1
# Copyright (C) 2006, 2007 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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
"""DirState objects record the state of a directory and its bzr metadata.
19
Pseudo EBNF grammar for the state file. Fields are separated by NULLs, and
20
lines by NL. The field delimiters are ommitted in the grammar, line delimiters
21
are not - this is done for clarity of reading. All string data is in utf8.
23
MINIKIND = "f" | "d" | "l" | "a" | "r";
26
WHOLE_NUMBER = {digit}, digit;
28
REVISION_ID = a non-empty utf8 string;
30
dirstate format = header line, full checksum, row count, parent details,
31
ghost_details, entries;
32
header line = "#bazaar dirstate flat format 2", NL;
33
full checksum = "adler32: ", ["-"], WHOLE_NUMBER, NL;
34
row count = "num_entries: ", digit, NL;
35
parent_details = WHOLE NUMBER, {REVISION_ID}* NL;
36
ghost_details = WHOLE NUMBER, {REVISION_ID}*, NL;
38
entry = entry_key, current_entry_details, {parent_entry_details};
39
entry_key = dirname, basename, fileid;
40
current_entry_details = common_entry_details, working_entry_details;
41
parent_entry_details = common_entry_details, history_entry_details;
42
common_entry_details = MINIKIND, fingerprint, size, executable
43
working_entry_details = packed_stat
44
history_entry_details = REVISION_ID;
47
fingerprint = a nonempty utf8 sequence with meaning defined by minikind.
49
Given this definition, the following is useful to know:
50
entry (aka row) - all the data for a given key.
51
entry[0]: The key (dirname, basename, fileid)
55
entry[1]: The tree(s) data for this path and id combination.
56
entry[1][0]: The current tree
57
entry[1][1]: The second tree
59
For an entry for a tree, we have (using tree 0 - current tree) to demonstrate:
60
entry[1][0][0]: minikind
61
entry[1][0][1]: fingerprint
63
entry[1][0][3]: executable
64
entry[1][0][4]: packed_stat
66
entry[1][1][4]: revision_id
68
There may be multiple rows at the root, one per id present in the root, so the
69
in memory root row is now:
70
self._dirblocks[0] -> ('', [entry ...]),
71
and the entries in there are
74
entries[0][2]: file_id
75
entries[1][0]: The tree data for the current tree for this fileid at /
79
'r' is a relocated entry: This path is not present in this tree with this id,
80
but the id can be found at another location. The fingerprint is used to
81
point to the target location.
82
'a' is an absent entry: In that tree the id is not present at this path.
83
'd' is a directory entry: This path in this tree is a directory with the
84
current file id. There is no fingerprint for directories.
85
'f' is a file entry: As for directory, but its a file. The fingerprint is a
87
'l' is a symlink entry: As for directory, but a symlink. The fingerprint is the
91
--- Format 1 had the following different definition: ---
92
rows = dirname, NULL, basename, NULL, MINIKIND, NULL, fileid_utf8, NULL,
93
WHOLE NUMBER (* size *), NULL, packed stat, NULL, sha1|symlink target,
95
PARENT ROW = NULL, revision_utf8, NULL, MINIKIND, NULL, dirname, NULL,
96
basename, NULL, WHOLE NUMBER (* size *), NULL, "y" | "n", NULL,
99
PARENT ROW's are emitted for every parent that is not in the ghosts details
100
line. That is, if the parents are foo, bar, baz, and the ghosts are bar, then
101
each row will have a PARENT ROW for foo and baz, but not for bar.
104
In any tree, a kind of 'moved' indicates that the fingerprint field
105
(which we treat as opaque data specific to the 'kind' anyway) has the
106
details for the id of this row in that tree.
108
I'm strongly tempted to add a id->path index as well, but I think that
109
where we need id->path mapping; we also usually read the whole file, so
110
I'm going to skip that for the moment, as we have the ability to locate
111
via bisect any path in any tree, and if we lookup things by path, we can
112
accumulate a id->path mapping as we go, which will tend to match what we
115
I plan to implement this asap, so please speak up now to alter/tweak the
116
design - and once we stabilise on this, I'll update the wiki page for
119
The rationale for all this is that we want fast operations for the
120
common case (diff/status/commit/merge on all files) and extremely fast
121
operations for the less common but still occurs a lot status/diff/commit
122
on specific files). Operations on specific files involve a scan for all
123
the children of a path, *in every involved tree*, which the current
124
format did not accommodate.
128
1) Fast end to end use for bzr's top 5 uses cases. (commmit/diff/status/merge/???)
129
2) fall back current object model as needed.
130
3) scale usably to the largest trees known today - say 50K entries. (mozilla
131
is an example of this)
135
Eventually reuse dirstate objects across locks IFF the dirstate file has not
136
been modified, but will require that we flush/ignore cached stat-hit data
137
because we wont want to restat all files on disk just because a lock was
138
acquired, yet we cannot trust the data after the previous lock was released.
140
Memory representation:
141
vector of all directories, and vector of the childen ?
143
root_entrie = (direntry for root, [parent_direntries_for_root]),
145
('', ['data for achild', 'data for bchild', 'data for cchild'])
146
('dir', ['achild', 'cchild', 'echild'])
148
- single bisect to find N subtrees from a path spec
149
- in-order for serialisation - this is 'dirblock' grouping.
150
- insertion of a file '/a' affects only the '/' child-vector, that is, to
151
insert 10K elements from scratch does not generates O(N^2) memoves of a
152
single vector, rather each individual, which tends to be limited to a
153
manageable number. Will scale badly on trees with 10K entries in a
154
single directory. compare with Inventory.InventoryDirectory which has
155
a dictionary for the children. No bisect capability, can only probe for
156
exact matches, or grab all elements and sorta.
157
- Whats the risk of error here? Once we have the base format being processed
158
we should have a net win regardless of optimality. So we are going to
159
go with what seems reasonably.
162
maybe we should do a test profile of these core structure - 10K simulated searches/lookups/etc?
164
Objects for each row?
165
The lifetime of Dirstate objects is current per lock, but see above for
166
possible extensions. The lifetime of a row from a dirstate is expected to be
167
very short in the optimistic case: which we are optimising for. For instance,
168
subtree status will determine from analysis of the disk data what rows need to
169
be examined at all, and will be able to determine from a single row whether
170
that file has altered or not, so we are aiming to process tens of thousands of
171
entries each second within the dirstate context, before exposing anything to
172
the larger codebase. This suggests we want the time for a single file
173
comparison to be < 0.1 milliseconds. That would give us 10000 paths per second
174
processed, and to scale to 100 thousand we'll another order of magnitude to do
175
that. Now, as the lifetime for all unchanged entries is the time to parse, stat
176
the file on disk, and then immediately discard, the overhead of object creation
177
becomes a significant cost.
179
Figures: Creating a tuple from from 3 elements was profiled at 0.0625
180
microseconds, whereas creating a object which is subclassed from tuple was
181
0.500 microseconds, and creating an object with 3 elements and slots was 3
182
microseconds long. 0.1 milliseconds is 100 microseconds, and ideally we'll get
183
down to 10 microseconds for the total processing - having 33% of that be object
184
creation is a huge overhead. There is a potential cost in using tuples within
185
each row which is that the conditional code to do comparisons may be slower
186
than method invocation, but method invocation is known to be slow due to stack
187
frame creation, so avoiding methods in these tight inner loops in unfortunately
188
desirable. We can consider a pyrex version of this with objects in future if
207
import bzrlib.inventory
208
from bzrlib import osutils
209
from bzrlib.osutils import (
217
class _Bisector(object):
218
"""This just keeps track of information as we are bisecting."""
221
class DirState(object):
222
"""Record directory and metadata state for fast access.
224
A dirstate is a specialised data structure for managing local working
225
tree state information. Its not yet well defined whether it is platform
226
specific, and if it is how we detect/parameterise that.
229
_kind_to_minikind = {'absent':'a', 'file':'f', 'directory':'d', 'relocated':'r', 'symlink':'l'}
230
_minikind_to_kind = {'a':'absent', 'f':'file', 'd':'directory', 'l':'symlink', 'r':'relocated'}
231
_to_yesno = {True:'y', False: 'n'} # TODO profile the performance gain
232
# of using int conversion rather than a dict here. AND BLAME ANDREW IF
235
# TODO: jam 20070221 Make sure we handle when there are duplicated records
236
# (like when we remove + add the same path, or we have a rename)
237
# TODO: jam 20070221 Figure out what to do if we have a record that exceeds
238
# the BISECT_PAGE_SIZE. For now, we just have to make it large enough
239
# that we are sure a single record will always fit.
240
BISECT_PAGE_SIZE = 4096
243
IN_MEMORY_UNMODIFIED = 1
244
IN_MEMORY_MODIFIED = 2
246
# A pack_stat (the x's) that is just noise and will never match the output
249
NULL_PARENT_DETAILS = ('a', '', 0, False, '')
251
def __init__(self, path):
252
"""Create a DirState object.
256
:attr _root_entrie: The root row of the directory/file information,
257
- contains the path to / - '', ''
258
- kind of 'directory',
259
- the file id of the root in utf8
262
- and no sha information.
263
:param path: The path at which the dirstate file on disk should live.
265
# _header_state and _dirblock_state represent the current state
266
# of the dirstate metadata and the per-row data respectiely.
267
# NOT_IN_MEMORY indicates that no data is in memory
268
# IN_MEMORY_UNMODIFIED indicates that what we have in memory
269
# is the same as is on disk
270
# IN_MEMORY_MODIFIED indicates that we have a modified version
271
# of what is on disk.
272
# In future we will add more granularity, for instance _dirblock_state
273
# will probably support partially-in-memory as a separate variable,
274
# allowing for partially-in-memory unmodified and partially-in-memory
276
self._header_state = DirState.NOT_IN_MEMORY
277
self._dirblock_state = DirState.NOT_IN_MEMORY
281
self._state_file = None
282
self._filename = path
283
self._lock_token = None
284
self._id_index = None
285
self._end_of_header = None
286
self._bisect_page_size = DirState.BISECT_PAGE_SIZE
288
def add(self, path, file_id, kind, stat, link_or_sha1):
289
"""Add a path to be tracked.
291
:param path: The path within the dirstate - '' is the root, 'foo' is the
292
path foo within the root, 'foo/bar' is the path bar within foo
294
:param file_id: The file id of the path being added.
295
:param kind: The kind of the path.
296
:param stat: The output of os.lstate for the path.
297
:param link_or_sha1: The sha value of the file, or the target of a
298
symlink. '' for directories.
301
# find the block its in.
302
# find the location in the block.
303
# check its not there
305
#------- copied from bzrlib.inventory.make_entry
306
# --- normalized_filename wants a unicode basename only, so get one.
307
dirname, basename = osutils.split(path)
308
# we dont import normalized_filename directly because we want to be
309
# able to change the implementation at runtime for tests.
310
norm_name, can_access = osutils.normalized_filename(basename)
311
if norm_name != basename:
315
raise errors.InvalidNormalization(path)
316
# now that we've normalised, we need the correct utf8 path and
317
# dirname and basename elements. This single encode and split should be
318
# faster than three separate encodes.
319
utf8path = (dirname + '/' + basename).strip('/').encode('utf8')
320
dirname, basename = osutils.split(utf8path)
321
assert file_id.__class__ == str, \
322
"must be a utf8 file_id not %s" % (type(file_id))
323
entry_key = (dirname, basename, file_id)
324
self._read_dirblocks_if_needed()
325
block_index, present = self._find_block_index_from_key(entry_key)
327
# TODO: This test is not complete - an empty directory, or a
328
# directory for a parent tree will fool it.
329
# some parent path has not been added - its an error to add this
331
raise errors.NotVersionedError(path, str(self))
332
block = self._dirblocks[block_index][1]
335
packed_stat = DirState.NULLSTAT
338
packed_stat = pack_stat(stat)
339
parent_info = self._empty_parent_info()
340
minikind = DirState._kind_to_minikind[kind]
342
entry_data = entry_key, [
343
(minikind, link_or_sha1, size, False, packed_stat),
345
elif kind == 'directory':
346
entry_data = entry_key, [
347
(minikind, '', 0, False, packed_stat),
349
elif kind == 'symlink':
350
entry_data = entry_key, [
351
(minikind, link_or_sha1, size, False, packed_stat),
354
raise errors.BzrError('unknown kind %r' % kind)
355
entry_index, present = self._find_entry_index(entry_key, block)
356
assert not present, "basename %r already added" % basename
357
block.insert(entry_index, entry_data)
359
if kind == 'directory':
360
# insert a new dirblock
361
self._ensure_block(block_index, entry_index, utf8path)
362
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
364
self._id_index.setdefault(entry_key[2], set()).add(entry_key)
366
def _bisect(self, dir_name_list):
367
"""Bisect through the disk structure for specific rows.
369
:param dir_name_list: A list of (dir, name) pairs.
370
:return: A dict mapping (dir, name) => entry for found entries. Missing
371
entries will not be in the map.
373
self._requires_lock()
374
# We need the file pointer to be right after the initial header block
375
self._read_header_if_needed()
376
# If _dirblock_state was in memory, we should just return info from
377
# there, this function is only meant to handle when we want to read
379
assert self._dirblock_state == DirState.NOT_IN_MEMORY
381
# The disk representation is generally info + '\0\n\0' at the end. But
382
# for bisecting, it is easier to treat this as '\0' + info + '\0\n'
383
# Because it means we can sync on the '\n'
384
state_file = self._state_file
385
file_size = os.fstat(state_file.fileno()).st_size
386
# We end up with 2 extra fields, we should have a trailing '\n' to
387
# ensure that we read the whole record, and we should have a precursur
388
# '' which ensures that we start after the previous '\n'
389
entry_field_count = self._fields_per_entry() + 1
391
low = self._end_of_header
392
high = file_size - 1 # Ignore the final '\0'
393
# Map from (dir, name) => entry
396
# Avoid infinite seeking
397
max_count = 30*len(dir_name_list)
399
# pending is a list of places to look.
400
# each entry is a tuple of low, high, dir_names
401
# low -> the first byte offset to read (inclusive)
402
# high -> the last byte offset (inclusive)
403
# dir_names -> The list of (dir, name) pairs that should be found in
404
# the [low, high] range
405
pending = [(low, high, dir_name_list)]
407
page_size = self._bisect_page_size
409
fields_to_entry = self._get_fields_to_entry()
412
low, high, cur_files = pending.pop()
414
if not cur_files or low >= high:
419
if count > max_count:
420
raise errors.BzrError('Too many seeks, most likely a bug.')
422
mid = max(low, (low+high-page_size)/2)
425
# limit the read size, so we don't end up reading data that we have
427
read_size = min(page_size, (high-mid)+1)
428
block = state_file.read(read_size)
431
entries = block.split('\n')
434
# We didn't find a '\n', so we cannot have found any records.
435
# So put this range back and try again. But we know we have to
436
# increase the page size, because a single read did not contain
437
# a record break (so records must be larger than page_size)
439
pending.append((low, high, cur_files))
442
# Check the first and last entries, in case they are partial, or if
443
# we don't care about the rest of this page
445
first_fields = entries[0].split('\0')
446
if len(first_fields) < entry_field_count:
447
# We didn't get the complete first entry
448
# so move start, and grab the next, which
449
# should be a full entry
450
start += len(entries[0])+1
451
first_fields = entries[1].split('\0')
454
if len(first_fields) <= 2:
455
# We didn't even get a filename here... what do we do?
456
# Try a large page size and repeat this query
458
pending.append((low, high, cur_files))
461
# Find what entries we are looking for, which occur before and
462
# after this first record.
464
first_dir_name = (first_fields[1], first_fields[2])
465
first_loc = bisect.bisect_left(cur_files, first_dir_name)
467
# These exist before the current location
468
pre = cur_files[:first_loc]
469
# These occur after the current location, which may be in the
470
# data we read, or might be after the last entry
471
post = cur_files[first_loc:]
473
if post and len(first_fields) >= entry_field_count:
474
# We have files after the first entry
476
# Parse the last entry
477
last_entry_num = len(entries)-1
478
last_fields = entries[last_entry_num].split('\0')
479
if len(last_fields) < entry_field_count:
480
# The very last hunk was not complete,
481
# read the previous hunk
482
after = mid + len(block) - len(entries[-1])
484
last_fields = entries[last_entry_num].split('\0')
486
after = mid + len(block)
488
last_dir_name = (last_fields[1], last_fields[2])
489
last_loc = bisect.bisect_right(post, last_dir_name)
491
middle_files = post[:last_loc]
492
post = post[last_loc:]
495
# We have files that should occur in this block
496
# (>= first, <= last)
497
# Either we will find them here, or we can mark them as
500
if middle_files[0] == first_dir_name:
501
# We might need to go before this location
502
pre.append(first_dir_name)
503
if middle_files[-1] == last_dir_name:
504
post.insert(0, last_dir_name)
506
# Find out what paths we have
507
paths = {first_dir_name:[first_fields]}
508
# last_dir_name might == first_dir_name so we need to be
509
# careful if we should append rather than overwrite
510
if last_entry_num != first_entry_num:
511
paths.setdefault(last_dir_name, []).append(last_fields)
512
for num in xrange(first_entry_num+1, last_entry_num):
513
# TODO: jam 20070223 We are already splitting here, so
514
# shouldn't we just split the whole thing rather
515
# than doing the split again in add_one_record?
516
fields = entries[num].split('\0')
517
dir_name = (fields[1], fields[2])
518
paths.setdefault(dir_name, []).append(fields)
520
for dir_name in middle_files:
521
for fields in paths.get(dir_name, []):
522
# offset by 1 because of the opening '\0'
523
# consider changing fields_to_entry to avoid the
525
entry = fields_to_entry(fields[1:])
526
found.setdefault(dir_name, []).append(entry)
528
# Now we have split up everything into pre, middle, and post, and
529
# we have handled everything that fell in 'middle'.
530
# We add 'post' first, so that we prefer to seek towards the
531
# beginning, so that we will tend to go as early as we need, and
532
# then only seek forward after that.
534
pending.append((after, high, post))
536
pending.append((low, start-1, pre))
538
# Consider that we may want to return the directory entries in sorted
539
# order. For now, we just return them in whatever order we found them,
540
# and leave it up to the caller if they care if it is ordered or not.
543
def _bisect_dirblocks(self, dir_list):
544
"""Bisect through the disk structure to find entries in given dirs.
546
_bisect_dirblocks is meant to find the contents of directories, which
547
differs from _bisect, which only finds individual entries.
549
:param dir_list: An sorted list of directory names ['', 'dir', 'foo'].
550
:return: A map from dir => entries_for_dir
552
# TODO: jam 20070223 A lot of the bisecting logic could be shared
553
# between this and _bisect. It would require parameterizing the
554
# inner loop with a function, though. We should evaluate the
555
# performance difference.
556
self._requires_lock()
557
# We need the file pointer to be right after the initial header block
558
self._read_header_if_needed()
559
# If _dirblock_state was in memory, we should just return info from
560
# there, this function is only meant to handle when we want to read
562
assert self._dirblock_state == DirState.NOT_IN_MEMORY
564
# The disk representation is generally info + '\0\n\0' at the end. But
565
# for bisecting, it is easier to treat this as '\0' + info + '\0\n'
566
# Because it means we can sync on the '\n'
567
state_file = self._state_file
568
file_size = os.fstat(state_file.fileno()).st_size
569
# We end up with 2 extra fields, we should have a trailing '\n' to
570
# ensure that we read the whole record, and we should have a precursur
571
# '' which ensures that we start after the previous '\n'
572
entry_field_count = self._fields_per_entry() + 1
574
low = self._end_of_header
575
high = file_size - 1 # Ignore the final '\0'
576
# Map from dir => entry
579
# Avoid infinite seeking
580
max_count = 30*len(dir_list)
582
# pending is a list of places to look.
583
# each entry is a tuple of low, high, dir_names
584
# low -> the first byte offset to read (inclusive)
585
# high -> the last byte offset (inclusive)
586
# dirs -> The list of directories that should be found in
587
# the [low, high] range
588
pending = [(low, high, dir_list)]
590
page_size = self._bisect_page_size
592
fields_to_entry = self._get_fields_to_entry()
595
low, high, cur_dirs = pending.pop()
597
if not cur_dirs or low >= high:
602
if count > max_count:
603
raise errors.BzrError('Too many seeks, most likely a bug.')
605
mid = max(low, (low+high-page_size)/2)
608
# limit the read size, so we don't end up reading data that we have
610
read_size = min(page_size, (high-mid)+1)
611
block = state_file.read(read_size)
614
entries = block.split('\n')
617
# We didn't find a '\n', so we cannot have found any records.
618
# So put this range back and try again. But we know we have to
619
# increase the page size, because a single read did not contain
620
# a record break (so records must be larger than page_size)
622
pending.append((low, high, cur_dirs))
625
# Check the first and last entries, in case they are partial, or if
626
# we don't care about the rest of this page
628
first_fields = entries[0].split('\0')
629
if len(first_fields) < entry_field_count:
630
# We didn't get the complete first entry
631
# so move start, and grab the next, which
632
# should be a full entry
633
start += len(entries[0])+1
634
first_fields = entries[1].split('\0')
637
if len(first_fields) <= 1:
638
# We didn't even get a dirname here... what do we do?
639
# Try a large page size and repeat this query
641
pending.append((low, high, cur_dirs))
644
# Find what entries we are looking for, which occur before and
645
# after this first record.
647
first_dir = first_fields[1]
648
first_loc = bisect.bisect_left(cur_dirs, first_dir)
650
# These exist before the current location
651
pre = cur_dirs[:first_loc]
652
# These occur after the current location, which may be in the
653
# data we read, or might be after the last entry
654
post = cur_dirs[first_loc:]
656
if post and len(first_fields) >= entry_field_count:
657
# We have records to look at after the first entry
659
# Parse the last entry
660
last_entry_num = len(entries)-1
661
last_fields = entries[last_entry_num].split('\0')
662
if len(last_fields) < entry_field_count:
663
# The very last hunk was not complete,
664
# read the previous hunk
665
after = mid + len(block) - len(entries[-1])
667
last_fields = entries[last_entry_num].split('\0')
669
after = mid + len(block)
671
last_dir = last_fields[1]
672
last_loc = bisect.bisect_right(post, last_dir)
674
middle_files = post[:last_loc]
675
post = post[last_loc:]
678
# We have files that should occur in this block
679
# (>= first, <= last)
680
# Either we will find them here, or we can mark them as
683
if middle_files[0] == first_dir:
684
# We might need to go before this location
685
pre.append(first_dir)
686
if middle_files[-1] == last_dir:
687
post.insert(0, last_dir)
689
# Find out what paths we have
690
paths = {first_dir:[first_fields]}
691
# last_dir might == first_dir so we need to be
692
# careful if we should append rather than overwrite
693
if last_entry_num != first_entry_num:
694
paths.setdefault(last_dir, []).append(last_fields)
695
for num in xrange(first_entry_num+1, last_entry_num):
696
# TODO: jam 20070223 We are already splitting here, so
697
# shouldn't we just split the whole thing rather
698
# than doing the split again in add_one_record?
699
fields = entries[num].split('\0')
700
paths.setdefault(fields[1], []).append(fields)
702
for cur_dir in middle_files:
703
for fields in paths.get(cur_dir, []):
704
# offset by 1 because of the opening '\0'
705
# consider changing fields_to_entry to avoid the
707
entry = fields_to_entry(fields[1:])
708
found.setdefault(cur_dir, []).append(entry)
710
# Now we have split up everything into pre, middle, and post, and
711
# we have handled everything that fell in 'middle'.
712
# We add 'post' first, so that we prefer to seek towards the
713
# beginning, so that we will tend to go as early as we need, and
714
# then only seek forward after that.
716
pending.append((after, high, post))
718
pending.append((low, start-1, pre))
722
def _bisect_recursive(self, dir_name_list):
723
"""Bisect for entries for all paths and their children.
725
This will use bisect to find all records for the supplied paths. It
726
will then continue to bisect for any records which are marked as
727
directories. (and renames?)
729
:param paths: A sorted list of (dir, name) pairs
730
eg: [('', 'a'), ('', 'f'), ('a/b', 'c')]
731
:return: A dictionary mapping (dir, name, file_id) => [tree_info]
733
# Map from (dir, name, file_id) => [tree_info]
736
found_dir_names = set()
738
# Directories that have been read
739
processed_dirs = set()
740
# Get the ball rolling with the first bisect for all entries.
741
newly_found = self._bisect(dir_name_list)
744
# Directories that need to be read
746
paths_to_search = set()
747
for entry_list in newly_found.itervalues():
748
for dir_name_id, trees_info in entry_list:
749
found[dir_name_id] = trees_info
750
found_dir_names.add(dir_name_id[:2])
752
for tree_info in trees_info:
753
minikind = tree_info[0]
756
# We already processed this one as a directory,
757
# we don't need to do the extra work again.
759
subdir, name, file_id = dir_name_id
760
path = osutils.pathjoin(subdir, name)
762
if path not in processed_dirs:
763
pending_dirs.add(path)
764
elif minikind == 'r':
765
# Rename, we need to directly search the target
766
# which is contained in the fingerprint column
767
dir_name = osutils.split(tree_info[1])
768
if dir_name[0] in pending_dirs:
769
# This entry will be found in the dir search
771
# TODO: We need to check if this entry has
772
# already been found. Otherwise we might be
773
# hitting infinite recursion.
774
if dir_name not in found_dir_names:
775
paths_to_search.add(dir_name)
776
# Now we have a list of paths to look for directly, and
777
# directory blocks that need to be read.
778
# newly_found is mixing the keys between (dir, name) and path
779
# entries, but that is okay, because we only really care about the
781
newly_found = self._bisect(sorted(paths_to_search))
782
newly_found.update(self._bisect_dirblocks(sorted(pending_dirs)))
783
processed_dirs.update(pending_dirs)
786
def _empty_parent_info(self):
787
return [DirState.NULL_PARENT_DETAILS] * (len(self._parents) -
790
def _ensure_block(self, parent_block_index, parent_row_index, dirname):
791
"""Ensure a block for dirname exists.
793
This function exists to let callers which know that there is a
794
directory dirname ensure that the block for it exists. This block can
795
fail to exist because of demand loading, or because a directory had no
796
children. In either case it is not an error. It is however an error to
797
call this if there is no parent entry for the directory, and thus the
798
function requires the coordinates of such an entry to be provided.
800
The root row is special cased and can be indicated with a parent block
803
:param parent_block_index: The index of the block in which dirname's row
805
:param parent_row_index: The index in the parent block where the row
807
:param dirname: The utf8 dirname to ensure there is a block for.
808
:return: The index for the block.
810
if dirname == '' and parent_row_index == 0 and parent_block_index == 0:
811
# This is the signature of the root row, and the
812
# contents-of-root row is always index 1
814
# the basename of the directory must be the end of its full name.
815
if not (parent_block_index == -1 and
816
parent_block_index == -1 and dirname == ''):
817
assert dirname.endswith(
818
self._dirblocks[parent_block_index][1][parent_row_index][0][1])
819
block_index, present = self._find_block_index_from_key((dirname, '', ''))
821
## In future, when doing partial parsing, this should load and
822
# populate the entire block.
823
self._dirblocks.insert(block_index, (dirname, []))
826
def _entries_to_current_state(self, new_entries):
827
"""Load new_entries into self.dirblocks.
829
Process new_entries into the current state object, making them the active
832
:param new_entries: A sorted list of entries. This function does not sort
833
to prevent unneeded overhead when callers have a sorted list already.
836
assert new_entries[0][0][0:2] == ('', ''), \
837
"Missing root row %r" % new_entries[0][0]
838
# The two blocks here are deliberate: the root block and the
839
# contents-of-root block.
840
self._dirblocks = [('', []), ('', [])]
841
current_block = self._dirblocks[0][1]
844
append_entry = current_block.append
845
for entry in new_entries:
846
if entry[0][0] != current_dirname:
847
# new block - different dirname
849
current_dirname = entry[0][0]
850
self._dirblocks.append((current_dirname, current_block))
851
append_entry = current_block.append
852
# append the entry to the current block
854
self._split_root_dirblock_into_contents()
856
def _split_root_dirblock_into_contents(self):
857
"""Split the root dirblocks into root and contents-of-root.
859
After parsing by path, we end up with root entries and contents-of-root
860
entries in the same block. This loop splits them out again.
862
# The above loop leaves the "root block" entries mixed with the
863
# "contents-of-root block". But we don't want an if check on
864
# all entries, so instead we just fix it up here.
865
assert self._dirblocks[1] == ('', [])
867
contents_of_root_block = []
868
for entry in self._dirblocks[0][1]:
869
if not entry[0][1]: # This is a root entry
870
root_block.append(entry)
872
contents_of_root_block.append(entry)
873
self._dirblocks[0] = ('', root_block)
874
self._dirblocks[1] = ('', contents_of_root_block)
876
def _entry_to_line(self, entry):
877
"""Serialize entry to a NULL delimited line ready for _get_output_lines.
879
:param entry: An entry_tuple as defined in the module docstring.
881
entire_entry = list(entry[0])
882
for tree_number, tree_data in enumerate(entry[1]):
883
# (minikind, fingerprint, size, executable, tree_specific_string)
884
entire_entry.extend(tree_data)
885
# 3 for the key, 5 for the fields per tree.
886
tree_offset = 3 + tree_number * 5
888
entire_entry[tree_offset + 0] = tree_data[0]
890
entire_entry[tree_offset + 2] = str(tree_data[2])
892
entire_entry[tree_offset + 3] = DirState._to_yesno[tree_data[3]]
893
return '\0'.join(entire_entry)
895
def _fields_per_entry(self):
896
"""How many null separated fields should be in each entry row.
898
Each line now has an extra '\n' field which is not used
899
so we just skip over it
902
+ number of fields per tree_data (5) * tree count
905
tree_count = 1 + self._num_present_parents()
906
return 3 + 5 * tree_count + 1
908
def _find_block(self, key, add_if_missing=False):
909
"""Return the block that key should be present in.
911
:param key: A dirstate entry key.
912
:return: The block tuple.
914
block_index, present = self._find_block_index_from_key(key)
916
if not add_if_missing:
917
# check to see if key is versioned itself - we might want to
918
# add it anyway, because dirs with no entries dont get a
919
# dirblock at parse time.
920
# This is an uncommon branch to take: most dirs have children,
921
# and most code works with versioned paths.
922
parent_base, parent_name = osutils.split(key[0])
923
if not self._get_block_entry_index(parent_base, parent_name, 0)[3]:
924
# some parent path has not been added - its an error to add
926
raise errors.NotVersionedError(key[0:2], str(self))
927
self._dirblocks.insert(block_index, (key[0], []))
928
return self._dirblocks[block_index]
930
def _find_block_index_from_key(self, key):
931
"""Find the dirblock index for a key.
933
:return: The block index, True if the block for the key is present.
935
if key[0:2] == ('', ''):
937
block_index = bisect.bisect_left(self._dirblocks, (key[0], []), 1)
938
# _right returns one-past-where-key is so we have to subtract
939
# one to use it. we use _right here because there are two
940
# '' blocks - the root, and the contents of root
941
# we always have a minimum of 2 in self._dirblocks: root and
942
# root-contents, and for '', we get 2 back, so this is
943
# simple and correct:
944
present = (block_index < len(self._dirblocks) and
945
self._dirblocks[block_index][0] == key[0])
946
return block_index, present
948
def _find_entry_index(self, key, block):
949
"""Find the entry index for a key in a block.
951
:return: The entry index, True if the entry for the key is present.
953
entry_index = bisect.bisect_left(block, (key, []))
954
present = (entry_index < len(block) and
955
block[entry_index][0] == key)
956
return entry_index, present
959
def from_tree(tree, dir_state_filename):
960
"""Create a dirstate from a bzr Tree.
962
:param tree: The tree which should provide parent information and
964
:return: a DirState object which is currently locked for writing.
965
(it was locked by DirState.initialize)
967
result = DirState.initialize(dir_state_filename)
971
parent_ids = tree.get_parent_ids()
972
num_parents = len(parent_ids)
974
for parent_id in parent_ids:
975
parent_tree = tree.branch.repository.revision_tree(parent_id)
976
parent_trees.append((parent_id, parent_tree))
977
parent_tree.lock_read()
978
result.set_parent_trees(parent_trees, [])
979
result.set_state_from_inventory(tree.inventory)
981
for revid, parent_tree in parent_trees:
985
# The caller won't have a chance to unlock this, so make sure we
991
def get_ghosts(self):
992
"""Return a list of the parent tree revision ids that are ghosts."""
993
self._read_header_if_needed()
997
"""Serialise the entire dirstate to a sequence of lines."""
998
if (self._header_state == DirState.IN_MEMORY_UNMODIFIED and
999
self._dirblock_state == DirState.IN_MEMORY_UNMODIFIED):
1000
# read whats on disk.
1001
self._state_file.seek(0)
1002
return self._state_file.readlines()
1004
lines.append(self._get_parents_line(self.get_parent_ids()))
1005
lines.append(self._get_ghosts_line(self._ghosts))
1006
# append the root line which is special cased
1007
lines.extend(map(self._entry_to_line, self._iter_entries()))
1008
return self._get_output_lines(lines)
1010
def _get_ghosts_line(self, ghost_ids):
1011
"""Create a line for the state file for ghost information."""
1012
return '\0'.join([str(len(ghost_ids))] + ghost_ids)
1014
def _get_parents_line(self, parent_ids):
1015
"""Create a line for the state file for parents information."""
1016
return '\0'.join([str(len(parent_ids))] + parent_ids)
1018
def _get_fields_to_entry(self):
1019
"""Get a function which converts entry fields into a entry record.
1021
This handles size and executable, as well as parent records.
1023
:return: A function which takes a list of fields, and returns an
1024
appropriate record for storing in memory.
1026
# This is intentionally unrolled for performance
1027
num_present_parents = self._num_present_parents()
1028
if num_present_parents == 0:
1029
def fields_to_entry_0_parents(fields, _int=int):
1030
path_name_file_id_key = (fields[0], fields[1], fields[2])
1031
return (path_name_file_id_key, [
1033
fields[3], # minikind
1034
fields[4], # fingerprint
1035
_int(fields[5]), # size
1036
fields[6] == 'y', # executable
1037
fields[7], # packed_stat or revision_id
1039
return fields_to_entry_0_parents
1040
elif num_present_parents == 1:
1041
def fields_to_entry_1_parent(fields, _int=int):
1042
path_name_file_id_key = (fields[0], fields[1], fields[2])
1043
return (path_name_file_id_key, [
1045
fields[3], # minikind
1046
fields[4], # fingerprint
1047
_int(fields[5]), # size
1048
fields[6] == 'y', # executable
1049
fields[7], # packed_stat or revision_id
1052
fields[8], # minikind
1053
fields[9], # fingerprint
1054
_int(fields[10]), # size
1055
fields[11] == 'y', # executable
1056
fields[12], # packed_stat or revision_id
1059
return fields_to_entry_1_parent
1060
elif num_present_parents == 2:
1061
def fields_to_entry_2_parents(fields, _int=int):
1062
path_name_file_id_key = (fields[0], fields[1], fields[2])
1063
return (path_name_file_id_key, [
1065
fields[3], # minikind
1066
fields[4], # fingerprint
1067
_int(fields[5]), # size
1068
fields[6] == 'y', # executable
1069
fields[7], # packed_stat or revision_id
1072
fields[8], # minikind
1073
fields[9], # fingerprint
1074
_int(fields[10]), # size
1075
fields[11] == 'y', # executable
1076
fields[12], # packed_stat or revision_id
1079
fields[13], # minikind
1080
fields[14], # fingerprint
1081
_int(fields[15]), # size
1082
fields[16] == 'y', # executable
1083
fields[17], # packed_stat or revision_id
1086
return fields_to_entry_2_parents
1088
def fields_to_entry_n_parents(fields, _int=int):
1089
path_name_file_id_key = (fields[0], fields[1], fields[2])
1090
trees = [(fields[cur], # minikind
1091
fields[cur+1], # fingerprint
1092
_int(fields[cur+2]), # size
1093
fields[cur+3] == 'y', # executable
1094
fields[cur+4], # stat or revision_id
1095
) for cur in xrange(3, len(fields)-1, 5)]
1096
return path_name_file_id_key, trees
1097
return fields_to_entry_n_parents
1099
def get_parent_ids(self):
1100
"""Return a list of the parent tree ids for the directory state."""
1101
self._read_header_if_needed()
1102
return list(self._parents)
1104
def _get_block_entry_index(self, dirname, basename, tree_index):
1105
"""Get the coordinates for a path in the state structure.
1107
:param dirname: The utf8 dirname to lookup.
1108
:param basename: The utf8 basename to lookup.
1109
:param tree_index: The index of the tree for which this lookup should
1111
:return: A tuple describing where the path is located, or should be
1112
inserted. The tuple contains four fields: the block index, the row
1113
index, anda two booleans are True when the directory is present, and
1114
when the entire path is present. There is no guarantee that either
1115
coordinate is currently reachable unless the found field for it is
1116
True. For instance, a directory not present in the searched tree
1117
may be returned with a value one greater than the current highest
1118
block offset. The directory present field will always be True when
1119
the path present field is True. The directory present field does
1120
NOT indicate that the directory is present in the searched tree,
1121
rather it indicates that there are at least some files in some
1124
self._read_dirblocks_if_needed()
1125
key = dirname, basename, ''
1126
block_index, present = self._find_block_index_from_key(key)
1128
# no such directory - return the dir index and 0 for the row.
1129
return block_index, 0, False, False
1130
block = self._dirblocks[block_index][1] # access the entries only
1131
entry_index, present = self._find_entry_index(key, block)
1132
# linear search through present entries at this path to find the one
1134
while entry_index < len(block) and block[entry_index][0][1] == basename:
1135
if block[entry_index][1][tree_index][0] not in \
1136
('a', 'r'): # absent, relocated
1137
return block_index, entry_index, True, True
1139
return block_index, entry_index, True, False
1141
def _get_entry(self, tree_index, fileid_utf8=None, path_utf8=None):
1142
"""Get the dirstate entry for path in tree tree_index
1144
If either file_id or path is supplied, it is used as the key to lookup.
1145
If both are supplied, the fastest lookup is used, and an error is
1146
raised if they do not both point at the same row.
1148
:param tree_index: The index of the tree we wish to locate this path
1149
in. If the path is present in that tree, the entry containing its
1150
details is returned, otherwise (None, None) is returned
1151
:param fileid_utf8: A utf8 file_id to look up.
1152
:param path_utf8: An utf8 path to be looked up.
1153
:return: The dirstate entry tuple for path, or (None, None)
1155
self._read_dirblocks_if_needed()
1156
if path_utf8 is not None:
1157
assert path_utf8.__class__ == str, 'path_utf8 is not a str: %s %s' % (type(path_utf8), path_utf8)
1158
# path lookups are faster
1159
dirname, basename = osutils.split(path_utf8)
1160
block_index, entry_index, dir_present, file_present = \
1161
self._get_block_entry_index(dirname, basename, tree_index)
1162
if not file_present:
1164
entry = self._dirblocks[block_index][1][entry_index]
1165
assert entry[0][2] and entry[1][tree_index][0] not in ('a', 'r'), 'unversioned entry?!?!'
1167
if entry[0][2] != fileid_utf8:
1168
raise errors.BzrError('integrity error ? : mismatching'
1169
' tree_index, file_id and path')
1172
possible_keys = self._get_id_index().get(fileid_utf8, None)
1173
if not possible_keys:
1175
for key in possible_keys:
1176
(block_index, entry_index, dir_present,
1177
file_present) = self._get_block_entry_index(key[0], key[1],
1180
entry = self._dirblocks[block_index][1][entry_index]
1181
assert entry[1][tree_index][0] not in ('a', 'r')
1182
assert key == entry[0]
1187
def initialize(path):
1188
"""Create a new dirstate on path.
1190
The new dirstate will be an empty tree - that is it has no parents,
1191
and only a root node - which has id ROOT_ID.
1193
The object will be write locked when returned to the caller,
1194
unless there was an exception in the writing, in which case it
1197
:param path: The name of the file for the dirstate.
1198
:return: A DirState object.
1200
# This constructs a new DirState object on a path, sets the _state_file
1201
# to a new empty file for that path. It then calls _set_data() with our
1202
# stock empty dirstate information - a root with ROOT_ID, no children,
1203
# and no parents. Finally it calls save() to ensure that this data will
1205
result = DirState(path)
1206
# root dir and root dir contents with no children.
1207
empty_tree_dirblocks = [('', []), ('', [])]
1208
# a new root directory, with a NULLSTAT.
1209
empty_tree_dirblocks[0][1].append(
1210
(('', '', bzrlib.inventory.ROOT_ID), [
1211
('d', '', 0, False, DirState.NULLSTAT),
1215
result._set_data([], empty_tree_dirblocks)
1222
def _inv_entry_to_details(self, inv_entry):
1223
"""Convert an inventory entry (from a revision tree) to state details.
1225
:param inv_entry: An inventory entry whose sha1 and link targets can be
1226
relied upon, and which has a revision set.
1227
:return: A details tuple - the details for a single tree at a path +
1230
kind = inv_entry.kind
1231
minikind = DirState._kind_to_minikind[kind]
1232
tree_data = inv_entry.revision
1233
assert len(tree_data) > 0, 'empty revision for the inv_entry.'
1234
if kind == 'directory':
1238
elif kind == 'symlink':
1239
fingerprint = inv_entry.symlink_target or ''
1242
elif kind == 'file':
1243
fingerprint = inv_entry.text_sha1 or ''
1244
size = inv_entry.text_size or 0
1245
executable = inv_entry.executable
1248
return (minikind, fingerprint, size, executable, tree_data)
1250
def _iter_entries(self):
1251
"""Iterate over all the entries in the dirstate.
1253
Each yelt item is an entry in the standard format described in the
1254
docstring of bzrlib.dirstate.
1256
self._read_dirblocks_if_needed()
1257
for directory in self._dirblocks:
1258
for entry in directory[1]:
1261
def _get_id_index(self):
1262
"""Get an id index of self._dirblocks."""
1263
if self._id_index is None:
1265
for key, tree_details in self._iter_entries():
1266
id_index.setdefault(key[2], set()).add(key)
1267
self._id_index = id_index
1268
return self._id_index
1270
def _get_output_lines(self, lines):
1271
"""format lines for final output.
1273
:param lines: A sequece of lines containing the parents list and the
1276
output_lines = ['#bazaar dirstate flat format 2\n']
1277
lines.append('') # a final newline
1278
inventory_text = '\0\n\0'.join(lines)
1279
output_lines.append('adler32: %s\n' % (zlib.adler32(inventory_text),))
1280
# -3, 1 for num parents, 1 for ghosts, 1 for final newline
1281
num_entries = len(lines)-3
1282
output_lines.append('num_entries: %s\n' % (num_entries,))
1283
output_lines.append(inventory_text)
1286
def _make_deleted_row(self, fileid_utf8, parents):
1287
"""Return a deleted for for fileid_utf8."""
1288
return ('/', 'RECYCLED.BIN', 'file', fileid_utf8, 0, DirState.NULLSTAT,
1291
def _num_present_parents(self):
1292
"""The number of parent entries in each record row."""
1293
return len(self._parents) - len(self._ghosts)
1297
"""Construct a DirState on the file at path path.
1299
:return: An unlocked DirState object, associated with the given path.
1301
result = DirState(path)
1304
def _read_dirblocks_if_needed(self):
1305
"""Read in all the dirblocks from the file if they are not in memory.
1307
This populates self._dirblocks, and sets self._dirblock_state to
1308
IN_MEMORY_UNMODIFIED. It is not currently ready for incremental block
1311
self._read_header_if_needed()
1312
if self._dirblock_state == DirState.NOT_IN_MEMORY:
1313
# move the _state_file pointer to after the header (in case bisect
1314
# has been called in the mean time)
1315
self._state_file.seek(self._end_of_header)
1316
text = self._state_file.read()
1317
# TODO: check the adler checksums. adler_measured = zlib.adler32(text)
1319
fields = text.split('\0')
1320
# Remove the last blank entry
1321
trailing = fields.pop()
1322
assert trailing == ''
1323
# consider turning fields into a tuple.
1325
# skip the first field which is the trailing null from the header.
1327
# Each line now has an extra '\n' field which is not used
1328
# so we just skip over it
1330
# 3 fields for the key
1331
# + number of fields per tree_data (5) * tree count
1333
num_present_parents = self._num_present_parents()
1334
tree_count = 1 + num_present_parents
1335
entry_size = self._fields_per_entry()
1336
expected_field_count = entry_size * self._num_entries
1337
if len(fields) - cur > expected_field_count:
1338
fields = fields[:expected_field_count + cur]
1339
trace.mutter('Unexpectedly long dirstate field count!')
1340
print "XXX: incorrectly truncated dirstate file bug triggered."
1341
field_count = len(fields)
1342
# this checks our adjustment, and also catches file too short.
1343
assert field_count - cur == expected_field_count, \
1344
'field count incorrect %s != %s, entry_size=%s, '\
1345
'num_entries=%s fields=%r' % (
1346
field_count - cur, expected_field_count, entry_size,
1347
self._num_entries, fields)
1349
if num_present_parents == 1:
1350
# Bind external functions to local names
1352
# We access all fields in order, so we can just iterate over
1353
# them. Grab an straight iterator over the fields. (We use an
1354
# iterator because we don't want to do a lot of additions, nor
1355
# do we want to do a lot of slicing)
1356
next = iter(fields).next
1357
# Move the iterator to the current position
1358
for x in xrange(cur):
1360
# The two blocks here are deliberate: the root block and the
1361
# contents-of-root block.
1362
self._dirblocks = [('', []), ('', [])]
1363
current_block = self._dirblocks[0][1]
1364
current_dirname = ''
1365
append_entry = current_block.append
1366
for count in xrange(self._num_entries):
1370
if dirname != current_dirname:
1371
# new block - different dirname
1373
current_dirname = dirname
1374
self._dirblocks.append((current_dirname, current_block))
1375
append_entry = current_block.append
1376
# we know current_dirname == dirname, so re-use it to avoid
1377
# creating new strings
1378
entry = ((current_dirname, name, file_id),
1381
next(), # fingerprint
1382
_int(next()), # size
1383
next() == 'y', # executable
1384
next(), # packed_stat or revision_id
1388
next(), # fingerprint
1389
_int(next()), # size
1390
next() == 'y', # executable
1391
next(), # packed_stat or revision_id
1395
assert trailing == '\n'
1396
# append the entry to the current block
1398
self._split_root_dirblock_into_contents()
1400
fields_to_entry = self._get_fields_to_entry()
1401
entries = [fields_to_entry(fields[pos:pos+entry_size])
1402
for pos in xrange(cur, field_count, entry_size)]
1403
self._entries_to_current_state(entries)
1404
self._dirblock_state = DirState.IN_MEMORY_UNMODIFIED
1406
def _read_header(self):
1407
"""This reads in the metadata header, and the parent ids.
1409
After reading in, the file should be positioned at the null
1410
just before the start of the first record in the file.
1412
:return: (expected adler checksum, number of entries, parent list)
1414
self._read_prelude()
1415
parent_line = self._state_file.readline()
1416
info = parent_line.split('\0')
1417
num_parents = int(info[0])
1418
assert num_parents == len(info)-2, 'incorrect parent info line'
1419
self._parents = info[1:-1]
1421
ghost_line = self._state_file.readline()
1422
info = ghost_line.split('\0')
1423
num_ghosts = int(info[1])
1424
assert num_ghosts == len(info)-3, 'incorrect ghost info line'
1425
self._ghosts = info[2:-1]
1426
self._header_state = DirState.IN_MEMORY_UNMODIFIED
1427
self._end_of_header = self._state_file.tell()
1429
def _read_header_if_needed(self):
1430
"""Read the header of the dirstate file if needed."""
1431
# inline this as it will be called a lot
1432
if not self._lock_token:
1433
raise errors.ObjectNotLocked(self)
1434
if self._header_state == DirState.NOT_IN_MEMORY:
1437
def _read_prelude(self):
1438
"""Read in the prelude header of the dirstate file
1440
This only reads in the stuff that is not connected to the adler
1441
checksum. The position will be correct to read in the rest of
1442
the file and check the checksum after this point.
1443
The next entry in the file should be the number of parents,
1444
and their ids. Followed by a newline.
1446
header = self._state_file.readline()
1447
assert header == '#bazaar dirstate flat format 2\n', \
1448
'invalid header line: %r' % (header,)
1449
adler_line = self._state_file.readline()
1450
assert adler_line.startswith('adler32: '), 'missing adler32 checksum'
1451
self.adler_expected = int(adler_line[len('adler32: '):-1])
1452
num_entries_line = self._state_file.readline()
1453
assert num_entries_line.startswith('num_entries: '), 'missing num_entries line'
1454
self._num_entries = int(num_entries_line[len('num_entries: '):-1])
1457
"""Save any pending changes created during this session.
1459
We reuse the existing file, because that prevents race conditions with
1460
file creation, and we expect to be using oslocks on it in the near
1461
future to prevent concurrent modification and reads - because dirstates
1462
incremental data aggretation is not compatible with reading a modified
1463
file, and replacing a file in use by another process is impossible on
1466
A dirstate in read only mode should be smart enough though to validate
1467
that the file has not changed, and otherwise discard its cache and
1468
start over, to allow for fine grained read lock duration, so 'status'
1469
wont block 'commit' - for example.
1471
if (self._header_state == DirState.IN_MEMORY_MODIFIED or
1472
self._dirblock_state == DirState.IN_MEMORY_MODIFIED):
1473
self._state_file.seek(0)
1474
self._state_file.writelines(self.get_lines())
1475
self._state_file.truncate()
1476
self._state_file.flush()
1477
self._header_state = DirState.IN_MEMORY_UNMODIFIED
1478
self._dirblock_state = DirState.IN_MEMORY_UNMODIFIED
1480
def _set_data(self, parent_ids, dirblocks):
1481
"""Set the full dirstate data in memory.
1483
This is an internal function used to completely replace the objects
1484
in memory state. It puts the dirstate into state 'full-dirty'.
1486
:param parent_ids: A list of parent tree revision ids.
1487
:param dirblocks: A list containing one tuple for each directory in the
1488
tree. Each tuple contains the directory path and a list of entries
1489
found in that directory.
1491
# our memory copy is now authoritative.
1492
self._dirblocks = dirblocks
1493
self._header_state = DirState.IN_MEMORY_MODIFIED
1494
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
1495
self._parents = list(parent_ids)
1496
self._id_index = None
1498
def set_path_id(self, path, new_id):
1499
"""Change the id of path to new_id in the current working tree.
1501
:param path: The path inside the tree to set - '' is the root, 'foo'
1502
is the path foo in the root.
1503
:param new_id: The new id to assign to the path. This must be a utf8
1504
file id (not unicode, and not None).
1506
# TODO: start warning here.
1507
assert new_id.__class__ == str
1508
self._read_dirblocks_if_needed()
1510
import pdb;pdb.set_trace()
1512
raise NotImplementedError(self.set_path_id)
1513
# TODO: check new id is unique
1514
entry = self._get_entry(0, path_utf8=path)
1515
# mark the old path absent, and insert a new root path
1516
self._make_absent(entry)
1517
self.update_minimal(('', '', new_id), 'd',
1518
path_utf8='', packed_stat=entry[1][0][4])
1519
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
1520
if self._id_index is not None:
1521
self._id_index.setdefault(new_id, set()).add(entry[0])
1523
def set_parent_trees(self, trees, ghosts):
1524
"""Set the parent trees for the dirstate.
1526
:param trees: A list of revision_id, tree tuples. tree must be provided
1527
even if the revision_id refers to a ghost: supply an empty tree in
1529
:param ghosts: A list of the revision_ids that are ghosts at the time
1532
# TODO: generate a list of parent indexes to preserve to save
1533
# processing specific parent trees. In the common case one tree will
1534
# be preserved - the left most parent.
1535
# TODO: if the parent tree is a dirstate, we might want to walk them
1536
# all by path in parallel for 'optimal' common-case performance.
1537
# generate new root row.
1538
self._read_dirblocks_if_needed()
1539
# TODO future sketch: Examine the existing parents to generate a change
1540
# map and then walk the new parent trees only, mapping them into the
1541
# dirstate. Walk the dirstate at the same time to remove unreferenced
1544
# sketch: loop over all entries in the dirstate, cherry picking
1545
# entries from the parent trees, if they are not ghost trees.
1546
# after we finish walking the dirstate, all entries not in the dirstate
1547
# are deletes, so we want to append them to the end as per the design
1548
# discussions. So do a set difference on ids with the parents to
1549
# get deletes, and add them to the end.
1550
# During the update process we need to answer the following questions:
1551
# - find other keys containing a fileid in order to create cross-path
1552
# links. We dont't trivially use the inventory from other trees
1553
# because this leads to either double touching, or to accessing
1555
# - find other keys containing a path
1556
# We accumulate each entry via this dictionary, including the root
1559
# we could do parallel iterators, but because file id data may be
1560
# scattered throughout, we dont save on index overhead: we have to look
1561
# at everything anyway. We can probably save cycles by reusing parent
1562
# data and doing an incremental update when adding an additional
1563
# parent, but for now the common cases are adding a new parent (merge),
1564
# and replacing completely (commit), and commit is more common: so
1565
# optimise merge later.
1567
# ---- start generation of full tree mapping data
1568
# what trees should we use?
1569
parent_trees = [tree for rev_id, tree in trees if rev_id not in ghosts]
1570
# how many trees do we end up with
1571
parent_count = len(parent_trees)
1573
# one: the current tree
1574
for entry in self._iter_entries():
1575
# skip entries not in the current tree
1576
if entry[1][0][0] in ('a', 'r'): # absent, relocated
1578
by_path[entry[0]] = [entry[1][0]] + \
1579
[DirState.NULL_PARENT_DETAILS] * parent_count
1580
id_index[entry[0][2]] = set([entry[0]])
1582
# now the parent trees:
1583
for tree_index, tree in enumerate(parent_trees):
1584
# the index is off by one, adjust it.
1585
tree_index = tree_index + 1
1586
# when we add new locations for a fileid we need these ranges for
1587
# any fileid in this tree as we set the by_path[id] to:
1588
# already_processed_tree_details + new_details + new_location_suffix
1589
# the suffix is from tree_index+1:parent_count+1.
1590
new_location_suffix = [DirState.NULL_PARENT_DETAILS] * (parent_count - tree_index)
1591
# now stitch in all the entries from this tree
1592
for path, entry in tree.inventory.iter_entries_by_dir():
1593
# here we process each trees details for each item in the tree.
1594
# we first update any existing entries for the id at other paths,
1595
# then we either create or update the entry for the id at the
1596
# right path, and finally we add (if needed) a mapping from
1597
# file_id to this path. We do it in this order to allow us to
1598
# avoid checking all known paths for the id when generating a
1599
# new entry at this path: by adding the id->path mapping last,
1600
# all the mappings are valid and have correct relocation
1601
# records where needed.
1602
file_id = entry.file_id
1603
path_utf8 = path.encode('utf8')
1604
dirname, basename = osutils.split(path_utf8)
1605
new_entry_key = (dirname, basename, file_id)
1606
# tree index consistency: All other paths for this id in this tree
1607
# index must point to the correct path.
1608
for entry_key in id_index.setdefault(file_id, set()):
1609
# TODO:PROFILING: It might be faster to just update
1610
# rather than checking if we need to, and then overwrite
1611
# the one we are located at.
1612
if entry_key != new_entry_key:
1613
# this file id is at a different path in one of the
1614
# other trees, so put absent pointers there
1615
# This is the vertical axis in the matrix, all pointing
1617
by_path[entry_key][tree_index] = ('r', path_utf8, 0, False, '')
1618
# by path consistency: Insert into an existing path record (trivial), or
1619
# add a new one with relocation pointers for the other tree indexes.
1620
if new_entry_key in id_index[file_id]:
1621
# there is already an entry where this data belongs, just insert it.
1622
by_path[new_entry_key][tree_index] = \
1623
self._inv_entry_to_details(entry)
1625
# add relocated entries to the horizontal axis - this row
1626
# mapping from path,id. We need to look up the correct path
1627
# for the indexes from 0 to tree_index -1
1629
for lookup_index in xrange(tree_index):
1630
# boundary case: this is the first occurence of file_id
1631
# so there are no id_indexs, possibly take this out of
1633
if not len(id_index[file_id]):
1634
new_details.append(DirState.NULL_PARENT_DETAILS)
1636
# grab any one entry, use it to find the right path.
1637
# TODO: optimise this to reduce memory use in highly
1638
# fragmented situations by reusing the relocation
1640
a_key = iter(id_index[file_id]).next()
1641
if by_path[a_key][lookup_index][0] in ('r', 'a'):
1642
# its a pointer or missing statement, use it as is.
1643
new_details.append(by_path[a_key][lookup_index])
1645
# we have the right key, make a pointer to it.
1646
real_path = ('/'.join(a_key[0:2])).strip('/')
1647
new_details.append(('r', real_path, 0, False, ''))
1648
new_details.append(self._inv_entry_to_details(entry))
1649
new_details.extend(new_location_suffix)
1650
by_path[new_entry_key] = new_details
1651
id_index[file_id].add(new_entry_key)
1652
# --- end generation of full tree mappings
1654
# sort and output all the entries
1655
new_entries = sorted(by_path.items())
1656
self._entries_to_current_state(new_entries)
1657
self._parents = [rev_id for rev_id, tree in trees]
1658
self._ghosts = list(ghosts)
1659
self._header_state = DirState.IN_MEMORY_MODIFIED
1660
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
1661
self._id_index = id_index
1663
def set_state_from_inventory(self, new_inv):
1664
"""Set new_inv as the current state.
1666
This API is called by tree transform, and will usually occur with
1667
existing parent trees.
1669
:param new_inv: The inventory object to set current state from.
1671
self._read_dirblocks_if_needed()
1673
# incremental algorithm:
1674
# two iterators: current data and new data, both in dirblock order.
1675
new_iterator = new_inv.iter_entries_by_dir()
1676
# we will be modifying the dirstate, so we need a stable iterator. In
1677
# future we might write one, for now we just clone the state into a
1678
# list - which is a shallow copy, so each
1679
old_iterator = iter(list(self._iter_entries()))
1680
# both must have roots so this is safe:
1681
current_new = new_iterator.next()
1682
current_old = old_iterator.next()
1683
def advance(iterator):
1685
return iterator.next()
1686
except StopIteration:
1688
while current_new or current_old:
1689
# skip entries in old that are not really there
1690
if current_old and current_old[1][0][0] in ('r', 'a'):
1691
# relocated or absent
1692
current_old = advance(old_iterator)
1695
# convert new into dirblock style
1696
new_path_utf8 = current_new[0].encode('utf8')
1697
new_dirname, new_basename = osutils.split(new_path_utf8)
1698
new_id = current_new[1].file_id
1699
new_entry_key = (new_dirname, new_basename, new_id)
1700
current_new_minikind = \
1701
DirState._kind_to_minikind[current_new[1].kind]
1703
# for safety disable variables
1704
new_path_utf8 = new_dirname = new_basename = new_id = new_entry_key = None
1705
# 5 cases, we dont have a value that is strictly greater than everything, so
1706
# we make both end conditions explicit
1708
# old is finished: insert current_new into the state.
1709
self.update_minimal(new_entry_key, current_new_minikind,
1710
executable=current_new[1].executable,
1711
path_utf8=new_path_utf8)
1712
current_new = advance(new_iterator)
1713
elif not current_new:
1715
self._make_absent(current_old)
1716
current_old = advance(old_iterator)
1717
elif new_entry_key == current_old[0]:
1718
# same - common case
1719
# TODO: update the record if anything significant has changed.
1720
# the minimal required trigger is if the execute bit or cached
1722
if (current_old[1][0][3] != current_new[1].executable or
1723
current_old[1][0][0] != current_new_minikind):
1724
self.update_minimal(current_old[0], current_new_minikind,
1725
executable=current_new[1].executable,
1726
path_utf8=new_path_utf8)
1727
# both sides are dealt with, move on
1728
current_old = advance(old_iterator)
1729
current_new = advance(new_iterator)
1730
elif new_entry_key < current_old[0]:
1732
# add a entry for this and advance new
1733
self.update_minimal(new_entry_key, current_new_minikind,
1734
executable=current_new[1].executable,
1735
path_utf8=new_path_utf8)
1736
current_new = advance(new_iterator)
1739
self._make_absent(current_old)
1740
current_old = advance(old_iterator)
1741
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
1742
self._id_index = None
1744
def _make_absent(self, current_old):
1745
"""Mark current_old - an entry - as absent for tree 0.
1747
:return: True if this was the last details entry for they entry key:
1748
that is, if the underlying block has had the entry removed, thus
1749
shrinking in length.
1751
# build up paths that this id will be left at after the change is made,
1752
# so we can update their cross references in tree 0
1753
all_remaining_keys = set()
1754
# Dont check the working tree, because its going.
1755
for details in current_old[1][1:]:
1756
if details[0] not in ('a', 'r'): # absent, relocated
1757
all_remaining_keys.add(current_old[0])
1758
elif details[0] == 'r': # relocated
1759
# record the key for the real path.
1760
all_remaining_keys.add(tuple(osutils.split(details[1])) + (current_old[0][2],))
1761
# absent rows are not present at any path.
1762
last_reference = current_old[0] not in all_remaining_keys
1764
# the current row consists entire of the current item (being marked
1765
# absent), and relocated or absent entries for the other trees:
1766
# Remove it, its meaningless.
1767
block = self._find_block(current_old[0])
1768
entry_index, present = self._find_entry_index(current_old[0], block[1])
1770
block[1].pop(entry_index)
1771
# if we have an id_index in use, remove this key from it for this id.
1772
if self._id_index is not None:
1773
self._id_index[current_old[0][2]].remove(current_old[0])
1774
# update all remaining keys for this id to record it as absent. The
1775
# existing details may either be the record we are making as deleted
1776
# (if there were other trees with the id present at this path), or may
1778
for update_key in all_remaining_keys:
1779
update_block_index, present = \
1780
self._find_block_index_from_key(update_key)
1782
update_entry_index, present = \
1783
self._find_entry_index(update_key, self._dirblocks[update_block_index][1])
1785
update_tree_details = self._dirblocks[update_block_index][1][update_entry_index][1]
1786
# it must not be absent at the moment
1787
assert update_tree_details[0][0] != 'a' # absent
1788
update_tree_details[0] = DirState.NULL_PARENT_DETAILS
1789
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
1790
return last_reference
1792
def update_minimal(self, key, minikind, executable=False, fingerprint='',
1793
packed_stat=None, size=0, path_utf8=None):
1794
"""Update an entry to the state in tree 0.
1796
This will either create a new entry at 'key' or update an existing one.
1797
It also makes sure that any other records which might mention this are
1800
:param key: (dir, name, file_id) for the new entry
1801
:param minikind: The type for the entry ('f' == 'file', 'd' ==
1803
:param executable: Should the executable bit be set?
1804
:param fingerprint: Simple fingerprint for new entry.
1805
:param packed_stat: packed stat value for new entry.
1806
:param size: Size information for new entry
1807
:param path_utf8: key[0] + '/' + key[1], just passed in to avoid doing
1810
block = self._find_block(key)[1]
1811
if packed_stat is None:
1812
packed_stat = DirState.NULLSTAT
1813
entry_index, present = self._find_entry_index(key, block)
1814
new_details = (minikind, fingerprint, size, executable, packed_stat)
1815
id_index = self._get_id_index()
1817
# new entry, synthesis cross reference here,
1818
existing_keys = id_index.setdefault(key[2], set())
1819
if not existing_keys:
1820
# not currently in the state, simplest case
1821
new_entry = key, [new_details] + self._empty_parent_info()
1823
# present at one or more existing other paths.
1824
# grab one of them and use it to generate parent
1825
# relocation/absent entries.
1826
new_entry = key, [new_details]
1827
for other_key in existing_keys:
1828
# change the record at other to be a pointer to this new
1829
# record. The loop looks similar to the change to
1830
# relocations when updating an existing record but its not:
1831
# the test for existing kinds is different: this can be
1832
# factored out to a helper though.
1833
other_block_index, present = self._find_block_index_from_key(other_key)
1835
other_entry_index, present = self._find_entry_index(other_key, self._dirblocks[other_block_index][1])
1837
assert path_utf8 is not None
1838
self._dirblocks[other_block_index][1][other_entry_index][1][0] = \
1839
('r', path_utf8, 0, False, '')
1841
num_present_parents = self._num_present_parents()
1842
for lookup_index in xrange(1, num_present_parents + 1):
1843
# grab any one entry, use it to find the right path.
1844
# TODO: optimise this to reduce memory use in highly
1845
# fragmented situations by reusing the relocation
1847
update_block_index, present = \
1848
self._find_block_index_from_key(other_key)
1850
update_entry_index, present = \
1851
self._find_entry_index(other_key, self._dirblocks[update_block_index][1])
1853
update_details = self._dirblocks[update_block_index][1][update_entry_index][1][lookup_index]
1854
if update_details[0] in ('r', 'a'): # relocated, absent
1855
# its a pointer or absent in lookup_index's tree, use
1857
new_entry[1].append(update_details)
1859
# we have the right key, make a pointer to it.
1860
pointer_path = osutils.pathjoin(*other_key[0:2])
1861
new_entry[1].append(('r', pointer_path, 0, False, ''))
1862
block.insert(entry_index, new_entry)
1863
existing_keys.add(key)
1865
# Does the new state matter?
1866
block[entry_index][1][0] = new_details
1867
# parents cannot be affected by what we do.
1868
# other occurences of this id can be found
1869
# from the id index.
1871
# tree index consistency: All other paths for this id in this tree
1872
# index must point to the correct path. We have to loop here because
1873
# we may have passed entries in the state with this file id already
1874
# that were absent - where parent entries are - and they need to be
1875
# converted to relocated.
1876
assert path_utf8 is not None
1877
for entry_key in id_index.setdefault(key[2], set()):
1878
# TODO:PROFILING: It might be faster to just update
1879
# rather than checking if we need to, and then overwrite
1880
# the one we are located at.
1881
if entry_key != key:
1882
# this file id is at a different path in one of the
1883
# other trees, so put absent pointers there
1884
# This is the vertical axis in the matrix, all pointing
1886
block_index, present = self._find_block_index_from_key(entry_key)
1888
entry_index, present = self._find_entry_index(entry_key, self._dirblocks[block_index][1])
1890
self._dirblocks[block_index][1][entry_index][1][0] = \
1891
('r', path_utf8, 0, False, '')
1892
# add a containing dirblock if needed.
1893
if new_details[0] == 'd':
1894
subdir_key = (osutils.pathjoin(*key[0:2]), '', '')
1895
block_index, present = self._find_block_index_from_key(subdir_key)
1897
self._dirblocks.insert(block_index, (subdir_key[0], []))
1899
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
1902
def _wipe_state(self):
1903
"""Forget all state information about the dirstate."""
1904
self._header_state = DirState.NOT_IN_MEMORY
1905
self._dirblock_state = DirState.NOT_IN_MEMORY
1908
self._dirblocks = []
1910
def lock_read(self):
1911
"""Acquire a read lock on the dirstate"""
1912
if self._lock_token is not None:
1913
raise errors.LockContention(self._lock_token)
1914
self._lock_token = lock.ReadLock(self._filename)
1915
self._state_file = self._lock_token.f
1918
def lock_write(self):
1919
"""Acquire a write lock on the dirstate"""
1920
if self._lock_token is not None:
1921
raise errors.LockContention(self._lock_token)
1922
self._lock_token = lock.WriteLock(self._filename)
1923
self._state_file = self._lock_token.f
1927
"""Drop any locks held on the dirstate"""
1928
if self._lock_token is None:
1929
raise errors.LockNotHeld(self)
1930
self._state_file = None
1931
self._lock_token.unlock()
1932
self._lock_token = None
1934
def _requires_lock(self):
1935
"""Checks that a lock is currently held by someone on the dirstate"""
1936
if not self._lock_token:
1937
raise errors.ObjectNotLocked(self)
1939
def pack_stat(st, _encode=base64.encodestring, _pack=struct.pack):
1940
"""Convert stat values into a packed representation."""
1941
# jam 20060614 it isn't really worth removing more entries if we
1942
# are going to leave it in packed form.
1943
# With only st_mtime and st_mode filesize is 5.5M and read time is 275ms
1944
# With all entries filesize is 5.9M and read time is mabye 280ms
1945
# well within the noise margin
1947
# base64.encode always adds a final newline, so strip it off
1948
return _encode(_pack('>llllll'
1949
, st.st_size, st.st_mtime, st.st_ctime
1950
, st.st_dev, st.st_ino, st.st_mode))[:-1]
added
removed
bzrlib/dirstate.py
