← Back to branch summary

~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/tests/test_btree_index.py

  • Committer: Vincent Ladeuil
  • Date: 2008-09-11 19:36:38 UTC
  • mfrom: (3703 +trunk)
  • mto: (3705.1.1 trunk2)
  • mto: This revision was merged to the branch mainline in revision 3708.
  • Revision ID: v.ladeuil+lp@free.fr-20080911193638-wtjyc1kcmacc6t1f
merge bzr.dev

Show diffs side-by-side

added added

removed removed

Lines of Context:
bzrlib/tests/test_btree_index.py
1
 
# Copyright (C) 2008-2011 Canonical Ltd
 
1
# Copyright (C) 2008 Canonical Ltd
2
2
#
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
12
12
#
13
13
# You should have received a copy of the GNU General Public License
14
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
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
16
16
#
17
17
 
18
18
"""Tests for btree indices."""
23
23
from bzrlib import (
24
24
    btree_index,
25
25
    errors,
26
 
    fifo_cache,
27
 
    lru_cache,
28
 
    osutils,
29
26
    tests,
30
 
    transport,
31
27
    )
32
28
from bzrlib.tests import (
33
29
    TestCaseWithTransport,
34
 
    scenarios,
35
 
    )
36
 
from bzrlib.tests import (
37
 
    features,
38
 
    )
39
 
 
40
 
 
41
 
load_tests = scenarios.load_tests_apply_scenarios
42
 
 
43
 
 
44
 
def btreeparser_scenarios():
 
30
    TestScenarioApplier,
 
31
    adapt_tests,
 
32
    condition_isinstance,
 
33
    split_suite_by_condition,
 
34
    )
 
35
from bzrlib.transport import get_transport
 
36
 
 
37
 
 
38
def load_tests(standard_tests, module, loader):
 
39
    # parameterise the TestBTreeNodes tests
 
40
    node_tests, others = split_suite_by_condition(standard_tests,
 
41
        condition_isinstance(TestBTreeNodes))
 
42
    applier = TestScenarioApplier()
45
43
    import bzrlib._btree_serializer_py as py_module
46
 
    scenarios = [('python', {'parse_btree': py_module})]
47
 
    if compiled_btreeparser_feature.available():
48
 
        scenarios.append(('C', 
49
 
            {'parse_btree': compiled_btreeparser_feature.module}))
50
 
    return scenarios
51
 
 
52
 
 
53
 
compiled_btreeparser_feature = features.ModuleAvailableFeature(
54
 
    'bzrlib._btree_serializer_pyx')
 
44
    applier.scenarios = [('python', {'parse_btree': py_module})]
 
45
    if CompiledBtreeParserFeature.available():
 
46
        # Is there a way to do this that gets missing feature failures rather
 
47
        # than no indication to the user?
 
48
        import bzrlib._btree_serializer_c as c_module
 
49
        applier.scenarios.append(('C', {'parse_btree': c_module}))
 
50
    adapt_tests(node_tests, applier, others)
 
51
    return others
 
52
 
 
53
 
 
54
class _CompiledBtreeParserFeature(tests.Feature):
 
55
    def _probe(self):
 
56
        try:
 
57
            import bzrlib._btree_serializer_c
 
58
        except ImportError:
 
59
            return False
 
60
        return True
 
61
 
 
62
    def feature_name(self):
 
63
        return 'bzrlib._btree_serializer_c'
 
64
 
 
65
CompiledBtreeParserFeature = _CompiledBtreeParserFeature()
55
66
 
56
67
 
57
68
class BTreeTestCase(TestCaseWithTransport):
60
71
 
61
72
    def setUp(self):
62
73
        TestCaseWithTransport.setUp(self)
63
 
        self.overrideAttr(btree_index, '_RESERVED_HEADER_BYTES', 100)
 
74
        self._original_header = btree_index._RESERVED_HEADER_BYTES
 
75
        def restore():
 
76
            btree_index._RESERVED_HEADER_BYTES = self._original_header
 
77
        self.addCleanup(restore)
 
78
        btree_index._RESERVED_HEADER_BYTES = 100
64
79
 
65
80
    def make_nodes(self, count, key_elements, reference_lists):
66
81
        """Generate count*key_elements sample nodes."""
100
115
 
101
116
    def shrink_page_size(self):
102
117
        """Shrink the default page size so that less fits in a page."""
103
 
        self.overrideAttr(btree_index, '_PAGE_SIZE')
 
118
        old_page_size = btree_index._PAGE_SIZE
 
119
        def cleanup():
 
120
            btree_index._PAGE_SIZE = old_page_size
 
121
        self.addCleanup(cleanup)
104
122
        btree_index._PAGE_SIZE = 2048
105
123
 
106
124
 
107
125
class TestBTreeBuilder(BTreeTestCase):
108
126
 
109
 
    def test_clear_cache(self):
110
 
        builder = btree_index.BTreeBuilder(reference_lists=0, key_elements=1)
111
 
        # This is a no-op, but we need the api to be consistent with other
112
 
        # BTreeGraphIndex apis.
113
 
        builder.clear_cache()
114
 
 
115
127
    def test_empty_1_0(self):
116
128
        builder = btree_index.BTreeBuilder(key_elements=1, reference_lists=0)
117
129
        # NamedTemporaryFile dies on builder.finish().read(). weird.
143
155
        temp_file = builder.finish()
144
156
        content = temp_file.read()
145
157
        del temp_file
146
 
        self.assertEqual(131, len(content))
 
158
        self.assertEqual(158, len(content))
147
159
        self.assertEqual(
148
160
            "B+Tree Graph Index 2\nnode_ref_lists=0\nkey_elements=1\nlen=5\n"
149
161
            "row_lengths=1\n",
167
179
        temp_file = builder.finish()
168
180
        content = temp_file.read()
169
181
        del temp_file
170
 
        self.assertEqual(238, len(content))
 
182
        self.assertEqual(264, len(content))
171
183
        self.assertEqual(
172
184
            "B+Tree Graph Index 2\nnode_ref_lists=2\nkey_elements=2\nlen=10\n"
173
185
            "row_lengths=1\n",
217
229
        leaf1_bytes = zlib.decompress(leaf1)
218
230
        sorted_node_keys = sorted(node[0] for node in nodes)
219
231
        node = btree_index._LeafNode(leaf1_bytes, 1, 0)
220
 
        self.assertEqual(231, len(node))
221
 
        self.assertEqual(sorted_node_keys[:231], node.all_keys())
 
232
        self.assertEqual(231, len(node.keys))
 
233
        self.assertEqual(sorted_node_keys[:231], sorted(node.keys))
222
234
        leaf2_bytes = zlib.decompress(leaf2)
223
235
        node = btree_index._LeafNode(leaf2_bytes, 1, 0)
224
 
        self.assertEqual(400 - 231, len(node))
225
 
        self.assertEqual(sorted_node_keys[231:], node.all_keys())
 
236
        self.assertEqual(400 - 231, len(node.keys))
 
237
        self.assertEqual(sorted_node_keys[231:], sorted(node.keys))
226
238
 
227
239
    def test_last_page_rounded_1_layer(self):
228
240
        builder = btree_index.BTreeBuilder(key_elements=1, reference_lists=0)
233
245
        temp_file = builder.finish()
234
246
        content = temp_file.read()
235
247
        del temp_file
236
 
        self.assertEqual(155, len(content))
 
248
        self.assertEqual(181, len(content))
237
249
        self.assertEqual(
238
250
            "B+Tree Graph Index 2\nnode_ref_lists=0\nkey_elements=1\nlen=10\n"
239
251
            "row_lengths=1\n",
242
254
        leaf2 = content[74:]
243
255
        leaf2_bytes = zlib.decompress(leaf2)
244
256
        node = btree_index._LeafNode(leaf2_bytes, 1, 0)
245
 
        self.assertEqual(10, len(node))
 
257
        self.assertEqual(10, len(node.keys))
246
258
        sorted_node_keys = sorted(node[0] for node in nodes)
247
 
        self.assertEqual(sorted_node_keys, node.all_keys())
 
259
        self.assertEqual(sorted_node_keys, sorted(node.keys))
248
260
 
249
261
    def test_last_page_not_rounded_2_layer(self):
250
262
        builder = btree_index.BTreeBuilder(key_elements=1, reference_lists=0)
264
276
        leaf2 = content[8192:]
265
277
        leaf2_bytes = zlib.decompress(leaf2)
266
278
        node = btree_index._LeafNode(leaf2_bytes, 1, 0)
267
 
        self.assertEqual(400 - 231, len(node))
 
279
        self.assertEqual(400 - 231, len(node.keys))
268
280
        sorted_node_keys = sorted(node[0] for node in nodes)
269
 
        self.assertEqual(sorted_node_keys[231:], node.all_keys())
 
281
        self.assertEqual(sorted_node_keys[231:], sorted(node.keys))
270
282
 
271
283
    def test_three_level_tree_details(self):
272
284
        # The left most pointer in the second internal node in a row should
281
293
 
282
294
        for node in nodes:
283
295
            builder.add_node(*node)
284
 
        t = transport.get_transport('trace+' + self.get_url(''))
285
 
        size = t.put_file('index', self.time(builder.finish))
 
296
        transport = get_transport('trace+' + self.get_url(''))
 
297
        size = transport.put_file('index', self.time(builder.finish))
286
298
        del builder
287
 
        index = btree_index.BTreeGraphIndex(t, 'index', size)
 
299
        index = btree_index.BTreeGraphIndex(transport, 'index', size)
288
300
        # Seed the metadata, we're using internal calls now.
289
301
        index.key_count()
290
302
        self.assertEqual(3, len(index._row_lengths),
303
315
        # in the second node it points at
304
316
        pos = index._row_offsets[2] + internal_node2.offset + 1
305
317
        leaf = index._get_leaf_nodes([pos])[pos]
306
 
        self.assertTrue(internal_node2.keys[0] in leaf)
 
318
        self.assertTrue(internal_node2.keys[0] in leaf.keys)
307
319
 
308
320
    def test_2_leaves_2_2(self):
309
321
        builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
341
353
        # Test the parts of the index that take up memory are doing so
342
354
        # predictably.
343
355
        self.assertEqual(1, len(builder._nodes))
 
356
        self.assertEqual(1, len(builder._keys))
344
357
        self.assertIs(None, builder._nodes_by_key)
345
358
        builder.add_node(*nodes[1])
346
359
        self.assertEqual(0, len(builder._nodes))
 
360
        self.assertEqual(0, len(builder._keys))
347
361
        self.assertIs(None, builder._nodes_by_key)
348
362
        self.assertEqual(1, len(builder._backing_indices))
349
363
        self.assertEqual(2, builder._backing_indices[0].key_count())
350
364
        # now back to memory
351
365
        builder.add_node(*nodes[2])
352
366
        self.assertEqual(1, len(builder._nodes))
 
367
        self.assertEqual(1, len(builder._keys))
353
368
        self.assertIs(None, builder._nodes_by_key)
354
369
        # And spills to a second backing index combing all
355
370
        builder.add_node(*nodes[3])
356
371
        self.assertEqual(0, len(builder._nodes))
 
372
        self.assertEqual(0, len(builder._keys))
357
373
        self.assertIs(None, builder._nodes_by_key)
358
374
        self.assertEqual(2, len(builder._backing_indices))
359
375
        self.assertEqual(None, builder._backing_indices[0])
362
378
        builder.add_node(*nodes[4])
363
379
        builder.add_node(*nodes[5])
364
380
        self.assertEqual(0, len(builder._nodes))
 
381
        self.assertEqual(0, len(builder._keys))
365
382
        self.assertIs(None, builder._nodes_by_key)
366
383
        self.assertEqual(2, len(builder._backing_indices))
367
384
        self.assertEqual(2, builder._backing_indices[0].key_count())
410
427
        self.assertEqual(None, builder._backing_indices[2])
411
428
        self.assertEqual(16, builder._backing_indices[3].key_count())
412
429
        # Now finish, and check we got a correctly ordered tree
413
 
        t = self.get_transport('')
414
 
        size = t.put_file('index', builder.finish())
415
 
        index = btree_index.BTreeGraphIndex(t, 'index', size)
416
 
        nodes = list(index.iter_all_entries())
417
 
        self.assertEqual(sorted(nodes), nodes)
418
 
        self.assertEqual(16, len(nodes))
419
 
 
420
 
    def test_spill_index_stress_1_1_no_combine(self):
421
 
        builder = btree_index.BTreeBuilder(key_elements=1, spill_at=2)
422
 
        builder.set_optimize(for_size=False, combine_backing_indices=False)
423
 
        nodes = [node[0:2] for node in self.make_nodes(16, 1, 0)]
424
 
        builder.add_node(*nodes[0])
425
 
        # Test the parts of the index that take up memory are doing so
426
 
        # predictably.
427
 
        self.assertEqual(1, len(builder._nodes))
428
 
        self.assertIs(None, builder._nodes_by_key)
429
 
        builder.add_node(*nodes[1])
430
 
        self.assertEqual(0, len(builder._nodes))
431
 
        self.assertIs(None, builder._nodes_by_key)
432
 
        self.assertEqual(1, len(builder._backing_indices))
433
 
        self.assertEqual(2, builder._backing_indices[0].key_count())
434
 
        # now back to memory
435
 
        builder.add_node(*nodes[2])
436
 
        self.assertEqual(1, len(builder._nodes))
437
 
        self.assertIs(None, builder._nodes_by_key)
438
 
        # And spills to a second backing index but doesn't combine
439
 
        builder.add_node(*nodes[3])
440
 
        self.assertEqual(0, len(builder._nodes))
441
 
        self.assertIs(None, builder._nodes_by_key)
442
 
        self.assertEqual(2, len(builder._backing_indices))
443
 
        for backing_index in builder._backing_indices:
444
 
            self.assertEqual(2, backing_index.key_count())
445
 
        # The next spills to the 3rd slot
446
 
        builder.add_node(*nodes[4])
447
 
        builder.add_node(*nodes[5])
448
 
        self.assertEqual(0, len(builder._nodes))
449
 
        self.assertIs(None, builder._nodes_by_key)
450
 
        self.assertEqual(3, len(builder._backing_indices))
451
 
        for backing_index in builder._backing_indices:
452
 
            self.assertEqual(2, backing_index.key_count())
453
 
        # Now spill a few more, and check that we don't combine
454
 
        builder.add_node(*nodes[6])
455
 
        builder.add_node(*nodes[7])
456
 
        builder.add_node(*nodes[8])
457
 
        builder.add_node(*nodes[9])
458
 
        builder.add_node(*nodes[10])
459
 
        builder.add_node(*nodes[11])
460
 
        builder.add_node(*nodes[12])
461
 
        self.assertEqual(6, len(builder._backing_indices))
462
 
        for backing_index in builder._backing_indices:
463
 
            self.assertEqual(2, backing_index.key_count())
464
 
        # Test that memory and disk are both used for query methods; and that
465
 
        # None is skipped over happily.
466
 
        self.assertEqual([(builder,) + node for node in sorted(nodes[:13])],
467
 
            list(builder.iter_all_entries()))
468
 
        # Two nodes - one memory one disk
469
 
        self.assertEqual(set([(builder,) + node for node in nodes[11:13]]),
470
 
            set(builder.iter_entries([nodes[12][0], nodes[11][0]])))
471
 
        self.assertEqual(13, builder.key_count())
472
 
        self.assertEqual(set([(builder,) + node for node in nodes[11:13]]),
473
 
            set(builder.iter_entries_prefix([nodes[12][0], nodes[11][0]])))
474
 
        builder.add_node(*nodes[13])
475
 
        builder.add_node(*nodes[14])
476
 
        builder.add_node(*nodes[15])
477
 
        self.assertEqual(8, len(builder._backing_indices))
478
 
        for backing_index in builder._backing_indices:
479
 
            self.assertEqual(2, backing_index.key_count())
480
 
        # Now finish, and check we got a correctly ordered tree
481
430
        transport = self.get_transport('')
482
431
        size = transport.put_file('index', builder.finish())
483
432
        index = btree_index.BTreeGraphIndex(transport, 'index', size)
485
434
        self.assertEqual(sorted(nodes), nodes)
486
435
        self.assertEqual(16, len(nodes))
487
436
 
488
 
    def test_set_optimize(self):
489
 
        builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
490
 
        builder.set_optimize(for_size=True)
491
 
        self.assertTrue(builder._optimize_for_size)
492
 
        builder.set_optimize(for_size=False)
493
 
        self.assertFalse(builder._optimize_for_size)
494
 
        # test that we can set combine_backing_indices without effecting
495
 
        # _optimize_for_size
496
 
        obj = object()
497
 
        builder._optimize_for_size = obj
498
 
        builder.set_optimize(combine_backing_indices=False)
499
 
        self.assertFalse(builder._combine_backing_indices)
500
 
        self.assertIs(obj, builder._optimize_for_size)
501
 
        builder.set_optimize(combine_backing_indices=True)
502
 
        self.assertTrue(builder._combine_backing_indices)
503
 
        self.assertIs(obj, builder._optimize_for_size)
504
 
 
505
437
    def test_spill_index_stress_2_2(self):
506
438
        # test that references and longer keys don't confuse things.
507
439
        builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2,
510
442
        builder.add_node(*nodes[0])
511
443
        # Test the parts of the index that take up memory are doing so
512
444
        # predictably.
 
445
        self.assertEqual(1, len(builder._keys))
513
446
        self.assertEqual(1, len(builder._nodes))
514
447
        self.assertIs(None, builder._nodes_by_key)
515
448
        builder.add_node(*nodes[1])
 
449
        self.assertEqual(0, len(builder._keys))
516
450
        self.assertEqual(0, len(builder._nodes))
517
451
        self.assertIs(None, builder._nodes_by_key)
518
452
        self.assertEqual(1, len(builder._backing_indices))
521
455
        old = dict(builder._get_nodes_by_key()) #Build up the nodes by key dict
522
456
        builder.add_node(*nodes[2])
523
457
        self.assertEqual(1, len(builder._nodes))
 
458
        self.assertEqual(1, len(builder._keys))
524
459
        self.assertIsNot(None, builder._nodes_by_key)
525
460
        self.assertNotEqual({}, builder._nodes_by_key)
526
461
        # We should have a new entry
528
463
        # And spills to a second backing index combing all
529
464
        builder.add_node(*nodes[3])
530
465
        self.assertEqual(0, len(builder._nodes))
 
466
        self.assertEqual(0, len(builder._keys))
531
467
        self.assertIs(None, builder._nodes_by_key)
532
468
        self.assertEqual(2, len(builder._backing_indices))
533
469
        self.assertEqual(None, builder._backing_indices[0])
536
472
        builder.add_node(*nodes[4])
537
473
        builder.add_node(*nodes[5])
538
474
        self.assertEqual(0, len(builder._nodes))
 
475
        self.assertEqual(0, len(builder._keys))
539
476
        self.assertIs(None, builder._nodes_by_key)
540
477
        self.assertEqual(2, len(builder._backing_indices))
541
478
        self.assertEqual(2, builder._backing_indices[0].key_count())
608
545
        for key, value, references in nodes:
609
546
            builder.add_node(key, value, references)
610
547
        stream = builder.finish()
611
 
        trans = transport.get_transport('trace+' + self.get_url())
 
548
        trans = get_transport('trace+' + self.get_url())
612
549
        size = trans.put_file('index', stream)
613
550
        return btree_index.BTreeGraphIndex(trans, 'index', size)
614
551
 
615
 
    def make_index_with_offset(self, ref_lists=1, key_elements=1, nodes=[],
616
 
                               offset=0):
617
 
        builder = btree_index.BTreeBuilder(key_elements=key_elements,
618
 
                                           reference_lists=ref_lists)
619
 
        builder.add_nodes(nodes)
620
 
        transport = self.get_transport('')
621
 
        # NamedTemporaryFile dies on builder.finish().read(). weird.
622
 
        temp_file = builder.finish()
623
 
        content = temp_file.read()
624
 
        del temp_file
625
 
        size = len(content)
626
 
        transport.put_bytes('index', (' '*offset)+content)
627
 
        return btree_index.BTreeGraphIndex(transport, 'index', size=size,
628
 
                                           offset=offset)
629
 
 
630
 
    def test_clear_cache(self):
631
 
        nodes = self.make_nodes(160, 2, 2)
632
 
        index = self.make_index(ref_lists=2, key_elements=2, nodes=nodes)
633
 
        self.assertEqual(1, len(list(index.iter_entries([nodes[30][0]]))))
634
 
        self.assertEqual([1, 4], index._row_lengths)
635
 
        self.assertIsNot(None, index._root_node)
636
 
        internal_node_pre_clear = index._internal_node_cache.keys()
637
 
        self.assertTrue(len(index._leaf_node_cache) > 0)
638
 
        index.clear_cache()
639
 
        # We don't touch _root_node or _internal_node_cache, both should be
640
 
        # small, and can save a round trip or two
641
 
        self.assertIsNot(None, index._root_node)
642
 
        # NOTE: We don't want to affect the _internal_node_cache, as we expect
643
 
        #       it will be small, and if we ever do touch this index again, it
644
 
        #       will save round-trips.  This assertion isn't very strong,
645
 
        #       becuase without a 3-level index, we don't have any internal
646
 
        #       nodes cached.
647
 
        self.assertEqual(internal_node_pre_clear,
648
 
                         index._internal_node_cache.keys())
649
 
        self.assertEqual(0, len(index._leaf_node_cache))
650
 
 
651
552
    def test_trivial_constructor(self):
652
 
        t = transport.get_transport('trace+' + self.get_url(''))
653
 
        index = btree_index.BTreeGraphIndex(t, 'index', None)
 
553
        transport = get_transport('trace+' + self.get_url(''))
 
554
        index = btree_index.BTreeGraphIndex(transport, 'index', None)
654
555
        # Checks the page size at load, but that isn't logged yet.
655
 
        self.assertEqual([], t._activity)
 
556
        self.assertEqual([], transport._activity)
656
557
 
657
558
    def test_with_size_constructor(self):
658
 
        t = transport.get_transport('trace+' + self.get_url(''))
659
 
        index = btree_index.BTreeGraphIndex(t, 'index', 1)
 
559
        transport = get_transport('trace+' + self.get_url(''))
 
560
        index = btree_index.BTreeGraphIndex(transport, 'index', 1)
660
561
        # Checks the page size at load, but that isn't logged yet.
661
 
        self.assertEqual([], t._activity)
 
562
        self.assertEqual([], transport._activity)
662
563
 
663
564
    def test_empty_key_count_no_size(self):
664
565
        builder = btree_index.BTreeBuilder(key_elements=1, reference_lists=0)
665
 
        t = transport.get_transport('trace+' + self.get_url(''))
666
 
        t.put_file('index', builder.finish())
667
 
        index = btree_index.BTreeGraphIndex(t, 'index', None)
668
 
        del t._activity[:]
669
 
        self.assertEqual([], t._activity)
 
566
        transport = get_transport('trace+' + self.get_url(''))
 
567
        transport.put_file('index', builder.finish())
 
568
        index = btree_index.BTreeGraphIndex(transport, 'index', None)
 
569
        del transport._activity[:]
 
570
        self.assertEqual([], transport._activity)
670
571
        self.assertEqual(0, index.key_count())
671
572
        # The entire index should have been requested (as we generally have the
672
573
        # size available, and doing many small readvs is inappropriate).
673
574
        # We can't tell how much was actually read here, but - check the code.
674
 
        self.assertEqual([('get', 'index')], t._activity)
 
575
        self.assertEqual([('get', 'index'),
 
576
            ('readv', 'index', [(0, 72)], False, None)],
 
577
            transport._activity)
675
578
 
676
579
    def test_empty_key_count(self):
677
580
        builder = btree_index.BTreeBuilder(key_elements=1, reference_lists=0)
678
 
        t = transport.get_transport('trace+' + self.get_url(''))
679
 
        size = t.put_file('index', builder.finish())
 
581
        transport = get_transport('trace+' + self.get_url(''))
 
582
        size = transport.put_file('index', builder.finish())
680
583
        self.assertEqual(72, size)
681
 
        index = btree_index.BTreeGraphIndex(t, 'index', size)
682
 
        del t._activity[:]
683
 
        self.assertEqual([], t._activity)
 
584
        index = btree_index.BTreeGraphIndex(transport, 'index', size)
 
585
        del transport._activity[:]
 
586
        self.assertEqual([], transport._activity)
684
587
        self.assertEqual(0, index.key_count())
685
588
        # The entire index should have been read, as 4K > size
686
589
        self.assertEqual([('readv', 'index', [(0, 72)], False, None)],
687
 
                         t._activity)
 
590
            transport._activity)
688
591
 
689
592
    def test_non_empty_key_count_2_2(self):
690
593
        builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
691
594
        nodes = self.make_nodes(35, 2, 2)
692
595
        for node in nodes:
693
596
            builder.add_node(*node)
694
 
        t = transport.get_transport('trace+' + self.get_url(''))
695
 
        size = t.put_file('index', builder.finish())
696
 
        index = btree_index.BTreeGraphIndex(t, 'index', size)
697
 
        del t._activity[:]
698
 
        self.assertEqual([], t._activity)
 
597
        transport = get_transport('trace+' + self.get_url(''))
 
598
        size = transport.put_file('index', builder.finish())
 
599
        index = btree_index.BTreeGraphIndex(transport, 'index', size)
 
600
        del transport._activity[:]
 
601
        self.assertEqual([], transport._activity)
699
602
        self.assertEqual(70, index.key_count())
700
603
        # The entire index should have been read, as it is one page long.
701
604
        self.assertEqual([('readv', 'index', [(0, size)], False, None)],
702
 
            t._activity)
703
 
        self.assertEqual(1173, size)
704
 
 
705
 
    def test_with_offset_no_size(self):
706
 
        index = self.make_index_with_offset(key_elements=1, ref_lists=1,
707
 
                                            offset=1234,
708
 
                                            nodes=self.make_nodes(200, 1, 1))
709
 
        index._size = None # throw away the size info
710
 
        self.assertEqual(200, index.key_count())
711
 
 
712
 
    def test_with_small_offset(self):
713
 
        index = self.make_index_with_offset(key_elements=1, ref_lists=1,
714
 
                                            offset=1234,
715
 
                                            nodes=self.make_nodes(200, 1, 1))
716
 
        self.assertEqual(200, index.key_count())
717
 
 
718
 
    def test_with_large_offset(self):
719
 
        index = self.make_index_with_offset(key_elements=1, ref_lists=1,
720
 
                                            offset=123456,
721
 
                                            nodes=self.make_nodes(200, 1, 1))
722
 
        self.assertEqual(200, index.key_count())
723
 
 
724
 
    def test__read_nodes_no_size_one_page_reads_once(self):
725
 
        self.make_index(nodes=[(('key',), 'value', ())])
726
 
        trans = transport.get_transport('trace+' + self.get_url())
727
 
        index = btree_index.BTreeGraphIndex(trans, 'index', None)
728
 
        del trans._activity[:]
729
 
        nodes = dict(index._read_nodes([0]))
730
 
        self.assertEqual([0], nodes.keys())
731
 
        node = nodes[0]
732
 
        self.assertEqual([('key',)], node.all_keys())
733
 
        self.assertEqual([('get', 'index')], trans._activity)
734
 
 
735
 
    def test__read_nodes_no_size_multiple_pages(self):
736
 
        index = self.make_index(2, 2, nodes=self.make_nodes(160, 2, 2))
737
 
        index.key_count()
738
 
        num_pages = index._row_offsets[-1]
739
 
        # Reopen with a traced transport and no size
740
 
        trans = transport.get_transport('trace+' + self.get_url())
741
 
        index = btree_index.BTreeGraphIndex(trans, 'index', None)
742
 
        del trans._activity[:]
743
 
        nodes = dict(index._read_nodes([0]))
744
 
        self.assertEqual(range(num_pages), nodes.keys())
 
605
            transport._activity)
 
606
        self.assertEqual(1199, size)
745
607
 
746
608
    def test_2_levels_key_count_2_2(self):
747
609
        builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
748
610
        nodes = self.make_nodes(160, 2, 2)
749
611
        for node in nodes:
750
612
            builder.add_node(*node)
751
 
        t = transport.get_transport('trace+' + self.get_url(''))
752
 
        size = t.put_file('index', builder.finish())
 
613
        transport = get_transport('trace+' + self.get_url(''))
 
614
        size = transport.put_file('index', builder.finish())
753
615
        self.assertEqual(17692, size)
754
 
        index = btree_index.BTreeGraphIndex(t, 'index', size)
755
 
        del t._activity[:]
756
 
        self.assertEqual([], t._activity)
 
616
        index = btree_index.BTreeGraphIndex(transport, 'index', size)
 
617
        del transport._activity[:]
 
618
        self.assertEqual([], transport._activity)
757
619
        self.assertEqual(320, index.key_count())
758
620
        # The entire index should not have been read.
759
621
        self.assertEqual([('readv', 'index', [(0, 4096)], False, None)],
760
 
                         t._activity)
 
622
            transport._activity)
761
623
 
762
624
    def test_validate_one_page(self):
763
625
        builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
764
626
        nodes = self.make_nodes(45, 2, 2)
765
627
        for node in nodes:
766
628
            builder.add_node(*node)
767
 
        t = transport.get_transport('trace+' + self.get_url(''))
768
 
        size = t.put_file('index', builder.finish())
769
 
        index = btree_index.BTreeGraphIndex(t, 'index', size)
770
 
        del t._activity[:]
771
 
        self.assertEqual([], t._activity)
 
629
        transport = get_transport('trace+' + self.get_url(''))
 
630
        size = transport.put_file('index', builder.finish())
 
631
        index = btree_index.BTreeGraphIndex(transport, 'index', size)
 
632
        del transport._activity[:]
 
633
        self.assertEqual([], transport._activity)
772
634
        index.validate()
773
635
        # The entire index should have been read linearly.
774
636
        self.assertEqual([('readv', 'index', [(0, size)], False, None)],
775
 
                         t._activity)
776
 
        self.assertEqual(1488, size)
 
637
            transport._activity)
 
638
        self.assertEqual(1514, size)
777
639
 
778
640
    def test_validate_two_pages(self):
779
641
        builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
780
642
        nodes = self.make_nodes(80, 2, 2)
781
643
        for node in nodes:
782
644
            builder.add_node(*node)
783
 
        t = transport.get_transport('trace+' + self.get_url(''))
784
 
        size = t.put_file('index', builder.finish())
 
645
        transport = get_transport('trace+' + self.get_url(''))
 
646
        size = transport.put_file('index', builder.finish())
785
647
        # Root page, 2 leaf pages
786
648
        self.assertEqual(9339, size)
787
 
        index = btree_index.BTreeGraphIndex(t, 'index', size)
788
 
        del t._activity[:]
789
 
        self.assertEqual([], t._activity)
 
649
        index = btree_index.BTreeGraphIndex(transport, 'index', size)
 
650
        del transport._activity[:]
 
651
        self.assertEqual([], transport._activity)
790
652
        index.validate()
791
653
        # The entire index should have been read linearly.
792
 
        self.assertEqual(
793
 
            [('readv', 'index', [(0, 4096)], False, None),
794
 
             ('readv', 'index', [(4096, 4096), (8192, 1147)], False, None)],
795
 
            t._activity)
 
654
        self.assertEqual([('readv', 'index', [(0, 4096)], False, None),
 
655
            ('readv', 'index', [(4096, 4096), (8192, 1147)], False, None)],
 
656
            transport._activity)
796
657
        # XXX: TODO: write some badly-ordered nodes, and some pointers-to-wrong
797
658
        # node and make validate find them.
798
659
 
799
660
    def test_eq_ne(self):
800
661
        # two indices are equal when constructed with the same parameters:
801
 
        t1 = transport.get_transport('trace+' + self.get_url(''))
802
 
        t2 = self.get_transport()
803
 
        self.assertTrue(
804
 
            btree_index.BTreeGraphIndex(t1, 'index', None) ==
805
 
            btree_index.BTreeGraphIndex(t1, 'index', None))
806
 
        self.assertTrue(
807
 
            btree_index.BTreeGraphIndex(t1, 'index', 20) ==
808
 
            btree_index.BTreeGraphIndex(t1, 'index', 20))
809
 
        self.assertFalse(
810
 
            btree_index.BTreeGraphIndex(t1, 'index', 20) ==
811
 
            btree_index.BTreeGraphIndex(t2, 'index', 20))
812
 
        self.assertFalse(
813
 
            btree_index.BTreeGraphIndex(t1, 'inde1', 20) ==
814
 
            btree_index.BTreeGraphIndex(t1, 'inde2', 20))
815
 
        self.assertFalse(
816
 
            btree_index.BTreeGraphIndex(t1, 'index', 10) ==
817
 
            btree_index.BTreeGraphIndex(t1, 'index', 20))
818
 
        self.assertFalse(
819
 
            btree_index.BTreeGraphIndex(t1, 'index', None) !=
820
 
            btree_index.BTreeGraphIndex(t1, 'index', None))
821
 
        self.assertFalse(
822
 
            btree_index.BTreeGraphIndex(t1, 'index', 20) !=
823
 
            btree_index.BTreeGraphIndex(t1, 'index', 20))
824
 
        self.assertTrue(
825
 
            btree_index.BTreeGraphIndex(t1, 'index', 20) !=
826
 
            btree_index.BTreeGraphIndex(t2, 'index', 20))
827
 
        self.assertTrue(
828
 
            btree_index.BTreeGraphIndex(t1, 'inde1', 20) !=
829
 
            btree_index.BTreeGraphIndex(t1, 'inde2', 20))
830
 
        self.assertTrue(
831
 
            btree_index.BTreeGraphIndex(t1, 'index', 10) !=
832
 
            btree_index.BTreeGraphIndex(t1, 'index', 20))
833
 
 
834
 
    def test_iter_all_only_root_no_size(self):
835
 
        self.make_index(nodes=[(('key',), 'value', ())])
836
 
        t = transport.get_transport('trace+' + self.get_url(''))
837
 
        index = btree_index.BTreeGraphIndex(t, 'index', None)
838
 
        del t._activity[:]
839
 
        self.assertEqual([(('key',), 'value')],
840
 
                         [x[1:] for x in index.iter_all_entries()])
841
 
        self.assertEqual([('get', 'index')], t._activity)
 
662
        transport1 = get_transport('trace+' + self.get_url(''))
 
663
        transport2 = get_transport(self.get_url(''))
 
664
        self.assertTrue(
 
665
            btree_index.BTreeGraphIndex(transport1, 'index', None) ==
 
666
            btree_index.BTreeGraphIndex(transport1, 'index', None))
 
667
        self.assertTrue(
 
668
            btree_index.BTreeGraphIndex(transport1, 'index', 20) ==
 
669
            btree_index.BTreeGraphIndex(transport1, 'index', 20))
 
670
        self.assertFalse(
 
671
            btree_index.BTreeGraphIndex(transport1, 'index', 20) ==
 
672
            btree_index.BTreeGraphIndex(transport2, 'index', 20))
 
673
        self.assertFalse(
 
674
            btree_index.BTreeGraphIndex(transport1, 'inde1', 20) ==
 
675
            btree_index.BTreeGraphIndex(transport1, 'inde2', 20))
 
676
        self.assertFalse(
 
677
            btree_index.BTreeGraphIndex(transport1, 'index', 10) ==
 
678
            btree_index.BTreeGraphIndex(transport1, 'index', 20))
 
679
        self.assertFalse(
 
680
            btree_index.BTreeGraphIndex(transport1, 'index', None) !=
 
681
            btree_index.BTreeGraphIndex(transport1, 'index', None))
 
682
        self.assertFalse(
 
683
            btree_index.BTreeGraphIndex(transport1, 'index', 20) !=
 
684
            btree_index.BTreeGraphIndex(transport1, 'index', 20))
 
685
        self.assertTrue(
 
686
            btree_index.BTreeGraphIndex(transport1, 'index', 20) !=
 
687
            btree_index.BTreeGraphIndex(transport2, 'index', 20))
 
688
        self.assertTrue(
 
689
            btree_index.BTreeGraphIndex(transport1, 'inde1', 20) !=
 
690
            btree_index.BTreeGraphIndex(transport1, 'inde2', 20))
 
691
        self.assertTrue(
 
692
            btree_index.BTreeGraphIndex(transport1, 'index', 10) !=
 
693
            btree_index.BTreeGraphIndex(transport1, 'index', 20))
842
694
 
843
695
    def test_iter_all_entries_reads(self):
844
696
        # iterating all entries reads the header, then does a linear
850
702
        nodes = self.make_nodes(10000, 2, 2)
851
703
        for node in nodes:
852
704
            builder.add_node(*node)
853
 
        t = transport.get_transport('trace+' + self.get_url(''))
854
 
        size = t.put_file('index', builder.finish())
 
705
        transport = get_transport('trace+' + self.get_url(''))
 
706
        size = transport.put_file('index', builder.finish())
855
707
        self.assertEqual(1303220, size, 'number of expected bytes in the'
856
708
                                        ' output changed')
857
709
        page_size = btree_index._PAGE_SIZE
858
710
        del builder
859
 
        index = btree_index.BTreeGraphIndex(t, 'index', size)
860
 
        del t._activity[:]
861
 
        self.assertEqual([], t._activity)
 
711
        index = btree_index.BTreeGraphIndex(transport, 'index', size)
 
712
        del transport._activity[:]
 
713
        self.assertEqual([], transport._activity)
862
714
        found_nodes = self.time(list, index.iter_all_entries())
863
715
        bare_nodes = []
864
716
        for node in found_nodes:
885
737
        readv_request[-1] = (readv_request[-1][0], 1303220 % page_size)
886
738
        expected = [('readv', 'index', [(0, page_size)], False, None),
887
739
             ('readv',  'index', readv_request, False, None)]
888
 
        if expected != t._activity:
 
740
        if expected != transport._activity:
889
741
            self.assertEqualDiff(pprint.pformat(expected),
890
742
                                 pprint.pformat(transport._activity))
891
743
 
905
757
        nodes = self.make_nodes(160, 2, 2)
906
758
        for node in nodes:
907
759
            builder.add_node(*node)
908
 
        t = transport.get_transport('trace+' + self.get_url(''))
909
 
        size = t.put_file('index', builder.finish())
 
760
        transport = get_transport('trace+' + self.get_url(''))
 
761
        size = transport.put_file('index', builder.finish())
910
762
        del builder
911
 
        index = btree_index.BTreeGraphIndex(t, 'index', size)
912
 
        del t._activity[:]
913
 
        self.assertEqual([], t._activity)
 
763
        index = btree_index.BTreeGraphIndex(transport, 'index', size)
 
764
        del transport._activity[:]
 
765
        self.assertEqual([], transport._activity)
914
766
        # search for one key
915
767
        found_nodes = list(index.iter_entries([nodes[30][0]]))
916
768
        bare_nodes = []
924
776
        # Should have read the root node, then one leaf page:
925
777
        self.assertEqual([('readv', 'index', [(0, 4096)], False, None),
926
778
             ('readv',  'index', [(8192, 4096), ], False, None)],
927
 
            t._activity)
 
779
            transport._activity)
928
780
 
929
781
    def test_iter_key_prefix_1_element_key_None(self):
930
782
        index = self.make_index()
981
833
            (index, ('name', 'fin2'), 'beta', ((), ))]),
982
834
            set(index.iter_entries_prefix([('name', None)])))
983
835
 
984
 
    # XXX: external_references tests are duplicated in test_index.  We
985
 
    # probably should have per_graph_index tests...
986
 
    def test_external_references_no_refs(self):
987
 
        index = self.make_index(ref_lists=0, nodes=[])
988
 
        self.assertRaises(ValueError, index.external_references, 0)
989
 
 
990
 
    def test_external_references_no_results(self):
991
 
        index = self.make_index(ref_lists=1, nodes=[
992
 
            (('key',), 'value', ([],))])
993
 
        self.assertEqual(set(), index.external_references(0))
994
 
 
995
 
    def test_external_references_missing_ref(self):
996
 
        missing_key = ('missing',)
997
 
        index = self.make_index(ref_lists=1, nodes=[
998
 
            (('key',), 'value', ([missing_key],))])
999
 
        self.assertEqual(set([missing_key]), index.external_references(0))
1000
 
 
1001
 
    def test_external_references_multiple_ref_lists(self):
1002
 
        missing_key = ('missing',)
1003
 
        index = self.make_index(ref_lists=2, nodes=[
1004
 
            (('key',), 'value', ([], [missing_key]))])
1005
 
        self.assertEqual(set([]), index.external_references(0))
1006
 
        self.assertEqual(set([missing_key]), index.external_references(1))
1007
 
 
1008
 
    def test_external_references_two_records(self):
1009
 
        index = self.make_index(ref_lists=1, nodes=[
1010
 
            (('key-1',), 'value', ([('key-2',)],)),
1011
 
            (('key-2',), 'value', ([],)),
1012
 
            ])
1013
 
        self.assertEqual(set([]), index.external_references(0))
1014
 
 
1015
 
    def test__find_ancestors_one_page(self):
1016
 
        key1 = ('key-1',)
1017
 
        key2 = ('key-2',)
1018
 
        index = self.make_index(ref_lists=1, key_elements=1, nodes=[
1019
 
            (key1, 'value', ([key2],)),
1020
 
            (key2, 'value', ([],)),
1021
 
            ])
1022
 
        parent_map = {}
1023
 
        missing_keys = set()
1024
 
        search_keys = index._find_ancestors([key1], 0, parent_map, missing_keys)
1025
 
        self.assertEqual({key1: (key2,), key2: ()}, parent_map)
1026
 
        self.assertEqual(set(), missing_keys)
1027
 
        self.assertEqual(set(), search_keys)
1028
 
 
1029
 
    def test__find_ancestors_one_page_w_missing(self):
1030
 
        key1 = ('key-1',)
1031
 
        key2 = ('key-2',)
1032
 
        key3 = ('key-3',)
1033
 
        index = self.make_index(ref_lists=1, key_elements=1, nodes=[
1034
 
            (key1, 'value', ([key2],)),
1035
 
            (key2, 'value', ([],)),
1036
 
            ])
1037
 
        parent_map = {}
1038
 
        missing_keys = set()
1039
 
        search_keys = index._find_ancestors([key2, key3], 0, parent_map,
1040
 
                                            missing_keys)
1041
 
        self.assertEqual({key2: ()}, parent_map)
1042
 
        # we know that key3 is missing because we read the page that it would
1043
 
        # otherwise be on
1044
 
        self.assertEqual(set([key3]), missing_keys)
1045
 
        self.assertEqual(set(), search_keys)
1046
 
 
1047
 
    def test__find_ancestors_one_parent_missing(self):
1048
 
        key1 = ('key-1',)
1049
 
        key2 = ('key-2',)
1050
 
        key3 = ('key-3',)
1051
 
        index = self.make_index(ref_lists=1, key_elements=1, nodes=[
1052
 
            (key1, 'value', ([key2],)),
1053
 
            (key2, 'value', ([key3],)),
1054
 
            ])
1055
 
        parent_map = {}
1056
 
        missing_keys = set()
1057
 
        search_keys = index._find_ancestors([key1], 0, parent_map,
1058
 
                                            missing_keys)
1059
 
        self.assertEqual({key1: (key2,), key2: (key3,)}, parent_map)
1060
 
        self.assertEqual(set(), missing_keys)
1061
 
        # all we know is that key3 wasn't present on the page we were reading
1062
 
        # but if you look, the last key is key2 which comes before key3, so we
1063
 
        # don't know whether key3 would land on this page or not.
1064
 
        self.assertEqual(set([key3]), search_keys)
1065
 
        search_keys = index._find_ancestors(search_keys, 0, parent_map,
1066
 
                                            missing_keys)
1067
 
        # passing it back in, we are sure it is 'missing'
1068
 
        self.assertEqual({key1: (key2,), key2: (key3,)}, parent_map)
1069
 
        self.assertEqual(set([key3]), missing_keys)
1070
 
        self.assertEqual(set([]), search_keys)
1071
 
 
1072
 
    def test__find_ancestors_dont_search_known(self):
1073
 
        key1 = ('key-1',)
1074
 
        key2 = ('key-2',)
1075
 
        key3 = ('key-3',)
1076
 
        index = self.make_index(ref_lists=1, key_elements=1, nodes=[
1077
 
            (key1, 'value', ([key2],)),
1078
 
            (key2, 'value', ([key3],)),
1079
 
            (key3, 'value', ([],)),
1080
 
            ])
1081
 
        # We already know about key2, so we won't try to search for key3
1082
 
        parent_map = {key2: (key3,)}
1083
 
        missing_keys = set()
1084
 
        search_keys = index._find_ancestors([key1], 0, parent_map,
1085
 
                                            missing_keys)
1086
 
        self.assertEqual({key1: (key2,), key2: (key3,)}, parent_map)
1087
 
        self.assertEqual(set(), missing_keys)
1088
 
        self.assertEqual(set(), search_keys)
1089
 
 
1090
 
    def test__find_ancestors_multiple_pages(self):
1091
 
        # We need to use enough keys that we actually cause a split
1092
 
        start_time = 1249671539
1093
 
        email = "joebob@example.com"
1094
 
        nodes = []
1095
 
        ref_lists = ((),)
1096
 
        rev_keys = []
1097
 
        for i in xrange(400):
1098
 
            rev_id = '%s-%s-%s' % (email,
1099
 
                                   osutils.compact_date(start_time + i),
1100
 
                                   osutils.rand_chars(16))
1101
 
            rev_key = (rev_id,)
1102
 
            nodes.append((rev_key, 'value', ref_lists))
1103
 
            # We have a ref 'list' of length 1, with a list of parents, with 1
1104
 
            # parent which is a key
1105
 
            ref_lists = ((rev_key,),)
1106
 
            rev_keys.append(rev_key)
1107
 
        index = self.make_index(ref_lists=1, key_elements=1, nodes=nodes)
1108
 
        self.assertEqual(400, index.key_count())
1109
 
        self.assertEqual(3, len(index._row_offsets))
1110
 
        nodes = dict(index._read_nodes([1, 2]))
1111
 
        l1 = nodes[1]
1112
 
        l2 = nodes[2]
1113
 
        min_l2_key = l2.min_key
1114
 
        max_l1_key = l1.max_key
1115
 
        self.assertTrue(max_l1_key < min_l2_key)
1116
 
        parents_min_l2_key = l2[min_l2_key][1][0]
1117
 
        self.assertEqual((l1.max_key,), parents_min_l2_key)
1118
 
        # Now, whatever key we select that would fall on the second page,
1119
 
        # should give us all the parents until the page break
1120
 
        key_idx = rev_keys.index(min_l2_key)
1121
 
        next_key = rev_keys[key_idx+1]
1122
 
        # So now when we get the parent map, we should get the key we are
1123
 
        # looking for, min_l2_key, and then a reference to go look for the
1124
 
        # parent of that key
1125
 
        parent_map = {}
1126
 
        missing_keys = set()
1127
 
        search_keys = index._find_ancestors([next_key], 0, parent_map,
1128
 
                                            missing_keys)
1129
 
        self.assertEqual([min_l2_key, next_key], sorted(parent_map))
1130
 
        self.assertEqual(set(), missing_keys)
1131
 
        self.assertEqual(set([max_l1_key]), search_keys)
1132
 
        parent_map = {}
1133
 
        search_keys = index._find_ancestors([max_l1_key], 0, parent_map,
1134
 
                                            missing_keys)
1135
 
        self.assertEqual(l1.all_keys(), sorted(parent_map))
1136
 
        self.assertEqual(set(), missing_keys)
1137
 
        self.assertEqual(set(), search_keys)
1138
 
 
1139
 
    def test__find_ancestors_empty_index(self):
1140
 
        index = self.make_index(ref_lists=1, key_elements=1, nodes=[])
1141
 
        parent_map = {}
1142
 
        missing_keys = set()
1143
 
        search_keys = index._find_ancestors([('one',), ('two',)], 0, parent_map,
1144
 
                                            missing_keys)
1145
 
        self.assertEqual(set(), search_keys)
1146
 
        self.assertEqual({}, parent_map)
1147
 
        self.assertEqual(set([('one',), ('two',)]), missing_keys)
1148
 
 
1149
 
    def test_supports_unlimited_cache(self):
1150
 
        builder = btree_index.BTreeBuilder(reference_lists=0, key_elements=1)
1151
 
        # We need enough nodes to cause a page split (so we have both an
1152
 
        # internal node and a couple leaf nodes. 500 seems to be enough.)
1153
 
        nodes = self.make_nodes(500, 1, 0)
1154
 
        for node in nodes:
1155
 
            builder.add_node(*node)
1156
 
        stream = builder.finish()
1157
 
        trans = self.get_transport()
1158
 
        size = trans.put_file('index', stream)
1159
 
        index = btree_index.BTreeGraphIndex(trans, 'index', size)
1160
 
        self.assertEqual(500, index.key_count())
1161
 
        # We have an internal node
1162
 
        self.assertEqual(2, len(index._row_lengths))
1163
 
        # We have at least 2 leaf nodes
1164
 
        self.assertTrue(index._row_lengths[-1] >= 2)
1165
 
        self.assertIsInstance(index._leaf_node_cache, lru_cache.LRUCache)
1166
 
        self.assertEqual(btree_index._NODE_CACHE_SIZE,
1167
 
                         index._leaf_node_cache._max_cache)
1168
 
        self.assertIsInstance(index._internal_node_cache, fifo_cache.FIFOCache)
1169
 
        self.assertEqual(100, index._internal_node_cache._max_cache)
1170
 
        # No change if unlimited_cache=False is passed
1171
 
        index = btree_index.BTreeGraphIndex(trans, 'index', size,
1172
 
                                            unlimited_cache=False)
1173
 
        self.assertIsInstance(index._leaf_node_cache, lru_cache.LRUCache)
1174
 
        self.assertEqual(btree_index._NODE_CACHE_SIZE,
1175
 
                         index._leaf_node_cache._max_cache)
1176
 
        self.assertIsInstance(index._internal_node_cache, fifo_cache.FIFOCache)
1177
 
        self.assertEqual(100, index._internal_node_cache._max_cache)
1178
 
        index = btree_index.BTreeGraphIndex(trans, 'index', size,
1179
 
                                            unlimited_cache=True)
1180
 
        self.assertIsInstance(index._leaf_node_cache, dict)
1181
 
        self.assertIs(type(index._internal_node_cache), dict)
1182
 
        # Exercise the lookup code
1183
 
        entries = set(index.iter_entries([n[0] for n in nodes]))
1184
 
        self.assertEqual(500, len(entries))
1185
 
 
1186
836
 
1187
837
class TestBTreeNodes(BTreeTestCase):
1188
838
 
1189
 
    scenarios = btreeparser_scenarios()
 
839
    def restore_parser(self):
 
840
        btree_index._btree_serializer = self.saved_parser
1190
841
 
1191
842
    def setUp(self):
1192
843
        BTreeTestCase.setUp(self)
1193
 
        self.overrideAttr(btree_index, '_btree_serializer', self.parse_btree)
 
844
        self.saved_parser = btree_index._btree_serializer
 
845
        self.addCleanup(self.restore_parser)
 
846
        btree_index._btree_serializer = self.parse_btree
1194
847
 
1195
848
    def test_LeafNode_1_0(self):
1196
849
        node_bytes = ("type=leaf\n"
1208
861
            ("2222222222222222222222222222222222222222",): ("value:2", ()),
1209
862
            ("3333333333333333333333333333333333333333",): ("value:3", ()),
1210
863
            ("4444444444444444444444444444444444444444",): ("value:4", ()),
1211
 
            }, dict(node.all_items()))
 
864
            }, node.keys)
1212
865
 
1213
866
    def test_LeafNode_2_2(self):
1214
867
        node_bytes = ("type=leaf\n"
1228
881
            ('11', '33'): ('value:3',
1229
882
                ((('11', 'ref22'),), (('11', 'ref22'), ('11', 'ref22')))),
1230
883
            ('11', '44'): ('value:4', ((), (('11', 'ref00'),)))
1231
 
            }, dict(node.all_items()))
 
884
            }, node.keys)
1232
885
 
1233
886
    def test_InternalNode_1(self):
1234
887
        node_bytes = ("type=internal\n"
1269
922
            ('11', '33'): ('value:3',
1270
923
                ((('11', 'ref22'),), (('11', 'ref22'), ('11', 'ref22')))),
1271
924
            ('11', '44'): ('value:4', ((), (('11', 'ref00'),)))
1272
 
            }, dict(node.all_items()))
 
925
            }, node.keys)
1273
926
 
1274
927
    def assertFlattened(self, expected, key, value, refs):
1275
928
        flat_key, flat_line = self.parse_btree._flatten_node(
1307
960
    def test_exists(self):
1308
961
        # This is just to let the user know if they don't have the feature
1309
962
        # available
1310
 
        self.requireFeature(compiled_btreeparser_feature)
 
963
        self.requireFeature(CompiledBtreeParserFeature)
1311
964
 
1312
965
 
1313
966
class TestMultiBisectRight(tests.TestCase):
1343
996
                                     (4, ['g', 'h'])],
1344
997
                                    ['a', 'b', 'd', 'e', 'g', 'h'],
1345
998
                                    ['c', 'd', 'f', 'g'])
1346
 
 
1347
 
 
1348
 
class TestExpandOffsets(tests.TestCase):
1349
 
 
1350
 
    def make_index(self, size, recommended_pages=None):
1351
 
        """Make an index with a generic size.
1352
 
 
1353
 
        This doesn't actually create anything on disk, it just primes a
1354
 
        BTreeGraphIndex with the recommended information.
1355
 
        """
1356
 
        index = btree_index.BTreeGraphIndex(
1357
 
            transport.get_transport('memory:///'), 'test-index', size=size)
1358
 
        if recommended_pages is not None:
1359
 
            index._recommended_pages = recommended_pages
1360
 
        return index
1361
 
 
1362
 
    def set_cached_offsets(self, index, cached_offsets):
1363
 
        """Monkeypatch to give a canned answer for _get_offsets_for...()."""
1364
 
        def _get_offsets_to_cached_pages():
1365
 
            cached = set(cached_offsets)
1366
 
            return cached
1367
 
        index._get_offsets_to_cached_pages = _get_offsets_to_cached_pages
1368
 
 
1369
 
    def prepare_index(self, index, node_ref_lists, key_length, key_count,
1370
 
                      row_lengths, cached_offsets):
1371
 
        """Setup the BTreeGraphIndex with some pre-canned information."""
1372
 
        index.node_ref_lists = node_ref_lists
1373
 
        index._key_length = key_length
1374
 
        index._key_count = key_count
1375
 
        index._row_lengths = row_lengths
1376
 
        index._compute_row_offsets()
1377
 
        index._root_node = btree_index._InternalNode('internal\noffset=0\n')
1378
 
        self.set_cached_offsets(index, cached_offsets)
1379
 
 
1380
 
    def make_100_node_index(self):
1381
 
        index = self.make_index(4096*100, 6)
1382
 
        # Consider we've already made a single request at the middle
1383
 
        self.prepare_index(index, node_ref_lists=0, key_length=1,
1384
 
                           key_count=1000, row_lengths=[1, 99],
1385
 
                           cached_offsets=[0, 50])
1386
 
        return index
1387
 
 
1388
 
    def make_1000_node_index(self):
1389
 
        index = self.make_index(4096*1000, 6)
1390
 
        # Pretend we've already made a single request in the middle
1391
 
        self.prepare_index(index, node_ref_lists=0, key_length=1,
1392
 
                           key_count=90000, row_lengths=[1, 9, 990],
1393
 
                           cached_offsets=[0, 5, 500])
1394
 
        return index
1395
 
 
1396
 
    def assertNumPages(self, expected_pages, index, size):
1397
 
        index._size = size
1398
 
        self.assertEqual(expected_pages, index._compute_total_pages_in_index())
1399
 
 
1400
 
    def assertExpandOffsets(self, expected, index, offsets):
1401
 
        self.assertEqual(expected, index._expand_offsets(offsets),
1402
 
                         'We did not get the expected value after expanding'
1403
 
                         ' %s' % (offsets,))
1404
 
 
1405
 
    def test_default_recommended_pages(self):
1406
 
        index = self.make_index(None)
1407
 
        # local transport recommends 4096 byte reads, which is 1 page
1408
 
        self.assertEqual(1, index._recommended_pages)
1409
 
 
1410
 
    def test__compute_total_pages_in_index(self):
1411
 
        index = self.make_index(None)
1412
 
        self.assertNumPages(1, index, 1024)
1413
 
        self.assertNumPages(1, index, 4095)
1414
 
        self.assertNumPages(1, index, 4096)
1415
 
        self.assertNumPages(2, index, 4097)
1416
 
        self.assertNumPages(2, index, 8192)
1417
 
        self.assertNumPages(76, index, 4096*75 + 10)
1418
 
 
1419
 
    def test__find_layer_start_and_stop(self):
1420
 
        index = self.make_1000_node_index()
1421
 
        self.assertEqual((0, 1), index._find_layer_first_and_end(0))
1422
 
        self.assertEqual((1, 10), index._find_layer_first_and_end(1))
1423
 
        self.assertEqual((1, 10), index._find_layer_first_and_end(9))
1424
 
        self.assertEqual((10, 1000), index._find_layer_first_and_end(10))
1425
 
        self.assertEqual((10, 1000), index._find_layer_first_and_end(99))
1426
 
        self.assertEqual((10, 1000), index._find_layer_first_and_end(999))
1427
 
 
1428
 
    def test_unknown_size(self):
1429
 
        # We should not expand if we don't know the file size
1430
 
        index = self.make_index(None, 10)
1431
 
        self.assertExpandOffsets([0], index, [0])
1432
 
        self.assertExpandOffsets([1, 4, 9], index, [1, 4, 9])
1433
 
 
1434
 
    def test_more_than_recommended(self):
1435
 
        index = self.make_index(4096*100, 2)
1436
 
        self.assertExpandOffsets([1, 10], index, [1, 10])
1437
 
        self.assertExpandOffsets([1, 10, 20], index, [1, 10, 20])
1438
 
 
1439
 
    def test_read_all_from_root(self):
1440
 
        index = self.make_index(4096*10, 20)
1441
 
        self.assertExpandOffsets(range(10), index, [0])
1442
 
 
1443
 
    def test_read_all_when_cached(self):
1444
 
        # We've read enough that we can grab all the rest in a single request
1445
 
        index = self.make_index(4096*10, 5)
1446
 
        self.prepare_index(index, node_ref_lists=0, key_length=1,
1447
 
                           key_count=1000, row_lengths=[1, 9],
1448
 
                           cached_offsets=[0, 1, 2, 5, 6])
1449
 
        # It should fill the remaining nodes, regardless of the one requested
1450
 
        self.assertExpandOffsets([3, 4, 7, 8, 9], index, [3])
1451
 
        self.assertExpandOffsets([3, 4, 7, 8, 9], index, [8])
1452
 
        self.assertExpandOffsets([3, 4, 7, 8, 9], index, [9])
1453
 
 
1454
 
    def test_no_root_node(self):
1455
 
        index = self.make_index(4096*10, 5)
1456
 
        self.assertExpandOffsets([0], index, [0])
1457
 
 
1458
 
    def test_include_neighbors(self):
1459
 
        index = self.make_100_node_index()
1460
 
        # We expand in both directions, until we have at least 'recommended'
1461
 
        # pages
1462
 
        self.assertExpandOffsets([9, 10, 11, 12, 13, 14, 15], index, [12])
1463
 
        self.assertExpandOffsets([88, 89, 90, 91, 92, 93, 94], index, [91])
1464
 
        # If we hit an 'edge' we continue in the other direction
1465
 
        self.assertExpandOffsets([1, 2, 3, 4, 5, 6], index, [2])
1466
 
        self.assertExpandOffsets([94, 95, 96, 97, 98, 99], index, [98])
1467
 
 
1468
 
        # Requesting many nodes will expand all locations equally
1469
 
        self.assertExpandOffsets([1, 2, 3, 80, 81, 82], index, [2, 81])
1470
 
        self.assertExpandOffsets([1, 2, 3, 9, 10, 11, 80, 81, 82], index,
1471
 
                               [2, 10, 81])
1472
 
 
1473
 
    def test_stop_at_cached(self):
1474
 
        index = self.make_100_node_index()
1475
 
        self.set_cached_offsets(index, [0, 10, 19])
1476
 
        self.assertExpandOffsets([11, 12, 13, 14, 15, 16], index, [11])
1477
 
        self.assertExpandOffsets([11, 12, 13, 14, 15, 16], index, [12])
1478
 
        self.assertExpandOffsets([12, 13, 14, 15, 16, 17, 18], index, [15])
1479
 
        self.assertExpandOffsets([13, 14, 15, 16, 17, 18], index, [16])
1480
 
        self.assertExpandOffsets([13, 14, 15, 16, 17, 18], index, [17])
1481
 
        self.assertExpandOffsets([13, 14, 15, 16, 17, 18], index, [18])
1482
 
 
1483
 
    def test_cannot_fully_expand(self):
1484
 
        index = self.make_100_node_index()
1485
 
        self.set_cached_offsets(index, [0, 10, 12])
1486
 
        # We don't go into an endless loop if we are bound by cached nodes
1487
 
        self.assertExpandOffsets([11], index, [11])
1488
 
 
1489
 
    def test_overlap(self):
1490
 
        index = self.make_100_node_index()
1491
 
        self.assertExpandOffsets([10, 11, 12, 13, 14, 15], index, [12, 13])
1492
 
        self.assertExpandOffsets([10, 11, 12, 13, 14, 15], index, [11, 14])
1493
 
 
1494
 
    def test_stay_within_layer(self):
1495
 
        index = self.make_1000_node_index()
1496
 
        # When expanding a request, we won't read nodes from the next layer
1497
 
        self.assertExpandOffsets([1, 2, 3, 4], index, [2])
1498
 
        self.assertExpandOffsets([6, 7, 8, 9], index, [6])
1499
 
        self.assertExpandOffsets([6, 7, 8, 9], index, [9])
1500
 
        self.assertExpandOffsets([10, 11, 12, 13, 14, 15], index, [10])
1501
 
        self.assertExpandOffsets([10, 11, 12, 13, 14, 15, 16], index, [13])
1502
 
 
1503
 
        self.set_cached_offsets(index, [0, 4, 12])
1504
 
        self.assertExpandOffsets([5, 6, 7, 8, 9], index, [7])
1505
 
        self.assertExpandOffsets([10, 11], index, [11])
1506
 
 
1507
 
    def test_small_requests_unexpanded(self):
1508
 
        index = self.make_100_node_index()
1509
 
        self.set_cached_offsets(index, [0])
1510
 
        self.assertExpandOffsets([1], index, [1])
1511
 
        self.assertExpandOffsets([50], index, [50])
1512
 
        # If we request more than one node, then we'll expand
1513
 
        self.assertExpandOffsets([49, 50, 51, 59, 60, 61], index, [50, 60])
1514
 
 
1515
 
        # The first pass does not expand
1516
 
        index = self.make_1000_node_index()
1517
 
        self.set_cached_offsets(index, [0])
1518
 
        self.assertExpandOffsets([1], index, [1])
1519
 
        self.set_cached_offsets(index, [0, 1])
1520
 
        self.assertExpandOffsets([100], index, [100])
1521
 
        self.set_cached_offsets(index, [0, 1, 100])
1522
 
        # But after the first depth, we will expand
1523
 
        self.assertExpandOffsets([2, 3, 4, 5, 6, 7], index, [2])
1524
 
        self.assertExpandOffsets([2, 3, 4, 5, 6, 7], index, [4])
1525
 
        self.set_cached_offsets(index, [0, 1, 2, 3, 4, 5, 6, 7, 100])
1526
 
        self.assertExpandOffsets([102, 103, 104, 105, 106, 107, 108], index,
1527
 
                                 [105])