f', 'b')])
1382
2173
class TestStacking(KnitTests):
1384
2175
def get_basis_and_test_knit(self):
1764
2562
multiparent.NewText(['foo\n']),
1765
2563
multiparent.ParentText(1, 0, 2, 1)])],
1767
self.assertEqual(4, len(basis.calls))
2565
self.assertEqual(3, len(basis.calls))
1768
2566
self.assertEqual([
1769
2567
("get_parent_map", set([key_left, key_right])),
1770
2568
("get_parent_map", set([key_left, key_right])),
1771
("get_parent_map", set([key_left, key_right])),
1774
last_call = basis.calls[3]
2571
last_call = basis.calls[-1]
1775
2572
self.assertEqual('get_record_stream', last_call[0])
1776
2573
self.assertEqual(set([key_left, key_right]), set(last_call[1]))
1777
self.assertEqual('unordered', last_call[2])
2574
self.assertEqual('topological', last_call[2])
1778
2575
self.assertEqual(True, last_call[3])
2578
class TestNetworkBehaviour(KnitTests):
2579
"""Tests for getting data out of/into knits over the network."""
2581
def test_include_delta_closure_generates_a_knit_delta_closure(self):
2582
vf = self.make_test_knit(name='test')
2583
# put in three texts, giving ft, delta, delta
2584
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2585
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2586
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2587
# But heuristics could interfere, so check what happened:
2588
self.assertEqual(['knit-ft-gz', 'knit-delta-gz', 'knit-delta-gz'],
2589
[record.storage_kind for record in
2590
vf.get_record_stream([('base',), ('d1',), ('d2',)],
2591
'topological', False)])
2592
# generate a stream of just the deltas include_delta_closure=True,
2593
# serialise to the network, and check that we get a delta closure on the wire.
2594
stream = vf.get_record_stream([('d1',), ('d2',)], 'topological', True)
2595
netb = [record.get_bytes_as(record.storage_kind) for record in stream]
2596
# The first bytes should be a memo from _ContentMapGenerator, and the
2597
# second bytes should be empty (because its a API proxy not something
2598
# for wire serialisation.
2599
self.assertEqual('', netb[1])
2601
kind, line_end = network_bytes_to_kind_and_offset(bytes)
2602
self.assertEqual('knit-delta-closure', kind)
2605
class TestContentMapGenerator(KnitTests):
2606
"""Tests for ContentMapGenerator"""
2608
def test_get_record_stream_gives_records(self):
2609
vf = self.make_test_knit(name='test')
2610
# put in three texts, giving ft, delta, delta
2611
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2612
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2613
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2614
keys = [('d1',), ('d2',)]
2615
generator = _VFContentMapGenerator(vf, keys,
2616
global_map=vf.get_parent_map(keys))
2617
for record in generator.get_record_stream():
2618
if record.key == ('d1',):
2619
self.assertEqual('d1\n', record.get_bytes_as('fulltext'))
2621
self.assertEqual('d2\n', record.get_bytes_as('fulltext'))
2623
def test_get_record_stream_kinds_are_raw(self):
2624
vf = self.make_test_knit(name='test')
2625
# put in three texts, giving ft, delta, delta
2626
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2627
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2628
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2629
keys = [('base',), ('d1',), ('d2',)]
2630
generator = _VFContentMapGenerator(vf, keys,
2631
global_map=vf.get_parent_map(keys))
2632
kinds = {('base',): 'knit-delta-closure',
2633
('d1',): 'knit-delta-closure-ref',
2634
('d2',): 'knit-delta-closure-ref',
2636
for record in generator.get_record_stream():
2637
self.assertEqual(kinds[record.key], record.storage_kind)
added
removed
bzrlib/tests/test_knit.py
