1
# Copyright (C) 2008, 2009 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18
"""Tests for btree indices."""
31
from bzrlib.tests import (
32
TestCaseWithTransport,
35
split_suite_by_condition,
37
from bzrlib.transport import get_transport
40
def load_tests(standard_tests, module, loader):
41
# parameterise the TestBTreeNodes tests
42
node_tests, others = split_suite_by_condition(standard_tests,
43
condition_isinstance(TestBTreeNodes))
44
import bzrlib._btree_serializer_py as py_module
45
scenarios = [('python', {'parse_btree': py_module})]
46
if compiled_btreeparser_feature.available():
47
scenarios.append(('C', {'parse_btree':
48
compiled_btreeparser_feature.module}))
49
return multiply_tests(node_tests, scenarios, others)
52
compiled_btreeparser_feature = tests.ModuleAvailableFeature(
53
'bzrlib._btree_serializer_pyx')
56
class BTreeTestCase(TestCaseWithTransport):
57
# test names here are suffixed by the key length and reference list count
61
TestCaseWithTransport.setUp(self)
62
self._original_header = btree_index._RESERVED_HEADER_BYTES
64
btree_index._RESERVED_HEADER_BYTES = self._original_header
65
self.addCleanup(restore)
66
btree_index._RESERVED_HEADER_BYTES = 100
68
def make_nodes(self, count, key_elements, reference_lists):
69
"""Generate count*key_elements sample nodes."""
71
for prefix_pos in range(key_elements):
73
prefix = (str(prefix_pos) * 40,)
76
for pos in xrange(count):
77
# TODO: This creates odd keys. When count == 100,000, it
78
# creates a 240 byte key
79
key = prefix + (str(pos) * 40,)
80
value = "value:%s" % pos
82
# generate some references
84
for list_pos in range(reference_lists):
85
# as many keys in each list as its index + the key depth
86
# mod 2 - this generates both 0 length lists and
87
# ones slightly longer than the number of lists.
88
# It also ensures we have non homogeneous lists.
90
for ref_pos in range(list_pos + pos % 2):
92
# refer to a nearby key
93
refs[-1].append(prefix + ("ref" + str(pos - 1) * 40,))
95
# serial of this ref in the ref list
96
refs[-1].append(prefix + ("ref" + str(ref_pos) * 40,))
97
refs[-1] = tuple(refs[-1])
101
keys.append((key, value, refs))
104
def shrink_page_size(self):
105
"""Shrink the default page size so that less fits in a page."""
106
old_page_size = btree_index._PAGE_SIZE
108
btree_index._PAGE_SIZE = old_page_size
109
self.addCleanup(cleanup)
110
btree_index._PAGE_SIZE = 2048
113
class TestBTreeBuilder(BTreeTestCase):
115
def test_clear_cache(self):
116
builder = btree_index.BTreeBuilder(reference_lists=0, key_elements=1)
117
# This is a no-op, but we need the api to be consistent with other
118
# BTreeGraphIndex apis.
119
builder.clear_cache()
121
def test_empty_1_0(self):
122
builder = btree_index.BTreeBuilder(key_elements=1, reference_lists=0)
123
# NamedTemporaryFile dies on builder.finish().read(). weird.
124
temp_file = builder.finish()
125
content = temp_file.read()
128
"B+Tree Graph Index 2\nnode_ref_lists=0\nkey_elements=1\nlen=0\n"
132
def test_empty_2_1(self):
133
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=1)
134
# NamedTemporaryFile dies on builder.finish().read(). weird.
135
temp_file = builder.finish()
136
content = temp_file.read()
139
"B+Tree Graph Index 2\nnode_ref_lists=1\nkey_elements=2\nlen=0\n"
143
def test_root_leaf_1_0(self):
144
builder = btree_index.BTreeBuilder(key_elements=1, reference_lists=0)
145
nodes = self.make_nodes(5, 1, 0)
147
builder.add_node(*node)
148
# NamedTemporaryFile dies on builder.finish().read(). weird.
149
temp_file = builder.finish()
150
content = temp_file.read()
152
self.assertEqual(131, len(content))
154
"B+Tree Graph Index 2\nnode_ref_lists=0\nkey_elements=1\nlen=5\n"
157
node_content = content[73:]
158
node_bytes = zlib.decompress(node_content)
159
expected_node = ("type=leaf\n"
160
"0000000000000000000000000000000000000000\x00\x00value:0\n"
161
"1111111111111111111111111111111111111111\x00\x00value:1\n"
162
"2222222222222222222222222222222222222222\x00\x00value:2\n"
163
"3333333333333333333333333333333333333333\x00\x00value:3\n"
164
"4444444444444444444444444444444444444444\x00\x00value:4\n")
165
self.assertEqual(expected_node, node_bytes)
167
def test_root_leaf_2_2(self):
168
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
169
nodes = self.make_nodes(5, 2, 2)
171
builder.add_node(*node)
172
# NamedTemporaryFile dies on builder.finish().read(). weird.
173
temp_file = builder.finish()
174
content = temp_file.read()
176
self.assertEqual(238, len(content))
178
"B+Tree Graph Index 2\nnode_ref_lists=2\nkey_elements=2\nlen=10\n"
181
node_content = content[74:]
182
node_bytes = zlib.decompress(node_content)
185
"0000000000000000000000000000000000000000\x000000000000000000000000000000000000000000\x00\t0000000000000000000000000000000000000000\x00ref0000000000000000000000000000000000000000\x00value:0\n"
186
"0000000000000000000000000000000000000000\x001111111111111111111111111111111111111111\x000000000000000000000000000000000000000000\x00ref0000000000000000000000000000000000000000\t0000000000000000000000000000000000000000\x00ref0000000000000000000000000000000000000000\r0000000000000000000000000000000000000000\x00ref0000000000000000000000000000000000000000\x00value:1\n"
187
"0000000000000000000000000000000000000000\x002222222222222222222222222222222222222222\x00\t0000000000000000000000000000000000000000\x00ref0000000000000000000000000000000000000000\x00value:2\n"
188
"0000000000000000000000000000000000000000\x003333333333333333333333333333333333333333\x000000000000000000000000000000000000000000\x00ref2222222222222222222222222222222222222222\t0000000000000000000000000000000000000000\x00ref2222222222222222222222222222222222222222\r0000000000000000000000000000000000000000\x00ref2222222222222222222222222222222222222222\x00value:3\n"
189
"0000000000000000000000000000000000000000\x004444444444444444444444444444444444444444\x00\t0000000000000000000000000000000000000000\x00ref0000000000000000000000000000000000000000\x00value:4\n"
190
"1111111111111111111111111111111111111111\x000000000000000000000000000000000000000000\x00\t1111111111111111111111111111111111111111\x00ref0000000000000000000000000000000000000000\x00value:0\n"
191
"1111111111111111111111111111111111111111\x001111111111111111111111111111111111111111\x001111111111111111111111111111111111111111\x00ref0000000000000000000000000000000000000000\t1111111111111111111111111111111111111111\x00ref0000000000000000000000000000000000000000\r1111111111111111111111111111111111111111\x00ref0000000000000000000000000000000000000000\x00value:1\n"
192
"1111111111111111111111111111111111111111\x002222222222222222222222222222222222222222\x00\t1111111111111111111111111111111111111111\x00ref0000000000000000000000000000000000000000\x00value:2\n"
193
"1111111111111111111111111111111111111111\x003333333333333333333333333333333333333333\x001111111111111111111111111111111111111111\x00ref2222222222222222222222222222222222222222\t1111111111111111111111111111111111111111\x00ref2222222222222222222222222222222222222222\r1111111111111111111111111111111111111111\x00ref2222222222222222222222222222222222222222\x00value:3\n"
194
"1111111111111111111111111111111111111111\x004444444444444444444444444444444444444444\x00\t1111111111111111111111111111111111111111\x00ref0000000000000000000000000000000000000000\x00value:4\n"
197
self.assertEqual(expected_node, node_bytes)
199
def test_2_leaves_1_0(self):
200
builder = btree_index.BTreeBuilder(key_elements=1, reference_lists=0)
201
nodes = self.make_nodes(400, 1, 0)
203
builder.add_node(*node)
204
# NamedTemporaryFile dies on builder.finish().read(). weird.
205
temp_file = builder.finish()
206
content = temp_file.read()
208
self.assertEqual(9283, len(content))
210
"B+Tree Graph Index 2\nnode_ref_lists=0\nkey_elements=1\nlen=400\n"
213
root = content[77:4096]
214
leaf1 = content[4096:8192]
215
leaf2 = content[8192:]
216
root_bytes = zlib.decompress(root)
220
) + ("307" * 40) + "\n"
221
self.assertEqual(expected_root, root_bytes)
222
# We already know serialisation works for leaves, check key selection:
223
leaf1_bytes = zlib.decompress(leaf1)
224
sorted_node_keys = sorted(node[0] for node in nodes)
225
node = btree_index._LeafNode(leaf1_bytes, 1, 0)
226
self.assertEqual(231, len(node.keys))
227
self.assertEqual(sorted_node_keys[:231], sorted(node.keys))
228
leaf2_bytes = zlib.decompress(leaf2)
229
node = btree_index._LeafNode(leaf2_bytes, 1, 0)
230
self.assertEqual(400 - 231, len(node.keys))
231
self.assertEqual(sorted_node_keys[231:], sorted(node.keys))
233
def test_last_page_rounded_1_layer(self):
234
builder = btree_index.BTreeBuilder(key_elements=1, reference_lists=0)
235
nodes = self.make_nodes(10, 1, 0)
237
builder.add_node(*node)
238
# NamedTemporaryFile dies on builder.finish().read(). weird.
239
temp_file = builder.finish()
240
content = temp_file.read()
242
self.assertEqual(155, len(content))
244
"B+Tree Graph Index 2\nnode_ref_lists=0\nkey_elements=1\nlen=10\n"
247
# Check thelast page is well formed
249
leaf2_bytes = zlib.decompress(leaf2)
250
node = btree_index._LeafNode(leaf2_bytes, 1, 0)
251
self.assertEqual(10, len(node.keys))
252
sorted_node_keys = sorted(node[0] for node in nodes)
253
self.assertEqual(sorted_node_keys, sorted(node.keys))
255
def test_last_page_not_rounded_2_layer(self):
256
builder = btree_index.BTreeBuilder(key_elements=1, reference_lists=0)
257
nodes = self.make_nodes(400, 1, 0)
259
builder.add_node(*node)
260
# NamedTemporaryFile dies on builder.finish().read(). weird.
261
temp_file = builder.finish()
262
content = temp_file.read()
264
self.assertEqual(9283, len(content))
266
"B+Tree Graph Index 2\nnode_ref_lists=0\nkey_elements=1\nlen=400\n"
269
# Check the last page is well formed
270
leaf2 = content[8192:]
271
leaf2_bytes = zlib.decompress(leaf2)
272
node = btree_index._LeafNode(leaf2_bytes, 1, 0)
273
self.assertEqual(400 - 231, len(node.keys))
274
sorted_node_keys = sorted(node[0] for node in nodes)
275
self.assertEqual(sorted_node_keys[231:], sorted(node.keys))
277
def test_three_level_tree_details(self):
278
# The left most pointer in the second internal node in a row should
279
# pointer to the second node that the internal node is for, _not_
280
# the first, otherwise the first node overlaps with the last node of
281
# the prior internal node on that row.
282
self.shrink_page_size()
283
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
284
# 40K nodes is enough to create a two internal nodes on the second
285
# level, with a 2K page size
286
nodes = self.make_nodes(20000, 2, 2)
289
builder.add_node(*node)
290
transport = get_transport('trace+' + self.get_url(''))
291
size = transport.put_file('index', self.time(builder.finish))
293
index = btree_index.BTreeGraphIndex(transport, 'index', size)
294
# Seed the metadata, we're using internal calls now.
296
self.assertEqual(3, len(index._row_lengths),
297
"Not enough rows: %r" % index._row_lengths)
298
self.assertEqual(4, len(index._row_offsets))
299
self.assertEqual(sum(index._row_lengths), index._row_offsets[-1])
300
internal_nodes = index._get_internal_nodes([0, 1, 2])
301
root_node = internal_nodes[0]
302
internal_node1 = internal_nodes[1]
303
internal_node2 = internal_nodes[2]
304
# The left most node node2 points at should be one after the right most
305
# node pointed at by node1.
306
self.assertEqual(internal_node2.offset, 1 + len(internal_node1.keys))
307
# The left most key of the second node pointed at by internal_node2
308
# should be its first key. We can check this by looking for its first key
309
# in the second node it points at
310
pos = index._row_offsets[2] + internal_node2.offset + 1
311
leaf = index._get_leaf_nodes([pos])[pos]
312
self.assertTrue(internal_node2.keys[0] in leaf.keys)
314
def test_2_leaves_2_2(self):
315
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
316
nodes = self.make_nodes(100, 2, 2)
318
builder.add_node(*node)
319
# NamedTemporaryFile dies on builder.finish().read(). weird.
320
temp_file = builder.finish()
321
content = temp_file.read()
323
self.assertEqual(12643, len(content))
325
"B+Tree Graph Index 2\nnode_ref_lists=2\nkey_elements=2\nlen=200\n"
328
root = content[77:4096]
329
leaf1 = content[4096:8192]
330
leaf2 = content[8192:12288]
331
leaf3 = content[12288:]
332
root_bytes = zlib.decompress(root)
336
+ ("0" * 40) + "\x00" + ("91" * 40) + "\n"
337
+ ("1" * 40) + "\x00" + ("81" * 40) + "\n"
339
self.assertEqual(expected_root, root_bytes)
340
# We assume the other leaf nodes have been written correctly - layering
343
def test_spill_index_stress_1_1(self):
344
builder = btree_index.BTreeBuilder(key_elements=1, spill_at=2)
345
nodes = [node[0:2] for node in self.make_nodes(16, 1, 0)]
346
builder.add_node(*nodes[0])
347
# Test the parts of the index that take up memory are doing so
349
self.assertEqual(1, len(builder._nodes))
350
self.assertIs(None, builder._nodes_by_key)
351
builder.add_node(*nodes[1])
352
self.assertEqual(0, len(builder._nodes))
353
self.assertIs(None, builder._nodes_by_key)
354
self.assertEqual(1, len(builder._backing_indices))
355
self.assertEqual(2, builder._backing_indices[0].key_count())
357
builder.add_node(*nodes[2])
358
self.assertEqual(1, len(builder._nodes))
359
self.assertIs(None, builder._nodes_by_key)
360
# And spills to a second backing index combing all
361
builder.add_node(*nodes[3])
362
self.assertEqual(0, len(builder._nodes))
363
self.assertIs(None, builder._nodes_by_key)
364
self.assertEqual(2, len(builder._backing_indices))
365
self.assertEqual(None, builder._backing_indices[0])
366
self.assertEqual(4, builder._backing_indices[1].key_count())
367
# The next spills to the 2-len slot
368
builder.add_node(*nodes[4])
369
builder.add_node(*nodes[5])
370
self.assertEqual(0, len(builder._nodes))
371
self.assertIs(None, builder._nodes_by_key)
372
self.assertEqual(2, len(builder._backing_indices))
373
self.assertEqual(2, builder._backing_indices[0].key_count())
374
self.assertEqual(4, builder._backing_indices[1].key_count())
375
# Next spill combines
376
builder.add_node(*nodes[6])
377
builder.add_node(*nodes[7])
378
self.assertEqual(3, len(builder._backing_indices))
379
self.assertEqual(None, builder._backing_indices[0])
380
self.assertEqual(None, builder._backing_indices[1])
381
self.assertEqual(8, builder._backing_indices[2].key_count())
382
# And so forth - counting up in binary.
383
builder.add_node(*nodes[8])
384
builder.add_node(*nodes[9])
385
self.assertEqual(3, len(builder._backing_indices))
386
self.assertEqual(2, builder._backing_indices[0].key_count())
387
self.assertEqual(None, builder._backing_indices[1])
388
self.assertEqual(8, builder._backing_indices[2].key_count())
389
builder.add_node(*nodes[10])
390
builder.add_node(*nodes[11])
391
self.assertEqual(3, len(builder._backing_indices))
392
self.assertEqual(None, builder._backing_indices[0])
393
self.assertEqual(4, builder._backing_indices[1].key_count())
394
self.assertEqual(8, builder._backing_indices[2].key_count())
395
builder.add_node(*nodes[12])
396
# Test that memory and disk are both used for query methods; and that
397
# None is skipped over happily.
398
self.assertEqual([(builder,) + node for node in sorted(nodes[:13])],
399
list(builder.iter_all_entries()))
400
# Two nodes - one memory one disk
401
self.assertEqual(set([(builder,) + node for node in nodes[11:13]]),
402
set(builder.iter_entries([nodes[12][0], nodes[11][0]])))
403
self.assertEqual(13, builder.key_count())
404
self.assertEqual(set([(builder,) + node for node in nodes[11:13]]),
405
set(builder.iter_entries_prefix([nodes[12][0], nodes[11][0]])))
406
builder.add_node(*nodes[13])
407
self.assertEqual(3, len(builder._backing_indices))
408
self.assertEqual(2, builder._backing_indices[0].key_count())
409
self.assertEqual(4, builder._backing_indices[1].key_count())
410
self.assertEqual(8, builder._backing_indices[2].key_count())
411
builder.add_node(*nodes[14])
412
builder.add_node(*nodes[15])
413
self.assertEqual(4, len(builder._backing_indices))
414
self.assertEqual(None, builder._backing_indices[0])
415
self.assertEqual(None, builder._backing_indices[1])
416
self.assertEqual(None, builder._backing_indices[2])
417
self.assertEqual(16, builder._backing_indices[3].key_count())
418
# Now finish, and check we got a correctly ordered tree
419
transport = self.get_transport('')
420
size = transport.put_file('index', builder.finish())
421
index = btree_index.BTreeGraphIndex(transport, 'index', size)
422
nodes = list(index.iter_all_entries())
423
self.assertEqual(sorted(nodes), nodes)
424
self.assertEqual(16, len(nodes))
426
def test_spill_index_stress_1_1_no_combine(self):
427
builder = btree_index.BTreeBuilder(key_elements=1, spill_at=2)
428
builder.set_optimize(for_size=False, combine_backing_indices=False)
429
nodes = [node[0:2] for node in self.make_nodes(16, 1, 0)]
430
builder.add_node(*nodes[0])
431
# Test the parts of the index that take up memory are doing so
433
self.assertEqual(1, len(builder._nodes))
434
self.assertIs(None, builder._nodes_by_key)
435
builder.add_node(*nodes[1])
436
self.assertEqual(0, len(builder._nodes))
437
self.assertIs(None, builder._nodes_by_key)
438
self.assertEqual(1, len(builder._backing_indices))
439
self.assertEqual(2, builder._backing_indices[0].key_count())
441
builder.add_node(*nodes[2])
442
self.assertEqual(1, len(builder._nodes))
443
self.assertIs(None, builder._nodes_by_key)
444
# And spills to a second backing index but doesn't combine
445
builder.add_node(*nodes[3])
446
self.assertEqual(0, len(builder._nodes))
447
self.assertIs(None, builder._nodes_by_key)
448
self.assertEqual(2, len(builder._backing_indices))
449
for backing_index in builder._backing_indices:
450
self.assertEqual(2, backing_index.key_count())
451
# The next spills to the 3rd slot
452
builder.add_node(*nodes[4])
453
builder.add_node(*nodes[5])
454
self.assertEqual(0, len(builder._nodes))
455
self.assertIs(None, builder._nodes_by_key)
456
self.assertEqual(3, len(builder._backing_indices))
457
for backing_index in builder._backing_indices:
458
self.assertEqual(2, backing_index.key_count())
459
# Now spill a few more, and check that we don't combine
460
builder.add_node(*nodes[6])
461
builder.add_node(*nodes[7])
462
builder.add_node(*nodes[8])
463
builder.add_node(*nodes[9])
464
builder.add_node(*nodes[10])
465
builder.add_node(*nodes[11])
466
builder.add_node(*nodes[12])
467
self.assertEqual(6, len(builder._backing_indices))
468
for backing_index in builder._backing_indices:
469
self.assertEqual(2, backing_index.key_count())
470
# Test that memory and disk are both used for query methods; and that
471
# None is skipped over happily.
472
self.assertEqual([(builder,) + node for node in sorted(nodes[:13])],
473
list(builder.iter_all_entries()))
474
# Two nodes - one memory one disk
475
self.assertEqual(set([(builder,) + node for node in nodes[11:13]]),
476
set(builder.iter_entries([nodes[12][0], nodes[11][0]])))
477
self.assertEqual(13, builder.key_count())
478
self.assertEqual(set([(builder,) + node for node in nodes[11:13]]),
479
set(builder.iter_entries_prefix([nodes[12][0], nodes[11][0]])))
480
builder.add_node(*nodes[13])
481
builder.add_node(*nodes[14])
482
builder.add_node(*nodes[15])
483
self.assertEqual(8, len(builder._backing_indices))
484
for backing_index in builder._backing_indices:
485
self.assertEqual(2, backing_index.key_count())
486
# Now finish, and check we got a correctly ordered tree
487
transport = self.get_transport('')
488
size = transport.put_file('index', builder.finish())
489
index = btree_index.BTreeGraphIndex(transport, 'index', size)
490
nodes = list(index.iter_all_entries())
491
self.assertEqual(sorted(nodes), nodes)
492
self.assertEqual(16, len(nodes))
494
def test_set_optimize(self):
495
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
496
builder.set_optimize(for_size=True)
497
self.assertTrue(builder._optimize_for_size)
498
builder.set_optimize(for_size=False)
499
self.assertFalse(builder._optimize_for_size)
500
# test that we can set combine_backing_indices without effecting
503
builder._optimize_for_size = obj
504
builder.set_optimize(combine_backing_indices=False)
505
self.assertFalse(builder._combine_backing_indices)
506
self.assertIs(obj, builder._optimize_for_size)
507
builder.set_optimize(combine_backing_indices=True)
508
self.assertTrue(builder._combine_backing_indices)
509
self.assertIs(obj, builder._optimize_for_size)
511
def test_spill_index_stress_2_2(self):
512
# test that references and longer keys don't confuse things.
513
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2,
515
nodes = self.make_nodes(16, 2, 2)
516
builder.add_node(*nodes[0])
517
# Test the parts of the index that take up memory are doing so
519
self.assertEqual(1, len(builder._nodes))
520
self.assertIs(None, builder._nodes_by_key)
521
builder.add_node(*nodes[1])
522
self.assertEqual(0, len(builder._nodes))
523
self.assertIs(None, builder._nodes_by_key)
524
self.assertEqual(1, len(builder._backing_indices))
525
self.assertEqual(2, builder._backing_indices[0].key_count())
527
old = dict(builder._get_nodes_by_key()) #Build up the nodes by key dict
528
builder.add_node(*nodes[2])
529
self.assertEqual(1, len(builder._nodes))
530
self.assertIsNot(None, builder._nodes_by_key)
531
self.assertNotEqual({}, builder._nodes_by_key)
532
# We should have a new entry
533
self.assertNotEqual(old, builder._nodes_by_key)
534
# And spills to a second backing index combing all
535
builder.add_node(*nodes[3])
536
self.assertEqual(0, len(builder._nodes))
537
self.assertIs(None, builder._nodes_by_key)
538
self.assertEqual(2, len(builder._backing_indices))
539
self.assertEqual(None, builder._backing_indices[0])
540
self.assertEqual(4, builder._backing_indices[1].key_count())
541
# The next spills to the 2-len slot
542
builder.add_node(*nodes[4])
543
builder.add_node(*nodes[5])
544
self.assertEqual(0, len(builder._nodes))
545
self.assertIs(None, builder._nodes_by_key)
546
self.assertEqual(2, len(builder._backing_indices))
547
self.assertEqual(2, builder._backing_indices[0].key_count())
548
self.assertEqual(4, builder._backing_indices[1].key_count())
549
# Next spill combines
550
builder.add_node(*nodes[6])
551
builder.add_node(*nodes[7])
552
self.assertEqual(3, len(builder._backing_indices))
553
self.assertEqual(None, builder._backing_indices[0])
554
self.assertEqual(None, builder._backing_indices[1])
555
self.assertEqual(8, builder._backing_indices[2].key_count())
556
# And so forth - counting up in binary.
557
builder.add_node(*nodes[8])
558
builder.add_node(*nodes[9])
559
self.assertEqual(3, len(builder._backing_indices))
560
self.assertEqual(2, builder._backing_indices[0].key_count())
561
self.assertEqual(None, builder._backing_indices[1])
562
self.assertEqual(8, builder._backing_indices[2].key_count())
563
builder.add_node(*nodes[10])
564
builder.add_node(*nodes[11])
565
self.assertEqual(3, len(builder._backing_indices))
566
self.assertEqual(None, builder._backing_indices[0])
567
self.assertEqual(4, builder._backing_indices[1].key_count())
568
self.assertEqual(8, builder._backing_indices[2].key_count())
569
builder.add_node(*nodes[12])
570
# Test that memory and disk are both used for query methods; and that
571
# None is skipped over happily.
572
self.assertEqual([(builder,) + node for node in sorted(nodes[:13])],
573
list(builder.iter_all_entries()))
574
# Two nodes - one memory one disk
575
self.assertEqual(set([(builder,) + node for node in nodes[11:13]]),
576
set(builder.iter_entries([nodes[12][0], nodes[11][0]])))
577
self.assertEqual(13, builder.key_count())
578
self.assertEqual(set([(builder,) + node for node in nodes[11:13]]),
579
set(builder.iter_entries_prefix([nodes[12][0], nodes[11][0]])))
580
builder.add_node(*nodes[13])
581
self.assertEqual(3, len(builder._backing_indices))
582
self.assertEqual(2, builder._backing_indices[0].key_count())
583
self.assertEqual(4, builder._backing_indices[1].key_count())
584
self.assertEqual(8, builder._backing_indices[2].key_count())
585
builder.add_node(*nodes[14])
586
builder.add_node(*nodes[15])
587
self.assertEqual(4, len(builder._backing_indices))
588
self.assertEqual(None, builder._backing_indices[0])
589
self.assertEqual(None, builder._backing_indices[1])
590
self.assertEqual(None, builder._backing_indices[2])
591
self.assertEqual(16, builder._backing_indices[3].key_count())
592
# Now finish, and check we got a correctly ordered tree
593
transport = self.get_transport('')
594
size = transport.put_file('index', builder.finish())
595
index = btree_index.BTreeGraphIndex(transport, 'index', size)
596
nodes = list(index.iter_all_entries())
597
self.assertEqual(sorted(nodes), nodes)
598
self.assertEqual(16, len(nodes))
600
def test_spill_index_duplicate_key_caught_on_finish(self):
601
builder = btree_index.BTreeBuilder(key_elements=1, spill_at=2)
602
nodes = [node[0:2] for node in self.make_nodes(16, 1, 0)]
603
builder.add_node(*nodes[0])
604
builder.add_node(*nodes[1])
605
builder.add_node(*nodes[0])
606
self.assertRaises(errors.BadIndexDuplicateKey, builder.finish)
609
class TestBTreeIndex(BTreeTestCase):
611
def make_index(self, ref_lists=0, key_elements=1, nodes=[]):
612
builder = btree_index.BTreeBuilder(reference_lists=ref_lists,
613
key_elements=key_elements)
614
for key, value, references in nodes:
615
builder.add_node(key, value, references)
616
stream = builder.finish()
617
trans = get_transport('trace+' + self.get_url())
618
size = trans.put_file('index', stream)
619
return btree_index.BTreeGraphIndex(trans, 'index', size)
621
def test_clear_cache(self):
622
nodes = self.make_nodes(160, 2, 2)
623
index = self.make_index(ref_lists=2, key_elements=2, nodes=nodes)
624
self.assertEqual(1, len(list(index.iter_entries([nodes[30][0]]))))
625
self.assertEqual([1, 4], index._row_lengths)
626
self.assertIsNot(None, index._root_node)
627
internal_node_pre_clear = index._internal_node_cache.keys()
628
self.assertTrue(len(index._leaf_node_cache) > 0)
630
# We don't touch _root_node or _internal_node_cache, both should be
631
# small, and can save a round trip or two
632
self.assertIsNot(None, index._root_node)
633
# NOTE: We don't want to affect the _internal_node_cache, as we expect
634
# it will be small, and if we ever do touch this index again, it
635
# will save round-trips. This assertion isn't very strong,
636
# becuase without a 3-level index, we don't have any internal
638
self.assertEqual(internal_node_pre_clear,
639
index._internal_node_cache.keys())
640
self.assertEqual(0, len(index._leaf_node_cache))
642
def test_trivial_constructor(self):
643
transport = get_transport('trace+' + self.get_url(''))
644
index = btree_index.BTreeGraphIndex(transport, 'index', None)
645
# Checks the page size at load, but that isn't logged yet.
646
self.assertEqual([], transport._activity)
648
def test_with_size_constructor(self):
649
transport = get_transport('trace+' + self.get_url(''))
650
index = btree_index.BTreeGraphIndex(transport, 'index', 1)
651
# Checks the page size at load, but that isn't logged yet.
652
self.assertEqual([], transport._activity)
654
def test_empty_key_count_no_size(self):
655
builder = btree_index.BTreeBuilder(key_elements=1, reference_lists=0)
656
transport = get_transport('trace+' + self.get_url(''))
657
transport.put_file('index', builder.finish())
658
index = btree_index.BTreeGraphIndex(transport, 'index', None)
659
del transport._activity[:]
660
self.assertEqual([], transport._activity)
661
self.assertEqual(0, index.key_count())
662
# The entire index should have been requested (as we generally have the
663
# size available, and doing many small readvs is inappropriate).
664
# We can't tell how much was actually read here, but - check the code.
665
self.assertEqual([('get', 'index')], transport._activity)
667
def test_empty_key_count(self):
668
builder = btree_index.BTreeBuilder(key_elements=1, reference_lists=0)
669
transport = get_transport('trace+' + self.get_url(''))
670
size = transport.put_file('index', builder.finish())
671
self.assertEqual(72, size)
672
index = btree_index.BTreeGraphIndex(transport, 'index', size)
673
del transport._activity[:]
674
self.assertEqual([], transport._activity)
675
self.assertEqual(0, index.key_count())
676
# The entire index should have been read, as 4K > size
677
self.assertEqual([('readv', 'index', [(0, 72)], False, None)],
680
def test_non_empty_key_count_2_2(self):
681
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
682
nodes = self.make_nodes(35, 2, 2)
684
builder.add_node(*node)
685
transport = get_transport('trace+' + self.get_url(''))
686
size = transport.put_file('index', builder.finish())
687
index = btree_index.BTreeGraphIndex(transport, 'index', size)
688
del transport._activity[:]
689
self.assertEqual([], transport._activity)
690
self.assertEqual(70, index.key_count())
691
# The entire index should have been read, as it is one page long.
692
self.assertEqual([('readv', 'index', [(0, size)], False, None)],
694
self.assertEqual(1173, size)
696
def test__read_nodes_no_size_one_page_reads_once(self):
697
self.make_index(nodes=[(('key',), 'value', ())])
698
trans = get_transport('trace+' + self.get_url())
699
index = btree_index.BTreeGraphIndex(trans, 'index', None)
700
del trans._activity[:]
701
nodes = dict(index._read_nodes([0]))
702
self.assertEqual([0], nodes.keys())
704
self.assertEqual([('key',)], node.keys.keys())
705
self.assertEqual([('get', 'index')], trans._activity)
707
def test__read_nodes_no_size_multiple_pages(self):
708
index = self.make_index(2, 2, nodes=self.make_nodes(160, 2, 2))
710
num_pages = index._row_offsets[-1]
711
# Reopen with a traced transport and no size
712
trans = get_transport('trace+' + self.get_url())
713
index = btree_index.BTreeGraphIndex(trans, 'index', None)
714
del trans._activity[:]
715
nodes = dict(index._read_nodes([0]))
716
self.assertEqual(range(num_pages), nodes.keys())
718
def test_2_levels_key_count_2_2(self):
719
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
720
nodes = self.make_nodes(160, 2, 2)
722
builder.add_node(*node)
723
transport = get_transport('trace+' + self.get_url(''))
724
size = transport.put_file('index', builder.finish())
725
self.assertEqual(17692, size)
726
index = btree_index.BTreeGraphIndex(transport, 'index', size)
727
del transport._activity[:]
728
self.assertEqual([], transport._activity)
729
self.assertEqual(320, index.key_count())
730
# The entire index should not have been read.
731
self.assertEqual([('readv', 'index', [(0, 4096)], False, None)],
734
def test_validate_one_page(self):
735
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
736
nodes = self.make_nodes(45, 2, 2)
738
builder.add_node(*node)
739
transport = get_transport('trace+' + self.get_url(''))
740
size = transport.put_file('index', builder.finish())
741
index = btree_index.BTreeGraphIndex(transport, 'index', size)
742
del transport._activity[:]
743
self.assertEqual([], transport._activity)
745
# The entire index should have been read linearly.
746
self.assertEqual([('readv', 'index', [(0, size)], False, None)],
748
self.assertEqual(1488, size)
750
def test_validate_two_pages(self):
751
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
752
nodes = self.make_nodes(80, 2, 2)
754
builder.add_node(*node)
755
transport = get_transport('trace+' + self.get_url(''))
756
size = transport.put_file('index', builder.finish())
757
# Root page, 2 leaf pages
758
self.assertEqual(9339, size)
759
index = btree_index.BTreeGraphIndex(transport, 'index', size)
760
del transport._activity[:]
761
self.assertEqual([], transport._activity)
763
# The entire index should have been read linearly.
764
self.assertEqual([('readv', 'index', [(0, 4096)], False, None),
765
('readv', 'index', [(4096, 4096), (8192, 1147)], False, None)],
767
# XXX: TODO: write some badly-ordered nodes, and some pointers-to-wrong
768
# node and make validate find them.
770
def test_eq_ne(self):
771
# two indices are equal when constructed with the same parameters:
772
transport1 = get_transport('trace+' + self.get_url(''))
773
transport2 = get_transport(self.get_url(''))
775
btree_index.BTreeGraphIndex(transport1, 'index', None) ==
776
btree_index.BTreeGraphIndex(transport1, 'index', None))
778
btree_index.BTreeGraphIndex(transport1, 'index', 20) ==
779
btree_index.BTreeGraphIndex(transport1, 'index', 20))
781
btree_index.BTreeGraphIndex(transport1, 'index', 20) ==
782
btree_index.BTreeGraphIndex(transport2, 'index', 20))
784
btree_index.BTreeGraphIndex(transport1, 'inde1', 20) ==
785
btree_index.BTreeGraphIndex(transport1, 'inde2', 20))
787
btree_index.BTreeGraphIndex(transport1, 'index', 10) ==
788
btree_index.BTreeGraphIndex(transport1, 'index', 20))
790
btree_index.BTreeGraphIndex(transport1, 'index', None) !=
791
btree_index.BTreeGraphIndex(transport1, 'index', None))
793
btree_index.BTreeGraphIndex(transport1, 'index', 20) !=
794
btree_index.BTreeGraphIndex(transport1, 'index', 20))
796
btree_index.BTreeGraphIndex(transport1, 'index', 20) !=
797
btree_index.BTreeGraphIndex(transport2, 'index', 20))
799
btree_index.BTreeGraphIndex(transport1, 'inde1', 20) !=
800
btree_index.BTreeGraphIndex(transport1, 'inde2', 20))
802
btree_index.BTreeGraphIndex(transport1, 'index', 10) !=
803
btree_index.BTreeGraphIndex(transport1, 'index', 20))
805
def test_iter_all_only_root_no_size(self):
806
self.make_index(nodes=[(('key',), 'value', ())])
807
trans = get_transport('trace+' + self.get_url(''))
808
index = btree_index.BTreeGraphIndex(trans, 'index', None)
809
del trans._activity[:]
810
self.assertEqual([(('key',), 'value')],
811
[x[1:] for x in index.iter_all_entries()])
812
self.assertEqual([('get', 'index')], trans._activity)
814
def test_iter_all_entries_reads(self):
815
# iterating all entries reads the header, then does a linear
817
self.shrink_page_size()
818
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
819
# 20k nodes is enough to create a two internal nodes on the second
820
# level, with a 2K page size
821
nodes = self.make_nodes(10000, 2, 2)
823
builder.add_node(*node)
824
transport = get_transport('trace+' + self.get_url(''))
825
size = transport.put_file('index', builder.finish())
826
self.assertEqual(1303220, size, 'number of expected bytes in the'
828
page_size = btree_index._PAGE_SIZE
830
index = btree_index.BTreeGraphIndex(transport, 'index', size)
831
del transport._activity[:]
832
self.assertEqual([], transport._activity)
833
found_nodes = self.time(list, index.iter_all_entries())
835
for node in found_nodes:
836
self.assertTrue(node[0] is index)
837
bare_nodes.append(node[1:])
838
self.assertEqual(3, len(index._row_lengths),
839
"Not enough rows: %r" % index._row_lengths)
840
# Should be as long as the nodes we supplied
841
self.assertEqual(20000, len(found_nodes))
842
# Should have the same content
843
self.assertEqual(set(nodes), set(bare_nodes))
844
# Should have done linear scan IO up the index, ignoring
845
# the internal nodes:
846
# The entire index should have been read
847
total_pages = sum(index._row_lengths)
848
self.assertEqual(total_pages, index._row_offsets[-1])
849
self.assertEqual(1303220, size)
850
# The start of the leaves
851
first_byte = index._row_offsets[-2] * page_size
853
for offset in range(first_byte, size, page_size):
854
readv_request.append((offset, page_size))
855
# The last page is truncated
856
readv_request[-1] = (readv_request[-1][0], 1303220 % page_size)
857
expected = [('readv', 'index', [(0, page_size)], False, None),
858
('readv', 'index', readv_request, False, None)]
859
if expected != transport._activity:
860
self.assertEqualDiff(pprint.pformat(expected),
861
pprint.pformat(transport._activity))
863
def _test_iter_entries_references_resolved(self):
864
index = self.make_index(1, nodes=[
865
(('name', ), 'data', ([('ref', ), ('ref', )], )),
866
(('ref', ), 'refdata', ([], ))])
867
self.assertEqual(set([(index, ('name', ), 'data', ((('ref',),('ref',)),)),
868
(index, ('ref', ), 'refdata', ((), ))]),
869
set(index.iter_entries([('name',), ('ref',)])))
871
def test_iter_entries_references_2_refs_resolved(self):
872
# iterating some entries reads just the pages needed. For now, to
873
# get it working and start measuring, only 4K pages are read.
874
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
875
# 80 nodes is enough to create a two-level index.
876
nodes = self.make_nodes(160, 2, 2)
878
builder.add_node(*node)
879
transport = get_transport('trace+' + self.get_url(''))
880
size = transport.put_file('index', builder.finish())
882
index = btree_index.BTreeGraphIndex(transport, 'index', size)
883
del transport._activity[:]
884
self.assertEqual([], transport._activity)
886
found_nodes = list(index.iter_entries([nodes[30][0]]))
888
for node in found_nodes:
889
self.assertTrue(node[0] is index)
890
bare_nodes.append(node[1:])
891
# Should be as long as the nodes we supplied
892
self.assertEqual(1, len(found_nodes))
893
# Should have the same content
894
self.assertEqual(nodes[30], bare_nodes[0])
895
# Should have read the root node, then one leaf page:
896
self.assertEqual([('readv', 'index', [(0, 4096)], False, None),
897
('readv', 'index', [(8192, 4096), ], False, None)],
900
def test_iter_key_prefix_1_element_key_None(self):
901
index = self.make_index()
902
self.assertRaises(errors.BadIndexKey, list,
903
index.iter_entries_prefix([(None, )]))
905
def test_iter_key_prefix_wrong_length(self):
906
index = self.make_index()
907
self.assertRaises(errors.BadIndexKey, list,
908
index.iter_entries_prefix([('foo', None)]))
909
index = self.make_index(key_elements=2)
910
self.assertRaises(errors.BadIndexKey, list,
911
index.iter_entries_prefix([('foo', )]))
912
self.assertRaises(errors.BadIndexKey, list,
913
index.iter_entries_prefix([('foo', None, None)]))
915
def test_iter_key_prefix_1_key_element_no_refs(self):
916
index = self.make_index( nodes=[
917
(('name', ), 'data', ()),
918
(('ref', ), 'refdata', ())])
919
self.assertEqual(set([(index, ('name', ), 'data'),
920
(index, ('ref', ), 'refdata')]),
921
set(index.iter_entries_prefix([('name', ), ('ref', )])))
923
def test_iter_key_prefix_1_key_element_refs(self):
924
index = self.make_index(1, nodes=[
925
(('name', ), 'data', ([('ref', )], )),
926
(('ref', ), 'refdata', ([], ))])
927
self.assertEqual(set([(index, ('name', ), 'data', ((('ref',),),)),
928
(index, ('ref', ), 'refdata', ((), ))]),
929
set(index.iter_entries_prefix([('name', ), ('ref', )])))
931
def test_iter_key_prefix_2_key_element_no_refs(self):
932
index = self.make_index(key_elements=2, nodes=[
933
(('name', 'fin1'), 'data', ()),
934
(('name', 'fin2'), 'beta', ()),
935
(('ref', 'erence'), 'refdata', ())])
936
self.assertEqual(set([(index, ('name', 'fin1'), 'data'),
937
(index, ('ref', 'erence'), 'refdata')]),
938
set(index.iter_entries_prefix([('name', 'fin1'), ('ref', 'erence')])))
939
self.assertEqual(set([(index, ('name', 'fin1'), 'data'),
940
(index, ('name', 'fin2'), 'beta')]),
941
set(index.iter_entries_prefix([('name', None)])))
943
def test_iter_key_prefix_2_key_element_refs(self):
944
index = self.make_index(1, key_elements=2, nodes=[
945
(('name', 'fin1'), 'data', ([('ref', 'erence')], )),
946
(('name', 'fin2'), 'beta', ([], )),
947
(('ref', 'erence'), 'refdata', ([], ))])
948
self.assertEqual(set([(index, ('name', 'fin1'), 'data', ((('ref', 'erence'),),)),
949
(index, ('ref', 'erence'), 'refdata', ((), ))]),
950
set(index.iter_entries_prefix([('name', 'fin1'), ('ref', 'erence')])))
951
self.assertEqual(set([(index, ('name', 'fin1'), 'data', ((('ref', 'erence'),),)),
952
(index, ('name', 'fin2'), 'beta', ((), ))]),
953
set(index.iter_entries_prefix([('name', None)])))
955
# XXX: external_references tests are duplicated in test_index. We
956
# probably should have per_graph_index tests...
957
def test_external_references_no_refs(self):
958
index = self.make_index(ref_lists=0, nodes=[])
959
self.assertRaises(ValueError, index.external_references, 0)
961
def test_external_references_no_results(self):
962
index = self.make_index(ref_lists=1, nodes=[
963
(('key',), 'value', ([],))])
964
self.assertEqual(set(), index.external_references(0))
966
def test_external_references_missing_ref(self):
967
missing_key = ('missing',)
968
index = self.make_index(ref_lists=1, nodes=[
969
(('key',), 'value', ([missing_key],))])
970
self.assertEqual(set([missing_key]), index.external_references(0))
972
def test_external_references_multiple_ref_lists(self):
973
missing_key = ('missing',)
974
index = self.make_index(ref_lists=2, nodes=[
975
(('key',), 'value', ([], [missing_key]))])
976
self.assertEqual(set([]), index.external_references(0))
977
self.assertEqual(set([missing_key]), index.external_references(1))
979
def test_external_references_two_records(self):
980
index = self.make_index(ref_lists=1, nodes=[
981
(('key-1',), 'value', ([('key-2',)],)),
982
(('key-2',), 'value', ([],)),
984
self.assertEqual(set([]), index.external_references(0))
986
def test__find_ancestors_one_page(self):
989
index = self.make_index(ref_lists=1, key_elements=1, nodes=[
990
(key1, 'value', ([key2],)),
991
(key2, 'value', ([],)),
995
search_keys = index._find_ancestors([key1], 0, parent_map, missing_keys)
996
self.assertEqual({key1: (key2,), key2: ()}, parent_map)
997
self.assertEqual(set(), missing_keys)
998
self.assertEqual(set(), search_keys)
1000
def test__find_ancestors_one_page_w_missing(self):
1004
index = self.make_index(ref_lists=1, key_elements=1, nodes=[
1005
(key1, 'value', ([key2],)),
1006
(key2, 'value', ([],)),
1009
missing_keys = set()
1010
search_keys = index._find_ancestors([key2, key3], 0, parent_map,
1012
self.assertEqual({key2: ()}, parent_map)
1013
# we know that key3 is missing because we read the page that it would
1015
self.assertEqual(set([key3]), missing_keys)
1016
self.assertEqual(set(), search_keys)
1018
def test__find_ancestors_one_parent_missing(self):
1022
index = self.make_index(ref_lists=1, key_elements=1, nodes=[
1023
(key1, 'value', ([key2],)),
1024
(key2, 'value', ([key3],)),
1027
missing_keys = set()
1028
search_keys = index._find_ancestors([key1], 0, parent_map,
1030
self.assertEqual({key1: (key2,), key2: (key3,)}, parent_map)
1031
self.assertEqual(set(), missing_keys)
1032
# all we know is that key3 wasn't present on the page we were reading
1033
# but if you look, the last key is key2 which comes before key3, so we
1034
# don't know whether key3 would land on this page or not.
1035
self.assertEqual(set([key3]), search_keys)
1036
search_keys = index._find_ancestors(search_keys, 0, parent_map,
1038
# passing it back in, we are sure it is 'missing'
1039
self.assertEqual({key1: (key2,), key2: (key3,)}, parent_map)
1040
self.assertEqual(set([key3]), missing_keys)
1041
self.assertEqual(set([]), search_keys)
1043
def test__find_ancestors_dont_search_known(self):
1047
index = self.make_index(ref_lists=1, key_elements=1, nodes=[
1048
(key1, 'value', ([key2],)),
1049
(key2, 'value', ([key3],)),
1050
(key3, 'value', ([],)),
1052
# We already know about key2, so we won't try to search for key3
1053
parent_map = {key2: (key3,)}
1054
missing_keys = set()
1055
search_keys = index._find_ancestors([key1], 0, parent_map,
1057
self.assertEqual({key1: (key2,), key2: (key3,)}, parent_map)
1058
self.assertEqual(set(), missing_keys)
1059
self.assertEqual(set(), search_keys)
1061
def test__find_ancestors_multiple_pages(self):
1062
# We need to use enough keys that we actually cause a split
1063
start_time = 1249671539
1064
email = "joebob@example.com"
1068
for i in xrange(400):
1069
rev_id = '%s-%s-%s' % (email,
1070
osutils.compact_date(start_time + i),
1071
osutils.rand_chars(16))
1073
nodes.append((rev_key, 'value', ref_lists))
1074
# We have a ref 'list' of length 1, with a list of parents, with 1
1075
# parent which is a key
1076
ref_lists = ((rev_key,),)
1077
rev_keys.append(rev_key)
1078
index = self.make_index(ref_lists=1, key_elements=1, nodes=nodes)
1079
self.assertEqual(400, index.key_count())
1080
self.assertEqual(3, len(index._row_offsets))
1081
nodes = dict(index._read_nodes([1, 2]))
1084
min_l2_key = l2.min_key
1085
max_l1_key = l1.max_key
1086
self.assertTrue(max_l1_key < min_l2_key)
1087
parents_min_l2_key = l2.keys[min_l2_key][1][0]
1088
self.assertEqual((l1.max_key,), parents_min_l2_key)
1089
# Now, whatever key we select that would fall on the second page,
1090
# should give us all the parents until the page break
1091
key_idx = rev_keys.index(min_l2_key)
1092
next_key = rev_keys[key_idx+1]
1093
# So now when we get the parent map, we should get the key we are
1094
# looking for, min_l2_key, and then a reference to go look for the
1095
# parent of that key
1097
missing_keys = set()
1098
search_keys = index._find_ancestors([next_key], 0, parent_map,
1100
self.assertEqual([min_l2_key, next_key], sorted(parent_map))
1101
self.assertEqual(set(), missing_keys)
1102
self.assertEqual(set([max_l1_key]), search_keys)
1104
search_keys = index._find_ancestors([max_l1_key], 0, parent_map,
1106
self.assertEqual(sorted(l1.keys), sorted(parent_map))
1107
self.assertEqual(set(), missing_keys)
1108
self.assertEqual(set(), search_keys)
1110
def test__find_ancestors_empty_index(self):
1111
index = self.make_index(ref_lists=1, key_elements=1, nodes=[])
1113
missing_keys = set()
1114
search_keys = index._find_ancestors([('one',), ('two',)], 0, parent_map,
1116
self.assertEqual(set(), search_keys)
1117
self.assertEqual({}, parent_map)
1118
self.assertEqual(set([('one',), ('two',)]), missing_keys)
1120
def test_supports_unlimited_cache(self):
1121
builder = btree_index.BTreeBuilder(reference_lists=0, key_elements=1)
1122
# We need enough nodes to cause a page split (so we have both an
1123
# internal node and a couple leaf nodes. 500 seems to be enough.)
1124
nodes = self.make_nodes(500, 1, 0)
1126
builder.add_node(*node)
1127
stream = builder.finish()
1128
trans = get_transport(self.get_url())
1129
size = trans.put_file('index', stream)
1130
index = btree_index.BTreeGraphIndex(trans, 'index', size)
1131
self.assertEqual(500, index.key_count())
1132
# We have an internal node
1133
self.assertEqual(2, len(index._row_lengths))
1134
# We have at least 2 leaf nodes
1135
self.assertTrue(index._row_lengths[-1] >= 2)
1136
self.assertIsInstance(index._leaf_node_cache, lru_cache.LRUCache)
1137
self.assertEqual(btree_index._NODE_CACHE_SIZE,
1138
index._leaf_node_cache._max_cache)
1139
self.assertIsInstance(index._internal_node_cache, fifo_cache.FIFOCache)
1140
self.assertEqual(100, index._internal_node_cache._max_cache)
1141
# No change if unlimited_cache=False is passed
1142
index = btree_index.BTreeGraphIndex(trans, 'index', size,
1143
unlimited_cache=False)
1144
self.assertIsInstance(index._leaf_node_cache, lru_cache.LRUCache)
1145
self.assertEqual(btree_index._NODE_CACHE_SIZE,
1146
index._leaf_node_cache._max_cache)
1147
self.assertIsInstance(index._internal_node_cache, fifo_cache.FIFOCache)
1148
self.assertEqual(100, index._internal_node_cache._max_cache)
1149
index = btree_index.BTreeGraphIndex(trans, 'index', size,
1150
unlimited_cache=True)
1151
self.assertIsInstance(index._leaf_node_cache, dict)
1152
self.assertIs(type(index._internal_node_cache), dict)
1153
# Exercise the lookup code
1154
entries = set(index.iter_entries([n[0] for n in nodes]))
1155
self.assertEqual(500, len(entries))
1158
class TestBTreeNodes(BTreeTestCase):
1160
def restore_parser(self):
1161
btree_index._btree_serializer = self.saved_parser
1164
BTreeTestCase.setUp(self)
1165
self.saved_parser = btree_index._btree_serializer
1166
self.addCleanup(self.restore_parser)
1167
btree_index._btree_serializer = self.parse_btree
1169
def test_LeafNode_1_0(self):
1170
node_bytes = ("type=leaf\n"
1171
"0000000000000000000000000000000000000000\x00\x00value:0\n"
1172
"1111111111111111111111111111111111111111\x00\x00value:1\n"
1173
"2222222222222222222222222222222222222222\x00\x00value:2\n"
1174
"3333333333333333333333333333333333333333\x00\x00value:3\n"
1175
"4444444444444444444444444444444444444444\x00\x00value:4\n")
1176
node = btree_index._LeafNode(node_bytes, 1, 0)
1177
# We do direct access, or don't care about order, to leaf nodes most of
1178
# the time, so a dict is useful:
1180
("0000000000000000000000000000000000000000",): ("value:0", ()),
1181
("1111111111111111111111111111111111111111",): ("value:1", ()),
1182
("2222222222222222222222222222222222222222",): ("value:2", ()),
1183
("3333333333333333333333333333333333333333",): ("value:3", ()),
1184
("4444444444444444444444444444444444444444",): ("value:4", ()),
1187
def test_LeafNode_2_2(self):
1188
node_bytes = ("type=leaf\n"
1189
"00\x0000\x00\t00\x00ref00\x00value:0\n"
1190
"00\x0011\x0000\x00ref00\t00\x00ref00\r01\x00ref01\x00value:1\n"
1191
"11\x0033\x0011\x00ref22\t11\x00ref22\r11\x00ref22\x00value:3\n"
1192
"11\x0044\x00\t11\x00ref00\x00value:4\n"
1195
node = btree_index._LeafNode(node_bytes, 2, 2)
1196
# We do direct access, or don't care about order, to leaf nodes most of
1197
# the time, so a dict is useful:
1199
('00', '00'): ('value:0', ((), (('00', 'ref00'),))),
1200
('00', '11'): ('value:1',
1201
((('00', 'ref00'),), (('00', 'ref00'), ('01', 'ref01')))),
1202
('11', '33'): ('value:3',
1203
((('11', 'ref22'),), (('11', 'ref22'), ('11', 'ref22')))),
1204
('11', '44'): ('value:4', ((), (('11', 'ref00'),)))
1207
def test_InternalNode_1(self):
1208
node_bytes = ("type=internal\n"
1210
"0000000000000000000000000000000000000000\n"
1211
"1111111111111111111111111111111111111111\n"
1212
"2222222222222222222222222222222222222222\n"
1213
"3333333333333333333333333333333333333333\n"
1214
"4444444444444444444444444444444444444444\n"
1216
node = btree_index._InternalNode(node_bytes)
1217
# We want to bisect to find the right children from this node, so a
1218
# vector is most useful.
1220
("0000000000000000000000000000000000000000",),
1221
("1111111111111111111111111111111111111111",),
1222
("2222222222222222222222222222222222222222",),
1223
("3333333333333333333333333333333333333333",),
1224
("4444444444444444444444444444444444444444",),
1226
self.assertEqual(1, node.offset)
1228
def test_LeafNode_2_2(self):
1229
node_bytes = ("type=leaf\n"
1230
"00\x0000\x00\t00\x00ref00\x00value:0\n"
1231
"00\x0011\x0000\x00ref00\t00\x00ref00\r01\x00ref01\x00value:1\n"
1232
"11\x0033\x0011\x00ref22\t11\x00ref22\r11\x00ref22\x00value:3\n"
1233
"11\x0044\x00\t11\x00ref00\x00value:4\n"
1236
node = btree_index._LeafNode(node_bytes, 2, 2)
1237
# We do direct access, or don't care about order, to leaf nodes most of
1238
# the time, so a dict is useful:
1240
('00', '00'): ('value:0', ((), (('00', 'ref00'),))),
1241
('00', '11'): ('value:1',
1242
((('00', 'ref00'),), (('00', 'ref00'), ('01', 'ref01')))),
1243
('11', '33'): ('value:3',
1244
((('11', 'ref22'),), (('11', 'ref22'), ('11', 'ref22')))),
1245
('11', '44'): ('value:4', ((), (('11', 'ref00'),)))
1248
def assertFlattened(self, expected, key, value, refs):
1249
flat_key, flat_line = self.parse_btree._flatten_node(
1250
(None, key, value, refs), bool(refs))
1251
self.assertEqual('\x00'.join(key), flat_key)
1252
self.assertEqual(expected, flat_line)
1254
def test__flatten_node(self):
1255
self.assertFlattened('key\0\0value\n', ('key',), 'value', [])
1256
self.assertFlattened('key\0tuple\0\0value str\n',
1257
('key', 'tuple'), 'value str', [])
1258
self.assertFlattened('key\0tuple\0triple\0\0value str\n',
1259
('key', 'tuple', 'triple'), 'value str', [])
1260
self.assertFlattened('k\0t\0s\0ref\0value str\n',
1261
('k', 't', 's'), 'value str', [[('ref',)]])
1262
self.assertFlattened('key\0tuple\0ref\0key\0value str\n',
1263
('key', 'tuple'), 'value str', [[('ref', 'key')]])
1264
self.assertFlattened("00\x0000\x00\t00\x00ref00\x00value:0\n",
1265
('00', '00'), 'value:0', ((), (('00', 'ref00'),)))
1266
self.assertFlattened(
1267
"00\x0011\x0000\x00ref00\t00\x00ref00\r01\x00ref01\x00value:1\n",
1268
('00', '11'), 'value:1',
1269
((('00', 'ref00'),), (('00', 'ref00'), ('01', 'ref01'))))
1270
self.assertFlattened(
1271
"11\x0033\x0011\x00ref22\t11\x00ref22\r11\x00ref22\x00value:3\n",
1272
('11', '33'), 'value:3',
1273
((('11', 'ref22'),), (('11', 'ref22'), ('11', 'ref22'))))
1274
self.assertFlattened(
1275
"11\x0044\x00\t11\x00ref00\x00value:4\n",
1276
('11', '44'), 'value:4', ((), (('11', 'ref00'),)))
1279
class TestCompiledBtree(tests.TestCase):
1281
def test_exists(self):
1282
# This is just to let the user know if they don't have the feature
1284
self.requireFeature(compiled_btreeparser_feature)
1287
class TestMultiBisectRight(tests.TestCase):
1289
def assertMultiBisectRight(self, offsets, search_keys, fixed_keys):
1290
self.assertEqual(offsets,
1291
btree_index.BTreeGraphIndex._multi_bisect_right(
1292
search_keys, fixed_keys))
1294
def test_after(self):
1295
self.assertMultiBisectRight([(1, ['b'])], ['b'], ['a'])
1296
self.assertMultiBisectRight([(3, ['e', 'f', 'g'])],
1297
['e', 'f', 'g'], ['a', 'b', 'c'])
1299
def test_before(self):
1300
self.assertMultiBisectRight([(0, ['a'])], ['a'], ['b'])
1301
self.assertMultiBisectRight([(0, ['a', 'b', 'c', 'd'])],
1302
['a', 'b', 'c', 'd'], ['e', 'f', 'g'])
1304
def test_exact(self):
1305
self.assertMultiBisectRight([(1, ['a'])], ['a'], ['a'])
1306
self.assertMultiBisectRight([(1, ['a']), (2, ['b'])], ['a', 'b'], ['a', 'b'])
1307
self.assertMultiBisectRight([(1, ['a']), (3, ['c'])],
1308
['a', 'c'], ['a', 'b', 'c'])
1310
def test_inbetween(self):
1311
self.assertMultiBisectRight([(1, ['b'])], ['b'], ['a', 'c'])
1312
self.assertMultiBisectRight([(1, ['b', 'c', 'd']), (2, ['f', 'g'])],
1313
['b', 'c', 'd', 'f', 'g'], ['a', 'e', 'h'])
1315
def test_mixed(self):
1316
self.assertMultiBisectRight([(0, ['a', 'b']), (2, ['d', 'e']),
1318
['a', 'b', 'd', 'e', 'g', 'h'],
1319
['c', 'd', 'f', 'g'])
1322
class TestExpandOffsets(tests.TestCase):
1324
def make_index(self, size, recommended_pages=None):
1325
"""Make an index with a generic size.
1327
This doesn't actually create anything on disk, it just primes a
1328
BTreeGraphIndex with the recommended information.
1330
index = btree_index.BTreeGraphIndex(get_transport('memory:///'),
1331
'test-index', size=size)
1332
if recommended_pages is not None:
1333
index._recommended_pages = recommended_pages
1336
def set_cached_offsets(self, index, cached_offsets):
1337
"""Monkeypatch to give a canned answer for _get_offsets_for...()."""
1338
def _get_offsets_to_cached_pages():
1339
cached = set(cached_offsets)
1341
index._get_offsets_to_cached_pages = _get_offsets_to_cached_pages
1343
def prepare_index(self, index, node_ref_lists, key_length, key_count,
1344
row_lengths, cached_offsets):
1345
"""Setup the BTreeGraphIndex with some pre-canned information."""
1346
index.node_ref_lists = node_ref_lists
1347
index._key_length = key_length
1348
index._key_count = key_count
1349
index._row_lengths = row_lengths
1350
index._compute_row_offsets()
1351
index._root_node = btree_index._InternalNode('internal\noffset=0\n')
1352
self.set_cached_offsets(index, cached_offsets)
1354
def make_100_node_index(self):
1355
index = self.make_index(4096*100, 6)
1356
# Consider we've already made a single request at the middle
1357
self.prepare_index(index, node_ref_lists=0, key_length=1,
1358
key_count=1000, row_lengths=[1, 99],
1359
cached_offsets=[0, 50])
1362
def make_1000_node_index(self):
1363
index = self.make_index(4096*1000, 6)
1364
# Pretend we've already made a single request in the middle
1365
self.prepare_index(index, node_ref_lists=0, key_length=1,
1366
key_count=90000, row_lengths=[1, 9, 990],
1367
cached_offsets=[0, 5, 500])
1370
def assertNumPages(self, expected_pages, index, size):
1372
self.assertEqual(expected_pages, index._compute_total_pages_in_index())
1374
def assertExpandOffsets(self, expected, index, offsets):
1375
self.assertEqual(expected, index._expand_offsets(offsets),
1376
'We did not get the expected value after expanding'
1379
def test_default_recommended_pages(self):
1380
index = self.make_index(None)
1381
# local transport recommends 4096 byte reads, which is 1 page
1382
self.assertEqual(1, index._recommended_pages)
1384
def test__compute_total_pages_in_index(self):
1385
index = self.make_index(None)
1386
self.assertNumPages(1, index, 1024)
1387
self.assertNumPages(1, index, 4095)
1388
self.assertNumPages(1, index, 4096)
1389
self.assertNumPages(2, index, 4097)
1390
self.assertNumPages(2, index, 8192)
1391
self.assertNumPages(76, index, 4096*75 + 10)
1393
def test__find_layer_start_and_stop(self):
1394
index = self.make_1000_node_index()
1395
self.assertEqual((0, 1), index._find_layer_first_and_end(0))
1396
self.assertEqual((1, 10), index._find_layer_first_and_end(1))
1397
self.assertEqual((1, 10), index._find_layer_first_and_end(9))
1398
self.assertEqual((10, 1000), index._find_layer_first_and_end(10))
1399
self.assertEqual((10, 1000), index._find_layer_first_and_end(99))
1400
self.assertEqual((10, 1000), index._find_layer_first_and_end(999))
1402
def test_unknown_size(self):
1403
# We should not expand if we don't know the file size
1404
index = self.make_index(None, 10)
1405
self.assertExpandOffsets([0], index, [0])
1406
self.assertExpandOffsets([1, 4, 9], index, [1, 4, 9])
1408
def test_more_than_recommended(self):
1409
index = self.make_index(4096*100, 2)
1410
self.assertExpandOffsets([1, 10], index, [1, 10])
1411
self.assertExpandOffsets([1, 10, 20], index, [1, 10, 20])
1413
def test_read_all_from_root(self):
1414
index = self.make_index(4096*10, 20)
1415
self.assertExpandOffsets(range(10), index, [0])
1417
def test_read_all_when_cached(self):
1418
# We've read enough that we can grab all the rest in a single request
1419
index = self.make_index(4096*10, 5)
1420
self.prepare_index(index, node_ref_lists=0, key_length=1,
1421
key_count=1000, row_lengths=[1, 9],
1422
cached_offsets=[0, 1, 2, 5, 6])
1423
# It should fill the remaining nodes, regardless of the one requested
1424
self.assertExpandOffsets([3, 4, 7, 8, 9], index, [3])
1425
self.assertExpandOffsets([3, 4, 7, 8, 9], index, [8])
1426
self.assertExpandOffsets([3, 4, 7, 8, 9], index, [9])
1428
def test_no_root_node(self):
1429
index = self.make_index(4096*10, 5)
1430
self.assertExpandOffsets([0], index, [0])
1432
def test_include_neighbors(self):
1433
index = self.make_100_node_index()
1434
# We expand in both directions, until we have at least 'recommended'
1436
self.assertExpandOffsets([9, 10, 11, 12, 13, 14, 15], index, [12])
1437
self.assertExpandOffsets([88, 89, 90, 91, 92, 93, 94], index, [91])
1438
# If we hit an 'edge' we continue in the other direction
1439
self.assertExpandOffsets([1, 2, 3, 4, 5, 6], index, [2])
1440
self.assertExpandOffsets([94, 95, 96, 97, 98, 99], index, [98])
1442
# Requesting many nodes will expand all locations equally
1443
self.assertExpandOffsets([1, 2, 3, 80, 81, 82], index, [2, 81])
1444
self.assertExpandOffsets([1, 2, 3, 9, 10, 11, 80, 81, 82], index,
1447
def test_stop_at_cached(self):
1448
index = self.make_100_node_index()
1449
self.set_cached_offsets(index, [0, 10, 19])
1450
self.assertExpandOffsets([11, 12, 13, 14, 15, 16], index, [11])
1451
self.assertExpandOffsets([11, 12, 13, 14, 15, 16], index, [12])
1452
self.assertExpandOffsets([12, 13, 14, 15, 16, 17, 18], index, [15])
1453
self.assertExpandOffsets([13, 14, 15, 16, 17, 18], index, [16])
1454
self.assertExpandOffsets([13, 14, 15, 16, 17, 18], index, [17])
1455
self.assertExpandOffsets([13, 14, 15, 16, 17, 18], index, [18])
1457
def test_cannot_fully_expand(self):
1458
index = self.make_100_node_index()
1459
self.set_cached_offsets(index, [0, 10, 12])
1460
# We don't go into an endless loop if we are bound by cached nodes
1461
self.assertExpandOffsets([11], index, [11])
1463
def test_overlap(self):
1464
index = self.make_100_node_index()
1465
self.assertExpandOffsets([10, 11, 12, 13, 14, 15], index, [12, 13])
1466
self.assertExpandOffsets([10, 11, 12, 13, 14, 15], index, [11, 14])
1468
def test_stay_within_layer(self):
1469
index = self.make_1000_node_index()
1470
# When expanding a request, we won't read nodes from the next layer
1471
self.assertExpandOffsets([1, 2, 3, 4], index, [2])
1472
self.assertExpandOffsets([6, 7, 8, 9], index, [6])
1473
self.assertExpandOffsets([6, 7, 8, 9], index, [9])
1474
self.assertExpandOffsets([10, 11, 12, 13, 14, 15], index, [10])
1475
self.assertExpandOffsets([10, 11, 12, 13, 14, 15, 16], index, [13])
1477
self.set_cached_offsets(index, [0, 4, 12])
1478
self.assertExpandOffsets([5, 6, 7, 8, 9], index, [7])
1479
self.assertExpandOffsets([10, 11], index, [11])
1481
def test_small_requests_unexpanded(self):
1482
index = self.make_100_node_index()
1483
self.set_cached_offsets(index, [0])
1484
self.assertExpandOffsets([1], index, [1])
1485
self.assertExpandOffsets([50], index, [50])
1486
# If we request more than one node, then we'll expand
1487
self.assertExpandOffsets([49, 50, 51, 59, 60, 61], index, [50, 60])
1489
# The first pass does not expand
1490
index = self.make_1000_node_index()
1491
self.set_cached_offsets(index, [0])
1492
self.assertExpandOffsets([1], index, [1])
1493
self.set_cached_offsets(index, [0, 1])
1494
self.assertExpandOffsets([100], index, [100])
1495
self.set_cached_offsets(index, [0, 1, 100])
1496
# But after the first depth, we will expand
1497
self.assertExpandOffsets([2, 3, 4, 5, 6, 7], index, [2])
1498
self.assertExpandOffsets([2, 3, 4, 5, 6, 7], index, [4])
1499
self.set_cached_offsets(index, [0, 1, 2, 3, 4, 5, 6, 7, 100])
1500
self.assertExpandOffsets([102, 103, 104, 105, 106, 107, 108], index,
added
removed
bzrlib/tests/test_btree_index.py
