1
# Copyright (C) 2008-2011 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."""
32
from bzrlib.tests import (
33
TestCaseWithTransport,
36
from bzrlib.tests import (
41
load_tests = scenarios.load_tests_apply_scenarios
44
def btreeparser_scenarios():
45
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}))
53
compiled_btreeparser_feature = features.ModuleAvailableFeature(
54
'bzrlib._btree_serializer_pyx')
57
class BTreeTestCase(TestCaseWithTransport):
58
# test names here are suffixed by the key length and reference list count
62
TestCaseWithTransport.setUp(self)
63
self.overrideAttr(btree_index, '_RESERVED_HEADER_BYTES', 100)
65
def make_nodes(self, count, key_elements, reference_lists):
66
"""Generate count*key_elements sample nodes."""
68
for prefix_pos in range(key_elements):
70
prefix = (str(prefix_pos) * 40,)
73
for pos in xrange(count):
74
# TODO: This creates odd keys. When count == 100,000, it
75
# creates a 240 byte key
76
key = prefix + (str(pos) * 40,)
77
value = "value:%s" % pos
79
# generate some references
81
for list_pos in range(reference_lists):
82
# as many keys in each list as its index + the key depth
83
# mod 2 - this generates both 0 length lists and
84
# ones slightly longer than the number of lists.
85
# It also ensures we have non homogeneous lists.
87
for ref_pos in range(list_pos + pos % 2):
89
# refer to a nearby key
90
refs[-1].append(prefix + ("ref" + str(pos - 1) * 40,))
92
# serial of this ref in the ref list
93
refs[-1].append(prefix + ("ref" + str(ref_pos) * 40,))
94
refs[-1] = tuple(refs[-1])
98
keys.append((key, value, refs))
101
def shrink_page_size(self):
102
"""Shrink the default page size so that less fits in a page."""
103
self.overrideAttr(btree_index, '_PAGE_SIZE')
104
btree_index._PAGE_SIZE = 2048
107
class TestBTreeBuilder(BTreeTestCase):
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()
115
def test_empty_1_0(self):
116
builder = btree_index.BTreeBuilder(key_elements=1, reference_lists=0)
117
# NamedTemporaryFile dies on builder.finish().read(). weird.
118
temp_file = builder.finish()
119
content = temp_file.read()
122
"B+Tree Graph Index 2\nnode_ref_lists=0\nkey_elements=1\nlen=0\n"
126
def test_empty_2_1(self):
127
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=1)
128
# NamedTemporaryFile dies on builder.finish().read(). weird.
129
temp_file = builder.finish()
130
content = temp_file.read()
133
"B+Tree Graph Index 2\nnode_ref_lists=1\nkey_elements=2\nlen=0\n"
137
def test_root_leaf_1_0(self):
138
builder = btree_index.BTreeBuilder(key_elements=1, reference_lists=0)
139
nodes = self.make_nodes(5, 1, 0)
141
builder.add_node(*node)
142
# NamedTemporaryFile dies on builder.finish().read(). weird.
143
temp_file = builder.finish()
144
content = temp_file.read()
146
self.assertEqual(131, len(content))
148
"B+Tree Graph Index 2\nnode_ref_lists=0\nkey_elements=1\nlen=5\n"
151
node_content = content[73:]
152
node_bytes = zlib.decompress(node_content)
153
expected_node = ("type=leaf\n"
154
"0000000000000000000000000000000000000000\x00\x00value:0\n"
155
"1111111111111111111111111111111111111111\x00\x00value:1\n"
156
"2222222222222222222222222222222222222222\x00\x00value:2\n"
157
"3333333333333333333333333333333333333333\x00\x00value:3\n"
158
"4444444444444444444444444444444444444444\x00\x00value:4\n")
159
self.assertEqual(expected_node, node_bytes)
161
def test_root_leaf_2_2(self):
162
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
163
nodes = self.make_nodes(5, 2, 2)
165
builder.add_node(*node)
166
# NamedTemporaryFile dies on builder.finish().read(). weird.
167
temp_file = builder.finish()
168
content = temp_file.read()
170
self.assertEqual(238, len(content))
172
"B+Tree Graph Index 2\nnode_ref_lists=2\nkey_elements=2\nlen=10\n"
175
node_content = content[74:]
176
node_bytes = zlib.decompress(node_content)
179
"0000000000000000000000000000000000000000\x000000000000000000000000000000000000000000\x00\t0000000000000000000000000000000000000000\x00ref0000000000000000000000000000000000000000\x00value:0\n"
180
"0000000000000000000000000000000000000000\x001111111111111111111111111111111111111111\x000000000000000000000000000000000000000000\x00ref0000000000000000000000000000000000000000\t0000000000000000000000000000000000000000\x00ref0000000000000000000000000000000000000000\r0000000000000000000000000000000000000000\x00ref0000000000000000000000000000000000000000\x00value:1\n"
181
"0000000000000000000000000000000000000000\x002222222222222222222222222222222222222222\x00\t0000000000000000000000000000000000000000\x00ref0000000000000000000000000000000000000000\x00value:2\n"
182
"0000000000000000000000000000000000000000\x003333333333333333333333333333333333333333\x000000000000000000000000000000000000000000\x00ref2222222222222222222222222222222222222222\t0000000000000000000000000000000000000000\x00ref2222222222222222222222222222222222222222\r0000000000000000000000000000000000000000\x00ref2222222222222222222222222222222222222222\x00value:3\n"
183
"0000000000000000000000000000000000000000\x004444444444444444444444444444444444444444\x00\t0000000000000000000000000000000000000000\x00ref0000000000000000000000000000000000000000\x00value:4\n"
184
"1111111111111111111111111111111111111111\x000000000000000000000000000000000000000000\x00\t1111111111111111111111111111111111111111\x00ref0000000000000000000000000000000000000000\x00value:0\n"
185
"1111111111111111111111111111111111111111\x001111111111111111111111111111111111111111\x001111111111111111111111111111111111111111\x00ref0000000000000000000000000000000000000000\t1111111111111111111111111111111111111111\x00ref0000000000000000000000000000000000000000\r1111111111111111111111111111111111111111\x00ref0000000000000000000000000000000000000000\x00value:1\n"
186
"1111111111111111111111111111111111111111\x002222222222222222222222222222222222222222\x00\t1111111111111111111111111111111111111111\x00ref0000000000000000000000000000000000000000\x00value:2\n"
187
"1111111111111111111111111111111111111111\x003333333333333333333333333333333333333333\x001111111111111111111111111111111111111111\x00ref2222222222222222222222222222222222222222\t1111111111111111111111111111111111111111\x00ref2222222222222222222222222222222222222222\r1111111111111111111111111111111111111111\x00ref2222222222222222222222222222222222222222\x00value:3\n"
188
"1111111111111111111111111111111111111111\x004444444444444444444444444444444444444444\x00\t1111111111111111111111111111111111111111\x00ref0000000000000000000000000000000000000000\x00value:4\n"
191
self.assertEqual(expected_node, node_bytes)
193
def test_2_leaves_1_0(self):
194
builder = btree_index.BTreeBuilder(key_elements=1, reference_lists=0)
195
nodes = self.make_nodes(400, 1, 0)
197
builder.add_node(*node)
198
# NamedTemporaryFile dies on builder.finish().read(). weird.
199
temp_file = builder.finish()
200
content = temp_file.read()
202
self.assertEqual(9283, len(content))
204
"B+Tree Graph Index 2\nnode_ref_lists=0\nkey_elements=1\nlen=400\n"
207
root = content[77:4096]
208
leaf1 = content[4096:8192]
209
leaf2 = content[8192:]
210
root_bytes = zlib.decompress(root)
214
) + ("307" * 40) + "\n"
215
self.assertEqual(expected_root, root_bytes)
216
# We already know serialisation works for leaves, check key selection:
217
leaf1_bytes = zlib.decompress(leaf1)
218
sorted_node_keys = sorted(node[0] for node in nodes)
219
node = btree_index._LeafNode(leaf1_bytes, 1, 0)
220
self.assertEqual(231, len(node))
221
self.assertEqual(sorted_node_keys[:231], node.all_keys())
222
leaf2_bytes = zlib.decompress(leaf2)
223
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())
227
def test_last_page_rounded_1_layer(self):
228
builder = btree_index.BTreeBuilder(key_elements=1, reference_lists=0)
229
nodes = self.make_nodes(10, 1, 0)
231
builder.add_node(*node)
232
# NamedTemporaryFile dies on builder.finish().read(). weird.
233
temp_file = builder.finish()
234
content = temp_file.read()
236
self.assertEqual(155, len(content))
238
"B+Tree Graph Index 2\nnode_ref_lists=0\nkey_elements=1\nlen=10\n"
241
# Check thelast page is well formed
243
leaf2_bytes = zlib.decompress(leaf2)
244
node = btree_index._LeafNode(leaf2_bytes, 1, 0)
245
self.assertEqual(10, len(node))
246
sorted_node_keys = sorted(node[0] for node in nodes)
247
self.assertEqual(sorted_node_keys, node.all_keys())
249
def test_last_page_not_rounded_2_layer(self):
250
builder = btree_index.BTreeBuilder(key_elements=1, reference_lists=0)
251
nodes = self.make_nodes(400, 1, 0)
253
builder.add_node(*node)
254
# NamedTemporaryFile dies on builder.finish().read(). weird.
255
temp_file = builder.finish()
256
content = temp_file.read()
258
self.assertEqual(9283, len(content))
260
"B+Tree Graph Index 2\nnode_ref_lists=0\nkey_elements=1\nlen=400\n"
263
# Check the last page is well formed
264
leaf2 = content[8192:]
265
leaf2_bytes = zlib.decompress(leaf2)
266
node = btree_index._LeafNode(leaf2_bytes, 1, 0)
267
self.assertEqual(400 - 231, len(node))
268
sorted_node_keys = sorted(node[0] for node in nodes)
269
self.assertEqual(sorted_node_keys[231:], node.all_keys())
271
def test_three_level_tree_details(self):
272
# The left most pointer in the second internal node in a row should
273
# pointer to the second node that the internal node is for, _not_
274
# the first, otherwise the first node overlaps with the last node of
275
# the prior internal node on that row.
276
self.shrink_page_size()
277
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
278
# 40K nodes is enough to create a two internal nodes on the second
279
# level, with a 2K page size
280
nodes = self.make_nodes(20000, 2, 2)
283
builder.add_node(*node)
284
t = transport.get_transport('trace+' + self.get_url(''))
285
size = t.put_file('index', self.time(builder.finish))
287
index = btree_index.BTreeGraphIndex(t, 'index', size)
288
# Seed the metadata, we're using internal calls now.
290
self.assertEqual(3, len(index._row_lengths),
291
"Not enough rows: %r" % index._row_lengths)
292
self.assertEqual(4, len(index._row_offsets))
293
self.assertEqual(sum(index._row_lengths), index._row_offsets[-1])
294
internal_nodes = index._get_internal_nodes([0, 1, 2])
295
root_node = internal_nodes[0]
296
internal_node1 = internal_nodes[1]
297
internal_node2 = internal_nodes[2]
298
# The left most node node2 points at should be one after the right most
299
# node pointed at by node1.
300
self.assertEqual(internal_node2.offset, 1 + len(internal_node1.keys))
301
# The left most key of the second node pointed at by internal_node2
302
# should be its first key. We can check this by looking for its first key
303
# in the second node it points at
304
pos = index._row_offsets[2] + internal_node2.offset + 1
305
leaf = index._get_leaf_nodes([pos])[pos]
306
self.assertTrue(internal_node2.keys[0] in leaf)
308
def test_2_leaves_2_2(self):
309
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
310
nodes = self.make_nodes(100, 2, 2)
312
builder.add_node(*node)
313
# NamedTemporaryFile dies on builder.finish().read(). weird.
314
temp_file = builder.finish()
315
content = temp_file.read()
317
self.assertEqual(12643, len(content))
319
"B+Tree Graph Index 2\nnode_ref_lists=2\nkey_elements=2\nlen=200\n"
322
root = content[77:4096]
323
leaf1 = content[4096:8192]
324
leaf2 = content[8192:12288]
325
leaf3 = content[12288:]
326
root_bytes = zlib.decompress(root)
330
+ ("0" * 40) + "\x00" + ("91" * 40) + "\n"
331
+ ("1" * 40) + "\x00" + ("81" * 40) + "\n"
333
self.assertEqual(expected_root, root_bytes)
334
# We assume the other leaf nodes have been written correctly - layering
337
def test_spill_index_stress_1_1(self):
338
builder = btree_index.BTreeBuilder(key_elements=1, spill_at=2)
339
nodes = [node[0:2] for node in self.make_nodes(16, 1, 0)]
340
builder.add_node(*nodes[0])
341
# Test the parts of the index that take up memory are doing so
343
self.assertEqual(1, len(builder._nodes))
344
self.assertIs(None, builder._nodes_by_key)
345
builder.add_node(*nodes[1])
346
self.assertEqual(0, len(builder._nodes))
347
self.assertIs(None, builder._nodes_by_key)
348
self.assertEqual(1, len(builder._backing_indices))
349
self.assertEqual(2, builder._backing_indices[0].key_count())
351
builder.add_node(*nodes[2])
352
self.assertEqual(1, len(builder._nodes))
353
self.assertIs(None, builder._nodes_by_key)
354
# And spills to a second backing index combing all
355
builder.add_node(*nodes[3])
356
self.assertEqual(0, len(builder._nodes))
357
self.assertIs(None, builder._nodes_by_key)
358
self.assertEqual(2, len(builder._backing_indices))
359
self.assertEqual(None, builder._backing_indices[0])
360
self.assertEqual(4, builder._backing_indices[1].key_count())
361
# The next spills to the 2-len slot
362
builder.add_node(*nodes[4])
363
builder.add_node(*nodes[5])
364
self.assertEqual(0, len(builder._nodes))
365
self.assertIs(None, builder._nodes_by_key)
366
self.assertEqual(2, len(builder._backing_indices))
367
self.assertEqual(2, builder._backing_indices[0].key_count())
368
self.assertEqual(4, builder._backing_indices[1].key_count())
369
# Next spill combines
370
builder.add_node(*nodes[6])
371
builder.add_node(*nodes[7])
372
self.assertEqual(3, len(builder._backing_indices))
373
self.assertEqual(None, builder._backing_indices[0])
374
self.assertEqual(None, builder._backing_indices[1])
375
self.assertEqual(8, builder._backing_indices[2].key_count())
376
# And so forth - counting up in binary.
377
builder.add_node(*nodes[8])
378
builder.add_node(*nodes[9])
379
self.assertEqual(3, len(builder._backing_indices))
380
self.assertEqual(2, builder._backing_indices[0].key_count())
381
self.assertEqual(None, builder._backing_indices[1])
382
self.assertEqual(8, builder._backing_indices[2].key_count())
383
builder.add_node(*nodes[10])
384
builder.add_node(*nodes[11])
385
self.assertEqual(3, len(builder._backing_indices))
386
self.assertEqual(None, builder._backing_indices[0])
387
self.assertEqual(4, builder._backing_indices[1].key_count())
388
self.assertEqual(8, builder._backing_indices[2].key_count())
389
builder.add_node(*nodes[12])
390
# Test that memory and disk are both used for query methods; and that
391
# None is skipped over happily.
392
self.assertEqual([(builder,) + node for node in sorted(nodes[:13])],
393
list(builder.iter_all_entries()))
394
# Two nodes - one memory one disk
395
self.assertEqual(set([(builder,) + node for node in nodes[11:13]]),
396
set(builder.iter_entries([nodes[12][0], nodes[11][0]])))
397
self.assertEqual(13, builder.key_count())
398
self.assertEqual(set([(builder,) + node for node in nodes[11:13]]),
399
set(builder.iter_entries_prefix([nodes[12][0], nodes[11][0]])))
400
builder.add_node(*nodes[13])
401
self.assertEqual(3, len(builder._backing_indices))
402
self.assertEqual(2, builder._backing_indices[0].key_count())
403
self.assertEqual(4, builder._backing_indices[1].key_count())
404
self.assertEqual(8, builder._backing_indices[2].key_count())
405
builder.add_node(*nodes[14])
406
builder.add_node(*nodes[15])
407
self.assertEqual(4, len(builder._backing_indices))
408
self.assertEqual(None, builder._backing_indices[0])
409
self.assertEqual(None, builder._backing_indices[1])
410
self.assertEqual(None, builder._backing_indices[2])
411
self.assertEqual(16, builder._backing_indices[3].key_count())
412
# 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))
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
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())
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
transport = self.get_transport('')
482
size = transport.put_file('index', builder.finish())
483
index = btree_index.BTreeGraphIndex(transport, 'index', size)
484
nodes = list(index.iter_all_entries())
485
self.assertEqual(sorted(nodes), nodes)
486
self.assertEqual(16, len(nodes))
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
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)
505
def test_spill_index_stress_2_2(self):
506
# test that references and longer keys don't confuse things.
507
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2,
509
nodes = self.make_nodes(16, 2, 2)
510
builder.add_node(*nodes[0])
511
# Test the parts of the index that take up memory are doing so
513
self.assertEqual(1, len(builder._nodes))
514
self.assertIs(None, builder._nodes_by_key)
515
builder.add_node(*nodes[1])
516
self.assertEqual(0, len(builder._nodes))
517
self.assertIs(None, builder._nodes_by_key)
518
self.assertEqual(1, len(builder._backing_indices))
519
self.assertEqual(2, builder._backing_indices[0].key_count())
521
old = dict(builder._get_nodes_by_key()) #Build up the nodes by key dict
522
builder.add_node(*nodes[2])
523
self.assertEqual(1, len(builder._nodes))
524
self.assertIsNot(None, builder._nodes_by_key)
525
self.assertNotEqual({}, builder._nodes_by_key)
526
# We should have a new entry
527
self.assertNotEqual(old, builder._nodes_by_key)
528
# And spills to a second backing index combing all
529
builder.add_node(*nodes[3])
530
self.assertEqual(0, len(builder._nodes))
531
self.assertIs(None, builder._nodes_by_key)
532
self.assertEqual(2, len(builder._backing_indices))
533
self.assertEqual(None, builder._backing_indices[0])
534
self.assertEqual(4, builder._backing_indices[1].key_count())
535
# The next spills to the 2-len slot
536
builder.add_node(*nodes[4])
537
builder.add_node(*nodes[5])
538
self.assertEqual(0, len(builder._nodes))
539
self.assertIs(None, builder._nodes_by_key)
540
self.assertEqual(2, len(builder._backing_indices))
541
self.assertEqual(2, builder._backing_indices[0].key_count())
542
self.assertEqual(4, builder._backing_indices[1].key_count())
543
# Next spill combines
544
builder.add_node(*nodes[6])
545
builder.add_node(*nodes[7])
546
self.assertEqual(3, len(builder._backing_indices))
547
self.assertEqual(None, builder._backing_indices[0])
548
self.assertEqual(None, builder._backing_indices[1])
549
self.assertEqual(8, builder._backing_indices[2].key_count())
550
# And so forth - counting up in binary.
551
builder.add_node(*nodes[8])
552
builder.add_node(*nodes[9])
553
self.assertEqual(3, len(builder._backing_indices))
554
self.assertEqual(2, builder._backing_indices[0].key_count())
555
self.assertEqual(None, builder._backing_indices[1])
556
self.assertEqual(8, builder._backing_indices[2].key_count())
557
builder.add_node(*nodes[10])
558
builder.add_node(*nodes[11])
559
self.assertEqual(3, len(builder._backing_indices))
560
self.assertEqual(None, builder._backing_indices[0])
561
self.assertEqual(4, builder._backing_indices[1].key_count())
562
self.assertEqual(8, builder._backing_indices[2].key_count())
563
builder.add_node(*nodes[12])
564
# Test that memory and disk are both used for query methods; and that
565
# None is skipped over happily.
566
self.assertEqual([(builder,) + node for node in sorted(nodes[:13])],
567
list(builder.iter_all_entries()))
568
# Two nodes - one memory one disk
569
self.assertEqual(set([(builder,) + node for node in nodes[11:13]]),
570
set(builder.iter_entries([nodes[12][0], nodes[11][0]])))
571
self.assertEqual(13, builder.key_count())
572
self.assertEqual(set([(builder,) + node for node in nodes[11:13]]),
573
set(builder.iter_entries_prefix([nodes[12][0], nodes[11][0]])))
574
builder.add_node(*nodes[13])
575
self.assertEqual(3, len(builder._backing_indices))
576
self.assertEqual(2, builder._backing_indices[0].key_count())
577
self.assertEqual(4, builder._backing_indices[1].key_count())
578
self.assertEqual(8, builder._backing_indices[2].key_count())
579
builder.add_node(*nodes[14])
580
builder.add_node(*nodes[15])
581
self.assertEqual(4, len(builder._backing_indices))
582
self.assertEqual(None, builder._backing_indices[0])
583
self.assertEqual(None, builder._backing_indices[1])
584
self.assertEqual(None, builder._backing_indices[2])
585
self.assertEqual(16, builder._backing_indices[3].key_count())
586
# Now finish, and check we got a correctly ordered tree
587
transport = self.get_transport('')
588
size = transport.put_file('index', builder.finish())
589
index = btree_index.BTreeGraphIndex(transport, 'index', size)
590
nodes = list(index.iter_all_entries())
591
self.assertEqual(sorted(nodes), nodes)
592
self.assertEqual(16, len(nodes))
594
def test_spill_index_duplicate_key_caught_on_finish(self):
595
builder = btree_index.BTreeBuilder(key_elements=1, spill_at=2)
596
nodes = [node[0:2] for node in self.make_nodes(16, 1, 0)]
597
builder.add_node(*nodes[0])
598
builder.add_node(*nodes[1])
599
builder.add_node(*nodes[0])
600
self.assertRaises(errors.BadIndexDuplicateKey, builder.finish)
603
class TestBTreeIndex(BTreeTestCase):
605
def make_index(self, ref_lists=0, key_elements=1, nodes=[]):
606
builder = btree_index.BTreeBuilder(reference_lists=ref_lists,
607
key_elements=key_elements)
608
for key, value, references in nodes:
609
builder.add_node(key, value, references)
610
stream = builder.finish()
611
trans = transport.get_transport('trace+' + self.get_url())
612
size = trans.put_file('index', stream)
613
return btree_index.BTreeGraphIndex(trans, 'index', size)
615
def make_index_with_offset(self, ref_lists=1, key_elements=1, nodes=[],
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()
626
transport.put_bytes('index', (' '*offset)+content)
627
return btree_index.BTreeGraphIndex(transport, 'index', size=size,
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)
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
647
self.assertEqual(internal_node_pre_clear,
648
index._internal_node_cache.keys())
649
self.assertEqual(0, len(index._leaf_node_cache))
651
def test_trivial_constructor(self):
652
t = transport.get_transport('trace+' + self.get_url(''))
653
index = btree_index.BTreeGraphIndex(t, 'index', None)
654
# Checks the page size at load, but that isn't logged yet.
655
self.assertEqual([], t._activity)
657
def test_with_size_constructor(self):
658
t = transport.get_transport('trace+' + self.get_url(''))
659
index = btree_index.BTreeGraphIndex(t, 'index', 1)
660
# Checks the page size at load, but that isn't logged yet.
661
self.assertEqual([], t._activity)
663
def test_empty_key_count_no_size(self):
664
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)
669
self.assertEqual([], t._activity)
670
self.assertEqual(0, index.key_count())
671
# The entire index should have been requested (as we generally have the
672
# size available, and doing many small readvs is inappropriate).
673
# We can't tell how much was actually read here, but - check the code.
674
self.assertEqual([('get', 'index')], t._activity)
676
def test_empty_key_count(self):
677
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())
680
self.assertEqual(72, size)
681
index = btree_index.BTreeGraphIndex(t, 'index', size)
683
self.assertEqual([], t._activity)
684
self.assertEqual(0, index.key_count())
685
# The entire index should have been read, as 4K > size
686
self.assertEqual([('readv', 'index', [(0, 72)], False, None)],
689
def test_non_empty_key_count_2_2(self):
690
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
691
nodes = self.make_nodes(35, 2, 2)
693
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)
698
self.assertEqual([], t._activity)
699
self.assertEqual(70, index.key_count())
700
# The entire index should have been read, as it is one page long.
701
self.assertEqual([('readv', 'index', [(0, size)], False, None)],
703
self.assertEqual(1173, size)
705
def test_with_offset_no_size(self):
706
index = self.make_index_with_offset(key_elements=1, ref_lists=1,
708
nodes=self.make_nodes(200, 1, 1))
709
index._size = None # throw away the size info
710
self.assertEqual(200, index.key_count())
712
def test_with_small_offset(self):
713
index = self.make_index_with_offset(key_elements=1, ref_lists=1,
715
nodes=self.make_nodes(200, 1, 1))
716
self.assertEqual(200, index.key_count())
718
def test_with_large_offset(self):
719
index = self.make_index_with_offset(key_elements=1, ref_lists=1,
721
nodes=self.make_nodes(200, 1, 1))
722
self.assertEqual(200, index.key_count())
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())
732
self.assertEqual([('key',)], node.all_keys())
733
self.assertEqual([('get', 'index')], trans._activity)
735
def test__read_nodes_no_size_multiple_pages(self):
736
index = self.make_index(2, 2, nodes=self.make_nodes(160, 2, 2))
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())
746
def test_2_levels_key_count_2_2(self):
747
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
748
nodes = self.make_nodes(160, 2, 2)
750
builder.add_node(*node)
751
t = transport.get_transport('trace+' + self.get_url(''))
752
size = t.put_file('index', builder.finish())
753
self.assertEqual(17692, size)
754
index = btree_index.BTreeGraphIndex(t, 'index', size)
756
self.assertEqual([], t._activity)
757
self.assertEqual(320, index.key_count())
758
# The entire index should not have been read.
759
self.assertEqual([('readv', 'index', [(0, 4096)], False, None)],
762
def test_validate_one_page(self):
763
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
764
nodes = self.make_nodes(45, 2, 2)
766
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)
771
self.assertEqual([], t._activity)
773
# The entire index should have been read linearly.
774
self.assertEqual([('readv', 'index', [(0, size)], False, None)],
776
self.assertEqual(1488, size)
778
def test_validate_two_pages(self):
779
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
780
nodes = self.make_nodes(80, 2, 2)
782
builder.add_node(*node)
783
t = transport.get_transport('trace+' + self.get_url(''))
784
size = t.put_file('index', builder.finish())
785
# Root page, 2 leaf pages
786
self.assertEqual(9339, size)
787
index = btree_index.BTreeGraphIndex(t, 'index', size)
789
self.assertEqual([], t._activity)
791
# The entire index should have been read linearly.
793
[('readv', 'index', [(0, 4096)], False, None),
794
('readv', 'index', [(4096, 4096), (8192, 1147)], False, None)],
796
# XXX: TODO: write some badly-ordered nodes, and some pointers-to-wrong
797
# node and make validate find them.
799
def test_eq_ne(self):
800
# two indices are equal when constructed with the same parameters:
801
t1 = transport.get_transport('trace+' + self.get_url(''))
802
t2 = self.get_transport()
804
btree_index.BTreeGraphIndex(t1, 'index', None) ==
805
btree_index.BTreeGraphIndex(t1, 'index', None))
807
btree_index.BTreeGraphIndex(t1, 'index', 20) ==
808
btree_index.BTreeGraphIndex(t1, 'index', 20))
810
btree_index.BTreeGraphIndex(t1, 'index', 20) ==
811
btree_index.BTreeGraphIndex(t2, 'index', 20))
813
btree_index.BTreeGraphIndex(t1, 'inde1', 20) ==
814
btree_index.BTreeGraphIndex(t1, 'inde2', 20))
816
btree_index.BTreeGraphIndex(t1, 'index', 10) ==
817
btree_index.BTreeGraphIndex(t1, 'index', 20))
819
btree_index.BTreeGraphIndex(t1, 'index', None) !=
820
btree_index.BTreeGraphIndex(t1, 'index', None))
822
btree_index.BTreeGraphIndex(t1, 'index', 20) !=
823
btree_index.BTreeGraphIndex(t1, 'index', 20))
825
btree_index.BTreeGraphIndex(t1, 'index', 20) !=
826
btree_index.BTreeGraphIndex(t2, 'index', 20))
828
btree_index.BTreeGraphIndex(t1, 'inde1', 20) !=
829
btree_index.BTreeGraphIndex(t1, 'inde2', 20))
831
btree_index.BTreeGraphIndex(t1, 'index', 10) !=
832
btree_index.BTreeGraphIndex(t1, 'index', 20))
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)
839
self.assertEqual([(('key',), 'value')],
840
[x[1:] for x in index.iter_all_entries()])
841
self.assertEqual([('get', 'index')], t._activity)
843
def test_iter_all_entries_reads(self):
844
# iterating all entries reads the header, then does a linear
846
self.shrink_page_size()
847
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
848
# 20k nodes is enough to create a two internal nodes on the second
849
# level, with a 2K page size
850
nodes = self.make_nodes(10000, 2, 2)
852
builder.add_node(*node)
853
t = transport.get_transport('trace+' + self.get_url(''))
854
size = t.put_file('index', builder.finish())
855
self.assertEqual(1303220, size, 'number of expected bytes in the'
857
page_size = btree_index._PAGE_SIZE
859
index = btree_index.BTreeGraphIndex(t, 'index', size)
861
self.assertEqual([], t._activity)
862
found_nodes = self.time(list, index.iter_all_entries())
864
for node in found_nodes:
865
self.assertTrue(node[0] is index)
866
bare_nodes.append(node[1:])
867
self.assertEqual(3, len(index._row_lengths),
868
"Not enough rows: %r" % index._row_lengths)
869
# Should be as long as the nodes we supplied
870
self.assertEqual(20000, len(found_nodes))
871
# Should have the same content
872
self.assertEqual(set(nodes), set(bare_nodes))
873
# Should have done linear scan IO up the index, ignoring
874
# the internal nodes:
875
# The entire index should have been read
876
total_pages = sum(index._row_lengths)
877
self.assertEqual(total_pages, index._row_offsets[-1])
878
self.assertEqual(1303220, size)
879
# The start of the leaves
880
first_byte = index._row_offsets[-2] * page_size
882
for offset in range(first_byte, size, page_size):
883
readv_request.append((offset, page_size))
884
# The last page is truncated
885
readv_request[-1] = (readv_request[-1][0], 1303220 % page_size)
886
expected = [('readv', 'index', [(0, page_size)], False, None),
887
('readv', 'index', readv_request, False, None)]
888
if expected != t._activity:
889
self.assertEqualDiff(pprint.pformat(expected),
890
pprint.pformat(transport._activity))
892
def _test_iter_entries_references_resolved(self):
893
index = self.make_index(1, nodes=[
894
(('name', ), 'data', ([('ref', ), ('ref', )], )),
895
(('ref', ), 'refdata', ([], ))])
896
self.assertEqual(set([(index, ('name', ), 'data', ((('ref',),('ref',)),)),
897
(index, ('ref', ), 'refdata', ((), ))]),
898
set(index.iter_entries([('name',), ('ref',)])))
900
def test_iter_entries_references_2_refs_resolved(self):
901
# iterating some entries reads just the pages needed. For now, to
902
# get it working and start measuring, only 4K pages are read.
903
builder = btree_index.BTreeBuilder(key_elements=2, reference_lists=2)
904
# 80 nodes is enough to create a two-level index.
905
nodes = self.make_nodes(160, 2, 2)
907
builder.add_node(*node)
908
t = transport.get_transport('trace+' + self.get_url(''))
909
size = t.put_file('index', builder.finish())
911
index = btree_index.BTreeGraphIndex(t, 'index', size)
913
self.assertEqual([], t._activity)
915
found_nodes = list(index.iter_entries([nodes[30][0]]))
917
for node in found_nodes:
918
self.assertTrue(node[0] is index)
919
bare_nodes.append(node[1:])
920
# Should be as long as the nodes we supplied
921
self.assertEqual(1, len(found_nodes))
922
# Should have the same content
923
self.assertEqual(nodes[30], bare_nodes[0])
924
# Should have read the root node, then one leaf page:
925
self.assertEqual([('readv', 'index', [(0, 4096)], False, None),
926
('readv', 'index', [(8192, 4096), ], False, None)],
929
def test_iter_key_prefix_1_element_key_None(self):
930
index = self.make_index()
931
self.assertRaises(errors.BadIndexKey, list,
932
index.iter_entries_prefix([(None, )]))
934
def test_iter_key_prefix_wrong_length(self):
935
index = self.make_index()
936
self.assertRaises(errors.BadIndexKey, list,
937
index.iter_entries_prefix([('foo', None)]))
938
index = self.make_index(key_elements=2)
939
self.assertRaises(errors.BadIndexKey, list,
940
index.iter_entries_prefix([('foo', )]))
941
self.assertRaises(errors.BadIndexKey, list,
942
index.iter_entries_prefix([('foo', None, None)]))
944
def test_iter_key_prefix_1_key_element_no_refs(self):
945
index = self.make_index( nodes=[
946
(('name', ), 'data', ()),
947
(('ref', ), 'refdata', ())])
948
self.assertEqual(set([(index, ('name', ), 'data'),
949
(index, ('ref', ), 'refdata')]),
950
set(index.iter_entries_prefix([('name', ), ('ref', )])))
952
def test_iter_key_prefix_1_key_element_refs(self):
953
index = self.make_index(1, nodes=[
954
(('name', ), 'data', ([('ref', )], )),
955
(('ref', ), 'refdata', ([], ))])
956
self.assertEqual(set([(index, ('name', ), 'data', ((('ref',),),)),
957
(index, ('ref', ), 'refdata', ((), ))]),
958
set(index.iter_entries_prefix([('name', ), ('ref', )])))
960
def test_iter_key_prefix_2_key_element_no_refs(self):
961
index = self.make_index(key_elements=2, nodes=[
962
(('name', 'fin1'), 'data', ()),
963
(('name', 'fin2'), 'beta', ()),
964
(('ref', 'erence'), 'refdata', ())])
965
self.assertEqual(set([(index, ('name', 'fin1'), 'data'),
966
(index, ('ref', 'erence'), 'refdata')]),
967
set(index.iter_entries_prefix([('name', 'fin1'), ('ref', 'erence')])))
968
self.assertEqual(set([(index, ('name', 'fin1'), 'data'),
969
(index, ('name', 'fin2'), 'beta')]),
970
set(index.iter_entries_prefix([('name', None)])))
972
def test_iter_key_prefix_2_key_element_refs(self):
973
index = self.make_index(1, key_elements=2, nodes=[
974
(('name', 'fin1'), 'data', ([('ref', 'erence')], )),
975
(('name', 'fin2'), 'beta', ([], )),
976
(('ref', 'erence'), 'refdata', ([], ))])
977
self.assertEqual(set([(index, ('name', 'fin1'), 'data', ((('ref', 'erence'),),)),
978
(index, ('ref', 'erence'), 'refdata', ((), ))]),
979
set(index.iter_entries_prefix([('name', 'fin1'), ('ref', 'erence')])))
980
self.assertEqual(set([(index, ('name', 'fin1'), 'data', ((('ref', 'erence'),),)),
981
(index, ('name', 'fin2'), 'beta', ((), ))]),
982
set(index.iter_entries_prefix([('name', None)])))
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)
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))
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))
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))
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', ([],)),
1013
self.assertEqual(set([]), index.external_references(0))
1015
def test__find_ancestors_one_page(self):
1018
index = self.make_index(ref_lists=1, key_elements=1, nodes=[
1019
(key1, 'value', ([key2],)),
1020
(key2, 'value', ([],)),
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)
1029
def test__find_ancestors_one_page_w_missing(self):
1033
index = self.make_index(ref_lists=1, key_elements=1, nodes=[
1034
(key1, 'value', ([key2],)),
1035
(key2, 'value', ([],)),
1038
missing_keys = set()
1039
search_keys = index._find_ancestors([key2, key3], 0, parent_map,
1041
self.assertEqual({key2: ()}, parent_map)
1042
# we know that key3 is missing because we read the page that it would
1044
self.assertEqual(set([key3]), missing_keys)
1045
self.assertEqual(set(), search_keys)
1047
def test__find_ancestors_one_parent_missing(self):
1051
index = self.make_index(ref_lists=1, key_elements=1, nodes=[
1052
(key1, 'value', ([key2],)),
1053
(key2, 'value', ([key3],)),
1056
missing_keys = set()
1057
search_keys = index._find_ancestors([key1], 0, parent_map,
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,
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)
1072
def test__find_ancestors_dont_search_known(self):
1076
index = self.make_index(ref_lists=1, key_elements=1, nodes=[
1077
(key1, 'value', ([key2],)),
1078
(key2, 'value', ([key3],)),
1079
(key3, 'value', ([],)),
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,
1086
self.assertEqual({key1: (key2,), key2: (key3,)}, parent_map)
1087
self.assertEqual(set(), missing_keys)
1088
self.assertEqual(set(), search_keys)
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"
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))
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]))
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
1126
missing_keys = set()
1127
search_keys = index._find_ancestors([next_key], 0, parent_map,
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)
1133
search_keys = index._find_ancestors([max_l1_key], 0, parent_map,
1135
self.assertEqual(l1.all_keys(), sorted(parent_map))
1136
self.assertEqual(set(), missing_keys)
1137
self.assertEqual(set(), search_keys)
1139
def test__find_ancestors_empty_index(self):
1140
index = self.make_index(ref_lists=1, key_elements=1, nodes=[])
1142
missing_keys = set()
1143
search_keys = index._find_ancestors([('one',), ('two',)], 0, parent_map,
1145
self.assertEqual(set(), search_keys)
1146
self.assertEqual({}, parent_map)
1147
self.assertEqual(set([('one',), ('two',)]), missing_keys)
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)
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))
1187
class TestBTreeNodes(BTreeTestCase):
1189
scenarios = btreeparser_scenarios()
1192
BTreeTestCase.setUp(self)
1193
self.overrideAttr(btree_index, '_btree_serializer', self.parse_btree)
1195
def test_LeafNode_1_0(self):
1196
node_bytes = ("type=leaf\n"
1197
"0000000000000000000000000000000000000000\x00\x00value:0\n"
1198
"1111111111111111111111111111111111111111\x00\x00value:1\n"
1199
"2222222222222222222222222222222222222222\x00\x00value:2\n"
1200
"3333333333333333333333333333333333333333\x00\x00value:3\n"
1201
"4444444444444444444444444444444444444444\x00\x00value:4\n")
1202
node = btree_index._LeafNode(node_bytes, 1, 0)
1203
# We do direct access, or don't care about order, to leaf nodes most of
1204
# the time, so a dict is useful:
1206
("0000000000000000000000000000000000000000",): ("value:0", ()),
1207
("1111111111111111111111111111111111111111",): ("value:1", ()),
1208
("2222222222222222222222222222222222222222",): ("value:2", ()),
1209
("3333333333333333333333333333333333333333",): ("value:3", ()),
1210
("4444444444444444444444444444444444444444",): ("value:4", ()),
1211
}, dict(node.all_items()))
1213
def test_LeafNode_2_2(self):
1214
node_bytes = ("type=leaf\n"
1215
"00\x0000\x00\t00\x00ref00\x00value:0\n"
1216
"00\x0011\x0000\x00ref00\t00\x00ref00\r01\x00ref01\x00value:1\n"
1217
"11\x0033\x0011\x00ref22\t11\x00ref22\r11\x00ref22\x00value:3\n"
1218
"11\x0044\x00\t11\x00ref00\x00value:4\n"
1221
node = btree_index._LeafNode(node_bytes, 2, 2)
1222
# We do direct access, or don't care about order, to leaf nodes most of
1223
# the time, so a dict is useful:
1225
('00', '00'): ('value:0', ((), (('00', 'ref00'),))),
1226
('00', '11'): ('value:1',
1227
((('00', 'ref00'),), (('00', 'ref00'), ('01', 'ref01')))),
1228
('11', '33'): ('value:3',
1229
((('11', 'ref22'),), (('11', 'ref22'), ('11', 'ref22')))),
1230
('11', '44'): ('value:4', ((), (('11', 'ref00'),)))
1231
}, dict(node.all_items()))
1233
def test_InternalNode_1(self):
1234
node_bytes = ("type=internal\n"
1236
"0000000000000000000000000000000000000000\n"
1237
"1111111111111111111111111111111111111111\n"
1238
"2222222222222222222222222222222222222222\n"
1239
"3333333333333333333333333333333333333333\n"
1240
"4444444444444444444444444444444444444444\n"
1242
node = btree_index._InternalNode(node_bytes)
1243
# We want to bisect to find the right children from this node, so a
1244
# vector is most useful.
1246
("0000000000000000000000000000000000000000",),
1247
("1111111111111111111111111111111111111111",),
1248
("2222222222222222222222222222222222222222",),
1249
("3333333333333333333333333333333333333333",),
1250
("4444444444444444444444444444444444444444",),
1252
self.assertEqual(1, node.offset)
1254
def test_LeafNode_2_2(self):
1255
node_bytes = ("type=leaf\n"
1256
"00\x0000\x00\t00\x00ref00\x00value:0\n"
1257
"00\x0011\x0000\x00ref00\t00\x00ref00\r01\x00ref01\x00value:1\n"
1258
"11\x0033\x0011\x00ref22\t11\x00ref22\r11\x00ref22\x00value:3\n"
1259
"11\x0044\x00\t11\x00ref00\x00value:4\n"
1262
node = btree_index._LeafNode(node_bytes, 2, 2)
1263
# We do direct access, or don't care about order, to leaf nodes most of
1264
# the time, so a dict is useful:
1266
('00', '00'): ('value:0', ((), (('00', 'ref00'),))),
1267
('00', '11'): ('value:1',
1268
((('00', 'ref00'),), (('00', 'ref00'), ('01', 'ref01')))),
1269
('11', '33'): ('value:3',
1270
((('11', 'ref22'),), (('11', 'ref22'), ('11', 'ref22')))),
1271
('11', '44'): ('value:4', ((), (('11', 'ref00'),)))
1272
}, dict(node.all_items()))
1274
def assertFlattened(self, expected, key, value, refs):
1275
flat_key, flat_line = self.parse_btree._flatten_node(
1276
(None, key, value, refs), bool(refs))
1277
self.assertEqual('\x00'.join(key), flat_key)
1278
self.assertEqual(expected, flat_line)
1280
def test__flatten_node(self):
1281
self.assertFlattened('key\0\0value\n', ('key',), 'value', [])
1282
self.assertFlattened('key\0tuple\0\0value str\n',
1283
('key', 'tuple'), 'value str', [])
1284
self.assertFlattened('key\0tuple\0triple\0\0value str\n',
1285
('key', 'tuple', 'triple'), 'value str', [])
1286
self.assertFlattened('k\0t\0s\0ref\0value str\n',
1287
('k', 't', 's'), 'value str', [[('ref',)]])
1288
self.assertFlattened('key\0tuple\0ref\0key\0value str\n',
1289
('key', 'tuple'), 'value str', [[('ref', 'key')]])
1290
self.assertFlattened("00\x0000\x00\t00\x00ref00\x00value:0\n",
1291
('00', '00'), 'value:0', ((), (('00', 'ref00'),)))
1292
self.assertFlattened(
1293
"00\x0011\x0000\x00ref00\t00\x00ref00\r01\x00ref01\x00value:1\n",
1294
('00', '11'), 'value:1',
1295
((('00', 'ref00'),), (('00', 'ref00'), ('01', 'ref01'))))
1296
self.assertFlattened(
1297
"11\x0033\x0011\x00ref22\t11\x00ref22\r11\x00ref22\x00value:3\n",
1298
('11', '33'), 'value:3',
1299
((('11', 'ref22'),), (('11', 'ref22'), ('11', 'ref22'))))
1300
self.assertFlattened(
1301
"11\x0044\x00\t11\x00ref00\x00value:4\n",
1302
('11', '44'), 'value:4', ((), (('11', 'ref00'),)))
1305
class TestCompiledBtree(tests.TestCase):
1307
def test_exists(self):
1308
# This is just to let the user know if they don't have the feature
1310
self.requireFeature(compiled_btreeparser_feature)
1313
class TestMultiBisectRight(tests.TestCase):
1315
def assertMultiBisectRight(self, offsets, search_keys, fixed_keys):
1316
self.assertEqual(offsets,
1317
btree_index.BTreeGraphIndex._multi_bisect_right(
1318
search_keys, fixed_keys))
1320
def test_after(self):
1321
self.assertMultiBisectRight([(1, ['b'])], ['b'], ['a'])
1322
self.assertMultiBisectRight([(3, ['e', 'f', 'g'])],
1323
['e', 'f', 'g'], ['a', 'b', 'c'])
1325
def test_before(self):
1326
self.assertMultiBisectRight([(0, ['a'])], ['a'], ['b'])
1327
self.assertMultiBisectRight([(0, ['a', 'b', 'c', 'd'])],
1328
['a', 'b', 'c', 'd'], ['e', 'f', 'g'])
1330
def test_exact(self):
1331
self.assertMultiBisectRight([(1, ['a'])], ['a'], ['a'])
1332
self.assertMultiBisectRight([(1, ['a']), (2, ['b'])], ['a', 'b'], ['a', 'b'])
1333
self.assertMultiBisectRight([(1, ['a']), (3, ['c'])],
1334
['a', 'c'], ['a', 'b', 'c'])
1336
def test_inbetween(self):
1337
self.assertMultiBisectRight([(1, ['b'])], ['b'], ['a', 'c'])
1338
self.assertMultiBisectRight([(1, ['b', 'c', 'd']), (2, ['f', 'g'])],
1339
['b', 'c', 'd', 'f', 'g'], ['a', 'e', 'h'])
1341
def test_mixed(self):
1342
self.assertMultiBisectRight([(0, ['a', 'b']), (2, ['d', 'e']),
1344
['a', 'b', 'd', 'e', 'g', 'h'],
1345
['c', 'd', 'f', 'g'])
1348
class TestExpandOffsets(tests.TestCase):
1350
def make_index(self, size, recommended_pages=None):
1351
"""Make an index with a generic size.
1353
This doesn't actually create anything on disk, it just primes a
1354
BTreeGraphIndex with the recommended information.
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
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)
1367
index._get_offsets_to_cached_pages = _get_offsets_to_cached_pages
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)
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])
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])
1396
def assertNumPages(self, expected_pages, index, size):
1398
self.assertEqual(expected_pages, index._compute_total_pages_in_index())
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'
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)
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)
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))
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])
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])
1439
def test_read_all_from_root(self):
1440
index = self.make_index(4096*10, 20)
1441
self.assertExpandOffsets(range(10), index, [0])
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])
1454
def test_no_root_node(self):
1455
index = self.make_index(4096*10, 5)
1456
self.assertExpandOffsets([0], index, [0])
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'
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])
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,
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])
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])
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])
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])
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])
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])
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,
added
removed
bzrlib/tests/test_btree_index.py
