1
# Copyright (C) 2006-2011 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
"""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.
25
MINIKIND = "f" | "d" | "l" | "a" | "r" | "t";
28
WHOLE_NUMBER = {digit}, digit;
30
REVISION_ID = a non-empty utf8 string;
32
dirstate format = header line, full checksum, row count, parent details,
33
ghost_details, entries;
34
header line = "#bazaar dirstate flat format 3", NL;
35
full checksum = "crc32: ", ["-"], WHOLE_NUMBER, NL;
36
row count = "num_entries: ", WHOLE_NUMBER, NL;
37
parent_details = WHOLE NUMBER, {REVISION_ID}* NL;
38
ghost_details = WHOLE NUMBER, {REVISION_ID}*, NL;
40
entry = entry_key, current_entry_details, {parent_entry_details};
41
entry_key = dirname, basename, fileid;
42
current_entry_details = common_entry_details, working_entry_details;
43
parent_entry_details = common_entry_details, history_entry_details;
44
common_entry_details = MINIKIND, fingerprint, size, executable
45
working_entry_details = packed_stat
46
history_entry_details = REVISION_ID;
49
fingerprint = a nonempty utf8 sequence with meaning defined by minikind.
51
Given this definition, the following is useful to know::
53
entry (aka row) - all the data for a given key.
54
entry[0]: The key (dirname, basename, fileid)
58
entry[1]: The tree(s) data for this path and id combination.
59
entry[1][0]: The current tree
60
entry[1][1]: The second tree
62
For an entry for a tree, we have (using tree 0 - current tree) to demonstrate::
64
entry[1][0][0]: minikind
65
entry[1][0][1]: fingerprint
67
entry[1][0][3]: executable
68
entry[1][0][4]: packed_stat
72
entry[1][1][4]: revision_id
74
There may be multiple rows at the root, one per id present in the root, so the
75
in memory root row is now::
77
self._dirblocks[0] -> ('', [entry ...]),
79
and the entries in there are::
83
entries[0][2]: file_id
84
entries[1][0]: The tree data for the current tree for this fileid at /
89
'r' is a relocated entry: This path is not present in this tree with this
90
id, but the id can be found at another location. The fingerprint is
91
used to point to the target location.
92
'a' is an absent entry: In that tree the id is not present at this path.
93
'd' is a directory entry: This path in this tree is a directory with the
94
current file id. There is no fingerprint for directories.
95
'f' is a file entry: As for directory, but it's a file. The fingerprint is
96
the sha1 value of the file's canonical form, i.e. after any read
97
filters have been applied to the convenience form stored in the working
99
'l' is a symlink entry: As for directory, but a symlink. The fingerprint is
101
't' is a reference to a nested subtree; the fingerprint is the referenced
106
The entries on disk and in memory are ordered according to the following keys::
108
directory, as a list of components
112
--- Format 1 had the following different definition: ---
116
rows = dirname, NULL, basename, NULL, MINIKIND, NULL, fileid_utf8, NULL,
117
WHOLE NUMBER (* size *), NULL, packed stat, NULL, sha1|symlink target,
119
PARENT ROW = NULL, revision_utf8, NULL, MINIKIND, NULL, dirname, NULL,
120
basename, NULL, WHOLE NUMBER (* size *), NULL, "y" | "n", NULL,
123
PARENT ROW's are emitted for every parent that is not in the ghosts details
124
line. That is, if the parents are foo, bar, baz, and the ghosts are bar, then
125
each row will have a PARENT ROW for foo and baz, but not for bar.
128
In any tree, a kind of 'moved' indicates that the fingerprint field
129
(which we treat as opaque data specific to the 'kind' anyway) has the
130
details for the id of this row in that tree.
132
I'm strongly tempted to add a id->path index as well, but I think that
133
where we need id->path mapping; we also usually read the whole file, so
134
I'm going to skip that for the moment, as we have the ability to locate
135
via bisect any path in any tree, and if we lookup things by path, we can
136
accumulate an id->path mapping as we go, which will tend to match what we
139
I plan to implement this asap, so please speak up now to alter/tweak the
140
design - and once we stabilise on this, I'll update the wiki page for
143
The rationale for all this is that we want fast operations for the
144
common case (diff/status/commit/merge on all files) and extremely fast
145
operations for the less common but still occurs a lot status/diff/commit
146
on specific files). Operations on specific files involve a scan for all
147
the children of a path, *in every involved tree*, which the current
148
format did not accommodate.
152
1. Fast end to end use for bzr's top 5 uses cases. (commmit/diff/status/merge/???)
153
2. fall back current object model as needed.
154
3. scale usably to the largest trees known today - say 50K entries. (mozilla
155
is an example of this)
160
Eventually reuse dirstate objects across locks IFF the dirstate file has not
161
been modified, but will require that we flush/ignore cached stat-hit data
162
because we won't want to restat all files on disk just because a lock was
163
acquired, yet we cannot trust the data after the previous lock was released.
165
Memory representation::
167
vector of all directories, and vector of the childen ?
169
root_entrie = (direntry for root, [parent_direntries_for_root]),
171
('', ['data for achild', 'data for bchild', 'data for cchild'])
172
('dir', ['achild', 'cchild', 'echild'])
174
- single bisect to find N subtrees from a path spec
175
- in-order for serialisation - this is 'dirblock' grouping.
176
- insertion of a file '/a' affects only the '/' child-vector, that is, to
177
insert 10K elements from scratch does not generates O(N^2) memoves of a
178
single vector, rather each individual, which tends to be limited to a
179
manageable number. Will scale badly on trees with 10K entries in a
180
single directory. compare with Inventory.InventoryDirectory which has
181
a dictionary for the children. No bisect capability, can only probe for
182
exact matches, or grab all elements and sort.
183
- What's the risk of error here? Once we have the base format being processed
184
we should have a net win regardless of optimality. So we are going to
185
go with what seems reasonable.
189
Maybe we should do a test profile of the core structure - 10K simulated
190
searches/lookups/etc?
192
Objects for each row?
193
The lifetime of Dirstate objects is current per lock, but see above for
194
possible extensions. The lifetime of a row from a dirstate is expected to be
195
very short in the optimistic case: which we are optimising for. For instance,
196
subtree status will determine from analysis of the disk data what rows need to
197
be examined at all, and will be able to determine from a single row whether
198
that file has altered or not, so we are aiming to process tens of thousands of
199
entries each second within the dirstate context, before exposing anything to
200
the larger codebase. This suggests we want the time for a single file
201
comparison to be < 0.1 milliseconds. That would give us 10000 paths per second
202
processed, and to scale to 100 thousand we'll another order of magnitude to do
203
that. Now, as the lifetime for all unchanged entries is the time to parse, stat
204
the file on disk, and then immediately discard, the overhead of object creation
205
becomes a significant cost.
207
Figures: Creating a tuple from 3 elements was profiled at 0.0625
208
microseconds, whereas creating a object which is subclassed from tuple was
209
0.500 microseconds, and creating an object with 3 elements and slots was 3
210
microseconds long. 0.1 milliseconds is 100 microseconds, and ideally we'll get
211
down to 10 microseconds for the total processing - having 33% of that be object
212
creation is a huge overhead. There is a potential cost in using tuples within
213
each row which is that the conditional code to do comparisons may be slower
214
than method invocation, but method invocation is known to be slow due to stack
215
frame creation, so avoiding methods in these tight inner loops in unfortunately
216
desirable. We can consider a pyrex version of this with objects in future if
221
from __future__ import absolute_import
227
from stat import S_IEXEC
247
# This is the Windows equivalent of ENOTDIR
248
# It is defined in pywin32.winerror, but we don't want a strong dependency for
249
# just an error code.
250
ERROR_PATH_NOT_FOUND = 3
251
ERROR_DIRECTORY = 267
254
class SHA1Provider(object):
255
"""An interface for getting sha1s of a file."""
257
def sha1(self, abspath):
258
"""Return the sha1 of a file given its absolute path.
260
:param abspath: May be a filesystem encoded absolute path
263
raise NotImplementedError(self.sha1)
265
def stat_and_sha1(self, abspath):
266
"""Return the stat and sha1 of a file given its absolute path.
268
:param abspath: May be a filesystem encoded absolute path
271
Note: the stat should be the stat of the physical file
272
while the sha may be the sha of its canonical content.
274
raise NotImplementedError(self.stat_and_sha1)
277
class DefaultSHA1Provider(SHA1Provider):
278
"""A SHA1Provider that reads directly from the filesystem."""
280
def sha1(self, abspath):
281
"""Return the sha1 of a file given its absolute path."""
282
return osutils.sha_file_by_name(abspath)
284
def stat_and_sha1(self, abspath):
285
"""Return the stat and sha1 of a file given its absolute path."""
286
file_obj = file(abspath, 'rb')
288
statvalue = os.fstat(file_obj.fileno())
289
sha1 = osutils.sha_file(file_obj)
292
return statvalue, sha1
295
class DirState(object):
296
"""Record directory and metadata state for fast access.
298
A dirstate is a specialised data structure for managing local working
299
tree state information. Its not yet well defined whether it is platform
300
specific, and if it is how we detect/parameterize that.
302
Dirstates use the usual lock_write, lock_read and unlock mechanisms.
303
Unlike most bzr disk formats, DirStates must be locked for reading, using
304
lock_read. (This is an os file lock internally.) This is necessary
305
because the file can be rewritten in place.
307
DirStates must be explicitly written with save() to commit changes; just
308
unlocking them does not write the changes to disk.
311
_kind_to_minikind = {
317
'tree-reference': 't',
319
_minikind_to_kind = {
325
't': 'tree-reference',
327
_stat_to_minikind = {
332
_to_yesno = {True:'y', False: 'n'} # TODO profile the performance gain
333
# of using int conversion rather than a dict here. AND BLAME ANDREW IF
336
# TODO: jam 20070221 Figure out what to do if we have a record that exceeds
337
# the BISECT_PAGE_SIZE. For now, we just have to make it large enough
338
# that we are sure a single record will always fit.
339
BISECT_PAGE_SIZE = 4096
342
IN_MEMORY_UNMODIFIED = 1
343
IN_MEMORY_MODIFIED = 2
344
IN_MEMORY_HASH_MODIFIED = 3 # Only hash-cache updates
346
# A pack_stat (the x's) that is just noise and will never match the output
349
NULL_PARENT_DETAILS = static_tuple.StaticTuple('a', '', 0, False, '')
351
HEADER_FORMAT_2 = '#bazaar dirstate flat format 2\n'
352
HEADER_FORMAT_3 = '#bazaar dirstate flat format 3\n'
354
def __init__(self, path, sha1_provider, worth_saving_limit=0):
355
"""Create a DirState object.
357
:param path: The path at which the dirstate file on disk should live.
358
:param sha1_provider: an object meeting the SHA1Provider interface.
359
:param worth_saving_limit: when the exact number of hash changed
360
entries is known, only bother saving the dirstate if more than
361
this count of entries have changed.
362
-1 means never save hash changes, 0 means always save hash changes.
364
# _header_state and _dirblock_state represent the current state
365
# of the dirstate metadata and the per-row data respectiely.
366
# NOT_IN_MEMORY indicates that no data is in memory
367
# IN_MEMORY_UNMODIFIED indicates that what we have in memory
368
# is the same as is on disk
369
# IN_MEMORY_MODIFIED indicates that we have a modified version
370
# of what is on disk.
371
# In future we will add more granularity, for instance _dirblock_state
372
# will probably support partially-in-memory as a separate variable,
373
# allowing for partially-in-memory unmodified and partially-in-memory
375
self._header_state = DirState.NOT_IN_MEMORY
376
self._dirblock_state = DirState.NOT_IN_MEMORY
377
# If true, an error has been detected while updating the dirstate, and
378
# for safety we're not going to commit to disk.
379
self._changes_aborted = False
383
self._state_file = None
384
self._filename = path
385
self._lock_token = None
386
self._lock_state = None
387
self._id_index = None
388
# a map from packed_stat to sha's.
389
self._packed_stat_index = None
390
self._end_of_header = None
391
self._cutoff_time = None
392
self._split_path_cache = {}
393
self._bisect_page_size = DirState.BISECT_PAGE_SIZE
394
self._sha1_provider = sha1_provider
395
if 'hashcache' in debug.debug_flags:
396
self._sha1_file = self._sha1_file_and_mutter
398
self._sha1_file = self._sha1_provider.sha1
399
# These two attributes provide a simple cache for lookups into the
400
# dirstate in-memory vectors. By probing respectively for the last
401
# block, and for the next entry, we save nearly 2 bisections per path
403
self._last_block_index = None
404
self._last_entry_index = None
405
# The set of known hash changes
406
self._known_hash_changes = set()
407
# How many hash changed entries can we have without saving
408
self._worth_saving_limit = worth_saving_limit
409
self._config_stack = config.LocationStack(urlutils.local_path_to_url(
414
(self.__class__.__name__, self._filename)
416
def _mark_modified(self, hash_changed_entries=None, header_modified=False):
417
"""Mark this dirstate as modified.
419
:param hash_changed_entries: if non-None, mark just these entries as
420
having their hash modified.
421
:param header_modified: mark the header modified as well, not just the
424
#trace.mutter_callsite(3, "modified hash entries: %s", hash_changed_entries)
425
if hash_changed_entries:
426
self._known_hash_changes.update([e[0] for e in hash_changed_entries])
427
if self._dirblock_state in (DirState.NOT_IN_MEMORY,
428
DirState.IN_MEMORY_UNMODIFIED):
429
# If the dirstate is already marked a IN_MEMORY_MODIFIED, then
430
# that takes precedence.
431
self._dirblock_state = DirState.IN_MEMORY_HASH_MODIFIED
433
# TODO: Since we now have a IN_MEMORY_HASH_MODIFIED state, we
434
# should fail noisily if someone tries to set
435
# IN_MEMORY_MODIFIED but we don't have a write-lock!
436
# We don't know exactly what changed so disable smart saving
437
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
439
self._header_state = DirState.IN_MEMORY_MODIFIED
441
def _mark_unmodified(self):
442
"""Mark this dirstate as unmodified."""
443
self._header_state = DirState.IN_MEMORY_UNMODIFIED
444
self._dirblock_state = DirState.IN_MEMORY_UNMODIFIED
445
self._known_hash_changes = set()
447
def add(self, path, file_id, kind, stat, fingerprint):
448
"""Add a path to be tracked.
450
:param path: The path within the dirstate - '' is the root, 'foo' is the
451
path foo within the root, 'foo/bar' is the path bar within foo
453
:param file_id: The file id of the path being added.
454
:param kind: The kind of the path, as a string like 'file',
456
:param stat: The output of os.lstat for the path.
457
:param fingerprint: The sha value of the file's canonical form (i.e.
458
after any read filters have been applied),
459
or the target of a symlink,
460
or the referenced revision id for tree-references,
461
or '' for directories.
464
# find the block its in.
465
# find the location in the block.
466
# check its not there
468
#------- copied from inventory.ensure_normalized_name - keep synced.
469
# --- normalized_filename wants a unicode basename only, so get one.
470
dirname, basename = osutils.split(path)
471
# we dont import normalized_filename directly because we want to be
472
# able to change the implementation at runtime for tests.
473
norm_name, can_access = osutils.normalized_filename(basename)
474
if norm_name != basename:
478
raise errors.InvalidNormalization(path)
479
# you should never have files called . or ..; just add the directory
480
# in the parent, or according to the special treatment for the root
481
if basename == '.' or basename == '..':
482
raise errors.InvalidEntryName(path)
483
# now that we've normalised, we need the correct utf8 path and
484
# dirname and basename elements. This single encode and split should be
485
# faster than three separate encodes.
486
utf8path = (dirname + '/' + basename).strip('/').encode('utf8')
487
dirname, basename = osutils.split(utf8path)
488
# uses __class__ for speed; the check is needed for safety
489
if file_id.__class__ is not str:
490
raise AssertionError(
491
"must be a utf8 file_id not %s" % (type(file_id), ))
492
# Make sure the file_id does not exist in this tree
494
file_id_entry = self._get_entry(0, fileid_utf8=file_id, include_deleted=True)
495
if file_id_entry != (None, None):
496
if file_id_entry[1][0][0] == 'a':
497
if file_id_entry[0] != (dirname, basename, file_id):
498
# set the old name's current operation to rename
499
self.update_minimal(file_id_entry[0],
505
rename_from = file_id_entry[0][0:2]
507
path = osutils.pathjoin(file_id_entry[0][0], file_id_entry[0][1])
508
kind = DirState._minikind_to_kind[file_id_entry[1][0][0]]
509
info = '%s:%s' % (kind, path)
510
raise errors.DuplicateFileId(file_id, info)
511
first_key = (dirname, basename, '')
512
block_index, present = self._find_block_index_from_key(first_key)
514
# check the path is not in the tree
515
block = self._dirblocks[block_index][1]
516
entry_index, _ = self._find_entry_index(first_key, block)
517
while (entry_index < len(block) and
518
block[entry_index][0][0:2] == first_key[0:2]):
519
if block[entry_index][1][0][0] not in 'ar':
520
# this path is in the dirstate in the current tree.
521
raise Exception, "adding already added path!"
524
# The block where we want to put the file is not present. But it
525
# might be because the directory was empty, or not loaded yet. Look
526
# for a parent entry, if not found, raise NotVersionedError
527
parent_dir, parent_base = osutils.split(dirname)
528
parent_block_idx, parent_entry_idx, _, parent_present = \
529
self._get_block_entry_index(parent_dir, parent_base, 0)
530
if not parent_present:
531
raise errors.NotVersionedError(path, str(self))
532
self._ensure_block(parent_block_idx, parent_entry_idx, dirname)
533
block = self._dirblocks[block_index][1]
534
entry_key = (dirname, basename, file_id)
537
packed_stat = DirState.NULLSTAT
540
packed_stat = pack_stat(stat)
541
parent_info = self._empty_parent_info()
542
minikind = DirState._kind_to_minikind[kind]
543
if rename_from is not None:
545
old_path_utf8 = '%s/%s' % rename_from
547
old_path_utf8 = rename_from[1]
548
parent_info[0] = ('r', old_path_utf8, 0, False, '')
550
entry_data = entry_key, [
551
(minikind, fingerprint, size, False, packed_stat),
553
elif kind == 'directory':
554
entry_data = entry_key, [
555
(minikind, '', 0, False, packed_stat),
557
elif kind == 'symlink':
558
entry_data = entry_key, [
559
(minikind, fingerprint, size, False, packed_stat),
561
elif kind == 'tree-reference':
562
entry_data = entry_key, [
563
(minikind, fingerprint, 0, False, packed_stat),
566
raise errors.BzrError('unknown kind %r' % kind)
567
entry_index, present = self._find_entry_index(entry_key, block)
569
block.insert(entry_index, entry_data)
571
if block[entry_index][1][0][0] != 'a':
572
raise AssertionError(" %r(%r) already added" % (basename, file_id))
573
block[entry_index][1][0] = entry_data[1][0]
575
if kind == 'directory':
576
# insert a new dirblock
577
self._ensure_block(block_index, entry_index, utf8path)
578
self._mark_modified()
580
self._add_to_id_index(self._id_index, entry_key)
582
def _bisect(self, paths):
583
"""Bisect through the disk structure for specific rows.
585
:param paths: A list of paths to find
586
:return: A dict mapping path => entries for found entries. Missing
587
entries will not be in the map.
588
The list is not sorted, and entries will be populated
589
based on when they were read.
591
self._requires_lock()
592
# We need the file pointer to be right after the initial header block
593
self._read_header_if_needed()
594
# If _dirblock_state was in memory, we should just return info from
595
# there, this function is only meant to handle when we want to read
597
if self._dirblock_state != DirState.NOT_IN_MEMORY:
598
raise AssertionError("bad dirblock state %r" % self._dirblock_state)
600
# The disk representation is generally info + '\0\n\0' at the end. But
601
# for bisecting, it is easier to treat this as '\0' + info + '\0\n'
602
# Because it means we can sync on the '\n'
603
state_file = self._state_file
604
file_size = os.fstat(state_file.fileno()).st_size
605
# We end up with 2 extra fields, we should have a trailing '\n' to
606
# ensure that we read the whole record, and we should have a precursur
607
# '' which ensures that we start after the previous '\n'
608
entry_field_count = self._fields_per_entry() + 1
610
low = self._end_of_header
611
high = file_size - 1 # Ignore the final '\0'
612
# Map from (dir, name) => entry
615
# Avoid infinite seeking
616
max_count = 30*len(paths)
618
# pending is a list of places to look.
619
# each entry is a tuple of low, high, dir_names
620
# low -> the first byte offset to read (inclusive)
621
# high -> the last byte offset (inclusive)
622
# dir_names -> The list of (dir, name) pairs that should be found in
623
# the [low, high] range
624
pending = [(low, high, paths)]
626
page_size = self._bisect_page_size
628
fields_to_entry = self._get_fields_to_entry()
631
low, high, cur_files = pending.pop()
633
if not cur_files or low >= high:
638
if count > max_count:
639
raise errors.BzrError('Too many seeks, most likely a bug.')
641
mid = max(low, (low+high-page_size)/2)
644
# limit the read size, so we don't end up reading data that we have
646
read_size = min(page_size, (high-mid)+1)
647
block = state_file.read(read_size)
650
entries = block.split('\n')
653
# We didn't find a '\n', so we cannot have found any records.
654
# So put this range back and try again. But we know we have to
655
# increase the page size, because a single read did not contain
656
# a record break (so records must be larger than page_size)
658
pending.append((low, high, cur_files))
661
# Check the first and last entries, in case they are partial, or if
662
# we don't care about the rest of this page
664
first_fields = entries[0].split('\0')
665
if len(first_fields) < entry_field_count:
666
# We didn't get the complete first entry
667
# so move start, and grab the next, which
668
# should be a full entry
669
start += len(entries[0])+1
670
first_fields = entries[1].split('\0')
673
if len(first_fields) <= 2:
674
# We didn't even get a filename here... what do we do?
675
# Try a large page size and repeat this query
677
pending.append((low, high, cur_files))
680
# Find what entries we are looking for, which occur before and
681
# after this first record.
684
first_path = first_fields[1] + '/' + first_fields[2]
686
first_path = first_fields[2]
687
first_loc = _bisect_path_left(cur_files, first_path)
689
# These exist before the current location
690
pre = cur_files[:first_loc]
691
# These occur after the current location, which may be in the
692
# data we read, or might be after the last entry
693
post = cur_files[first_loc:]
695
if post and len(first_fields) >= entry_field_count:
696
# We have files after the first entry
698
# Parse the last entry
699
last_entry_num = len(entries)-1
700
last_fields = entries[last_entry_num].split('\0')
701
if len(last_fields) < entry_field_count:
702
# The very last hunk was not complete,
703
# read the previous hunk
704
after = mid + len(block) - len(entries[-1])
706
last_fields = entries[last_entry_num].split('\0')
708
after = mid + len(block)
711
last_path = last_fields[1] + '/' + last_fields[2]
713
last_path = last_fields[2]
714
last_loc = _bisect_path_right(post, last_path)
716
middle_files = post[:last_loc]
717
post = post[last_loc:]
720
# We have files that should occur in this block
721
# (>= first, <= last)
722
# Either we will find them here, or we can mark them as
725
if middle_files[0] == first_path:
726
# We might need to go before this location
727
pre.append(first_path)
728
if middle_files[-1] == last_path:
729
post.insert(0, last_path)
731
# Find out what paths we have
732
paths = {first_path:[first_fields]}
733
# last_path might == first_path so we need to be
734
# careful if we should append rather than overwrite
735
if last_entry_num != first_entry_num:
736
paths.setdefault(last_path, []).append(last_fields)
737
for num in xrange(first_entry_num+1, last_entry_num):
738
# TODO: jam 20070223 We are already splitting here, so
739
# shouldn't we just split the whole thing rather
740
# than doing the split again in add_one_record?
741
fields = entries[num].split('\0')
743
path = fields[1] + '/' + fields[2]
746
paths.setdefault(path, []).append(fields)
748
for path in middle_files:
749
for fields in paths.get(path, []):
750
# offset by 1 because of the opening '\0'
751
# consider changing fields_to_entry to avoid the
753
entry = fields_to_entry(fields[1:])
754
found.setdefault(path, []).append(entry)
756
# Now we have split up everything into pre, middle, and post, and
757
# we have handled everything that fell in 'middle'.
758
# We add 'post' first, so that we prefer to seek towards the
759
# beginning, so that we will tend to go as early as we need, and
760
# then only seek forward after that.
762
pending.append((after, high, post))
764
pending.append((low, start-1, pre))
766
# Consider that we may want to return the directory entries in sorted
767
# order. For now, we just return them in whatever order we found them,
768
# and leave it up to the caller if they care if it is ordered or not.
771
def _bisect_dirblocks(self, dir_list):
772
"""Bisect through the disk structure to find entries in given dirs.
774
_bisect_dirblocks is meant to find the contents of directories, which
775
differs from _bisect, which only finds individual entries.
777
:param dir_list: A sorted list of directory names ['', 'dir', 'foo'].
778
:return: A map from dir => entries_for_dir
780
# TODO: jam 20070223 A lot of the bisecting logic could be shared
781
# between this and _bisect. It would require parameterizing the
782
# inner loop with a function, though. We should evaluate the
783
# performance difference.
784
self._requires_lock()
785
# We need the file pointer to be right after the initial header block
786
self._read_header_if_needed()
787
# If _dirblock_state was in memory, we should just return info from
788
# there, this function is only meant to handle when we want to read
790
if self._dirblock_state != DirState.NOT_IN_MEMORY:
791
raise AssertionError("bad dirblock state %r" % self._dirblock_state)
792
# The disk representation is generally info + '\0\n\0' at the end. But
793
# for bisecting, it is easier to treat this as '\0' + info + '\0\n'
794
# Because it means we can sync on the '\n'
795
state_file = self._state_file
796
file_size = os.fstat(state_file.fileno()).st_size
797
# We end up with 2 extra fields, we should have a trailing '\n' to
798
# ensure that we read the whole record, and we should have a precursur
799
# '' which ensures that we start after the previous '\n'
800
entry_field_count = self._fields_per_entry() + 1
802
low = self._end_of_header
803
high = file_size - 1 # Ignore the final '\0'
804
# Map from dir => entry
807
# Avoid infinite seeking
808
max_count = 30*len(dir_list)
810
# pending is a list of places to look.
811
# each entry is a tuple of low, high, dir_names
812
# low -> the first byte offset to read (inclusive)
813
# high -> the last byte offset (inclusive)
814
# dirs -> The list of directories that should be found in
815
# the [low, high] range
816
pending = [(low, high, dir_list)]
818
page_size = self._bisect_page_size
820
fields_to_entry = self._get_fields_to_entry()
823
low, high, cur_dirs = pending.pop()
825
if not cur_dirs or low >= high:
830
if count > max_count:
831
raise errors.BzrError('Too many seeks, most likely a bug.')
833
mid = max(low, (low+high-page_size)/2)
836
# limit the read size, so we don't end up reading data that we have
838
read_size = min(page_size, (high-mid)+1)
839
block = state_file.read(read_size)
842
entries = block.split('\n')
845
# We didn't find a '\n', so we cannot have found any records.
846
# So put this range back and try again. But we know we have to
847
# increase the page size, because a single read did not contain
848
# a record break (so records must be larger than page_size)
850
pending.append((low, high, cur_dirs))
853
# Check the first and last entries, in case they are partial, or if
854
# we don't care about the rest of this page
856
first_fields = entries[0].split('\0')
857
if len(first_fields) < entry_field_count:
858
# We didn't get the complete first entry
859
# so move start, and grab the next, which
860
# should be a full entry
861
start += len(entries[0])+1
862
first_fields = entries[1].split('\0')
865
if len(first_fields) <= 1:
866
# We didn't even get a dirname here... what do we do?
867
# Try a large page size and repeat this query
869
pending.append((low, high, cur_dirs))
872
# Find what entries we are looking for, which occur before and
873
# after this first record.
875
first_dir = first_fields[1]
876
first_loc = bisect.bisect_left(cur_dirs, first_dir)
878
# These exist before the current location
879
pre = cur_dirs[:first_loc]
880
# These occur after the current location, which may be in the
881
# data we read, or might be after the last entry
882
post = cur_dirs[first_loc:]
884
if post and len(first_fields) >= entry_field_count:
885
# We have records to look at after the first entry
887
# Parse the last entry
888
last_entry_num = len(entries)-1
889
last_fields = entries[last_entry_num].split('\0')
890
if len(last_fields) < entry_field_count:
891
# The very last hunk was not complete,
892
# read the previous hunk
893
after = mid + len(block) - len(entries[-1])
895
last_fields = entries[last_entry_num].split('\0')
897
after = mid + len(block)
899
last_dir = last_fields[1]
900
last_loc = bisect.bisect_right(post, last_dir)
902
middle_files = post[:last_loc]
903
post = post[last_loc:]
906
# We have files that should occur in this block
907
# (>= first, <= last)
908
# Either we will find them here, or we can mark them as
911
if middle_files[0] == first_dir:
912
# We might need to go before this location
913
pre.append(first_dir)
914
if middle_files[-1] == last_dir:
915
post.insert(0, last_dir)
917
# Find out what paths we have
918
paths = {first_dir:[first_fields]}
919
# last_dir might == first_dir so we need to be
920
# careful if we should append rather than overwrite
921
if last_entry_num != first_entry_num:
922
paths.setdefault(last_dir, []).append(last_fields)
923
for num in xrange(first_entry_num+1, last_entry_num):
924
# TODO: jam 20070223 We are already splitting here, so
925
# shouldn't we just split the whole thing rather
926
# than doing the split again in add_one_record?
927
fields = entries[num].split('\0')
928
paths.setdefault(fields[1], []).append(fields)
930
for cur_dir in middle_files:
931
for fields in paths.get(cur_dir, []):
932
# offset by 1 because of the opening '\0'
933
# consider changing fields_to_entry to avoid the
935
entry = fields_to_entry(fields[1:])
936
found.setdefault(cur_dir, []).append(entry)
938
# Now we have split up everything into pre, middle, and post, and
939
# we have handled everything that fell in 'middle'.
940
# We add 'post' first, so that we prefer to seek towards the
941
# beginning, so that we will tend to go as early as we need, and
942
# then only seek forward after that.
944
pending.append((after, high, post))
946
pending.append((low, start-1, pre))
950
def _bisect_recursive(self, paths):
951
"""Bisect for entries for all paths and their children.
953
This will use bisect to find all records for the supplied paths. It
954
will then continue to bisect for any records which are marked as
955
directories. (and renames?)
957
:param paths: A sorted list of (dir, name) pairs
958
eg: [('', 'a'), ('', 'f'), ('a/b', 'c')]
959
:return: A dictionary mapping (dir, name, file_id) => [tree_info]
961
# Map from (dir, name, file_id) => [tree_info]
964
found_dir_names = set()
966
# Directories that have been read
967
processed_dirs = set()
968
# Get the ball rolling with the first bisect for all entries.
969
newly_found = self._bisect(paths)
972
# Directories that need to be read
974
paths_to_search = set()
975
for entry_list in newly_found.itervalues():
976
for dir_name_id, trees_info in entry_list:
977
found[dir_name_id] = trees_info
978
found_dir_names.add(dir_name_id[:2])
980
for tree_info in trees_info:
981
minikind = tree_info[0]
984
# We already processed this one as a directory,
985
# we don't need to do the extra work again.
987
subdir, name, file_id = dir_name_id
988
path = osutils.pathjoin(subdir, name)
990
if path not in processed_dirs:
991
pending_dirs.add(path)
992
elif minikind == 'r':
993
# Rename, we need to directly search the target
994
# which is contained in the fingerprint column
995
dir_name = osutils.split(tree_info[1])
996
if dir_name[0] in pending_dirs:
997
# This entry will be found in the dir search
999
if dir_name not in found_dir_names:
1000
paths_to_search.add(tree_info[1])
1001
# Now we have a list of paths to look for directly, and
1002
# directory blocks that need to be read.
1003
# newly_found is mixing the keys between (dir, name) and path
1004
# entries, but that is okay, because we only really care about the
1006
newly_found = self._bisect(sorted(paths_to_search))
1007
newly_found.update(self._bisect_dirblocks(sorted(pending_dirs)))
1008
processed_dirs.update(pending_dirs)
1011
def _discard_merge_parents(self):
1012
"""Discard any parents trees beyond the first.
1014
Note that if this fails the dirstate is corrupted.
1016
After this function returns the dirstate contains 2 trees, neither of
1019
self._read_header_if_needed()
1020
parents = self.get_parent_ids()
1021
if len(parents) < 1:
1023
# only require all dirblocks if we are doing a full-pass removal.
1024
self._read_dirblocks_if_needed()
1025
dead_patterns = set([('a', 'r'), ('a', 'a'), ('r', 'r'), ('r', 'a')])
1026
def iter_entries_removable():
1027
for block in self._dirblocks:
1028
deleted_positions = []
1029
for pos, entry in enumerate(block[1]):
1031
if (entry[1][0][0], entry[1][1][0]) in dead_patterns:
1032
deleted_positions.append(pos)
1033
if deleted_positions:
1034
if len(deleted_positions) == len(block[1]):
1037
for pos in reversed(deleted_positions):
1039
# if the first parent is a ghost:
1040
if parents[0] in self.get_ghosts():
1041
empty_parent = [DirState.NULL_PARENT_DETAILS]
1042
for entry in iter_entries_removable():
1043
entry[1][1:] = empty_parent
1045
for entry in iter_entries_removable():
1049
self._parents = [parents[0]]
1050
self._mark_modified(header_modified=True)
1052
def _empty_parent_info(self):
1053
return [DirState.NULL_PARENT_DETAILS] * (len(self._parents) -
1056
def _ensure_block(self, parent_block_index, parent_row_index, dirname):
1057
"""Ensure a block for dirname exists.
1059
This function exists to let callers which know that there is a
1060
directory dirname ensure that the block for it exists. This block can
1061
fail to exist because of demand loading, or because a directory had no
1062
children. In either case it is not an error. It is however an error to
1063
call this if there is no parent entry for the directory, and thus the
1064
function requires the coordinates of such an entry to be provided.
1066
The root row is special cased and can be indicated with a parent block
1069
:param parent_block_index: The index of the block in which dirname's row
1071
:param parent_row_index: The index in the parent block where the row
1073
:param dirname: The utf8 dirname to ensure there is a block for.
1074
:return: The index for the block.
1076
if dirname == '' and parent_row_index == 0 and parent_block_index == 0:
1077
# This is the signature of the root row, and the
1078
# contents-of-root row is always index 1
1080
# the basename of the directory must be the end of its full name.
1081
if not (parent_block_index == -1 and
1082
parent_block_index == -1 and dirname == ''):
1083
if not dirname.endswith(
1084
self._dirblocks[parent_block_index][1][parent_row_index][0][1]):
1085
raise AssertionError("bad dirname %r" % dirname)
1086
block_index, present = self._find_block_index_from_key((dirname, '', ''))
1088
## In future, when doing partial parsing, this should load and
1089
# populate the entire block.
1090
self._dirblocks.insert(block_index, (dirname, []))
1093
def _entries_to_current_state(self, new_entries):
1094
"""Load new_entries into self.dirblocks.
1096
Process new_entries into the current state object, making them the active
1097
state. The entries are grouped together by directory to form dirblocks.
1099
:param new_entries: A sorted list of entries. This function does not sort
1100
to prevent unneeded overhead when callers have a sorted list already.
1103
if new_entries[0][0][0:2] != ('', ''):
1104
raise AssertionError(
1105
"Missing root row %r" % (new_entries[0][0],))
1106
# The two blocks here are deliberate: the root block and the
1107
# contents-of-root block.
1108
self._dirblocks = [('', []), ('', [])]
1109
current_block = self._dirblocks[0][1]
1110
current_dirname = ''
1112
append_entry = current_block.append
1113
for entry in new_entries:
1114
if entry[0][0] != current_dirname:
1115
# new block - different dirname
1117
current_dirname = entry[0][0]
1118
self._dirblocks.append((current_dirname, current_block))
1119
append_entry = current_block.append
1120
# append the entry to the current block
1122
self._split_root_dirblock_into_contents()
1124
def _split_root_dirblock_into_contents(self):
1125
"""Split the root dirblocks into root and contents-of-root.
1127
After parsing by path, we end up with root entries and contents-of-root
1128
entries in the same block. This loop splits them out again.
1130
# The above loop leaves the "root block" entries mixed with the
1131
# "contents-of-root block". But we don't want an if check on
1132
# all entries, so instead we just fix it up here.
1133
if self._dirblocks[1] != ('', []):
1134
raise ValueError("bad dirblock start %r" % (self._dirblocks[1],))
1136
contents_of_root_block = []
1137
for entry in self._dirblocks[0][1]:
1138
if not entry[0][1]: # This is a root entry
1139
root_block.append(entry)
1141
contents_of_root_block.append(entry)
1142
self._dirblocks[0] = ('', root_block)
1143
self._dirblocks[1] = ('', contents_of_root_block)
1145
def _entries_for_path(self, path):
1146
"""Return a list with all the entries that match path for all ids."""
1147
dirname, basename = os.path.split(path)
1148
key = (dirname, basename, '')
1149
block_index, present = self._find_block_index_from_key(key)
1151
# the block which should contain path is absent.
1154
block = self._dirblocks[block_index][1]
1155
entry_index, _ = self._find_entry_index(key, block)
1156
# we may need to look at multiple entries at this path: walk while the specific_files match.
1157
while (entry_index < len(block) and
1158
block[entry_index][0][0:2] == key[0:2]):
1159
result.append(block[entry_index])
1163
def _entry_to_line(self, entry):
1164
"""Serialize entry to a NULL delimited line ready for _get_output_lines.
1166
:param entry: An entry_tuple as defined in the module docstring.
1168
entire_entry = list(entry[0])
1169
for tree_number, tree_data in enumerate(entry[1]):
1170
# (minikind, fingerprint, size, executable, tree_specific_string)
1171
entire_entry.extend(tree_data)
1172
# 3 for the key, 5 for the fields per tree.
1173
tree_offset = 3 + tree_number * 5
1175
entire_entry[tree_offset + 0] = tree_data[0]
1177
entire_entry[tree_offset + 2] = str(tree_data[2])
1179
entire_entry[tree_offset + 3] = DirState._to_yesno[tree_data[3]]
1180
return '\0'.join(entire_entry)
1182
def _fields_per_entry(self):
1183
"""How many null separated fields should be in each entry row.
1185
Each line now has an extra '\\n' field which is not used
1186
so we just skip over it
1189
3 fields for the key
1190
+ number of fields per tree_data (5) * tree count
1193
tree_count = 1 + self._num_present_parents()
1194
return 3 + 5 * tree_count + 1
1196
def _find_block(self, key, add_if_missing=False):
1197
"""Return the block that key should be present in.
1199
:param key: A dirstate entry key.
1200
:return: The block tuple.
1202
block_index, present = self._find_block_index_from_key(key)
1204
if not add_if_missing:
1205
# check to see if key is versioned itself - we might want to
1206
# add it anyway, because dirs with no entries dont get a
1207
# dirblock at parse time.
1208
# This is an uncommon branch to take: most dirs have children,
1209
# and most code works with versioned paths.
1210
parent_base, parent_name = osutils.split(key[0])
1211
if not self._get_block_entry_index(parent_base, parent_name, 0)[3]:
1212
# some parent path has not been added - its an error to add
1214
raise errors.NotVersionedError(key[0:2], str(self))
1215
self._dirblocks.insert(block_index, (key[0], []))
1216
return self._dirblocks[block_index]
1218
def _find_block_index_from_key(self, key):
1219
"""Find the dirblock index for a key.
1221
:return: The block index, True if the block for the key is present.
1223
if key[0:2] == ('', ''):
1226
if (self._last_block_index is not None and
1227
self._dirblocks[self._last_block_index][0] == key[0]):
1228
return self._last_block_index, True
1231
block_index = bisect_dirblock(self._dirblocks, key[0], 1,
1232
cache=self._split_path_cache)
1233
# _right returns one-past-where-key is so we have to subtract
1234
# one to use it. we use _right here because there are two
1235
# '' blocks - the root, and the contents of root
1236
# we always have a minimum of 2 in self._dirblocks: root and
1237
# root-contents, and for '', we get 2 back, so this is
1238
# simple and correct:
1239
present = (block_index < len(self._dirblocks) and
1240
self._dirblocks[block_index][0] == key[0])
1241
self._last_block_index = block_index
1242
# Reset the entry index cache to the beginning of the block.
1243
self._last_entry_index = -1
1244
return block_index, present
1246
def _find_entry_index(self, key, block):
1247
"""Find the entry index for a key in a block.
1249
:return: The entry index, True if the entry for the key is present.
1251
len_block = len(block)
1253
if self._last_entry_index is not None:
1255
entry_index = self._last_entry_index + 1
1256
# A hit is when the key is after the last slot, and before or
1257
# equal to the next slot.
1258
if ((entry_index > 0 and block[entry_index - 1][0] < key) and
1259
key <= block[entry_index][0]):
1260
self._last_entry_index = entry_index
1261
present = (block[entry_index][0] == key)
1262
return entry_index, present
1265
entry_index = bisect.bisect_left(block, (key, []))
1266
present = (entry_index < len_block and
1267
block[entry_index][0] == key)
1268
self._last_entry_index = entry_index
1269
return entry_index, present
1272
def from_tree(tree, dir_state_filename, sha1_provider=None):
1273
"""Create a dirstate from a bzr Tree.
1275
:param tree: The tree which should provide parent information and
1277
:param sha1_provider: an object meeting the SHA1Provider interface.
1278
If None, a DefaultSHA1Provider is used.
1279
:return: a DirState object which is currently locked for writing.
1280
(it was locked by DirState.initialize)
1282
result = DirState.initialize(dir_state_filename,
1283
sha1_provider=sha1_provider)
1287
parent_ids = tree.get_parent_ids()
1288
num_parents = len(parent_ids)
1290
for parent_id in parent_ids:
1291
parent_tree = tree.branch.repository.revision_tree(parent_id)
1292
parent_trees.append((parent_id, parent_tree))
1293
parent_tree.lock_read()
1294
result.set_parent_trees(parent_trees, [])
1295
result.set_state_from_inventory(tree.root_inventory)
1297
for revid, parent_tree in parent_trees:
1298
parent_tree.unlock()
1301
# The caller won't have a chance to unlock this, so make sure we
1307
def _check_delta_is_valid(self, delta):
1308
return list(inventory._check_delta_unique_ids(
1309
inventory._check_delta_unique_old_paths(
1310
inventory._check_delta_unique_new_paths(
1311
inventory._check_delta_ids_match_entry(
1312
inventory._check_delta_ids_are_valid(
1313
inventory._check_delta_new_path_entry_both_or_None(delta)))))))
1315
def update_by_delta(self, delta):
1316
"""Apply an inventory delta to the dirstate for tree 0
1318
This is the workhorse for apply_inventory_delta in dirstate based
1321
:param delta: An inventory delta. See Inventory.apply_delta for
1324
self._read_dirblocks_if_needed()
1325
encode = cache_utf8.encode
1328
# Accumulate parent references (path_utf8, id), to check for parentless
1329
# items or items placed under files/links/tree-references. We get
1330
# references from every item in the delta that is not a deletion and
1331
# is not itself the root.
1333
# Added ids must not be in the dirstate already. This set holds those
1336
# This loop transforms the delta to single atomic operations that can
1337
# be executed and validated.
1338
delta = sorted(self._check_delta_is_valid(delta), reverse=True)
1339
for old_path, new_path, file_id, inv_entry in delta:
1340
if (file_id in insertions) or (file_id in removals):
1341
self._raise_invalid(old_path or new_path, file_id,
1343
if old_path is not None:
1344
old_path = old_path.encode('utf-8')
1345
removals[file_id] = old_path
1347
new_ids.add(file_id)
1348
if new_path is not None:
1349
if inv_entry is None:
1350
self._raise_invalid(new_path, file_id,
1351
"new_path with no entry")
1352
new_path = new_path.encode('utf-8')
1353
dirname_utf8, basename = osutils.split(new_path)
1355
parents.add((dirname_utf8, inv_entry.parent_id))
1356
key = (dirname_utf8, basename, file_id)
1357
minikind = DirState._kind_to_minikind[inv_entry.kind]
1359
fingerprint = inv_entry.reference_revision or ''
1362
insertions[file_id] = (key, minikind, inv_entry.executable,
1363
fingerprint, new_path)
1364
# Transform moves into delete+add pairs
1365
if None not in (old_path, new_path):
1366
for child in self._iter_child_entries(0, old_path):
1367
if child[0][2] in insertions or child[0][2] in removals:
1369
child_dirname = child[0][0]
1370
child_basename = child[0][1]
1371
minikind = child[1][0][0]
1372
fingerprint = child[1][0][4]
1373
executable = child[1][0][3]
1374
old_child_path = osutils.pathjoin(child_dirname,
1376
removals[child[0][2]] = old_child_path
1377
child_suffix = child_dirname[len(old_path):]
1378
new_child_dirname = (new_path + child_suffix)
1379
key = (new_child_dirname, child_basename, child[0][2])
1380
new_child_path = osutils.pathjoin(new_child_dirname,
1382
insertions[child[0][2]] = (key, minikind, executable,
1383
fingerprint, new_child_path)
1384
self._check_delta_ids_absent(new_ids, delta, 0)
1386
self._apply_removals(removals.iteritems())
1387
self._apply_insertions(insertions.values())
1389
self._after_delta_check_parents(parents, 0)
1390
except errors.BzrError, e:
1391
self._changes_aborted = True
1392
if 'integrity error' not in str(e):
1394
# _get_entry raises BzrError when a request is inconsistent; we
1395
# want such errors to be shown as InconsistentDelta - and that
1396
# fits the behaviour we trigger.
1397
raise errors.InconsistentDeltaDelta(delta,
1398
"error from _get_entry. %s" % (e,))
1400
def _apply_removals(self, removals):
1401
for file_id, path in sorted(removals, reverse=True,
1402
key=operator.itemgetter(1)):
1403
dirname, basename = osutils.split(path)
1404
block_i, entry_i, d_present, f_present = \
1405
self._get_block_entry_index(dirname, basename, 0)
1407
entry = self._dirblocks[block_i][1][entry_i]
1409
self._raise_invalid(path, file_id,
1410
"Wrong path for old path.")
1411
if not f_present or entry[1][0][0] in 'ar':
1412
self._raise_invalid(path, file_id,
1413
"Wrong path for old path.")
1414
if file_id != entry[0][2]:
1415
self._raise_invalid(path, file_id,
1416
"Attempt to remove path has wrong id - found %r."
1418
self._make_absent(entry)
1419
# See if we have a malformed delta: deleting a directory must not
1420
# leave crud behind. This increases the number of bisects needed
1421
# substantially, but deletion or renames of large numbers of paths
1422
# is rare enough it shouldn't be an issue (famous last words?) RBC
1424
block_i, entry_i, d_present, f_present = \
1425
self._get_block_entry_index(path, '', 0)
1427
# The dir block is still present in the dirstate; this could
1428
# be due to it being in a parent tree, or a corrupt delta.
1429
for child_entry in self._dirblocks[block_i][1]:
1430
if child_entry[1][0][0] not in ('r', 'a'):
1431
self._raise_invalid(path, entry[0][2],
1432
"The file id was deleted but its children were "
1435
def _apply_insertions(self, adds):
1437
for key, minikind, executable, fingerprint, path_utf8 in sorted(adds):
1438
self.update_minimal(key, minikind, executable, fingerprint,
1439
path_utf8=path_utf8)
1440
except errors.NotVersionedError:
1441
self._raise_invalid(path_utf8.decode('utf8'), key[2],
1444
def update_basis_by_delta(self, delta, new_revid):
1445
"""Update the parents of this tree after a commit.
1447
This gives the tree one parent, with revision id new_revid. The
1448
inventory delta is applied to the current basis tree to generate the
1449
inventory for the parent new_revid, and all other parent trees are
1452
Note that an exception during the operation of this method will leave
1453
the dirstate in a corrupt state where it should not be saved.
1455
:param new_revid: The new revision id for the trees parent.
1456
:param delta: An inventory delta (see apply_inventory_delta) describing
1457
the changes from the current left most parent revision to new_revid.
1459
self._read_dirblocks_if_needed()
1460
self._discard_merge_parents()
1461
if self._ghosts != []:
1462
raise NotImplementedError(self.update_basis_by_delta)
1463
if len(self._parents) == 0:
1464
# setup a blank tree, the most simple way.
1465
empty_parent = DirState.NULL_PARENT_DETAILS
1466
for entry in self._iter_entries():
1467
entry[1].append(empty_parent)
1468
self._parents.append(new_revid)
1470
self._parents[0] = new_revid
1472
delta = sorted(self._check_delta_is_valid(delta), reverse=True)
1476
# The paths this function accepts are unicode and must be encoded as we
1478
encode = cache_utf8.encode
1479
inv_to_entry = self._inv_entry_to_details
1480
# delta is now (deletes, changes), (adds) in reverse lexographical
1482
# deletes in reverse lexographic order are safe to process in situ.
1483
# renames are not, as a rename from any path could go to a path
1484
# lexographically lower, so we transform renames into delete, add pairs,
1485
# expanding them recursively as needed.
1486
# At the same time, to reduce interface friction we convert the input
1487
# inventory entries to dirstate.
1488
root_only = ('', '')
1489
# Accumulate parent references (path_utf8, id), to check for parentless
1490
# items or items placed under files/links/tree-references. We get
1491
# references from every item in the delta that is not a deletion and
1492
# is not itself the root.
1494
# Added ids must not be in the dirstate already. This set holds those
1497
for old_path, new_path, file_id, inv_entry in delta:
1498
if inv_entry is not None and file_id != inv_entry.file_id:
1499
self._raise_invalid(new_path, file_id,
1500
"mismatched entry file_id %r" % inv_entry)
1501
if new_path is None:
1502
new_path_utf8 = None
1504
if inv_entry is None:
1505
self._raise_invalid(new_path, file_id,
1506
"new_path with no entry")
1507
new_path_utf8 = encode(new_path)
1508
# note the parent for validation
1509
dirname_utf8, basename_utf8 = osutils.split(new_path_utf8)
1511
parents.add((dirname_utf8, inv_entry.parent_id))
1512
if old_path is None:
1513
old_path_utf8 = None
1515
old_path_utf8 = encode(old_path)
1516
if old_path is None:
1517
adds.append((None, new_path_utf8, file_id,
1518
inv_to_entry(inv_entry), True))
1519
new_ids.add(file_id)
1520
elif new_path is None:
1521
deletes.append((old_path_utf8, None, file_id, None, True))
1522
elif (old_path, new_path) == root_only:
1523
# change things in-place
1524
# Note: the case of a parent directory changing its file_id
1525
# tends to break optimizations here, because officially
1526
# the file has actually been moved, it just happens to
1527
# end up at the same path. If we can figure out how to
1528
# handle that case, we can avoid a lot of add+delete
1529
# pairs for objects that stay put.
1530
# elif old_path == new_path:
1531
changes.append((old_path_utf8, new_path_utf8, file_id,
1532
inv_to_entry(inv_entry)))
1535
# Because renames must preserve their children we must have
1536
# processed all relocations and removes before hand. The sort
1537
# order ensures we've examined the child paths, but we also
1538
# have to execute the removals, or the split to an add/delete
1539
# pair will result in the deleted item being reinserted, or
1540
# renamed items being reinserted twice - and possibly at the
1541
# wrong place. Splitting into a delete/add pair also simplifies
1542
# the handling of entries with ('f', ...), ('r' ...) because
1543
# the target of the 'r' is old_path here, and we add that to
1544
# deletes, meaning that the add handler does not need to check
1545
# for 'r' items on every pass.
1546
self._update_basis_apply_deletes(deletes)
1548
# Split into an add/delete pair recursively.
1549
adds.append((old_path_utf8, new_path_utf8, file_id,
1550
inv_to_entry(inv_entry), False))
1551
# Expunge deletes that we've seen so that deleted/renamed
1552
# children of a rename directory are handled correctly.
1553
new_deletes = reversed(list(
1554
self._iter_child_entries(1, old_path_utf8)))
1555
# Remove the current contents of the tree at orig_path, and
1556
# reinsert at the correct new path.
1557
for entry in new_deletes:
1558
child_dirname, child_basename, child_file_id = entry[0]
1560
source_path = child_dirname + '/' + child_basename
1562
source_path = child_basename
1565
new_path_utf8 + source_path[len(old_path_utf8):]
1567
if old_path_utf8 == '':
1568
raise AssertionError("cannot rename directory to"
1570
target_path = source_path[len(old_path_utf8) + 1:]
1571
adds.append((None, target_path, entry[0][2], entry[1][1], False))
1573
(source_path, target_path, entry[0][2], None, False))
1575
(old_path_utf8, new_path_utf8, file_id, None, False))
1577
self._check_delta_ids_absent(new_ids, delta, 1)
1579
# Finish expunging deletes/first half of renames.
1580
self._update_basis_apply_deletes(deletes)
1581
# Reinstate second half of renames and new paths.
1582
self._update_basis_apply_adds(adds)
1583
# Apply in-situ changes.
1584
self._update_basis_apply_changes(changes)
1586
self._after_delta_check_parents(parents, 1)
1587
except errors.BzrError, e:
1588
self._changes_aborted = True
1589
if 'integrity error' not in str(e):
1591
# _get_entry raises BzrError when a request is inconsistent; we
1592
# want such errors to be shown as InconsistentDelta - and that
1593
# fits the behaviour we trigger.
1594
raise errors.InconsistentDeltaDelta(delta,
1595
"error from _get_entry. %s" % (e,))
1597
self._mark_modified(header_modified=True)
1598
self._id_index = None
1601
def _check_delta_ids_absent(self, new_ids, delta, tree_index):
1602
"""Check that none of the file_ids in new_ids are present in a tree."""
1605
id_index = self._get_id_index()
1606
for file_id in new_ids:
1607
for key in id_index.get(file_id, ()):
1608
block_i, entry_i, d_present, f_present = \
1609
self._get_block_entry_index(key[0], key[1], tree_index)
1611
# In a different tree
1613
entry = self._dirblocks[block_i][1][entry_i]
1614
if entry[0][2] != file_id:
1615
# Different file_id, so not what we want.
1617
self._raise_invalid(("%s/%s" % key[0:2]).decode('utf8'), file_id,
1618
"This file_id is new in the delta but already present in "
1621
def _raise_invalid(self, path, file_id, reason):
1622
self._changes_aborted = True
1623
raise errors.InconsistentDelta(path, file_id, reason)
1625
def _update_basis_apply_adds(self, adds):
1626
"""Apply a sequence of adds to tree 1 during update_basis_by_delta.
1628
They may be adds, or renames that have been split into add/delete
1631
:param adds: A sequence of adds. Each add is a tuple:
1632
(None, new_path_utf8, file_id, (entry_details), real_add). real_add
1633
is False when the add is the second half of a remove-and-reinsert
1634
pair created to handle renames and deletes.
1636
# Adds are accumulated partly from renames, so can be in any input
1638
# TODO: we may want to sort in dirblocks order. That way each entry
1639
# will end up in the same directory, allowing the _get_entry
1640
# fast-path for looking up 2 items in the same dir work.
1641
adds.sort(key=lambda x: x[1])
1642
# adds is now in lexographic order, which places all parents before
1643
# their children, so we can process it linearly.
1645
st = static_tuple.StaticTuple
1646
for old_path, new_path, file_id, new_details, real_add in adds:
1647
dirname, basename = osutils.split(new_path)
1648
entry_key = st(dirname, basename, file_id)
1649
block_index, present = self._find_block_index_from_key(entry_key)
1651
# The block where we want to put the file is not present.
1652
# However, it might have just been an empty directory. Look for
1653
# the parent in the basis-so-far before throwing an error.
1654
parent_dir, parent_base = osutils.split(dirname)
1655
parent_block_idx, parent_entry_idx, _, parent_present = \
1656
self._get_block_entry_index(parent_dir, parent_base, 1)
1657
if not parent_present:
1658
self._raise_invalid(new_path, file_id,
1659
"Unable to find block for this record."
1660
" Was the parent added?")
1661
self._ensure_block(parent_block_idx, parent_entry_idx, dirname)
1663
block = self._dirblocks[block_index][1]
1664
entry_index, present = self._find_entry_index(entry_key, block)
1666
if old_path is not None:
1667
self._raise_invalid(new_path, file_id,
1668
'considered a real add but still had old_path at %s'
1671
entry = block[entry_index]
1672
basis_kind = entry[1][1][0]
1673
if basis_kind == 'a':
1674
entry[1][1] = new_details
1675
elif basis_kind == 'r':
1676
raise NotImplementedError()
1678
self._raise_invalid(new_path, file_id,
1679
"An entry was marked as a new add"
1680
" but the basis target already existed")
1682
# The exact key was not found in the block. However, we need to
1683
# check if there is a key next to us that would have matched.
1684
# We only need to check 2 locations, because there are only 2
1686
for maybe_index in range(entry_index-1, entry_index+1):
1687
if maybe_index < 0 or maybe_index >= len(block):
1689
maybe_entry = block[maybe_index]
1690
if maybe_entry[0][:2] != (dirname, basename):
1691
# Just a random neighbor
1693
if maybe_entry[0][2] == file_id:
1694
raise AssertionError(
1695
'_find_entry_index didnt find a key match'
1696
' but walking the data did, for %s'
1698
basis_kind = maybe_entry[1][1][0]
1699
if basis_kind not in 'ar':
1700
self._raise_invalid(new_path, file_id,
1701
"we have an add record for path, but the path"
1702
" is already present with another file_id %s"
1703
% (maybe_entry[0][2],))
1705
entry = (entry_key, [DirState.NULL_PARENT_DETAILS,
1707
block.insert(entry_index, entry)
1709
active_kind = entry[1][0][0]
1710
if active_kind == 'a':
1711
# The active record shows up as absent, this could be genuine,
1712
# or it could be present at some other location. We need to
1714
id_index = self._get_id_index()
1715
# The id_index may not be perfectly accurate for tree1, because
1716
# we haven't been keeping it updated. However, it should be
1717
# fine for tree0, and that gives us enough info for what we
1719
keys = id_index.get(file_id, ())
1721
block_i, entry_i, d_present, f_present = \
1722
self._get_block_entry_index(key[0], key[1], 0)
1725
active_entry = self._dirblocks[block_i][1][entry_i]
1726
if (active_entry[0][2] != file_id):
1727
# Some other file is at this path, we don't need to
1730
real_active_kind = active_entry[1][0][0]
1731
if real_active_kind in 'ar':
1732
# We found a record, which was not *this* record,
1733
# which matches the file_id, but is not actually
1734
# present. Something seems *really* wrong.
1735
self._raise_invalid(new_path, file_id,
1736
"We found a tree0 entry that doesnt make sense")
1737
# Now, we've found a tree0 entry which matches the file_id
1738
# but is at a different location. So update them to be
1740
active_dir, active_name = active_entry[0][:2]
1742
active_path = active_dir + '/' + active_name
1744
active_path = active_name
1745
active_entry[1][1] = st('r', new_path, 0, False, '')
1746
entry[1][0] = st('r', active_path, 0, False, '')
1747
elif active_kind == 'r':
1748
raise NotImplementedError()
1750
new_kind = new_details[0]
1752
self._ensure_block(block_index, entry_index, new_path)
1754
def _update_basis_apply_changes(self, changes):
1755
"""Apply a sequence of changes to tree 1 during update_basis_by_delta.
1757
:param adds: A sequence of changes. Each change is a tuple:
1758
(path_utf8, path_utf8, file_id, (entry_details))
1761
for old_path, new_path, file_id, new_details in changes:
1762
# the entry for this file_id must be in tree 0.
1763
entry = self._get_entry(1, file_id, new_path)
1764
if entry[0] is None or entry[1][1][0] in 'ar':
1765
self._raise_invalid(new_path, file_id,
1766
'changed entry considered not present')
1767
entry[1][1] = new_details
1769
def _update_basis_apply_deletes(self, deletes):
1770
"""Apply a sequence of deletes to tree 1 during update_basis_by_delta.
1772
They may be deletes, or renames that have been split into add/delete
1775
:param deletes: A sequence of deletes. Each delete is a tuple:
1776
(old_path_utf8, new_path_utf8, file_id, None, real_delete).
1777
real_delete is True when the desired outcome is an actual deletion
1778
rather than the rename handling logic temporarily deleting a path
1779
during the replacement of a parent.
1781
null = DirState.NULL_PARENT_DETAILS
1782
for old_path, new_path, file_id, _, real_delete in deletes:
1783
if real_delete != (new_path is None):
1784
self._raise_invalid(old_path, file_id, "bad delete delta")
1785
# the entry for this file_id must be in tree 1.
1786
dirname, basename = osutils.split(old_path)
1787
block_index, entry_index, dir_present, file_present = \
1788
self._get_block_entry_index(dirname, basename, 1)
1789
if not file_present:
1790
self._raise_invalid(old_path, file_id,
1791
'basis tree does not contain removed entry')
1792
entry = self._dirblocks[block_index][1][entry_index]
1793
# The state of the entry in the 'active' WT
1794
active_kind = entry[1][0][0]
1795
if entry[0][2] != file_id:
1796
self._raise_invalid(old_path, file_id,
1797
'mismatched file_id in tree 1')
1799
old_kind = entry[1][1][0]
1800
if active_kind in 'ar':
1801
# The active tree doesn't have this file_id.
1802
# The basis tree is changing this record. If this is a
1803
# rename, then we don't want the record here at all
1804
# anymore. If it is just an in-place change, we want the
1805
# record here, but we'll add it if we need to. So we just
1807
if active_kind == 'r':
1808
active_path = entry[1][0][1]
1809
active_entry = self._get_entry(0, file_id, active_path)
1810
if active_entry[1][1][0] != 'r':
1811
self._raise_invalid(old_path, file_id,
1812
"Dirstate did not have matching rename entries")
1813
elif active_entry[1][0][0] in 'ar':
1814
self._raise_invalid(old_path, file_id,
1815
"Dirstate had a rename pointing at an inactive"
1817
active_entry[1][1] = null
1818
del self._dirblocks[block_index][1][entry_index]
1820
# This was a directory, and the active tree says it
1821
# doesn't exist, and now the basis tree says it doesn't
1822
# exist. Remove its dirblock if present
1824
present) = self._find_block_index_from_key(
1827
dir_block = self._dirblocks[dir_block_index][1]
1829
# This entry is empty, go ahead and just remove it
1830
del self._dirblocks[dir_block_index]
1832
# There is still an active record, so just mark this
1835
block_i, entry_i, d_present, f_present = \
1836
self._get_block_entry_index(old_path, '', 1)
1838
dir_block = self._dirblocks[block_i][1]
1839
for child_entry in dir_block:
1840
child_basis_kind = child_entry[1][1][0]
1841
if child_basis_kind not in 'ar':
1842
self._raise_invalid(old_path, file_id,
1843
"The file id was deleted but its children were "
1846
def _after_delta_check_parents(self, parents, index):
1847
"""Check that parents required by the delta are all intact.
1849
:param parents: An iterable of (path_utf8, file_id) tuples which are
1850
required to be present in tree 'index' at path_utf8 with id file_id
1852
:param index: The column in the dirstate to check for parents in.
1854
for dirname_utf8, file_id in parents:
1855
# Get the entry - the ensures that file_id, dirname_utf8 exists and
1856
# has the right file id.
1857
entry = self._get_entry(index, file_id, dirname_utf8)
1858
if entry[1] is None:
1859
self._raise_invalid(dirname_utf8.decode('utf8'),
1860
file_id, "This parent is not present.")
1861
# Parents of things must be directories
1862
if entry[1][index][0] != 'd':
1863
self._raise_invalid(dirname_utf8.decode('utf8'),
1864
file_id, "This parent is not a directory.")
1866
def _observed_sha1(self, entry, sha1, stat_value,
1867
_stat_to_minikind=_stat_to_minikind):
1868
"""Note the sha1 of a file.
1870
:param entry: The entry the sha1 is for.
1871
:param sha1: The observed sha1.
1872
:param stat_value: The os.lstat for the file.
1875
minikind = _stat_to_minikind[stat_value.st_mode & 0170000]
1880
if self._cutoff_time is None:
1881
self._sha_cutoff_time()
1882
if (stat_value.st_mtime < self._cutoff_time
1883
and stat_value.st_ctime < self._cutoff_time):
1884
entry[1][0] = ('f', sha1, stat_value.st_size, entry[1][0][3],
1885
pack_stat(stat_value))
1886
self._mark_modified([entry])
1888
def _sha_cutoff_time(self):
1889
"""Return cutoff time.
1891
Files modified more recently than this time are at risk of being
1892
undetectably modified and so can't be cached.
1894
# Cache the cutoff time as long as we hold a lock.
1895
# time.time() isn't super expensive (approx 3.38us), but
1896
# when you call it 50,000 times it adds up.
1897
# For comparison, os.lstat() costs 7.2us if it is hot.
1898
self._cutoff_time = int(time.time()) - 3
1899
return self._cutoff_time
1901
def _lstat(self, abspath, entry):
1902
"""Return the os.lstat value for this path."""
1903
return os.lstat(abspath)
1905
def _sha1_file_and_mutter(self, abspath):
1906
# when -Dhashcache is turned on, this is monkey-patched in to log
1908
trace.mutter("dirstate sha1 " + abspath)
1909
return self._sha1_provider.sha1(abspath)
1911
def _is_executable(self, mode, old_executable):
1912
"""Is this file executable?"""
1913
return bool(S_IEXEC & mode)
1915
def _is_executable_win32(self, mode, old_executable):
1916
"""On win32 the executable bit is stored in the dirstate."""
1917
return old_executable
1919
if sys.platform == 'win32':
1920
_is_executable = _is_executable_win32
1922
def _read_link(self, abspath, old_link):
1923
"""Read the target of a symlink"""
1924
# TODO: jam 200700301 On Win32, this could just return the value
1925
# already in memory. However, this really needs to be done at a
1926
# higher level, because there either won't be anything on disk,
1927
# or the thing on disk will be a file.
1928
fs_encoding = osutils._fs_enc
1929
if isinstance(abspath, unicode):
1930
# abspath is defined as the path to pass to lstat. readlink is
1931
# buggy in python < 2.6 (it doesn't encode unicode path into FS
1932
# encoding), so we need to encode ourselves knowing that unicode
1933
# paths are produced by UnicodeDirReader on purpose.
1934
abspath = abspath.encode(fs_encoding)
1935
target = os.readlink(abspath)
1936
if fs_encoding not in ('utf-8', 'ascii'):
1937
# Change encoding if needed
1938
target = target.decode(fs_encoding).encode('UTF-8')
1941
def get_ghosts(self):
1942
"""Return a list of the parent tree revision ids that are ghosts."""
1943
self._read_header_if_needed()
1946
def get_lines(self):
1947
"""Serialise the entire dirstate to a sequence of lines."""
1948
if (self._header_state == DirState.IN_MEMORY_UNMODIFIED and
1949
self._dirblock_state == DirState.IN_MEMORY_UNMODIFIED):
1950
# read what's on disk.
1951
self._state_file.seek(0)
1952
return self._state_file.readlines()
1954
lines.append(self._get_parents_line(self.get_parent_ids()))
1955
lines.append(self._get_ghosts_line(self._ghosts))
1956
lines.extend(self._get_entry_lines())
1957
return self._get_output_lines(lines)
1959
def _get_ghosts_line(self, ghost_ids):
1960
"""Create a line for the state file for ghost information."""
1961
return '\0'.join([str(len(ghost_ids))] + ghost_ids)
1963
def _get_parents_line(self, parent_ids):
1964
"""Create a line for the state file for parents information."""
1965
return '\0'.join([str(len(parent_ids))] + parent_ids)
1967
def _get_entry_lines(self):
1968
"""Create lines for entries."""
1969
return map(self._entry_to_line, self._iter_entries())
1971
def _get_fields_to_entry(self):
1972
"""Get a function which converts entry fields into a entry record.
1974
This handles size and executable, as well as parent records.
1976
:return: A function which takes a list of fields, and returns an
1977
appropriate record for storing in memory.
1979
# This is intentionally unrolled for performance
1980
num_present_parents = self._num_present_parents()
1981
if num_present_parents == 0:
1982
def fields_to_entry_0_parents(fields, _int=int):
1983
path_name_file_id_key = (fields[0], fields[1], fields[2])
1984
return (path_name_file_id_key, [
1986
fields[3], # minikind
1987
fields[4], # fingerprint
1988
_int(fields[5]), # size
1989
fields[6] == 'y', # executable
1990
fields[7], # packed_stat or revision_id
1992
return fields_to_entry_0_parents
1993
elif num_present_parents == 1:
1994
def fields_to_entry_1_parent(fields, _int=int):
1995
path_name_file_id_key = (fields[0], fields[1], fields[2])
1996
return (path_name_file_id_key, [
1998
fields[3], # minikind
1999
fields[4], # fingerprint
2000
_int(fields[5]), # size
2001
fields[6] == 'y', # executable
2002
fields[7], # packed_stat or revision_id
2005
fields[8], # minikind
2006
fields[9], # fingerprint
2007
_int(fields[10]), # size
2008
fields[11] == 'y', # executable
2009
fields[12], # packed_stat or revision_id
2012
return fields_to_entry_1_parent
2013
elif num_present_parents == 2:
2014
def fields_to_entry_2_parents(fields, _int=int):
2015
path_name_file_id_key = (fields[0], fields[1], fields[2])
2016
return (path_name_file_id_key, [
2018
fields[3], # minikind
2019
fields[4], # fingerprint
2020
_int(fields[5]), # size
2021
fields[6] == 'y', # executable
2022
fields[7], # packed_stat or revision_id
2025
fields[8], # minikind
2026
fields[9], # fingerprint
2027
_int(fields[10]), # size
2028
fields[11] == 'y', # executable
2029
fields[12], # packed_stat or revision_id
2032
fields[13], # minikind
2033
fields[14], # fingerprint
2034
_int(fields[15]), # size
2035
fields[16] == 'y', # executable
2036
fields[17], # packed_stat or revision_id
2039
return fields_to_entry_2_parents
2041
def fields_to_entry_n_parents(fields, _int=int):
2042
path_name_file_id_key = (fields[0], fields[1], fields[2])
2043
trees = [(fields[cur], # minikind
2044
fields[cur+1], # fingerprint
2045
_int(fields[cur+2]), # size
2046
fields[cur+3] == 'y', # executable
2047
fields[cur+4], # stat or revision_id
2048
) for cur in xrange(3, len(fields)-1, 5)]
2049
return path_name_file_id_key, trees
2050
return fields_to_entry_n_parents
2052
def get_parent_ids(self):
2053
"""Return a list of the parent tree ids for the directory state."""
2054
self._read_header_if_needed()
2055
return list(self._parents)
2057
def _get_block_entry_index(self, dirname, basename, tree_index):
2058
"""Get the coordinates for a path in the state structure.
2060
:param dirname: The utf8 dirname to lookup.
2061
:param basename: The utf8 basename to lookup.
2062
:param tree_index: The index of the tree for which this lookup should
2064
:return: A tuple describing where the path is located, or should be
2065
inserted. The tuple contains four fields: the block index, the row
2066
index, the directory is present (boolean), the entire path is
2067
present (boolean). There is no guarantee that either
2068
coordinate is currently reachable unless the found field for it is
2069
True. For instance, a directory not present in the searched tree
2070
may be returned with a value one greater than the current highest
2071
block offset. The directory present field will always be True when
2072
the path present field is True. The directory present field does
2073
NOT indicate that the directory is present in the searched tree,
2074
rather it indicates that there are at least some files in some
2077
self._read_dirblocks_if_needed()
2078
key = dirname, basename, ''
2079
block_index, present = self._find_block_index_from_key(key)
2081
# no such directory - return the dir index and 0 for the row.
2082
return block_index, 0, False, False
2083
block = self._dirblocks[block_index][1] # access the entries only
2084
entry_index, present = self._find_entry_index(key, block)
2085
# linear search through entries at this path to find the one
2087
while entry_index < len(block) and block[entry_index][0][1] == basename:
2088
if block[entry_index][1][tree_index][0] not in 'ar':
2089
# neither absent or relocated
2090
return block_index, entry_index, True, True
2092
return block_index, entry_index, True, False
2094
def _get_entry(self, tree_index, fileid_utf8=None, path_utf8=None,
2095
include_deleted=False):
2096
"""Get the dirstate entry for path in tree tree_index.
2098
If either file_id or path is supplied, it is used as the key to lookup.
2099
If both are supplied, the fastest lookup is used, and an error is
2100
raised if they do not both point at the same row.
2102
:param tree_index: The index of the tree we wish to locate this path
2103
in. If the path is present in that tree, the entry containing its
2104
details is returned, otherwise (None, None) is returned
2105
0 is the working tree, higher indexes are successive parent
2107
:param fileid_utf8: A utf8 file_id to look up.
2108
:param path_utf8: An utf8 path to be looked up.
2109
:param include_deleted: If True, and performing a lookup via
2110
fileid_utf8 rather than path_utf8, return an entry for deleted
2112
:return: The dirstate entry tuple for path, or (None, None)
2114
self._read_dirblocks_if_needed()
2115
if path_utf8 is not None:
2116
if type(path_utf8) is not str:
2117
raise errors.BzrError('path_utf8 is not a str: %s %r'
2118
% (type(path_utf8), path_utf8))
2119
# path lookups are faster
2120
dirname, basename = osutils.split(path_utf8)
2121
block_index, entry_index, dir_present, file_present = \
2122
self._get_block_entry_index(dirname, basename, tree_index)
2123
if not file_present:
2125
entry = self._dirblocks[block_index][1][entry_index]
2126
if not (entry[0][2] and entry[1][tree_index][0] not in ('a', 'r')):
2127
raise AssertionError('unversioned entry?')
2129
if entry[0][2] != fileid_utf8:
2130
self._changes_aborted = True
2131
raise errors.BzrError('integrity error ? : mismatching'
2132
' tree_index, file_id and path')
2135
possible_keys = self._get_id_index().get(fileid_utf8, ())
2136
if not possible_keys:
2138
for key in possible_keys:
2139
block_index, present = \
2140
self._find_block_index_from_key(key)
2141
# strange, probably indicates an out of date
2142
# id index - for now, allow this.
2145
# WARNING: DO not change this code to use _get_block_entry_index
2146
# as that function is not suitable: it does not use the key
2147
# to lookup, and thus the wrong coordinates are returned.
2148
block = self._dirblocks[block_index][1]
2149
entry_index, present = self._find_entry_index(key, block)
2151
entry = self._dirblocks[block_index][1][entry_index]
2152
# TODO: We might want to assert that entry[0][2] ==
2154
if entry[1][tree_index][0] in 'fdlt':
2155
# this is the result we are looking for: the
2156
# real home of this file_id in this tree.
2158
if entry[1][tree_index][0] == 'a':
2159
# there is no home for this entry in this tree
2163
if entry[1][tree_index][0] != 'r':
2164
raise AssertionError(
2165
"entry %r has invalid minikind %r for tree %r" \
2167
entry[1][tree_index][0],
2169
real_path = entry[1][tree_index][1]
2170
return self._get_entry(tree_index, fileid_utf8=fileid_utf8,
2171
path_utf8=real_path)
2175
def initialize(cls, path, sha1_provider=None):
2176
"""Create a new dirstate on path.
2178
The new dirstate will be an empty tree - that is it has no parents,
2179
and only a root node - which has id ROOT_ID.
2181
:param path: The name of the file for the dirstate.
2182
:param sha1_provider: an object meeting the SHA1Provider interface.
2183
If None, a DefaultSHA1Provider is used.
2184
:return: A write-locked DirState object.
2186
# This constructs a new DirState object on a path, sets the _state_file
2187
# to a new empty file for that path. It then calls _set_data() with our
2188
# stock empty dirstate information - a root with ROOT_ID, no children,
2189
# and no parents. Finally it calls save() to ensure that this data will
2191
if sha1_provider is None:
2192
sha1_provider = DefaultSHA1Provider()
2193
result = cls(path, sha1_provider)
2194
# root dir and root dir contents with no children.
2195
empty_tree_dirblocks = [('', []), ('', [])]
2196
# a new root directory, with a NULLSTAT.
2197
empty_tree_dirblocks[0][1].append(
2198
(('', '', inventory.ROOT_ID), [
2199
('d', '', 0, False, DirState.NULLSTAT),
2203
result._set_data([], empty_tree_dirblocks)
2211
def _inv_entry_to_details(inv_entry):
2212
"""Convert an inventory entry (from a revision tree) to state details.
2214
:param inv_entry: An inventory entry whose sha1 and link targets can be
2215
relied upon, and which has a revision set.
2216
:return: A details tuple - the details for a single tree at a path +
2219
kind = inv_entry.kind
2220
minikind = DirState._kind_to_minikind[kind]
2221
tree_data = inv_entry.revision
2222
if kind == 'directory':
2226
elif kind == 'symlink':
2227
if inv_entry.symlink_target is None:
2230
fingerprint = inv_entry.symlink_target.encode('utf8')
2233
elif kind == 'file':
2234
fingerprint = inv_entry.text_sha1 or ''
2235
size = inv_entry.text_size or 0
2236
executable = inv_entry.executable
2237
elif kind == 'tree-reference':
2238
fingerprint = inv_entry.reference_revision or ''
2242
raise Exception("can't pack %s" % inv_entry)
2243
return static_tuple.StaticTuple(minikind, fingerprint, size,
2244
executable, tree_data)
2246
def _iter_child_entries(self, tree_index, path_utf8):
2247
"""Iterate over all the entries that are children of path_utf.
2249
This only returns entries that are present (not in 'a', 'r') in
2250
tree_index. tree_index data is not refreshed, so if tree 0 is used,
2251
results may differ from that obtained if paths were statted to
2252
determine what ones were directories.
2254
Asking for the children of a non-directory will return an empty
2258
next_pending_dirs = [path_utf8]
2260
while next_pending_dirs:
2261
pending_dirs = next_pending_dirs
2262
next_pending_dirs = []
2263
for path in pending_dirs:
2264
block_index, present = self._find_block_index_from_key(
2266
if block_index == 0:
2268
if len(self._dirblocks) == 1:
2269
# asked for the children of the root with no other
2273
# children of a non-directory asked for.
2275
block = self._dirblocks[block_index]
2276
for entry in block[1]:
2277
kind = entry[1][tree_index][0]
2278
if kind not in absent:
2282
path = entry[0][0] + '/' + entry[0][1]
2285
next_pending_dirs.append(path)
2287
def _iter_entries(self):
2288
"""Iterate over all the entries in the dirstate.
2290
Each yelt item is an entry in the standard format described in the
2291
docstring of bzrlib.dirstate.
2293
self._read_dirblocks_if_needed()
2294
for directory in self._dirblocks:
2295
for entry in directory[1]:
2298
def _get_id_index(self):
2299
"""Get an id index of self._dirblocks.
2301
This maps from file_id => [(directory, name, file_id)] entries where
2302
that file_id appears in one of the trees.
2304
if self._id_index is None:
2306
for key, tree_details in self._iter_entries():
2307
self._add_to_id_index(id_index, key)
2308
self._id_index = id_index
2309
return self._id_index
2311
def _add_to_id_index(self, id_index, entry_key):
2312
"""Add this entry to the _id_index mapping."""
2313
# This code used to use a set for every entry in the id_index. However,
2314
# it is *rare* to have more than one entry. So a set is a large
2315
# overkill. And even when we do, we won't ever have more than the
2316
# number of parent trees. Which is still a small number (rarely >2). As
2317
# such, we use a simple tuple, and do our own uniqueness checks. While
2318
# the 'in' check is O(N) since N is nicely bounded it shouldn't ever
2319
# cause quadratic failure.
2320
file_id = entry_key[2]
2321
entry_key = static_tuple.StaticTuple.from_sequence(entry_key)
2322
if file_id not in id_index:
2323
id_index[file_id] = static_tuple.StaticTuple(entry_key,)
2325
entry_keys = id_index[file_id]
2326
if entry_key not in entry_keys:
2327
id_index[file_id] = entry_keys + (entry_key,)
2329
def _remove_from_id_index(self, id_index, entry_key):
2330
"""Remove this entry from the _id_index mapping.
2332
It is an programming error to call this when the entry_key is not
2335
file_id = entry_key[2]
2336
entry_keys = list(id_index[file_id])
2337
entry_keys.remove(entry_key)
2338
id_index[file_id] = static_tuple.StaticTuple.from_sequence(entry_keys)
2340
def _get_output_lines(self, lines):
2341
"""Format lines for final output.
2343
:param lines: A sequence of lines containing the parents list and the
2346
output_lines = [DirState.HEADER_FORMAT_3]
2347
lines.append('') # a final newline
2348
inventory_text = '\0\n\0'.join(lines)
2349
output_lines.append('crc32: %s\n' % (zlib.crc32(inventory_text),))
2350
# -3, 1 for num parents, 1 for ghosts, 1 for final newline
2351
num_entries = len(lines)-3
2352
output_lines.append('num_entries: %s\n' % (num_entries,))
2353
output_lines.append(inventory_text)
2356
def _make_deleted_row(self, fileid_utf8, parents):
2357
"""Return a deleted row for fileid_utf8."""
2358
return ('/', 'RECYCLED.BIN', 'file', fileid_utf8, 0, DirState.NULLSTAT,
2361
def _num_present_parents(self):
2362
"""The number of parent entries in each record row."""
2363
return len(self._parents) - len(self._ghosts)
2366
def on_file(cls, path, sha1_provider=None, worth_saving_limit=0):
2367
"""Construct a DirState on the file at path "path".
2369
:param path: The path at which the dirstate file on disk should live.
2370
:param sha1_provider: an object meeting the SHA1Provider interface.
2371
If None, a DefaultSHA1Provider is used.
2372
:param worth_saving_limit: when the exact number of hash changed
2373
entries is known, only bother saving the dirstate if more than
2374
this count of entries have changed. -1 means never save.
2375
:return: An unlocked DirState object, associated with the given path.
2377
if sha1_provider is None:
2378
sha1_provider = DefaultSHA1Provider()
2379
result = cls(path, sha1_provider,
2380
worth_saving_limit=worth_saving_limit)
2383
def _read_dirblocks_if_needed(self):
2384
"""Read in all the dirblocks from the file if they are not in memory.
2386
This populates self._dirblocks, and sets self._dirblock_state to
2387
IN_MEMORY_UNMODIFIED. It is not currently ready for incremental block
2390
self._read_header_if_needed()
2391
if self._dirblock_state == DirState.NOT_IN_MEMORY:
2392
_read_dirblocks(self)
2394
def _read_header(self):
2395
"""This reads in the metadata header, and the parent ids.
2397
After reading in, the file should be positioned at the null
2398
just before the start of the first record in the file.
2400
:return: (expected crc checksum, number of entries, parent list)
2402
self._read_prelude()
2403
parent_line = self._state_file.readline()
2404
info = parent_line.split('\0')
2405
num_parents = int(info[0])
2406
self._parents = info[1:-1]
2407
ghost_line = self._state_file.readline()
2408
info = ghost_line.split('\0')
2409
num_ghosts = int(info[1])
2410
self._ghosts = info[2:-1]
2411
self._header_state = DirState.IN_MEMORY_UNMODIFIED
2412
self._end_of_header = self._state_file.tell()
2414
def _read_header_if_needed(self):
2415
"""Read the header of the dirstate file if needed."""
2416
# inline this as it will be called a lot
2417
if not self._lock_token:
2418
raise errors.ObjectNotLocked(self)
2419
if self._header_state == DirState.NOT_IN_MEMORY:
2422
def _read_prelude(self):
2423
"""Read in the prelude header of the dirstate file.
2425
This only reads in the stuff that is not connected to the crc
2426
checksum. The position will be correct to read in the rest of
2427
the file and check the checksum after this point.
2428
The next entry in the file should be the number of parents,
2429
and their ids. Followed by a newline.
2431
header = self._state_file.readline()
2432
if header != DirState.HEADER_FORMAT_3:
2433
raise errors.BzrError(
2434
'invalid header line: %r' % (header,))
2435
crc_line = self._state_file.readline()
2436
if not crc_line.startswith('crc32: '):
2437
raise errors.BzrError('missing crc32 checksum: %r' % crc_line)
2438
self.crc_expected = int(crc_line[len('crc32: '):-1])
2439
num_entries_line = self._state_file.readline()
2440
if not num_entries_line.startswith('num_entries: '):
2441
raise errors.BzrError('missing num_entries line')
2442
self._num_entries = int(num_entries_line[len('num_entries: '):-1])
2444
def sha1_from_stat(self, path, stat_result):
2445
"""Find a sha1 given a stat lookup."""
2446
return self._get_packed_stat_index().get(pack_stat(stat_result), None)
2448
def _get_packed_stat_index(self):
2449
"""Get a packed_stat index of self._dirblocks."""
2450
if self._packed_stat_index is None:
2452
for key, tree_details in self._iter_entries():
2453
if tree_details[0][0] == 'f':
2454
index[tree_details[0][4]] = tree_details[0][1]
2455
self._packed_stat_index = index
2456
return self._packed_stat_index
2459
"""Save any pending changes created during this session.
2461
We reuse the existing file, because that prevents race conditions with
2462
file creation, and use oslocks on it to prevent concurrent modification
2463
and reads - because dirstate's incremental data aggregation is not
2464
compatible with reading a modified file, and replacing a file in use by
2465
another process is impossible on Windows.
2467
A dirstate in read only mode should be smart enough though to validate
2468
that the file has not changed, and otherwise discard its cache and
2469
start over, to allow for fine grained read lock duration, so 'status'
2470
wont block 'commit' - for example.
2472
if self._changes_aborted:
2473
# Should this be a warning? For now, I'm expecting that places that
2474
# mark it inconsistent will warn, making a warning here redundant.
2475
trace.mutter('Not saving DirState because '
2476
'_changes_aborted is set.')
2478
# TODO: Since we now distinguish IN_MEMORY_MODIFIED from
2479
# IN_MEMORY_HASH_MODIFIED, we should only fail quietly if we fail
2480
# to save an IN_MEMORY_HASH_MODIFIED, and fail *noisily* if we
2481
# fail to save IN_MEMORY_MODIFIED
2482
if not self._worth_saving():
2485
grabbed_write_lock = False
2486
if self._lock_state != 'w':
2487
grabbed_write_lock, new_lock = self._lock_token.temporary_write_lock()
2488
# Switch over to the new lock, as the old one may be closed.
2489
# TODO: jam 20070315 We should validate the disk file has
2490
# not changed contents, since temporary_write_lock may
2491
# not be an atomic operation.
2492
self._lock_token = new_lock
2493
self._state_file = new_lock.f
2494
if not grabbed_write_lock:
2495
# We couldn't grab a write lock, so we switch back to a read one
2498
lines = self.get_lines()
2499
self._state_file.seek(0)
2500
self._state_file.writelines(lines)
2501
self._state_file.truncate()
2502
self._state_file.flush()
2503
self._maybe_fdatasync()
2504
self._mark_unmodified()
2506
if grabbed_write_lock:
2507
self._lock_token = self._lock_token.restore_read_lock()
2508
self._state_file = self._lock_token.f
2509
# TODO: jam 20070315 We should validate the disk file has
2510
# not changed contents. Since restore_read_lock may
2511
# not be an atomic operation.
2513
def _maybe_fdatasync(self):
2514
"""Flush to disk if possible and if not configured off."""
2515
if self._config_stack.get('dirstate.fdatasync'):
2516
osutils.fdatasync(self._state_file.fileno())
2518
def _worth_saving(self):
2519
"""Is it worth saving the dirstate or not?"""
2520
if (self._header_state == DirState.IN_MEMORY_MODIFIED
2521
or self._dirblock_state == DirState.IN_MEMORY_MODIFIED):
2523
if self._dirblock_state == DirState.IN_MEMORY_HASH_MODIFIED:
2524
if self._worth_saving_limit == -1:
2525
# We never save hash changes when the limit is -1
2527
# If we're using smart saving and only a small number of
2528
# entries have changed their hash, don't bother saving. John has
2529
# suggested using a heuristic here based on the size of the
2530
# changed files and/or tree. For now, we go with a configurable
2531
# number of changes, keeping the calculation time
2532
# as low overhead as possible. (This also keeps all existing
2533
# tests passing as the default is 0, i.e. always save.)
2534
if len(self._known_hash_changes) >= self._worth_saving_limit:
2538
def _set_data(self, parent_ids, dirblocks):
2539
"""Set the full dirstate data in memory.
2541
This is an internal function used to completely replace the objects
2542
in memory state. It puts the dirstate into state 'full-dirty'.
2544
:param parent_ids: A list of parent tree revision ids.
2545
:param dirblocks: A list containing one tuple for each directory in the
2546
tree. Each tuple contains the directory path and a list of entries
2547
found in that directory.
2549
# our memory copy is now authoritative.
2550
self._dirblocks = dirblocks
2551
self._mark_modified(header_modified=True)
2552
self._parents = list(parent_ids)
2553
self._id_index = None
2554
self._packed_stat_index = None
2556
def set_path_id(self, path, new_id):
2557
"""Change the id of path to new_id in the current working tree.
2559
:param path: The path inside the tree to set - '' is the root, 'foo'
2560
is the path foo in the root.
2561
:param new_id: The new id to assign to the path. This must be a utf8
2562
file id (not unicode, and not None).
2564
self._read_dirblocks_if_needed()
2566
# TODO: logic not written
2567
raise NotImplementedError(self.set_path_id)
2568
# TODO: check new id is unique
2569
entry = self._get_entry(0, path_utf8=path)
2570
if entry[0][2] == new_id:
2571
# Nothing to change.
2573
# mark the old path absent, and insert a new root path
2574
self._make_absent(entry)
2575
self.update_minimal(('', '', new_id), 'd',
2576
path_utf8='', packed_stat=entry[1][0][4])
2577
self._mark_modified()
2579
def set_parent_trees(self, trees, ghosts):
2580
"""Set the parent trees for the dirstate.
2582
:param trees: A list of revision_id, tree tuples. tree must be provided
2583
even if the revision_id refers to a ghost: supply an empty tree in
2585
:param ghosts: A list of the revision_ids that are ghosts at the time
2588
# TODO: generate a list of parent indexes to preserve to save
2589
# processing specific parent trees. In the common case one tree will
2590
# be preserved - the left most parent.
2591
# TODO: if the parent tree is a dirstate, we might want to walk them
2592
# all by path in parallel for 'optimal' common-case performance.
2593
# generate new root row.
2594
self._read_dirblocks_if_needed()
2595
# TODO future sketch: Examine the existing parents to generate a change
2596
# map and then walk the new parent trees only, mapping them into the
2597
# dirstate. Walk the dirstate at the same time to remove unreferenced
2600
# sketch: loop over all entries in the dirstate, cherry picking
2601
# entries from the parent trees, if they are not ghost trees.
2602
# after we finish walking the dirstate, all entries not in the dirstate
2603
# are deletes, so we want to append them to the end as per the design
2604
# discussions. So do a set difference on ids with the parents to
2605
# get deletes, and add them to the end.
2606
# During the update process we need to answer the following questions:
2607
# - find other keys containing a fileid in order to create cross-path
2608
# links. We dont't trivially use the inventory from other trees
2609
# because this leads to either double touching, or to accessing
2611
# - find other keys containing a path
2612
# We accumulate each entry via this dictionary, including the root
2615
# we could do parallel iterators, but because file id data may be
2616
# scattered throughout, we dont save on index overhead: we have to look
2617
# at everything anyway. We can probably save cycles by reusing parent
2618
# data and doing an incremental update when adding an additional
2619
# parent, but for now the common cases are adding a new parent (merge),
2620
# and replacing completely (commit), and commit is more common: so
2621
# optimise merge later.
2623
# ---- start generation of full tree mapping data
2624
# what trees should we use?
2625
parent_trees = [tree for rev_id, tree in trees if rev_id not in ghosts]
2626
# how many trees do we end up with
2627
parent_count = len(parent_trees)
2628
st = static_tuple.StaticTuple
2630
# one: the current tree
2631
for entry in self._iter_entries():
2632
# skip entries not in the current tree
2633
if entry[1][0][0] in 'ar': # absent, relocated
2635
by_path[entry[0]] = [entry[1][0]] + \
2636
[DirState.NULL_PARENT_DETAILS] * parent_count
2637
# TODO: Possibly inline this, since we know it isn't present yet
2638
# id_index[entry[0][2]] = (entry[0],)
2639
self._add_to_id_index(id_index, entry[0])
2641
# now the parent trees:
2642
for tree_index, tree in enumerate(parent_trees):
2643
# the index is off by one, adjust it.
2644
tree_index = tree_index + 1
2645
# when we add new locations for a fileid we need these ranges for
2646
# any fileid in this tree as we set the by_path[id] to:
2647
# already_processed_tree_details + new_details + new_location_suffix
2648
# the suffix is from tree_index+1:parent_count+1.
2649
new_location_suffix = [DirState.NULL_PARENT_DETAILS] * (parent_count - tree_index)
2650
# now stitch in all the entries from this tree
2652
for path, entry in tree.iter_entries_by_dir():
2653
# here we process each trees details for each item in the tree.
2654
# we first update any existing entries for the id at other paths,
2655
# then we either create or update the entry for the id at the
2656
# right path, and finally we add (if needed) a mapping from
2657
# file_id to this path. We do it in this order to allow us to
2658
# avoid checking all known paths for the id when generating a
2659
# new entry at this path: by adding the id->path mapping last,
2660
# all the mappings are valid and have correct relocation
2661
# records where needed.
2662
file_id = entry.file_id
2663
path_utf8 = path.encode('utf8')
2664
dirname, basename = osutils.split(path_utf8)
2665
if dirname == last_dirname:
2666
# Try to re-use objects as much as possible
2667
dirname = last_dirname
2669
last_dirname = dirname
2670
new_entry_key = st(dirname, basename, file_id)
2671
# tree index consistency: All other paths for this id in this tree
2672
# index must point to the correct path.
2673
entry_keys = id_index.get(file_id, ())
2674
for entry_key in entry_keys:
2675
# TODO:PROFILING: It might be faster to just update
2676
# rather than checking if we need to, and then overwrite
2677
# the one we are located at.
2678
if entry_key != new_entry_key:
2679
# this file id is at a different path in one of the
2680
# other trees, so put absent pointers there
2681
# This is the vertical axis in the matrix, all pointing
2683
by_path[entry_key][tree_index] = st('r', path_utf8, 0,
2685
# by path consistency: Insert into an existing path record
2686
# (trivial), or add a new one with relocation pointers for the
2687
# other tree indexes.
2688
if new_entry_key in entry_keys:
2689
# there is already an entry where this data belongs, just
2691
by_path[new_entry_key][tree_index] = \
2692
self._inv_entry_to_details(entry)
2694
# add relocated entries to the horizontal axis - this row
2695
# mapping from path,id. We need to look up the correct path
2696
# for the indexes from 0 to tree_index -1
2698
for lookup_index in xrange(tree_index):
2699
# boundary case: this is the first occurence of file_id
2700
# so there are no id_indexes, possibly take this out of
2702
if not len(entry_keys):
2703
new_details.append(DirState.NULL_PARENT_DETAILS)
2705
# grab any one entry, use it to find the right path.
2706
a_key = iter(entry_keys).next()
2707
if by_path[a_key][lookup_index][0] in ('r', 'a'):
2708
# its a pointer or missing statement, use it as
2710
new_details.append(by_path[a_key][lookup_index])
2712
# we have the right key, make a pointer to it.
2713
real_path = ('/'.join(a_key[0:2])).strip('/')
2714
new_details.append(st('r', real_path, 0, False,
2716
new_details.append(self._inv_entry_to_details(entry))
2717
new_details.extend(new_location_suffix)
2718
by_path[new_entry_key] = new_details
2719
self._add_to_id_index(id_index, new_entry_key)
2720
# --- end generation of full tree mappings
2722
# sort and output all the entries
2723
new_entries = self._sort_entries(by_path.items())
2724
self._entries_to_current_state(new_entries)
2725
self._parents = [rev_id for rev_id, tree in trees]
2726
self._ghosts = list(ghosts)
2727
self._mark_modified(header_modified=True)
2728
self._id_index = id_index
2730
def _sort_entries(self, entry_list):
2731
"""Given a list of entries, sort them into the right order.
2733
This is done when constructing a new dirstate from trees - normally we
2734
try to keep everything in sorted blocks all the time, but sometimes
2735
it's easier to sort after the fact.
2737
# When sorting, we usually have 10x more entries than directories. (69k
2738
# total entries, 4k directories). So cache the results of splitting.
2739
# Saving time and objects. Also, use StaticTuple to avoid putting all
2740
# of these object into python's garbage collector.
2742
def _key(entry, _split_dirs=split_dirs, _st=static_tuple.StaticTuple):
2743
# sort by: directory parts, file name, file id
2744
dirpath, fname, file_id = entry[0]
2746
split = _split_dirs[dirpath]
2748
split = _st.from_sequence(dirpath.split('/'))
2749
_split_dirs[dirpath] = split
2750
return _st(split, fname, file_id)
2751
return sorted(entry_list, key=_key)
2753
def set_state_from_inventory(self, new_inv):
2754
"""Set new_inv as the current state.
2756
This API is called by tree transform, and will usually occur with
2757
existing parent trees.
2759
:param new_inv: The inventory object to set current state from.
2761
if 'evil' in debug.debug_flags:
2762
trace.mutter_callsite(1,
2763
"set_state_from_inventory called; please mutate the tree instead")
2764
tracing = 'dirstate' in debug.debug_flags
2766
trace.mutter("set_state_from_inventory trace:")
2767
self._read_dirblocks_if_needed()
2769
# Two iterators: current data and new data, both in dirblock order.
2770
# We zip them together, which tells about entries that are new in the
2771
# inventory, or removed in the inventory, or present in both and
2774
# You might think we could just synthesize a new dirstate directly
2775
# since we're processing it in the right order. However, we need to
2776
# also consider there may be any number of parent trees and relocation
2777
# pointers, and we don't want to duplicate that here.
2778
new_iterator = new_inv.iter_entries_by_dir()
2779
# we will be modifying the dirstate, so we need a stable iterator. In
2780
# future we might write one, for now we just clone the state into a
2781
# list using a copy so that we see every original item and don't have
2782
# to adjust the position when items are inserted or deleted in the
2783
# underlying dirstate.
2784
old_iterator = iter(list(self._iter_entries()))
2785
# both must have roots so this is safe:
2786
current_new = new_iterator.next()
2787
current_old = old_iterator.next()
2788
def advance(iterator):
2790
return iterator.next()
2791
except StopIteration:
2793
while current_new or current_old:
2794
# skip entries in old that are not really there
2795
if current_old and current_old[1][0][0] in 'ar':
2796
# relocated or absent
2797
current_old = advance(old_iterator)
2800
# convert new into dirblock style
2801
new_path_utf8 = current_new[0].encode('utf8')
2802
new_dirname, new_basename = osutils.split(new_path_utf8)
2803
new_id = current_new[1].file_id
2804
new_entry_key = (new_dirname, new_basename, new_id)
2805
current_new_minikind = \
2806
DirState._kind_to_minikind[current_new[1].kind]
2807
if current_new_minikind == 't':
2808
fingerprint = current_new[1].reference_revision or ''
2810
# We normally only insert or remove records, or update
2811
# them when it has significantly changed. Then we want to
2812
# erase its fingerprint. Unaffected records should
2813
# normally not be updated at all.
2816
# for safety disable variables
2817
new_path_utf8 = new_dirname = new_basename = new_id = \
2818
new_entry_key = None
2819
# 5 cases, we dont have a value that is strictly greater than everything, so
2820
# we make both end conditions explicit
2822
# old is finished: insert current_new into the state.
2824
trace.mutter("Appending from new '%s'.",
2825
new_path_utf8.decode('utf8'))
2826
self.update_minimal(new_entry_key, current_new_minikind,
2827
executable=current_new[1].executable,
2828
path_utf8=new_path_utf8, fingerprint=fingerprint,
2830
current_new = advance(new_iterator)
2831
elif not current_new:
2834
trace.mutter("Truncating from old '%s/%s'.",
2835
current_old[0][0].decode('utf8'),
2836
current_old[0][1].decode('utf8'))
2837
self._make_absent(current_old)
2838
current_old = advance(old_iterator)
2839
elif new_entry_key == current_old[0]:
2840
# same - common case
2841
# We're looking at the same path and id in both the dirstate
2842
# and inventory, so just need to update the fields in the
2843
# dirstate from the one in the inventory.
2844
# TODO: update the record if anything significant has changed.
2845
# the minimal required trigger is if the execute bit or cached
2847
if (current_old[1][0][3] != current_new[1].executable or
2848
current_old[1][0][0] != current_new_minikind):
2850
trace.mutter("Updating in-place change '%s'.",
2851
new_path_utf8.decode('utf8'))
2852
self.update_minimal(current_old[0], current_new_minikind,
2853
executable=current_new[1].executable,
2854
path_utf8=new_path_utf8, fingerprint=fingerprint,
2856
# both sides are dealt with, move on
2857
current_old = advance(old_iterator)
2858
current_new = advance(new_iterator)
2859
elif (cmp_by_dirs(new_dirname, current_old[0][0]) < 0
2860
or (new_dirname == current_old[0][0]
2861
and new_entry_key[1:] < current_old[0][1:])):
2863
# add a entry for this and advance new
2865
trace.mutter("Inserting from new '%s'.",
2866
new_path_utf8.decode('utf8'))
2867
self.update_minimal(new_entry_key, current_new_minikind,
2868
executable=current_new[1].executable,
2869
path_utf8=new_path_utf8, fingerprint=fingerprint,
2871
current_new = advance(new_iterator)
2873
# we've advanced past the place where the old key would be,
2874
# without seeing it in the new list. so it must be gone.
2876
trace.mutter("Deleting from old '%s/%s'.",
2877
current_old[0][0].decode('utf8'),
2878
current_old[0][1].decode('utf8'))
2879
self._make_absent(current_old)
2880
current_old = advance(old_iterator)
2881
self._mark_modified()
2882
self._id_index = None
2883
self._packed_stat_index = None
2885
trace.mutter("set_state_from_inventory complete.")
2887
def set_state_from_scratch(self, working_inv, parent_trees, parent_ghosts):
2888
"""Wipe the currently stored state and set it to something new.
2890
This is a hard-reset for the data we are working with.
2892
# Technically, we really want a write lock, but until we write, we
2893
# don't really need it.
2894
self._requires_lock()
2895
# root dir and root dir contents with no children. We have to have a
2896
# root for set_state_from_inventory to work correctly.
2897
empty_root = (('', '', inventory.ROOT_ID),
2898
[('d', '', 0, False, DirState.NULLSTAT)])
2899
empty_tree_dirblocks = [('', [empty_root]), ('', [])]
2900
self._set_data([], empty_tree_dirblocks)
2901
self.set_state_from_inventory(working_inv)
2902
self.set_parent_trees(parent_trees, parent_ghosts)
2904
def _make_absent(self, current_old):
2905
"""Mark current_old - an entry - as absent for tree 0.
2907
:return: True if this was the last details entry for the entry key:
2908
that is, if the underlying block has had the entry removed, thus
2909
shrinking in length.
2911
# build up paths that this id will be left at after the change is made,
2912
# so we can update their cross references in tree 0
2913
all_remaining_keys = set()
2914
# Dont check the working tree, because it's going.
2915
for details in current_old[1][1:]:
2916
if details[0] not in 'ar': # absent, relocated
2917
all_remaining_keys.add(current_old[0])
2918
elif details[0] == 'r': # relocated
2919
# record the key for the real path.
2920
all_remaining_keys.add(tuple(osutils.split(details[1])) + (current_old[0][2],))
2921
# absent rows are not present at any path.
2922
last_reference = current_old[0] not in all_remaining_keys
2924
# the current row consists entire of the current item (being marked
2925
# absent), and relocated or absent entries for the other trees:
2926
# Remove it, its meaningless.
2927
block = self._find_block(current_old[0])
2928
entry_index, present = self._find_entry_index(current_old[0], block[1])
2930
raise AssertionError('could not find entry for %s' % (current_old,))
2931
block[1].pop(entry_index)
2932
# if we have an id_index in use, remove this key from it for this id.
2933
if self._id_index is not None:
2934
self._remove_from_id_index(self._id_index, current_old[0])
2935
# update all remaining keys for this id to record it as absent. The
2936
# existing details may either be the record we are marking as deleted
2937
# (if there were other trees with the id present at this path), or may
2939
for update_key in all_remaining_keys:
2940
update_block_index, present = \
2941
self._find_block_index_from_key(update_key)
2943
raise AssertionError('could not find block for %s' % (update_key,))
2944
update_entry_index, present = \
2945
self._find_entry_index(update_key, self._dirblocks[update_block_index][1])
2947
raise AssertionError('could not find entry for %s' % (update_key,))
2948
update_tree_details = self._dirblocks[update_block_index][1][update_entry_index][1]
2949
# it must not be absent at the moment
2950
if update_tree_details[0][0] == 'a': # absent
2951
raise AssertionError('bad row %r' % (update_tree_details,))
2952
update_tree_details[0] = DirState.NULL_PARENT_DETAILS
2953
self._mark_modified()
2954
return last_reference
2956
def update_minimal(self, key, minikind, executable=False, fingerprint='',
2957
packed_stat=None, size=0, path_utf8=None, fullscan=False):
2958
"""Update an entry to the state in tree 0.
2960
This will either create a new entry at 'key' or update an existing one.
2961
It also makes sure that any other records which might mention this are
2964
:param key: (dir, name, file_id) for the new entry
2965
:param minikind: The type for the entry ('f' == 'file', 'd' ==
2967
:param executable: Should the executable bit be set?
2968
:param fingerprint: Simple fingerprint for new entry: canonical-form
2969
sha1 for files, referenced revision id for subtrees, etc.
2970
:param packed_stat: Packed stat value for new entry.
2971
:param size: Size information for new entry
2972
:param path_utf8: key[0] + '/' + key[1], just passed in to avoid doing
2974
:param fullscan: If True then a complete scan of the dirstate is being
2975
done and checking for duplicate rows should not be done. This
2976
should only be set by set_state_from_inventory and similar methods.
2978
If packed_stat and fingerprint are not given, they're invalidated in
2981
block = self._find_block(key)[1]
2982
if packed_stat is None:
2983
packed_stat = DirState.NULLSTAT
2984
# XXX: Some callers pass '' as the packed_stat, and it seems to be
2985
# sometimes present in the dirstate - this seems oddly inconsistent.
2987
entry_index, present = self._find_entry_index(key, block)
2988
new_details = (minikind, fingerprint, size, executable, packed_stat)
2989
id_index = self._get_id_index()
2991
# New record. Check there isn't a entry at this path already.
2993
low_index, _ = self._find_entry_index(key[0:2] + ('',), block)
2994
while low_index < len(block):
2995
entry = block[low_index]
2996
if entry[0][0:2] == key[0:2]:
2997
if entry[1][0][0] not in 'ar':
2998
# This entry has the same path (but a different id) as
2999
# the new entry we're adding, and is present in ths
3001
self._raise_invalid(
3002
("%s/%s" % key[0:2]).decode('utf8'), key[2],
3003
"Attempt to add item at path already occupied by "
3004
"id %r" % entry[0][2])
3008
# new entry, synthesis cross reference here,
3009
existing_keys = id_index.get(key[2], ())
3010
if not existing_keys:
3011
# not currently in the state, simplest case
3012
new_entry = key, [new_details] + self._empty_parent_info()
3014
# present at one or more existing other paths.
3015
# grab one of them and use it to generate parent
3016
# relocation/absent entries.
3017
new_entry = key, [new_details]
3018
# existing_keys can be changed as we iterate.
3019
for other_key in tuple(existing_keys):
3020
# change the record at other to be a pointer to this new
3021
# record. The loop looks similar to the change to
3022
# relocations when updating an existing record but its not:
3023
# the test for existing kinds is different: this can be
3024
# factored out to a helper though.
3025
other_block_index, present = self._find_block_index_from_key(
3028
raise AssertionError('could not find block for %s' % (
3030
other_block = self._dirblocks[other_block_index][1]
3031
other_entry_index, present = self._find_entry_index(
3032
other_key, other_block)
3034
raise AssertionError(
3035
'update_minimal: could not find other entry for %s'
3037
if path_utf8 is None:
3038
raise AssertionError('no path')
3039
# Turn this other location into a reference to the new
3040
# location. This also updates the aliased iterator
3041
# (current_old in set_state_from_inventory) so that the old
3042
# entry, if not already examined, is skipped over by that
3044
other_entry = other_block[other_entry_index]
3045
other_entry[1][0] = ('r', path_utf8, 0, False, '')
3046
if self._maybe_remove_row(other_block, other_entry_index,
3048
# If the row holding this was removed, we need to
3049
# recompute where this entry goes
3050
entry_index, _ = self._find_entry_index(key, block)
3053
# adds a tuple to the new details for each column
3054
# - either by copying an existing relocation pointer inside that column
3055
# - or by creating a new pointer to the right row inside that column
3056
num_present_parents = self._num_present_parents()
3057
if num_present_parents:
3058
# TODO: This re-evaluates the existing_keys set, do we need
3059
# to do that ourselves?
3060
other_key = list(existing_keys)[0]
3061
for lookup_index in xrange(1, num_present_parents + 1):
3062
# grab any one entry, use it to find the right path.
3063
# TODO: optimise this to reduce memory use in highly
3064
# fragmented situations by reusing the relocation
3066
update_block_index, present = \
3067
self._find_block_index_from_key(other_key)
3069
raise AssertionError('could not find block for %s' % (other_key,))
3070
update_entry_index, present = \
3071
self._find_entry_index(other_key, self._dirblocks[update_block_index][1])
3073
raise AssertionError('update_minimal: could not find entry for %s' % (other_key,))
3074
update_details = self._dirblocks[update_block_index][1][update_entry_index][1][lookup_index]
3075
if update_details[0] in 'ar': # relocated, absent
3076
# its a pointer or absent in lookup_index's tree, use
3078
new_entry[1].append(update_details)
3080
# we have the right key, make a pointer to it.
3081
pointer_path = osutils.pathjoin(*other_key[0:2])
3082
new_entry[1].append(('r', pointer_path, 0, False, ''))
3083
block.insert(entry_index, new_entry)
3084
self._add_to_id_index(id_index, key)
3086
# Does the new state matter?
3087
block[entry_index][1][0] = new_details
3088
# parents cannot be affected by what we do.
3089
# other occurences of this id can be found
3090
# from the id index.
3092
# tree index consistency: All other paths for this id in this tree
3093
# index must point to the correct path. We have to loop here because
3094
# we may have passed entries in the state with this file id already
3095
# that were absent - where parent entries are - and they need to be
3096
# converted to relocated.
3097
if path_utf8 is None:
3098
raise AssertionError('no path')
3099
existing_keys = id_index.get(key[2], ())
3100
if key not in existing_keys:
3101
raise AssertionError('We found the entry in the blocks, but'
3102
' the key is not in the id_index.'
3103
' key: %s, existing_keys: %s' % (key, existing_keys))
3104
for entry_key in existing_keys:
3105
# TODO:PROFILING: It might be faster to just update
3106
# rather than checking if we need to, and then overwrite
3107
# the one we are located at.
3108
if entry_key != key:
3109
# this file id is at a different path in one of the
3110
# other trees, so put absent pointers there
3111
# This is the vertical axis in the matrix, all pointing
3113
block_index, present = self._find_block_index_from_key(entry_key)
3115
raise AssertionError('not present: %r', entry_key)
3116
entry_index, present = self._find_entry_index(entry_key, self._dirblocks[block_index][1])
3118
raise AssertionError('not present: %r', entry_key)
3119
self._dirblocks[block_index][1][entry_index][1][0] = \
3120
('r', path_utf8, 0, False, '')
3121
# add a containing dirblock if needed.
3122
if new_details[0] == 'd':
3123
subdir_key = (osutils.pathjoin(*key[0:2]), '', '')
3124
block_index, present = self._find_block_index_from_key(subdir_key)
3126
self._dirblocks.insert(block_index, (subdir_key[0], []))
3128
self._mark_modified()
3130
def _maybe_remove_row(self, block, index, id_index):
3131
"""Remove index if it is absent or relocated across the row.
3133
id_index is updated accordingly.
3134
:return: True if we removed the row, False otherwise
3136
present_in_row = False
3137
entry = block[index]
3138
for column in entry[1]:
3139
if column[0] not in 'ar':
3140
present_in_row = True
3142
if not present_in_row:
3144
self._remove_from_id_index(id_index, entry[0])
3148
def _validate(self):
3149
"""Check that invariants on the dirblock are correct.
3151
This can be useful in debugging; it shouldn't be necessary in
3154
This must be called with a lock held.
3156
# NOTE: This must always raise AssertionError not just assert,
3157
# otherwise it may not behave properly under python -O
3159
# TODO: All entries must have some content that's not 'a' or 'r',
3160
# otherwise it could just be removed.
3162
# TODO: All relocations must point directly to a real entry.
3164
# TODO: No repeated keys.
3167
from pprint import pformat
3168
self._read_dirblocks_if_needed()
3169
if len(self._dirblocks) > 0:
3170
if not self._dirblocks[0][0] == '':
3171
raise AssertionError(
3172
"dirblocks don't start with root block:\n" + \
3173
pformat(self._dirblocks))
3174
if len(self._dirblocks) > 1:
3175
if not self._dirblocks[1][0] == '':
3176
raise AssertionError(
3177
"dirblocks missing root directory:\n" + \
3178
pformat(self._dirblocks))
3179
# the dirblocks are sorted by their path components, name, and dir id
3180
dir_names = [d[0].split('/')
3181
for d in self._dirblocks[1:]]
3182
if dir_names != sorted(dir_names):
3183
raise AssertionError(
3184
"dir names are not in sorted order:\n" + \
3185
pformat(self._dirblocks) + \
3188
for dirblock in self._dirblocks:
3189
# within each dirblock, the entries are sorted by filename and
3191
for entry in dirblock[1]:
3192
if dirblock[0] != entry[0][0]:
3193
raise AssertionError(
3195
"doesn't match directory name in\n%r" %
3196
(entry, pformat(dirblock)))
3197
if dirblock[1] != sorted(dirblock[1]):
3198
raise AssertionError(
3199
"dirblock for %r is not sorted:\n%s" % \
3200
(dirblock[0], pformat(dirblock)))
3202
def check_valid_parent():
3203
"""Check that the current entry has a valid parent.
3205
This makes sure that the parent has a record,
3206
and that the parent isn't marked as "absent" in the
3207
current tree. (It is invalid to have a non-absent file in an absent
3210
if entry[0][0:2] == ('', ''):
3211
# There should be no parent for the root row
3213
parent_entry = self._get_entry(tree_index, path_utf8=entry[0][0])
3214
if parent_entry == (None, None):
3215
raise AssertionError(
3216
"no parent entry for: %s in tree %s"
3217
% (this_path, tree_index))
3218
if parent_entry[1][tree_index][0] != 'd':
3219
raise AssertionError(
3220
"Parent entry for %s is not marked as a valid"
3221
" directory. %s" % (this_path, parent_entry,))
3223
# For each file id, for each tree: either
3224
# the file id is not present at all; all rows with that id in the
3225
# key have it marked as 'absent'
3226
# OR the file id is present under exactly one name; any other entries
3227
# that mention that id point to the correct name.
3229
# We check this with a dict per tree pointing either to the present
3230
# name, or None if absent.
3231
tree_count = self._num_present_parents() + 1
3232
id_path_maps = [dict() for i in range(tree_count)]
3233
# Make sure that all renamed entries point to the correct location.
3234
for entry in self._iter_entries():
3235
file_id = entry[0][2]
3236
this_path = osutils.pathjoin(entry[0][0], entry[0][1])
3237
if len(entry[1]) != tree_count:
3238
raise AssertionError(
3239
"wrong number of entry details for row\n%s" \
3240
",\nexpected %d" % \
3241
(pformat(entry), tree_count))
3242
absent_positions = 0
3243
for tree_index, tree_state in enumerate(entry[1]):
3244
this_tree_map = id_path_maps[tree_index]
3245
minikind = tree_state[0]
3246
if minikind in 'ar':
3247
absent_positions += 1
3248
# have we seen this id before in this column?
3249
if file_id in this_tree_map:
3250
previous_path, previous_loc = this_tree_map[file_id]
3251
# any later mention of this file must be consistent with
3252
# what was said before
3254
if previous_path is not None:
3255
raise AssertionError(
3256
"file %s is absent in row %r but also present " \
3258
(file_id, entry, previous_path))
3259
elif minikind == 'r':
3260
target_location = tree_state[1]
3261
if previous_path != target_location:
3262
raise AssertionError(
3263
"file %s relocation in row %r but also at %r" \
3264
% (file_id, entry, previous_path))
3266
# a file, directory, etc - may have been previously
3267
# pointed to by a relocation, which must point here
3268
if previous_path != this_path:
3269
raise AssertionError(
3270
"entry %r inconsistent with previous path %r "
3272
(entry, previous_path, previous_loc))
3273
check_valid_parent()
3276
# absent; should not occur anywhere else
3277
this_tree_map[file_id] = None, this_path
3278
elif minikind == 'r':
3279
# relocation, must occur at expected location
3280
this_tree_map[file_id] = tree_state[1], this_path
3282
this_tree_map[file_id] = this_path, this_path
3283
check_valid_parent()
3284
if absent_positions == tree_count:
3285
raise AssertionError(
3286
"entry %r has no data for any tree." % (entry,))
3287
if self._id_index is not None:
3288
for file_id, entry_keys in self._id_index.iteritems():
3289
for entry_key in entry_keys:
3290
# Check that the entry in the map is pointing to the same
3292
if entry_key[2] != file_id:
3293
raise AssertionError(
3294
'file_id %r did not match entry key %s'
3295
% (file_id, entry_key))
3296
# And that from this entry key, we can look up the original
3298
block_index, present = self._find_block_index_from_key(entry_key)
3300
raise AssertionError('missing block for entry key: %r', entry_key)
3301
entry_index, present = self._find_entry_index(entry_key, self._dirblocks[block_index][1])
3303
raise AssertionError('missing entry for key: %r', entry_key)
3304
if len(entry_keys) != len(set(entry_keys)):
3305
raise AssertionError(
3306
'id_index contained non-unique data for %s'
3309
def _wipe_state(self):
3310
"""Forget all state information about the dirstate."""
3311
self._header_state = DirState.NOT_IN_MEMORY
3312
self._dirblock_state = DirState.NOT_IN_MEMORY
3313
self._changes_aborted = False
3316
self._dirblocks = []
3317
self._id_index = None
3318
self._packed_stat_index = None
3319
self._end_of_header = None
3320
self._cutoff_time = None
3321
self._split_path_cache = {}
3323
def lock_read(self):
3324
"""Acquire a read lock on the dirstate."""
3325
if self._lock_token is not None:
3326
raise errors.LockContention(self._lock_token)
3327
# TODO: jam 20070301 Rather than wiping completely, if the blocks are
3328
# already in memory, we could read just the header and check for
3329
# any modification. If not modified, we can just leave things
3331
self._lock_token = lock.ReadLock(self._filename)
3332
self._lock_state = 'r'
3333
self._state_file = self._lock_token.f
3336
def lock_write(self):
3337
"""Acquire a write lock on the dirstate."""
3338
if self._lock_token is not None:
3339
raise errors.LockContention(self._lock_token)
3340
# TODO: jam 20070301 Rather than wiping completely, if the blocks are
3341
# already in memory, we could read just the header and check for
3342
# any modification. If not modified, we can just leave things
3344
self._lock_token = lock.WriteLock(self._filename)
3345
self._lock_state = 'w'
3346
self._state_file = self._lock_token.f
3350
"""Drop any locks held on the dirstate."""
3351
if self._lock_token is None:
3352
raise errors.LockNotHeld(self)
3353
# TODO: jam 20070301 Rather than wiping completely, if the blocks are
3354
# already in memory, we could read just the header and check for
3355
# any modification. If not modified, we can just leave things
3357
self._state_file = None
3358
self._lock_state = None
3359
self._lock_token.unlock()
3360
self._lock_token = None
3361
self._split_path_cache = {}
3363
def _requires_lock(self):
3364
"""Check that a lock is currently held by someone on the dirstate."""
3365
if not self._lock_token:
3366
raise errors.ObjectNotLocked(self)
3369
def py_update_entry(state, entry, abspath, stat_value,
3370
_stat_to_minikind=DirState._stat_to_minikind):
3371
"""Update the entry based on what is actually on disk.
3373
This function only calculates the sha if it needs to - if the entry is
3374
uncachable, or clearly different to the first parent's entry, no sha
3375
is calculated, and None is returned.
3377
:param state: The dirstate this entry is in.
3378
:param entry: This is the dirblock entry for the file in question.
3379
:param abspath: The path on disk for this file.
3380
:param stat_value: The stat value done on the path.
3381
:return: None, or The sha1 hexdigest of the file (40 bytes) or link
3382
target of a symlink.
3385
minikind = _stat_to_minikind[stat_value.st_mode & 0170000]
3389
packed_stat = pack_stat(stat_value)
3390
(saved_minikind, saved_link_or_sha1, saved_file_size,
3391
saved_executable, saved_packed_stat) = entry[1][0]
3393
if minikind == 'd' and saved_minikind == 't':
3395
if (minikind == saved_minikind
3396
and packed_stat == saved_packed_stat):
3397
# The stat hasn't changed since we saved, so we can re-use the
3402
# size should also be in packed_stat
3403
if saved_file_size == stat_value.st_size:
3404
return saved_link_or_sha1
3406
# If we have gotten this far, that means that we need to actually
3407
# process this entry.
3411
executable = state._is_executable(stat_value.st_mode,
3413
if state._cutoff_time is None:
3414
state._sha_cutoff_time()
3415
if (stat_value.st_mtime < state._cutoff_time
3416
and stat_value.st_ctime < state._cutoff_time
3417
and len(entry[1]) > 1
3418
and entry[1][1][0] != 'a'):
3419
# Could check for size changes for further optimised
3420
# avoidance of sha1's. However the most prominent case of
3421
# over-shaing is during initial add, which this catches.
3422
# Besides, if content filtering happens, size and sha
3423
# are calculated at the same time, so checking just the size
3424
# gains nothing w.r.t. performance.
3425
link_or_sha1 = state._sha1_file(abspath)
3426
entry[1][0] = ('f', link_or_sha1, stat_value.st_size,
3427
executable, packed_stat)
3429
entry[1][0] = ('f', '', stat_value.st_size,
3430
executable, DirState.NULLSTAT)
3431
worth_saving = False
3432
elif minikind == 'd':
3434
entry[1][0] = ('d', '', 0, False, packed_stat)
3435
if saved_minikind != 'd':
3436
# This changed from something into a directory. Make sure we
3437
# have a directory block for it. This doesn't happen very
3438
# often, so this doesn't have to be super fast.
3439
block_index, entry_index, dir_present, file_present = \
3440
state._get_block_entry_index(entry[0][0], entry[0][1], 0)
3441
state._ensure_block(block_index, entry_index,
3442
osutils.pathjoin(entry[0][0], entry[0][1]))
3444
worth_saving = False
3445
elif minikind == 'l':
3446
if saved_minikind == 'l':
3447
worth_saving = False
3448
link_or_sha1 = state._read_link(abspath, saved_link_or_sha1)
3449
if state._cutoff_time is None:
3450
state._sha_cutoff_time()
3451
if (stat_value.st_mtime < state._cutoff_time
3452
and stat_value.st_ctime < state._cutoff_time):
3453
entry[1][0] = ('l', link_or_sha1, stat_value.st_size,
3456
entry[1][0] = ('l', '', stat_value.st_size,
3457
False, DirState.NULLSTAT)
3459
state._mark_modified([entry])
3463
class ProcessEntryPython(object):
3465
__slots__ = ["old_dirname_to_file_id", "new_dirname_to_file_id",
3466
"last_source_parent", "last_target_parent", "include_unchanged",
3467
"partial", "use_filesystem_for_exec", "utf8_decode",
3468
"searched_specific_files", "search_specific_files",
3469
"searched_exact_paths", "search_specific_file_parents", "seen_ids",
3470
"state", "source_index", "target_index", "want_unversioned", "tree"]
3472
def __init__(self, include_unchanged, use_filesystem_for_exec,
3473
search_specific_files, state, source_index, target_index,
3474
want_unversioned, tree):
3475
self.old_dirname_to_file_id = {}
3476
self.new_dirname_to_file_id = {}
3477
# Are we doing a partial iter_changes?
3478
self.partial = search_specific_files != set([''])
3479
# Using a list so that we can access the values and change them in
3480
# nested scope. Each one is [path, file_id, entry]
3481
self.last_source_parent = [None, None]
3482
self.last_target_parent = [None, None]
3483
self.include_unchanged = include_unchanged
3484
self.use_filesystem_for_exec = use_filesystem_for_exec
3485
self.utf8_decode = cache_utf8._utf8_decode
3486
# for all search_indexs in each path at or under each element of
3487
# search_specific_files, if the detail is relocated: add the id, and
3488
# add the relocated path as one to search if its not searched already.
3489
# If the detail is not relocated, add the id.
3490
self.searched_specific_files = set()
3491
# When we search exact paths without expanding downwards, we record
3493
self.searched_exact_paths = set()
3494
self.search_specific_files = search_specific_files
3495
# The parents up to the root of the paths we are searching.
3496
# After all normal paths are returned, these specific items are returned.
3497
self.search_specific_file_parents = set()
3498
# The ids we've sent out in the delta.
3499
self.seen_ids = set()
3501
self.source_index = source_index
3502
self.target_index = target_index
3503
if target_index != 0:
3504
# A lot of code in here depends on target_index == 0
3505
raise errors.BzrError('unsupported target index')
3506
self.want_unversioned = want_unversioned
3509
def _process_entry(self, entry, path_info, pathjoin=osutils.pathjoin):
3510
"""Compare an entry and real disk to generate delta information.
3512
:param path_info: top_relpath, basename, kind, lstat, abspath for
3513
the path of entry. If None, then the path is considered absent in
3514
the target (Perhaps we should pass in a concrete entry for this ?)
3515
Basename is returned as a utf8 string because we expect this
3516
tuple will be ignored, and don't want to take the time to
3518
:return: (iter_changes_result, changed). If the entry has not been
3519
handled then changed is None. Otherwise it is False if no content
3520
or metadata changes have occurred, and True if any content or
3521
metadata change has occurred. If self.include_unchanged is True then
3522
if changed is not None, iter_changes_result will always be a result
3523
tuple. Otherwise, iter_changes_result is None unless changed is
3526
if self.source_index is None:
3527
source_details = DirState.NULL_PARENT_DETAILS
3529
source_details = entry[1][self.source_index]
3530
target_details = entry[1][self.target_index]
3531
target_minikind = target_details[0]
3532
if path_info is not None and target_minikind in 'fdlt':
3533
if not (self.target_index == 0):
3534
raise AssertionError()
3535
link_or_sha1 = update_entry(self.state, entry,
3536
abspath=path_info[4], stat_value=path_info[3])
3537
# The entry may have been modified by update_entry
3538
target_details = entry[1][self.target_index]
3539
target_minikind = target_details[0]
3542
file_id = entry[0][2]
3543
source_minikind = source_details[0]
3544
if source_minikind in 'fdltr' and target_minikind in 'fdlt':
3545
# claimed content in both: diff
3546
# r | fdlt | | add source to search, add id path move and perform
3547
# | | | diff check on source-target
3548
# r | fdlt | a | dangling file that was present in the basis.
3550
if source_minikind in 'r':
3551
# add the source to the search path to find any children it
3552
# has. TODO ? : only add if it is a container ?
3553
if not osutils.is_inside_any(self.searched_specific_files,
3555
self.search_specific_files.add(source_details[1])
3556
# generate the old path; this is needed for stating later
3558
old_path = source_details[1]
3559
old_dirname, old_basename = os.path.split(old_path)
3560
path = pathjoin(entry[0][0], entry[0][1])
3561
old_entry = self.state._get_entry(self.source_index,
3563
# update the source details variable to be the real
3565
if old_entry == (None, None):
3566
raise errors.CorruptDirstate(self.state._filename,
3567
"entry '%s/%s' is considered renamed from %r"
3568
" but source does not exist\n"
3569
"entry: %s" % (entry[0][0], entry[0][1], old_path, entry))
3570
source_details = old_entry[1][self.source_index]
3571
source_minikind = source_details[0]
3573
old_dirname = entry[0][0]
3574
old_basename = entry[0][1]
3575
old_path = path = None
3576
if path_info is None:
3577
# the file is missing on disk, show as removed.
3578
content_change = True
3582
# source and target are both versioned and disk file is present.
3583
target_kind = path_info[2]
3584
if target_kind == 'directory':
3586
old_path = path = pathjoin(old_dirname, old_basename)
3587
self.new_dirname_to_file_id[path] = file_id
3588
if source_minikind != 'd':
3589
content_change = True
3591
# directories have no fingerprint
3592
content_change = False
3594
elif target_kind == 'file':
3595
if source_minikind != 'f':
3596
content_change = True
3598
# Check the sha. We can't just rely on the size as
3599
# content filtering may mean differ sizes actually
3600
# map to the same content
3601
if link_or_sha1 is None:
3603
statvalue, link_or_sha1 = \
3604
self.state._sha1_provider.stat_and_sha1(
3606
self.state._observed_sha1(entry, link_or_sha1,
3608
content_change = (link_or_sha1 != source_details[1])
3609
# Target details is updated at update_entry time
3610
if self.use_filesystem_for_exec:
3611
# We don't need S_ISREG here, because we are sure
3612
# we are dealing with a file.
3613
target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
3615
target_exec = target_details[3]
3616
elif target_kind == 'symlink':
3617
if source_minikind != 'l':
3618
content_change = True
3620
content_change = (link_or_sha1 != source_details[1])
3622
elif target_kind == 'tree-reference':
3623
if source_minikind != 't':
3624
content_change = True
3626
content_change = False
3630
path = pathjoin(old_dirname, old_basename)
3631
raise errors.BadFileKindError(path, path_info[2])
3632
if source_minikind == 'd':
3634
old_path = path = pathjoin(old_dirname, old_basename)
3635
self.old_dirname_to_file_id[old_path] = file_id
3636
# parent id is the entry for the path in the target tree
3637
if old_basename and old_dirname == self.last_source_parent[0]:
3638
source_parent_id = self.last_source_parent[1]
3641
source_parent_id = self.old_dirname_to_file_id[old_dirname]
3643
source_parent_entry = self.state._get_entry(self.source_index,
3644
path_utf8=old_dirname)
3645
source_parent_id = source_parent_entry[0][2]
3646
if source_parent_id == entry[0][2]:
3647
# This is the root, so the parent is None
3648
source_parent_id = None
3650
self.last_source_parent[0] = old_dirname
3651
self.last_source_parent[1] = source_parent_id
3652
new_dirname = entry[0][0]
3653
if entry[0][1] and new_dirname == self.last_target_parent[0]:
3654
target_parent_id = self.last_target_parent[1]
3657
target_parent_id = self.new_dirname_to_file_id[new_dirname]
3659
# TODO: We don't always need to do the lookup, because the
3660
# parent entry will be the same as the source entry.
3661
target_parent_entry = self.state._get_entry(self.target_index,
3662
path_utf8=new_dirname)
3663
if target_parent_entry == (None, None):
3664
raise AssertionError(
3665
"Could not find target parent in wt: %s\nparent of: %s"
3666
% (new_dirname, entry))
3667
target_parent_id = target_parent_entry[0][2]
3668
if target_parent_id == entry[0][2]:
3669
# This is the root, so the parent is None
3670
target_parent_id = None
3672
self.last_target_parent[0] = new_dirname
3673
self.last_target_parent[1] = target_parent_id
3675
source_exec = source_details[3]
3676
changed = (content_change
3677
or source_parent_id != target_parent_id
3678
or old_basename != entry[0][1]
3679
or source_exec != target_exec
3681
if not changed and not self.include_unchanged:
3684
if old_path is None:
3685
old_path = path = pathjoin(old_dirname, old_basename)
3686
old_path_u = self.utf8_decode(old_path)[0]
3689
old_path_u = self.utf8_decode(old_path)[0]
3690
if old_path == path:
3693
path_u = self.utf8_decode(path)[0]
3694
source_kind = DirState._minikind_to_kind[source_minikind]
3695
return (entry[0][2],
3696
(old_path_u, path_u),
3699
(source_parent_id, target_parent_id),
3700
(self.utf8_decode(old_basename)[0], self.utf8_decode(entry[0][1])[0]),
3701
(source_kind, target_kind),
3702
(source_exec, target_exec)), changed
3703
elif source_minikind in 'a' and target_minikind in 'fdlt':
3704
# looks like a new file
3705
path = pathjoin(entry[0][0], entry[0][1])
3706
# parent id is the entry for the path in the target tree
3707
# TODO: these are the same for an entire directory: cache em.
3708
parent_id = self.state._get_entry(self.target_index,
3709
path_utf8=entry[0][0])[0][2]
3710
if parent_id == entry[0][2]:
3712
if path_info is not None:
3714
if self.use_filesystem_for_exec:
3715
# We need S_ISREG here, because we aren't sure if this
3718
stat.S_ISREG(path_info[3].st_mode)
3719
and stat.S_IEXEC & path_info[3].st_mode)
3721
target_exec = target_details[3]
3722
return (entry[0][2],
3723
(None, self.utf8_decode(path)[0]),
3727
(None, self.utf8_decode(entry[0][1])[0]),
3728
(None, path_info[2]),
3729
(None, target_exec)), True
3731
# Its a missing file, report it as such.
3732
return (entry[0][2],
3733
(None, self.utf8_decode(path)[0]),
3737
(None, self.utf8_decode(entry[0][1])[0]),
3739
(None, False)), True
3740
elif source_minikind in 'fdlt' and target_minikind in 'a':
3741
# unversioned, possibly, or possibly not deleted: we dont care.
3742
# if its still on disk, *and* theres no other entry at this
3743
# path [we dont know this in this routine at the moment -
3744
# perhaps we should change this - then it would be an unknown.
3745
old_path = pathjoin(entry[0][0], entry[0][1])
3746
# parent id is the entry for the path in the target tree
3747
parent_id = self.state._get_entry(self.source_index, path_utf8=entry[0][0])[0][2]
3748
if parent_id == entry[0][2]:
3750
return (entry[0][2],
3751
(self.utf8_decode(old_path)[0], None),
3755
(self.utf8_decode(entry[0][1])[0], None),
3756
(DirState._minikind_to_kind[source_minikind], None),
3757
(source_details[3], None)), True
3758
elif source_minikind in 'fdlt' and target_minikind in 'r':
3759
# a rename; could be a true rename, or a rename inherited from
3760
# a renamed parent. TODO: handle this efficiently. Its not
3761
# common case to rename dirs though, so a correct but slow
3762
# implementation will do.
3763
if not osutils.is_inside_any(self.searched_specific_files, target_details[1]):
3764
self.search_specific_files.add(target_details[1])
3765
elif source_minikind in 'ra' and target_minikind in 'ra':
3766
# neither of the selected trees contain this file,
3767
# so skip over it. This is not currently directly tested, but
3768
# is indirectly via test_too_much.TestCommands.test_conflicts.
3771
raise AssertionError("don't know how to compare "
3772
"source_minikind=%r, target_minikind=%r"
3773
% (source_minikind, target_minikind))
3779
def _gather_result_for_consistency(self, result):
3780
"""Check a result we will yield to make sure we are consistent later.
3782
This gathers result's parents into a set to output later.
3784
:param result: A result tuple.
3786
if not self.partial or not result[0]:
3788
self.seen_ids.add(result[0])
3789
new_path = result[1][1]
3791
# Not the root and not a delete: queue up the parents of the path.
3792
self.search_specific_file_parents.update(
3793
osutils.parent_directories(new_path.encode('utf8')))
3794
# Add the root directory which parent_directories does not
3796
self.search_specific_file_parents.add('')
3798
def iter_changes(self):
3799
"""Iterate over the changes."""
3800
utf8_decode = cache_utf8._utf8_decode
3801
_cmp_by_dirs = cmp_by_dirs
3802
_process_entry = self._process_entry
3803
search_specific_files = self.search_specific_files
3804
searched_specific_files = self.searched_specific_files
3805
splitpath = osutils.splitpath
3807
# compare source_index and target_index at or under each element of search_specific_files.
3808
# follow the following comparison table. Note that we only want to do diff operations when
3809
# the target is fdl because thats when the walkdirs logic will have exposed the pathinfo
3813
# Source | Target | disk | action
3814
# r | fdlt | | add source to search, add id path move and perform
3815
# | | | diff check on source-target
3816
# r | fdlt | a | dangling file that was present in the basis.
3818
# r | a | | add source to search
3820
# r | r | | this path is present in a non-examined tree, skip.
3821
# r | r | a | this path is present in a non-examined tree, skip.
3822
# a | fdlt | | add new id
3823
# a | fdlt | a | dangling locally added file, skip
3824
# a | a | | not present in either tree, skip
3825
# a | a | a | not present in any tree, skip
3826
# a | r | | not present in either tree at this path, skip as it
3827
# | | | may not be selected by the users list of paths.
3828
# a | r | a | not present in either tree at this path, skip as it
3829
# | | | may not be selected by the users list of paths.
3830
# fdlt | fdlt | | content in both: diff them
3831
# fdlt | fdlt | a | deleted locally, but not unversioned - show as deleted ?
3832
# fdlt | a | | unversioned: output deleted id for now
3833
# fdlt | a | a | unversioned and deleted: output deleted id
3834
# fdlt | r | | relocated in this tree, so add target to search.
3835
# | | | Dont diff, we will see an r,fd; pair when we reach
3836
# | | | this id at the other path.
3837
# fdlt | r | a | relocated in this tree, so add target to search.
3838
# | | | Dont diff, we will see an r,fd; pair when we reach
3839
# | | | this id at the other path.
3841
# TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
3842
# keeping a cache of directories that we have seen.
3844
while search_specific_files:
3845
# TODO: the pending list should be lexically sorted? the
3846
# interface doesn't require it.
3847
current_root = search_specific_files.pop()
3848
current_root_unicode = current_root.decode('utf8')
3849
searched_specific_files.add(current_root)
3850
# process the entries for this containing directory: the rest will be
3851
# found by their parents recursively.
3852
root_entries = self.state._entries_for_path(current_root)
3853
root_abspath = self.tree.abspath(current_root_unicode)
3855
root_stat = os.lstat(root_abspath)
3857
if e.errno == errno.ENOENT:
3858
# the path does not exist: let _process_entry know that.
3859
root_dir_info = None
3861
# some other random error: hand it up.
3864
root_dir_info = ('', current_root,
3865
osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
3867
if root_dir_info[2] == 'directory':
3868
if self.tree._directory_is_tree_reference(
3869
current_root.decode('utf8')):
3870
root_dir_info = root_dir_info[:2] + \
3871
('tree-reference',) + root_dir_info[3:]
3873
if not root_entries and not root_dir_info:
3874
# this specified path is not present at all, skip it.
3876
path_handled = False
3877
for entry in root_entries:
3878
result, changed = _process_entry(entry, root_dir_info)
3879
if changed is not None:
3882
self._gather_result_for_consistency(result)
3883
if changed or self.include_unchanged:
3885
if self.want_unversioned and not path_handled and root_dir_info:
3886
new_executable = bool(
3887
stat.S_ISREG(root_dir_info[3].st_mode)
3888
and stat.S_IEXEC & root_dir_info[3].st_mode)
3890
(None, current_root_unicode),
3894
(None, splitpath(current_root_unicode)[-1]),
3895
(None, root_dir_info[2]),
3896
(None, new_executable)
3898
initial_key = (current_root, '', '')
3899
block_index, _ = self.state._find_block_index_from_key(initial_key)
3900
if block_index == 0:
3901
# we have processed the total root already, but because the
3902
# initial key matched it we should skip it here.
3904
if root_dir_info and root_dir_info[2] == 'tree-reference':
3905
current_dir_info = None
3907
dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
3909
current_dir_info = dir_iterator.next()
3911
# on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
3912
# python 2.5 has e.errno == EINVAL,
3913
# and e.winerror == ERROR_DIRECTORY
3914
e_winerror = getattr(e, 'winerror', None)
3915
win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
3916
# there may be directories in the inventory even though
3917
# this path is not a file on disk: so mark it as end of
3919
if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
3920
current_dir_info = None
3921
elif (sys.platform == 'win32'
3922
and (e.errno in win_errors
3923
or e_winerror in win_errors)):
3924
current_dir_info = None
3928
if current_dir_info[0][0] == '':
3929
# remove .bzr from iteration
3930
bzr_index = bisect.bisect_left(current_dir_info[1], ('.bzr',))
3931
if current_dir_info[1][bzr_index][0] != '.bzr':
3932
raise AssertionError()
3933
del current_dir_info[1][bzr_index]
3934
# walk until both the directory listing and the versioned metadata
3936
if (block_index < len(self.state._dirblocks) and
3937
osutils.is_inside(current_root, self.state._dirblocks[block_index][0])):
3938
current_block = self.state._dirblocks[block_index]
3940
current_block = None
3941
while (current_dir_info is not None or
3942
current_block is not None):
3943
if (current_dir_info and current_block
3944
and current_dir_info[0][0] != current_block[0]):
3945
if _cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
3946
# filesystem data refers to paths not covered by the dirblock.
3947
# this has two possibilities:
3948
# A) it is versioned but empty, so there is no block for it
3949
# B) it is not versioned.
3951
# if (A) then we need to recurse into it to check for
3952
# new unknown files or directories.
3953
# if (B) then we should ignore it, because we don't
3954
# recurse into unknown directories.
3956
while path_index < len(current_dir_info[1]):
3957
current_path_info = current_dir_info[1][path_index]
3958
if self.want_unversioned:
3959
if current_path_info[2] == 'directory':
3960
if self.tree._directory_is_tree_reference(
3961
current_path_info[0].decode('utf8')):
3962
current_path_info = current_path_info[:2] + \
3963
('tree-reference',) + current_path_info[3:]
3964
new_executable = bool(
3965
stat.S_ISREG(current_path_info[3].st_mode)
3966
and stat.S_IEXEC & current_path_info[3].st_mode)
3968
(None, utf8_decode(current_path_info[0])[0]),
3972
(None, utf8_decode(current_path_info[1])[0]),
3973
(None, current_path_info[2]),
3974
(None, new_executable))
3975
# dont descend into this unversioned path if it is
3977
if current_path_info[2] in ('directory',
3979
del current_dir_info[1][path_index]
3983
# This dir info has been handled, go to the next
3985
current_dir_info = dir_iterator.next()
3986
except StopIteration:
3987
current_dir_info = None
3989
# We have a dirblock entry for this location, but there
3990
# is no filesystem path for this. This is most likely
3991
# because a directory was removed from the disk.
3992
# We don't have to report the missing directory,
3993
# because that should have already been handled, but we
3994
# need to handle all of the files that are contained
3996
for current_entry in current_block[1]:
3997
# entry referring to file not present on disk.
3998
# advance the entry only, after processing.
3999
result, changed = _process_entry(current_entry, None)
4000
if changed is not None:
4002
self._gather_result_for_consistency(result)
4003
if changed or self.include_unchanged:
4006
if (block_index < len(self.state._dirblocks) and
4007
osutils.is_inside(current_root,
4008
self.state._dirblocks[block_index][0])):
4009
current_block = self.state._dirblocks[block_index]
4011
current_block = None
4014
if current_block and entry_index < len(current_block[1]):
4015
current_entry = current_block[1][entry_index]
4017
current_entry = None
4018
advance_entry = True
4020
if current_dir_info and path_index < len(current_dir_info[1]):
4021
current_path_info = current_dir_info[1][path_index]
4022
if current_path_info[2] == 'directory':
4023
if self.tree._directory_is_tree_reference(
4024
current_path_info[0].decode('utf8')):
4025
current_path_info = current_path_info[:2] + \
4026
('tree-reference',) + current_path_info[3:]
4028
current_path_info = None
4030
path_handled = False
4031
while (current_entry is not None or
4032
current_path_info is not None):
4033
if current_entry is None:
4034
# the check for path_handled when the path is advanced
4035
# will yield this path if needed.
4037
elif current_path_info is None:
4038
# no path is fine: the per entry code will handle it.
4039
result, changed = _process_entry(current_entry, current_path_info)
4040
if changed is not None:
4042
self._gather_result_for_consistency(result)
4043
if changed or self.include_unchanged:
4045
elif (current_entry[0][1] != current_path_info[1]
4046
or current_entry[1][self.target_index][0] in 'ar'):
4047
# The current path on disk doesn't match the dirblock
4048
# record. Either the dirblock is marked as absent, or
4049
# the file on disk is not present at all in the
4050
# dirblock. Either way, report about the dirblock
4051
# entry, and let other code handle the filesystem one.
4053
# Compare the basename for these files to determine
4055
if current_path_info[1] < current_entry[0][1]:
4056
# extra file on disk: pass for now, but only
4057
# increment the path, not the entry
4058
advance_entry = False
4060
# entry referring to file not present on disk.
4061
# advance the entry only, after processing.
4062
result, changed = _process_entry(current_entry, None)
4063
if changed is not None:
4065
self._gather_result_for_consistency(result)
4066
if changed or self.include_unchanged:
4068
advance_path = False
4070
result, changed = _process_entry(current_entry, current_path_info)
4071
if changed is not None:
4074
self._gather_result_for_consistency(result)
4075
if changed or self.include_unchanged:
4077
if advance_entry and current_entry is not None:
4079
if entry_index < len(current_block[1]):
4080
current_entry = current_block[1][entry_index]
4082
current_entry = None
4084
advance_entry = True # reset the advance flaga
4085
if advance_path and current_path_info is not None:
4086
if not path_handled:
4087
# unversioned in all regards
4088
if self.want_unversioned:
4089
new_executable = bool(
4090
stat.S_ISREG(current_path_info[3].st_mode)
4091
and stat.S_IEXEC & current_path_info[3].st_mode)
4093
relpath_unicode = utf8_decode(current_path_info[0])[0]
4094
except UnicodeDecodeError:
4095
raise errors.BadFilenameEncoding(
4096
current_path_info[0], osutils._fs_enc)
4098
(None, relpath_unicode),
4102
(None, utf8_decode(current_path_info[1])[0]),
4103
(None, current_path_info[2]),
4104
(None, new_executable))
4105
# dont descend into this unversioned path if it is
4107
if current_path_info[2] in ('directory'):
4108
del current_dir_info[1][path_index]
4110
# dont descend the disk iterator into any tree
4112
if current_path_info[2] == 'tree-reference':
4113
del current_dir_info[1][path_index]
4116
if path_index < len(current_dir_info[1]):
4117
current_path_info = current_dir_info[1][path_index]
4118
if current_path_info[2] == 'directory':
4119
if self.tree._directory_is_tree_reference(
4120
current_path_info[0].decode('utf8')):
4121
current_path_info = current_path_info[:2] + \
4122
('tree-reference',) + current_path_info[3:]
4124
current_path_info = None
4125
path_handled = False
4127
advance_path = True # reset the advance flagg.
4128
if current_block is not None:
4130
if (block_index < len(self.state._dirblocks) and
4131
osutils.is_inside(current_root, self.state._dirblocks[block_index][0])):
4132
current_block = self.state._dirblocks[block_index]
4134
current_block = None
4135
if current_dir_info is not None:
4137
current_dir_info = dir_iterator.next()
4138
except StopIteration:
4139
current_dir_info = None
4140
for result in self._iter_specific_file_parents():
4143
def _iter_specific_file_parents(self):
4144
"""Iter over the specific file parents."""
4145
while self.search_specific_file_parents:
4146
# Process the parent directories for the paths we were iterating.
4147
# Even in extremely large trees this should be modest, so currently
4148
# no attempt is made to optimise.
4149
path_utf8 = self.search_specific_file_parents.pop()
4150
if osutils.is_inside_any(self.searched_specific_files, path_utf8):
4151
# We've examined this path.
4153
if path_utf8 in self.searched_exact_paths:
4154
# We've examined this path.
4156
path_entries = self.state._entries_for_path(path_utf8)
4157
# We need either one or two entries. If the path in
4158
# self.target_index has moved (so the entry in source_index is in
4159
# 'ar') then we need to also look for the entry for this path in
4160
# self.source_index, to output the appropriate delete-or-rename.
4161
selected_entries = []
4163
for candidate_entry in path_entries:
4164
# Find entries present in target at this path:
4165
if candidate_entry[1][self.target_index][0] not in 'ar':
4167
selected_entries.append(candidate_entry)
4168
# Find entries present in source at this path:
4169
elif (self.source_index is not None and
4170
candidate_entry[1][self.source_index][0] not in 'ar'):
4172
if candidate_entry[1][self.target_index][0] == 'a':
4173
# Deleted, emit it here.
4174
selected_entries.append(candidate_entry)
4176
# renamed, emit it when we process the directory it
4178
self.search_specific_file_parents.add(
4179
candidate_entry[1][self.target_index][1])
4181
raise AssertionError(
4182
"Missing entry for specific path parent %r, %r" % (
4183
path_utf8, path_entries))
4184
path_info = self._path_info(path_utf8, path_utf8.decode('utf8'))
4185
for entry in selected_entries:
4186
if entry[0][2] in self.seen_ids:
4188
result, changed = self._process_entry(entry, path_info)
4190
raise AssertionError(
4191
"Got entry<->path mismatch for specific path "
4192
"%r entry %r path_info %r " % (
4193
path_utf8, entry, path_info))
4194
# Only include changes - we're outside the users requested
4197
self._gather_result_for_consistency(result)
4198
if (result[6][0] == 'directory' and
4199
result[6][1] != 'directory'):
4200
# This stopped being a directory, the old children have
4202
if entry[1][self.source_index][0] == 'r':
4203
# renamed, take the source path
4204
entry_path_utf8 = entry[1][self.source_index][1]
4206
entry_path_utf8 = path_utf8
4207
initial_key = (entry_path_utf8, '', '')
4208
block_index, _ = self.state._find_block_index_from_key(
4210
if block_index == 0:
4211
# The children of the root are in block index 1.
4213
current_block = None
4214
if block_index < len(self.state._dirblocks):
4215
current_block = self.state._dirblocks[block_index]
4216
if not osutils.is_inside(
4217
entry_path_utf8, current_block[0]):
4218
# No entries for this directory at all.
4219
current_block = None
4220
if current_block is not None:
4221
for entry in current_block[1]:
4222
if entry[1][self.source_index][0] in 'ar':
4223
# Not in the source tree, so doesn't have to be
4226
# Path of the entry itself.
4228
self.search_specific_file_parents.add(
4229
osutils.pathjoin(*entry[0][:2]))
4230
if changed or self.include_unchanged:
4232
self.searched_exact_paths.add(path_utf8)
4234
def _path_info(self, utf8_path, unicode_path):
4235
"""Generate path_info for unicode_path.
4237
:return: None if unicode_path does not exist, or a path_info tuple.
4239
abspath = self.tree.abspath(unicode_path)
4241
stat = os.lstat(abspath)
4243
if e.errno == errno.ENOENT:
4244
# the path does not exist.
4248
utf8_basename = utf8_path.rsplit('/', 1)[-1]
4249
dir_info = (utf8_path, utf8_basename,
4250
osutils.file_kind_from_stat_mode(stat.st_mode), stat,
4252
if dir_info[2] == 'directory':
4253
if self.tree._directory_is_tree_reference(
4255
self.root_dir_info = self.root_dir_info[:2] + \
4256
('tree-reference',) + self.root_dir_info[3:]
4260
# Try to load the compiled form if possible
4262
from bzrlib._dirstate_helpers_pyx import (
4269
ProcessEntryC as _process_entry,
4270
update_entry as update_entry,
4272
except ImportError, e:
4273
osutils.failed_to_load_extension(e)
4274
from bzrlib._dirstate_helpers_py import (
4282
# FIXME: It would be nice to be able to track moved lines so that the
4283
# corresponding python code can be moved to the _dirstate_helpers_py
4284
# module. I don't want to break the history for this important piece of
4285
# code so I left the code here -- vila 20090622
4286
update_entry = py_update_entry
4287
_process_entry = ProcessEntryPython
added
removed
bzrlib/dirstate.py
