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" | "t";
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 = "crc32: ", ["-"], 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
89
't' is a reference to a nested subtree; the fingerprint is the referenced
94
The entries on disk and in memory are ordered according to the following keys:
96
directory, as a list of components
100
--- Format 1 had the following different definition: ---
101
rows = dirname, NULL, basename, NULL, MINIKIND, NULL, fileid_utf8, NULL,
102
WHOLE NUMBER (* size *), NULL, packed stat, NULL, sha1|symlink target,
104
PARENT ROW = NULL, revision_utf8, NULL, MINIKIND, NULL, dirname, NULL,
105
basename, NULL, WHOLE NUMBER (* size *), NULL, "y" | "n", NULL,
108
PARENT ROW's are emitted for every parent that is not in the ghosts details
109
line. That is, if the parents are foo, bar, baz, and the ghosts are bar, then
110
each row will have a PARENT ROW for foo and baz, but not for bar.
113
In any tree, a kind of 'moved' indicates that the fingerprint field
114
(which we treat as opaque data specific to the 'kind' anyway) has the
115
details for the id of this row in that tree.
117
I'm strongly tempted to add a id->path index as well, but I think that
118
where we need id->path mapping; we also usually read the whole file, so
119
I'm going to skip that for the moment, as we have the ability to locate
120
via bisect any path in any tree, and if we lookup things by path, we can
121
accumulate a id->path mapping as we go, which will tend to match what we
124
I plan to implement this asap, so please speak up now to alter/tweak the
125
design - and once we stabilise on this, I'll update the wiki page for
128
The rationale for all this is that we want fast operations for the
129
common case (diff/status/commit/merge on all files) and extremely fast
130
operations for the less common but still occurs a lot status/diff/commit
131
on specific files). Operations on specific files involve a scan for all
132
the children of a path, *in every involved tree*, which the current
133
format did not accommodate.
137
1) Fast end to end use for bzr's top 5 uses cases. (commmit/diff/status/merge/???)
138
2) fall back current object model as needed.
139
3) scale usably to the largest trees known today - say 50K entries. (mozilla
140
is an example of this)
144
Eventually reuse dirstate objects across locks IFF the dirstate file has not
145
been modified, but will require that we flush/ignore cached stat-hit data
146
because we wont want to restat all files on disk just because a lock was
147
acquired, yet we cannot trust the data after the previous lock was released.
149
Memory representation:
150
vector of all directories, and vector of the childen ?
152
root_entrie = (direntry for root, [parent_direntries_for_root]),
154
('', ['data for achild', 'data for bchild', 'data for cchild'])
155
('dir', ['achild', 'cchild', 'echild'])
157
- single bisect to find N subtrees from a path spec
158
- in-order for serialisation - this is 'dirblock' grouping.
159
- insertion of a file '/a' affects only the '/' child-vector, that is, to
160
insert 10K elements from scratch does not generates O(N^2) memoves of a
161
single vector, rather each individual, which tends to be limited to a
162
manageable number. Will scale badly on trees with 10K entries in a
163
single directory. compare with Inventory.InventoryDirectory which has
164
a dictionary for the children. No bisect capability, can only probe for
165
exact matches, or grab all elements and sorta.
166
- Whats the risk of error here? Once we have the base format being processed
167
we should have a net win regardless of optimality. So we are going to
168
go with what seems reasonably.
171
maybe we should do a test profile of these core structure - 10K simulated searches/lookups/etc?
173
Objects for each row?
174
The lifetime of Dirstate objects is current per lock, but see above for
175
possible extensions. The lifetime of a row from a dirstate is expected to be
176
very short in the optimistic case: which we are optimising for. For instance,
177
subtree status will determine from analysis of the disk data what rows need to
178
be examined at all, and will be able to determine from a single row whether
179
that file has altered or not, so we are aiming to process tens of thousands of
180
entries each second within the dirstate context, before exposing anything to
181
the larger codebase. This suggests we want the time for a single file
182
comparison to be < 0.1 milliseconds. That would give us 10000 paths per second
183
processed, and to scale to 100 thousand we'll another order of magnitude to do
184
that. Now, as the lifetime for all unchanged entries is the time to parse, stat
185
the file on disk, and then immediately discard, the overhead of object creation
186
becomes a significant cost.
188
Figures: Creating a tuple from from 3 elements was profiled at 0.0625
189
microseconds, whereas creating a object which is subclassed from tuple was
190
0.500 microseconds, and creating an object with 3 elements and slots was 3
191
microseconds long. 0.1 milliseconds is 100 microseconds, and ideally we'll get
192
down to 10 microseconds for the total processing - having 33% of that be object
193
creation is a huge overhead. There is a potential cost in using tuples within
194
each row which is that the conditional code to do comparisons may be slower
195
than method invocation, but method invocation is known to be slow due to stack
196
frame creation, so avoiding methods in these tight inner loops in unfortunately
197
desirable. We can consider a pyrex version of this with objects in future if
207
from stat import S_IEXEC
222
class _Bisector(object):
223
"""This just keeps track of information as we are bisecting."""
226
class DirState(object):
227
"""Record directory and metadata state for fast access.
229
A dirstate is a specialised data structure for managing local working
230
tree state information. Its not yet well defined whether it is platform
231
specific, and if it is how we detect/parameterise that.
233
Dirstates use the usual lock_write, lock_read and unlock mechanisms.
234
Unlike most bzr disk formats, DirStates must be locked for reading, using
235
lock_read. (This is an os file lock internally.) This is necessary
236
because the file can be rewritten in place.
238
DirStates must be explicitly written with save() to commit changes; just
239
unlocking them does not write the changes to disk.
242
_kind_to_minikind = {
248
'tree-reference': 't',
250
_minikind_to_kind = {
256
't': 'tree-reference',
258
_to_yesno = {True:'y', False: 'n'} # TODO profile the performance gain
259
# of using int conversion rather than a dict here. AND BLAME ANDREW IF
262
# TODO: jam 20070221 Figure out what to do if we have a record that exceeds
263
# the BISECT_PAGE_SIZE. For now, we just have to make it large enough
264
# that we are sure a single record will always fit.
265
BISECT_PAGE_SIZE = 4096
268
IN_MEMORY_UNMODIFIED = 1
269
IN_MEMORY_MODIFIED = 2
271
# A pack_stat (the x's) that is just noise and will never match the output
274
NULL_PARENT_DETAILS = ('a', '', 0, False, '')
276
HEADER_FORMAT_2 = '#bazaar dirstate flat format 2\n'
277
HEADER_FORMAT_3 = '#bazaar dirstate flat format 3\n'
279
def __init__(self, path):
280
"""Create a DirState object.
284
:attr _root_entrie: The root row of the directory/file information,
285
- contains the path to / - '', ''
286
- kind of 'directory',
287
- the file id of the root in utf8
290
- and no sha information.
291
:param path: The path at which the dirstate file on disk should live.
293
# _header_state and _dirblock_state represent the current state
294
# of the dirstate metadata and the per-row data respectiely.
295
# NOT_IN_MEMORY indicates that no data is in memory
296
# IN_MEMORY_UNMODIFIED indicates that what we have in memory
297
# is the same as is on disk
298
# IN_MEMORY_MODIFIED indicates that we have a modified version
299
# of what is on disk.
300
# In future we will add more granularity, for instance _dirblock_state
301
# will probably support partially-in-memory as a separate variable,
302
# allowing for partially-in-memory unmodified and partially-in-memory
304
self._header_state = DirState.NOT_IN_MEMORY
305
self._dirblock_state = DirState.NOT_IN_MEMORY
309
self._state_file = None
310
self._filename = path
311
self._lock_token = None
312
self._lock_state = None
313
self._id_index = None
314
self._end_of_header = None
315
self._cutoff_time = None
316
self._split_path_cache = {}
317
self._bisect_page_size = DirState.BISECT_PAGE_SIZE
321
(self.__class__.__name__, self._filename)
323
def add(self, path, file_id, kind, stat, fingerprint):
324
"""Add a path to be tracked.
326
:param path: The path within the dirstate - '' is the root, 'foo' is the
327
path foo within the root, 'foo/bar' is the path bar within foo
329
:param file_id: The file id of the path being added.
330
:param kind: The kind of the path, as a string like 'file',
332
:param stat: The output of os.lstat for the path.
333
:param fingerprint: The sha value of the file,
334
or the target of a symlink,
335
or the referenced revision id for tree-references,
336
or '' for directories.
339
# find the block its in.
340
# find the location in the block.
341
# check its not there
343
#------- copied from inventory.make_entry
344
# --- normalized_filename wants a unicode basename only, so get one.
345
dirname, basename = osutils.split(path)
346
# we dont import normalized_filename directly because we want to be
347
# able to change the implementation at runtime for tests.
348
norm_name, can_access = osutils.normalized_filename(basename)
349
if norm_name != basename:
353
raise errors.InvalidNormalization(path)
354
# you should never have files called . or ..; just add the directory
355
# in the parent, or according to the special treatment for the root
356
if basename == '.' or basename == '..':
357
raise errors.InvalidEntryName(path)
358
# now that we've normalised, we need the correct utf8 path and
359
# dirname and basename elements. This single encode and split should be
360
# faster than three separate encodes.
361
utf8path = (dirname + '/' + basename).strip('/').encode('utf8')
362
dirname, basename = osutils.split(utf8path)
363
assert file_id.__class__ == str, \
364
"must be a utf8 file_id not %s" % (type(file_id))
365
# Make sure the file_id does not exist in this tree
366
file_id_entry = self._get_entry(0, fileid_utf8=file_id)
367
if file_id_entry != (None, None):
368
path = osutils.pathjoin(file_id_entry[0][0], file_id_entry[0][1])
369
kind = DirState._minikind_to_kind[file_id_entry[1][0][0]]
370
info = '%s:%s' % (kind, path)
371
raise errors.DuplicateFileId(file_id, info)
372
first_key = (dirname, basename, '')
373
block_index, present = self._find_block_index_from_key(first_key)
375
# check the path is not in the tree
376
block = self._dirblocks[block_index][1]
377
entry_index, _ = self._find_entry_index(first_key, block)
378
while (entry_index < len(block) and
379
block[entry_index][0][0:2] == first_key[0:2]):
380
if block[entry_index][1][0][0] not in 'ar':
381
# this path is in the dirstate in the current tree.
382
raise Exception, "adding already added path!"
385
# The block where we want to put the file is not present. But it
386
# might be because the directory was empty, or not loaded yet. Look
387
# for a parent entry, if not found, raise NotVersionedError
388
parent_dir, parent_base = osutils.split(dirname)
389
parent_block_idx, parent_entry_idx, _, parent_present = \
390
self._get_block_entry_index(parent_dir, parent_base, 0)
391
if not parent_present:
392
raise errors.NotVersionedError(path, str(self))
393
self._ensure_block(parent_block_idx, parent_entry_idx, dirname)
394
block = self._dirblocks[block_index][1]
395
entry_key = (dirname, basename, file_id)
398
packed_stat = DirState.NULLSTAT
401
packed_stat = pack_stat(stat)
402
parent_info = self._empty_parent_info()
403
minikind = DirState._kind_to_minikind[kind]
405
entry_data = entry_key, [
406
(minikind, fingerprint, size, False, packed_stat),
408
elif kind == 'directory':
409
entry_data = entry_key, [
410
(minikind, '', 0, False, packed_stat),
412
elif kind == 'symlink':
413
entry_data = entry_key, [
414
(minikind, fingerprint, size, False, packed_stat),
416
elif kind == 'tree-reference':
417
entry_data = entry_key, [
418
(minikind, fingerprint, 0, False, packed_stat),
421
raise errors.BzrError('unknown kind %r' % kind)
422
entry_index, present = self._find_entry_index(entry_key, block)
424
block.insert(entry_index, entry_data)
426
assert block[entry_index][1][0][0] == 'a', " %r(%r) already added" % (basename, file_id)
427
block[entry_index][1][0] = entry_data[1][0]
429
if kind == 'directory':
430
# insert a new dirblock
431
self._ensure_block(block_index, entry_index, utf8path)
432
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
434
self._id_index.setdefault(entry_key[2], set()).add(entry_key)
436
def _bisect(self, dir_name_list):
437
"""Bisect through the disk structure for specific rows.
439
:param dir_name_list: A list of (dir, name) pairs.
440
:return: A dict mapping (dir, name) => entry for found entries. Missing
441
entries will not be in the map.
443
self._requires_lock()
444
# We need the file pointer to be right after the initial header block
445
self._read_header_if_needed()
446
# If _dirblock_state was in memory, we should just return info from
447
# there, this function is only meant to handle when we want to read
449
assert self._dirblock_state == DirState.NOT_IN_MEMORY
451
# The disk representation is generally info + '\0\n\0' at the end. But
452
# for bisecting, it is easier to treat this as '\0' + info + '\0\n'
453
# Because it means we can sync on the '\n'
454
state_file = self._state_file
455
file_size = os.fstat(state_file.fileno()).st_size
456
# We end up with 2 extra fields, we should have a trailing '\n' to
457
# ensure that we read the whole record, and we should have a precursur
458
# '' which ensures that we start after the previous '\n'
459
entry_field_count = self._fields_per_entry() + 1
461
low = self._end_of_header
462
high = file_size - 1 # Ignore the final '\0'
463
# Map from (dir, name) => entry
466
# Avoid infinite seeking
467
max_count = 30*len(dir_name_list)
469
# pending is a list of places to look.
470
# each entry is a tuple of low, high, dir_names
471
# low -> the first byte offset to read (inclusive)
472
# high -> the last byte offset (inclusive)
473
# dir_names -> The list of (dir, name) pairs that should be found in
474
# the [low, high] range
475
pending = [(low, high, dir_name_list)]
477
page_size = self._bisect_page_size
479
fields_to_entry = self._get_fields_to_entry()
482
low, high, cur_files = pending.pop()
484
if not cur_files or low >= high:
489
if count > max_count:
490
raise errors.BzrError('Too many seeks, most likely a bug.')
492
mid = max(low, (low+high-page_size)/2)
495
# limit the read size, so we don't end up reading data that we have
497
read_size = min(page_size, (high-mid)+1)
498
block = state_file.read(read_size)
501
entries = block.split('\n')
504
# We didn't find a '\n', so we cannot have found any records.
505
# So put this range back and try again. But we know we have to
506
# increase the page size, because a single read did not contain
507
# a record break (so records must be larger than page_size)
509
pending.append((low, high, cur_files))
512
# Check the first and last entries, in case they are partial, or if
513
# we don't care about the rest of this page
515
first_fields = entries[0].split('\0')
516
if len(first_fields) < entry_field_count:
517
# We didn't get the complete first entry
518
# so move start, and grab the next, which
519
# should be a full entry
520
start += len(entries[0])+1
521
first_fields = entries[1].split('\0')
524
if len(first_fields) <= 2:
525
# We didn't even get a filename here... what do we do?
526
# Try a large page size and repeat this query
528
pending.append((low, high, cur_files))
531
# Find what entries we are looking for, which occur before and
532
# after this first record.
534
first_dir_name = (first_fields[1], first_fields[2])
535
first_loc = bisect.bisect_left(cur_files, first_dir_name)
537
# These exist before the current location
538
pre = cur_files[:first_loc]
539
# These occur after the current location, which may be in the
540
# data we read, or might be after the last entry
541
post = cur_files[first_loc:]
543
if post and len(first_fields) >= entry_field_count:
544
# We have files after the first entry
546
# Parse the last entry
547
last_entry_num = len(entries)-1
548
last_fields = entries[last_entry_num].split('\0')
549
if len(last_fields) < entry_field_count:
550
# The very last hunk was not complete,
551
# read the previous hunk
552
after = mid + len(block) - len(entries[-1])
554
last_fields = entries[last_entry_num].split('\0')
556
after = mid + len(block)
558
last_dir_name = (last_fields[1], last_fields[2])
559
last_loc = bisect.bisect_right(post, last_dir_name)
561
middle_files = post[:last_loc]
562
post = post[last_loc:]
565
# We have files that should occur in this block
566
# (>= first, <= last)
567
# Either we will find them here, or we can mark them as
570
if middle_files[0] == first_dir_name:
571
# We might need to go before this location
572
pre.append(first_dir_name)
573
if middle_files[-1] == last_dir_name:
574
post.insert(0, last_dir_name)
576
# Find out what paths we have
577
paths = {first_dir_name:[first_fields]}
578
# last_dir_name might == first_dir_name so we need to be
579
# careful if we should append rather than overwrite
580
if last_entry_num != first_entry_num:
581
paths.setdefault(last_dir_name, []).append(last_fields)
582
for num in xrange(first_entry_num+1, last_entry_num):
583
# TODO: jam 20070223 We are already splitting here, so
584
# shouldn't we just split the whole thing rather
585
# than doing the split again in add_one_record?
586
fields = entries[num].split('\0')
587
dir_name = (fields[1], fields[2])
588
paths.setdefault(dir_name, []).append(fields)
590
for dir_name in middle_files:
591
for fields in paths.get(dir_name, []):
592
# offset by 1 because of the opening '\0'
593
# consider changing fields_to_entry to avoid the
595
entry = fields_to_entry(fields[1:])
596
found.setdefault(dir_name, []).append(entry)
598
# Now we have split up everything into pre, middle, and post, and
599
# we have handled everything that fell in 'middle'.
600
# We add 'post' first, so that we prefer to seek towards the
601
# beginning, so that we will tend to go as early as we need, and
602
# then only seek forward after that.
604
pending.append((after, high, post))
606
pending.append((low, start-1, pre))
608
# Consider that we may want to return the directory entries in sorted
609
# order. For now, we just return them in whatever order we found them,
610
# and leave it up to the caller if they care if it is ordered or not.
613
def _bisect_dirblocks(self, dir_list):
614
"""Bisect through the disk structure to find entries in given dirs.
616
_bisect_dirblocks is meant to find the contents of directories, which
617
differs from _bisect, which only finds individual entries.
619
:param dir_list: An sorted list of directory names ['', 'dir', 'foo'].
620
:return: A map from dir => entries_for_dir
622
# TODO: jam 20070223 A lot of the bisecting logic could be shared
623
# between this and _bisect. It would require parameterizing the
624
# inner loop with a function, though. We should evaluate the
625
# performance difference.
626
self._requires_lock()
627
# We need the file pointer to be right after the initial header block
628
self._read_header_if_needed()
629
# If _dirblock_state was in memory, we should just return info from
630
# there, this function is only meant to handle when we want to read
632
assert self._dirblock_state == DirState.NOT_IN_MEMORY
634
# The disk representation is generally info + '\0\n\0' at the end. But
635
# for bisecting, it is easier to treat this as '\0' + info + '\0\n'
636
# Because it means we can sync on the '\n'
637
state_file = self._state_file
638
file_size = os.fstat(state_file.fileno()).st_size
639
# We end up with 2 extra fields, we should have a trailing '\n' to
640
# ensure that we read the whole record, and we should have a precursur
641
# '' which ensures that we start after the previous '\n'
642
entry_field_count = self._fields_per_entry() + 1
644
low = self._end_of_header
645
high = file_size - 1 # Ignore the final '\0'
646
# Map from dir => entry
649
# Avoid infinite seeking
650
max_count = 30*len(dir_list)
652
# pending is a list of places to look.
653
# each entry is a tuple of low, high, dir_names
654
# low -> the first byte offset to read (inclusive)
655
# high -> the last byte offset (inclusive)
656
# dirs -> The list of directories that should be found in
657
# the [low, high] range
658
pending = [(low, high, dir_list)]
660
page_size = self._bisect_page_size
662
fields_to_entry = self._get_fields_to_entry()
665
low, high, cur_dirs = pending.pop()
667
if not cur_dirs or low >= high:
672
if count > max_count:
673
raise errors.BzrError('Too many seeks, most likely a bug.')
675
mid = max(low, (low+high-page_size)/2)
678
# limit the read size, so we don't end up reading data that we have
680
read_size = min(page_size, (high-mid)+1)
681
block = state_file.read(read_size)
684
entries = block.split('\n')
687
# We didn't find a '\n', so we cannot have found any records.
688
# So put this range back and try again. But we know we have to
689
# increase the page size, because a single read did not contain
690
# a record break (so records must be larger than page_size)
692
pending.append((low, high, cur_dirs))
695
# Check the first and last entries, in case they are partial, or if
696
# we don't care about the rest of this page
698
first_fields = entries[0].split('\0')
699
if len(first_fields) < entry_field_count:
700
# We didn't get the complete first entry
701
# so move start, and grab the next, which
702
# should be a full entry
703
start += len(entries[0])+1
704
first_fields = entries[1].split('\0')
707
if len(first_fields) <= 1:
708
# We didn't even get a dirname here... what do we do?
709
# Try a large page size and repeat this query
711
pending.append((low, high, cur_dirs))
714
# Find what entries we are looking for, which occur before and
715
# after this first record.
717
first_dir = first_fields[1]
718
first_loc = bisect.bisect_left(cur_dirs, first_dir)
720
# These exist before the current location
721
pre = cur_dirs[:first_loc]
722
# These occur after the current location, which may be in the
723
# data we read, or might be after the last entry
724
post = cur_dirs[first_loc:]
726
if post and len(first_fields) >= entry_field_count:
727
# We have records to look at after the first entry
729
# Parse the last entry
730
last_entry_num = len(entries)-1
731
last_fields = entries[last_entry_num].split('\0')
732
if len(last_fields) < entry_field_count:
733
# The very last hunk was not complete,
734
# read the previous hunk
735
after = mid + len(block) - len(entries[-1])
737
last_fields = entries[last_entry_num].split('\0')
739
after = mid + len(block)
741
last_dir = last_fields[1]
742
last_loc = bisect.bisect_right(post, last_dir)
744
middle_files = post[:last_loc]
745
post = post[last_loc:]
748
# We have files that should occur in this block
749
# (>= first, <= last)
750
# Either we will find them here, or we can mark them as
753
if middle_files[0] == first_dir:
754
# We might need to go before this location
755
pre.append(first_dir)
756
if middle_files[-1] == last_dir:
757
post.insert(0, last_dir)
759
# Find out what paths we have
760
paths = {first_dir:[first_fields]}
761
# last_dir might == first_dir so we need to be
762
# careful if we should append rather than overwrite
763
if last_entry_num != first_entry_num:
764
paths.setdefault(last_dir, []).append(last_fields)
765
for num in xrange(first_entry_num+1, last_entry_num):
766
# TODO: jam 20070223 We are already splitting here, so
767
# shouldn't we just split the whole thing rather
768
# than doing the split again in add_one_record?
769
fields = entries[num].split('\0')
770
paths.setdefault(fields[1], []).append(fields)
772
for cur_dir in middle_files:
773
for fields in paths.get(cur_dir, []):
774
# offset by 1 because of the opening '\0'
775
# consider changing fields_to_entry to avoid the
777
entry = fields_to_entry(fields[1:])
778
found.setdefault(cur_dir, []).append(entry)
780
# Now we have split up everything into pre, middle, and post, and
781
# we have handled everything that fell in 'middle'.
782
# We add 'post' first, so that we prefer to seek towards the
783
# beginning, so that we will tend to go as early as we need, and
784
# then only seek forward after that.
786
pending.append((after, high, post))
788
pending.append((low, start-1, pre))
792
def _bisect_recursive(self, dir_name_list):
793
"""Bisect for entries for all paths and their children.
795
This will use bisect to find all records for the supplied paths. It
796
will then continue to bisect for any records which are marked as
797
directories. (and renames?)
799
:param paths: A sorted list of (dir, name) pairs
800
eg: [('', 'a'), ('', 'f'), ('a/b', 'c')]
801
:return: A dictionary mapping (dir, name, file_id) => [tree_info]
803
# Map from (dir, name, file_id) => [tree_info]
806
found_dir_names = set()
808
# Directories that have been read
809
processed_dirs = set()
810
# Get the ball rolling with the first bisect for all entries.
811
newly_found = self._bisect(dir_name_list)
814
# Directories that need to be read
816
paths_to_search = set()
817
for entry_list in newly_found.itervalues():
818
for dir_name_id, trees_info in entry_list:
819
found[dir_name_id] = trees_info
820
found_dir_names.add(dir_name_id[:2])
822
for tree_info in trees_info:
823
minikind = tree_info[0]
826
# We already processed this one as a directory,
827
# we don't need to do the extra work again.
829
subdir, name, file_id = dir_name_id
830
path = osutils.pathjoin(subdir, name)
832
if path not in processed_dirs:
833
pending_dirs.add(path)
834
elif minikind == 'r':
835
# Rename, we need to directly search the target
836
# which is contained in the fingerprint column
837
dir_name = osutils.split(tree_info[1])
838
if dir_name[0] in pending_dirs:
839
# This entry will be found in the dir search
841
# TODO: We need to check if this entry has
842
# already been found. Otherwise we might be
843
# hitting infinite recursion.
844
if dir_name not in found_dir_names:
845
paths_to_search.add(dir_name)
846
# Now we have a list of paths to look for directly, and
847
# directory blocks that need to be read.
848
# newly_found is mixing the keys between (dir, name) and path
849
# entries, but that is okay, because we only really care about the
851
newly_found = self._bisect(sorted(paths_to_search))
852
newly_found.update(self._bisect_dirblocks(sorted(pending_dirs)))
853
processed_dirs.update(pending_dirs)
856
def _empty_parent_info(self):
857
return [DirState.NULL_PARENT_DETAILS] * (len(self._parents) -
860
def _ensure_block(self, parent_block_index, parent_row_index, dirname):
861
"""Ensure a block for dirname exists.
863
This function exists to let callers which know that there is a
864
directory dirname ensure that the block for it exists. This block can
865
fail to exist because of demand loading, or because a directory had no
866
children. In either case it is not an error. It is however an error to
867
call this if there is no parent entry for the directory, and thus the
868
function requires the coordinates of such an entry to be provided.
870
The root row is special cased and can be indicated with a parent block
873
:param parent_block_index: The index of the block in which dirname's row
875
:param parent_row_index: The index in the parent block where the row
877
:param dirname: The utf8 dirname to ensure there is a block for.
878
:return: The index for the block.
880
if dirname == '' and parent_row_index == 0 and parent_block_index == 0:
881
# This is the signature of the root row, and the
882
# contents-of-root row is always index 1
884
# the basename of the directory must be the end of its full name.
885
if not (parent_block_index == -1 and
886
parent_block_index == -1 and dirname == ''):
887
assert dirname.endswith(
888
self._dirblocks[parent_block_index][1][parent_row_index][0][1])
889
block_index, present = self._find_block_index_from_key((dirname, '', ''))
891
## In future, when doing partial parsing, this should load and
892
# populate the entire block.
893
self._dirblocks.insert(block_index, (dirname, []))
896
def _entries_to_current_state(self, new_entries):
897
"""Load new_entries into self.dirblocks.
899
Process new_entries into the current state object, making them the active
900
state. The entries are grouped together by directory to form dirblocks.
902
:param new_entries: A sorted list of entries. This function does not sort
903
to prevent unneeded overhead when callers have a sorted list already.
906
assert new_entries[0][0][0:2] == ('', ''), \
907
"Missing root row %r" % (new_entries[0][0],)
908
# The two blocks here are deliberate: the root block and the
909
# contents-of-root block.
910
self._dirblocks = [('', []), ('', [])]
911
current_block = self._dirblocks[0][1]
914
append_entry = current_block.append
915
for entry in new_entries:
916
if entry[0][0] != current_dirname:
917
# new block - different dirname
919
current_dirname = entry[0][0]
920
self._dirblocks.append((current_dirname, current_block))
921
append_entry = current_block.append
922
# append the entry to the current block
924
self._split_root_dirblock_into_contents()
926
def _split_root_dirblock_into_contents(self):
927
"""Split the root dirblocks into root and contents-of-root.
929
After parsing by path, we end up with root entries and contents-of-root
930
entries in the same block. This loop splits them out again.
932
# The above loop leaves the "root block" entries mixed with the
933
# "contents-of-root block". But we don't want an if check on
934
# all entries, so instead we just fix it up here.
935
assert self._dirblocks[1] == ('', [])
937
contents_of_root_block = []
938
for entry in self._dirblocks[0][1]:
939
if not entry[0][1]: # This is a root entry
940
root_block.append(entry)
942
contents_of_root_block.append(entry)
943
self._dirblocks[0] = ('', root_block)
944
self._dirblocks[1] = ('', contents_of_root_block)
946
def _entry_to_line(self, entry):
947
"""Serialize entry to a NULL delimited line ready for _get_output_lines.
949
:param entry: An entry_tuple as defined in the module docstring.
951
entire_entry = list(entry[0])
952
for tree_number, tree_data in enumerate(entry[1]):
953
# (minikind, fingerprint, size, executable, tree_specific_string)
954
entire_entry.extend(tree_data)
955
# 3 for the key, 5 for the fields per tree.
956
tree_offset = 3 + tree_number * 5
958
entire_entry[tree_offset + 0] = tree_data[0]
960
entire_entry[tree_offset + 2] = str(tree_data[2])
962
entire_entry[tree_offset + 3] = DirState._to_yesno[tree_data[3]]
963
return '\0'.join(entire_entry)
965
def _fields_per_entry(self):
966
"""How many null separated fields should be in each entry row.
968
Each line now has an extra '\n' field which is not used
969
so we just skip over it
972
+ number of fields per tree_data (5) * tree count
975
tree_count = 1 + self._num_present_parents()
976
return 3 + 5 * tree_count + 1
978
def _find_block(self, key, add_if_missing=False):
979
"""Return the block that key should be present in.
981
:param key: A dirstate entry key.
982
:return: The block tuple.
984
block_index, present = self._find_block_index_from_key(key)
986
if not add_if_missing:
987
# check to see if key is versioned itself - we might want to
988
# add it anyway, because dirs with no entries dont get a
989
# dirblock at parse time.
990
# This is an uncommon branch to take: most dirs have children,
991
# and most code works with versioned paths.
992
parent_base, parent_name = osutils.split(key[0])
993
if not self._get_block_entry_index(parent_base, parent_name, 0)[3]:
994
# some parent path has not been added - its an error to add
996
raise errors.NotVersionedError(key[0:2], str(self))
997
self._dirblocks.insert(block_index, (key[0], []))
998
return self._dirblocks[block_index]
1000
def _find_block_index_from_key(self, key):
1001
"""Find the dirblock index for a key.
1003
:return: The block index, True if the block for the key is present.
1005
if key[0:2] == ('', ''):
1007
block_index = bisect_dirblock(self._dirblocks, key[0], 1,
1008
cache=self._split_path_cache)
1009
# _right returns one-past-where-key is so we have to subtract
1010
# one to use it. we use _right here because there are two
1011
# '' blocks - the root, and the contents of root
1012
# we always have a minimum of 2 in self._dirblocks: root and
1013
# root-contents, and for '', we get 2 back, so this is
1014
# simple and correct:
1015
present = (block_index < len(self._dirblocks) and
1016
self._dirblocks[block_index][0] == key[0])
1017
return block_index, present
1019
def _find_entry_index(self, key, block):
1020
"""Find the entry index for a key in a block.
1022
:return: The entry index, True if the entry for the key is present.
1024
entry_index = bisect.bisect_left(block, (key, []))
1025
present = (entry_index < len(block) and
1026
block[entry_index][0] == key)
1027
return entry_index, present
1030
def from_tree(tree, dir_state_filename):
1031
"""Create a dirstate from a bzr Tree.
1033
:param tree: The tree which should provide parent information and
1035
:return: a DirState object which is currently locked for writing.
1036
(it was locked by DirState.initialize)
1038
result = DirState.initialize(dir_state_filename)
1042
parent_ids = tree.get_parent_ids()
1043
num_parents = len(parent_ids)
1045
for parent_id in parent_ids:
1046
parent_tree = tree.branch.repository.revision_tree(parent_id)
1047
parent_trees.append((parent_id, parent_tree))
1048
parent_tree.lock_read()
1049
result.set_parent_trees(parent_trees, [])
1050
result.set_state_from_inventory(tree.inventory)
1052
for revid, parent_tree in parent_trees:
1053
parent_tree.unlock()
1056
# The caller won't have a chance to unlock this, so make sure we
1062
def update_entry(self, entry, abspath, stat_value=None):
1063
"""Update the entry based on what is actually on disk.
1065
:param entry: This is the dirblock entry for the file in question.
1066
:param abspath: The path on disk for this file.
1067
:param stat_value: (optional) if we already have done a stat on the
1069
:return: The sha1 hexdigest of the file (40 bytes) or link target of a
1072
# This code assumes that the entry passed in is directly held in one of
1073
# the internal _dirblocks. So the dirblock state must have already been
1075
assert self._dirblock_state != DirState.NOT_IN_MEMORY
1076
if stat_value is None:
1078
# We could inline os.lstat but the common case is that
1079
# stat_value will be passed in, not read here.
1080
stat_value = self._lstat(abspath, entry)
1081
except (OSError, IOError), e:
1082
if e.errno in (errno.ENOENT, errno.EACCES,
1084
# The entry is missing, consider it gone
1088
kind = osutils.file_kind_from_stat_mode(stat_value.st_mode)
1090
minikind = DirState._kind_to_minikind[kind]
1091
except KeyError: # Unknown kind
1093
packed_stat = pack_stat(stat_value)
1094
(saved_minikind, saved_link_or_sha1, saved_file_size,
1095
saved_executable, saved_packed_stat) = entry[1][0]
1097
if (minikind == saved_minikind
1098
and packed_stat == saved_packed_stat
1099
# size should also be in packed_stat
1100
and saved_file_size == stat_value.st_size):
1101
# The stat hasn't changed since we saved, so we can potentially
1102
# re-use the saved sha hash.
1106
if self._cutoff_time is None:
1107
self._sha_cutoff_time()
1109
if (stat_value.st_mtime < self._cutoff_time
1110
and stat_value.st_ctime < self._cutoff_time):
1111
# Return the existing fingerprint
1112
return saved_link_or_sha1
1114
# If we have gotten this far, that means that we need to actually
1115
# process this entry.
1118
link_or_sha1 = self._sha1_file(abspath, entry)
1119
executable = self._is_executable(stat_value.st_mode,
1121
entry[1][0] = ('f', link_or_sha1, stat_value.st_size,
1122
executable, packed_stat)
1123
elif minikind == 'd':
1125
entry[1][0] = ('d', '', 0, False, packed_stat)
1126
if saved_minikind != 'd':
1127
# This changed from something into a directory. Make sure we
1128
# have a directory block for it. This doesn't happen very
1129
# often, so this doesn't have to be super fast.
1130
block_index, entry_index, dir_present, file_present = \
1131
self._get_block_entry_index(entry[0][0], entry[0][1], 0)
1132
self._ensure_block(block_index, entry_index,
1133
osutils.pathjoin(entry[0][0], entry[0][1]))
1134
elif minikind == 'l':
1135
link_or_sha1 = self._read_link(abspath, saved_link_or_sha1)
1136
entry[1][0] = ('l', link_or_sha1, stat_value.st_size,
1138
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
1141
def _sha_cutoff_time(self):
1142
"""Return cutoff time.
1144
Files modified more recently than this time are at risk of being
1145
undetectably modified and so can't be cached.
1147
# Cache the cutoff time as long as we hold a lock.
1148
# time.time() isn't super expensive (approx 3.38us), but
1149
# when you call it 50,000 times it adds up.
1150
# For comparison, os.lstat() costs 7.2us if it is hot.
1151
self._cutoff_time = int(time.time()) - 3
1152
return self._cutoff_time
1154
def _lstat(self, abspath, entry):
1155
"""Return the os.lstat value for this path."""
1156
return os.lstat(abspath)
1158
def _sha1_file(self, abspath, entry):
1159
"""Calculate the SHA1 of a file by reading the full text"""
1160
f = file(abspath, 'rb', buffering=65000)
1162
return osutils.sha_file(f)
1166
def _is_executable(self, mode, old_executable):
1167
"""Is this file executable?"""
1168
return bool(S_IEXEC & mode)
1170
def _is_executable_win32(self, mode, old_executable):
1171
"""On win32 the executable bit is stored in the dirstate."""
1172
return old_executable
1174
if sys.platform == 'win32':
1175
_is_executable = _is_executable_win32
1177
def _read_link(self, abspath, old_link):
1178
"""Read the target of a symlink"""
1179
# TODO: jam 200700301 On Win32, this could just return the value
1180
# already in memory. However, this really needs to be done at a
1181
# higher level, because there either won't be anything on disk,
1182
# or the thing on disk will be a file.
1183
return os.readlink(abspath)
1185
def get_ghosts(self):
1186
"""Return a list of the parent tree revision ids that are ghosts."""
1187
self._read_header_if_needed()
1190
def get_lines(self):
1191
"""Serialise the entire dirstate to a sequence of lines."""
1192
if (self._header_state == DirState.IN_MEMORY_UNMODIFIED and
1193
self._dirblock_state == DirState.IN_MEMORY_UNMODIFIED):
1194
# read whats on disk.
1195
self._state_file.seek(0)
1196
return self._state_file.readlines()
1198
lines.append(self._get_parents_line(self.get_parent_ids()))
1199
lines.append(self._get_ghosts_line(self._ghosts))
1200
# append the root line which is special cased
1201
lines.extend(map(self._entry_to_line, self._iter_entries()))
1202
return self._get_output_lines(lines)
1204
def _get_ghosts_line(self, ghost_ids):
1205
"""Create a line for the state file for ghost information."""
1206
return '\0'.join([str(len(ghost_ids))] + ghost_ids)
1208
def _get_parents_line(self, parent_ids):
1209
"""Create a line for the state file for parents information."""
1210
return '\0'.join([str(len(parent_ids))] + parent_ids)
1212
def _get_fields_to_entry(self):
1213
"""Get a function which converts entry fields into a entry record.
1215
This handles size and executable, as well as parent records.
1217
:return: A function which takes a list of fields, and returns an
1218
appropriate record for storing in memory.
1220
# This is intentionally unrolled for performance
1221
num_present_parents = self._num_present_parents()
1222
if num_present_parents == 0:
1223
def fields_to_entry_0_parents(fields, _int=int):
1224
path_name_file_id_key = (fields[0], fields[1], fields[2])
1225
return (path_name_file_id_key, [
1227
fields[3], # minikind
1228
fields[4], # fingerprint
1229
_int(fields[5]), # size
1230
fields[6] == 'y', # executable
1231
fields[7], # packed_stat or revision_id
1233
return fields_to_entry_0_parents
1234
elif num_present_parents == 1:
1235
def fields_to_entry_1_parent(fields, _int=int):
1236
path_name_file_id_key = (fields[0], fields[1], fields[2])
1237
return (path_name_file_id_key, [
1239
fields[3], # minikind
1240
fields[4], # fingerprint
1241
_int(fields[5]), # size
1242
fields[6] == 'y', # executable
1243
fields[7], # packed_stat or revision_id
1246
fields[8], # minikind
1247
fields[9], # fingerprint
1248
_int(fields[10]), # size
1249
fields[11] == 'y', # executable
1250
fields[12], # packed_stat or revision_id
1253
return fields_to_entry_1_parent
1254
elif num_present_parents == 2:
1255
def fields_to_entry_2_parents(fields, _int=int):
1256
path_name_file_id_key = (fields[0], fields[1], fields[2])
1257
return (path_name_file_id_key, [
1259
fields[3], # minikind
1260
fields[4], # fingerprint
1261
_int(fields[5]), # size
1262
fields[6] == 'y', # executable
1263
fields[7], # packed_stat or revision_id
1266
fields[8], # minikind
1267
fields[9], # fingerprint
1268
_int(fields[10]), # size
1269
fields[11] == 'y', # executable
1270
fields[12], # packed_stat or revision_id
1273
fields[13], # minikind
1274
fields[14], # fingerprint
1275
_int(fields[15]), # size
1276
fields[16] == 'y', # executable
1277
fields[17], # packed_stat or revision_id
1280
return fields_to_entry_2_parents
1282
def fields_to_entry_n_parents(fields, _int=int):
1283
path_name_file_id_key = (fields[0], fields[1], fields[2])
1284
trees = [(fields[cur], # minikind
1285
fields[cur+1], # fingerprint
1286
_int(fields[cur+2]), # size
1287
fields[cur+3] == 'y', # executable
1288
fields[cur+4], # stat or revision_id
1289
) for cur in xrange(3, len(fields)-1, 5)]
1290
return path_name_file_id_key, trees
1291
return fields_to_entry_n_parents
1293
def get_parent_ids(self):
1294
"""Return a list of the parent tree ids for the directory state."""
1295
self._read_header_if_needed()
1296
return list(self._parents)
1298
def _get_block_entry_index(self, dirname, basename, tree_index):
1299
"""Get the coordinates for a path in the state structure.
1301
:param dirname: The utf8 dirname to lookup.
1302
:param basename: The utf8 basename to lookup.
1303
:param tree_index: The index of the tree for which this lookup should
1305
:return: A tuple describing where the path is located, or should be
1306
inserted. The tuple contains four fields: the block index, the row
1307
index, anda two booleans are True when the directory is present, and
1308
when the entire path is present. There is no guarantee that either
1309
coordinate is currently reachable unless the found field for it is
1310
True. For instance, a directory not present in the searched tree
1311
may be returned with a value one greater than the current highest
1312
block offset. The directory present field will always be True when
1313
the path present field is True. The directory present field does
1314
NOT indicate that the directory is present in the searched tree,
1315
rather it indicates that there are at least some files in some
1318
self._read_dirblocks_if_needed()
1319
key = dirname, basename, ''
1320
block_index, present = self._find_block_index_from_key(key)
1322
# no such directory - return the dir index and 0 for the row.
1323
return block_index, 0, False, False
1324
block = self._dirblocks[block_index][1] # access the entries only
1325
entry_index, present = self._find_entry_index(key, block)
1326
# linear search through present entries at this path to find the one
1328
while entry_index < len(block) and block[entry_index][0][1] == basename:
1329
if block[entry_index][1][tree_index][0] not in \
1330
('a', 'r'): # absent, relocated
1331
return block_index, entry_index, True, True
1333
return block_index, entry_index, True, False
1335
def _get_entry(self, tree_index, fileid_utf8=None, path_utf8=None):
1336
"""Get the dirstate entry for path in tree tree_index
1338
If either file_id or path is supplied, it is used as the key to lookup.
1339
If both are supplied, the fastest lookup is used, and an error is
1340
raised if they do not both point at the same row.
1342
:param tree_index: The index of the tree we wish to locate this path
1343
in. If the path is present in that tree, the entry containing its
1344
details is returned, otherwise (None, None) is returned
1345
0 is the working tree, higher indexes are successive parent
1347
:param fileid_utf8: A utf8 file_id to look up.
1348
:param path_utf8: An utf8 path to be looked up.
1349
:return: The dirstate entry tuple for path, or (None, None)
1351
self._read_dirblocks_if_needed()
1352
if path_utf8 is not None:
1353
assert path_utf8.__class__ == str, 'path_utf8 is not a str: %s %s' % (type(path_utf8), path_utf8)
1354
# path lookups are faster
1355
dirname, basename = osutils.split(path_utf8)
1356
block_index, entry_index, dir_present, file_present = \
1357
self._get_block_entry_index(dirname, basename, tree_index)
1358
if not file_present:
1360
entry = self._dirblocks[block_index][1][entry_index]
1361
assert entry[0][2] and entry[1][tree_index][0] not in ('a', 'r'), 'unversioned entry?!?!'
1363
if entry[0][2] != fileid_utf8:
1364
raise errors.BzrError('integrity error ? : mismatching'
1365
' tree_index, file_id and path')
1368
assert fileid_utf8 is not None
1369
possible_keys = self._get_id_index().get(fileid_utf8, None)
1370
if not possible_keys:
1372
for key in possible_keys:
1373
block_index, present = \
1374
self._find_block_index_from_key(key)
1375
# strange, probably indicates an out of date
1376
# id index - for now, allow this.
1379
# WARNING: DO not change this code to use _get_block_entry_index
1380
# as that function is not suitable: it does not use the key
1381
# to lookup, and thus the wront coordinates are returned.
1382
block = self._dirblocks[block_index][1]
1383
entry_index, present = self._find_entry_index(key, block)
1385
entry = self._dirblocks[block_index][1][entry_index]
1386
if entry[1][tree_index][0] in 'fdlt':
1387
# this is the result we are looking for: the
1388
# real home of this file_id in this tree.
1390
if entry[1][tree_index][0] == 'a':
1391
# there is no home for this entry in this tree
1393
assert entry[1][tree_index][0] == 'r', \
1394
"entry %r has invalid minikind %r for tree %r" \
1396
entry[1][tree_index][0],
1398
real_path = entry[1][tree_index][1]
1399
return self._get_entry(tree_index, fileid_utf8=fileid_utf8,
1400
path_utf8=real_path)
1404
def initialize(cls, path):
1405
"""Create a new dirstate on path.
1407
The new dirstate will be an empty tree - that is it has no parents,
1408
and only a root node - which has id ROOT_ID.
1410
:param path: The name of the file for the dirstate.
1411
:return: A write-locked DirState object.
1413
# This constructs a new DirState object on a path, sets the _state_file
1414
# to a new empty file for that path. It then calls _set_data() with our
1415
# stock empty dirstate information - a root with ROOT_ID, no children,
1416
# and no parents. Finally it calls save() to ensure that this data will
1419
# root dir and root dir contents with no children.
1420
empty_tree_dirblocks = [('', []), ('', [])]
1421
# a new root directory, with a NULLSTAT.
1422
empty_tree_dirblocks[0][1].append(
1423
(('', '', inventory.ROOT_ID), [
1424
('d', '', 0, False, DirState.NULLSTAT),
1428
result._set_data([], empty_tree_dirblocks)
1435
def _inv_entry_to_details(self, inv_entry):
1436
"""Convert an inventory entry (from a revision tree) to state details.
1438
:param inv_entry: An inventory entry whose sha1 and link targets can be
1439
relied upon, and which has a revision set.
1440
:return: A details tuple - the details for a single tree at a path +
1443
kind = inv_entry.kind
1444
minikind = DirState._kind_to_minikind[kind]
1445
tree_data = inv_entry.revision
1446
assert len(tree_data) > 0, 'empty revision for the inv_entry.'
1447
if kind == 'directory':
1451
elif kind == 'symlink':
1452
fingerprint = inv_entry.symlink_target or ''
1455
elif kind == 'file':
1456
fingerprint = inv_entry.text_sha1 or ''
1457
size = inv_entry.text_size or 0
1458
executable = inv_entry.executable
1459
elif kind == 'tree-reference':
1460
fingerprint = inv_entry.reference_revision or ''
1464
raise Exception("can't pack %s" % inv_entry)
1465
return (minikind, fingerprint, size, executable, tree_data)
1467
def _iter_entries(self):
1468
"""Iterate over all the entries in the dirstate.
1470
Each yelt item is an entry in the standard format described in the
1471
docstring of bzrlib.dirstate.
1473
self._read_dirblocks_if_needed()
1474
for directory in self._dirblocks:
1475
for entry in directory[1]:
1478
def _get_id_index(self):
1479
"""Get an id index of self._dirblocks."""
1480
if self._id_index is None:
1482
for key, tree_details in self._iter_entries():
1483
id_index.setdefault(key[2], set()).add(key)
1484
self._id_index = id_index
1485
return self._id_index
1487
def _get_output_lines(self, lines):
1488
"""format lines for final output.
1490
:param lines: A sequece of lines containing the parents list and the
1493
output_lines = [DirState.HEADER_FORMAT_3]
1494
lines.append('') # a final newline
1495
inventory_text = '\0\n\0'.join(lines)
1496
output_lines.append('crc32: %s\n' % (zlib.crc32(inventory_text),))
1497
# -3, 1 for num parents, 1 for ghosts, 1 for final newline
1498
num_entries = len(lines)-3
1499
output_lines.append('num_entries: %s\n' % (num_entries,))
1500
output_lines.append(inventory_text)
1503
def _make_deleted_row(self, fileid_utf8, parents):
1504
"""Return a deleted for for fileid_utf8."""
1505
return ('/', 'RECYCLED.BIN', 'file', fileid_utf8, 0, DirState.NULLSTAT,
1508
def _num_present_parents(self):
1509
"""The number of parent entries in each record row."""
1510
return len(self._parents) - len(self._ghosts)
1514
"""Construct a DirState on the file at path path.
1516
:return: An unlocked DirState object, associated with the given path.
1518
result = DirState(path)
1521
def _read_dirblocks_if_needed(self):
1522
"""Read in all the dirblocks from the file if they are not in memory.
1524
This populates self._dirblocks, and sets self._dirblock_state to
1525
IN_MEMORY_UNMODIFIED. It is not currently ready for incremental block
1528
self._read_header_if_needed()
1529
if self._dirblock_state == DirState.NOT_IN_MEMORY:
1530
# move the _state_file pointer to after the header (in case bisect
1531
# has been called in the mean time)
1532
self._state_file.seek(self._end_of_header)
1533
text = self._state_file.read()
1534
# TODO: check the crc checksums. crc_measured = zlib.crc32(text)
1536
fields = text.split('\0')
1537
# Remove the last blank entry
1538
trailing = fields.pop()
1539
assert trailing == ''
1540
# consider turning fields into a tuple.
1542
# skip the first field which is the trailing null from the header.
1544
# Each line now has an extra '\n' field which is not used
1545
# so we just skip over it
1547
# 3 fields for the key
1548
# + number of fields per tree_data (5) * tree count
1550
num_present_parents = self._num_present_parents()
1551
tree_count = 1 + num_present_parents
1552
entry_size = self._fields_per_entry()
1553
expected_field_count = entry_size * self._num_entries
1554
field_count = len(fields)
1555
# this checks our adjustment, and also catches file too short.
1556
assert field_count - cur == expected_field_count, \
1557
'field count incorrect %s != %s, entry_size=%s, '\
1558
'num_entries=%s fields=%r' % (
1559
field_count - cur, expected_field_count, entry_size,
1560
self._num_entries, fields)
1562
if num_present_parents == 1:
1563
# Bind external functions to local names
1565
# We access all fields in order, so we can just iterate over
1566
# them. Grab an straight iterator over the fields. (We use an
1567
# iterator because we don't want to do a lot of additions, nor
1568
# do we want to do a lot of slicing)
1569
next = iter(fields).next
1570
# Move the iterator to the current position
1571
for x in xrange(cur):
1573
# The two blocks here are deliberate: the root block and the
1574
# contents-of-root block.
1575
self._dirblocks = [('', []), ('', [])]
1576
current_block = self._dirblocks[0][1]
1577
current_dirname = ''
1578
append_entry = current_block.append
1579
for count in xrange(self._num_entries):
1583
if dirname != current_dirname:
1584
# new block - different dirname
1586
current_dirname = dirname
1587
self._dirblocks.append((current_dirname, current_block))
1588
append_entry = current_block.append
1589
# we know current_dirname == dirname, so re-use it to avoid
1590
# creating new strings
1591
entry = ((current_dirname, name, file_id),
1594
next(), # fingerprint
1595
_int(next()), # size
1596
next() == 'y', # executable
1597
next(), # packed_stat or revision_id
1601
next(), # fingerprint
1602
_int(next()), # size
1603
next() == 'y', # executable
1604
next(), # packed_stat or revision_id
1608
assert trailing == '\n'
1609
# append the entry to the current block
1611
self._split_root_dirblock_into_contents()
1613
fields_to_entry = self._get_fields_to_entry()
1614
entries = [fields_to_entry(fields[pos:pos+entry_size])
1615
for pos in xrange(cur, field_count, entry_size)]
1616
self._entries_to_current_state(entries)
1617
# To convert from format 2 => format 3
1618
# self._dirblocks = sorted(self._dirblocks,
1619
# key=lambda blk:blk[0].split('/'))
1620
# To convert from format 3 => format 2
1621
# self._dirblocks = sorted(self._dirblocks)
1622
self._dirblock_state = DirState.IN_MEMORY_UNMODIFIED
1624
def _read_header(self):
1625
"""This reads in the metadata header, and the parent ids.
1627
After reading in, the file should be positioned at the null
1628
just before the start of the first record in the file.
1630
:return: (expected crc checksum, number of entries, parent list)
1632
self._read_prelude()
1633
parent_line = self._state_file.readline()
1634
info = parent_line.split('\0')
1635
num_parents = int(info[0])
1636
assert num_parents == len(info)-2, 'incorrect parent info line'
1637
self._parents = info[1:-1]
1639
ghost_line = self._state_file.readline()
1640
info = ghost_line.split('\0')
1641
num_ghosts = int(info[1])
1642
assert num_ghosts == len(info)-3, 'incorrect ghost info line'
1643
self._ghosts = info[2:-1]
1644
self._header_state = DirState.IN_MEMORY_UNMODIFIED
1645
self._end_of_header = self._state_file.tell()
1647
def _read_header_if_needed(self):
1648
"""Read the header of the dirstate file if needed."""
1649
# inline this as it will be called a lot
1650
if not self._lock_token:
1651
raise errors.ObjectNotLocked(self)
1652
if self._header_state == DirState.NOT_IN_MEMORY:
1655
def _read_prelude(self):
1656
"""Read in the prelude header of the dirstate file
1658
This only reads in the stuff that is not connected to the crc
1659
checksum. The position will be correct to read in the rest of
1660
the file and check the checksum after this point.
1661
The next entry in the file should be the number of parents,
1662
and their ids. Followed by a newline.
1664
header = self._state_file.readline()
1665
assert header == DirState.HEADER_FORMAT_3, \
1666
'invalid header line: %r' % (header,)
1667
crc_line = self._state_file.readline()
1668
assert crc_line.startswith('crc32: '), 'missing crc32 checksum'
1669
self.crc_expected = int(crc_line[len('crc32: '):-1])
1670
num_entries_line = self._state_file.readline()
1671
assert num_entries_line.startswith('num_entries: '), 'missing num_entries line'
1672
self._num_entries = int(num_entries_line[len('num_entries: '):-1])
1675
"""Save any pending changes created during this session.
1677
We reuse the existing file, because that prevents race conditions with
1678
file creation, and use oslocks on it to prevent concurrent modification
1679
and reads - because dirstates incremental data aggretation is not
1680
compatible with reading a modified file, and replacing a file in use by
1681
another process is impossible on windows.
1683
A dirstate in read only mode should be smart enough though to validate
1684
that the file has not changed, and otherwise discard its cache and
1685
start over, to allow for fine grained read lock duration, so 'status'
1686
wont block 'commit' - for example.
1688
if (self._header_state == DirState.IN_MEMORY_MODIFIED or
1689
self._dirblock_state == DirState.IN_MEMORY_MODIFIED):
1691
grabbed_write_lock = False
1692
if self._lock_state != 'w':
1693
grabbed_write_lock, new_lock = self._lock_token.temporary_write_lock()
1694
# Switch over to the new lock, as the old one may be closed.
1695
# TODO: jam 20070315 We should validate the disk file has
1696
# not changed contents. Since temporary_write_lock may
1697
# not be an atomic operation.
1698
self._lock_token = new_lock
1699
self._state_file = new_lock.f
1700
if not grabbed_write_lock:
1701
# We couldn't grab a write lock, so we switch back to a read one
1704
self._state_file.seek(0)
1705
self._state_file.writelines(self.get_lines())
1706
self._state_file.truncate()
1707
self._state_file.flush()
1708
self._header_state = DirState.IN_MEMORY_UNMODIFIED
1709
self._dirblock_state = DirState.IN_MEMORY_UNMODIFIED
1711
if grabbed_write_lock:
1712
self._lock_token = self._lock_token.restore_read_lock()
1713
self._state_file = self._lock_token.f
1714
# TODO: jam 20070315 We should validate the disk file has
1715
# not changed contents. Since restore_read_lock may
1716
# not be an atomic operation.
1718
def _set_data(self, parent_ids, dirblocks):
1719
"""Set the full dirstate data in memory.
1721
This is an internal function used to completely replace the objects
1722
in memory state. It puts the dirstate into state 'full-dirty'.
1724
:param parent_ids: A list of parent tree revision ids.
1725
:param dirblocks: A list containing one tuple for each directory in the
1726
tree. Each tuple contains the directory path and a list of entries
1727
found in that directory.
1729
# our memory copy is now authoritative.
1730
self._dirblocks = dirblocks
1731
self._header_state = DirState.IN_MEMORY_MODIFIED
1732
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
1733
self._parents = list(parent_ids)
1734
self._id_index = None
1736
def set_path_id(self, path, new_id):
1737
"""Change the id of path to new_id in the current working tree.
1739
:param path: The path inside the tree to set - '' is the root, 'foo'
1740
is the path foo in the root.
1741
:param new_id: The new id to assign to the path. This must be a utf8
1742
file id (not unicode, and not None).
1744
assert new_id.__class__ == str, \
1745
"path_id %r is not a plain string" % (new_id,)
1746
self._read_dirblocks_if_needed()
1749
raise NotImplementedError(self.set_path_id)
1750
# TODO: check new id is unique
1751
entry = self._get_entry(0, path_utf8=path)
1752
if entry[0][2] == new_id:
1753
# Nothing to change.
1755
# mark the old path absent, and insert a new root path
1756
self._make_absent(entry)
1757
self.update_minimal(('', '', new_id), 'd',
1758
path_utf8='', packed_stat=entry[1][0][4])
1759
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
1760
if self._id_index is not None:
1761
self._id_index.setdefault(new_id, set()).add(entry[0])
1763
def set_parent_trees(self, trees, ghosts):
1764
"""Set the parent trees for the dirstate.
1766
:param trees: A list of revision_id, tree tuples. tree must be provided
1767
even if the revision_id refers to a ghost: supply an empty tree in
1769
:param ghosts: A list of the revision_ids that are ghosts at the time
1773
# TODO: generate a list of parent indexes to preserve to save
1774
# processing specific parent trees. In the common case one tree will
1775
# be preserved - the left most parent.
1776
# TODO: if the parent tree is a dirstate, we might want to walk them
1777
# all by path in parallel for 'optimal' common-case performance.
1778
# generate new root row.
1779
self._read_dirblocks_if_needed()
1780
# TODO future sketch: Examine the existing parents to generate a change
1781
# map and then walk the new parent trees only, mapping them into the
1782
# dirstate. Walk the dirstate at the same time to remove unreferenced
1785
# sketch: loop over all entries in the dirstate, cherry picking
1786
# entries from the parent trees, if they are not ghost trees.
1787
# after we finish walking the dirstate, all entries not in the dirstate
1788
# are deletes, so we want to append them to the end as per the design
1789
# discussions. So do a set difference on ids with the parents to
1790
# get deletes, and add them to the end.
1791
# During the update process we need to answer the following questions:
1792
# - find other keys containing a fileid in order to create cross-path
1793
# links. We dont't trivially use the inventory from other trees
1794
# because this leads to either double touching, or to accessing
1796
# - find other keys containing a path
1797
# We accumulate each entry via this dictionary, including the root
1800
# we could do parallel iterators, but because file id data may be
1801
# scattered throughout, we dont save on index overhead: we have to look
1802
# at everything anyway. We can probably save cycles by reusing parent
1803
# data and doing an incremental update when adding an additional
1804
# parent, but for now the common cases are adding a new parent (merge),
1805
# and replacing completely (commit), and commit is more common: so
1806
# optimise merge later.
1808
# ---- start generation of full tree mapping data
1809
# what trees should we use?
1810
parent_trees = [tree for rev_id, tree in trees if rev_id not in ghosts]
1811
# how many trees do we end up with
1812
parent_count = len(parent_trees)
1814
# one: the current tree
1815
for entry in self._iter_entries():
1816
# skip entries not in the current tree
1817
if entry[1][0][0] in ('a', 'r'): # absent, relocated
1819
by_path[entry[0]] = [entry[1][0]] + \
1820
[DirState.NULL_PARENT_DETAILS] * parent_count
1821
id_index[entry[0][2]] = set([entry[0]])
1823
# now the parent trees:
1824
for tree_index, tree in enumerate(parent_trees):
1825
# the index is off by one, adjust it.
1826
tree_index = tree_index + 1
1827
# when we add new locations for a fileid we need these ranges for
1828
# any fileid in this tree as we set the by_path[id] to:
1829
# already_processed_tree_details + new_details + new_location_suffix
1830
# the suffix is from tree_index+1:parent_count+1.
1831
new_location_suffix = [DirState.NULL_PARENT_DETAILS] * (parent_count - tree_index)
1832
# now stitch in all the entries from this tree
1833
for path, entry in tree.inventory.iter_entries_by_dir():
1834
# here we process each trees details for each item in the tree.
1835
# we first update any existing entries for the id at other paths,
1836
# then we either create or update the entry for the id at the
1837
# right path, and finally we add (if needed) a mapping from
1838
# file_id to this path. We do it in this order to allow us to
1839
# avoid checking all known paths for the id when generating a
1840
# new entry at this path: by adding the id->path mapping last,
1841
# all the mappings are valid and have correct relocation
1842
# records where needed.
1843
file_id = entry.file_id
1844
path_utf8 = path.encode('utf8')
1845
dirname, basename = osutils.split(path_utf8)
1846
new_entry_key = (dirname, basename, file_id)
1847
# tree index consistency: All other paths for this id in this tree
1848
# index must point to the correct path.
1849
for entry_key in id_index.setdefault(file_id, set()):
1850
# TODO:PROFILING: It might be faster to just update
1851
# rather than checking if we need to, and then overwrite
1852
# the one we are located at.
1853
if entry_key != new_entry_key:
1854
# this file id is at a different path in one of the
1855
# other trees, so put absent pointers there
1856
# This is the vertical axis in the matrix, all pointing
1858
by_path[entry_key][tree_index] = ('r', path_utf8, 0, False, '')
1859
# by path consistency: Insert into an existing path record (trivial), or
1860
# add a new one with relocation pointers for the other tree indexes.
1861
if new_entry_key in id_index[file_id]:
1862
# there is already an entry where this data belongs, just insert it.
1863
by_path[new_entry_key][tree_index] = \
1864
self._inv_entry_to_details(entry)
1866
# add relocated entries to the horizontal axis - this row
1867
# mapping from path,id. We need to look up the correct path
1868
# for the indexes from 0 to tree_index -1
1870
for lookup_index in xrange(tree_index):
1871
# boundary case: this is the first occurence of file_id
1872
# so there are no id_indexs, possibly take this out of
1874
if not len(id_index[file_id]):
1875
new_details.append(DirState.NULL_PARENT_DETAILS)
1877
# grab any one entry, use it to find the right path.
1878
# TODO: optimise this to reduce memory use in highly
1879
# fragmented situations by reusing the relocation
1881
a_key = iter(id_index[file_id]).next()
1882
if by_path[a_key][lookup_index][0] in ('r', 'a'):
1883
# its a pointer or missing statement, use it as is.
1884
new_details.append(by_path[a_key][lookup_index])
1886
# we have the right key, make a pointer to it.
1887
real_path = ('/'.join(a_key[0:2])).strip('/')
1888
new_details.append(('r', real_path, 0, False, ''))
1889
new_details.append(self._inv_entry_to_details(entry))
1890
new_details.extend(new_location_suffix)
1891
by_path[new_entry_key] = new_details
1892
id_index[file_id].add(new_entry_key)
1893
# --- end generation of full tree mappings
1895
# sort and output all the entries
1896
new_entries = self._sort_entries(by_path.items())
1897
self._entries_to_current_state(new_entries)
1898
self._parents = [rev_id for rev_id, tree in trees]
1899
self._ghosts = list(ghosts)
1900
self._header_state = DirState.IN_MEMORY_MODIFIED
1901
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
1902
self._id_index = id_index
1905
def _sort_entries(self, entry_list):
1906
"""Given a list of entries, sort them into the right order.
1908
This is done when constructing a new dirstate from trees - normally we
1909
try to keep everything in sorted blocks all the time, but sometimes
1910
it's easier to sort after the fact.
1912
# TODO: Might be faster to do a schwartzian transform?
1914
# sort by: directory parts, file name, file id
1915
return entry[0][0].split('/'), entry[0][1], entry[0][2]
1916
return sorted(entry_list, key=_key)
1918
def set_state_from_inventory(self, new_inv):
1919
"""Set new_inv as the current state.
1921
This API is called by tree transform, and will usually occur with
1922
existing parent trees.
1924
:param new_inv: The inventory object to set current state from.
1926
self._read_dirblocks_if_needed()
1928
# incremental algorithm:
1929
# two iterators: current data and new data, both in dirblock order.
1930
new_iterator = new_inv.iter_entries_by_dir()
1931
# we will be modifying the dirstate, so we need a stable iterator. In
1932
# future we might write one, for now we just clone the state into a
1933
# list - which is a shallow copy, so each
1934
old_iterator = iter(list(self._iter_entries()))
1935
# both must have roots so this is safe:
1936
current_new = new_iterator.next()
1937
current_old = old_iterator.next()
1938
def advance(iterator):
1940
return iterator.next()
1941
except StopIteration:
1943
while current_new or current_old:
1944
# skip entries in old that are not really there
1945
if current_old and current_old[1][0][0] in ('r', 'a'):
1946
# relocated or absent
1947
current_old = advance(old_iterator)
1950
# convert new into dirblock style
1951
new_path_utf8 = current_new[0].encode('utf8')
1952
new_dirname, new_basename = osutils.split(new_path_utf8)
1953
new_id = current_new[1].file_id
1954
new_entry_key = (new_dirname, new_basename, new_id)
1955
current_new_minikind = \
1956
DirState._kind_to_minikind[current_new[1].kind]
1957
if current_new_minikind == 't':
1958
fingerprint = current_new[1].reference_revision
1962
# for safety disable variables
1963
new_path_utf8 = new_dirname = new_basename = new_id = new_entry_key = None
1964
# 5 cases, we dont have a value that is strictly greater than everything, so
1965
# we make both end conditions explicit
1967
# old is finished: insert current_new into the state.
1968
self.update_minimal(new_entry_key, current_new_minikind,
1969
executable=current_new[1].executable,
1970
path_utf8=new_path_utf8, fingerprint=fingerprint)
1971
current_new = advance(new_iterator)
1972
elif not current_new:
1974
self._make_absent(current_old)
1975
current_old = advance(old_iterator)
1976
elif new_entry_key == current_old[0]:
1977
# same - common case
1978
# TODO: update the record if anything significant has changed.
1979
# the minimal required trigger is if the execute bit or cached
1981
if (current_old[1][0][3] != current_new[1].executable or
1982
current_old[1][0][0] != current_new_minikind):
1983
self.update_minimal(current_old[0], current_new_minikind,
1984
executable=current_new[1].executable,
1985
path_utf8=new_path_utf8, fingerprint=fingerprint)
1986
# both sides are dealt with, move on
1987
current_old = advance(old_iterator)
1988
current_new = advance(new_iterator)
1989
elif new_entry_key < current_old[0]:
1991
# add a entry for this and advance new
1992
self.update_minimal(new_entry_key, current_new_minikind,
1993
executable=current_new[1].executable,
1994
path_utf8=new_path_utf8, fingerprint=fingerprint)
1995
current_new = advance(new_iterator)
1998
self._make_absent(current_old)
1999
current_old = advance(old_iterator)
2000
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
2001
self._id_index = None
2003
def _make_absent(self, current_old):
2004
"""Mark current_old - an entry - as absent for tree 0.
2006
:return: True if this was the last details entry for they entry key:
2007
that is, if the underlying block has had the entry removed, thus
2008
shrinking in length.
2010
# build up paths that this id will be left at after the change is made,
2011
# so we can update their cross references in tree 0
2012
all_remaining_keys = set()
2013
# Dont check the working tree, because its going.
2014
for details in current_old[1][1:]:
2015
if details[0] not in ('a', 'r'): # absent, relocated
2016
all_remaining_keys.add(current_old[0])
2017
elif details[0] == 'r': # relocated
2018
# record the key for the real path.
2019
all_remaining_keys.add(tuple(osutils.split(details[1])) + (current_old[0][2],))
2020
# absent rows are not present at any path.
2021
last_reference = current_old[0] not in all_remaining_keys
2023
# the current row consists entire of the current item (being marked
2024
# absent), and relocated or absent entries for the other trees:
2025
# Remove it, its meaningless.
2026
block = self._find_block(current_old[0])
2027
entry_index, present = self._find_entry_index(current_old[0], block[1])
2028
assert present, 'could not find entry for %s' % (current_old,)
2029
block[1].pop(entry_index)
2030
# if we have an id_index in use, remove this key from it for this id.
2031
if self._id_index is not None:
2032
self._id_index[current_old[0][2]].remove(current_old[0])
2033
# update all remaining keys for this id to record it as absent. The
2034
# existing details may either be the record we are making as deleted
2035
# (if there were other trees with the id present at this path), or may
2037
for update_key in all_remaining_keys:
2038
update_block_index, present = \
2039
self._find_block_index_from_key(update_key)
2040
assert present, 'could not find block for %s' % (update_key,)
2041
update_entry_index, present = \
2042
self._find_entry_index(update_key, self._dirblocks[update_block_index][1])
2043
assert present, 'could not find entry for %s' % (update_key,)
2044
update_tree_details = self._dirblocks[update_block_index][1][update_entry_index][1]
2045
# it must not be absent at the moment
2046
assert update_tree_details[0][0] != 'a' # absent
2047
update_tree_details[0] = DirState.NULL_PARENT_DETAILS
2048
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
2049
return last_reference
2051
def update_minimal(self, key, minikind, executable=False, fingerprint='',
2052
packed_stat=None, size=0, path_utf8=None):
2053
"""Update an entry to the state in tree 0.
2055
This will either create a new entry at 'key' or update an existing one.
2056
It also makes sure that any other records which might mention this are
2059
:param key: (dir, name, file_id) for the new entry
2060
:param minikind: The type for the entry ('f' == 'file', 'd' ==
2062
:param executable: Should the executable bit be set?
2063
:param fingerprint: Simple fingerprint for new entry.
2064
:param packed_stat: packed stat value for new entry.
2065
:param size: Size information for new entry
2066
:param path_utf8: key[0] + '/' + key[1], just passed in to avoid doing
2069
block = self._find_block(key)[1]
2070
if packed_stat is None:
2071
packed_stat = DirState.NULLSTAT
2072
entry_index, present = self._find_entry_index(key, block)
2073
new_details = (minikind, fingerprint, size, executable, packed_stat)
2074
id_index = self._get_id_index()
2076
# new entry, synthesis cross reference here,
2077
existing_keys = id_index.setdefault(key[2], set())
2078
if not existing_keys:
2079
# not currently in the state, simplest case
2080
new_entry = key, [new_details] + self._empty_parent_info()
2082
# present at one or more existing other paths.
2083
# grab one of them and use it to generate parent
2084
# relocation/absent entries.
2085
new_entry = key, [new_details]
2086
for other_key in existing_keys:
2087
# change the record at other to be a pointer to this new
2088
# record. The loop looks similar to the change to
2089
# relocations when updating an existing record but its not:
2090
# the test for existing kinds is different: this can be
2091
# factored out to a helper though.
2092
other_block_index, present = self._find_block_index_from_key(other_key)
2093
assert present, 'could not find block for %s' % (other_key,)
2094
other_entry_index, present = self._find_entry_index(other_key,
2095
self._dirblocks[other_block_index][1])
2096
assert present, 'could not find entry for %s' % (other_key,)
2097
assert path_utf8 is not None
2098
self._dirblocks[other_block_index][1][other_entry_index][1][0] = \
2099
('r', path_utf8, 0, False, '')
2101
num_present_parents = self._num_present_parents()
2102
for lookup_index in xrange(1, num_present_parents + 1):
2103
# grab any one entry, use it to find the right path.
2104
# TODO: optimise this to reduce memory use in highly
2105
# fragmented situations by reusing the relocation
2107
update_block_index, present = \
2108
self._find_block_index_from_key(other_key)
2109
assert present, 'could not find block for %s' % (other_key,)
2110
update_entry_index, present = \
2111
self._find_entry_index(other_key, self._dirblocks[update_block_index][1])
2112
assert present, 'could not find entry for %s' % (other_key,)
2113
update_details = self._dirblocks[update_block_index][1][update_entry_index][1][lookup_index]
2114
if update_details[0] in ('r', 'a'): # relocated, absent
2115
# its a pointer or absent in lookup_index's tree, use
2117
new_entry[1].append(update_details)
2119
# we have the right key, make a pointer to it.
2120
pointer_path = osutils.pathjoin(*other_key[0:2])
2121
new_entry[1].append(('r', pointer_path, 0, False, ''))
2122
block.insert(entry_index, new_entry)
2123
existing_keys.add(key)
2125
# Does the new state matter?
2126
block[entry_index][1][0] = new_details
2127
# parents cannot be affected by what we do.
2128
# other occurences of this id can be found
2129
# from the id index.
2131
# tree index consistency: All other paths for this id in this tree
2132
# index must point to the correct path. We have to loop here because
2133
# we may have passed entries in the state with this file id already
2134
# that were absent - where parent entries are - and they need to be
2135
# converted to relocated.
2136
assert path_utf8 is not None
2137
for entry_key in id_index.setdefault(key[2], set()):
2138
# TODO:PROFILING: It might be faster to just update
2139
# rather than checking if we need to, and then overwrite
2140
# the one we are located at.
2141
if entry_key != key:
2142
# this file id is at a different path in one of the
2143
# other trees, so put absent pointers there
2144
# This is the vertical axis in the matrix, all pointing
2146
block_index, present = self._find_block_index_from_key(entry_key)
2148
entry_index, present = self._find_entry_index(entry_key, self._dirblocks[block_index][1])
2150
self._dirblocks[block_index][1][entry_index][1][0] = \
2151
('r', path_utf8, 0, False, '')
2152
# add a containing dirblock if needed.
2153
if new_details[0] == 'd':
2154
subdir_key = (osutils.pathjoin(*key[0:2]), '', '')
2155
block_index, present = self._find_block_index_from_key(subdir_key)
2157
self._dirblocks.insert(block_index, (subdir_key[0], []))
2159
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
2161
def _validate(self):
2162
"""Check that invariants on the dirblock are correct.
2164
This can be useful in debugging; it shouldn't be necessary in
2167
This must be called with a lock held.
2169
# NOTE: This must always raise AssertionError not just assert,
2170
# otherwise it may not behave properly under python -O
2172
# TODO: All entries must have some content that's not 'a' or 'r',
2173
# otherwise it could just be removed.
2175
# TODO: All relocations must point directly to a real entry.
2177
# TODO: No repeated keys.
2180
from pprint import pformat
2181
self._read_dirblocks_if_needed()
2182
if len(self._dirblocks) > 0:
2183
if not self._dirblocks[0][0] == '':
2184
raise AssertionError(
2185
"dirblocks don't start with root block:\n" + \
2187
if len(self._dirblocks) > 1:
2188
if not self._dirblocks[1][0] == '':
2189
raise AssertionError(
2190
"dirblocks missing root directory:\n" + \
2192
# the dirblocks are sorted by their path components, name, and dir id
2193
dir_names = [d[0].split('/')
2194
for d in self._dirblocks[1:]]
2195
if dir_names != sorted(dir_names):
2196
raise AssertionError(
2197
"dir names are not in sorted order:\n" + \
2198
pformat(self._dirblocks) + \
2201
for dirblock in self._dirblocks:
2202
# within each dirblock, the entries are sorted by filename and
2204
for entry in dirblock[1]:
2205
if dirblock[0] != entry[0][0]:
2206
raise AssertionError(
2208
"doesn't match directory name in\n%r" %
2209
(entry, pformat(dirblock)))
2210
if dirblock[1] != sorted(dirblock[1]):
2211
raise AssertionError(
2212
"dirblock for %r is not sorted:\n%s" % \
2213
(dirblock[0], pformat(dirblock)))
2216
def check_valid_parent():
2217
"""Check that the current entry has a valid parent.
2219
This makes sure that the parent has a record,
2220
and that the parent isn't marked as "absent" in the
2221
current tree. (It is invalid to have a non-absent file in an absent
2224
if entry[0][0:2] == ('', ''):
2225
# There should be no parent for the root row
2227
parent_entry = self._get_entry(tree_index, path_utf8=entry[0][0])
2228
if parent_entry == (None, None):
2229
raise AssertionError(
2230
"no parent entry for: %s in tree %s"
2231
% (this_path, tree_index))
2232
if parent_entry[1][tree_index][0] != 'd':
2233
raise AssertionError(
2234
"Parent entry for %s is not marked as a valid"
2235
" directory. %s" % (this_path, parent_entry,))
2237
# For each file id, for each tree: either
2238
# the file id is not present at all; all rows with that id in the
2239
# key have it marked as 'absent'
2240
# OR the file id is present under exactly one name; any other entries
2241
# that mention that id point to the correct name.
2243
# We check this with a dict per tree pointing either to the present
2244
# name, or None if absent.
2245
tree_count = self._num_present_parents() + 1
2246
id_path_maps = [dict() for i in range(tree_count)]
2247
# Make sure that all renamed entries point to the correct location.
2248
for entry in self._iter_entries():
2249
file_id = entry[0][2]
2250
this_path = osutils.pathjoin(entry[0][0], entry[0][1])
2251
if len(entry[1]) != tree_count:
2252
raise AssertionError(
2253
"wrong number of entry details for row\n%s" \
2254
",\nexpected %d" % \
2255
(pformat(entry), tree_count))
2256
for tree_index, tree_state in enumerate(entry[1]):
2257
this_tree_map = id_path_maps[tree_index]
2258
minikind = tree_state[0]
2259
# have we seen this id before in this column?
2260
if file_id in this_tree_map:
2261
previous_path = this_tree_map[file_id]
2262
# any later mention of this file must be consistent with
2263
# what was said before
2265
if previous_path is not None:
2266
raise AssertionError(
2267
"file %s is absent in row %r but also present " \
2269
(file_id, entry, previous_path))
2270
elif minikind == 'r':
2271
target_location = tree_state[1]
2272
if previous_path != target_location:
2273
raise AssertionError(
2274
"file %s relocation in row %r but also at %r" \
2275
% (file_id, entry, previous_path))
2277
# a file, directory, etc - may have been previously
2278
# pointed to by a relocation, which must point here
2279
if previous_path != this_path:
2280
raise AssertionError(
2281
"entry %r inconsistent with previous path %r" % \
2282
(entry, previous_path))
2283
check_valid_parent()
2286
# absent; should not occur anywhere else
2287
this_tree_map[file_id] = None
2288
elif minikind == 'r':
2289
# relocation, must occur at expected location
2290
this_tree_map[file_id] = tree_state[1]
2292
this_tree_map[file_id] = this_path
2293
check_valid_parent()
2295
def _wipe_state(self):
2296
"""Forget all state information about the dirstate."""
2297
self._header_state = DirState.NOT_IN_MEMORY
2298
self._dirblock_state = DirState.NOT_IN_MEMORY
2301
self._dirblocks = []
2302
self._id_index = None
2303
self._end_of_header = None
2304
self._cutoff_time = None
2305
self._split_path_cache = {}
2307
def lock_read(self):
2308
"""Acquire a read lock on the dirstate"""
2309
if self._lock_token is not None:
2310
raise errors.LockContention(self._lock_token)
2311
# TODO: jam 20070301 Rather than wiping completely, if the blocks are
2312
# already in memory, we could read just the header and check for
2313
# any modification. If not modified, we can just leave things
2315
self._lock_token = lock.ReadLock(self._filename)
2316
self._lock_state = 'r'
2317
self._state_file = self._lock_token.f
2320
def lock_write(self):
2321
"""Acquire a write lock on the dirstate"""
2322
if self._lock_token is not None:
2323
raise errors.LockContention(self._lock_token)
2324
# TODO: jam 20070301 Rather than wiping completely, if the blocks are
2325
# already in memory, we could read just the header and check for
2326
# any modification. If not modified, we can just leave things
2328
self._lock_token = lock.WriteLock(self._filename)
2329
self._lock_state = 'w'
2330
self._state_file = self._lock_token.f
2334
"""Drop any locks held on the dirstate"""
2335
if self._lock_token is None:
2336
raise errors.LockNotHeld(self)
2337
# TODO: jam 20070301 Rather than wiping completely, if the blocks are
2338
# already in memory, we could read just the header and check for
2339
# any modification. If not modified, we can just leave things
2341
self._state_file = None
2342
self._lock_state = None
2343
self._lock_token.unlock()
2344
self._lock_token = None
2345
self._split_path_cache = {}
2347
def _requires_lock(self):
2348
"""Checks that a lock is currently held by someone on the dirstate"""
2349
if not self._lock_token:
2350
raise errors.ObjectNotLocked(self)
2353
def bisect_dirblock(dirblocks, dirname, lo=0, hi=None, cache={}):
2354
"""Return the index where to insert dirname into the dirblocks.
2356
The return value idx is such that all directories blocks in dirblock[:idx]
2357
have names < dirname, and all blocks in dirblock[idx:] have names >=
2360
Optional args lo (default 0) and hi (default len(dirblocks)) bound the
2361
slice of a to be searched.
2366
dirname_split = cache[dirname]
2368
dirname_split = dirname.split('/')
2369
cache[dirname] = dirname_split
2372
# Grab the dirname for the current dirblock
2373
cur = dirblocks[mid][0]
2375
cur_split = cache[cur]
2377
cur_split = cur.split('/')
2378
cache[cur] = cur_split
2379
if cur_split < dirname_split: lo = mid+1
2385
def pack_stat(st, _encode=base64.encodestring, _pack=struct.pack):
2386
"""Convert stat values into a packed representation."""
2387
# jam 20060614 it isn't really worth removing more entries if we
2388
# are going to leave it in packed form.
2389
# With only st_mtime and st_mode filesize is 5.5M and read time is 275ms
2390
# With all entries filesize is 5.9M and read time is mabye 280ms
2391
# well within the noise margin
2393
# base64.encode always adds a final newline, so strip it off
2394
return _encode(_pack('>LLLLLL'
2395
, st.st_size, int(st.st_mtime), int(st.st_ctime)
2396
, st.st_dev, st.st_ino & 0xFFFFFFFF, st.st_mode))[:-1]
added
removed
bzrlib/dirstate.py
