~bzr-pqm/bzr/bzr.dev

5752.3.8 by John Arbash Meinel
Merge bzr.dev 5764 to resolve release-notes (aka NEWS) conflicts
1
# Copyright (C) 2008-2011 Canonical Ltd
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
2
#
3
# This program is free software; you can redistribute it and/or modify
3641.3.29 by John Arbash Meinel
Cleanup the copyright headers
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.
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
7
#
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.
12
#
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
4183.7.1 by Sabin Iacob
update FSF mailing address
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
16
#
17
6379.6.7 by Jelmer Vernooij
Move importing from future until after doc string, otherwise the doc string will disappear.
18
"""B+Tree indices"""
19
6379.6.1 by Jelmer Vernooij
Import absolute_import in a few places.
20
from __future__ import absolute_import
21
4708.1.1 by John Arbash Meinel
Use a cStringIO.StringIO for 1-page btree indexes.
22
import cStringIO
5753.2.2 by Jelmer Vernooij
Remove some unnecessary imports, clean up lazy imports.
23
24
from bzrlib.lazy_import import lazy_import
25
lazy_import(globals(), """
5753.2.4 by Jelmer Vernooij
Review feedback from John.
26
import bisect
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
27
import math
28
import tempfile
29
import zlib
5753.2.2 by Jelmer Vernooij
Remove some unnecessary imports, clean up lazy imports.
30
""")
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
31
32
from bzrlib import (
33
    chunk_writer,
34
    debug,
35
    errors,
4208.1.2 by John Arbash Meinel
Switch to using a FIFOCache.
36
    fifo_cache,
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
37
    index,
38
    lru_cache,
39
    osutils,
4789.28.1 by John Arbash Meinel
Use StaticTuple as part of the builder process.
40
    static_tuple,
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
41
    trace,
5273.1.7 by Vincent Ladeuil
No more use of the get_transport imported *symbol*, all uses are through
42
    transport,
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
43
    )
44
from bzrlib.index import _OPTION_NODE_REFS, _OPTION_KEY_ELEMENTS, _OPTION_LEN
45
46
3641.3.3 by John Arbash Meinel
Change the header to indicate these indexes are
47
_BTSIGNATURE = "B+Tree Graph Index 2\n"
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
48
_OPTION_ROW_LENGTHS = "row_lengths="
49
_LEAF_FLAG = "type=leaf\n"
50
_INTERNAL_FLAG = "type=internal\n"
51
_INTERNAL_OFFSET = "offset="
52
53
_RESERVED_HEADER_BYTES = 120
54
_PAGE_SIZE = 4096
55
56
# 4K per page: 4MB - 1000 entries
57
_NODE_CACHE_SIZE = 1000
58
59
60
class _BuilderRow(object):
61
    """The stored state accumulated while writing out a row in the index.
62
63
    :ivar spool: A temporary file used to accumulate nodes for this row
64
        in the tree.
65
    :ivar nodes: The count of nodes emitted so far.
66
    """
67
68
    def __init__(self):
69
        """Create a _BuilderRow."""
70
        self.nodes = 0
4708.1.1 by John Arbash Meinel
Use a cStringIO.StringIO for 1-page btree indexes.
71
        self.spool = None# tempfile.TemporaryFile(prefix='bzr-index-row-')
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
72
        self.writer = None
73
74
    def finish_node(self, pad=True):
75
        byte_lines, _, padding = self.writer.finish()
76
        if self.nodes == 0:
4708.1.1 by John Arbash Meinel
Use a cStringIO.StringIO for 1-page btree indexes.
77
            self.spool = cStringIO.StringIO()
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
78
            # padded note:
79
            self.spool.write("\x00" * _RESERVED_HEADER_BYTES)
4708.1.1 by John Arbash Meinel
Use a cStringIO.StringIO for 1-page btree indexes.
80
        elif self.nodes == 1:
81
            # We got bigger than 1 node, switch to a temp file
82
            spool = tempfile.TemporaryFile(prefix='bzr-index-row-')
83
            spool.write(self.spool.getvalue())
84
            self.spool = spool
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
85
        skipped_bytes = 0
86
        if not pad and padding:
87
            del byte_lines[-1]
88
            skipped_bytes = padding
89
        self.spool.writelines(byte_lines)
3644.2.3 by John Arbash Meinel
Do a bit more work to get all the tests to pass.
90
        remainder = (self.spool.tell() + skipped_bytes) % _PAGE_SIZE
91
        if remainder != 0:
92
            raise AssertionError("incorrect node length: %d, %d"
93
                                 % (self.spool.tell(), remainder))
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
94
        self.nodes += 1
95
        self.writer = None
96
97
98
class _InternalBuilderRow(_BuilderRow):
99
    """The stored state accumulated while writing out internal rows."""
100
101
    def finish_node(self, pad=True):
102
        if not pad:
103
            raise AssertionError("Must pad internal nodes only.")
104
        _BuilderRow.finish_node(self)
105
106
107
class _LeafBuilderRow(_BuilderRow):
108
    """The stored state accumulated while writing out a leaf rows."""
109
110
111
class BTreeBuilder(index.GraphIndexBuilder):
112
    """A Builder for B+Tree based Graph indices.
113
114
    The resulting graph has the structure:
115
116
    _SIGNATURE OPTIONS NODES
117
    _SIGNATURE     := 'B+Tree Graph Index 1' NEWLINE
118
    OPTIONS        := REF_LISTS KEY_ELEMENTS LENGTH
119
    REF_LISTS      := 'node_ref_lists=' DIGITS NEWLINE
120
    KEY_ELEMENTS   := 'key_elements=' DIGITS NEWLINE
121
    LENGTH         := 'len=' DIGITS NEWLINE
122
    ROW_LENGTHS    := 'row_lengths' DIGITS (COMMA DIGITS)*
123
    NODES          := NODE_COMPRESSED*
124
    NODE_COMPRESSED:= COMPRESSED_BYTES{4096}
125
    NODE_RAW       := INTERNAL | LEAF
126
    INTERNAL       := INTERNAL_FLAG POINTERS
127
    LEAF           := LEAF_FLAG ROWS
128
    KEY_ELEMENT    := Not-whitespace-utf8
129
    KEY            := KEY_ELEMENT (NULL KEY_ELEMENT)*
130
    ROWS           := ROW*
131
    ROW            := KEY NULL ABSENT? NULL REFERENCES NULL VALUE NEWLINE
132
    ABSENT         := 'a'
133
    REFERENCES     := REFERENCE_LIST (TAB REFERENCE_LIST){node_ref_lists - 1}
134
    REFERENCE_LIST := (REFERENCE (CR REFERENCE)*)?
135
    REFERENCE      := KEY
136
    VALUE          := no-newline-no-null-bytes
137
    """
138
139
    def __init__(self, reference_lists=0, key_elements=1, spill_at=100000):
140
        """See GraphIndexBuilder.__init__.
141
142
        :param spill_at: Optional parameter controlling the maximum number
143
            of nodes that BTreeBuilder will hold in memory.
144
        """
145
        index.GraphIndexBuilder.__init__(self, reference_lists=reference_lists,
146
            key_elements=key_elements)
147
        self._spill_at = spill_at
148
        self._backing_indices = []
3644.2.11 by John Arbash Meinel
Document the new form of _nodes and remove an unnecessary cast.
149
        # A map of {key: (node_refs, value)}
150
        self._nodes = {}
3644.2.1 by John Arbash Meinel
Change the IndexBuilders to not generate the nodes_by_key unless needed.
151
        # Indicate it hasn't been built yet
152
        self._nodes_by_key = None
3777.5.2 by John Arbash Meinel
Change the name to ChunkWriter.set_optimize()
153
        self._optimize_for_size = False
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
154
155
    def add_node(self, key, value, references=()):
156
        """Add a node to the index.
157
158
        If adding the node causes the builder to reach its spill_at threshold,
159
        disk spilling will be triggered.
160
161
        :param key: The key. keys are non-empty tuples containing
162
            as many whitespace-free utf8 bytestrings as the key length
163
            defined for this index.
164
        :param references: An iterable of iterables of keys. Each is a
165
            reference to another key.
166
        :param value: The value to associate with the key. It may be any
5891.1.3 by Andrew Bennetts
Move docstring formatting fixes.
167
            bytes as long as it does not contain \\0 or \\n.
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
168
        """
4789.28.2 by John Arbash Meinel
Get rid of the GraphIndexBuilder/BTreeBuilder._keys attribute.
169
        # Ensure that 'key' is a StaticTuple
170
        key = static_tuple.StaticTuple.from_sequence(key).intern()
3644.2.9 by John Arbash Meinel
Refactor some code.
171
        # we don't care about absent_references
172
        node_refs, _ = self._check_key_ref_value(key, references, value)
3644.2.2 by John Arbash Meinel
the new btree index doesn't have 'absent' keys in its _nodes
173
        if key in self._nodes:
3644.2.1 by John Arbash Meinel
Change the IndexBuilders to not generate the nodes_by_key unless needed.
174
            raise errors.BadIndexDuplicateKey(key, self)
4789.28.1 by John Arbash Meinel
Use StaticTuple as part of the builder process.
175
        self._nodes[key] = static_tuple.StaticTuple(node_refs, value)
3644.2.9 by John Arbash Meinel
Refactor some code.
176
        if self._nodes_by_key is not None and self._key_length > 1:
177
            self._update_nodes_by_key(key, value, node_refs)
4789.28.2 by John Arbash Meinel
Get rid of the GraphIndexBuilder/BTreeBuilder._keys attribute.
178
        if len(self._nodes) < self._spill_at:
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
179
            return
3644.2.9 by John Arbash Meinel
Refactor some code.
180
        self._spill_mem_keys_to_disk()
181
182
    def _spill_mem_keys_to_disk(self):
183
        """Write the in memory keys down to disk to cap memory consumption.
184
185
        If we already have some keys written to disk, we will combine them so
186
        as to preserve the sorted order.  The algorithm for combining uses
187
        powers of two.  So on the first spill, write all mem nodes into a
188
        single index. On the second spill, combine the mem nodes with the nodes
189
        on disk to create a 2x sized disk index and get rid of the first index.
190
        On the third spill, create a single new disk index, which will contain
191
        the mem nodes, and preserve the existing 2x sized index.  On the fourth,
192
        combine mem with the first and second indexes, creating a new one of
193
        size 4x. On the fifth create a single new one, etc.
194
        """
4168.3.6 by John Arbash Meinel
Add 'combine_backing_indices' as a flag for GraphIndex.set_optimize().
195
        if self._combine_backing_indices:
4168.3.5 by John Arbash Meinel
Check that setting _combine_spilled_indices has the expected effect.
196
            (new_backing_file, size,
197
             backing_pos) = self._spill_mem_keys_and_combine()
198
        else:
199
            new_backing_file, size = self._spill_mem_keys_without_combining()
200
        # Note: The transport here isn't strictly needed, because we will use
201
        #       direct access to the new_backing._file object
6083.1.1 by Jelmer Vernooij
Use get_transport_from_{url,path} in more places.
202
        new_backing = BTreeGraphIndex(transport.get_transport_from_path('.'),
5273.1.7 by Vincent Ladeuil
No more use of the get_transport imported *symbol*, all uses are through
203
                                      '<temp>', size)
4168.3.5 by John Arbash Meinel
Check that setting _combine_spilled_indices has the expected effect.
204
        # GC will clean up the file
205
        new_backing._file = new_backing_file
4168.3.6 by John Arbash Meinel
Add 'combine_backing_indices' as a flag for GraphIndex.set_optimize().
206
        if self._combine_backing_indices:
4168.3.5 by John Arbash Meinel
Check that setting _combine_spilled_indices has the expected effect.
207
            if len(self._backing_indices) == backing_pos:
208
                self._backing_indices.append(None)
209
            self._backing_indices[backing_pos] = new_backing
210
            for backing_pos in range(backing_pos):
211
                self._backing_indices[backing_pos] = None
212
        else:
213
            self._backing_indices.append(new_backing)
214
        self._nodes = {}
215
        self._nodes_by_key = None
216
217
    def _spill_mem_keys_without_combining(self):
218
        return self._write_nodes(self._iter_mem_nodes(), allow_optimize=False)
219
220
    def _spill_mem_keys_and_combine(self):
4168.3.4 by John Arbash Meinel
Restore the ability to spill, but prepare a flag to disable it.
221
        iterators_to_combine = [self._iter_mem_nodes()]
222
        pos = -1
223
        for pos, backing in enumerate(self._backing_indices):
224
            if backing is None:
225
                pos -= 1
226
                break
227
            iterators_to_combine.append(backing.iter_all_entries())
228
        backing_pos = pos + 1
229
        new_backing_file, size = \
230
            self._write_nodes(self._iter_smallest(iterators_to_combine),
231
                              allow_optimize=False)
4168.3.5 by John Arbash Meinel
Check that setting _combine_spilled_indices has the expected effect.
232
        return new_backing_file, size, backing_pos
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
233
234
    def add_nodes(self, nodes):
235
        """Add nodes to the index.
236
237
        :param nodes: An iterable of (key, node_refs, value) entries to add.
238
        """
239
        if self.reference_lists:
240
            for (key, value, node_refs) in nodes:
241
                self.add_node(key, value, node_refs)
242
        else:
243
            for (key, value) in nodes:
244
                self.add_node(key, value)
245
246
    def _iter_mem_nodes(self):
247
        """Iterate over the nodes held in memory."""
3644.2.8 by John Arbash Meinel
Two quick tweaks.
248
        nodes = self._nodes
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
249
        if self.reference_lists:
3644.2.8 by John Arbash Meinel
Two quick tweaks.
250
            for key in sorted(nodes):
251
                references, value = nodes[key]
252
                yield self, key, value, references
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
253
        else:
3644.2.8 by John Arbash Meinel
Two quick tweaks.
254
            for key in sorted(nodes):
255
                references, value = nodes[key]
256
                yield self, key, value
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
257
258
    def _iter_smallest(self, iterators_to_combine):
3641.3.9 by John Arbash Meinel
Special case around _iter_smallest when we have only
259
        if len(iterators_to_combine) == 1:
260
            for value in iterators_to_combine[0]:
261
                yield value
262
            return
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
263
        current_values = []
264
        for iterator in iterators_to_combine:
265
            try:
266
                current_values.append(iterator.next())
267
            except StopIteration:
268
                current_values.append(None)
269
        last = None
270
        while True:
271
            # Decorate candidates with the value to allow 2.4's min to be used.
272
            candidates = [(item[1][1], item) for item
273
                in enumerate(current_values) if item[1] is not None]
274
            if not len(candidates):
275
                return
276
            selected = min(candidates)
277
            # undecorate back to (pos, node)
278
            selected = selected[1]
279
            if last == selected[1][1]:
280
                raise errors.BadIndexDuplicateKey(last, self)
281
            last = selected[1][1]
282
            # Yield, with self as the index
283
            yield (self,) + selected[1][1:]
284
            pos = selected[0]
285
            try:
286
                current_values[pos] = iterators_to_combine[pos].next()
287
            except StopIteration:
288
                current_values[pos] = None
289
4168.2.1 by John Arbash Meinel
Disable optimizations when spilling content to disk.
290
    def _add_key(self, string_key, line, rows, allow_optimize=True):
3641.3.8 by John Arbash Meinel
Move the add_key helper function into a separate func
291
        """Add a key to the current chunk.
292
293
        :param string_key: The key to add.
3641.3.11 by John Arbash Meinel
Start working on an alternate way to track compressed_chunk state.
294
        :param line: The fully serialised key and value.
4168.2.1 by John Arbash Meinel
Disable optimizations when spilling content to disk.
295
        :param allow_optimize: If set to False, prevent setting the optimize
296
            flag when writing out. This is used by the _spill_mem_keys_to_disk
297
            functionality.
3641.3.8 by John Arbash Meinel
Move the add_key helper function into a separate func
298
        """
6178.2.9 by Shannon Weyrick
A version of the patch, based on suggestions from John Meinel, which detects an empty page differently to avoid false positives.
299
        new_leaf = False
3641.3.8 by John Arbash Meinel
Move the add_key helper function into a separate func
300
        if rows[-1].writer is None:
301
            # opening a new leaf chunk;
6178.2.9 by Shannon Weyrick
A version of the patch, based on suggestions from John Meinel, which detects an empty page differently to avoid false positives.
302
            new_leaf = True
3641.3.8 by John Arbash Meinel
Move the add_key helper function into a separate func
303
            for pos, internal_row in enumerate(rows[:-1]):
304
                # flesh out any internal nodes that are needed to
3641.3.11 by John Arbash Meinel
Start working on an alternate way to track compressed_chunk state.
305
                # preserve the height of the tree
3641.3.8 by John Arbash Meinel
Move the add_key helper function into a separate func
306
                if internal_row.writer is None:
307
                    length = _PAGE_SIZE
308
                    if internal_row.nodes == 0:
309
                        length -= _RESERVED_HEADER_BYTES # padded
4168.2.1 by John Arbash Meinel
Disable optimizations when spilling content to disk.
310
                    if allow_optimize:
311
                        optimize_for_size = self._optimize_for_size
312
                    else:
313
                        optimize_for_size = False
3777.5.2 by John Arbash Meinel
Change the name to ChunkWriter.set_optimize()
314
                    internal_row.writer = chunk_writer.ChunkWriter(length, 0,
4168.2.1 by John Arbash Meinel
Disable optimizations when spilling content to disk.
315
                        optimize_for_size=optimize_for_size)
3641.3.8 by John Arbash Meinel
Move the add_key helper function into a separate func
316
                    internal_row.writer.write(_INTERNAL_FLAG)
317
                    internal_row.writer.write(_INTERNAL_OFFSET +
318
                        str(rows[pos + 1].nodes) + "\n")
319
            # add a new leaf
320
            length = _PAGE_SIZE
321
            if rows[-1].nodes == 0:
322
                length -= _RESERVED_HEADER_BYTES # padded
3777.5.2 by John Arbash Meinel
Change the name to ChunkWriter.set_optimize()
323
            rows[-1].writer = chunk_writer.ChunkWriter(length,
324
                optimize_for_size=self._optimize_for_size)
3641.3.8 by John Arbash Meinel
Move the add_key helper function into a separate func
325
            rows[-1].writer.write(_LEAF_FLAG)
3641.3.11 by John Arbash Meinel
Start working on an alternate way to track compressed_chunk state.
326
        if rows[-1].writer.write(line):
6178.2.9 by Shannon Weyrick
A version of the patch, based on suggestions from John Meinel, which detects an empty page differently to avoid false positives.
327
            # if we failed to write, despite having an empty page to write to,
328
            # then line is too big. raising the error avoids infinite recursion
329
            # searching for a suitably large page that will not be found.
330
            if new_leaf:
331
                raise errors.BadIndexKey(string_key)
3641.3.8 by John Arbash Meinel
Move the add_key helper function into a separate func
332
            # this key did not fit in the node:
333
            rows[-1].finish_node()
3641.3.11 by John Arbash Meinel
Start working on an alternate way to track compressed_chunk state.
334
            key_line = string_key + "\n"
3641.3.8 by John Arbash Meinel
Move the add_key helper function into a separate func
335
            new_row = True
336
            for row in reversed(rows[:-1]):
337
                # Mark the start of the next node in the node above. If it
4031.3.1 by Frank Aspell
Fixing various typos
338
                # doesn't fit then propagate upwards until we find one that
3641.3.8 by John Arbash Meinel
Move the add_key helper function into a separate func
339
                # it does fit into.
3641.3.11 by John Arbash Meinel
Start working on an alternate way to track compressed_chunk state.
340
                if row.writer.write(key_line):
3641.3.8 by John Arbash Meinel
Move the add_key helper function into a separate func
341
                    row.finish_node()
342
                else:
343
                    # We've found a node that can handle the pointer.
344
                    new_row = False
345
                    break
346
            # If we reached the current root without being able to mark the
347
            # division point, then we need a new root:
348
            if new_row:
349
                # We need a new row
350
                if 'index' in debug.debug_flags:
351
                    trace.mutter('Inserting new global row.')
352
                new_row = _InternalBuilderRow()
353
                reserved_bytes = 0
354
                rows.insert(0, new_row)
355
                # This will be padded, hence the -100
356
                new_row.writer = chunk_writer.ChunkWriter(
357
                    _PAGE_SIZE - _RESERVED_HEADER_BYTES,
3777.5.2 by John Arbash Meinel
Change the name to ChunkWriter.set_optimize()
358
                    reserved_bytes,
359
                    optimize_for_size=self._optimize_for_size)
3641.3.8 by John Arbash Meinel
Move the add_key helper function into a separate func
360
                new_row.writer.write(_INTERNAL_FLAG)
361
                new_row.writer.write(_INTERNAL_OFFSET +
362
                    str(rows[1].nodes - 1) + "\n")
3641.3.11 by John Arbash Meinel
Start working on an alternate way to track compressed_chunk state.
363
                new_row.writer.write(key_line)
6178.2.4 by Shannon Weyrick
raise BadIndexKey instead of skipping
364
            self._add_key(string_key, line, rows, allow_optimize=allow_optimize)
3641.3.8 by John Arbash Meinel
Move the add_key helper function into a separate func
365
4168.2.1 by John Arbash Meinel
Disable optimizations when spilling content to disk.
366
    def _write_nodes(self, node_iterator, allow_optimize=True):
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
367
        """Write node_iterator out as a B+Tree.
368
369
        :param node_iterator: An iterator of sorted nodes. Each node should
370
            match the output given by iter_all_entries.
4168.2.1 by John Arbash Meinel
Disable optimizations when spilling content to disk.
371
        :param allow_optimize: If set to False, prevent setting the optimize
372
            flag when writing out. This is used by the _spill_mem_keys_to_disk
373
            functionality.
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
374
        :return: A file handle for a temporary file containing a B+Tree for
375
            the nodes.
376
        """
377
        # The index rows - rows[0] is the root, rows[1] is the layer under it
378
        # etc.
379
        rows = []
380
        # forward sorted by key. In future we may consider topological sorting,
381
        # at the cost of table scans for direct lookup, or a second index for
382
        # direct lookup
383
        key_count = 0
384
        # A stack with the number of nodes of each size. 0 is the root node
385
        # and must always be 1 (if there are any nodes in the tree).
386
        self.row_lengths = []
387
        # Loop over all nodes adding them to the bottom row
388
        # (rows[-1]). When we finish a chunk in a row,
4031.3.1 by Frank Aspell
Fixing various typos
389
        # propagate the key that didn't fit (comes after the chunk) to the
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
390
        # row above, transitively.
391
        for node in node_iterator:
392
            if key_count == 0:
393
                # First key triggers the first row
394
                rows.append(_LeafBuilderRow())
6178.2.4 by Shannon Weyrick
raise BadIndexKey instead of skipping
395
            key_count += 1
3641.3.30 by John Arbash Meinel
Rename _parse_btree to _btree_serializer
396
            string_key, line = _btree_serializer._flatten_node(node,
397
                                    self.reference_lists)
6178.2.4 by Shannon Weyrick
raise BadIndexKey instead of skipping
398
            self._add_key(string_key, line, rows, allow_optimize=allow_optimize)
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
399
        for row in reversed(rows):
400
            pad = (type(row) != _LeafBuilderRow)
401
            row.finish_node(pad=pad)
402
        lines = [_BTSIGNATURE]
403
        lines.append(_OPTION_NODE_REFS + str(self.reference_lists) + '\n')
404
        lines.append(_OPTION_KEY_ELEMENTS + str(self._key_length) + '\n')
405
        lines.append(_OPTION_LEN + str(key_count) + '\n')
406
        row_lengths = [row.nodes for row in rows]
407
        lines.append(_OPTION_ROW_LENGTHS + ','.join(map(str, row_lengths)) + '\n')
4708.1.1 by John Arbash Meinel
Use a cStringIO.StringIO for 1-page btree indexes.
408
        if row_lengths and row_lengths[-1] > 1:
409
            result = tempfile.NamedTemporaryFile(prefix='bzr-index-')
410
        else:
411
            result = cStringIO.StringIO()
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
412
        result.writelines(lines)
413
        position = sum(map(len, lines))
414
        root_row = True
415
        if position > _RESERVED_HEADER_BYTES:
416
            raise AssertionError("Could not fit the header in the"
417
                                 " reserved space: %d > %d"
418
                                 % (position, _RESERVED_HEADER_BYTES))
419
        # write the rows out:
420
        for row in rows:
421
            reserved = _RESERVED_HEADER_BYTES # reserved space for first node
422
            row.spool.flush()
423
            row.spool.seek(0)
424
            # copy nodes to the finalised file.
425
            # Special case the first node as it may be prefixed
426
            node = row.spool.read(_PAGE_SIZE)
427
            result.write(node[reserved:])
4771.3.1 by John Arbash Meinel
We don't have to pad 'short' records.
428
            if len(node) == _PAGE_SIZE:
429
                result.write("\x00" * (reserved - position))
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
430
            position = 0 # Only the root row actually has an offset
431
            copied_len = osutils.pumpfile(row.spool, result)
432
            if copied_len != (row.nodes - 1) * _PAGE_SIZE:
433
                if type(row) != _LeafBuilderRow:
3644.2.3 by John Arbash Meinel
Do a bit more work to get all the tests to pass.
434
                    raise AssertionError("Incorrect amount of data copied"
435
                        " expected: %d, got: %d"
436
                        % ((row.nodes - 1) * _PAGE_SIZE,
437
                           copied_len))
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
438
        result.flush()
439
        size = result.tell()
440
        result.seek(0)
441
        return result, size
442
443
    def finish(self):
444
        """Finalise the index.
445
446
        :return: A file handle for a temporary file containing the nodes added
447
            to the index.
448
        """
449
        return self._write_nodes(self.iter_all_entries())[0]
450
451
    def iter_all_entries(self):
452
        """Iterate over all keys within the index
453
4343.2.2 by John Arbash Meinel
Fix an important doc bug about the api of iter_all_entries()
454
        :return: An iterable of (index, key, value, reference_lists). There is
455
            no defined order for the result iteration - it will be in the most
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
456
            efficient order for the index (in this case dictionary hash order).
457
        """
458
        if 'evil' in debug.debug_flags:
459
            trace.mutter_callsite(3,
460
                "iter_all_entries scales with size of history.")
461
        # Doing serial rather than ordered would be faster; but this shouldn't
462
        # be getting called routinely anyway.
3644.2.8 by John Arbash Meinel
Two quick tweaks.
463
        iterators = [self._iter_mem_nodes()]
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
464
        for backing in self._backing_indices:
465
            if backing is not None:
466
                iterators.append(backing.iter_all_entries())
3641.3.9 by John Arbash Meinel
Special case around _iter_smallest when we have only
467
        if len(iterators) == 1:
468
            return iterators[0]
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
469
        return self._iter_smallest(iterators)
470
471
    def iter_entries(self, keys):
472
        """Iterate over keys within the index.
473
474
        :param keys: An iterable providing the keys to be retrieved.
475
        :return: An iterable of (index, key, value, reference_lists). There is no
476
            defined order for the result iteration - it will be in the most
477
            efficient order for the index (keys iteration order in this case).
478
        """
479
        keys = set(keys)
4789.28.2 by John Arbash Meinel
Get rid of the GraphIndexBuilder/BTreeBuilder._keys attribute.
480
        # Note: We don't use keys.intersection() here. If you read the C api,
481
        #       set.intersection(other) special cases when other is a set and
482
        #       will iterate the smaller of the two and lookup in the other.
483
        #       It does *not* do this for any other type (even dict, unlike
484
        #       some other set functions.) Since we expect keys is generally <<
485
        #       self._nodes, it is faster to iterate over it in a list
486
        #       comprehension
487
        nodes = self._nodes
488
        local_keys = [key for key in keys if key in nodes]
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
489
        if self.reference_lists:
3847.2.2 by John Arbash Meinel
Rather than skipping the difference_update entirely, just restrict it to the intersection keys.
490
            for key in local_keys:
4789.28.2 by John Arbash Meinel
Get rid of the GraphIndexBuilder/BTreeBuilder._keys attribute.
491
                node = nodes[key]
3644.2.1 by John Arbash Meinel
Change the IndexBuilders to not generate the nodes_by_key unless needed.
492
                yield self, key, node[1], node[0]
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
493
        else:
3847.2.2 by John Arbash Meinel
Rather than skipping the difference_update entirely, just restrict it to the intersection keys.
494
            for key in local_keys:
4789.28.2 by John Arbash Meinel
Get rid of the GraphIndexBuilder/BTreeBuilder._keys attribute.
495
                node = nodes[key]
3644.2.1 by John Arbash Meinel
Change the IndexBuilders to not generate the nodes_by_key unless needed.
496
                yield self, key, node[1]
3847.2.1 by John Arbash Meinel
Shortcut BTreeBuilder.iter_entries when there are no backing indices.
497
        # Find things that are in backing indices that have not been handled
498
        # yet.
3847.2.3 by John Arbash Meinel
Bring back the shortcut
499
        if not self._backing_indices:
500
            return # We won't find anything there either
3847.2.2 by John Arbash Meinel
Rather than skipping the difference_update entirely, just restrict it to the intersection keys.
501
        # Remove all of the keys that we found locally
502
        keys.difference_update(local_keys)
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
503
        for backing in self._backing_indices:
504
            if backing is None:
505
                continue
506
            if not keys:
507
                return
508
            for node in backing.iter_entries(keys):
509
                keys.remove(node[1])
510
                yield (self,) + node[1:]
511
512
    def iter_entries_prefix(self, keys):
513
        """Iterate over keys within the index using prefix matching.
514
515
        Prefix matching is applied within the tuple of a key, not to within
516
        the bytestring of each key element. e.g. if you have the keys ('foo',
517
        'bar'), ('foobar', 'gam') and do a prefix search for ('foo', None) then
518
        only the former key is returned.
519
520
        :param keys: An iterable providing the key prefixes to be retrieved.
521
            Each key prefix takes the form of a tuple the length of a key, but
522
            with the last N elements 'None' rather than a regular bytestring.
523
            The first element cannot be 'None'.
524
        :return: An iterable as per iter_all_entries, but restricted to the
525
            keys with a matching prefix to those supplied. No additional keys
526
            will be returned, and every match that is in the index will be
527
            returned.
528
        """
529
        # XXX: To much duplication with the GraphIndex class; consider finding
530
        # a good place to pull out the actual common logic.
531
        keys = set(keys)
532
        if not keys:
533
            return
534
        for backing in self._backing_indices:
535
            if backing is None:
536
                continue
537
            for node in backing.iter_entries_prefix(keys):
538
                yield (self,) + node[1:]
539
        if self._key_length == 1:
540
            for key in keys:
541
                # sanity check
542
                if key[0] is None:
543
                    raise errors.BadIndexKey(key)
544
                if len(key) != self._key_length:
545
                    raise errors.BadIndexKey(key)
546
                try:
547
                    node = self._nodes[key]
548
                except KeyError:
549
                    continue
550
                if self.reference_lists:
3644.2.1 by John Arbash Meinel
Change the IndexBuilders to not generate the nodes_by_key unless needed.
551
                    yield self, key, node[1], node[0]
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
552
                else:
3644.2.1 by John Arbash Meinel
Change the IndexBuilders to not generate the nodes_by_key unless needed.
553
                    yield self, key, node[1]
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
554
            return
555
        for key in keys:
556
            # sanity check
557
            if key[0] is None:
558
                raise errors.BadIndexKey(key)
559
            if len(key) != self._key_length:
560
                raise errors.BadIndexKey(key)
561
            # find what it refers to:
3644.2.1 by John Arbash Meinel
Change the IndexBuilders to not generate the nodes_by_key unless needed.
562
            key_dict = self._get_nodes_by_key()
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
563
            elements = list(key)
564
            # find the subdict to return
565
            try:
566
                while len(elements) and elements[0] is not None:
567
                    key_dict = key_dict[elements[0]]
568
                    elements.pop(0)
569
            except KeyError:
570
                # a non-existant lookup.
571
                continue
572
            if len(elements):
573
                dicts = [key_dict]
574
                while dicts:
575
                    key_dict = dicts.pop(-1)
576
                    # can't be empty or would not exist
577
                    item, value = key_dict.iteritems().next()
578
                    if type(value) == dict:
579
                        # push keys
580
                        dicts.extend(key_dict.itervalues())
581
                    else:
582
                        # yield keys
583
                        for value in key_dict.itervalues():
5088.1.1 by Jelmer Vernooij
Force value to a tuple before concatenating with tuple that contains variables
584
                            yield (self, ) + tuple(value)
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
585
            else:
586
                yield (self, ) + key_dict
587
3644.2.1 by John Arbash Meinel
Change the IndexBuilders to not generate the nodes_by_key unless needed.
588
    def _get_nodes_by_key(self):
589
        if self._nodes_by_key is None:
590
            nodes_by_key = {}
591
            if self.reference_lists:
592
                for key, (references, value) in self._nodes.iteritems():
593
                    key_dict = nodes_by_key
594
                    for subkey in key[:-1]:
595
                        key_dict = key_dict.setdefault(subkey, {})
596
                    key_dict[key[-1]] = key, value, references
597
            else:
598
                for key, (references, value) in self._nodes.iteritems():
599
                    key_dict = nodes_by_key
600
                    for subkey in key[:-1]:
601
                        key_dict = key_dict.setdefault(subkey, {})
602
                    key_dict[key[-1]] = key, value
603
            self._nodes_by_key = nodes_by_key
604
        return self._nodes_by_key
605
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
606
    def key_count(self):
607
        """Return an estimate of the number of keys in this index.
608
609
        For InMemoryGraphIndex the estimate is exact.
610
        """
4789.28.2 by John Arbash Meinel
Get rid of the GraphIndexBuilder/BTreeBuilder._keys attribute.
611
        return len(self._nodes) + sum(backing.key_count() for backing in
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
612
            self._backing_indices if backing is not None)
613
614
    def validate(self):
615
        """In memory index's have no known corruption at the moment."""
616
617
5365.5.12 by John Arbash Meinel
Make _LeafNode inherit from dict (is-a rather than have-a)
618
class _LeafNode(dict):
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
619
    """A leaf node for a serialised B+Tree index."""
620
5365.5.23 by John Arbash Meinel
A __sizeof__ check that ensure we are getting what we are looking for.
621
    __slots__ = ('min_key', 'max_key', '_keys')
4274.1.2 by John Arbash Meinel
Add slots to _LeafNode and _InternalNode.
622
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
623
    def __init__(self, bytes, key_length, ref_list_length):
624
        """Parse bytes to create a leaf node object."""
625
        # splitlines mangles the \r delimiters.. don't use it.
4593.4.2 by John Arbash Meinel
Removing the min(keys) and max(keys) calls saves 100ms in the inner loop
626
        key_list = _btree_serializer._parse_leaf_lines(bytes,
627
            key_length, ref_list_length)
628
        if key_list:
4593.4.4 by John Arbash Meinel
Trying out a few more tweaks.
629
            self.min_key = key_list[0][0]
630
            self.max_key = key_list[-1][0]
4593.4.2 by John Arbash Meinel
Removing the min(keys) and max(keys) calls saves 100ms in the inner loop
631
        else:
632
            self.min_key = self.max_key = None
5365.5.12 by John Arbash Meinel
Make _LeafNode inherit from dict (is-a rather than have-a)
633
        super(_LeafNode, self).__init__(key_list)
5365.5.23 by John Arbash Meinel
A __sizeof__ check that ensure we are getting what we are looking for.
634
        self._keys = dict(self)
5365.5.1 by John Arbash Meinel
Implement a custom parser for chk btree leaves.
635
636
    def all_items(self):
637
        """Return a sorted list of (key, (value, refs)) items"""
5365.5.12 by John Arbash Meinel
Make _LeafNode inherit from dict (is-a rather than have-a)
638
        items = self.items()
5365.5.1 by John Arbash Meinel
Implement a custom parser for chk btree leaves.
639
        items.sort()
640
        return items
641
642
    def all_keys(self):
643
        """Return a sorted list of all keys."""
5365.5.12 by John Arbash Meinel
Make _LeafNode inherit from dict (is-a rather than have-a)
644
        keys = self.keys()
5365.5.1 by John Arbash Meinel
Implement a custom parser for chk btree leaves.
645
        keys.sort()
646
        return keys
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
647
648
649
class _InternalNode(object):
650
    """An internal node for a serialised B+Tree index."""
651
4274.1.2 by John Arbash Meinel
Add slots to _LeafNode and _InternalNode.
652
    __slots__ = ('keys', 'offset')
653
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
654
    def __init__(self, bytes):
655
        """Parse bytes to create an internal node object."""
656
        # splitlines mangles the \r delimiters.. don't use it.
657
        self.keys = self._parse_lines(bytes.split('\n'))
658
659
    def _parse_lines(self, lines):
660
        nodes = []
661
        self.offset = int(lines[1][7:])
4789.28.1 by John Arbash Meinel
Use StaticTuple as part of the builder process.
662
        as_st = static_tuple.StaticTuple.from_sequence
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
663
        for line in lines[2:]:
664
            if line == '':
665
                break
4789.28.1 by John Arbash Meinel
Use StaticTuple as part of the builder process.
666
            nodes.append(as_st(map(intern, line.split('\0'))).intern())
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
667
        return nodes
668
669
670
class BTreeGraphIndex(object):
671
    """Access to nodes via the standard GraphIndex interface for B+Tree's.
672
673
    Individual nodes are held in a LRU cache. This holds the root node in
674
    memory except when very large walks are done.
675
    """
676
5074.4.1 by John Arbash Meinel
Add an offset flag to BTreeGraphIndex.
677
    def __init__(self, transport, name, size, unlimited_cache=False,
678
                 offset=0):
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
679
        """Create a B+Tree index object on the index name.
680
681
        :param transport: The transport to read data for the index from.
682
        :param name: The file name of the index on transport.
683
        :param size: Optional size of the index in bytes. This allows
684
            compatibility with the GraphIndex API, as well as ensuring that
685
            the initial read (to read the root node header) can be done
686
            without over-reading even on empty indices, and on small indices
687
            allows single-IO to read the entire index.
4634.71.1 by John Arbash Meinel
Work around bug #402623 by allowing BTreeGraphIndex(...,unlimited_cache=True).
688
        :param unlimited_cache: If set to True, then instead of using an
689
            LRUCache with size _NODE_CACHE_SIZE, we will use a dict and always
690
            cache all leaf nodes.
5074.4.1 by John Arbash Meinel
Add an offset flag to BTreeGraphIndex.
691
        :param offset: The start of the btree index data isn't byte 0 of the
692
            file. Instead it starts at some point later.
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
693
        """
694
        self._transport = transport
695
        self._name = name
696
        self._size = size
697
        self._file = None
3763.8.7 by John Arbash Meinel
A bit of doc updates, start putting in tests for current behavior.
698
        self._recommended_pages = self._compute_recommended_pages()
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
699
        self._root_node = None
5074.4.1 by John Arbash Meinel
Add an offset flag to BTreeGraphIndex.
700
        self._base_offset = offset
5365.5.18 by John Arbash Meinel
Expose the new leaf node factory across the stack.
701
        self._leaf_factory = _LeafNode
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
702
        # Default max size is 100,000 leave values
703
        self._leaf_value_cache = None # lru_cache.LRUCache(100*1000)
4634.71.1 by John Arbash Meinel
Work around bug #402623 by allowing BTreeGraphIndex(...,unlimited_cache=True).
704
        if unlimited_cache:
705
            self._leaf_node_cache = {}
706
            self._internal_node_cache = {}
707
        else:
708
            self._leaf_node_cache = lru_cache.LRUCache(_NODE_CACHE_SIZE)
709
            # We use a FIFO here just to prevent possible blowout. However, a
710
            # 300k record btree has only 3k leaf nodes, and only 20 internal
711
            # nodes. A value of 100 scales to ~100*100*100 = 1M records.
712
            self._internal_node_cache = fifo_cache.FIFOCache(100)
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
713
        self._key_count = None
714
        self._row_lengths = None
715
        self._row_offsets = None # Start of each row, [-1] is the end
716
717
    def __eq__(self, other):
718
        """Equal when self and other were created with the same parameters."""
719
        return (
720
            type(self) == type(other) and
721
            self._transport == other._transport and
722
            self._name == other._name and
723
            self._size == other._size)
724
725
    def __ne__(self, other):
726
        return not self.__eq__(other)
727
3763.8.12 by John Arbash Meinel
Code cleanup.
728
    def _get_and_cache_nodes(self, nodes):
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
729
        """Read nodes and cache them in the lru.
730
731
        The nodes list supplied is sorted and then read from disk, each node
732
        being inserted it into the _node_cache.
733
734
        Note: Asking for more nodes than the _node_cache can contain will
735
        result in some of the results being immediately discarded, to prevent
736
        this an assertion is raised if more nodes are asked for than are
737
        cachable.
738
739
        :return: A dict of {node_pos: node}
740
        """
741
        found = {}
3763.8.1 by John Arbash Meinel
Playing around with expanding requests for btree index nodes into neighboring nodes.
742
        start_of_leaves = None
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
743
        for node_pos, node in self._read_nodes(sorted(nodes)):
744
            if node_pos == 0: # Special case
745
                self._root_node = node
746
            else:
3763.8.1 by John Arbash Meinel
Playing around with expanding requests for btree index nodes into neighboring nodes.
747
                if start_of_leaves is None:
748
                    start_of_leaves = self._row_offsets[-2]
749
                if node_pos < start_of_leaves:
4634.71.2 by John Arbash Meinel
If we are going to sometimes use a dict, we have to conform to just the dict interface.
750
                    self._internal_node_cache[node_pos] = node
3763.8.1 by John Arbash Meinel
Playing around with expanding requests for btree index nodes into neighboring nodes.
751
                else:
4634.71.2 by John Arbash Meinel
If we are going to sometimes use a dict, we have to conform to just the dict interface.
752
                    self._leaf_node_cache[node_pos] = node
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
753
            found[node_pos] = node
754
        return found
755
3763.8.7 by John Arbash Meinel
A bit of doc updates, start putting in tests for current behavior.
756
    def _compute_recommended_pages(self):
3763.8.15 by John Arbash Meinel
Review comments from Martin. Code clarity/variable name/docstring updates.
757
        """Convert transport's recommended_page_size into btree pages.
758
759
        recommended_page_size is in bytes, we want to know how many _PAGE_SIZE
760
        pages fit in that length.
761
        """
3763.8.7 by John Arbash Meinel
A bit of doc updates, start putting in tests for current behavior.
762
        recommended_read = self._transport.recommended_page_size()
763
        recommended_pages = int(math.ceil(recommended_read /
764
                                          float(_PAGE_SIZE)))
765
        return recommended_pages
766
3763.8.15 by John Arbash Meinel
Review comments from Martin. Code clarity/variable name/docstring updates.
767
    def _compute_total_pages_in_index(self):
768
        """How many pages are in the index.
769
770
        If we have read the header we will use the value stored there.
771
        Otherwise it will be computed based on the length of the index.
772
        """
3763.8.7 by John Arbash Meinel
A bit of doc updates, start putting in tests for current behavior.
773
        if self._size is None:
3763.8.15 by John Arbash Meinel
Review comments from Martin. Code clarity/variable name/docstring updates.
774
            raise AssertionError('_compute_total_pages_in_index should not be'
775
                                 ' called when self._size is None')
3763.8.7 by John Arbash Meinel
A bit of doc updates, start putting in tests for current behavior.
776
        if self._root_node is not None:
777
            # This is the number of pages as defined by the header
778
            return self._row_offsets[-1]
779
        # This is the number of pages as defined by the size of the index. They
780
        # should be indentical.
3763.8.15 by John Arbash Meinel
Review comments from Martin. Code clarity/variable name/docstring updates.
781
        total_pages = int(math.ceil(self._size / float(_PAGE_SIZE)))
782
        return total_pages
3763.8.7 by John Arbash Meinel
A bit of doc updates, start putting in tests for current behavior.
783
3763.8.15 by John Arbash Meinel
Review comments from Martin. Code clarity/variable name/docstring updates.
784
    def _expand_offsets(self, offsets):
785
        """Find extra pages to download.
3763.8.1 by John Arbash Meinel
Playing around with expanding requests for btree index nodes into neighboring nodes.
786
787
        The idea is that we always want to make big-enough requests (like 64kB
788
        for http), so that we don't waste round trips. So given the entries
3763.8.15 by John Arbash Meinel
Review comments from Martin. Code clarity/variable name/docstring updates.
789
        that we already have cached and the new pages being downloaded figure
3763.8.1 by John Arbash Meinel
Playing around with expanding requests for btree index nodes into neighboring nodes.
790
        out what other pages we might want to read.
791
3763.8.15 by John Arbash Meinel
Review comments from Martin. Code clarity/variable name/docstring updates.
792
        See also doc/developers/btree_index_prefetch.txt for more details.
793
794
        :param offsets: The offsets to be read
795
        :return: A list of offsets to download
3763.8.1 by John Arbash Meinel
Playing around with expanding requests for btree index nodes into neighboring nodes.
796
        """
3763.8.7 by John Arbash Meinel
A bit of doc updates, start putting in tests for current behavior.
797
        if 'index' in debug.debug_flags:
3763.8.15 by John Arbash Meinel
Review comments from Martin. Code clarity/variable name/docstring updates.
798
            trace.mutter('expanding: %s\toffsets: %s', self._name, offsets)
3763.8.1 by John Arbash Meinel
Playing around with expanding requests for btree index nodes into neighboring nodes.
799
3763.8.15 by John Arbash Meinel
Review comments from Martin. Code clarity/variable name/docstring updates.
800
        if len(offsets) >= self._recommended_pages:
3763.8.1 by John Arbash Meinel
Playing around with expanding requests for btree index nodes into neighboring nodes.
801
            # Don't add more, we are already requesting more than enough
3763.8.7 by John Arbash Meinel
A bit of doc updates, start putting in tests for current behavior.
802
            if 'index' in debug.debug_flags:
803
                trace.mutter('  not expanding large request (%s >= %s)',
3763.8.15 by John Arbash Meinel
Review comments from Martin. Code clarity/variable name/docstring updates.
804
                             len(offsets), self._recommended_pages)
805
            return offsets
3763.8.1 by John Arbash Meinel
Playing around with expanding requests for btree index nodes into neighboring nodes.
806
        if self._size is None:
807
            # Don't try anything, because we don't know where the file ends
3763.8.7 by John Arbash Meinel
A bit of doc updates, start putting in tests for current behavior.
808
            if 'index' in debug.debug_flags:
809
                trace.mutter('  not expanding without knowing index size')
3763.8.15 by John Arbash Meinel
Review comments from Martin. Code clarity/variable name/docstring updates.
810
            return offsets
811
        total_pages = self._compute_total_pages_in_index()
812
        cached_offsets = self._get_offsets_to_cached_pages()
3763.8.1 by John Arbash Meinel
Playing around with expanding requests for btree index nodes into neighboring nodes.
813
        # If reading recommended_pages would read the rest of the index, just
814
        # do so.
3763.8.15 by John Arbash Meinel
Review comments from Martin. Code clarity/variable name/docstring updates.
815
        if total_pages - len(cached_offsets) <= self._recommended_pages:
3763.8.7 by John Arbash Meinel
A bit of doc updates, start putting in tests for current behavior.
816
            # Read whatever is left
3763.8.15 by John Arbash Meinel
Review comments from Martin. Code clarity/variable name/docstring updates.
817
            if cached_offsets:
818
                expanded = [x for x in xrange(total_pages)
819
                               if x not in cached_offsets]
3763.8.7 by John Arbash Meinel
A bit of doc updates, start putting in tests for current behavior.
820
            else:
3763.8.15 by John Arbash Meinel
Review comments from Martin. Code clarity/variable name/docstring updates.
821
                expanded = range(total_pages)
3763.8.7 by John Arbash Meinel
A bit of doc updates, start putting in tests for current behavior.
822
            if 'index' in debug.debug_flags:
823
                trace.mutter('  reading all unread pages: %s', expanded)
824
            return expanded
3763.8.1 by John Arbash Meinel
Playing around with expanding requests for btree index nodes into neighboring nodes.
825
3763.8.7 by John Arbash Meinel
A bit of doc updates, start putting in tests for current behavior.
826
        if self._root_node is None:
827
            # ATM on the first read of the root node of a large index, we don't
828
            # bother pre-reading any other pages. This is because the
829
            # likelyhood of actually reading interesting pages is very low.
830
            # See doc/developers/btree_index_prefetch.txt for a discussion, and
831
            # a possible implementation when we are guessing that the second
832
            # layer index is small
3763.8.15 by John Arbash Meinel
Review comments from Martin. Code clarity/variable name/docstring updates.
833
            final_offsets = offsets
3763.8.1 by John Arbash Meinel
Playing around with expanding requests for btree index nodes into neighboring nodes.
834
        else:
3763.8.14 by John Arbash Meinel
Add in a shortcut when we haven't cached much yet.
835
            tree_depth = len(self._row_lengths)
3763.8.15 by John Arbash Meinel
Review comments from Martin. Code clarity/variable name/docstring updates.
836
            if len(cached_offsets) < tree_depth and len(offsets) == 1:
3763.8.14 by John Arbash Meinel
Add in a shortcut when we haven't cached much yet.
837
                # We haven't read enough to justify expansion
838
                # If we are only going to read the root node, and 1 leaf node,
839
                # then it isn't worth expanding our request. Once we've read at
840
                # least 2 nodes, then we are probably doing a search, and we
841
                # start expanding our requests.
842
                if 'index' in debug.debug_flags:
843
                    trace.mutter('  not expanding on first reads')
3763.8.15 by John Arbash Meinel
Review comments from Martin. Code clarity/variable name/docstring updates.
844
                return offsets
845
            final_offsets = self._expand_to_neighbors(offsets, cached_offsets,
846
                                                      total_pages)
847
848
        final_offsets = sorted(final_offsets)
3763.8.7 by John Arbash Meinel
A bit of doc updates, start putting in tests for current behavior.
849
        if 'index' in debug.debug_flags:
3763.8.15 by John Arbash Meinel
Review comments from Martin. Code clarity/variable name/docstring updates.
850
            trace.mutter('expanded:  %s', final_offsets)
851
        return final_offsets
852
853
    def _expand_to_neighbors(self, offsets, cached_offsets, total_pages):
854
        """Expand requests to neighbors until we have enough pages.
855
856
        This is called from _expand_offsets after policy has determined that we
857
        want to expand.
858
        We only want to expand requests within a given layer. We cheat a little
859
        bit and assume all requests will be in the same layer. This is true
860
        given the current design, but if it changes this algorithm may perform
861
        oddly.
862
863
        :param offsets: requested offsets
864
        :param cached_offsets: offsets for pages we currently have cached
865
        :return: A set() of offsets after expansion
866
        """
867
        final_offsets = set(offsets)
868
        first = end = None
869
        new_tips = set(final_offsets)
870
        while len(final_offsets) < self._recommended_pages and new_tips:
871
            next_tips = set()
872
            for pos in new_tips:
873
                if first is None:
874
                    first, end = self._find_layer_first_and_end(pos)
875
                previous = pos - 1
876
                if (previous > 0
877
                    and previous not in cached_offsets
878
                    and previous not in final_offsets
879
                    and previous >= first):
880
                    next_tips.add(previous)
881
                after = pos + 1
882
                if (after < total_pages
883
                    and after not in cached_offsets
884
                    and after not in final_offsets
885
                    and after < end):
886
                    next_tips.add(after)
887
                # This would keep us from going bigger than
888
                # recommended_pages by only expanding the first offsets.
889
                # However, if we are making a 'wide' request, it is
890
                # reasonable to expand all points equally.
891
                # if len(final_offsets) > recommended_pages:
892
                #     break
893
            final_offsets.update(next_tips)
894
            new_tips = next_tips
895
        return final_offsets
3763.8.1 by John Arbash Meinel
Playing around with expanding requests for btree index nodes into neighboring nodes.
896
4744.2.6 by John Arbash Meinel
Start exposing an GraphIndex.clear_cache() member.
897
    def clear_cache(self):
898
        """Clear out any cached/memoized values.
899
900
        This can be called at any time, but generally it is used when we have
901
        extracted some information, but don't expect to be requesting any more
902
        from this index.
903
        """
904
        # Note that we don't touch self._root_node or self._internal_node_cache
905
        # We don't expect either of those to be big, and it can save
906
        # round-trips in the future. We may re-evaluate this if InternalNode
907
        # memory starts to be an issue.
908
        self._leaf_node_cache.clear()
909
4011.5.3 by Andrew Bennetts
Implement and test external_references on GraphIndex and BTreeGraphIndex.
910
    def external_references(self, ref_list_num):
911
        if self._root_node is None:
912
            self._get_root_node()
913
        if ref_list_num + 1 > self.node_ref_lists:
914
            raise ValueError('No ref list %d, index has %d ref lists'
915
                % (ref_list_num, self.node_ref_lists))
916
        keys = set()
917
        refs = set()
918
        for node in self.iter_all_entries():
919
            keys.add(node[1])
920
            refs.update(node[3][ref_list_num])
921
        return refs - keys
922
3763.8.12 by John Arbash Meinel
Code cleanup.
923
    def _find_layer_first_and_end(self, offset):
924
        """Find the start/stop nodes for the layer corresponding to offset.
925
926
        :return: (first, end)
927
            first is the first node in this layer
928
            end is the first node of the next layer
929
        """
930
        first = end = 0
931
        for roffset in self._row_offsets:
932
            first = end
933
            end = roffset
934
            if offset < roffset:
935
                break
936
        return first, end
937
3763.8.15 by John Arbash Meinel
Review comments from Martin. Code clarity/variable name/docstring updates.
938
    def _get_offsets_to_cached_pages(self):
3763.8.12 by John Arbash Meinel
Code cleanup.
939
        """Determine what nodes we already have cached."""
3763.8.15 by John Arbash Meinel
Review comments from Martin. Code clarity/variable name/docstring updates.
940
        cached_offsets = set(self._internal_node_cache.keys())
941
        cached_offsets.update(self._leaf_node_cache.keys())
3763.8.12 by John Arbash Meinel
Code cleanup.
942
        if self._root_node is not None:
3763.8.15 by John Arbash Meinel
Review comments from Martin. Code clarity/variable name/docstring updates.
943
            cached_offsets.add(0)
944
        return cached_offsets
3763.8.12 by John Arbash Meinel
Code cleanup.
945
946
    def _get_root_node(self):
947
        if self._root_node is None:
948
            # We may not have a root node yet
949
            self._get_internal_nodes([0])
950
        return self._root_node
951
3641.5.18 by John Arbash Meinel
Clean out the global state, good for prototyping and tuning, bad for production code.
952
    def _get_nodes(self, cache, node_indexes):
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
953
        found = {}
954
        needed = []
955
        for idx in node_indexes:
956
            if idx == 0 and self._root_node is not None:
957
                found[0] = self._root_node
958
                continue
959
            try:
960
                found[idx] = cache[idx]
961
            except KeyError:
962
                needed.append(idx)
3763.8.1 by John Arbash Meinel
Playing around with expanding requests for btree index nodes into neighboring nodes.
963
        if not needed:
964
            return found
3763.8.15 by John Arbash Meinel
Review comments from Martin. Code clarity/variable name/docstring updates.
965
        needed = self._expand_offsets(needed)
3763.8.12 by John Arbash Meinel
Code cleanup.
966
        found.update(self._get_and_cache_nodes(needed))
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
967
        return found
968
969
    def _get_internal_nodes(self, node_indexes):
970
        """Get a node, from cache or disk.
971
972
        After getting it, the node will be cached.
973
        """
3641.5.18 by John Arbash Meinel
Clean out the global state, good for prototyping and tuning, bad for production code.
974
        return self._get_nodes(self._internal_node_cache, node_indexes)
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
975
3805.4.6 by John Arbash Meinel
refactor for clarity.
976
    def _cache_leaf_values(self, nodes):
977
        """Cache directly from key => value, skipping the btree."""
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
978
        if self._leaf_value_cache is not None:
3805.4.6 by John Arbash Meinel
refactor for clarity.
979
            for node in nodes.itervalues():
5365.5.1 by John Arbash Meinel
Implement a custom parser for chk btree leaves.
980
                for key, value in node.all_items():
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
981
                    if key in self._leaf_value_cache:
982
                        # Don't add the rest of the keys, we've seen this node
983
                        # before.
984
                        break
985
                    self._leaf_value_cache[key] = value
3805.4.6 by John Arbash Meinel
refactor for clarity.
986
987
    def _get_leaf_nodes(self, node_indexes):
988
        """Get a bunch of nodes, from cache or disk."""
989
        found = self._get_nodes(self._leaf_node_cache, node_indexes)
990
        self._cache_leaf_values(found)
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
991
        return found
992
993
    def iter_all_entries(self):
994
        """Iterate over all keys within the index.
995
996
        :return: An iterable of (index, key, value) or (index, key, value, reference_lists).
997
            The former tuple is used when there are no reference lists in the
998
            index, making the API compatible with simple key:value index types.
999
            There is no defined order for the result iteration - it will be in
1000
            the most efficient order for the index.
1001
        """
1002
        if 'evil' in debug.debug_flags:
1003
            trace.mutter_callsite(3,
1004
                "iter_all_entries scales with size of history.")
1005
        if not self.key_count():
1006
            return
3823.5.2 by John Arbash Meinel
It turns out that we read the pack-names file 3-times because
1007
        if self._row_offsets[-1] == 1:
1008
            # There is only the root node, and we read that via key_count()
1009
            if self.node_ref_lists:
5365.5.1 by John Arbash Meinel
Implement a custom parser for chk btree leaves.
1010
                for key, (value, refs) in self._root_node.all_items():
3823.5.2 by John Arbash Meinel
It turns out that we read the pack-names file 3-times because
1011
                    yield (self, key, value, refs)
1012
            else:
5365.5.1 by John Arbash Meinel
Implement a custom parser for chk btree leaves.
1013
                for key, (value, refs) in self._root_node.all_items():
3823.5.2 by John Arbash Meinel
It turns out that we read the pack-names file 3-times because
1014
                    yield (self, key, value)
1015
            return
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
1016
        start_of_leaves = self._row_offsets[-2]
1017
        end_of_leaves = self._row_offsets[-1]
3824.1.2 by John Arbash Meinel
iter_all_entries() shouldn't need to re-read the page.
1018
        needed_offsets = range(start_of_leaves, end_of_leaves)
1019
        if needed_offsets == [0]:
1020
            # Special case when we only have a root node, as we have already
1021
            # read everything
1022
            nodes = [(0, self._root_node)]
1023
        else:
1024
            nodes = self._read_nodes(needed_offsets)
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
1025
        # We iterate strictly in-order so that we can use this function
1026
        # for spilling index builds to disk.
1027
        if self.node_ref_lists:
3824.1.2 by John Arbash Meinel
iter_all_entries() shouldn't need to re-read the page.
1028
            for _, node in nodes:
5365.5.1 by John Arbash Meinel
Implement a custom parser for chk btree leaves.
1029
                for key, (value, refs) in node.all_items():
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
1030
                    yield (self, key, value, refs)
1031
        else:
3824.1.2 by John Arbash Meinel
iter_all_entries() shouldn't need to re-read the page.
1032
            for _, node in nodes:
5365.5.1 by John Arbash Meinel
Implement a custom parser for chk btree leaves.
1033
                for key, (value, refs) in node.all_items():
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
1034
                    yield (self, key, value)
1035
1036
    @staticmethod
1037
    def _multi_bisect_right(in_keys, fixed_keys):
1038
        """Find the positions where each 'in_key' would fit in fixed_keys.
1039
1040
        This is equivalent to doing "bisect_right" on each in_key into
1041
        fixed_keys
1042
1043
        :param in_keys: A sorted list of keys to match with fixed_keys
1044
        :param fixed_keys: A sorted list of keys to match against
1045
        :return: A list of (integer position, [key list]) tuples.
1046
        """
1047
        if not in_keys:
1048
            return []
1049
        if not fixed_keys:
1050
            # no pointers in the fixed_keys list, which means everything must
1051
            # fall to the left.
1052
            return [(0, in_keys)]
1053
1054
        # TODO: Iterating both lists will generally take M + N steps
1055
        #       Bisecting each key will generally take M * log2 N steps.
1056
        #       If we had an efficient way to compare, we could pick the method
1057
        #       based on which has the fewer number of steps.
1058
        #       There is also the argument that bisect_right is a compiled
1059
        #       function, so there is even more to be gained.
1060
        # iter_steps = len(in_keys) + len(fixed_keys)
1061
        # bisect_steps = len(in_keys) * math.log(len(fixed_keys), 2)
1062
        if len(in_keys) == 1: # Bisect will always be faster for M = 1
5753.2.4 by Jelmer Vernooij
Review feedback from John.
1063
            return [(bisect.bisect_right(fixed_keys, in_keys[0]), in_keys)]
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
1064
        # elif bisect_steps < iter_steps:
1065
        #     offsets = {}
1066
        #     for key in in_keys:
1067
        #         offsets.setdefault(bisect_right(fixed_keys, key),
1068
        #                            []).append(key)
1069
        #     return [(o, offsets[o]) for o in sorted(offsets)]
1070
        in_keys_iter = iter(in_keys)
1071
        fixed_keys_iter = enumerate(fixed_keys)
1072
        cur_in_key = in_keys_iter.next()
1073
        cur_fixed_offset, cur_fixed_key = fixed_keys_iter.next()
1074
1075
        class InputDone(Exception): pass
1076
        class FixedDone(Exception): pass
1077
1078
        output = []
1079
        cur_out = []
1080
1081
        # TODO: Another possibility is that rather than iterating on each side,
1082
        #       we could use a combination of bisecting and iterating. For
1083
        #       example, while cur_in_key < fixed_key, bisect to find its
1084
        #       point, then iterate all matching keys, then bisect (restricted
1085
        #       to only the remainder) for the next one, etc.
1086
        try:
1087
            while True:
1088
                if cur_in_key < cur_fixed_key:
1089
                    cur_keys = []
1090
                    cur_out = (cur_fixed_offset, cur_keys)
1091
                    output.append(cur_out)
1092
                    while cur_in_key < cur_fixed_key:
1093
                        cur_keys.append(cur_in_key)
1094
                        try:
1095
                            cur_in_key = in_keys_iter.next()
1096
                        except StopIteration:
1097
                            raise InputDone
1098
                    # At this point cur_in_key must be >= cur_fixed_key
1099
                # step the cur_fixed_key until we pass the cur key, or walk off
1100
                # the end
1101
                while cur_in_key >= cur_fixed_key:
1102
                    try:
1103
                        cur_fixed_offset, cur_fixed_key = fixed_keys_iter.next()
1104
                    except StopIteration:
1105
                        raise FixedDone
1106
        except InputDone:
1107
            # We consumed all of the input, nothing more to do
1108
            pass
1109
        except FixedDone:
1110
            # There was some input left, but we consumed all of fixed, so we
1111
            # have to add one more for the tail
1112
            cur_keys = [cur_in_key]
1113
            cur_keys.extend(in_keys_iter)
1114
            cur_out = (len(fixed_keys), cur_keys)
1115
            output.append(cur_out)
1116
        return output
1117
4593.4.5 by John Arbash Meinel
Start adding some tests.
1118
    def _walk_through_internal_nodes(self, keys):
1119
        """Take the given set of keys, and find the corresponding LeafNodes.
1120
1121
        :param keys: An unsorted iterable of keys to search for
1122
        :return: (nodes, index_and_keys)
1123
            nodes is a dict mapping {index: LeafNode}
1124
            keys_at_index is a list of tuples of [(index, [keys for Leaf])]
1125
        """
1126
        # 6 seconds spent in miss_torture using the sorted() line.
1127
        # Even with out of order disk IO it seems faster not to sort it when
1128
        # large queries are being made.
1129
        keys_at_index = [(0, sorted(keys))]
1130
1131
        for row_pos, next_row_start in enumerate(self._row_offsets[1:-1]):
1132
            node_indexes = [idx for idx, s_keys in keys_at_index]
1133
            nodes = self._get_internal_nodes(node_indexes)
1134
1135
            next_nodes_and_keys = []
1136
            for node_index, sub_keys in keys_at_index:
1137
                node = nodes[node_index]
1138
                positions = self._multi_bisect_right(sub_keys, node.keys)
1139
                node_offset = next_row_start + node.offset
1140
                next_nodes_and_keys.extend([(node_offset + pos, s_keys)
1141
                                           for pos, s_keys in positions])
1142
            keys_at_index = next_nodes_and_keys
1143
        # We should now be at the _LeafNodes
1144
        node_indexes = [idx for idx, s_keys in keys_at_index]
1145
1146
        # TODO: We may *not* want to always read all the nodes in one
1147
        #       big go. Consider setting a max size on this.
1148
        nodes = self._get_leaf_nodes(node_indexes)
1149
        return nodes, keys_at_index
1150
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
1151
    def iter_entries(self, keys):
1152
        """Iterate over keys within the index.
1153
1154
        :param keys: An iterable providing the keys to be retrieved.
1155
        :return: An iterable as per iter_all_entries, but restricted to the
1156
            keys supplied. No additional keys will be returned, and every
1157
            key supplied that is in the index will be returned.
1158
        """
1159
        # 6 seconds spent in miss_torture using the sorted() line.
1160
        # Even with out of order disk IO it seems faster not to sort it when
1161
        # large queries are being made.
1162
        # However, now that we are doing multi-way bisecting, we need the keys
1163
        # in sorted order anyway. We could change the multi-way code to not
1164
        # require sorted order. (For example, it bisects for the first node,
1165
        # does an in-order search until a key comes before the current point,
1166
        # which it then bisects for, etc.)
1167
        keys = frozenset(keys)
1168
        if not keys:
1169
            return
1170
1171
        if not self.key_count():
1172
            return
1173
1174
        needed_keys = []
1175
        if self._leaf_value_cache is None:
1176
            needed_keys = keys
1177
        else:
1178
            for key in keys:
1179
                value = self._leaf_value_cache.get(key, None)
1180
                if value is not None:
1181
                    # This key is known not to be here, skip it
1182
                    value, refs = value
1183
                    if self.node_ref_lists:
1184
                        yield (self, key, value, refs)
1185
                    else:
1186
                        yield (self, key, value)
1187
                else:
1188
                    needed_keys.append(key)
1189
1190
        last_key = None
1191
        needed_keys = keys
1192
        if not needed_keys:
1193
            return
4593.4.5 by John Arbash Meinel
Start adding some tests.
1194
        nodes, nodes_and_keys = self._walk_through_internal_nodes(needed_keys)
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
1195
        for node_index, sub_keys in nodes_and_keys:
1196
            if not sub_keys:
1197
                continue
1198
            node = nodes[node_index]
1199
            for next_sub_key in sub_keys:
5365.5.1 by John Arbash Meinel
Implement a custom parser for chk btree leaves.
1200
                if next_sub_key in node:
1201
                    value, refs = node[next_sub_key]
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
1202
                    if self.node_ref_lists:
1203
                        yield (self, next_sub_key, value, refs)
1204
                    else:
1205
                        yield (self, next_sub_key, value)
1206
4593.4.12 by John Arbash Meinel
Name the specific index api _find_ancestors, and the public CombinedGraphIndex api find_ancestry()
1207
    def _find_ancestors(self, keys, ref_list_num, parent_map, missing_keys):
4593.4.11 by John Arbash Meinel
Snapshot the work in progress.
1208
        """Find the parent_map information for the set of keys.
1209
1210
        This populates the parent_map dict and missing_keys set based on the
1211
        queried keys. It also can fill out an arbitrary number of parents that
1212
        it finds while searching for the supplied keys.
1213
1214
        It is unlikely that you want to call this directly. See
4593.4.12 by John Arbash Meinel
Name the specific index api _find_ancestors, and the public CombinedGraphIndex api find_ancestry()
1215
        "CombinedGraphIndex.find_ancestry()" for a more appropriate API.
4593.4.11 by John Arbash Meinel
Snapshot the work in progress.
1216
1217
        :param keys: A keys whose ancestry we want to return
1218
            Every key will either end up in 'parent_map' or 'missing_keys'.
4593.4.1 by John Arbash Meinel
Implement a function on btree that inlines the get_parent_map loop.
1219
        :param ref_list_num: This index in the ref_lists is the parents we
1220
            care about.
4593.4.11 by John Arbash Meinel
Snapshot the work in progress.
1221
        :param parent_map: {key: parent_keys} for keys that are present in this
1222
            index. This may contain more entries than were in 'keys', that are
1223
            reachable ancestors of the keys requested.
4593.4.5 by John Arbash Meinel
Start adding some tests.
1224
        :param missing_keys: keys which are known to be missing in this index.
4593.4.11 by John Arbash Meinel
Snapshot the work in progress.
1225
            This may include parents that were not directly requested, but we
1226
            were able to determine that they are not present in this index.
1227
        :return: search_keys    parents that were found but not queried to know
1228
            if they are missing or present. Callers can re-query this index for
1229
            those keys, and they will be placed into parent_map or missing_keys
4593.4.1 by John Arbash Meinel
Implement a function on btree that inlines the get_parent_map loop.
1230
        """
1231
        if not self.key_count():
1232
            # We use key_count() to trigger reading the root node and
1233
            # determining info about this BTreeGraphIndex
1234
            # If we don't have any keys, then everything is missing
4593.4.11 by John Arbash Meinel
Snapshot the work in progress.
1235
            missing_keys.update(keys)
1236
            return set()
4593.4.1 by John Arbash Meinel
Implement a function on btree that inlines the get_parent_map loop.
1237
        if ref_list_num >= self.node_ref_lists:
1238
            raise ValueError('No ref list %d, index has %d ref lists'
1239
                % (ref_list_num, self.node_ref_lists))
1240
1241
        # The main trick we are trying to accomplish is that when we find a
1242
        # key listing its parents, we expect that the parent key is also likely
1243
        # to sit on the same page. Allowing us to expand parents quickly
1244
        # without suffering the full stack of bisecting, etc.
4593.4.5 by John Arbash Meinel
Start adding some tests.
1245
        nodes, nodes_and_keys = self._walk_through_internal_nodes(keys)
1246
4593.4.1 by John Arbash Meinel
Implement a function on btree that inlines the get_parent_map loop.
1247
        # These are parent keys which could not be immediately resolved on the
1248
        # page where the child was present. Note that we may already be
1249
        # searching for that key, and it may actually be present [or known
1250
        # missing] on one of the other pages we are reading.
1251
        # TODO:
1252
        #   We could try searching for them in the immediate previous or next
1253
        #   page. If they occur "later" we could put them in a pending lookup
1254
        #   set, and then for each node we read thereafter we could check to
1255
        #   see if they are present.
1256
        #   However, we don't know the impact of keeping this list of things
1257
        #   that I'm going to search for every node I come across from here on
1258
        #   out.
1259
        #   It doesn't handle the case when the parent key is missing on a
1260
        #   page that we *don't* read. So we already have to handle being
1261
        #   re-entrant for that.
1262
        #   Since most keys contain a date string, they are more likely to be
1263
        #   found earlier in the file than later, but we would know that right
1264
        #   away (key < min_key), and wouldn't keep searching it on every other
1265
        #   page that we read.
1266
        #   Mostly, it is an idea, one which should be benchmarked.
1267
        parents_not_on_page = set()
1268
1269
        for node_index, sub_keys in nodes_and_keys:
1270
            if not sub_keys:
1271
                continue
1272
            # sub_keys is all of the keys we are looking for that should exist
1273
            # on this page, if they aren't here, then they won't be found
1274
            node = nodes[node_index]
1275
            parents_to_check = set()
1276
            for next_sub_key in sub_keys:
5365.5.1 by John Arbash Meinel
Implement a custom parser for chk btree leaves.
1277
                if next_sub_key not in node:
4593.4.5 by John Arbash Meinel
Start adding some tests.
1278
                    # This one is just not present in the index at all
1279
                    missing_keys.add(next_sub_key)
1280
                else:
5365.5.1 by John Arbash Meinel
Implement a custom parser for chk btree leaves.
1281
                    value, refs = node[next_sub_key]
4593.4.1 by John Arbash Meinel
Implement a function on btree that inlines the get_parent_map loop.
1282
                    parent_keys = refs[ref_list_num]
1283
                    parent_map[next_sub_key] = parent_keys
1284
                    parents_to_check.update(parent_keys)
1285
            # Don't look for things we've already found
1286
            parents_to_check = parents_to_check.difference(parent_map)
4593.4.4 by John Arbash Meinel
Trying out a few more tweaks.
1287
            # this can be used to test the benefit of having the check loop
1288
            # inlined.
1289
            # parents_not_on_page.update(parents_to_check)
1290
            # continue
4593.4.1 by John Arbash Meinel
Implement a function on btree that inlines the get_parent_map loop.
1291
            while parents_to_check:
1292
                next_parents_to_check = set()
1293
                for key in parents_to_check:
5365.5.1 by John Arbash Meinel
Implement a custom parser for chk btree leaves.
1294
                    if key in node:
1295
                        value, refs = node[key]
4593.4.1 by John Arbash Meinel
Implement a function on btree that inlines the get_parent_map loop.
1296
                        parent_keys = refs[ref_list_num]
1297
                        parent_map[key] = parent_keys
1298
                        next_parents_to_check.update(parent_keys)
1299
                    else:
4593.4.4 by John Arbash Meinel
Trying out a few more tweaks.
1300
                        # This parent either is genuinely missing, or should be
1301
                        # found on another page. Perf test whether it is better
1302
                        # to check if this node should fit on this page or not.
1303
                        # in the 'everything-in-one-pack' scenario, this *not*
1304
                        # doing the check is 237ms vs 243ms.
1305
                        # So slightly better, but I assume the standard 'lots
1306
                        # of packs' is going to show a reasonable improvement
1307
                        # from the check, because it avoids 'going around
1308
                        # again' for everything that is in another index
4593.4.5 by John Arbash Meinel
Start adding some tests.
1309
                        # parents_not_on_page.add(key)
1310
                        # Missing for some reason
1311
                        if key < node.min_key:
1312
                            # in the case of bzr.dev, 3.4k/5.3k misses are
1313
                            # 'earlier' misses (65%)
1314
                            parents_not_on_page.add(key)
1315
                        elif key > node.max_key:
1316
                            # This parent key would be present on a different
1317
                            # LeafNode
1318
                            parents_not_on_page.add(key)
1319
                        else:
1320
                            # assert key != node.min_key and key != node.max_key
1321
                            # If it was going to be present, it would be on
1322
                            # *this* page, so mark it missing.
1323
                            missing_keys.add(key)
4593.4.1 by John Arbash Meinel
Implement a function on btree that inlines the get_parent_map loop.
1324
                parents_to_check = next_parents_to_check.difference(parent_map)
4593.4.4 by John Arbash Meinel
Trying out a few more tweaks.
1325
                # Might want to do another .difference() from missing_keys
4593.4.1 by John Arbash Meinel
Implement a function on btree that inlines the get_parent_map loop.
1326
        # parents_not_on_page could have been found on a different page, or be
1327
        # known to be missing. So cull out everything that has already been
1328
        # found.
4593.4.5 by John Arbash Meinel
Start adding some tests.
1329
        search_keys = parents_not_on_page.difference(
4593.4.1 by John Arbash Meinel
Implement a function on btree that inlines the get_parent_map loop.
1330
            parent_map).difference(missing_keys)
4593.4.5 by John Arbash Meinel
Start adding some tests.
1331
        return search_keys
4593.4.1 by John Arbash Meinel
Implement a function on btree that inlines the get_parent_map loop.
1332
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
1333
    def iter_entries_prefix(self, keys):
1334
        """Iterate over keys within the index using prefix matching.
1335
1336
        Prefix matching is applied within the tuple of a key, not to within
1337
        the bytestring of each key element. e.g. if you have the keys ('foo',
1338
        'bar'), ('foobar', 'gam') and do a prefix search for ('foo', None) then
1339
        only the former key is returned.
1340
1341
        WARNING: Note that this method currently causes a full index parse
1342
        unconditionally (which is reasonably appropriate as it is a means for
1343
        thunking many small indices into one larger one and still supplies
1344
        iter_all_entries at the thunk layer).
1345
1346
        :param keys: An iterable providing the key prefixes to be retrieved.
1347
            Each key prefix takes the form of a tuple the length of a key, but
1348
            with the last N elements 'None' rather than a regular bytestring.
1349
            The first element cannot be 'None'.
1350
        :return: An iterable as per iter_all_entries, but restricted to the
1351
            keys with a matching prefix to those supplied. No additional keys
1352
            will be returned, and every match that is in the index will be
1353
            returned.
1354
        """
1355
        keys = sorted(set(keys))
1356
        if not keys:
1357
            return
1358
        # Load if needed to check key lengths
1359
        if self._key_count is None:
1360
            self._get_root_node()
1361
        # TODO: only access nodes that can satisfy the prefixes we are looking
1362
        # for. For now, to meet API usage (as this function is not used by
1363
        # current bzrlib) just suck the entire index and iterate in memory.
1364
        nodes = {}
1365
        if self.node_ref_lists:
1366
            if self._key_length == 1:
1367
                for _1, key, value, refs in self.iter_all_entries():
1368
                    nodes[key] = value, refs
1369
            else:
1370
                nodes_by_key = {}
1371
                for _1, key, value, refs in self.iter_all_entries():
1372
                    key_value = key, value, refs
1373
                    # For a key of (foo, bar, baz) create
1374
                    # _nodes_by_key[foo][bar][baz] = key_value
1375
                    key_dict = nodes_by_key
1376
                    for subkey in key[:-1]:
1377
                        key_dict = key_dict.setdefault(subkey, {})
1378
                    key_dict[key[-1]] = key_value
1379
        else:
1380
            if self._key_length == 1:
1381
                for _1, key, value in self.iter_all_entries():
1382
                    nodes[key] = value
1383
            else:
1384
                nodes_by_key = {}
1385
                for _1, key, value in self.iter_all_entries():
1386
                    key_value = key, value
1387
                    # For a key of (foo, bar, baz) create
1388
                    # _nodes_by_key[foo][bar][baz] = key_value
1389
                    key_dict = nodes_by_key
1390
                    for subkey in key[:-1]:
1391
                        key_dict = key_dict.setdefault(subkey, {})
1392
                    key_dict[key[-1]] = key_value
1393
        if self._key_length == 1:
1394
            for key in keys:
1395
                # sanity check
1396
                if key[0] is None:
1397
                    raise errors.BadIndexKey(key)
1398
                if len(key) != self._key_length:
1399
                    raise errors.BadIndexKey(key)
1400
                try:
1401
                    if self.node_ref_lists:
1402
                        value, node_refs = nodes[key]
1403
                        yield self, key, value, node_refs
1404
                    else:
1405
                        yield self, key, nodes[key]
1406
                except KeyError:
1407
                    pass
1408
            return
1409
        for key in keys:
1410
            # sanity check
1411
            if key[0] is None:
1412
                raise errors.BadIndexKey(key)
1413
            if len(key) != self._key_length:
1414
                raise errors.BadIndexKey(key)
1415
            # find what it refers to:
1416
            key_dict = nodes_by_key
1417
            elements = list(key)
1418
            # find the subdict whose contents should be returned.
1419
            try:
1420
                while len(elements) and elements[0] is not None:
1421
                    key_dict = key_dict[elements[0]]
1422
                    elements.pop(0)
1423
            except KeyError:
1424
                # a non-existant lookup.
1425
                continue
1426
            if len(elements):
1427
                dicts = [key_dict]
1428
                while dicts:
1429
                    key_dict = dicts.pop(-1)
1430
                    # can't be empty or would not exist
1431
                    item, value = key_dict.iteritems().next()
1432
                    if type(value) == dict:
1433
                        # push keys
1434
                        dicts.extend(key_dict.itervalues())
1435
                    else:
1436
                        # yield keys
1437
                        for value in key_dict.itervalues():
1438
                            # each value is the key:value:node refs tuple
1439
                            # ready to yield.
1440
                            yield (self, ) + value
1441
            else:
1442
                # the last thing looked up was a terminal element
1443
                yield (self, ) + key_dict
1444
1445
    def key_count(self):
1446
        """Return an estimate of the number of keys in this index.
1447
1448
        For BTreeGraphIndex the estimate is exact as it is contained in the
1449
        header.
1450
        """
1451
        if self._key_count is None:
1452
            self._get_root_node()
1453
        return self._key_count
1454
3763.8.7 by John Arbash Meinel
A bit of doc updates, start putting in tests for current behavior.
1455
    def _compute_row_offsets(self):
1456
        """Fill out the _row_offsets attribute based on _row_lengths."""
1457
        offsets = []
1458
        row_offset = 0
1459
        for row in self._row_lengths:
1460
            offsets.append(row_offset)
1461
            row_offset += row
1462
        offsets.append(row_offset)
1463
        self._row_offsets = offsets
1464
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
1465
    def _parse_header_from_bytes(self, bytes):
1466
        """Parse the header from a region of bytes.
1467
1468
        :param bytes: The data to parse.
1469
        :return: An offset, data tuple such as readv yields, for the unparsed
1470
            data. (which may be of length 0).
1471
        """
1472
        signature = bytes[0:len(self._signature())]
1473
        if not signature == self._signature():
1474
            raise errors.BadIndexFormatSignature(self._name, BTreeGraphIndex)
1475
        lines = bytes[len(self._signature()):].splitlines()
1476
        options_line = lines[0]
1477
        if not options_line.startswith(_OPTION_NODE_REFS):
1478
            raise errors.BadIndexOptions(self)
1479
        try:
1480
            self.node_ref_lists = int(options_line[len(_OPTION_NODE_REFS):])
1481
        except ValueError:
1482
            raise errors.BadIndexOptions(self)
1483
        options_line = lines[1]
1484
        if not options_line.startswith(_OPTION_KEY_ELEMENTS):
1485
            raise errors.BadIndexOptions(self)
1486
        try:
1487
            self._key_length = int(options_line[len(_OPTION_KEY_ELEMENTS):])
1488
        except ValueError:
1489
            raise errors.BadIndexOptions(self)
1490
        options_line = lines[2]
1491
        if not options_line.startswith(_OPTION_LEN):
1492
            raise errors.BadIndexOptions(self)
1493
        try:
1494
            self._key_count = int(options_line[len(_OPTION_LEN):])
1495
        except ValueError:
1496
            raise errors.BadIndexOptions(self)
1497
        options_line = lines[3]
1498
        if not options_line.startswith(_OPTION_ROW_LENGTHS):
1499
            raise errors.BadIndexOptions(self)
1500
        try:
1501
            self._row_lengths = map(int, [length for length in
1502
                options_line[len(_OPTION_ROW_LENGTHS):].split(',')
1503
                if len(length)])
1504
        except ValueError:
1505
            raise errors.BadIndexOptions(self)
3763.8.7 by John Arbash Meinel
A bit of doc updates, start putting in tests for current behavior.
1506
        self._compute_row_offsets()
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
1507
1508
        # calculate the bytes we have processed
1509
        header_end = (len(signature) + sum(map(len, lines[0:4])) + 4)
1510
        return header_end, bytes[header_end:]
1511
1512
    def _read_nodes(self, nodes):
1513
        """Read some nodes from disk into the LRU cache.
1514
1515
        This performs a readv to get the node data into memory, and parses each
3868.1.1 by Martin Pool
merge John's patch to avoid re-reading pack-names file
1516
        node, then yields it to the caller. The nodes are requested in the
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
1517
        supplied order. If possible doing sort() on the list before requesting
1518
        a read may improve performance.
1519
1520
        :param nodes: The nodes to read. 0 - first node, 1 - second node etc.
1521
        :return: None
1522
        """
3868.1.1 by Martin Pool
merge John's patch to avoid re-reading pack-names file
1523
        # may be the byte string of the whole file
3823.5.2 by John Arbash Meinel
It turns out that we read the pack-names file 3-times because
1524
        bytes = None
3868.1.1 by Martin Pool
merge John's patch to avoid re-reading pack-names file
1525
        # list of (offset, length) regions of the file that should, evenually
1526
        # be read in to data_ranges, either from 'bytes' or from the transport
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
1527
        ranges = []
5074.4.1 by John Arbash Meinel
Add an offset flag to BTreeGraphIndex.
1528
        base_offset = self._base_offset
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
1529
        for index in nodes:
5074.4.1 by John Arbash Meinel
Add an offset flag to BTreeGraphIndex.
1530
            offset = (index * _PAGE_SIZE)
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
1531
            size = _PAGE_SIZE
1532
            if index == 0:
1533
                # Root node - special case
1534
                if self._size:
1535
                    size = min(_PAGE_SIZE, self._size)
1536
                else:
3824.1.1 by John Arbash Meinel
Fix _read_nodes() to only issue a single read if there is no known size.
1537
                    # The only case where we don't know the size, is for very
1538
                    # small indexes. So we read the whole thing
3823.5.2 by John Arbash Meinel
It turns out that we read the pack-names file 3-times because
1539
                    bytes = self._transport.get_bytes(self._name)
5074.4.1 by John Arbash Meinel
Add an offset flag to BTreeGraphIndex.
1540
                    num_bytes = len(bytes)
1541
                    self._size = num_bytes - base_offset
3868.1.1 by Martin Pool
merge John's patch to avoid re-reading pack-names file
1542
                    # the whole thing should be parsed out of 'bytes'
5074.4.1 by John Arbash Meinel
Add an offset flag to BTreeGraphIndex.
1543
                    ranges = [(start, min(_PAGE_SIZE, num_bytes - start))
1544
                        for start in xrange(base_offset, num_bytes, _PAGE_SIZE)]
3823.5.2 by John Arbash Meinel
It turns out that we read the pack-names file 3-times because
1545
                    break
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
1546
            else:
3763.8.6 by John Arbash Meinel
Fix the logic a bit, and add a bit more tweaking opportunities
1547
                if offset > self._size:
1548
                    raise AssertionError('tried to read past the end'
1549
                                         ' of the file %s > %s'
1550
                                         % (offset, self._size))
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
1551
                size = min(size, self._size - offset)
5074.4.1 by John Arbash Meinel
Add an offset flag to BTreeGraphIndex.
1552
            ranges.append((base_offset + offset, size))
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
1553
        if not ranges:
1554
            return
3868.1.1 by Martin Pool
merge John's patch to avoid re-reading pack-names file
1555
        elif bytes is not None:
1556
            # already have the whole file
5074.4.1 by John Arbash Meinel
Add an offset flag to BTreeGraphIndex.
1557
            data_ranges = [(start, bytes[start:start+size])
1558
                           for start, size in ranges]
3824.1.1 by John Arbash Meinel
Fix _read_nodes() to only issue a single read if there is no known size.
1559
        elif self._file is None:
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
1560
            data_ranges = self._transport.readv(self._name, ranges)
1561
        else:
1562
            data_ranges = []
1563
            for offset, size in ranges:
1564
                self._file.seek(offset)
1565
                data_ranges.append((offset, self._file.read(size)))
1566
        for offset, data in data_ranges:
5074.4.1 by John Arbash Meinel
Add an offset flag to BTreeGraphIndex.
1567
            offset -= base_offset
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
1568
            if offset == 0:
1569
                # extract the header
1570
                offset, data = self._parse_header_from_bytes(data)
1571
                if len(data) == 0:
1572
                    continue
1573
            bytes = zlib.decompress(data)
1574
            if bytes.startswith(_LEAF_FLAG):
5365.5.18 by John Arbash Meinel
Expose the new leaf node factory across the stack.
1575
                node = self._leaf_factory(bytes, self._key_length,
1576
                                          self.node_ref_lists)
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
1577
            elif bytes.startswith(_INTERNAL_FLAG):
1578
                node = _InternalNode(bytes)
1579
            else:
1580
                raise AssertionError("Unknown node type for %r" % bytes)
1581
            yield offset / _PAGE_SIZE, node
1582
1583
    def _signature(self):
1584
        """The file signature for this index type."""
1585
        return _BTSIGNATURE
1586
1587
    def validate(self):
1588
        """Validate that everything in the index can be accessed."""
1589
        # just read and parse every node.
1590
        self._get_root_node()
1591
        if len(self._row_lengths) > 1:
1592
            start_node = self._row_offsets[1]
1593
        else:
1594
            # We shouldn't be reading anything anyway
1595
            start_node = 1
1596
        node_end = self._row_offsets[-1]
1597
        for node in self._read_nodes(range(start_node, node_end)):
1598
            pass
1599
1600
5365.5.18 by John Arbash Meinel
Expose the new leaf node factory across the stack.
1601
_gcchk_factory = _LeafNode
1602
3641.3.1 by John Arbash Meinel
Bring in the btree_index and chunk_writer code and their tests.
1603
try:
4459.2.1 by Vincent Ladeuil
Use a consistent scheme for naming pyrex source files.
1604
    from bzrlib import _btree_serializer_pyx as _btree_serializer
5365.5.18 by John Arbash Meinel
Expose the new leaf node factory across the stack.
1605
    _gcchk_factory = _btree_serializer._parse_into_chk
4574.3.6 by Martin Pool
More warnings when failing to load extensions
1606
except ImportError, e:
4574.3.8 by Martin Pool
Only mutter extension load errors when they occur, and record for later
1607
    osutils.failed_to_load_extension(e)
3641.3.30 by John Arbash Meinel
Rename _parse_btree to _btree_serializer
1608
    from bzrlib import _btree_serializer_py as _btree_serializer