1
# Copyright (C) 2007-2011 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
"""Indexing facilities."""
23
'GraphIndexPrefixAdapter',
27
from bisect import bisect_right
28
from io import BytesIO
31
from ..lazy_import import lazy_import
32
lazy_import(globals(), """
35
revision as _mod_revision,
43
from ..static_tuple import StaticTuple
45
_HEADER_READV = (0, 200)
46
_OPTION_KEY_ELEMENTS = b"key_elements="
48
_OPTION_NODE_REFS = b"node_ref_lists="
49
_SIGNATURE = b"Bazaar Graph Index 1\n"
52
class BadIndexFormatSignature(errors.BzrError):
54
_fmt = "%(value)s is not an index of type %(_type)s."
56
def __init__(self, value, _type):
57
errors.BzrError.__init__(self)
62
class BadIndexData(errors.BzrError):
64
_fmt = "Error in data for index %(value)s."
66
def __init__(self, value):
67
errors.BzrError.__init__(self)
71
class BadIndexDuplicateKey(errors.BzrError):
73
_fmt = "The key '%(key)s' is already in index '%(index)s'."
75
def __init__(self, key, index):
76
errors.BzrError.__init__(self)
81
class BadIndexKey(errors.BzrError):
83
_fmt = "The key '%(key)s' is not a valid key."
85
def __init__(self, key):
86
errors.BzrError.__init__(self)
90
class BadIndexOptions(errors.BzrError):
92
_fmt = "Could not parse options for index %(value)s."
94
def __init__(self, value):
95
errors.BzrError.__init__(self)
99
class BadIndexValue(errors.BzrError):
101
_fmt = "The value '%(value)s' is not a valid value."
103
def __init__(self, value):
104
errors.BzrError.__init__(self)
108
_whitespace_re = re.compile(b'[\t\n\x0b\x0c\r\x00 ]')
109
_newline_null_re = re.compile(b'[\n\0]')
112
def _has_key_from_parent_map(self, key):
113
"""Check if this index has one key.
115
If it's possible to check for multiple keys at once through
116
calling get_parent_map that should be faster.
118
return (key in self.get_parent_map([key]))
121
def _missing_keys_from_parent_map(self, keys):
122
return set(keys) - set(self.get_parent_map(keys))
125
class GraphIndexBuilder(object):
126
"""A builder that can build a GraphIndex.
128
The resulting graph has the structure::
130
_SIGNATURE OPTIONS NODES NEWLINE
131
_SIGNATURE := 'Bazaar Graph Index 1' NEWLINE
132
OPTIONS := 'node_ref_lists=' DIGITS NEWLINE
134
NODE := KEY NULL ABSENT? NULL REFERENCES NULL VALUE NEWLINE
135
KEY := Not-whitespace-utf8
137
REFERENCES := REFERENCE_LIST (TAB REFERENCE_LIST){node_ref_lists - 1}
138
REFERENCE_LIST := (REFERENCE (CR REFERENCE)*)?
139
REFERENCE := DIGITS ; digits is the byte offset in the index of the
141
VALUE := no-newline-no-null-bytes
144
def __init__(self, reference_lists=0, key_elements=1):
145
"""Create a GraphIndex builder.
147
:param reference_lists: The number of node references lists for each
149
:param key_elements: The number of bytestrings in each key.
151
self.reference_lists = reference_lists
152
# A dict of {key: (absent, ref_lists, value)}
154
# Keys that are referenced but not actually present in this index
155
self._absent_keys = set()
156
self._nodes_by_key = None
157
self._key_length = key_elements
158
self._optimize_for_size = False
159
self._combine_backing_indices = True
161
def _check_key(self, key):
162
"""Raise BadIndexKey if key is not a valid key for this index."""
163
if type(key) not in (tuple, StaticTuple):
164
raise BadIndexKey(key)
165
if self._key_length != len(key):
166
raise BadIndexKey(key)
168
if not element or type(element) != bytes or _whitespace_re.search(element) is not None:
169
raise BadIndexKey(key)
171
def _external_references(self):
172
"""Return references that are not present in this index.
176
# TODO: JAM 2008-11-21 This makes an assumption about how the reference
177
# lists are used. It is currently correct for pack-0.92 through
178
# 1.9, which use the node references (3rd column) second
179
# reference list as the compression parent. Perhaps this should
180
# be moved into something higher up the stack, since it
181
# makes assumptions about how the index is used.
182
if self.reference_lists > 1:
183
for node in self.iter_all_entries():
185
refs.update(node[3][1])
188
# If reference_lists == 0 there can be no external references, and
189
# if reference_lists == 1, then there isn't a place to store the
193
def _get_nodes_by_key(self):
194
if self._nodes_by_key is None:
196
if self.reference_lists:
197
for key, (absent, references, value) in self._nodes.items():
200
key_dict = nodes_by_key
201
for subkey in key[:-1]:
202
key_dict = key_dict.setdefault(subkey, {})
203
key_dict[key[-1]] = key, value, references
205
for key, (absent, references, value) in self._nodes.items():
208
key_dict = nodes_by_key
209
for subkey in key[:-1]:
210
key_dict = key_dict.setdefault(subkey, {})
211
key_dict[key[-1]] = key, value
212
self._nodes_by_key = nodes_by_key
213
return self._nodes_by_key
215
def _update_nodes_by_key(self, key, value, node_refs):
216
"""Update the _nodes_by_key dict with a new key.
218
For a key of (foo, bar, baz) create
219
_nodes_by_key[foo][bar][baz] = key_value
221
if self._nodes_by_key is None:
223
key_dict = self._nodes_by_key
224
if self.reference_lists:
225
key_value = StaticTuple(key, value, node_refs)
227
key_value = StaticTuple(key, value)
228
for subkey in key[:-1]:
229
key_dict = key_dict.setdefault(subkey, {})
230
key_dict[key[-1]] = key_value
232
def _check_key_ref_value(self, key, references, value):
233
"""Check that 'key' and 'references' are all valid.
235
:param key: A key tuple. Must conform to the key interface (be a tuple,
236
be of the right length, not have any whitespace or nulls in any key
238
:param references: An iterable of reference lists. Something like
239
[[(ref, key)], [(ref, key), (other, key)]]
240
:param value: The value associate with this key. Must not contain
241
newlines or null characters.
242
:return: (node_refs, absent_references)
244
* node_refs: basically a packed form of 'references' where all
246
* absent_references: reference keys that are not in self._nodes.
247
This may contain duplicates if the same key is referenced in
250
as_st = StaticTuple.from_sequence
252
if _newline_null_re.search(value) is not None:
253
raise BadIndexValue(value)
254
if len(references) != self.reference_lists:
255
raise BadIndexValue(references)
257
absent_references = []
258
for reference_list in references:
259
for reference in reference_list:
260
# If reference *is* in self._nodes, then we know it has already
262
if reference not in self._nodes:
263
self._check_key(reference)
264
absent_references.append(reference)
265
reference_list = as_st([as_st(ref).intern()
266
for ref in reference_list])
267
node_refs.append(reference_list)
268
return as_st(node_refs), absent_references
270
def add_node(self, key, value, references=()):
271
"""Add a node to the index.
273
:param key: The key. keys are non-empty tuples containing
274
as many whitespace-free utf8 bytestrings as the key length
275
defined for this index.
276
:param references: An iterable of iterables of keys. Each is a
277
reference to another key.
278
:param value: The value to associate with the key. It may be any
279
bytes as long as it does not contain \\0 or \\n.
282
absent_references) = self._check_key_ref_value(key, references, value)
283
if key in self._nodes and self._nodes[key][0] != b'a':
284
raise BadIndexDuplicateKey(key, self)
285
for reference in absent_references:
286
# There may be duplicates, but I don't think it is worth worrying
288
self._nodes[reference] = (b'a', (), b'')
289
self._absent_keys.update(absent_references)
290
self._absent_keys.discard(key)
291
self._nodes[key] = (b'', node_refs, value)
292
if self._nodes_by_key is not None and self._key_length > 1:
293
self._update_nodes_by_key(key, value, node_refs)
295
def clear_cache(self):
296
"""See GraphIndex.clear_cache()
298
This is a no-op, but we need the api to conform to a generic 'Index'
305
:returns: cBytesIO holding the full context of the index as it
306
should be written to disk.
309
lines.append(b'%s%d\n' % (_OPTION_NODE_REFS, self.reference_lists))
310
lines.append(b'%s%d\n' % (_OPTION_KEY_ELEMENTS, self._key_length))
311
key_count = len(self._nodes) - len(self._absent_keys)
312
lines.append(b'%s%d\n' % (_OPTION_LEN, key_count))
313
prefix_length = sum(len(x) for x in lines)
314
# references are byte offsets. To avoid having to do nasty
315
# polynomial work to resolve offsets (references to later in the
316
# file cannot be determined until all the inbetween references have
317
# been calculated too) we pad the offsets with 0's to make them be
318
# of consistent length. Using binary offsets would break the trivial
320
# to calculate the width of zero's needed we do three passes:
321
# one to gather all the non-reference data and the number of references.
322
# one to pad all the data with reference-length and determine entry
326
# forward sorted by key. In future we may consider topological sorting,
327
# at the cost of table scans for direct lookup, or a second index for
329
nodes = sorted(self._nodes.items())
330
# if we do not prepass, we don't know how long it will be up front.
331
expected_bytes = None
332
# we only need to pre-pass if we have reference lists at all.
333
if self.reference_lists:
335
non_ref_bytes = prefix_length
337
# TODO use simple multiplication for the constants in this loop.
338
for key, (absent, references, value) in nodes:
339
# record the offset known *so far* for this key:
340
# the non reference bytes to date, and the total references to
341
# date - saves reaccumulating on the second pass
342
key_offset_info.append((key, non_ref_bytes, total_references))
343
# key is literal, value is literal, there are 3 null's, 1 NL
344
# key is variable length tuple, \x00 between elements
345
non_ref_bytes += sum(len(element) for element in key)
346
if self._key_length > 1:
347
non_ref_bytes += self._key_length - 1
348
# value is literal bytes, there are 3 null's, 1 NL.
349
non_ref_bytes += len(value) + 3 + 1
350
# one byte for absent if set.
353
elif self.reference_lists:
354
# (ref_lists -1) tabs
355
non_ref_bytes += self.reference_lists - 1
356
# (ref-1 cr's per ref_list)
357
for ref_list in references:
358
# how many references across the whole file?
359
total_references += len(ref_list)
360
# accrue reference separators
362
non_ref_bytes += len(ref_list) - 1
363
# how many digits are needed to represent the total byte count?
365
possible_total_bytes = non_ref_bytes + total_references * digits
366
while 10 ** digits < possible_total_bytes:
368
possible_total_bytes = non_ref_bytes + total_references * digits
369
expected_bytes = possible_total_bytes + 1 # terminating newline
370
# resolve key addresses.
372
for key, non_ref_bytes, total_references in key_offset_info:
373
key_addresses[key] = non_ref_bytes + total_references * digits
375
format_string = b'%%0%dd' % digits
376
for key, (absent, references, value) in nodes:
377
flattened_references = []
378
for ref_list in references:
380
for reference in ref_list:
381
ref_addresses.append(format_string %
382
key_addresses[reference])
383
flattened_references.append(b'\r'.join(ref_addresses))
384
string_key = b'\x00'.join(key)
385
lines.append(b"%s\x00%s\x00%s\x00%s\n" % (string_key, absent,
386
b'\t'.join(flattened_references), value))
388
result = BytesIO(b''.join(lines))
389
if expected_bytes and len(result.getvalue()) != expected_bytes:
390
raise errors.BzrError('Failed index creation. Internal error:'
391
' mismatched output length and expected length: %d %d' %
392
(len(result.getvalue()), expected_bytes))
395
def set_optimize(self, for_size=None, combine_backing_indices=None):
396
"""Change how the builder tries to optimize the result.
398
:param for_size: Tell the builder to try and make the index as small as
400
:param combine_backing_indices: If the builder spills to disk to save
401
memory, should the on-disk indices be combined. Set to True if you
402
are going to be probing the index, but to False if you are not. (If
403
you are not querying, then the time spent combining is wasted.)
406
# GraphIndexBuilder itself doesn't pay attention to the flag yet, but
408
if for_size is not None:
409
self._optimize_for_size = for_size
410
if combine_backing_indices is not None:
411
self._combine_backing_indices = combine_backing_indices
413
def find_ancestry(self, keys, ref_list_num):
414
"""See CombinedGraphIndex.find_ancestry()"""
420
for _, key, value, ref_lists in self.iter_entries(pending):
421
parent_keys = ref_lists[ref_list_num]
422
parent_map[key] = parent_keys
423
next_pending.update([p for p in parent_keys if p not in
425
missing_keys.update(pending.difference(parent_map))
426
pending = next_pending
427
return parent_map, missing_keys
430
class GraphIndex(object):
431
"""An index for data with embedded graphs.
433
The index maps keys to a list of key reference lists, and a value.
434
Each node has the same number of key reference lists. Each key reference
435
list can be empty or an arbitrary length. The value is an opaque NULL
436
terminated string without any newlines. The storage of the index is
437
hidden in the interface: keys and key references are always tuples of
438
bytestrings, never the internal representation (e.g. dictionary offsets).
440
It is presumed that the index will not be mutated - it is static data.
442
Successive iter_all_entries calls will read the entire index each time.
443
Additionally, iter_entries calls will read the index linearly until the
444
desired keys are found. XXX: This must be fixed before the index is
445
suitable for production use. :XXX
448
def __init__(self, transport, name, size, unlimited_cache=False, offset=0):
449
"""Open an index called name on transport.
451
:param transport: A breezy.transport.Transport.
452
:param name: A path to provide to transport API calls.
453
:param size: The size of the index in bytes. This is used for bisection
454
logic to perform partial index reads. While the size could be
455
obtained by statting the file this introduced an additional round
456
trip as well as requiring stat'able transports, both of which are
457
avoided by having it supplied. If size is None, then bisection
458
support will be disabled and accessing the index will just stream
460
:param offset: Instead of starting the index data at offset 0, start it
461
at an arbitrary offset.
463
self._transport = transport
465
# Becomes a dict of key:(value, reference-list-byte-locations) used by
466
# the bisection interface to store parsed but not resolved keys.
467
self._bisect_nodes = None
468
# Becomes a dict of key:(value, reference-list-keys) which are ready to
469
# be returned directly to callers.
471
# a sorted list of slice-addresses for the parsed bytes of the file.
472
# e.g. (0,1) would mean that byte 0 is parsed.
473
self._parsed_byte_map = []
474
# a sorted list of keys matching each slice address for parsed bytes
475
# e.g. (None, 'foo@bar') would mean that the first byte contained no
476
# key, and the end byte of the slice is the of the data for 'foo@bar'
477
self._parsed_key_map = []
478
self._key_count = None
479
self._keys_by_offset = None
480
self._nodes_by_key = None
482
# The number of bytes we've read so far in trying to process this file
484
self._base_offset = offset
486
def __eq__(self, other):
487
"""Equal when self and other were created with the same parameters."""
489
isinstance(self, type(other)) and
490
self._transport == other._transport and
491
self._name == other._name and
492
self._size == other._size)
494
def __ne__(self, other):
495
return not self.__eq__(other)
497
def __lt__(self, other):
498
# We don't really care about the order, just that there is an order.
499
if (not isinstance(other, GraphIndex) and
500
not isinstance(other, InMemoryGraphIndex)):
501
raise TypeError(other)
502
return hash(self) < hash(other)
505
return hash((type(self), self._transport, self._name, self._size))
508
return "%s(%r)" % (self.__class__.__name__,
509
self._transport.abspath(self._name))
511
def _buffer_all(self, stream=None):
512
"""Buffer all the index data.
514
Mutates self._nodes and self.keys_by_offset.
516
if self._nodes is not None:
517
# We already did this
519
if 'index' in debug.debug_flags:
520
trace.mutter('Reading entire index %s',
521
self._transport.abspath(self._name))
523
stream = self._transport.get(self._name)
524
if self._base_offset != 0:
525
# This is wasteful, but it is better than dealing with
526
# adjusting all the offsets, etc.
527
stream = BytesIO(stream.read()[self._base_offset:])
529
self._read_prefix(stream)
530
self._expected_elements = 3 + self._key_length
532
# raw data keyed by offset
533
self._keys_by_offset = {}
534
# ready-to-return key:value or key:value, node_ref_lists
536
self._nodes_by_key = None
539
lines = stream.read().split(b'\n')
543
_, _, _, trailers = self._parse_lines(lines, pos)
544
for key, absent, references, value in self._keys_by_offset.values():
547
# resolve references:
548
if self.node_ref_lists:
549
node_value = (value, self._resolve_references(references))
552
self._nodes[key] = node_value
553
# cache the keys for quick set intersections
555
# there must be one line - the empty trailer line.
556
raise BadIndexData(self)
558
def clear_cache(self):
559
"""Clear out any cached/memoized values.
561
This can be called at any time, but generally it is used when we have
562
extracted some information, but don't expect to be requesting any more
566
def external_references(self, ref_list_num):
567
"""Return references that are not present in this index.
570
if ref_list_num + 1 > self.node_ref_lists:
571
raise ValueError('No ref list %d, index has %d ref lists'
572
% (ref_list_num, self.node_ref_lists))
575
for key, (value, ref_lists) in nodes.items():
576
ref_list = ref_lists[ref_list_num]
577
refs.update([ref for ref in ref_list if ref not in nodes])
580
def _get_nodes_by_key(self):
581
if self._nodes_by_key is None:
583
if self.node_ref_lists:
584
for key, (value, references) in self._nodes.items():
585
key_dict = nodes_by_key
586
for subkey in key[:-1]:
587
key_dict = key_dict.setdefault(subkey, {})
588
key_dict[key[-1]] = key, value, references
590
for key, value in self._nodes.items():
591
key_dict = nodes_by_key
592
for subkey in key[:-1]:
593
key_dict = key_dict.setdefault(subkey, {})
594
key_dict[key[-1]] = key, value
595
self._nodes_by_key = nodes_by_key
596
return self._nodes_by_key
598
def iter_all_entries(self):
599
"""Iterate over all keys within the index.
601
:return: An iterable of (index, key, value) or (index, key, value, reference_lists).
602
The former tuple is used when there are no reference lists in the
603
index, making the API compatible with simple key:value index types.
604
There is no defined order for the result iteration - it will be in
605
the most efficient order for the index.
607
if 'evil' in debug.debug_flags:
608
trace.mutter_callsite(3,
609
"iter_all_entries scales with size of history.")
610
if self._nodes is None:
612
if self.node_ref_lists:
613
for key, (value, node_ref_lists) in self._nodes.items():
614
yield self, key, value, node_ref_lists
616
for key, value in self._nodes.items():
617
yield self, key, value
619
def _read_prefix(self, stream):
620
signature = stream.read(len(self._signature()))
621
if not signature == self._signature():
622
raise BadIndexFormatSignature(self._name, GraphIndex)
623
options_line = stream.readline()
624
if not options_line.startswith(_OPTION_NODE_REFS):
625
raise BadIndexOptions(self)
627
self.node_ref_lists = int(options_line[len(_OPTION_NODE_REFS):-1])
629
raise BadIndexOptions(self)
630
options_line = stream.readline()
631
if not options_line.startswith(_OPTION_KEY_ELEMENTS):
632
raise BadIndexOptions(self)
634
self._key_length = int(options_line[len(_OPTION_KEY_ELEMENTS):-1])
636
raise BadIndexOptions(self)
637
options_line = stream.readline()
638
if not options_line.startswith(_OPTION_LEN):
639
raise BadIndexOptions(self)
641
self._key_count = int(options_line[len(_OPTION_LEN):-1])
643
raise BadIndexOptions(self)
645
def _resolve_references(self, references):
646
"""Return the resolved key references for references.
648
References are resolved by looking up the location of the key in the
649
_keys_by_offset map and substituting the key name, preserving ordering.
651
:param references: An iterable of iterables of key locations. e.g.
653
:return: A tuple of tuples of keys.
656
for ref_list in references:
658
tuple([self._keys_by_offset[ref][0] for ref in ref_list]))
659
return tuple(node_refs)
662
def _find_index(range_map, key):
663
"""Helper for the _parsed_*_index calls.
665
Given a range map - [(start, end), ...], finds the index of the range
666
in the map for key if it is in the map, and if it is not there, the
667
immediately preceeding range in the map.
669
result = bisect_right(range_map, key) - 1
670
if result + 1 < len(range_map):
671
# check the border condition, it may be in result + 1
672
if range_map[result + 1][0] == key[0]:
676
def _parsed_byte_index(self, offset):
677
"""Return the index of the entry immediately before offset.
679
e.g. if the parsed map has regions 0,10 and 11,12 parsed, meaning that
680
there is one unparsed byte (the 11th, addressed as[10]). then:
681
asking for 0 will return 0
682
asking for 10 will return 0
683
asking for 11 will return 1
684
asking for 12 will return 1
687
return self._find_index(self._parsed_byte_map, key)
689
def _parsed_key_index(self, key):
690
"""Return the index of the entry immediately before key.
692
e.g. if the parsed map has regions (None, 'a') and ('b','c') parsed,
693
meaning that keys from None to 'a' inclusive, and 'b' to 'c' inclusive
694
have been parsed, then:
695
asking for '' will return 0
696
asking for 'a' will return 0
697
asking for 'b' will return 1
698
asking for 'e' will return 1
700
search_key = (key, b'')
701
return self._find_index(self._parsed_key_map, search_key)
703
def _is_parsed(self, offset):
704
"""Returns True if offset has been parsed."""
705
index = self._parsed_byte_index(offset)
706
if index == len(self._parsed_byte_map):
707
return offset < self._parsed_byte_map[index - 1][1]
708
start, end = self._parsed_byte_map[index]
709
return offset >= start and offset < end
711
def _iter_entries_from_total_buffer(self, keys):
712
"""Iterate over keys when the entire index is parsed."""
713
# Note: See the note in BTreeBuilder.iter_entries for why we don't use
714
# .intersection() here
716
keys = [key for key in keys if key in nodes]
717
if self.node_ref_lists:
719
value, node_refs = nodes[key]
720
yield self, key, value, node_refs
723
yield self, key, nodes[key]
725
def iter_entries(self, keys):
726
"""Iterate over keys within the index.
728
:param keys: An iterable providing the keys to be retrieved.
729
:return: An iterable as per iter_all_entries, but restricted to the
730
keys supplied. No additional keys will be returned, and every
731
key supplied that is in the index will be returned.
736
if self._size is None and self._nodes is None:
739
# We fit about 20 keys per minimum-read (4K), so if we are looking for
740
# more than 1/20th of the index its likely (assuming homogenous key
741
# spread) that we'll read the entire index. If we're going to do that,
742
# buffer the whole thing. A better analysis might take key spread into
743
# account - but B+Tree indices are better anyway.
744
# We could look at all data read, and use a threshold there, which will
745
# trigger on ancestry walks, but that is not yet fully mapped out.
746
if self._nodes is None and len(keys) * 20 > self.key_count():
748
if self._nodes is not None:
749
return self._iter_entries_from_total_buffer(keys)
751
return (result[1] for result in bisect_multi.bisect_multi_bytes(
752
self._lookup_keys_via_location, self._size, keys))
754
def iter_entries_prefix(self, keys):
755
"""Iterate over keys within the index using prefix matching.
757
Prefix matching is applied within the tuple of a key, not to within
758
the bytestring of each key element. e.g. if you have the keys ('foo',
759
'bar'), ('foobar', 'gam') and do a prefix search for ('foo', None) then
760
only the former key is returned.
762
WARNING: Note that this method currently causes a full index parse
763
unconditionally (which is reasonably appropriate as it is a means for
764
thunking many small indices into one larger one and still supplies
765
iter_all_entries at the thunk layer).
767
:param keys: An iterable providing the key prefixes to be retrieved.
768
Each key prefix takes the form of a tuple the length of a key, but
769
with the last N elements 'None' rather than a regular bytestring.
770
The first element cannot be 'None'.
771
:return: An iterable as per iter_all_entries, but restricted to the
772
keys with a matching prefix to those supplied. No additional keys
773
will be returned, and every match that is in the index will be
779
# load data - also finds key lengths
780
if self._nodes is None:
782
if self._key_length == 1:
784
_sanity_check_key(self, key)
785
if self.node_ref_lists:
786
value, node_refs = self._nodes[key]
787
yield self, key, value, node_refs
789
yield self, key, self._nodes[key]
791
nodes_by_key = self._get_nodes_by_key()
792
for entry in _iter_entries_prefix(self, nodes_by_key, keys):
795
def _find_ancestors(self, keys, ref_list_num, parent_map, missing_keys):
796
"""See BTreeIndex._find_ancestors."""
797
# The api can be implemented as a trivial overlay on top of
798
# iter_entries, it is not an efficient implementation, but it at least
802
for index, key, value, refs in self.iter_entries(keys):
803
parent_keys = refs[ref_list_num]
805
parent_map[key] = parent_keys
806
search_keys.update(parent_keys)
807
# Figure out what, if anything, was missing
808
missing_keys.update(set(keys).difference(found_keys))
809
search_keys = search_keys.difference(parent_map)
813
"""Return an estimate of the number of keys in this index.
815
For GraphIndex the estimate is exact.
817
if self._key_count is None:
818
self._read_and_parse([_HEADER_READV])
819
return self._key_count
821
def _lookup_keys_via_location(self, location_keys):
822
"""Public interface for implementing bisection.
824
If _buffer_all has been called, then all the data for the index is in
825
memory, and this method should not be called, as it uses a separate
826
cache because it cannot pre-resolve all indices, which buffer_all does
829
:param location_keys: A list of location(byte offset), key tuples.
830
:return: A list of (location_key, result) tuples as expected by
831
breezy.bisect_multi.bisect_multi_bytes.
833
# Possible improvements:
834
# - only bisect lookup each key once
835
# - sort the keys first, and use that to reduce the bisection window
837
# this progresses in three parts:
840
# attempt to answer the question from the now in memory data.
841
# build the readv request
842
# for each location, ask for 800 bytes - much more than rows we've seen
845
for location, key in location_keys:
846
# can we answer from cache?
847
if self._bisect_nodes and key in self._bisect_nodes:
848
# We have the key parsed.
850
index = self._parsed_key_index(key)
851
if (len(self._parsed_key_map) and
852
self._parsed_key_map[index][0] <= key and
853
(self._parsed_key_map[index][1] >= key or
854
# end of the file has been parsed
855
self._parsed_byte_map[index][1] == self._size)):
856
# the key has been parsed, so no lookup is needed even if its
859
# - if we have examined this part of the file already - yes
860
index = self._parsed_byte_index(location)
861
if (len(self._parsed_byte_map) and
862
self._parsed_byte_map[index][0] <= location and
863
self._parsed_byte_map[index][1] > location):
864
# the byte region has been parsed, so no read is needed.
867
if location + length > self._size:
868
length = self._size - location
869
# todo, trim out parsed locations.
871
readv_ranges.append((location, length))
872
# read the header if needed
873
if self._bisect_nodes is None:
874
readv_ranges.append(_HEADER_READV)
875
self._read_and_parse(readv_ranges)
877
if self._nodes is not None:
878
# _read_and_parse triggered a _buffer_all because we requested the
880
for location, key in location_keys:
881
if key not in self._nodes: # not present
882
result.append(((location, key), False))
883
elif self.node_ref_lists:
884
value, refs = self._nodes[key]
885
result.append(((location, key),
886
(self, key, value, refs)))
888
result.append(((location, key),
889
(self, key, self._nodes[key])))
892
# - figure out <, >, missing, present
893
# - result present references so we can return them.
894
# keys that we cannot answer until we resolve references
895
pending_references = []
896
pending_locations = set()
897
for location, key in location_keys:
898
# can we answer from cache?
899
if key in self._bisect_nodes:
900
# the key has been parsed, so no lookup is needed
901
if self.node_ref_lists:
902
# the references may not have been all parsed.
903
value, refs = self._bisect_nodes[key]
904
wanted_locations = []
905
for ref_list in refs:
907
if ref not in self._keys_by_offset:
908
wanted_locations.append(ref)
910
pending_locations.update(wanted_locations)
911
pending_references.append((location, key))
913
result.append(((location, key), (self, key,
914
value, self._resolve_references(refs))))
916
result.append(((location, key),
917
(self, key, self._bisect_nodes[key])))
920
# has the region the key should be in, been parsed?
921
index = self._parsed_key_index(key)
922
if (self._parsed_key_map[index][0] <= key and
923
(self._parsed_key_map[index][1] >= key or
924
# end of the file has been parsed
925
self._parsed_byte_map[index][1] == self._size)):
926
result.append(((location, key), False))
928
# no, is the key above or below the probed location:
929
# get the range of the probed & parsed location
930
index = self._parsed_byte_index(location)
931
# if the key is below the start of the range, its below
932
if key < self._parsed_key_map[index][0]:
936
result.append(((location, key), direction))
938
# lookup data to resolve references
939
for location in pending_locations:
941
if location + length > self._size:
942
length = self._size - location
943
# TODO: trim out parsed locations (e.g. if the 800 is into the
944
# parsed region trim it, and dont use the adjust_for_latency
947
readv_ranges.append((location, length))
948
self._read_and_parse(readv_ranges)
949
if self._nodes is not None:
950
# The _read_and_parse triggered a _buffer_all, grab the data and
952
for location, key in pending_references:
953
value, refs = self._nodes[key]
954
result.append(((location, key), (self, key, value, refs)))
956
for location, key in pending_references:
957
# answer key references we had to look-up-late.
958
value, refs = self._bisect_nodes[key]
959
result.append(((location, key), (self, key,
960
value, self._resolve_references(refs))))
963
def _parse_header_from_bytes(self, bytes):
964
"""Parse the header from a region of bytes.
966
:param bytes: The data to parse.
967
:return: An offset, data tuple such as readv yields, for the unparsed
968
data. (which may length 0).
970
signature = bytes[0:len(self._signature())]
971
if not signature == self._signature():
972
raise BadIndexFormatSignature(self._name, GraphIndex)
973
lines = bytes[len(self._signature()):].splitlines()
974
options_line = lines[0]
975
if not options_line.startswith(_OPTION_NODE_REFS):
976
raise BadIndexOptions(self)
978
self.node_ref_lists = int(options_line[len(_OPTION_NODE_REFS):])
980
raise BadIndexOptions(self)
981
options_line = lines[1]
982
if not options_line.startswith(_OPTION_KEY_ELEMENTS):
983
raise BadIndexOptions(self)
985
self._key_length = int(options_line[len(_OPTION_KEY_ELEMENTS):])
987
raise BadIndexOptions(self)
988
options_line = lines[2]
989
if not options_line.startswith(_OPTION_LEN):
990
raise BadIndexOptions(self)
992
self._key_count = int(options_line[len(_OPTION_LEN):])
994
raise BadIndexOptions(self)
995
# calculate the bytes we have processed
996
header_end = (len(signature) + len(lines[0]) + len(lines[1]) +
998
self._parsed_bytes(0, (), header_end, ())
999
# setup parsing state
1000
self._expected_elements = 3 + self._key_length
1001
# raw data keyed by offset
1002
self._keys_by_offset = {}
1003
# keys with the value and node references
1004
self._bisect_nodes = {}
1005
return header_end, bytes[header_end:]
1007
def _parse_region(self, offset, data):
1008
"""Parse node data returned from a readv operation.
1010
:param offset: The byte offset the data starts at.
1011
:param data: The data to parse.
1015
end = offset + len(data)
1016
high_parsed = offset
1018
# Trivial test - if the current index's end is within the
1019
# low-matching parsed range, we're done.
1020
index = self._parsed_byte_index(high_parsed)
1021
if end < self._parsed_byte_map[index][1]:
1023
# print "[%d:%d]" % (offset, end), \
1024
# self._parsed_byte_map[index:index + 2]
1025
high_parsed, last_segment = self._parse_segment(
1026
offset, data, end, index)
1030
def _parse_segment(self, offset, data, end, index):
1031
"""Parse one segment of data.
1033
:param offset: Where 'data' begins in the file.
1034
:param data: Some data to parse a segment of.
1035
:param end: Where data ends
1036
:param index: The current index into the parsed bytes map.
1037
:return: True if the parsed segment is the last possible one in the
1039
:return: high_parsed_byte, last_segment.
1040
high_parsed_byte is the location of the highest parsed byte in this
1041
segment, last_segment is True if the parsed segment is the last
1042
possible one in the data block.
1044
# default is to use all data
1046
# accomodate overlap with data before this.
1047
if offset < self._parsed_byte_map[index][1]:
1048
# overlaps the lower parsed region
1049
# skip the parsed data
1050
trim_start = self._parsed_byte_map[index][1] - offset
1051
# don't trim the start for \n
1052
start_adjacent = True
1053
elif offset == self._parsed_byte_map[index][1]:
1054
# abuts the lower parsed region
1057
# do not trim anything
1058
start_adjacent = True
1060
# does not overlap the lower parsed region
1063
# but trim the leading \n
1064
start_adjacent = False
1065
if end == self._size:
1066
# lines up to the end of all data:
1069
# do not strip to the last \n
1072
elif index + 1 == len(self._parsed_byte_map):
1073
# at the end of the parsed data
1076
# but strip to the last \n
1077
end_adjacent = False
1079
elif end == self._parsed_byte_map[index + 1][0]:
1080
# buts up against the next parsed region
1083
# do not strip to the last \n
1086
elif end > self._parsed_byte_map[index + 1][0]:
1087
# overlaps into the next parsed region
1088
# only consider the unparsed data
1089
trim_end = self._parsed_byte_map[index + 1][0] - offset
1090
# do not strip to the last \n as we know its an entire record
1092
last_segment = end < self._parsed_byte_map[index + 1][1]
1094
# does not overlap into the next region
1097
# but strip to the last \n
1098
end_adjacent = False
1100
# now find bytes to discard if needed
1101
if not start_adjacent:
1102
# work around python bug in rfind
1103
if trim_start is None:
1104
trim_start = data.find(b'\n') + 1
1106
trim_start = data.find(b'\n', trim_start) + 1
1107
if not (trim_start != 0):
1108
raise AssertionError('no \n was present')
1109
# print 'removing start', offset, trim_start, repr(data[:trim_start])
1110
if not end_adjacent:
1111
# work around python bug in rfind
1112
if trim_end is None:
1113
trim_end = data.rfind(b'\n') + 1
1115
trim_end = data.rfind(b'\n', None, trim_end) + 1
1116
if not (trim_end != 0):
1117
raise AssertionError('no \n was present')
1118
# print 'removing end', offset, trim_end, repr(data[trim_end:])
1119
# adjust offset and data to the parseable data.
1120
trimmed_data = data[trim_start:trim_end]
1121
if not (trimmed_data):
1122
raise AssertionError('read unneeded data [%d:%d] from [%d:%d]'
1123
% (trim_start, trim_end, offset, offset + len(data)))
1125
offset += trim_start
1126
# print "parsing", repr(trimmed_data)
1127
# splitlines mangles the \r delimiters.. don't use it.
1128
lines = trimmed_data.split(b'\n')
1131
first_key, last_key, nodes, _ = self._parse_lines(lines, pos)
1132
for key, value in nodes:
1133
self._bisect_nodes[key] = value
1134
self._parsed_bytes(offset, first_key,
1135
offset + len(trimmed_data), last_key)
1136
return offset + len(trimmed_data), last_segment
1138
def _parse_lines(self, lines, pos):
1145
# must be at the end
1147
if not (self._size == pos + 1):
1148
raise AssertionError("%s %s" % (self._size, pos))
1151
elements = line.split(b'\0')
1152
if len(elements) != self._expected_elements:
1153
raise BadIndexData(self)
1154
# keys are tuples. Each element is a string that may occur many
1155
# times, so we intern them to save space. AB, RC, 200807
1156
key = tuple([element for element in elements[:self._key_length]])
1157
if first_key is None:
1159
absent, references, value = elements[-3:]
1161
for ref_string in references.split(b'\t'):
1162
ref_lists.append(tuple([
1163
int(ref) for ref in ref_string.split(b'\r') if ref
1165
ref_lists = tuple(ref_lists)
1166
self._keys_by_offset[pos] = (key, absent, ref_lists, value)
1167
pos += len(line) + 1 # +1 for the \n
1170
if self.node_ref_lists:
1171
node_value = (value, ref_lists)
1174
nodes.append((key, node_value))
1175
# print "parsed ", key
1176
return first_key, key, nodes, trailers
1178
def _parsed_bytes(self, start, start_key, end, end_key):
1179
"""Mark the bytes from start to end as parsed.
1181
Calling self._parsed_bytes(1,2) will mark one byte (the one at offset
1184
:param start: The start of the parsed region.
1185
:param end: The end of the parsed region.
1187
index = self._parsed_byte_index(start)
1188
new_value = (start, end)
1189
new_key = (start_key, end_key)
1191
# first range parsed is always the beginning.
1192
self._parsed_byte_map.insert(index, new_value)
1193
self._parsed_key_map.insert(index, new_key)
1197
# extend lower region
1198
# extend higher region
1199
# combine two regions
1200
if (index + 1 < len(self._parsed_byte_map) and
1201
self._parsed_byte_map[index][1] == start and
1202
self._parsed_byte_map[index + 1][0] == end):
1203
# combine two regions
1204
self._parsed_byte_map[index] = (self._parsed_byte_map[index][0],
1205
self._parsed_byte_map[index + 1][1])
1206
self._parsed_key_map[index] = (self._parsed_key_map[index][0],
1207
self._parsed_key_map[index + 1][1])
1208
del self._parsed_byte_map[index + 1]
1209
del self._parsed_key_map[index + 1]
1210
elif self._parsed_byte_map[index][1] == start:
1211
# extend the lower entry
1212
self._parsed_byte_map[index] = (
1213
self._parsed_byte_map[index][0], end)
1214
self._parsed_key_map[index] = (
1215
self._parsed_key_map[index][0], end_key)
1216
elif (index + 1 < len(self._parsed_byte_map) and
1217
self._parsed_byte_map[index + 1][0] == end):
1218
# extend the higher entry
1219
self._parsed_byte_map[index + 1] = (
1220
start, self._parsed_byte_map[index + 1][1])
1221
self._parsed_key_map[index + 1] = (
1222
start_key, self._parsed_key_map[index + 1][1])
1225
self._parsed_byte_map.insert(index + 1, new_value)
1226
self._parsed_key_map.insert(index + 1, new_key)
1228
def _read_and_parse(self, readv_ranges):
1229
"""Read the ranges and parse the resulting data.
1231
:param readv_ranges: A prepared readv range list.
1233
if not readv_ranges:
1235
if self._nodes is None and self._bytes_read * 2 >= self._size:
1236
# We've already read more than 50% of the file and we are about to
1237
# request more data, just _buffer_all() and be done
1241
base_offset = self._base_offset
1242
if base_offset != 0:
1243
# Rewrite the ranges for the offset
1244
readv_ranges = [(start + base_offset, size)
1245
for start, size in readv_ranges]
1246
readv_data = self._transport.readv(self._name, readv_ranges, True,
1247
self._size + self._base_offset)
1249
for offset, data in readv_data:
1250
offset -= base_offset
1251
self._bytes_read += len(data)
1253
# transport.readv() expanded to extra data which isn't part of
1255
data = data[-offset:]
1257
if offset == 0 and len(data) == self._size:
1258
# We read the whole range, most likely because the
1259
# Transport upcast our readv ranges into one long request
1260
# for enough total data to grab the whole index.
1261
self._buffer_all(BytesIO(data))
1263
if self._bisect_nodes is None:
1264
# this must be the start
1265
if not (offset == 0):
1266
raise AssertionError()
1267
offset, data = self._parse_header_from_bytes(data)
1268
# print readv_ranges, "[%d:%d]" % (offset, offset + len(data))
1269
self._parse_region(offset, data)
1271
def _signature(self):
1272
"""The file signature for this index type."""
1276
"""Validate that everything in the index can be accessed."""
1277
# iter_all validates completely at the moment, so just do that.
1278
for node in self.iter_all_entries():
1282
class CombinedGraphIndex(object):
1283
"""A GraphIndex made up from smaller GraphIndices.
1285
The backing indices must implement GraphIndex, and are presumed to be
1288
Queries against the combined index will be made against the first index,
1289
and then the second and so on. The order of indices can thus influence
1290
performance significantly. For example, if one index is on local disk and a
1291
second on a remote server, the local disk index should be before the other
1294
Also, queries tend to need results from the same indices as previous
1295
queries. So the indices will be reordered after every query to put the
1296
indices that had the result(s) of that query first (while otherwise
1297
preserving the relative ordering).
1300
def __init__(self, indices, reload_func=None):
1301
"""Create a CombinedGraphIndex backed by indices.
1303
:param indices: An ordered list of indices to query for data.
1304
:param reload_func: A function to call if we find we are missing an
1305
index. Should have the form reload_func() => True/False to indicate
1306
if reloading actually changed anything.
1308
self._indices = indices
1309
self._reload_func = reload_func
1310
# Sibling indices are other CombinedGraphIndex that we should call
1311
# _move_to_front_by_name on when we auto-reorder ourself.
1312
self._sibling_indices = []
1313
# A list of names that corresponds to the instances in self._indices,
1314
# so _index_names[0] is always the name for _indices[0], etc. Sibling
1315
# indices must all use the same set of names as each other.
1316
self._index_names = [None] * len(self._indices)
1320
self.__class__.__name__,
1321
', '.join(map(repr, self._indices)))
1323
def clear_cache(self):
1324
"""See GraphIndex.clear_cache()"""
1325
for index in self._indices:
1328
def get_parent_map(self, keys):
1329
"""See graph.StackedParentsProvider.get_parent_map"""
1330
search_keys = set(keys)
1331
if _mod_revision.NULL_REVISION in search_keys:
1332
search_keys.discard(_mod_revision.NULL_REVISION)
1333
found_parents = {_mod_revision.NULL_REVISION: []}
1336
for index, key, value, refs in self.iter_entries(search_keys):
1339
parents = (_mod_revision.NULL_REVISION,)
1340
found_parents[key] = parents
1341
return found_parents
1343
__contains__ = _has_key_from_parent_map
1345
def insert_index(self, pos, index, name=None):
1346
"""Insert a new index in the list of indices to query.
1348
:param pos: The position to insert the index.
1349
:param index: The index to insert.
1350
:param name: a name for this index, e.g. a pack name. These names can
1351
be used to reflect index reorderings to related CombinedGraphIndex
1352
instances that use the same names. (see set_sibling_indices)
1354
self._indices.insert(pos, index)
1355
self._index_names.insert(pos, name)
1357
def iter_all_entries(self):
1358
"""Iterate over all keys within the index
1360
Duplicate keys across child indices are presumed to have the same
1361
value and are only reported once.
1363
:return: An iterable of (index, key, reference_lists, value).
1364
There is no defined order for the result iteration - it will be in
1365
the most efficient order for the index.
1370
for index in self._indices:
1371
for node in index.iter_all_entries():
1372
if node[1] not in seen_keys:
1374
seen_keys.add(node[1])
1376
except errors.NoSuchFile as e:
1377
if not self._try_reload(e):
1380
def iter_entries(self, keys):
1381
"""Iterate over keys within the index.
1383
Duplicate keys across child indices are presumed to have the same
1384
value and are only reported once.
1386
:param keys: An iterable providing the keys to be retrieved.
1387
:return: An iterable of (index, key, reference_lists, value). There is
1388
no defined order for the result iteration - it will be in the most
1389
efficient order for the index.
1395
for index in self._indices:
1399
for node in index.iter_entries(keys):
1400
keys.remove(node[1])
1404
hit_indices.append(index)
1406
except errors.NoSuchFile as e:
1407
if not self._try_reload(e):
1409
self._move_to_front(hit_indices)
1411
def iter_entries_prefix(self, keys):
1412
"""Iterate over keys within the index using prefix matching.
1414
Duplicate keys across child indices are presumed to have the same
1415
value and are only reported once.
1417
Prefix matching is applied within the tuple of a key, not to within
1418
the bytestring of each key element. e.g. if you have the keys ('foo',
1419
'bar'), ('foobar', 'gam') and do a prefix search for ('foo', None) then
1420
only the former key is returned.
1422
:param keys: An iterable providing the key prefixes to be retrieved.
1423
Each key prefix takes the form of a tuple the length of a key, but
1424
with the last N elements 'None' rather than a regular bytestring.
1425
The first element cannot be 'None'.
1426
:return: An iterable as per iter_all_entries, but restricted to the
1427
keys with a matching prefix to those supplied. No additional keys
1428
will be returned, and every match that is in the index will be
1438
for index in self._indices:
1440
for node in index.iter_entries_prefix(keys):
1441
if node[1] in seen_keys:
1443
seen_keys.add(node[1])
1447
hit_indices.append(index)
1449
except errors.NoSuchFile as e:
1450
if not self._try_reload(e):
1452
self._move_to_front(hit_indices)
1454
def _move_to_front(self, hit_indices):
1455
"""Rearrange self._indices so that hit_indices are first.
1457
Order is maintained as much as possible, e.g. the first unhit index
1458
will be the first index in _indices after the hit_indices, and the
1459
hit_indices will be present in exactly the order they are passed to
1462
_move_to_front propagates to all objects in self._sibling_indices by
1463
calling _move_to_front_by_name.
1465
if self._indices[:len(hit_indices)] == hit_indices:
1466
# The 'hit_indices' are already at the front (and in the same
1467
# order), no need to re-order
1469
hit_names = self._move_to_front_by_index(hit_indices)
1470
for sibling_idx in self._sibling_indices:
1471
sibling_idx._move_to_front_by_name(hit_names)
1473
def _move_to_front_by_index(self, hit_indices):
1474
"""Core logic for _move_to_front.
1476
Returns a list of names corresponding to the hit_indices param.
1478
indices_info = zip(self._index_names, self._indices)
1479
if 'index' in debug.debug_flags:
1480
indices_info = list(indices_info)
1481
trace.mutter('CombinedGraphIndex reordering: currently %r, '
1482
'promoting %r', indices_info, hit_indices)
1485
new_hit_indices = []
1488
for offset, (name, idx) in enumerate(indices_info):
1489
if idx in hit_indices:
1490
hit_names.append(name)
1491
new_hit_indices.append(idx)
1492
if len(new_hit_indices) == len(hit_indices):
1493
# We've found all of the hit entries, everything else is
1495
unhit_names.extend(self._index_names[offset + 1:])
1496
unhit_indices.extend(self._indices[offset + 1:])
1499
unhit_names.append(name)
1500
unhit_indices.append(idx)
1502
self._indices = new_hit_indices + unhit_indices
1503
self._index_names = hit_names + unhit_names
1504
if 'index' in debug.debug_flags:
1505
trace.mutter('CombinedGraphIndex reordered: %r', self._indices)
1508
def _move_to_front_by_name(self, hit_names):
1509
"""Moves indices named by 'hit_names' to front of the search order, as
1510
described in _move_to_front.
1512
# Translate names to index instances, and then call
1513
# _move_to_front_by_index.
1514
indices_info = zip(self._index_names, self._indices)
1516
for name, idx in indices_info:
1517
if name in hit_names:
1518
hit_indices.append(idx)
1519
self._move_to_front_by_index(hit_indices)
1521
def find_ancestry(self, keys, ref_list_num):
1522
"""Find the complete ancestry for the given set of keys.
1524
Note that this is a whole-ancestry request, so it should be used
1527
:param keys: An iterable of keys to look for
1528
:param ref_list_num: The reference list which references the parents
1530
:return: (parent_map, missing_keys)
1532
# XXX: make this call _move_to_front?
1533
missing_keys = set()
1535
keys_to_lookup = set(keys)
1537
while keys_to_lookup:
1538
# keys that *all* indexes claim are missing, stop searching them
1540
all_index_missing = None
1541
# print 'gen\tidx\tsub\tn_keys\tn_pmap\tn_miss'
1542
# print '%4d\t\t\t%4d\t%5d\t%5d' % (generation, len(keys_to_lookup),
1544
# len(missing_keys))
1545
for index_idx, index in enumerate(self._indices):
1546
# TODO: we should probably be doing something with
1547
# 'missing_keys' since we've already determined that
1548
# those revisions have not been found anywhere
1549
index_missing_keys = set()
1550
# Find all of the ancestry we can from this index
1551
# keep looking until the search_keys set is empty, which means
1552
# things we didn't find should be in index_missing_keys
1553
search_keys = keys_to_lookup
1555
# print ' \t%2d\t\t%4d\t%5d\t%5d' % (
1556
# index_idx, len(search_keys),
1557
# len(parent_map), len(index_missing_keys))
1560
# TODO: ref_list_num should really be a parameter, since
1561
# CombinedGraphIndex does not know what the ref lists
1563
search_keys = index._find_ancestors(search_keys,
1564
ref_list_num, parent_map, index_missing_keys)
1565
# print ' \t \t%2d\t%4d\t%5d\t%5d' % (
1566
# sub_generation, len(search_keys),
1567
# len(parent_map), len(index_missing_keys))
1568
# Now set whatever was missing to be searched in the next index
1569
keys_to_lookup = index_missing_keys
1570
if all_index_missing is None:
1571
all_index_missing = set(index_missing_keys)
1573
all_index_missing.intersection_update(index_missing_keys)
1574
if not keys_to_lookup:
1576
if all_index_missing is None:
1577
# There were no indexes, so all search keys are 'missing'
1578
missing_keys.update(keys_to_lookup)
1579
keys_to_lookup = None
1581
missing_keys.update(all_index_missing)
1582
keys_to_lookup.difference_update(all_index_missing)
1583
return parent_map, missing_keys
1585
def key_count(self):
1586
"""Return an estimate of the number of keys in this index.
1588
For CombinedGraphIndex this is approximated by the sum of the keys of
1589
the child indices. As child indices may have duplicate keys this can
1590
have a maximum error of the number of child indices * largest number of
1595
return sum((index.key_count() for index in self._indices), 0)
1596
except errors.NoSuchFile as e:
1597
if not self._try_reload(e):
1600
missing_keys = _missing_keys_from_parent_map
1602
def _try_reload(self, error):
1603
"""We just got a NoSuchFile exception.
1605
Try to reload the indices, if it fails, just raise the current
1608
if self._reload_func is None:
1611
'Trying to reload after getting exception: %s', str(error))
1612
if not self._reload_func():
1613
# We tried to reload, but nothing changed, so we fail anyway
1614
trace.mutter('_reload_func indicated nothing has changed.'
1615
' Raising original exception.')
1619
def set_sibling_indices(self, sibling_combined_graph_indices):
1620
"""Set the CombinedGraphIndex objects to reorder after reordering self.
1622
self._sibling_indices = sibling_combined_graph_indices
1625
"""Validate that everything in the index can be accessed."""
1628
for index in self._indices:
1631
except errors.NoSuchFile as e:
1632
if not self._try_reload(e):
1636
class InMemoryGraphIndex(GraphIndexBuilder):
1637
"""A GraphIndex which operates entirely out of memory and is mutable.
1639
This is designed to allow the accumulation of GraphIndex entries during a
1640
single write operation, where the accumulated entries need to be immediately
1641
available - for example via a CombinedGraphIndex.
1644
def add_nodes(self, nodes):
1645
"""Add nodes to the index.
1647
:param nodes: An iterable of (key, node_refs, value) entries to add.
1649
if self.reference_lists:
1650
for (key, value, node_refs) in nodes:
1651
self.add_node(key, value, node_refs)
1653
for (key, value) in nodes:
1654
self.add_node(key, value)
1656
def iter_all_entries(self):
1657
"""Iterate over all keys within the index
1659
:return: An iterable of (index, key, reference_lists, value). There is no
1660
defined order for the result iteration - it will be in the most
1661
efficient order for the index (in this case dictionary hash order).
1663
if 'evil' in debug.debug_flags:
1664
trace.mutter_callsite(3,
1665
"iter_all_entries scales with size of history.")
1666
if self.reference_lists:
1667
for key, (absent, references, value) in self._nodes.items():
1669
yield self, key, value, references
1671
for key, (absent, references, value) in self._nodes.items():
1673
yield self, key, value
1675
def iter_entries(self, keys):
1676
"""Iterate over keys within the index.
1678
:param keys: An iterable providing the keys to be retrieved.
1679
:return: An iterable of (index, key, value, reference_lists). There is no
1680
defined order for the result iteration - it will be in the most
1681
efficient order for the index (keys iteration order in this case).
1683
# Note: See BTreeBuilder.iter_entries for an explanation of why we
1684
# aren't using set().intersection() here
1686
keys = [key for key in keys if key in nodes]
1687
if self.reference_lists:
1691
yield self, key, node[2], node[1]
1696
yield self, key, node[2]
1698
def iter_entries_prefix(self, keys):
1699
"""Iterate over keys within the index using prefix matching.
1701
Prefix matching is applied within the tuple of a key, not to within
1702
the bytestring of each key element. e.g. if you have the keys ('foo',
1703
'bar'), ('foobar', 'gam') and do a prefix search for ('foo', None) then
1704
only the former key is returned.
1706
:param keys: An iterable providing the key prefixes to be retrieved.
1707
Each key prefix takes the form of a tuple the length of a key, but
1708
with the last N elements 'None' rather than a regular bytestring.
1709
The first element cannot be 'None'.
1710
:return: An iterable as per iter_all_entries, but restricted to the
1711
keys with a matching prefix to those supplied. No additional keys
1712
will be returned, and every match that is in the index will be
1718
if self._key_length == 1:
1720
_sanity_check_key(self, key)
1721
node = self._nodes[key]
1724
if self.reference_lists:
1725
yield self, key, node[2], node[1]
1727
yield self, key, node[2]
1729
nodes_by_key = self._get_nodes_by_key()
1730
for entry in _iter_entries_prefix(self, nodes_by_key, keys):
1733
def key_count(self):
1734
"""Return an estimate of the number of keys in this index.
1736
For InMemoryGraphIndex the estimate is exact.
1738
return len(self._nodes) - len(self._absent_keys)
1741
"""In memory index's have no known corruption at the moment."""
1743
def __lt__(self, other):
1744
# We don't really care about the order, just that there is an order.
1745
if (not isinstance(other, GraphIndex) and
1746
not isinstance(other, InMemoryGraphIndex)):
1747
raise TypeError(other)
1748
return hash(self) < hash(other)
1751
class GraphIndexPrefixAdapter(object):
1752
"""An adapter between GraphIndex with different key lengths.
1754
Queries against this will emit queries against the adapted Graph with the
1755
prefix added, queries for all items use iter_entries_prefix. The returned
1756
nodes will have their keys and node references adjusted to remove the
1757
prefix. Finally, an add_nodes_callback can be supplied - when called the
1758
nodes and references being added will have prefix prepended.
1761
def __init__(self, adapted, prefix, missing_key_length,
1762
add_nodes_callback=None):
1763
"""Construct an adapter against adapted with prefix."""
1764
self.adapted = adapted
1765
self.prefix_key = prefix + (None,) * missing_key_length
1766
self.prefix = prefix
1767
self.prefix_len = len(prefix)
1768
self.add_nodes_callback = add_nodes_callback
1770
def add_nodes(self, nodes):
1771
"""Add nodes to the index.
1773
:param nodes: An iterable of (key, node_refs, value) entries to add.
1775
# save nodes in case its an iterator
1776
nodes = tuple(nodes)
1777
translated_nodes = []
1779
# Add prefix_key to each reference node_refs is a tuple of tuples,
1780
# so split it apart, and add prefix_key to the internal reference
1781
for (key, value, node_refs) in nodes:
1782
adjusted_references = (
1783
tuple(tuple(self.prefix + ref_node for ref_node in ref_list)
1784
for ref_list in node_refs))
1785
translated_nodes.append((self.prefix + key, value,
1786
adjusted_references))
1788
# XXX: TODO add an explicit interface for getting the reference list
1789
# status, to handle this bit of user-friendliness in the API more
1791
for (key, value) in nodes:
1792
translated_nodes.append((self.prefix + key, value))
1793
self.add_nodes_callback(translated_nodes)
1795
def add_node(self, key, value, references=()):
1796
"""Add a node to the index.
1798
:param key: The key. keys are non-empty tuples containing
1799
as many whitespace-free utf8 bytestrings as the key length
1800
defined for this index.
1801
:param references: An iterable of iterables of keys. Each is a
1802
reference to another key.
1803
:param value: The value to associate with the key. It may be any
1804
bytes as long as it does not contain \0 or \n.
1806
self.add_nodes(((key, value, references), ))
1808
def _strip_prefix(self, an_iter):
1809
"""Strip prefix data from nodes and return it."""
1810
for node in an_iter:
1812
if node[1][:self.prefix_len] != self.prefix:
1813
raise BadIndexData(self)
1814
for ref_list in node[3]:
1815
for ref_node in ref_list:
1816
if ref_node[:self.prefix_len] != self.prefix:
1817
raise BadIndexData(self)
1818
yield node[0], node[1][self.prefix_len:], node[2], (
1819
tuple(tuple(ref_node[self.prefix_len:] for ref_node in ref_list)
1820
for ref_list in node[3]))
1822
def iter_all_entries(self):
1823
"""Iterate over all keys within the index
1825
iter_all_entries is implemented against the adapted index using
1826
iter_entries_prefix.
1828
:return: An iterable of (index, key, reference_lists, value). There is no
1829
defined order for the result iteration - it will be in the most
1830
efficient order for the index (in this case dictionary hash order).
1832
return self._strip_prefix(self.adapted.iter_entries_prefix([self.prefix_key]))
1834
def iter_entries(self, keys):
1835
"""Iterate over keys within the index.
1837
:param keys: An iterable providing the keys to be retrieved.
1838
:return: An iterable of (index, key, value, reference_lists). There is no
1839
defined order for the result iteration - it will be in the most
1840
efficient order for the index (keys iteration order in this case).
1842
return self._strip_prefix(self.adapted.iter_entries(
1843
self.prefix + key for key in keys))
1845
def iter_entries_prefix(self, keys):
1846
"""Iterate over keys within the index using prefix matching.
1848
Prefix matching is applied within the tuple of a key, not to within
1849
the bytestring of each key element. e.g. if you have the keys ('foo',
1850
'bar'), ('foobar', 'gam') and do a prefix search for ('foo', None) then
1851
only the former key is returned.
1853
:param keys: An iterable providing the key prefixes to be retrieved.
1854
Each key prefix takes the form of a tuple the length of a key, but
1855
with the last N elements 'None' rather than a regular bytestring.
1856
The first element cannot be 'None'.
1857
:return: An iterable as per iter_all_entries, but restricted to the
1858
keys with a matching prefix to those supplied. No additional keys
1859
will be returned, and every match that is in the index will be
1862
return self._strip_prefix(self.adapted.iter_entries_prefix(
1863
self.prefix + key for key in keys))
1865
def key_count(self):
1866
"""Return an estimate of the number of keys in this index.
1868
For GraphIndexPrefixAdapter this is relatively expensive - key
1869
iteration with the prefix is done.
1871
return len(list(self.iter_all_entries()))
1874
"""Call the adapted's validate."""
1875
self.adapted.validate()
1878
def _sanity_check_key(index_or_builder, key):
1879
"""Raise BadIndexKey if key cannot be used for prefix matching."""
1881
raise BadIndexKey(key)
1882
if len(key) != index_or_builder._key_length:
1883
raise BadIndexKey(key)
1886
def _iter_entries_prefix(index_or_builder, nodes_by_key, keys):
1887
"""Helper for implementing prefix matching iterators."""
1889
_sanity_check_key(index_or_builder, key)
1890
# find what it refers to:
1891
key_dict = nodes_by_key
1892
elements = list(key)
1893
# find the subdict whose contents should be returned.
1895
while len(elements) and elements[0] is not None:
1896
key_dict = key_dict[elements[0]]
1899
# a non-existant lookup.
1904
values_view = dicts.pop().values()
1905
# can't be empty or would not exist
1906
value = next(iter(values_view))
1907
if isinstance(value, dict):
1908
# still descending, push values
1909
dicts.extend(values_view)
1911
# at leaf tuples, yield values
1912
for value in values_view:
1913
# each value is the key:value:node refs tuple
1915
yield (index_or_builder, ) + value
1917
# the last thing looked up was a terminal element
1918
yield (index_or_builder, ) + key_dict
added
removed
breezy/bzr/index.py
