467
359
self.assertEqual(['1234567890'], list(access.get_raw_records(memos)))
469
def test_missing_index_raises_retry(self):
470
memos = self.make_pack_file()
471
transport = self.get_transport()
472
reload_called, reload_func = self.make_reload_func()
473
# Note that the index key has changed from 'foo' to 'bar'
474
access = _DirectPackAccess({'bar':(transport, 'packname')},
475
reload_func=reload_func)
476
e = self.assertListRaises(errors.RetryWithNewPacks,
477
access.get_raw_records, memos)
478
# Because a key was passed in which does not match our index list, we
479
# assume that the listing was already reloaded
480
self.assertTrue(e.reload_occurred)
481
self.assertIsInstance(e.exc_info, tuple)
482
self.assertIs(e.exc_info[0], KeyError)
483
self.assertIsInstance(e.exc_info[1], KeyError)
485
def test_missing_index_raises_key_error_with_no_reload(self):
486
memos = self.make_pack_file()
487
transport = self.get_transport()
488
# Note that the index key has changed from 'foo' to 'bar'
489
access = _DirectPackAccess({'bar':(transport, 'packname')})
490
e = self.assertListRaises(KeyError, access.get_raw_records, memos)
492
def test_missing_file_raises_retry(self):
493
memos = self.make_pack_file()
494
transport = self.get_transport()
495
reload_called, reload_func = self.make_reload_func()
496
# Note that the 'filename' has been changed to 'different-packname'
497
access = _DirectPackAccess({'foo':(transport, 'different-packname')},
498
reload_func=reload_func)
499
e = self.assertListRaises(errors.RetryWithNewPacks,
500
access.get_raw_records, memos)
501
# The file has gone missing, so we assume we need to reload
502
self.assertFalse(e.reload_occurred)
503
self.assertIsInstance(e.exc_info, tuple)
504
self.assertIs(e.exc_info[0], errors.NoSuchFile)
505
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
506
self.assertEqual('different-packname', e.exc_info[1].path)
508
def test_missing_file_raises_no_such_file_with_no_reload(self):
509
memos = self.make_pack_file()
510
transport = self.get_transport()
511
# Note that the 'filename' has been changed to 'different-packname'
512
access = _DirectPackAccess({'foo':(transport, 'different-packname')})
513
e = self.assertListRaises(errors.NoSuchFile,
514
access.get_raw_records, memos)
516
def test_failing_readv_raises_retry(self):
517
memos = self.make_pack_file()
518
transport = self.get_transport()
519
failing_transport = MockReadvFailingTransport(
520
[transport.get_bytes('packname')])
521
reload_called, reload_func = self.make_reload_func()
522
access = _DirectPackAccess({'foo':(failing_transport, 'packname')},
523
reload_func=reload_func)
524
# Asking for a single record will not trigger the Mock failure
525
self.assertEqual(['1234567890'],
526
list(access.get_raw_records(memos[:1])))
527
self.assertEqual(['12345'],
528
list(access.get_raw_records(memos[1:2])))
529
# A multiple offset readv() will fail mid-way through
530
e = self.assertListRaises(errors.RetryWithNewPacks,
531
access.get_raw_records, memos)
532
# The file has gone missing, so we assume we need to reload
533
self.assertFalse(e.reload_occurred)
534
self.assertIsInstance(e.exc_info, tuple)
535
self.assertIs(e.exc_info[0], errors.NoSuchFile)
536
self.assertIsInstance(e.exc_info[1], errors.NoSuchFile)
537
self.assertEqual('packname', e.exc_info[1].path)
539
def test_failing_readv_raises_no_such_file_with_no_reload(self):
540
memos = self.make_pack_file()
541
transport = self.get_transport()
542
failing_transport = MockReadvFailingTransport(
543
[transport.get_bytes('packname')])
544
reload_called, reload_func = self.make_reload_func()
545
access = _DirectPackAccess({'foo':(failing_transport, 'packname')})
546
# Asking for a single record will not trigger the Mock failure
547
self.assertEqual(['1234567890'],
548
list(access.get_raw_records(memos[:1])))
549
self.assertEqual(['12345'],
550
list(access.get_raw_records(memos[1:2])))
551
# A multiple offset readv() will fail mid-way through
552
e = self.assertListRaises(errors.NoSuchFile,
553
access.get_raw_records, memos)
555
def test_reload_or_raise_no_reload(self):
556
access = _DirectPackAccess({}, reload_func=None)
557
retry_exc = self.make_retry_exception()
558
# Without a reload_func, we will just re-raise the original exception
559
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
561
def test_reload_or_raise_reload_changed(self):
562
reload_called, reload_func = self.make_reload_func(return_val=True)
563
access = _DirectPackAccess({}, reload_func=reload_func)
564
retry_exc = self.make_retry_exception()
565
access.reload_or_raise(retry_exc)
566
self.assertEqual([1], reload_called)
567
retry_exc.reload_occurred=True
568
access.reload_or_raise(retry_exc)
569
self.assertEqual([2], reload_called)
571
def test_reload_or_raise_reload_no_change(self):
572
reload_called, reload_func = self.make_reload_func(return_val=False)
573
access = _DirectPackAccess({}, reload_func=reload_func)
574
retry_exc = self.make_retry_exception()
575
# If reload_occurred is False, then we consider it an error to have
576
# reload_func() return False (no changes).
577
self.assertRaises(_TestException, access.reload_or_raise, retry_exc)
578
self.assertEqual([1], reload_called)
579
retry_exc.reload_occurred=True
580
# If reload_occurred is True, then we assume nothing changed because
581
# it had changed earlier, but didn't change again
582
access.reload_or_raise(retry_exc)
583
self.assertEqual([2], reload_called)
585
def test_annotate_retries(self):
586
vf, reload_counter = self.make_vf_for_retrying()
587
# It is a little bit bogus to annotate the Revision VF, but it works,
588
# as we have ancestry stored there
590
reload_lines = vf.annotate(key)
591
self.assertEqual([1, 1, 0], reload_counter)
592
plain_lines = vf.annotate(key)
593
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
594
if reload_lines != plain_lines:
595
self.fail('Annotation was not identical with reloading.')
596
# Now delete the packs-in-use, which should trigger another reload, but
597
# this time we just raise an exception because we can't recover
598
for trans, name in vf._access._indices.itervalues():
600
self.assertRaises(errors.NoSuchFile, vf.annotate, key)
601
self.assertEqual([2, 1, 1], reload_counter)
603
def test__get_record_map_retries(self):
604
vf, reload_counter = self.make_vf_for_retrying()
605
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
606
records = vf._get_record_map(keys)
607
self.assertEqual(keys, sorted(records.keys()))
608
self.assertEqual([1, 1, 0], reload_counter)
609
# Now delete the packs-in-use, which should trigger another reload, but
610
# this time we just raise an exception because we can't recover
611
for trans, name in vf._access._indices.itervalues():
613
self.assertRaises(errors.NoSuchFile, vf._get_record_map, keys)
614
self.assertEqual([2, 1, 1], reload_counter)
616
def test_get_record_stream_retries(self):
617
vf, reload_counter = self.make_vf_for_retrying()
618
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
619
record_stream = vf.get_record_stream(keys, 'topological', False)
620
record = record_stream.next()
621
self.assertEqual(('rev-1',), record.key)
622
self.assertEqual([0, 0, 0], reload_counter)
623
record = record_stream.next()
624
self.assertEqual(('rev-2',), record.key)
625
self.assertEqual([1, 1, 0], reload_counter)
626
record = record_stream.next()
627
self.assertEqual(('rev-3',), record.key)
628
self.assertEqual([1, 1, 0], reload_counter)
629
# Now delete all pack files, and see that we raise the right error
630
for trans, name in vf._access._indices.itervalues():
632
self.assertListRaises(errors.NoSuchFile,
633
vf.get_record_stream, keys, 'topological', False)
635
def test_iter_lines_added_or_present_in_keys_retries(self):
636
vf, reload_counter = self.make_vf_for_retrying()
637
keys = [('rev-1',), ('rev-2',), ('rev-3',)]
638
# Unfortunately, iter_lines_added_or_present_in_keys iterates the
639
# result in random order (determined by the iteration order from a
640
# set()), so we don't have any solid way to trigger whether data is
641
# read before or after. However we tried to delete the middle node to
642
# exercise the code well.
643
# What we care about is that all lines are always yielded, but not
646
reload_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
647
self.assertEqual([1, 1, 0], reload_counter)
648
# Now do it again, to make sure the result is equivalent
649
plain_lines = sorted(vf.iter_lines_added_or_present_in_keys(keys))
650
self.assertEqual([1, 1, 0], reload_counter) # No extra reloading
651
self.assertEqual(plain_lines, reload_lines)
652
self.assertEqual(21, len(plain_lines))
653
# Now delete all pack files, and see that we raise the right error
654
for trans, name in vf._access._indices.itervalues():
656
self.assertListRaises(errors.NoSuchFile,
657
vf.iter_lines_added_or_present_in_keys, keys)
658
self.assertEqual([2, 1, 1], reload_counter)
660
def test_get_record_stream_yields_disk_sorted_order(self):
661
# if we get 'unordered' pick a semi-optimal order for reading. The
662
# order should be grouped by pack file, and then by position in file
663
repo = self.make_repository('test', format='pack-0.92')
665
self.addCleanup(repo.unlock)
666
repo.start_write_group()
668
vf.add_lines(('f-id', 'rev-5'), [('f-id', 'rev-4')], ['lines\n'])
669
vf.add_lines(('f-id', 'rev-1'), [], ['lines\n'])
670
vf.add_lines(('f-id', 'rev-2'), [('f-id', 'rev-1')], ['lines\n'])
671
repo.commit_write_group()
672
# We inserted them as rev-5, rev-1, rev-2, we should get them back in
674
stream = vf.get_record_stream([('f-id', 'rev-1'), ('f-id', 'rev-5'),
675
('f-id', 'rev-2')], 'unordered', False)
676
keys = [r.key for r in stream]
677
self.assertEqual([('f-id', 'rev-5'), ('f-id', 'rev-1'),
678
('f-id', 'rev-2')], keys)
679
repo.start_write_group()
680
vf.add_lines(('f-id', 'rev-4'), [('f-id', 'rev-3')], ['lines\n'])
681
vf.add_lines(('f-id', 'rev-3'), [('f-id', 'rev-2')], ['lines\n'])
682
vf.add_lines(('f-id', 'rev-6'), [('f-id', 'rev-5')], ['lines\n'])
683
repo.commit_write_group()
684
# Request in random order, to make sure the output order isn't based on
686
request_keys = set(('f-id', 'rev-%d' % i) for i in range(1, 7))
687
stream = vf.get_record_stream(request_keys, 'unordered', False)
688
keys = [r.key for r in stream]
689
# We want to get the keys back in disk order, but it doesn't matter
690
# which pack we read from first. So this can come back in 2 orders
691
alt1 = [('f-id', 'rev-%d' % i) for i in [4, 3, 6, 5, 1, 2]]
692
alt2 = [('f-id', 'rev-%d' % i) for i in [5, 1, 2, 4, 3, 6]]
693
if keys != alt1 and keys != alt2:
694
self.fail('Returned key order did not match either expected order.'
695
' expected %s or %s, not %s'
696
% (alt1, alt2, keys))
699
362
class LowLevelKnitDataTests(TestCase):
1295
895
class LowLevelKnitIndexTests_c(LowLevelKnitIndexTests):
1297
_test_needs_features = [compiled_knit_feature]
897
_test_needs_features = [CompiledKnitFeature]
1299
899
def get_knit_index(self, transport, name, mode):
1300
900
mapper = ConstantMapper(name)
1301
from bzrlib._knit_load_data_pyx import _load_data_c
1302
self.overrideAttr(knit, '_load_data', _load_data_c)
901
orig = knit._load_data
903
knit._load_data = orig
904
self.addCleanup(reset)
905
from bzrlib._knit_load_data_c import _load_data_c
906
knit._load_data = _load_data_c
1303
907
allow_writes = lambda: mode == 'w'
1304
return _KndxIndex(transport, mapper, lambda:None,
1305
allow_writes, lambda:True)
1308
class Test_KnitAnnotator(TestCaseWithMemoryTransport):
1310
def make_annotator(self):
1311
factory = knit.make_pack_factory(True, True, 1)
1312
vf = factory(self.get_transport())
1313
return knit._KnitAnnotator(vf)
1315
def test__expand_fulltext(self):
1316
ann = self.make_annotator()
1317
rev_key = ('rev-id',)
1318
ann._num_compression_children[rev_key] = 1
1319
res = ann._expand_record(rev_key, (('parent-id',),), None,
1320
['line1\n', 'line2\n'], ('fulltext', True))
1321
# The content object and text lines should be cached appropriately
1322
self.assertEqual(['line1\n', 'line2'], res)
1323
content_obj = ann._content_objects[rev_key]
1324
self.assertEqual(['line1\n', 'line2\n'], content_obj._lines)
1325
self.assertEqual(res, content_obj.text())
1326
self.assertEqual(res, ann._text_cache[rev_key])
1328
def test__expand_delta_comp_parent_not_available(self):
1329
# Parent isn't available yet, so we return nothing, but queue up this
1330
# node for later processing
1331
ann = self.make_annotator()
1332
rev_key = ('rev-id',)
1333
parent_key = ('parent-id',)
1334
record = ['0,1,1\n', 'new-line\n']
1335
details = ('line-delta', False)
1336
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1338
self.assertEqual(None, res)
1339
self.assertTrue(parent_key in ann._pending_deltas)
1340
pending = ann._pending_deltas[parent_key]
1341
self.assertEqual(1, len(pending))
1342
self.assertEqual((rev_key, (parent_key,), record, details), pending[0])
1344
def test__expand_record_tracks_num_children(self):
1345
ann = self.make_annotator()
1346
rev_key = ('rev-id',)
1347
rev2_key = ('rev2-id',)
1348
parent_key = ('parent-id',)
1349
record = ['0,1,1\n', 'new-line\n']
1350
details = ('line-delta', False)
1351
ann._num_compression_children[parent_key] = 2
1352
ann._expand_record(parent_key, (), None, ['line1\n', 'line2\n'],
1353
('fulltext', False))
1354
res = ann._expand_record(rev_key, (parent_key,), parent_key,
1356
self.assertEqual({parent_key: 1}, ann._num_compression_children)
1357
# Expanding the second child should remove the content object, and the
1358
# num_compression_children entry
1359
res = ann._expand_record(rev2_key, (parent_key,), parent_key,
1361
self.assertFalse(parent_key in ann._content_objects)
1362
self.assertEqual({}, ann._num_compression_children)
1363
# We should not cache the content_objects for rev2 and rev, because
1364
# they do not have compression children of their own.
1365
self.assertEqual({}, ann._content_objects)
1367
def test__expand_delta_records_blocks(self):
1368
ann = self.make_annotator()
1369
rev_key = ('rev-id',)
1370
parent_key = ('parent-id',)
1371
record = ['0,1,1\n', 'new-line\n']
1372
details = ('line-delta', True)
1373
ann._num_compression_children[parent_key] = 2
1374
ann._expand_record(parent_key, (), None,
1375
['line1\n', 'line2\n', 'line3\n'],
1376
('fulltext', False))
1377
ann._expand_record(rev_key, (parent_key,), parent_key, record, details)
1378
self.assertEqual({(rev_key, parent_key): [(1, 1, 1), (3, 3, 0)]},
1379
ann._matching_blocks)
1380
rev2_key = ('rev2-id',)
1381
record = ['0,1,1\n', 'new-line\n']
1382
details = ('line-delta', False)
1383
ann._expand_record(rev2_key, (parent_key,), parent_key, record, details)
1384
self.assertEqual([(1, 1, 2), (3, 3, 0)],
1385
ann._matching_blocks[(rev2_key, parent_key)])
1387
def test__get_parent_ann_uses_matching_blocks(self):
1388
ann = self.make_annotator()
1389
rev_key = ('rev-id',)
1390
parent_key = ('parent-id',)
1391
parent_ann = [(parent_key,)]*3
1392
block_key = (rev_key, parent_key)
1393
ann._annotations_cache[parent_key] = parent_ann
1394
ann._matching_blocks[block_key] = [(0, 1, 1), (3, 3, 0)]
1395
# We should not try to access any parent_lines content, because we know
1396
# we already have the matching blocks
1397
par_ann, blocks = ann._get_parent_annotations_and_matches(rev_key,
1398
['1\n', '2\n', '3\n'], parent_key)
1399
self.assertEqual(parent_ann, par_ann)
1400
self.assertEqual([(0, 1, 1), (3, 3, 0)], blocks)
1401
self.assertEqual({}, ann._matching_blocks)
1403
def test__process_pending(self):
1404
ann = self.make_annotator()
1405
rev_key = ('rev-id',)
1408
record = ['0,1,1\n', 'new-line\n']
1409
details = ('line-delta', False)
1410
p1_record = ['line1\n', 'line2\n']
1411
ann._num_compression_children[p1_key] = 1
1412
res = ann._expand_record(rev_key, (p1_key,p2_key), p1_key,
1414
self.assertEqual(None, res)
1415
# self.assertTrue(p1_key in ann._pending_deltas)
1416
self.assertEqual({}, ann._pending_annotation)
1417
# Now insert p1, and we should be able to expand the delta
1418
res = ann._expand_record(p1_key, (), None, p1_record,
1419
('fulltext', False))
1420
self.assertEqual(p1_record, res)
1421
ann._annotations_cache[p1_key] = [(p1_key,)]*2
1422
res = ann._process_pending(p1_key)
1423
self.assertEqual([], res)
1424
self.assertFalse(p1_key in ann._pending_deltas)
1425
self.assertTrue(p2_key in ann._pending_annotation)
1426
self.assertEqual({p2_key: [(rev_key, (p1_key, p2_key))]},
1427
ann._pending_annotation)
1428
# Now fill in parent 2, and pending annotation should be satisfied
1429
res = ann._expand_record(p2_key, (), None, [], ('fulltext', False))
1430
ann._annotations_cache[p2_key] = []
1431
res = ann._process_pending(p2_key)
1432
self.assertEqual([rev_key], res)
1433
self.assertEqual({}, ann._pending_annotation)
1434
self.assertEqual({}, ann._pending_deltas)
1436
def test_record_delta_removes_basis(self):
1437
ann = self.make_annotator()
1438
ann._expand_record(('parent-id',), (), None,
1439
['line1\n', 'line2\n'], ('fulltext', False))
1440
ann._num_compression_children['parent-id'] = 2
1442
def test_annotate_special_text(self):
1443
ann = self.make_annotator()
1445
rev1_key = ('rev-1',)
1446
rev2_key = ('rev-2',)
1447
rev3_key = ('rev-3',)
1448
spec_key = ('special:',)
1449
vf.add_lines(rev1_key, [], ['initial content\n'])
1450
vf.add_lines(rev2_key, [rev1_key], ['initial content\n',
1453
vf.add_lines(rev3_key, [rev1_key], ['initial content\n',
1456
spec_text = ('initial content\n'
1460
ann.add_special_text(spec_key, [rev2_key, rev3_key], spec_text)
1461
anns, lines = ann.annotate(spec_key)
1462
self.assertEqual([(rev1_key,),
1463
(rev2_key, rev3_key),
1467
self.assertEqualDiff(spec_text, ''.join(lines))
908
return _KndxIndex(transport, mapper, lambda:None, allow_writes, lambda:True)
1470
911
class KnitTests(TestCaseWithTransport):
1781
1199
# change options in the second record
1782
1200
self.assertRaises(errors.KnitCorrupt, index.add_records,
1783
1201
[(('tip',), 'fulltext,no-eol', (None, 0, 100), [('parent',)]),
1784
(('tip',), 'line-delta', (None, 0, 100), [('parent',)])])
1202
(('tip',), 'no-eol,line-delta', (None, 0, 100), [('parent',)])])
1785
1203
self.assertEqual([], self.caught_entries)
1787
def make_g_index_missing_compression_parent(self):
1788
graph_index = self.make_g_index('missing_comp', 2,
1789
[(('tip', ), ' 100 78',
1790
([('missing-parent', ), ('ghost', )], [('missing-parent', )]))])
1793
def make_g_index_missing_parent(self):
1794
graph_index = self.make_g_index('missing_parent', 2,
1795
[(('parent', ), ' 100 78', ([], [])),
1796
(('tip', ), ' 100 78',
1797
([('parent', ), ('missing-parent', )], [('parent', )])),
1801
def make_g_index_no_external_refs(self):
1802
graph_index = self.make_g_index('no_external_refs', 2,
1803
[(('rev', ), ' 100 78',
1804
([('parent', ), ('ghost', )], []))])
1807
def test_add_good_unvalidated_index(self):
1808
unvalidated = self.make_g_index_no_external_refs()
1809
combined = CombinedGraphIndex([unvalidated])
1810
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1811
index.scan_unvalidated_index(unvalidated)
1812
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1814
def test_add_missing_compression_parent_unvalidated_index(self):
1815
unvalidated = self.make_g_index_missing_compression_parent()
1816
combined = CombinedGraphIndex([unvalidated])
1817
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1818
index.scan_unvalidated_index(unvalidated)
1819
# This also checks that its only the compression parent that is
1820
# examined, otherwise 'ghost' would also be reported as a missing
1823
frozenset([('missing-parent',)]),
1824
index.get_missing_compression_parents())
1826
def test_add_missing_noncompression_parent_unvalidated_index(self):
1827
unvalidated = self.make_g_index_missing_parent()
1828
combined = CombinedGraphIndex([unvalidated])
1829
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1830
track_external_parent_refs=True)
1831
index.scan_unvalidated_index(unvalidated)
1833
frozenset([('missing-parent',)]), index.get_missing_parents())
1835
def test_track_external_parent_refs(self):
1836
g_index = self.make_g_index('empty', 2, [])
1837
combined = CombinedGraphIndex([g_index])
1838
index = _KnitGraphIndex(combined, lambda: True, deltas=True,
1839
add_callback=self.catch_add, track_external_parent_refs=True)
1840
self.caught_entries = []
1842
(('new-key',), 'fulltext,no-eol', (None, 50, 60),
1843
[('parent-1',), ('parent-2',)])])
1845
frozenset([('parent-1',), ('parent-2',)]),
1846
index.get_missing_parents())
1848
def test_add_unvalidated_index_with_present_external_references(self):
1849
index = self.two_graph_index(deltas=True)
1850
# Ugly hack to get at one of the underlying GraphIndex objects that
1851
# two_graph_index built.
1852
unvalidated = index._graph_index._indices[1]
1853
# 'parent' is an external ref of _indices[1] (unvalidated), but is
1854
# present in _indices[0].
1855
index.scan_unvalidated_index(unvalidated)
1856
self.assertEqual(frozenset(), index.get_missing_compression_parents())
1858
def make_new_missing_parent_g_index(self, name):
1859
missing_parent = name + '-missing-parent'
1860
graph_index = self.make_g_index(name, 2,
1861
[((name + 'tip', ), ' 100 78',
1862
([(missing_parent, ), ('ghost', )], [(missing_parent, )]))])
1865
def test_add_mulitiple_unvalidated_indices_with_missing_parents(self):
1866
g_index_1 = self.make_new_missing_parent_g_index('one')
1867
g_index_2 = self.make_new_missing_parent_g_index('two')
1868
combined = CombinedGraphIndex([g_index_1, g_index_2])
1869
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1870
index.scan_unvalidated_index(g_index_1)
1871
index.scan_unvalidated_index(g_index_2)
1873
frozenset([('one-missing-parent',), ('two-missing-parent',)]),
1874
index.get_missing_compression_parents())
1876
def test_add_mulitiple_unvalidated_indices_with_mutual_dependencies(self):
1877
graph_index_a = self.make_g_index('one', 2,
1878
[(('parent-one', ), ' 100 78', ([('non-compression-parent',)], [])),
1879
(('child-of-two', ), ' 100 78',
1880
([('parent-two',)], [('parent-two',)]))])
1881
graph_index_b = self.make_g_index('two', 2,
1882
[(('parent-two', ), ' 100 78', ([('non-compression-parent',)], [])),
1883
(('child-of-one', ), ' 100 78',
1884
([('parent-one',)], [('parent-one',)]))])
1885
combined = CombinedGraphIndex([graph_index_a, graph_index_b])
1886
index = _KnitGraphIndex(combined, lambda: True, deltas=True)
1887
index.scan_unvalidated_index(graph_index_a)
1888
index.scan_unvalidated_index(graph_index_b)
1890
frozenset([]), index.get_missing_compression_parents())
1893
1206
class TestNoParentsGraphIndexKnit(KnitTests):
1894
1207
"""Tests for knits using _KnitGraphIndex with no parents."""
2066
1371
[(('tip',), 'fulltext,no-eol', (None, 0, 100), []),
2067
1372
(('tip',), 'no-eol,line-delta', (None, 0, 100), [])])
2068
1373
self.assertEqual([], self.caught_entries)
2071
class TestKnitVersionedFiles(KnitTests):
2073
def assertGroupKeysForIo(self, exp_groups, keys, non_local_keys,
2074
positions, _min_buffer_size=None):
2075
kvf = self.make_test_knit()
2076
if _min_buffer_size is None:
2077
_min_buffer_size = knit._STREAM_MIN_BUFFER_SIZE
2078
self.assertEqual(exp_groups, kvf._group_keys_for_io(keys,
2079
non_local_keys, positions,
2080
_min_buffer_size=_min_buffer_size))
2082
def assertSplitByPrefix(self, expected_map, expected_prefix_order,
2084
split, prefix_order = KnitVersionedFiles._split_by_prefix(keys)
2085
self.assertEqual(expected_map, split)
2086
self.assertEqual(expected_prefix_order, prefix_order)
2088
def test__group_keys_for_io(self):
2089
ft_detail = ('fulltext', False)
2090
ld_detail = ('line-delta', False)
2098
f_a: (ft_detail, (f_a, 0, 100), None),
2099
f_b: (ld_detail, (f_b, 100, 21), f_a),
2100
f_c: (ld_detail, (f_c, 180, 15), f_b),
2101
g_a: (ft_detail, (g_a, 121, 35), None),
2102
g_b: (ld_detail, (g_b, 156, 12), g_a),
2103
g_c: (ld_detail, (g_c, 195, 13), g_a),
2105
self.assertGroupKeysForIo([([f_a], set())],
2106
[f_a], [], positions)
2107
self.assertGroupKeysForIo([([f_a], set([f_a]))],
2108
[f_a], [f_a], positions)
2109
self.assertGroupKeysForIo([([f_a, f_b], set([]))],
2110
[f_a, f_b], [], positions)
2111
self.assertGroupKeysForIo([([f_a, f_b], set([f_b]))],
2112
[f_a, f_b], [f_b], positions)
2113
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2114
[f_a, g_a, f_b, g_b], [], positions)
2115
self.assertGroupKeysForIo([([f_a, f_b, g_a, g_b], set())],
2116
[f_a, g_a, f_b, g_b], [], positions,
2117
_min_buffer_size=150)
2118
self.assertGroupKeysForIo([([f_a, f_b], set()), ([g_a, g_b], set())],
2119
[f_a, g_a, f_b, g_b], [], positions,
2120
_min_buffer_size=100)
2121
self.assertGroupKeysForIo([([f_c], set()), ([g_b], set())],
2122
[f_c, g_b], [], positions,
2123
_min_buffer_size=125)
2124
self.assertGroupKeysForIo([([g_b, f_c], set())],
2125
[g_b, f_c], [], positions,
2126
_min_buffer_size=125)
2128
def test__split_by_prefix(self):
2129
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2130
'g': [('g', 'b'), ('g', 'a')],
2132
[('f', 'a'), ('g', 'b'),
2133
('g', 'a'), ('f', 'b')])
2135
self.assertSplitByPrefix({'f': [('f', 'a'), ('f', 'b')],
2136
'g': [('g', 'b'), ('g', 'a')],
2138
[('f', 'a'), ('f', 'b'),
2139
('g', 'b'), ('g', 'a')])
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')],
2149
'': [('a',), ('b',)]
2151
[('f', 'a'), ('g', 'b'),
2153
('g', 'a'), ('f', 'b')])
2156
class TestStacking(KnitTests):
2158
def get_basis_and_test_knit(self):
2159
basis = self.make_test_knit(name='basis')
2160
basis = RecordingVersionedFilesDecorator(basis)
2161
test = self.make_test_knit(name='test')
2162
test.add_fallback_versioned_files(basis)
2165
def test_add_fallback_versioned_files(self):
2166
basis = self.make_test_knit(name='basis')
2167
test = self.make_test_knit(name='test')
2168
# It must not error; other tests test that the fallback is referred to
2169
# when accessing data.
2170
test.add_fallback_versioned_files(basis)
2172
def test_add_lines(self):
2173
# lines added to the test are not added to the basis
2174
basis, test = self.get_basis_and_test_knit()
2176
key_basis = ('bar',)
2177
key_cross_border = ('quux',)
2178
key_delta = ('zaphod',)
2179
test.add_lines(key, (), ['foo\n'])
2180
self.assertEqual({}, basis.get_parent_map([key]))
2181
# lines added to the test that reference across the stack do a
2183
basis.add_lines(key_basis, (), ['foo\n'])
2185
test.add_lines(key_cross_border, (key_basis,), ['foo\n'])
2186
self.assertEqual('fulltext', test._index.get_method(key_cross_border))
2187
# we don't even need to look at the basis to see that this should be
2188
# stored as a fulltext
2189
self.assertEqual([], basis.calls)
2190
# Subsequent adds do delta.
2192
test.add_lines(key_delta, (key_cross_border,), ['foo\n'])
2193
self.assertEqual('line-delta', test._index.get_method(key_delta))
2194
self.assertEqual([], basis.calls)
2196
def test_annotate(self):
2197
# annotations from the test knit are answered without asking the basis
2198
basis, test = self.get_basis_and_test_knit()
2200
key_basis = ('bar',)
2201
key_missing = ('missing',)
2202
test.add_lines(key, (), ['foo\n'])
2203
details = test.annotate(key)
2204
self.assertEqual([(key, 'foo\n')], details)
2205
self.assertEqual([], basis.calls)
2206
# But texts that are not in the test knit are looked for in the basis
2208
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2210
details = test.annotate(key_basis)
2211
self.assertEqual([(key_basis, 'foo\n'), (key_basis, 'bar\n')], details)
2212
# Not optimised to date:
2213
# self.assertEqual([("annotate", key_basis)], basis.calls)
2214
self.assertEqual([('get_parent_map', set([key_basis])),
2215
('get_parent_map', set([key_basis])),
2216
('get_record_stream', [key_basis], 'topological', True)],
2219
def test_check(self):
2220
# At the moment checking a stacked knit does implicitly check the
2222
basis, test = self.get_basis_and_test_knit()
2225
def test_get_parent_map(self):
2226
# parents in the test knit are answered without asking the basis
2227
basis, test = self.get_basis_and_test_knit()
2229
key_basis = ('bar',)
2230
key_missing = ('missing',)
2231
test.add_lines(key, (), [])
2232
parent_map = test.get_parent_map([key])
2233
self.assertEqual({key: ()}, parent_map)
2234
self.assertEqual([], basis.calls)
2235
# But parents that are not in the test knit are looked for in the basis
2236
basis.add_lines(key_basis, (), [])
2238
parent_map = test.get_parent_map([key, key_basis, key_missing])
2239
self.assertEqual({key: (),
2240
key_basis: ()}, parent_map)
2241
self.assertEqual([("get_parent_map", set([key_basis, key_missing]))],
2244
def test_get_record_stream_unordered_fulltexts(self):
2245
# records from the test knit are answered without asking the basis:
2246
basis, test = self.get_basis_and_test_knit()
2248
key_basis = ('bar',)
2249
key_missing = ('missing',)
2250
test.add_lines(key, (), ['foo\n'])
2251
records = list(test.get_record_stream([key], 'unordered', True))
2252
self.assertEqual(1, len(records))
2253
self.assertEqual([], basis.calls)
2254
# Missing (from test knit) objects are retrieved from the basis:
2255
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2257
records = list(test.get_record_stream([key_basis, key_missing],
2259
self.assertEqual(2, len(records))
2260
calls = list(basis.calls)
2261
for record in records:
2262
self.assertSubset([record.key], (key_basis, key_missing))
2263
if record.key == key_missing:
2264
self.assertIsInstance(record, AbsentContentFactory)
2266
reference = list(basis.get_record_stream([key_basis],
2267
'unordered', True))[0]
2268
self.assertEqual(reference.key, record.key)
2269
self.assertEqual(reference.sha1, record.sha1)
2270
self.assertEqual(reference.storage_kind, record.storage_kind)
2271
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2272
record.get_bytes_as(record.storage_kind))
2273
self.assertEqual(reference.get_bytes_as('fulltext'),
2274
record.get_bytes_as('fulltext'))
2275
# It's not strictly minimal, but it seems reasonable for now for it to
2276
# ask which fallbacks have which parents.
2278
("get_parent_map", set([key_basis, key_missing])),
2279
("get_record_stream", [key_basis], 'unordered', True)],
2282
def test_get_record_stream_ordered_fulltexts(self):
2283
# ordering is preserved down into the fallback store.
2284
basis, test = self.get_basis_and_test_knit()
2286
key_basis = ('bar',)
2287
key_basis_2 = ('quux',)
2288
key_missing = ('missing',)
2289
test.add_lines(key, (key_basis,), ['foo\n'])
2290
# Missing (from test knit) objects are retrieved from the basis:
2291
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2292
basis.add_lines(key_basis_2, (), ['quux\n'])
2294
# ask for in non-topological order
2295
records = list(test.get_record_stream(
2296
[key, key_basis, key_missing, key_basis_2], 'topological', True))
2297
self.assertEqual(4, len(records))
2299
for record in records:
2300
self.assertSubset([record.key],
2301
(key_basis, key_missing, key_basis_2, key))
2302
if record.key == key_missing:
2303
self.assertIsInstance(record, AbsentContentFactory)
2305
results.append((record.key, record.sha1, record.storage_kind,
2306
record.get_bytes_as('fulltext')))
2307
calls = list(basis.calls)
2308
order = [record[0] for record in results]
2309
self.assertEqual([key_basis_2, key_basis, key], order)
2310
for result in results:
2311
if result[0] == key:
2315
record = source.get_record_stream([result[0]], 'unordered',
2317
self.assertEqual(record.key, result[0])
2318
self.assertEqual(record.sha1, result[1])
2319
# We used to check that the storage kind matched, but actually it
2320
# depends on whether it was sourced from the basis, or in a single
2321
# group, because asking for full texts returns proxy objects to a
2322
# _ContentMapGenerator object; so checking the kind is unneeded.
2323
self.assertEqual(record.get_bytes_as('fulltext'), result[3])
2324
# It's not strictly minimal, but it seems reasonable for now for it to
2325
# ask which fallbacks have which parents.
2327
("get_parent_map", set([key_basis, key_basis_2, key_missing])),
2328
# topological is requested from the fallback, because that is what
2329
# was requested at the top level.
2330
("get_record_stream", [key_basis_2, key_basis], 'topological', True)],
2333
def test_get_record_stream_unordered_deltas(self):
2334
# records from the test knit are answered without asking the basis:
2335
basis, test = self.get_basis_and_test_knit()
2337
key_basis = ('bar',)
2338
key_missing = ('missing',)
2339
test.add_lines(key, (), ['foo\n'])
2340
records = list(test.get_record_stream([key], 'unordered', False))
2341
self.assertEqual(1, len(records))
2342
self.assertEqual([], basis.calls)
2343
# Missing (from test knit) objects are retrieved from the basis:
2344
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2346
records = list(test.get_record_stream([key_basis, key_missing],
2347
'unordered', False))
2348
self.assertEqual(2, len(records))
2349
calls = list(basis.calls)
2350
for record in records:
2351
self.assertSubset([record.key], (key_basis, key_missing))
2352
if record.key == key_missing:
2353
self.assertIsInstance(record, AbsentContentFactory)
2355
reference = list(basis.get_record_stream([key_basis],
2356
'unordered', False))[0]
2357
self.assertEqual(reference.key, record.key)
2358
self.assertEqual(reference.sha1, record.sha1)
2359
self.assertEqual(reference.storage_kind, record.storage_kind)
2360
self.assertEqual(reference.get_bytes_as(reference.storage_kind),
2361
record.get_bytes_as(record.storage_kind))
2362
# It's not strictly minimal, but it seems reasonable for now for it to
2363
# ask which fallbacks have which parents.
2365
("get_parent_map", set([key_basis, key_missing])),
2366
("get_record_stream", [key_basis], 'unordered', False)],
2369
def test_get_record_stream_ordered_deltas(self):
2370
# ordering is preserved down into the fallback store.
2371
basis, test = self.get_basis_and_test_knit()
2373
key_basis = ('bar',)
2374
key_basis_2 = ('quux',)
2375
key_missing = ('missing',)
2376
test.add_lines(key, (key_basis,), ['foo\n'])
2377
# Missing (from test knit) objects are retrieved from the basis:
2378
basis.add_lines(key_basis, (key_basis_2,), ['foo\n', 'bar\n'])
2379
basis.add_lines(key_basis_2, (), ['quux\n'])
2381
# ask for in non-topological order
2382
records = list(test.get_record_stream(
2383
[key, key_basis, key_missing, key_basis_2], 'topological', False))
2384
self.assertEqual(4, len(records))
2386
for record in records:
2387
self.assertSubset([record.key],
2388
(key_basis, key_missing, key_basis_2, key))
2389
if record.key == key_missing:
2390
self.assertIsInstance(record, AbsentContentFactory)
2392
results.append((record.key, record.sha1, record.storage_kind,
2393
record.get_bytes_as(record.storage_kind)))
2394
calls = list(basis.calls)
2395
order = [record[0] for record in results]
2396
self.assertEqual([key_basis_2, key_basis, key], order)
2397
for result in results:
2398
if result[0] == key:
2402
record = source.get_record_stream([result[0]], 'unordered',
2404
self.assertEqual(record.key, result[0])
2405
self.assertEqual(record.sha1, result[1])
2406
self.assertEqual(record.storage_kind, result[2])
2407
self.assertEqual(record.get_bytes_as(record.storage_kind), result[3])
2408
# It's not strictly minimal, but it seems reasonable for now for it to
2409
# ask which fallbacks have which parents.
2411
("get_parent_map", set([key_basis, key_basis_2, key_missing])),
2412
("get_record_stream", [key_basis_2, key_basis], 'topological', False)],
2415
def test_get_sha1s(self):
2416
# sha1's in the test knit are answered without asking the basis
2417
basis, test = self.get_basis_and_test_knit()
2419
key_basis = ('bar',)
2420
key_missing = ('missing',)
2421
test.add_lines(key, (), ['foo\n'])
2422
key_sha1sum = osutils.sha('foo\n').hexdigest()
2423
sha1s = test.get_sha1s([key])
2424
self.assertEqual({key: key_sha1sum}, sha1s)
2425
self.assertEqual([], basis.calls)
2426
# But texts that are not in the test knit are looked for in the basis
2427
# directly (rather than via text reconstruction) so that remote servers
2428
# etc don't have to answer with full content.
2429
basis.add_lines(key_basis, (), ['foo\n', 'bar\n'])
2430
basis_sha1sum = osutils.sha('foo\nbar\n').hexdigest()
2432
sha1s = test.get_sha1s([key, key_missing, key_basis])
2433
self.assertEqual({key: key_sha1sum,
2434
key_basis: basis_sha1sum}, sha1s)
2435
self.assertEqual([("get_sha1s", set([key_basis, key_missing]))],
2438
def test_insert_record_stream(self):
2439
# records are inserted as normal; insert_record_stream builds on
2440
# add_lines, so a smoke test should be all that's needed:
2442
key_basis = ('bar',)
2443
key_delta = ('zaphod',)
2444
basis, test = self.get_basis_and_test_knit()
2445
source = self.make_test_knit(name='source')
2446
basis.add_lines(key_basis, (), ['foo\n'])
2448
source.add_lines(key_basis, (), ['foo\n'])
2449
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2450
stream = source.get_record_stream([key_delta], 'unordered', False)
2451
test.insert_record_stream(stream)
2452
# XXX: this does somewhat too many calls in making sure of whether it
2453
# has to recreate the full text.
2454
self.assertEqual([("get_parent_map", set([key_basis])),
2455
('get_parent_map', set([key_basis])),
2456
('get_record_stream', [key_basis], 'unordered', True)],
2458
self.assertEqual({key_delta:(key_basis,)},
2459
test.get_parent_map([key_delta]))
2460
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2461
'unordered', True).next().get_bytes_as('fulltext'))
2463
def test_iter_lines_added_or_present_in_keys(self):
2464
# Lines from the basis are returned, and lines for a given key are only
2468
# all sources are asked for keys:
2469
basis, test = self.get_basis_and_test_knit()
2470
basis.add_lines(key1, (), ["foo"])
2472
lines = list(test.iter_lines_added_or_present_in_keys([key1]))
2473
self.assertEqual([("foo\n", key1)], lines)
2474
self.assertEqual([("iter_lines_added_or_present_in_keys", set([key1]))],
2476
# keys in both are not duplicated:
2477
test.add_lines(key2, (), ["bar\n"])
2478
basis.add_lines(key2, (), ["bar\n"])
2480
lines = list(test.iter_lines_added_or_present_in_keys([key2]))
2481
self.assertEqual([("bar\n", key2)], lines)
2482
self.assertEqual([], basis.calls)
2484
def test_keys(self):
2487
# all sources are asked for keys:
2488
basis, test = self.get_basis_and_test_knit()
2490
self.assertEqual(set(), set(keys))
2491
self.assertEqual([("keys",)], basis.calls)
2492
# keys from a basis are returned:
2493
basis.add_lines(key1, (), [])
2496
self.assertEqual(set([key1]), set(keys))
2497
self.assertEqual([("keys",)], basis.calls)
2498
# keys in both are not duplicated:
2499
test.add_lines(key2, (), [])
2500
basis.add_lines(key2, (), [])
2503
self.assertEqual(2, len(keys))
2504
self.assertEqual(set([key1, key2]), set(keys))
2505
self.assertEqual([("keys",)], basis.calls)
2507
def test_add_mpdiffs(self):
2508
# records are inserted as normal; add_mpdiff builds on
2509
# add_lines, so a smoke test should be all that's needed:
2511
key_basis = ('bar',)
2512
key_delta = ('zaphod',)
2513
basis, test = self.get_basis_and_test_knit()
2514
source = self.make_test_knit(name='source')
2515
basis.add_lines(key_basis, (), ['foo\n'])
2517
source.add_lines(key_basis, (), ['foo\n'])
2518
source.add_lines(key_delta, (key_basis,), ['bar\n'])
2519
diffs = source.make_mpdiffs([key_delta])
2520
test.add_mpdiffs([(key_delta, (key_basis,),
2521
source.get_sha1s([key_delta])[key_delta], diffs[0])])
2522
self.assertEqual([("get_parent_map", set([key_basis])),
2523
('get_record_stream', [key_basis], 'unordered', True),],
2525
self.assertEqual({key_delta:(key_basis,)},
2526
test.get_parent_map([key_delta]))
2527
self.assertEqual('bar\n', test.get_record_stream([key_delta],
2528
'unordered', True).next().get_bytes_as('fulltext'))
2530
def test_make_mpdiffs(self):
2531
# Generating an mpdiff across a stacking boundary should detect parent
2535
key_right = ('zaphod',)
2536
basis, test = self.get_basis_and_test_knit()
2537
basis.add_lines(key_left, (), ['bar\n'])
2538
basis.add_lines(key_right, (), ['zaphod\n'])
2540
test.add_lines(key, (key_left, key_right),
2541
['bar\n', 'foo\n', 'zaphod\n'])
2542
diffs = test.make_mpdiffs([key])
2544
multiparent.MultiParent([multiparent.ParentText(0, 0, 0, 1),
2545
multiparent.NewText(['foo\n']),
2546
multiparent.ParentText(1, 0, 2, 1)])],
2548
self.assertEqual(3, len(basis.calls))
2550
("get_parent_map", set([key_left, key_right])),
2551
("get_parent_map", set([key_left, key_right])),
2554
last_call = basis.calls[-1]
2555
self.assertEqual('get_record_stream', last_call[0])
2556
self.assertEqual(set([key_left, key_right]), set(last_call[1]))
2557
self.assertEqual('topological', last_call[2])
2558
self.assertEqual(True, last_call[3])
2561
class TestNetworkBehaviour(KnitTests):
2562
"""Tests for getting data out of/into knits over the network."""
2564
def test_include_delta_closure_generates_a_knit_delta_closure(self):
2565
vf = self.make_test_knit(name='test')
2566
# put in three texts, giving ft, delta, delta
2567
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2568
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2569
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2570
# But heuristics could interfere, so check what happened:
2571
self.assertEqual(['knit-ft-gz', 'knit-delta-gz', 'knit-delta-gz'],
2572
[record.storage_kind for record in
2573
vf.get_record_stream([('base',), ('d1',), ('d2',)],
2574
'topological', False)])
2575
# generate a stream of just the deltas include_delta_closure=True,
2576
# serialise to the network, and check that we get a delta closure on the wire.
2577
stream = vf.get_record_stream([('d1',), ('d2',)], 'topological', True)
2578
netb = [record.get_bytes_as(record.storage_kind) for record in stream]
2579
# The first bytes should be a memo from _ContentMapGenerator, and the
2580
# second bytes should be empty (because its a API proxy not something
2581
# for wire serialisation.
2582
self.assertEqual('', netb[1])
2584
kind, line_end = network_bytes_to_kind_and_offset(bytes)
2585
self.assertEqual('knit-delta-closure', kind)
2588
class TestContentMapGenerator(KnitTests):
2589
"""Tests for ContentMapGenerator"""
2591
def test_get_record_stream_gives_records(self):
2592
vf = self.make_test_knit(name='test')
2593
# put in three texts, giving ft, delta, delta
2594
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2595
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2596
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2597
keys = [('d1',), ('d2',)]
2598
generator = _VFContentMapGenerator(vf, keys,
2599
global_map=vf.get_parent_map(keys))
2600
for record in generator.get_record_stream():
2601
if record.key == ('d1',):
2602
self.assertEqual('d1\n', record.get_bytes_as('fulltext'))
2604
self.assertEqual('d2\n', record.get_bytes_as('fulltext'))
2606
def test_get_record_stream_kinds_are_raw(self):
2607
vf = self.make_test_knit(name='test')
2608
# put in three texts, giving ft, delta, delta
2609
vf.add_lines(('base',), (), ['base\n', 'content\n'])
2610
vf.add_lines(('d1',), (('base',),), ['d1\n'])
2611
vf.add_lines(('d2',), (('d1',),), ['d2\n'])
2612
keys = [('base',), ('d1',), ('d2',)]
2613
generator = _VFContentMapGenerator(vf, keys,
2614
global_map=vf.get_parent_map(keys))
2615
kinds = {('base',): 'knit-delta-closure',
2616
('d1',): 'knit-delta-closure-ref',
2617
('d2',): 'knit-delta-closure-ref',
2619
for record in generator.get_record_stream():
2620
self.assertEqual(kinds[record.key], record.storage_kind)
added
removed
bzrlib/tests/test_knit.py
