1
# Copyright (C) 2006, 2007, 2008 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.
23
MINIKIND = "f" | "d" | "l" | "a" | "r" | "t";
26
WHOLE_NUMBER = {digit}, digit;
28
REVISION_ID = a non-empty utf8 string;
30
dirstate format = header line, full checksum, row count, parent details,
31
ghost_details, entries;
32
header line = "#bazaar dirstate flat format 3", NL;
33
full checksum = "crc32: ", ["-"], WHOLE_NUMBER, NL;
34
row count = "num_entries: ", WHOLE_NUMBER, NL;
35
parent_details = WHOLE NUMBER, {REVISION_ID}* NL;
36
ghost_details = WHOLE NUMBER, {REVISION_ID}*, NL;
38
entry = entry_key, current_entry_details, {parent_entry_details};
39
entry_key = dirname, basename, fileid;
40
current_entry_details = common_entry_details, working_entry_details;
41
parent_entry_details = common_entry_details, history_entry_details;
42
common_entry_details = MINIKIND, fingerprint, size, executable
43
working_entry_details = packed_stat
44
history_entry_details = REVISION_ID;
47
fingerprint = a nonempty utf8 sequence with meaning defined by minikind.
49
Given this definition, the following is useful to know:
50
entry (aka row) - all the data for a given key.
51
entry[0]: The key (dirname, basename, fileid)
55
entry[1]: The tree(s) data for this path and id combination.
56
entry[1][0]: The current tree
57
entry[1][1]: The second tree
59
For an entry for a tree, we have (using tree 0 - current tree) to demonstrate:
60
entry[1][0][0]: minikind
61
entry[1][0][1]: fingerprint
63
entry[1][0][3]: executable
64
entry[1][0][4]: packed_stat
66
entry[1][1][4]: revision_id
68
There may be multiple rows at the root, one per id present in the root, so the
69
in memory root row is now:
70
self._dirblocks[0] -> ('', [entry ...]),
71
and the entries in there are
74
entries[0][2]: file_id
75
entries[1][0]: The tree data for the current tree for this fileid at /
79
'r' is a relocated entry: This path is not present in this tree with this id,
80
but the id can be found at another location. The fingerprint is used to
81
point to the target location.
82
'a' is an absent entry: In that tree the id is not present at this path.
83
'd' is a directory entry: This path in this tree is a directory with the
84
current file id. There is no fingerprint for directories.
85
'f' is a file entry: As for directory, but it's a file. The fingerprint is the
86
sha1 value of the file's canonical form, i.e. after any read filters have
87
been applied to the convenience form stored in the working tree.
88
'l' is a symlink entry: As for directory, but a symlink. The fingerprint is the
90
't' is a reference to a nested subtree; the fingerprint is the referenced
95
The entries on disk and in memory are ordered according to the following keys:
97
directory, as a list of components
101
--- Format 1 had the following different definition: ---
102
rows = dirname, NULL, basename, NULL, MINIKIND, NULL, fileid_utf8, NULL,
103
WHOLE NUMBER (* size *), NULL, packed stat, NULL, sha1|symlink target,
105
PARENT ROW = NULL, revision_utf8, NULL, MINIKIND, NULL, dirname, NULL,
106
basename, NULL, WHOLE NUMBER (* size *), NULL, "y" | "n", NULL,
109
PARENT ROW's are emitted for every parent that is not in the ghosts details
110
line. That is, if the parents are foo, bar, baz, and the ghosts are bar, then
111
each row will have a PARENT ROW for foo and baz, but not for bar.
114
In any tree, a kind of 'moved' indicates that the fingerprint field
115
(which we treat as opaque data specific to the 'kind' anyway) has the
116
details for the id of this row in that tree.
118
I'm strongly tempted to add a id->path index as well, but I think that
119
where we need id->path mapping; we also usually read the whole file, so
120
I'm going to skip that for the moment, as we have the ability to locate
121
via bisect any path in any tree, and if we lookup things by path, we can
122
accumulate an id->path mapping as we go, which will tend to match what we
125
I plan to implement this asap, so please speak up now to alter/tweak the
126
design - and once we stabilise on this, I'll update the wiki page for
129
The rationale for all this is that we want fast operations for the
130
common case (diff/status/commit/merge on all files) and extremely fast
131
operations for the less common but still occurs a lot status/diff/commit
132
on specific files). Operations on specific files involve a scan for all
133
the children of a path, *in every involved tree*, which the current
134
format did not accommodate.
138
1) Fast end to end use for bzr's top 5 uses cases. (commmit/diff/status/merge/???)
139
2) fall back current object model as needed.
140
3) scale usably to the largest trees known today - say 50K entries. (mozilla
141
is an example of this)
145
Eventually reuse dirstate objects across locks IFF the dirstate file has not
146
been modified, but will require that we flush/ignore cached stat-hit data
147
because we won't want to restat all files on disk just because a lock was
148
acquired, yet we cannot trust the data after the previous lock was released.
150
Memory representation:
151
vector of all directories, and vector of the childen ?
153
root_entrie = (direntry for root, [parent_direntries_for_root]),
155
('', ['data for achild', 'data for bchild', 'data for cchild'])
156
('dir', ['achild', 'cchild', 'echild'])
158
- single bisect to find N subtrees from a path spec
159
- in-order for serialisation - this is 'dirblock' grouping.
160
- insertion of a file '/a' affects only the '/' child-vector, that is, to
161
insert 10K elements from scratch does not generates O(N^2) memoves of a
162
single vector, rather each individual, which tends to be limited to a
163
manageable number. Will scale badly on trees with 10K entries in a
164
single directory. compare with Inventory.InventoryDirectory which has
165
a dictionary for the children. No bisect capability, can only probe for
166
exact matches, or grab all elements and sort.
167
- What's the risk of error here? Once we have the base format being processed
168
we should have a net win regardless of optimality. So we are going to
169
go with what seems reasonable.
172
Maybe we should do a test profile of the core structure - 10K simulated
173
searches/lookups/etc?
175
Objects for each row?
176
The lifetime of Dirstate objects is current per lock, but see above for
177
possible extensions. The lifetime of a row from a dirstate is expected to be
178
very short in the optimistic case: which we are optimising for. For instance,
179
subtree status will determine from analysis of the disk data what rows need to
180
be examined at all, and will be able to determine from a single row whether
181
that file has altered or not, so we are aiming to process tens of thousands of
182
entries each second within the dirstate context, before exposing anything to
183
the larger codebase. This suggests we want the time for a single file
184
comparison to be < 0.1 milliseconds. That would give us 10000 paths per second
185
processed, and to scale to 100 thousand we'll another order of magnitude to do
186
that. Now, as the lifetime for all unchanged entries is the time to parse, stat
187
the file on disk, and then immediately discard, the overhead of object creation
188
becomes a significant cost.
190
Figures: Creating a tuple from 3 elements was profiled at 0.0625
191
microseconds, whereas creating a object which is subclassed from tuple was
192
0.500 microseconds, and creating an object with 3 elements and slots was 3
193
microseconds long. 0.1 milliseconds is 100 microseconds, and ideally we'll get
194
down to 10 microseconds for the total processing - having 33% of that be object
195
creation is a huge overhead. There is a potential cost in using tuples within
196
each row which is that the conditional code to do comparisons may be slower
197
than method invocation, but method invocation is known to be slow due to stack
198
frame creation, so avoiding methods in these tight inner loops in unfortunately
199
desirable. We can consider a pyrex version of this with objects in future if
209
from stat import S_IEXEC
227
# This is the Windows equivalent of ENOTDIR
228
# It is defined in pywin32.winerror, but we don't want a strong dependency for
229
# just an error code.
230
ERROR_PATH_NOT_FOUND = 3
231
ERROR_DIRECTORY = 267
234
if not getattr(struct, '_compile', None):
235
# Cannot pre-compile the dirstate pack_stat
236
def pack_stat(st, _encode=binascii.b2a_base64, _pack=struct.pack):
237
"""Convert stat values into a packed representation."""
238
return _encode(_pack('>LLLLLL', st.st_size, int(st.st_mtime),
239
int(st.st_ctime), st.st_dev, st.st_ino & 0xFFFFFFFF,
242
# compile the struct compiler we need, so as to only do it once
243
from _struct import Struct
244
_compiled_pack = Struct('>LLLLLL').pack
245
def pack_stat(st, _encode=binascii.b2a_base64, _pack=_compiled_pack):
246
"""Convert stat values into a packed representation."""
247
# jam 20060614 it isn't really worth removing more entries if we
248
# are going to leave it in packed form.
249
# With only st_mtime and st_mode filesize is 5.5M and read time is 275ms
250
# With all entries, filesize is 5.9M and read time is maybe 280ms
251
# well within the noise margin
253
# base64 encoding always adds a final newline, so strip it off
254
# The current version
255
return _encode(_pack(st.st_size, int(st.st_mtime), int(st.st_ctime),
256
st.st_dev, st.st_ino & 0xFFFFFFFF, st.st_mode))[:-1]
257
# This is 0.060s / 1.520s faster by not encoding as much information
258
# return _encode(_pack('>LL', int(st.st_mtime), st.st_mode))[:-1]
259
# This is not strictly faster than _encode(_pack())[:-1]
260
# return '%X.%X.%X.%X.%X.%X' % (
261
# st.st_size, int(st.st_mtime), int(st.st_ctime),
262
# st.st_dev, st.st_ino, st.st_mode)
263
# Similar to the _encode(_pack('>LL'))
264
# return '%X.%X' % (int(st.st_mtime), st.st_mode)
267
class SHA1Provider(object):
268
"""An interface for getting sha1s of a file."""
270
def sha1(self, abspath):
271
"""Return the sha1 of a file given its absolute path.
273
:param abspath: May be a filesystem encoded absolute path
276
raise NotImplementedError(self.sha1)
278
def stat_and_sha1(self, abspath):
279
"""Return the stat and sha1 of a file given its absolute path.
281
:param abspath: May be a filesystem encoded absolute path
284
Note: the stat should be the stat of the physical file
285
while the sha may be the sha of its canonical content.
287
raise NotImplementedError(self.stat_and_sha1)
290
class DefaultSHA1Provider(SHA1Provider):
291
"""A SHA1Provider that reads directly from the filesystem."""
293
def sha1(self, abspath):
294
"""Return the sha1 of a file given its absolute path."""
295
return osutils.sha_file_by_name(abspath)
297
def stat_and_sha1(self, abspath):
298
"""Return the stat and sha1 of a file given its absolute path."""
299
file_obj = file(abspath, 'rb')
301
statvalue = os.fstat(file_obj.fileno())
302
sha1 = osutils.sha_file(file_obj)
305
return statvalue, sha1
308
class DirState(object):
309
"""Record directory and metadata state for fast access.
311
A dirstate is a specialised data structure for managing local working
312
tree state information. Its not yet well defined whether it is platform
313
specific, and if it is how we detect/parameterize that.
315
Dirstates use the usual lock_write, lock_read and unlock mechanisms.
316
Unlike most bzr disk formats, DirStates must be locked for reading, using
317
lock_read. (This is an os file lock internally.) This is necessary
318
because the file can be rewritten in place.
320
DirStates must be explicitly written with save() to commit changes; just
321
unlocking them does not write the changes to disk.
324
_kind_to_minikind = {
330
'tree-reference': 't',
332
_minikind_to_kind = {
338
't': 'tree-reference',
340
_stat_to_minikind = {
345
_to_yesno = {True:'y', False: 'n'} # TODO profile the performance gain
346
# of using int conversion rather than a dict here. AND BLAME ANDREW IF
349
# TODO: jam 20070221 Figure out what to do if we have a record that exceeds
350
# the BISECT_PAGE_SIZE. For now, we just have to make it large enough
351
# that we are sure a single record will always fit.
352
BISECT_PAGE_SIZE = 4096
355
IN_MEMORY_UNMODIFIED = 1
356
IN_MEMORY_MODIFIED = 2
358
# A pack_stat (the x's) that is just noise and will never match the output
361
NULL_PARENT_DETAILS = ('a', '', 0, False, '')
363
HEADER_FORMAT_2 = '#bazaar dirstate flat format 2\n'
364
HEADER_FORMAT_3 = '#bazaar dirstate flat format 3\n'
366
def __init__(self, path, sha1_provider):
367
"""Create a DirState object.
369
:param path: The path at which the dirstate file on disk should live.
370
:param sha1_provider: an object meeting the SHA1Provider interface.
372
# _header_state and _dirblock_state represent the current state
373
# of the dirstate metadata and the per-row data respectiely.
374
# NOT_IN_MEMORY indicates that no data is in memory
375
# IN_MEMORY_UNMODIFIED indicates that what we have in memory
376
# is the same as is on disk
377
# IN_MEMORY_MODIFIED indicates that we have a modified version
378
# of what is on disk.
379
# In future we will add more granularity, for instance _dirblock_state
380
# will probably support partially-in-memory as a separate variable,
381
# allowing for partially-in-memory unmodified and partially-in-memory
383
self._header_state = DirState.NOT_IN_MEMORY
384
self._dirblock_state = DirState.NOT_IN_MEMORY
385
# If true, an error has been detected while updating the dirstate, and
386
# for safety we're not going to commit to disk.
387
self._changes_aborted = False
391
self._state_file = None
392
self._filename = path
393
self._lock_token = None
394
self._lock_state = None
395
self._id_index = None
396
# a map from packed_stat to sha's.
397
self._packed_stat_index = None
398
self._end_of_header = None
399
self._cutoff_time = None
400
self._split_path_cache = {}
401
self._bisect_page_size = DirState.BISECT_PAGE_SIZE
402
self._sha1_provider = sha1_provider
403
if 'hashcache' in debug.debug_flags:
404
self._sha1_file = self._sha1_file_and_mutter
406
self._sha1_file = self._sha1_provider.sha1
407
# These two attributes provide a simple cache for lookups into the
408
# dirstate in-memory vectors. By probing respectively for the last
409
# block, and for the next entry, we save nearly 2 bisections per path
411
self._last_block_index = None
412
self._last_entry_index = None
416
(self.__class__.__name__, self._filename)
418
def add(self, path, file_id, kind, stat, fingerprint):
419
"""Add a path to be tracked.
421
:param path: The path within the dirstate - '' is the root, 'foo' is the
422
path foo within the root, 'foo/bar' is the path bar within foo
424
:param file_id: The file id of the path being added.
425
:param kind: The kind of the path, as a string like 'file',
427
:param stat: The output of os.lstat for the path.
428
:param fingerprint: The sha value of the file's canonical form (i.e.
429
after any read filters have been applied),
430
or the target of a symlink,
431
or the referenced revision id for tree-references,
432
or '' for directories.
435
# find the block its in.
436
# find the location in the block.
437
# check its not there
439
#------- copied from inventory.ensure_normalized_name - keep synced.
440
# --- normalized_filename wants a unicode basename only, so get one.
441
dirname, basename = osutils.split(path)
442
# we dont import normalized_filename directly because we want to be
443
# able to change the implementation at runtime for tests.
444
norm_name, can_access = osutils.normalized_filename(basename)
445
if norm_name != basename:
449
raise errors.InvalidNormalization(path)
450
# you should never have files called . or ..; just add the directory
451
# in the parent, or according to the special treatment for the root
452
if basename == '.' or basename == '..':
453
raise errors.InvalidEntryName(path)
454
# now that we've normalised, we need the correct utf8 path and
455
# dirname and basename elements. This single encode and split should be
456
# faster than three separate encodes.
457
utf8path = (dirname + '/' + basename).strip('/').encode('utf8')
458
dirname, basename = osutils.split(utf8path)
459
# uses __class__ for speed; the check is needed for safety
460
if file_id.__class__ is not str:
461
raise AssertionError(
462
"must be a utf8 file_id not %s" % (type(file_id), ))
463
# Make sure the file_id does not exist in this tree
465
file_id_entry = self._get_entry(0, fileid_utf8=file_id, include_deleted=True)
466
if file_id_entry != (None, None):
467
if file_id_entry[1][0][0] == 'a':
468
if file_id_entry[0] != (dirname, basename, file_id):
469
# set the old name's current operation to rename
470
self.update_minimal(file_id_entry[0],
476
rename_from = file_id_entry[0][0:2]
478
path = osutils.pathjoin(file_id_entry[0][0], file_id_entry[0][1])
479
kind = DirState._minikind_to_kind[file_id_entry[1][0][0]]
480
info = '%s:%s' % (kind, path)
481
raise errors.DuplicateFileId(file_id, info)
482
first_key = (dirname, basename, '')
483
block_index, present = self._find_block_index_from_key(first_key)
485
# check the path is not in the tree
486
block = self._dirblocks[block_index][1]
487
entry_index, _ = self._find_entry_index(first_key, block)
488
while (entry_index < len(block) and
489
block[entry_index][0][0:2] == first_key[0:2]):
490
if block[entry_index][1][0][0] not in 'ar':
491
# this path is in the dirstate in the current tree.
492
raise Exception, "adding already added path!"
495
# The block where we want to put the file is not present. But it
496
# might be because the directory was empty, or not loaded yet. Look
497
# for a parent entry, if not found, raise NotVersionedError
498
parent_dir, parent_base = osutils.split(dirname)
499
parent_block_idx, parent_entry_idx, _, parent_present = \
500
self._get_block_entry_index(parent_dir, parent_base, 0)
501
if not parent_present:
502
raise errors.NotVersionedError(path, str(self))
503
self._ensure_block(parent_block_idx, parent_entry_idx, dirname)
504
block = self._dirblocks[block_index][1]
505
entry_key = (dirname, basename, file_id)
508
packed_stat = DirState.NULLSTAT
511
packed_stat = pack_stat(stat)
512
parent_info = self._empty_parent_info()
513
minikind = DirState._kind_to_minikind[kind]
514
if rename_from is not None:
516
old_path_utf8 = '%s/%s' % rename_from
518
old_path_utf8 = rename_from[1]
519
parent_info[0] = ('r', old_path_utf8, 0, False, '')
521
entry_data = entry_key, [
522
(minikind, fingerprint, size, False, packed_stat),
524
elif kind == 'directory':
525
entry_data = entry_key, [
526
(minikind, '', 0, False, packed_stat),
528
elif kind == 'symlink':
529
entry_data = entry_key, [
530
(minikind, fingerprint, size, False, packed_stat),
532
elif kind == 'tree-reference':
533
entry_data = entry_key, [
534
(minikind, fingerprint, 0, False, packed_stat),
537
raise errors.BzrError('unknown kind %r' % kind)
538
entry_index, present = self._find_entry_index(entry_key, block)
540
block.insert(entry_index, entry_data)
542
if block[entry_index][1][0][0] != 'a':
543
raise AssertionError(" %r(%r) already added" % (basename, file_id))
544
block[entry_index][1][0] = entry_data[1][0]
546
if kind == 'directory':
547
# insert a new dirblock
548
self._ensure_block(block_index, entry_index, utf8path)
549
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
551
self._id_index.setdefault(entry_key[2], set()).add(entry_key)
553
def _bisect(self, paths):
554
"""Bisect through the disk structure for specific rows.
556
:param paths: A list of paths to find
557
:return: A dict mapping path => entries for found entries. Missing
558
entries will not be in the map.
559
The list is not sorted, and entries will be populated
560
based on when they were read.
562
self._requires_lock()
563
# We need the file pointer to be right after the initial header block
564
self._read_header_if_needed()
565
# If _dirblock_state was in memory, we should just return info from
566
# there, this function is only meant to handle when we want to read
568
if self._dirblock_state != DirState.NOT_IN_MEMORY:
569
raise AssertionError("bad dirblock state %r" % self._dirblock_state)
571
# The disk representation is generally info + '\0\n\0' at the end. But
572
# for bisecting, it is easier to treat this as '\0' + info + '\0\n'
573
# Because it means we can sync on the '\n'
574
state_file = self._state_file
575
file_size = os.fstat(state_file.fileno()).st_size
576
# We end up with 2 extra fields, we should have a trailing '\n' to
577
# ensure that we read the whole record, and we should have a precursur
578
# '' which ensures that we start after the previous '\n'
579
entry_field_count = self._fields_per_entry() + 1
581
low = self._end_of_header
582
high = file_size - 1 # Ignore the final '\0'
583
# Map from (dir, name) => entry
586
# Avoid infinite seeking
587
max_count = 30*len(paths)
589
# pending is a list of places to look.
590
# each entry is a tuple of low, high, dir_names
591
# low -> the first byte offset to read (inclusive)
592
# high -> the last byte offset (inclusive)
593
# dir_names -> The list of (dir, name) pairs that should be found in
594
# the [low, high] range
595
pending = [(low, high, paths)]
597
page_size = self._bisect_page_size
599
fields_to_entry = self._get_fields_to_entry()
602
low, high, cur_files = pending.pop()
604
if not cur_files or low >= high:
609
if count > max_count:
610
raise errors.BzrError('Too many seeks, most likely a bug.')
612
mid = max(low, (low+high-page_size)/2)
615
# limit the read size, so we don't end up reading data that we have
617
read_size = min(page_size, (high-mid)+1)
618
block = state_file.read(read_size)
621
entries = block.split('\n')
624
# We didn't find a '\n', so we cannot have found any records.
625
# So put this range back and try again. But we know we have to
626
# increase the page size, because a single read did not contain
627
# a record break (so records must be larger than page_size)
629
pending.append((low, high, cur_files))
632
# Check the first and last entries, in case they are partial, or if
633
# we don't care about the rest of this page
635
first_fields = entries[0].split('\0')
636
if len(first_fields) < entry_field_count:
637
# We didn't get the complete first entry
638
# so move start, and grab the next, which
639
# should be a full entry
640
start += len(entries[0])+1
641
first_fields = entries[1].split('\0')
644
if len(first_fields) <= 2:
645
# We didn't even get a filename here... what do we do?
646
# Try a large page size and repeat this query
648
pending.append((low, high, cur_files))
651
# Find what entries we are looking for, which occur before and
652
# after this first record.
655
first_path = first_fields[1] + '/' + first_fields[2]
657
first_path = first_fields[2]
658
first_loc = _bisect_path_left(cur_files, first_path)
660
# These exist before the current location
661
pre = cur_files[:first_loc]
662
# These occur after the current location, which may be in the
663
# data we read, or might be after the last entry
664
post = cur_files[first_loc:]
666
if post and len(first_fields) >= entry_field_count:
667
# We have files after the first entry
669
# Parse the last entry
670
last_entry_num = len(entries)-1
671
last_fields = entries[last_entry_num].split('\0')
672
if len(last_fields) < entry_field_count:
673
# The very last hunk was not complete,
674
# read the previous hunk
675
after = mid + len(block) - len(entries[-1])
677
last_fields = entries[last_entry_num].split('\0')
679
after = mid + len(block)
682
last_path = last_fields[1] + '/' + last_fields[2]
684
last_path = last_fields[2]
685
last_loc = _bisect_path_right(post, last_path)
687
middle_files = post[:last_loc]
688
post = post[last_loc:]
691
# We have files that should occur in this block
692
# (>= first, <= last)
693
# Either we will find them here, or we can mark them as
696
if middle_files[0] == first_path:
697
# We might need to go before this location
698
pre.append(first_path)
699
if middle_files[-1] == last_path:
700
post.insert(0, last_path)
702
# Find out what paths we have
703
paths = {first_path:[first_fields]}
704
# last_path might == first_path so we need to be
705
# careful if we should append rather than overwrite
706
if last_entry_num != first_entry_num:
707
paths.setdefault(last_path, []).append(last_fields)
708
for num in xrange(first_entry_num+1, last_entry_num):
709
# TODO: jam 20070223 We are already splitting here, so
710
# shouldn't we just split the whole thing rather
711
# than doing the split again in add_one_record?
712
fields = entries[num].split('\0')
714
path = fields[1] + '/' + fields[2]
717
paths.setdefault(path, []).append(fields)
719
for path in middle_files:
720
for fields in paths.get(path, []):
721
# offset by 1 because of the opening '\0'
722
# consider changing fields_to_entry to avoid the
724
entry = fields_to_entry(fields[1:])
725
found.setdefault(path, []).append(entry)
727
# Now we have split up everything into pre, middle, and post, and
728
# we have handled everything that fell in 'middle'.
729
# We add 'post' first, so that we prefer to seek towards the
730
# beginning, so that we will tend to go as early as we need, and
731
# then only seek forward after that.
733
pending.append((after, high, post))
735
pending.append((low, start-1, pre))
737
# Consider that we may want to return the directory entries in sorted
738
# order. For now, we just return them in whatever order we found them,
739
# and leave it up to the caller if they care if it is ordered or not.
742
def _bisect_dirblocks(self, dir_list):
743
"""Bisect through the disk structure to find entries in given dirs.
745
_bisect_dirblocks is meant to find the contents of directories, which
746
differs from _bisect, which only finds individual entries.
748
:param dir_list: A sorted list of directory names ['', 'dir', 'foo'].
749
:return: A map from dir => entries_for_dir
751
# TODO: jam 20070223 A lot of the bisecting logic could be shared
752
# between this and _bisect. It would require parameterizing the
753
# inner loop with a function, though. We should evaluate the
754
# performance difference.
755
self._requires_lock()
756
# We need the file pointer to be right after the initial header block
757
self._read_header_if_needed()
758
# If _dirblock_state was in memory, we should just return info from
759
# there, this function is only meant to handle when we want to read
761
if self._dirblock_state != DirState.NOT_IN_MEMORY:
762
raise AssertionError("bad dirblock state %r" % self._dirblock_state)
763
# The disk representation is generally info + '\0\n\0' at the end. But
764
# for bisecting, it is easier to treat this as '\0' + info + '\0\n'
765
# Because it means we can sync on the '\n'
766
state_file = self._state_file
767
file_size = os.fstat(state_file.fileno()).st_size
768
# We end up with 2 extra fields, we should have a trailing '\n' to
769
# ensure that we read the whole record, and we should have a precursur
770
# '' which ensures that we start after the previous '\n'
771
entry_field_count = self._fields_per_entry() + 1
773
low = self._end_of_header
774
high = file_size - 1 # Ignore the final '\0'
775
# Map from dir => entry
778
# Avoid infinite seeking
779
max_count = 30*len(dir_list)
781
# pending is a list of places to look.
782
# each entry is a tuple of low, high, dir_names
783
# low -> the first byte offset to read (inclusive)
784
# high -> the last byte offset (inclusive)
785
# dirs -> The list of directories that should be found in
786
# the [low, high] range
787
pending = [(low, high, dir_list)]
789
page_size = self._bisect_page_size
791
fields_to_entry = self._get_fields_to_entry()
794
low, high, cur_dirs = pending.pop()
796
if not cur_dirs or low >= high:
801
if count > max_count:
802
raise errors.BzrError('Too many seeks, most likely a bug.')
804
mid = max(low, (low+high-page_size)/2)
807
# limit the read size, so we don't end up reading data that we have
809
read_size = min(page_size, (high-mid)+1)
810
block = state_file.read(read_size)
813
entries = block.split('\n')
816
# We didn't find a '\n', so we cannot have found any records.
817
# So put this range back and try again. But we know we have to
818
# increase the page size, because a single read did not contain
819
# a record break (so records must be larger than page_size)
821
pending.append((low, high, cur_dirs))
824
# Check the first and last entries, in case they are partial, or if
825
# we don't care about the rest of this page
827
first_fields = entries[0].split('\0')
828
if len(first_fields) < entry_field_count:
829
# We didn't get the complete first entry
830
# so move start, and grab the next, which
831
# should be a full entry
832
start += len(entries[0])+1
833
first_fields = entries[1].split('\0')
836
if len(first_fields) <= 1:
837
# We didn't even get a dirname here... what do we do?
838
# Try a large page size and repeat this query
840
pending.append((low, high, cur_dirs))
843
# Find what entries we are looking for, which occur before and
844
# after this first record.
846
first_dir = first_fields[1]
847
first_loc = bisect.bisect_left(cur_dirs, first_dir)
849
# These exist before the current location
850
pre = cur_dirs[:first_loc]
851
# These occur after the current location, which may be in the
852
# data we read, or might be after the last entry
853
post = cur_dirs[first_loc:]
855
if post and len(first_fields) >= entry_field_count:
856
# We have records to look at after the first entry
858
# Parse the last entry
859
last_entry_num = len(entries)-1
860
last_fields = entries[last_entry_num].split('\0')
861
if len(last_fields) < entry_field_count:
862
# The very last hunk was not complete,
863
# read the previous hunk
864
after = mid + len(block) - len(entries[-1])
866
last_fields = entries[last_entry_num].split('\0')
868
after = mid + len(block)
870
last_dir = last_fields[1]
871
last_loc = bisect.bisect_right(post, last_dir)
873
middle_files = post[:last_loc]
874
post = post[last_loc:]
877
# We have files that should occur in this block
878
# (>= first, <= last)
879
# Either we will find them here, or we can mark them as
882
if middle_files[0] == first_dir:
883
# We might need to go before this location
884
pre.append(first_dir)
885
if middle_files[-1] == last_dir:
886
post.insert(0, last_dir)
888
# Find out what paths we have
889
paths = {first_dir:[first_fields]}
890
# last_dir might == first_dir so we need to be
891
# careful if we should append rather than overwrite
892
if last_entry_num != first_entry_num:
893
paths.setdefault(last_dir, []).append(last_fields)
894
for num in xrange(first_entry_num+1, last_entry_num):
895
# TODO: jam 20070223 We are already splitting here, so
896
# shouldn't we just split the whole thing rather
897
# than doing the split again in add_one_record?
898
fields = entries[num].split('\0')
899
paths.setdefault(fields[1], []).append(fields)
901
for cur_dir in middle_files:
902
for fields in paths.get(cur_dir, []):
903
# offset by 1 because of the opening '\0'
904
# consider changing fields_to_entry to avoid the
906
entry = fields_to_entry(fields[1:])
907
found.setdefault(cur_dir, []).append(entry)
909
# Now we have split up everything into pre, middle, and post, and
910
# we have handled everything that fell in 'middle'.
911
# We add 'post' first, so that we prefer to seek towards the
912
# beginning, so that we will tend to go as early as we need, and
913
# then only seek forward after that.
915
pending.append((after, high, post))
917
pending.append((low, start-1, pre))
921
def _bisect_recursive(self, paths):
922
"""Bisect for entries for all paths and their children.
924
This will use bisect to find all records for the supplied paths. It
925
will then continue to bisect for any records which are marked as
926
directories. (and renames?)
928
:param paths: A sorted list of (dir, name) pairs
929
eg: [('', 'a'), ('', 'f'), ('a/b', 'c')]
930
:return: A dictionary mapping (dir, name, file_id) => [tree_info]
932
# Map from (dir, name, file_id) => [tree_info]
935
found_dir_names = set()
937
# Directories that have been read
938
processed_dirs = set()
939
# Get the ball rolling with the first bisect for all entries.
940
newly_found = self._bisect(paths)
943
# Directories that need to be read
945
paths_to_search = set()
946
for entry_list in newly_found.itervalues():
947
for dir_name_id, trees_info in entry_list:
948
found[dir_name_id] = trees_info
949
found_dir_names.add(dir_name_id[:2])
951
for tree_info in trees_info:
952
minikind = tree_info[0]
955
# We already processed this one as a directory,
956
# we don't need to do the extra work again.
958
subdir, name, file_id = dir_name_id
959
path = osutils.pathjoin(subdir, name)
961
if path not in processed_dirs:
962
pending_dirs.add(path)
963
elif minikind == 'r':
964
# Rename, we need to directly search the target
965
# which is contained in the fingerprint column
966
dir_name = osutils.split(tree_info[1])
967
if dir_name[0] in pending_dirs:
968
# This entry will be found in the dir search
970
if dir_name not in found_dir_names:
971
paths_to_search.add(tree_info[1])
972
# Now we have a list of paths to look for directly, and
973
# directory blocks that need to be read.
974
# newly_found is mixing the keys between (dir, name) and path
975
# entries, but that is okay, because we only really care about the
977
newly_found = self._bisect(sorted(paths_to_search))
978
newly_found.update(self._bisect_dirblocks(sorted(pending_dirs)))
979
processed_dirs.update(pending_dirs)
982
def _discard_merge_parents(self):
983
"""Discard any parents trees beyond the first.
985
Note that if this fails the dirstate is corrupted.
987
After this function returns the dirstate contains 2 trees, neither of
990
self._read_header_if_needed()
991
parents = self.get_parent_ids()
994
# only require all dirblocks if we are doing a full-pass removal.
995
self._read_dirblocks_if_needed()
996
dead_patterns = set([('a', 'r'), ('a', 'a'), ('r', 'r'), ('r', 'a')])
997
def iter_entries_removable():
998
for block in self._dirblocks:
999
deleted_positions = []
1000
for pos, entry in enumerate(block[1]):
1002
if (entry[1][0][0], entry[1][1][0]) in dead_patterns:
1003
deleted_positions.append(pos)
1004
if deleted_positions:
1005
if len(deleted_positions) == len(block[1]):
1008
for pos in reversed(deleted_positions):
1010
# if the first parent is a ghost:
1011
if parents[0] in self.get_ghosts():
1012
empty_parent = [DirState.NULL_PARENT_DETAILS]
1013
for entry in iter_entries_removable():
1014
entry[1][1:] = empty_parent
1016
for entry in iter_entries_removable():
1020
self._parents = [parents[0]]
1021
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
1022
self._header_state = DirState.IN_MEMORY_MODIFIED
1024
def _empty_parent_info(self):
1025
return [DirState.NULL_PARENT_DETAILS] * (len(self._parents) -
1028
def _ensure_block(self, parent_block_index, parent_row_index, dirname):
1029
"""Ensure a block for dirname exists.
1031
This function exists to let callers which know that there is a
1032
directory dirname ensure that the block for it exists. This block can
1033
fail to exist because of demand loading, or because a directory had no
1034
children. In either case it is not an error. It is however an error to
1035
call this if there is no parent entry for the directory, and thus the
1036
function requires the coordinates of such an entry to be provided.
1038
The root row is special cased and can be indicated with a parent block
1041
:param parent_block_index: The index of the block in which dirname's row
1043
:param parent_row_index: The index in the parent block where the row
1045
:param dirname: The utf8 dirname to ensure there is a block for.
1046
:return: The index for the block.
1048
if dirname == '' and parent_row_index == 0 and parent_block_index == 0:
1049
# This is the signature of the root row, and the
1050
# contents-of-root row is always index 1
1052
# the basename of the directory must be the end of its full name.
1053
if not (parent_block_index == -1 and
1054
parent_block_index == -1 and dirname == ''):
1055
if not dirname.endswith(
1056
self._dirblocks[parent_block_index][1][parent_row_index][0][1]):
1057
raise AssertionError("bad dirname %r" % dirname)
1058
block_index, present = self._find_block_index_from_key((dirname, '', ''))
1060
## In future, when doing partial parsing, this should load and
1061
# populate the entire block.
1062
self._dirblocks.insert(block_index, (dirname, []))
1065
def _entries_to_current_state(self, new_entries):
1066
"""Load new_entries into self.dirblocks.
1068
Process new_entries into the current state object, making them the active
1069
state. The entries are grouped together by directory to form dirblocks.
1071
:param new_entries: A sorted list of entries. This function does not sort
1072
to prevent unneeded overhead when callers have a sorted list already.
1075
if new_entries[0][0][0:2] != ('', ''):
1076
raise AssertionError(
1077
"Missing root row %r" % (new_entries[0][0],))
1078
# The two blocks here are deliberate: the root block and the
1079
# contents-of-root block.
1080
self._dirblocks = [('', []), ('', [])]
1081
current_block = self._dirblocks[0][1]
1082
current_dirname = ''
1084
append_entry = current_block.append
1085
for entry in new_entries:
1086
if entry[0][0] != current_dirname:
1087
# new block - different dirname
1089
current_dirname = entry[0][0]
1090
self._dirblocks.append((current_dirname, current_block))
1091
append_entry = current_block.append
1092
# append the entry to the current block
1094
self._split_root_dirblock_into_contents()
1096
def _split_root_dirblock_into_contents(self):
1097
"""Split the root dirblocks into root and contents-of-root.
1099
After parsing by path, we end up with root entries and contents-of-root
1100
entries in the same block. This loop splits them out again.
1102
# The above loop leaves the "root block" entries mixed with the
1103
# "contents-of-root block". But we don't want an if check on
1104
# all entries, so instead we just fix it up here.
1105
if self._dirblocks[1] != ('', []):
1106
raise ValueError("bad dirblock start %r" % (self._dirblocks[1],))
1108
contents_of_root_block = []
1109
for entry in self._dirblocks[0][1]:
1110
if not entry[0][1]: # This is a root entry
1111
root_block.append(entry)
1113
contents_of_root_block.append(entry)
1114
self._dirblocks[0] = ('', root_block)
1115
self._dirblocks[1] = ('', contents_of_root_block)
1117
def _entries_for_path(self, path):
1118
"""Return a list with all the entries that match path for all ids."""
1119
dirname, basename = os.path.split(path)
1120
key = (dirname, basename, '')
1121
block_index, present = self._find_block_index_from_key(key)
1123
# the block which should contain path is absent.
1126
block = self._dirblocks[block_index][1]
1127
entry_index, _ = self._find_entry_index(key, block)
1128
# we may need to look at multiple entries at this path: walk while the specific_files match.
1129
while (entry_index < len(block) and
1130
block[entry_index][0][0:2] == key[0:2]):
1131
result.append(block[entry_index])
1135
def _entry_to_line(self, entry):
1136
"""Serialize entry to a NULL delimited line ready for _get_output_lines.
1138
:param entry: An entry_tuple as defined in the module docstring.
1140
entire_entry = list(entry[0])
1141
for tree_number, tree_data in enumerate(entry[1]):
1142
# (minikind, fingerprint, size, executable, tree_specific_string)
1143
entire_entry.extend(tree_data)
1144
# 3 for the key, 5 for the fields per tree.
1145
tree_offset = 3 + tree_number * 5
1147
entire_entry[tree_offset + 0] = tree_data[0]
1149
entire_entry[tree_offset + 2] = str(tree_data[2])
1151
entire_entry[tree_offset + 3] = DirState._to_yesno[tree_data[3]]
1152
return '\0'.join(entire_entry)
1154
def _fields_per_entry(self):
1155
"""How many null separated fields should be in each entry row.
1157
Each line now has an extra '\n' field which is not used
1158
so we just skip over it
1160
3 fields for the key
1161
+ number of fields per tree_data (5) * tree count
1164
tree_count = 1 + self._num_present_parents()
1165
return 3 + 5 * tree_count + 1
1167
def _find_block(self, key, add_if_missing=False):
1168
"""Return the block that key should be present in.
1170
:param key: A dirstate entry key.
1171
:return: The block tuple.
1173
block_index, present = self._find_block_index_from_key(key)
1175
if not add_if_missing:
1176
# check to see if key is versioned itself - we might want to
1177
# add it anyway, because dirs with no entries dont get a
1178
# dirblock at parse time.
1179
# This is an uncommon branch to take: most dirs have children,
1180
# and most code works with versioned paths.
1181
parent_base, parent_name = osutils.split(key[0])
1182
if not self._get_block_entry_index(parent_base, parent_name, 0)[3]:
1183
# some parent path has not been added - its an error to add
1185
raise errors.NotVersionedError(key[0:2], str(self))
1186
self._dirblocks.insert(block_index, (key[0], []))
1187
return self._dirblocks[block_index]
1189
def _find_block_index_from_key(self, key):
1190
"""Find the dirblock index for a key.
1192
:return: The block index, True if the block for the key is present.
1194
if key[0:2] == ('', ''):
1197
if (self._last_block_index is not None and
1198
self._dirblocks[self._last_block_index][0] == key[0]):
1199
return self._last_block_index, True
1202
block_index = bisect_dirblock(self._dirblocks, key[0], 1,
1203
cache=self._split_path_cache)
1204
# _right returns one-past-where-key is so we have to subtract
1205
# one to use it. we use _right here because there are two
1206
# '' blocks - the root, and the contents of root
1207
# we always have a minimum of 2 in self._dirblocks: root and
1208
# root-contents, and for '', we get 2 back, so this is
1209
# simple and correct:
1210
present = (block_index < len(self._dirblocks) and
1211
self._dirblocks[block_index][0] == key[0])
1212
self._last_block_index = block_index
1213
# Reset the entry index cache to the beginning of the block.
1214
self._last_entry_index = -1
1215
return block_index, present
1217
def _find_entry_index(self, key, block):
1218
"""Find the entry index for a key in a block.
1220
:return: The entry index, True if the entry for the key is present.
1222
len_block = len(block)
1224
if self._last_entry_index is not None:
1226
entry_index = self._last_entry_index + 1
1227
# A hit is when the key is after the last slot, and before or
1228
# equal to the next slot.
1229
if ((entry_index > 0 and block[entry_index - 1][0] < key) and
1230
key <= block[entry_index][0]):
1231
self._last_entry_index = entry_index
1232
present = (block[entry_index][0] == key)
1233
return entry_index, present
1236
entry_index = bisect.bisect_left(block, (key, []))
1237
present = (entry_index < len_block and
1238
block[entry_index][0] == key)
1239
self._last_entry_index = entry_index
1240
return entry_index, present
1243
def from_tree(tree, dir_state_filename, sha1_provider=None):
1244
"""Create a dirstate from a bzr Tree.
1246
:param tree: The tree which should provide parent information and
1248
:param sha1_provider: an object meeting the SHA1Provider interface.
1249
If None, a DefaultSHA1Provider is used.
1250
:return: a DirState object which is currently locked for writing.
1251
(it was locked by DirState.initialize)
1253
result = DirState.initialize(dir_state_filename,
1254
sha1_provider=sha1_provider)
1258
parent_ids = tree.get_parent_ids()
1259
num_parents = len(parent_ids)
1261
for parent_id in parent_ids:
1262
parent_tree = tree.branch.repository.revision_tree(parent_id)
1263
parent_trees.append((parent_id, parent_tree))
1264
parent_tree.lock_read()
1265
result.set_parent_trees(parent_trees, [])
1266
result.set_state_from_inventory(tree.inventory)
1268
for revid, parent_tree in parent_trees:
1269
parent_tree.unlock()
1272
# The caller won't have a chance to unlock this, so make sure we
1278
def update_by_delta(self, delta):
1279
"""Apply an inventory delta to the dirstate for tree 0
1281
This is the workhorse for apply_inventory_delta in dirstate based
1284
:param delta: An inventory delta. See Inventory.apply_delta for
1287
self._read_dirblocks_if_needed()
1288
encode = cache_utf8.encode
1291
# Accumulate parent references (path_utf8, id), to check for parentless
1292
# items or items placed under files/links/tree-references. We get
1293
# references from every item in the delta that is not a deletion and
1294
# is not itself the root.
1296
# Added ids must not be in the dirstate already. This set holds those
1299
# This loop transforms the delta to single atomic operations that can
1300
# be executed and validated.
1301
for old_path, new_path, file_id, inv_entry in sorted(
1302
inventory._check_delta_unique_old_paths(
1303
inventory._check_delta_unique_new_paths(
1304
inventory._check_delta_ids_match_entry(
1305
inventory._check_delta_new_path_entry_both_or_None(delta)))),
1307
if (file_id in insertions) or (file_id in removals):
1308
raise errors.InconsistentDelta(old_path or new_path, file_id,
1310
if old_path is not None:
1311
old_path = old_path.encode('utf-8')
1312
removals[file_id] = old_path
1314
new_ids.add(file_id)
1315
if new_path is not None:
1316
if inv_entry is None:
1317
raise errors.InconsistentDelta(new_path, file_id,
1318
"new_path with no entry")
1319
new_path = new_path.encode('utf-8')
1320
dirname, basename = osutils.split(new_path)
1321
dirname_utf8 = encode(dirname)
1323
parents.add((dirname_utf8, inv_entry.parent_id))
1324
key = (dirname, basename, file_id)
1325
minikind = DirState._kind_to_minikind[inv_entry.kind]
1327
fingerprint = inv_entry.reference_revision
1330
insertions[file_id] = (key, minikind, inv_entry.executable,
1331
fingerprint, new_path)
1332
# Transform moves into delete+add pairs
1333
if None not in (old_path, new_path):
1334
for child in self._iter_child_entries(0, old_path):
1335
if child[0][2] in insertions or child[0][2] in removals:
1337
child_dirname = child[0][0]
1338
child_basename = child[0][1]
1339
minikind = child[1][0][0]
1340
fingerprint = child[1][0][4]
1341
executable = child[1][0][3]
1342
old_child_path = osutils.pathjoin(child[0][0],
1344
removals[child[0][2]] = old_child_path
1345
child_suffix = child_dirname[len(old_path):]
1346
new_child_dirname = (new_path + child_suffix)
1347
key = (new_child_dirname, child_basename, child[0][2])
1348
new_child_path = os.path.join(new_child_dirname,
1350
insertions[child[0][2]] = (key, minikind, executable,
1351
fingerprint, new_child_path)
1352
self._check_delta_ids_absent(new_ids, delta, 0)
1354
self._apply_removals(removals.iteritems())
1355
self._apply_insertions(insertions.values())
1357
self._after_delta_check_parents(parents, 0)
1358
except errors.BzrError, e:
1359
if 'integrity error' not in str(e):
1361
# _get_entry raises BzrError when a request is inconsistent; we
1362
# want such errors to be shown as InconsistentDelta - and that
1363
# fits the behaviour we trigger.
1364
self._changes_aborted = True
1365
raise errors.InconsistentDeltaDelta(delta, "error from _get_entry.")
1367
def _apply_removals(self, removals):
1368
for file_id, path in sorted(removals, reverse=True,
1369
key=operator.itemgetter(1)):
1370
dirname, basename = osutils.split(path)
1371
block_i, entry_i, d_present, f_present = \
1372
self._get_block_entry_index(dirname, basename, 0)
1374
entry = self._dirblocks[block_i][1][entry_i]
1376
raise errors.InconsistentDelta(path, file_id,
1377
"Wrong path for old path.")
1378
if not f_present or entry[1][0][0] in 'ar':
1379
raise errors.InconsistentDelta(path, file_id,
1380
"Wrong path for old path.")
1381
if file_id != entry[0][2]:
1382
raise errors.InconsistentDelta(path, file_id,
1383
"Attempt to remove wrong has wrong id - found %r."
1385
self._make_absent(entry)
1386
# See if we have a malformed delta: deleting a directory must not
1387
# leave crud behind. This increases the number of bisects needed
1388
# substantially, but deletion or renames of large numbers of paths
1389
# is rare enough it shouldn't be an issue (famous last words?) RBC
1391
block_i, entry_i, d_present, f_present = \
1392
self._get_block_entry_index(path, '', 0)
1394
# The dir block is still present in the dirstate; this could
1395
# be due to it being in a parent tree, or a corrupt delta.
1396
for child_entry in self._dirblocks[block_i][1]:
1397
if child_entry[1][0][0] not in ('r', 'a'):
1398
raise errors.InconsistentDelta(path, entry[0][2],
1399
"The file id was deleted but its children were "
1402
def _apply_insertions(self, adds):
1404
for key, minikind, executable, fingerprint, path_utf8 in sorted(adds):
1405
self.update_minimal(key, minikind, executable, fingerprint,
1406
path_utf8=path_utf8)
1407
except errors.NotVersionedError:
1408
raise errors.InconsistentDelta(path_utf8.decode('utf8'), key[2],
1411
def update_basis_by_delta(self, delta, new_revid):
1412
"""Update the parents of this tree after a commit.
1414
This gives the tree one parent, with revision id new_revid. The
1415
inventory delta is applied to the current basis tree to generate the
1416
inventory for the parent new_revid, and all other parent trees are
1419
Note that an exception during the operation of this method will leave
1420
the dirstate in a corrupt state where it should not be saved.
1422
Finally, we expect all changes to be synchronising the basis tree with
1425
:param new_revid: The new revision id for the trees parent.
1426
:param delta: An inventory delta (see apply_inventory_delta) describing
1427
the changes from the current left most parent revision to new_revid.
1429
self._read_dirblocks_if_needed()
1430
self._discard_merge_parents()
1431
if self._ghosts != []:
1432
raise NotImplementedError(self.update_basis_by_delta)
1433
if len(self._parents) == 0:
1434
# setup a blank tree, the most simple way.
1435
empty_parent = DirState.NULL_PARENT_DETAILS
1436
for entry in self._iter_entries():
1437
entry[1].append(empty_parent)
1438
self._parents.append(new_revid)
1440
self._parents[0] = new_revid
1442
delta = sorted(delta, reverse=True)
1446
# The paths this function accepts are unicode and must be encoded as we
1448
encode = cache_utf8.encode
1449
inv_to_entry = self._inv_entry_to_details
1450
# delta is now (deletes, changes), (adds) in reverse lexographical
1452
# deletes in reverse lexographic order are safe to process in situ.
1453
# renames are not, as a rename from any path could go to a path
1454
# lexographically lower, so we transform renames into delete, add pairs,
1455
# expanding them recursively as needed.
1456
# At the same time, to reduce interface friction we convert the input
1457
# inventory entries to dirstate.
1458
root_only = ('', '')
1459
# Accumulate parent references (path_utf8, id), to check for parentless
1460
# items or items placed under files/links/tree-references. We get
1461
# references from every item in the delta that is not a deletion and
1462
# is not itself the root.
1464
# Added ids must not be in the dirstate already. This set holds those
1467
for old_path, new_path, file_id, inv_entry in delta:
1468
if inv_entry is not None and file_id != inv_entry.file_id:
1469
raise errors.InconsistentDelta(new_path, file_id,
1470
"mismatched entry file_id %r" % inv_entry)
1471
if new_path is not None:
1472
if inv_entry is None:
1473
raise errors.InconsistentDelta(new_path, file_id,
1474
"new_path with no entry")
1475
new_path_utf8 = encode(new_path)
1476
# note the parent for validation
1477
dirname_utf8, basename_utf8 = osutils.split(new_path_utf8)
1479
parents.add((dirname_utf8, inv_entry.parent_id))
1480
if old_path is None:
1481
adds.append((None, encode(new_path), file_id,
1482
inv_to_entry(inv_entry), True))
1483
new_ids.add(file_id)
1484
elif new_path is None:
1485
deletes.append((encode(old_path), None, file_id, None, True))
1486
elif (old_path, new_path) != root_only:
1488
# Because renames must preserve their children we must have
1489
# processed all relocations and removes before hand. The sort
1490
# order ensures we've examined the child paths, but we also
1491
# have to execute the removals, or the split to an add/delete
1492
# pair will result in the deleted item being reinserted, or
1493
# renamed items being reinserted twice - and possibly at the
1494
# wrong place. Splitting into a delete/add pair also simplifies
1495
# the handling of entries with ('f', ...), ('r' ...) because
1496
# the target of the 'r' is old_path here, and we add that to
1497
# deletes, meaning that the add handler does not need to check
1498
# for 'r' items on every pass.
1499
self._update_basis_apply_deletes(deletes)
1501
# Split into an add/delete pair recursively.
1502
adds.append((None, new_path_utf8, file_id,
1503
inv_to_entry(inv_entry), False))
1504
# Expunge deletes that we've seen so that deleted/renamed
1505
# children of a rename directory are handled correctly.
1506
new_deletes = reversed(list(self._iter_child_entries(1,
1508
# Remove the current contents of the tree at orig_path, and
1509
# reinsert at the correct new path.
1510
for entry in new_deletes:
1512
source_path = entry[0][0] + '/' + entry[0][1]
1514
source_path = entry[0][1]
1516
target_path = new_path_utf8 + source_path[len(old_path):]
1519
raise AssertionError("cannot rename directory to"
1521
target_path = source_path[len(old_path) + 1:]
1522
adds.append((None, target_path, entry[0][2], entry[1][1], False))
1524
(source_path, target_path, entry[0][2], None, False))
1526
(encode(old_path), new_path, file_id, None, False))
1528
# changes to just the root should not require remove/insertion
1530
changes.append((encode(old_path), encode(new_path), file_id,
1531
inv_to_entry(inv_entry)))
1532
self._check_delta_ids_absent(new_ids, delta, 1)
1534
# Finish expunging deletes/first half of renames.
1535
self._update_basis_apply_deletes(deletes)
1536
# Reinstate second half of renames and new paths.
1537
self._update_basis_apply_adds(adds)
1538
# Apply in-situ changes.
1539
self._update_basis_apply_changes(changes)
1541
self._after_delta_check_parents(parents, 1)
1542
except errors.BzrError, e:
1543
if 'integrity error' not in str(e):
1545
# _get_entry raises BzrError when a request is inconsistent; we
1546
# want such errors to be shown as InconsistentDelta - and that
1547
# fits the behaviour we trigger. Partof this is driven by dirstate
1548
# only supporting deltas that turn the basis into a closer fit to
1550
self._changes_aborted = True
1551
raise errors.InconsistentDeltaDelta(delta, "error from _get_entry.")
1553
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
1554
self._header_state = DirState.IN_MEMORY_MODIFIED
1555
self._id_index = None
1558
def _check_delta_ids_absent(self, new_ids, delta, tree_index):
1559
"""Check that none of the file_ids in new_ids are present in a tree."""
1562
id_index = self._get_id_index()
1563
for file_id in new_ids:
1564
for key in id_index.get(file_id, []):
1565
block_i, entry_i, d_present, f_present = \
1566
self._get_block_entry_index(key[0], key[1], tree_index)
1568
# In a different tree
1570
entry = self._dirblocks[block_i][1][entry_i]
1571
if entry[0][2] != file_id:
1572
# Different file_id, so not what we want.
1574
# NB: No changes made before this helper is called, so no need
1575
# to set the _changes_aborted flag.
1576
raise errors.InconsistentDelta(
1577
("%s/%s" % key[0:2]).decode('utf8'), file_id,
1578
"This file_id is new in the delta but already present in "
1581
def _update_basis_apply_adds(self, adds):
1582
"""Apply a sequence of adds to tree 1 during update_basis_by_delta.
1584
They may be adds, or renames that have been split into add/delete
1587
:param adds: A sequence of adds. Each add is a tuple:
1588
(None, new_path_utf8, file_id, (entry_details), real_add). real_add
1589
is False when the add is the second half of a remove-and-reinsert
1590
pair created to handle renames and deletes.
1592
# Adds are accumulated partly from renames, so can be in any input
1595
# adds is now in lexographic order, which places all parents before
1596
# their children, so we can process it linearly.
1598
for old_path, new_path, file_id, new_details, real_add in adds:
1599
# the entry for this file_id must be in tree 0.
1600
entry = self._get_entry(0, file_id, new_path)
1601
if entry[0] is None or entry[0][2] != file_id:
1602
self._changes_aborted = True
1603
raise errors.InconsistentDelta(new_path, file_id,
1604
'working tree does not contain new entry')
1605
if real_add and entry[1][1][0] not in absent:
1606
self._changes_aborted = True
1607
raise errors.InconsistentDelta(new_path, file_id,
1608
'The entry was considered to be a genuinely new record,'
1609
' but there was already an old record for it.')
1610
# We don't need to update the target of an 'r' because the handling
1611
# of renames turns all 'r' situations into a delete at the original
1613
entry[1][1] = new_details
1615
def _update_basis_apply_changes(self, changes):
1616
"""Apply a sequence of changes to tree 1 during update_basis_by_delta.
1618
:param adds: A sequence of changes. Each change is a tuple:
1619
(path_utf8, path_utf8, file_id, (entry_details))
1622
for old_path, new_path, file_id, new_details in changes:
1623
# the entry for this file_id must be in tree 0.
1624
entry = self._get_entry(0, file_id, new_path)
1625
if entry[0] is None or entry[0][2] != file_id:
1626
self._changes_aborted = True
1627
raise errors.InconsistentDelta(new_path, file_id,
1628
'working tree does not contain new entry')
1629
if (entry[1][0][0] in absent or
1630
entry[1][1][0] in absent):
1631
self._changes_aborted = True
1632
raise errors.InconsistentDelta(new_path, file_id,
1633
'changed considered absent')
1634
entry[1][1] = new_details
1636
def _update_basis_apply_deletes(self, deletes):
1637
"""Apply a sequence of deletes to tree 1 during update_basis_by_delta.
1639
They may be deletes, or renames that have been split into add/delete
1642
:param deletes: A sequence of deletes. Each delete is a tuple:
1643
(old_path_utf8, new_path_utf8, file_id, None, real_delete).
1644
real_delete is True when the desired outcome is an actual deletion
1645
rather than the rename handling logic temporarily deleting a path
1646
during the replacement of a parent.
1648
null = DirState.NULL_PARENT_DETAILS
1649
for old_path, new_path, file_id, _, real_delete in deletes:
1650
if real_delete != (new_path is None):
1651
raise AssertionError("bad delete delta")
1652
# the entry for this file_id must be in tree 1.
1653
dirname, basename = osutils.split(old_path)
1654
block_index, entry_index, dir_present, file_present = \
1655
self._get_block_entry_index(dirname, basename, 1)
1656
if not file_present:
1657
self._changes_aborted = True
1658
raise errors.InconsistentDelta(old_path, file_id,
1659
'basis tree does not contain removed entry')
1660
entry = self._dirblocks[block_index][1][entry_index]
1661
if entry[0][2] != file_id:
1662
self._changes_aborted = True
1663
raise errors.InconsistentDelta(old_path, file_id,
1664
'mismatched file_id in tree 1')
1666
if entry[1][0][0] != 'a':
1667
self._changes_aborted = True
1668
raise errors.InconsistentDelta(old_path, file_id,
1669
'This was marked as a real delete, but the WT state'
1670
' claims that it still exists and is versioned.')
1671
del self._dirblocks[block_index][1][entry_index]
1673
if entry[1][0][0] == 'a':
1674
self._changes_aborted = True
1675
raise errors.InconsistentDelta(old_path, file_id,
1676
'The entry was considered a rename, but the source path'
1677
' is marked as absent.')
1678
# For whatever reason, we were asked to rename an entry
1679
# that was originally marked as deleted. This could be
1680
# because we are renaming the parent directory, and the WT
1681
# current state has the file marked as deleted.
1682
elif entry[1][0][0] == 'r':
1683
# implement the rename
1684
del self._dirblocks[block_index][1][entry_index]
1686
# it is being resurrected here, so blank it out temporarily.
1687
self._dirblocks[block_index][1][entry_index][1][1] = null
1689
def _after_delta_check_parents(self, parents, index):
1690
"""Check that parents required by the delta are all intact.
1692
:param parents: An iterable of (path_utf8, file_id) tuples which are
1693
required to be present in tree 'index' at path_utf8 with id file_id
1695
:param index: The column in the dirstate to check for parents in.
1697
for dirname_utf8, file_id in parents:
1698
# Get the entry - the ensures that file_id, dirname_utf8 exists and
1699
# has the right file id.
1700
entry = self._get_entry(index, file_id, dirname_utf8)
1701
if entry[1] is None:
1702
self._changes_aborted = True
1703
raise errors.InconsistentDelta(dirname_utf8.decode('utf8'),
1704
file_id, "This parent is not present.")
1705
# Parents of things must be directories
1706
if entry[1][index][0] != 'd':
1707
self._changes_aborted = True
1708
raise errors.InconsistentDelta(dirname_utf8.decode('utf8'),
1709
file_id, "This parent is not a directory.")
1711
def _observed_sha1(self, entry, sha1, stat_value,
1712
_stat_to_minikind=_stat_to_minikind, _pack_stat=pack_stat):
1713
"""Note the sha1 of a file.
1715
:param entry: The entry the sha1 is for.
1716
:param sha1: The observed sha1.
1717
:param stat_value: The os.lstat for the file.
1720
minikind = _stat_to_minikind[stat_value.st_mode & 0170000]
1724
packed_stat = _pack_stat(stat_value)
1726
if self._cutoff_time is None:
1727
self._sha_cutoff_time()
1728
if (stat_value.st_mtime < self._cutoff_time
1729
and stat_value.st_ctime < self._cutoff_time):
1730
entry[1][0] = ('f', sha1, entry[1][0][2], entry[1][0][3],
1732
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
1734
def _sha_cutoff_time(self):
1735
"""Return cutoff time.
1737
Files modified more recently than this time are at risk of being
1738
undetectably modified and so can't be cached.
1740
# Cache the cutoff time as long as we hold a lock.
1741
# time.time() isn't super expensive (approx 3.38us), but
1742
# when you call it 50,000 times it adds up.
1743
# For comparison, os.lstat() costs 7.2us if it is hot.
1744
self._cutoff_time = int(time.time()) - 3
1745
return self._cutoff_time
1747
def _lstat(self, abspath, entry):
1748
"""Return the os.lstat value for this path."""
1749
return os.lstat(abspath)
1751
def _sha1_file_and_mutter(self, abspath):
1752
# when -Dhashcache is turned on, this is monkey-patched in to log
1754
trace.mutter("dirstate sha1 " + abspath)
1755
return self._sha1_provider.sha1(abspath)
1757
def _is_executable(self, mode, old_executable):
1758
"""Is this file executable?"""
1759
return bool(S_IEXEC & mode)
1761
def _is_executable_win32(self, mode, old_executable):
1762
"""On win32 the executable bit is stored in the dirstate."""
1763
return old_executable
1765
if sys.platform == 'win32':
1766
_is_executable = _is_executable_win32
1768
def _read_link(self, abspath, old_link):
1769
"""Read the target of a symlink"""
1770
# TODO: jam 200700301 On Win32, this could just return the value
1771
# already in memory. However, this really needs to be done at a
1772
# higher level, because there either won't be anything on disk,
1773
# or the thing on disk will be a file.
1774
fs_encoding = osutils._fs_enc
1775
if isinstance(abspath, unicode):
1776
# abspath is defined as the path to pass to lstat. readlink is
1777
# buggy in python < 2.6 (it doesn't encode unicode path into FS
1778
# encoding), so we need to encode ourselves knowing that unicode
1779
# paths are produced by UnicodeDirReader on purpose.
1780
abspath = abspath.encode(fs_encoding)
1781
target = os.readlink(abspath)
1782
if fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1783
# Change encoding if needed
1784
target = target.decode(fs_encoding).encode('UTF-8')
1787
def get_ghosts(self):
1788
"""Return a list of the parent tree revision ids that are ghosts."""
1789
self._read_header_if_needed()
1792
def get_lines(self):
1793
"""Serialise the entire dirstate to a sequence of lines."""
1794
if (self._header_state == DirState.IN_MEMORY_UNMODIFIED and
1795
self._dirblock_state == DirState.IN_MEMORY_UNMODIFIED):
1796
# read whats on disk.
1797
self._state_file.seek(0)
1798
return self._state_file.readlines()
1800
lines.append(self._get_parents_line(self.get_parent_ids()))
1801
lines.append(self._get_ghosts_line(self._ghosts))
1802
# append the root line which is special cased
1803
lines.extend(map(self._entry_to_line, self._iter_entries()))
1804
return self._get_output_lines(lines)
1806
def _get_ghosts_line(self, ghost_ids):
1807
"""Create a line for the state file for ghost information."""
1808
return '\0'.join([str(len(ghost_ids))] + ghost_ids)
1810
def _get_parents_line(self, parent_ids):
1811
"""Create a line for the state file for parents information."""
1812
return '\0'.join([str(len(parent_ids))] + parent_ids)
1814
def _get_fields_to_entry(self):
1815
"""Get a function which converts entry fields into a entry record.
1817
This handles size and executable, as well as parent records.
1819
:return: A function which takes a list of fields, and returns an
1820
appropriate record for storing in memory.
1822
# This is intentionally unrolled for performance
1823
num_present_parents = self._num_present_parents()
1824
if num_present_parents == 0:
1825
def fields_to_entry_0_parents(fields, _int=int):
1826
path_name_file_id_key = (fields[0], fields[1], fields[2])
1827
return (path_name_file_id_key, [
1829
fields[3], # minikind
1830
fields[4], # fingerprint
1831
_int(fields[5]), # size
1832
fields[6] == 'y', # executable
1833
fields[7], # packed_stat or revision_id
1835
return fields_to_entry_0_parents
1836
elif num_present_parents == 1:
1837
def fields_to_entry_1_parent(fields, _int=int):
1838
path_name_file_id_key = (fields[0], fields[1], fields[2])
1839
return (path_name_file_id_key, [
1841
fields[3], # minikind
1842
fields[4], # fingerprint
1843
_int(fields[5]), # size
1844
fields[6] == 'y', # executable
1845
fields[7], # packed_stat or revision_id
1848
fields[8], # minikind
1849
fields[9], # fingerprint
1850
_int(fields[10]), # size
1851
fields[11] == 'y', # executable
1852
fields[12], # packed_stat or revision_id
1855
return fields_to_entry_1_parent
1856
elif num_present_parents == 2:
1857
def fields_to_entry_2_parents(fields, _int=int):
1858
path_name_file_id_key = (fields[0], fields[1], fields[2])
1859
return (path_name_file_id_key, [
1861
fields[3], # minikind
1862
fields[4], # fingerprint
1863
_int(fields[5]), # size
1864
fields[6] == 'y', # executable
1865
fields[7], # packed_stat or revision_id
1868
fields[8], # minikind
1869
fields[9], # fingerprint
1870
_int(fields[10]), # size
1871
fields[11] == 'y', # executable
1872
fields[12], # packed_stat or revision_id
1875
fields[13], # minikind
1876
fields[14], # fingerprint
1877
_int(fields[15]), # size
1878
fields[16] == 'y', # executable
1879
fields[17], # packed_stat or revision_id
1882
return fields_to_entry_2_parents
1884
def fields_to_entry_n_parents(fields, _int=int):
1885
path_name_file_id_key = (fields[0], fields[1], fields[2])
1886
trees = [(fields[cur], # minikind
1887
fields[cur+1], # fingerprint
1888
_int(fields[cur+2]), # size
1889
fields[cur+3] == 'y', # executable
1890
fields[cur+4], # stat or revision_id
1891
) for cur in xrange(3, len(fields)-1, 5)]
1892
return path_name_file_id_key, trees
1893
return fields_to_entry_n_parents
1895
def get_parent_ids(self):
1896
"""Return a list of the parent tree ids for the directory state."""
1897
self._read_header_if_needed()
1898
return list(self._parents)
1900
def _get_block_entry_index(self, dirname, basename, tree_index):
1901
"""Get the coordinates for a path in the state structure.
1903
:param dirname: The utf8 dirname to lookup.
1904
:param basename: The utf8 basename to lookup.
1905
:param tree_index: The index of the tree for which this lookup should
1907
:return: A tuple describing where the path is located, or should be
1908
inserted. The tuple contains four fields: the block index, the row
1909
index, the directory is present (boolean), the entire path is
1910
present (boolean). There is no guarantee that either
1911
coordinate is currently reachable unless the found field for it is
1912
True. For instance, a directory not present in the searched tree
1913
may be returned with a value one greater than the current highest
1914
block offset. The directory present field will always be True when
1915
the path present field is True. The directory present field does
1916
NOT indicate that the directory is present in the searched tree,
1917
rather it indicates that there are at least some files in some
1920
self._read_dirblocks_if_needed()
1921
key = dirname, basename, ''
1922
block_index, present = self._find_block_index_from_key(key)
1924
# no such directory - return the dir index and 0 for the row.
1925
return block_index, 0, False, False
1926
block = self._dirblocks[block_index][1] # access the entries only
1927
entry_index, present = self._find_entry_index(key, block)
1928
# linear search through entries at this path to find the one
1930
while entry_index < len(block) and block[entry_index][0][1] == basename:
1931
if block[entry_index][1][tree_index][0] not in 'ar':
1932
# neither absent or relocated
1933
return block_index, entry_index, True, True
1935
return block_index, entry_index, True, False
1937
def _get_entry(self, tree_index, fileid_utf8=None, path_utf8=None, include_deleted=False):
1938
"""Get the dirstate entry for path in tree tree_index.
1940
If either file_id or path is supplied, it is used as the key to lookup.
1941
If both are supplied, the fastest lookup is used, and an error is
1942
raised if they do not both point at the same row.
1944
:param tree_index: The index of the tree we wish to locate this path
1945
in. If the path is present in that tree, the entry containing its
1946
details is returned, otherwise (None, None) is returned
1947
0 is the working tree, higher indexes are successive parent
1949
:param fileid_utf8: A utf8 file_id to look up.
1950
:param path_utf8: An utf8 path to be looked up.
1951
:param include_deleted: If True, and performing a lookup via
1952
fileid_utf8 rather than path_utf8, return an entry for deleted
1954
:return: The dirstate entry tuple for path, or (None, None)
1956
self._read_dirblocks_if_needed()
1957
if path_utf8 is not None:
1958
if type(path_utf8) is not str:
1959
raise errors.BzrError('path_utf8 is not a str: %s %r'
1960
% (type(path_utf8), path_utf8))
1961
# path lookups are faster
1962
dirname, basename = osutils.split(path_utf8)
1963
block_index, entry_index, dir_present, file_present = \
1964
self._get_block_entry_index(dirname, basename, tree_index)
1965
if not file_present:
1967
entry = self._dirblocks[block_index][1][entry_index]
1968
if not (entry[0][2] and entry[1][tree_index][0] not in ('a', 'r')):
1969
raise AssertionError('unversioned entry?')
1971
if entry[0][2] != fileid_utf8:
1972
self._changes_aborted = True
1973
raise errors.BzrError('integrity error ? : mismatching'
1974
' tree_index, file_id and path')
1977
possible_keys = self._get_id_index().get(fileid_utf8, None)
1978
if not possible_keys:
1980
for key in possible_keys:
1981
block_index, present = \
1982
self._find_block_index_from_key(key)
1983
# strange, probably indicates an out of date
1984
# id index - for now, allow this.
1987
# WARNING: DO not change this code to use _get_block_entry_index
1988
# as that function is not suitable: it does not use the key
1989
# to lookup, and thus the wrong coordinates are returned.
1990
block = self._dirblocks[block_index][1]
1991
entry_index, present = self._find_entry_index(key, block)
1993
entry = self._dirblocks[block_index][1][entry_index]
1994
if entry[1][tree_index][0] in 'fdlt':
1995
# this is the result we are looking for: the
1996
# real home of this file_id in this tree.
1998
if entry[1][tree_index][0] == 'a':
1999
# there is no home for this entry in this tree
2003
if entry[1][tree_index][0] != 'r':
2004
raise AssertionError(
2005
"entry %r has invalid minikind %r for tree %r" \
2007
entry[1][tree_index][0],
2009
real_path = entry[1][tree_index][1]
2010
return self._get_entry(tree_index, fileid_utf8=fileid_utf8,
2011
path_utf8=real_path)
2015
def initialize(cls, path, sha1_provider=None):
2016
"""Create a new dirstate on path.
2018
The new dirstate will be an empty tree - that is it has no parents,
2019
and only a root node - which has id ROOT_ID.
2021
:param path: The name of the file for the dirstate.
2022
:param sha1_provider: an object meeting the SHA1Provider interface.
2023
If None, a DefaultSHA1Provider is used.
2024
:return: A write-locked DirState object.
2026
# This constructs a new DirState object on a path, sets the _state_file
2027
# to a new empty file for that path. It then calls _set_data() with our
2028
# stock empty dirstate information - a root with ROOT_ID, no children,
2029
# and no parents. Finally it calls save() to ensure that this data will
2031
if sha1_provider is None:
2032
sha1_provider = DefaultSHA1Provider()
2033
result = cls(path, sha1_provider)
2034
# root dir and root dir contents with no children.
2035
empty_tree_dirblocks = [('', []), ('', [])]
2036
# a new root directory, with a NULLSTAT.
2037
empty_tree_dirblocks[0][1].append(
2038
(('', '', inventory.ROOT_ID), [
2039
('d', '', 0, False, DirState.NULLSTAT),
2043
result._set_data([], empty_tree_dirblocks)
2051
def _inv_entry_to_details(inv_entry):
2052
"""Convert an inventory entry (from a revision tree) to state details.
2054
:param inv_entry: An inventory entry whose sha1 and link targets can be
2055
relied upon, and which has a revision set.
2056
:return: A details tuple - the details for a single tree at a path +
2059
kind = inv_entry.kind
2060
minikind = DirState._kind_to_minikind[kind]
2061
tree_data = inv_entry.revision
2062
if kind == 'directory':
2066
elif kind == 'symlink':
2067
if inv_entry.symlink_target is None:
2070
fingerprint = inv_entry.symlink_target.encode('utf8')
2073
elif kind == 'file':
2074
fingerprint = inv_entry.text_sha1 or ''
2075
size = inv_entry.text_size or 0
2076
executable = inv_entry.executable
2077
elif kind == 'tree-reference':
2078
fingerprint = inv_entry.reference_revision or ''
2082
raise Exception("can't pack %s" % inv_entry)
2083
return (minikind, fingerprint, size, executable, tree_data)
2085
def _iter_child_entries(self, tree_index, path_utf8):
2086
"""Iterate over all the entries that are children of path_utf.
2088
This only returns entries that are present (not in 'a', 'r') in
2089
tree_index. tree_index data is not refreshed, so if tree 0 is used,
2090
results may differ from that obtained if paths were statted to
2091
determine what ones were directories.
2093
Asking for the children of a non-directory will return an empty
2097
next_pending_dirs = [path_utf8]
2099
while next_pending_dirs:
2100
pending_dirs = next_pending_dirs
2101
next_pending_dirs = []
2102
for path in pending_dirs:
2103
block_index, present = self._find_block_index_from_key(
2105
if block_index == 0:
2107
if len(self._dirblocks) == 1:
2108
# asked for the children of the root with no other
2112
# children of a non-directory asked for.
2114
block = self._dirblocks[block_index]
2115
for entry in block[1]:
2116
kind = entry[1][tree_index][0]
2117
if kind not in absent:
2121
path = entry[0][0] + '/' + entry[0][1]
2124
next_pending_dirs.append(path)
2126
def _iter_entries(self):
2127
"""Iterate over all the entries in the dirstate.
2129
Each yelt item is an entry in the standard format described in the
2130
docstring of bzrlib.dirstate.
2132
self._read_dirblocks_if_needed()
2133
for directory in self._dirblocks:
2134
for entry in directory[1]:
2137
def _get_id_index(self):
2138
"""Get an id index of self._dirblocks."""
2139
if self._id_index is None:
2141
for key, tree_details in self._iter_entries():
2142
id_index.setdefault(key[2], set()).add(key)
2143
self._id_index = id_index
2144
return self._id_index
2146
def _get_output_lines(self, lines):
2147
"""Format lines for final output.
2149
:param lines: A sequence of lines containing the parents list and the
2152
output_lines = [DirState.HEADER_FORMAT_3]
2153
lines.append('') # a final newline
2154
inventory_text = '\0\n\0'.join(lines)
2155
output_lines.append('crc32: %s\n' % (zlib.crc32(inventory_text),))
2156
# -3, 1 for num parents, 1 for ghosts, 1 for final newline
2157
num_entries = len(lines)-3
2158
output_lines.append('num_entries: %s\n' % (num_entries,))
2159
output_lines.append(inventory_text)
2162
def _make_deleted_row(self, fileid_utf8, parents):
2163
"""Return a deleted row for fileid_utf8."""
2164
return ('/', 'RECYCLED.BIN', 'file', fileid_utf8, 0, DirState.NULLSTAT,
2167
def _num_present_parents(self):
2168
"""The number of parent entries in each record row."""
2169
return len(self._parents) - len(self._ghosts)
2172
def on_file(path, sha1_provider=None):
2173
"""Construct a DirState on the file at path "path".
2175
:param path: The path at which the dirstate file on disk should live.
2176
:param sha1_provider: an object meeting the SHA1Provider interface.
2177
If None, a DefaultSHA1Provider is used.
2178
:return: An unlocked DirState object, associated with the given path.
2180
if sha1_provider is None:
2181
sha1_provider = DefaultSHA1Provider()
2182
result = DirState(path, sha1_provider)
2185
def _read_dirblocks_if_needed(self):
2186
"""Read in all the dirblocks from the file if they are not in memory.
2188
This populates self._dirblocks, and sets self._dirblock_state to
2189
IN_MEMORY_UNMODIFIED. It is not currently ready for incremental block
2192
self._read_header_if_needed()
2193
if self._dirblock_state == DirState.NOT_IN_MEMORY:
2194
_read_dirblocks(self)
2196
def _read_header(self):
2197
"""This reads in the metadata header, and the parent ids.
2199
After reading in, the file should be positioned at the null
2200
just before the start of the first record in the file.
2202
:return: (expected crc checksum, number of entries, parent list)
2204
self._read_prelude()
2205
parent_line = self._state_file.readline()
2206
info = parent_line.split('\0')
2207
num_parents = int(info[0])
2208
self._parents = info[1:-1]
2209
ghost_line = self._state_file.readline()
2210
info = ghost_line.split('\0')
2211
num_ghosts = int(info[1])
2212
self._ghosts = info[2:-1]
2213
self._header_state = DirState.IN_MEMORY_UNMODIFIED
2214
self._end_of_header = self._state_file.tell()
2216
def _read_header_if_needed(self):
2217
"""Read the header of the dirstate file if needed."""
2218
# inline this as it will be called a lot
2219
if not self._lock_token:
2220
raise errors.ObjectNotLocked(self)
2221
if self._header_state == DirState.NOT_IN_MEMORY:
2224
def _read_prelude(self):
2225
"""Read in the prelude header of the dirstate file.
2227
This only reads in the stuff that is not connected to the crc
2228
checksum. The position will be correct to read in the rest of
2229
the file and check the checksum after this point.
2230
The next entry in the file should be the number of parents,
2231
and their ids. Followed by a newline.
2233
header = self._state_file.readline()
2234
if header != DirState.HEADER_FORMAT_3:
2235
raise errors.BzrError(
2236
'invalid header line: %r' % (header,))
2237
crc_line = self._state_file.readline()
2238
if not crc_line.startswith('crc32: '):
2239
raise errors.BzrError('missing crc32 checksum: %r' % crc_line)
2240
self.crc_expected = int(crc_line[len('crc32: '):-1])
2241
num_entries_line = self._state_file.readline()
2242
if not num_entries_line.startswith('num_entries: '):
2243
raise errors.BzrError('missing num_entries line')
2244
self._num_entries = int(num_entries_line[len('num_entries: '):-1])
2246
def sha1_from_stat(self, path, stat_result, _pack_stat=pack_stat):
2247
"""Find a sha1 given a stat lookup."""
2248
return self._get_packed_stat_index().get(_pack_stat(stat_result), None)
2250
def _get_packed_stat_index(self):
2251
"""Get a packed_stat index of self._dirblocks."""
2252
if self._packed_stat_index is None:
2254
for key, tree_details in self._iter_entries():
2255
if tree_details[0][0] == 'f':
2256
index[tree_details[0][4]] = tree_details[0][1]
2257
self._packed_stat_index = index
2258
return self._packed_stat_index
2261
"""Save any pending changes created during this session.
2263
We reuse the existing file, because that prevents race conditions with
2264
file creation, and use oslocks on it to prevent concurrent modification
2265
and reads - because dirstate's incremental data aggregation is not
2266
compatible with reading a modified file, and replacing a file in use by
2267
another process is impossible on Windows.
2269
A dirstate in read only mode should be smart enough though to validate
2270
that the file has not changed, and otherwise discard its cache and
2271
start over, to allow for fine grained read lock duration, so 'status'
2272
wont block 'commit' - for example.
2274
if self._changes_aborted:
2275
# Should this be a warning? For now, I'm expecting that places that
2276
# mark it inconsistent will warn, making a warning here redundant.
2277
trace.mutter('Not saving DirState because '
2278
'_changes_aborted is set.')
2280
if (self._header_state == DirState.IN_MEMORY_MODIFIED or
2281
self._dirblock_state == DirState.IN_MEMORY_MODIFIED):
2283
grabbed_write_lock = False
2284
if self._lock_state != 'w':
2285
grabbed_write_lock, new_lock = self._lock_token.temporary_write_lock()
2286
# Switch over to the new lock, as the old one may be closed.
2287
# TODO: jam 20070315 We should validate the disk file has
2288
# not changed contents. Since temporary_write_lock may
2289
# not be an atomic operation.
2290
self._lock_token = new_lock
2291
self._state_file = new_lock.f
2292
if not grabbed_write_lock:
2293
# We couldn't grab a write lock, so we switch back to a read one
2296
self._state_file.seek(0)
2297
self._state_file.writelines(self.get_lines())
2298
self._state_file.truncate()
2299
self._state_file.flush()
2300
self._header_state = DirState.IN_MEMORY_UNMODIFIED
2301
self._dirblock_state = DirState.IN_MEMORY_UNMODIFIED
2303
if grabbed_write_lock:
2304
self._lock_token = self._lock_token.restore_read_lock()
2305
self._state_file = self._lock_token.f
2306
# TODO: jam 20070315 We should validate the disk file has
2307
# not changed contents. Since restore_read_lock may
2308
# not be an atomic operation.
2310
def _set_data(self, parent_ids, dirblocks):
2311
"""Set the full dirstate data in memory.
2313
This is an internal function used to completely replace the objects
2314
in memory state. It puts the dirstate into state 'full-dirty'.
2316
:param parent_ids: A list of parent tree revision ids.
2317
:param dirblocks: A list containing one tuple for each directory in the
2318
tree. Each tuple contains the directory path and a list of entries
2319
found in that directory.
2321
# our memory copy is now authoritative.
2322
self._dirblocks = dirblocks
2323
self._header_state = DirState.IN_MEMORY_MODIFIED
2324
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
2325
self._parents = list(parent_ids)
2326
self._id_index = None
2327
self._packed_stat_index = None
2329
def set_path_id(self, path, new_id):
2330
"""Change the id of path to new_id in the current working tree.
2332
:param path: The path inside the tree to set - '' is the root, 'foo'
2333
is the path foo in the root.
2334
:param new_id: The new id to assign to the path. This must be a utf8
2335
file id (not unicode, and not None).
2337
self._read_dirblocks_if_needed()
2339
# TODO: logic not written
2340
raise NotImplementedError(self.set_path_id)
2341
# TODO: check new id is unique
2342
entry = self._get_entry(0, path_utf8=path)
2343
if entry[0][2] == new_id:
2344
# Nothing to change.
2346
# mark the old path absent, and insert a new root path
2347
self._make_absent(entry)
2348
self.update_minimal(('', '', new_id), 'd',
2349
path_utf8='', packed_stat=entry[1][0][4])
2350
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
2351
if self._id_index is not None:
2352
self._id_index.setdefault(new_id, set()).add(entry[0])
2354
def set_parent_trees(self, trees, ghosts):
2355
"""Set the parent trees for the dirstate.
2357
:param trees: A list of revision_id, tree tuples. tree must be provided
2358
even if the revision_id refers to a ghost: supply an empty tree in
2360
:param ghosts: A list of the revision_ids that are ghosts at the time
2363
# TODO: generate a list of parent indexes to preserve to save
2364
# processing specific parent trees. In the common case one tree will
2365
# be preserved - the left most parent.
2366
# TODO: if the parent tree is a dirstate, we might want to walk them
2367
# all by path in parallel for 'optimal' common-case performance.
2368
# generate new root row.
2369
self._read_dirblocks_if_needed()
2370
# TODO future sketch: Examine the existing parents to generate a change
2371
# map and then walk the new parent trees only, mapping them into the
2372
# dirstate. Walk the dirstate at the same time to remove unreferenced
2375
# sketch: loop over all entries in the dirstate, cherry picking
2376
# entries from the parent trees, if they are not ghost trees.
2377
# after we finish walking the dirstate, all entries not in the dirstate
2378
# are deletes, so we want to append them to the end as per the design
2379
# discussions. So do a set difference on ids with the parents to
2380
# get deletes, and add them to the end.
2381
# During the update process we need to answer the following questions:
2382
# - find other keys containing a fileid in order to create cross-path
2383
# links. We dont't trivially use the inventory from other trees
2384
# because this leads to either double touching, or to accessing
2386
# - find other keys containing a path
2387
# We accumulate each entry via this dictionary, including the root
2390
# we could do parallel iterators, but because file id data may be
2391
# scattered throughout, we dont save on index overhead: we have to look
2392
# at everything anyway. We can probably save cycles by reusing parent
2393
# data and doing an incremental update when adding an additional
2394
# parent, but for now the common cases are adding a new parent (merge),
2395
# and replacing completely (commit), and commit is more common: so
2396
# optimise merge later.
2398
# ---- start generation of full tree mapping data
2399
# what trees should we use?
2400
parent_trees = [tree for rev_id, tree in trees if rev_id not in ghosts]
2401
# how many trees do we end up with
2402
parent_count = len(parent_trees)
2404
# one: the current tree
2405
for entry in self._iter_entries():
2406
# skip entries not in the current tree
2407
if entry[1][0][0] in 'ar': # absent, relocated
2409
by_path[entry[0]] = [entry[1][0]] + \
2410
[DirState.NULL_PARENT_DETAILS] * parent_count
2411
id_index[entry[0][2]] = set([entry[0]])
2413
# now the parent trees:
2414
for tree_index, tree in enumerate(parent_trees):
2415
# the index is off by one, adjust it.
2416
tree_index = tree_index + 1
2417
# when we add new locations for a fileid we need these ranges for
2418
# any fileid in this tree as we set the by_path[id] to:
2419
# already_processed_tree_details + new_details + new_location_suffix
2420
# the suffix is from tree_index+1:parent_count+1.
2421
new_location_suffix = [DirState.NULL_PARENT_DETAILS] * (parent_count - tree_index)
2422
# now stitch in all the entries from this tree
2423
for path, entry in tree.inventory.iter_entries_by_dir():
2424
# here we process each trees details for each item in the tree.
2425
# we first update any existing entries for the id at other paths,
2426
# then we either create or update the entry for the id at the
2427
# right path, and finally we add (if needed) a mapping from
2428
# file_id to this path. We do it in this order to allow us to
2429
# avoid checking all known paths for the id when generating a
2430
# new entry at this path: by adding the id->path mapping last,
2431
# all the mappings are valid and have correct relocation
2432
# records where needed.
2433
file_id = entry.file_id
2434
path_utf8 = path.encode('utf8')
2435
dirname, basename = osutils.split(path_utf8)
2436
new_entry_key = (dirname, basename, file_id)
2437
# tree index consistency: All other paths for this id in this tree
2438
# index must point to the correct path.
2439
for entry_key in id_index.setdefault(file_id, set()):
2440
# TODO:PROFILING: It might be faster to just update
2441
# rather than checking if we need to, and then overwrite
2442
# the one we are located at.
2443
if entry_key != new_entry_key:
2444
# this file id is at a different path in one of the
2445
# other trees, so put absent pointers there
2446
# This is the vertical axis in the matrix, all pointing
2448
by_path[entry_key][tree_index] = ('r', path_utf8, 0, False, '')
2449
# by path consistency: Insert into an existing path record (trivial), or
2450
# add a new one with relocation pointers for the other tree indexes.
2451
if new_entry_key in id_index[file_id]:
2452
# there is already an entry where this data belongs, just insert it.
2453
by_path[new_entry_key][tree_index] = \
2454
self._inv_entry_to_details(entry)
2456
# add relocated entries to the horizontal axis - this row
2457
# mapping from path,id. We need to look up the correct path
2458
# for the indexes from 0 to tree_index -1
2460
for lookup_index in xrange(tree_index):
2461
# boundary case: this is the first occurence of file_id
2462
# so there are no id_indexs, possibly take this out of
2464
if not len(id_index[file_id]):
2465
new_details.append(DirState.NULL_PARENT_DETAILS)
2467
# grab any one entry, use it to find the right path.
2468
# TODO: optimise this to reduce memory use in highly
2469
# fragmented situations by reusing the relocation
2471
a_key = iter(id_index[file_id]).next()
2472
if by_path[a_key][lookup_index][0] in ('r', 'a'):
2473
# its a pointer or missing statement, use it as is.
2474
new_details.append(by_path[a_key][lookup_index])
2476
# we have the right key, make a pointer to it.
2477
real_path = ('/'.join(a_key[0:2])).strip('/')
2478
new_details.append(('r', real_path, 0, False, ''))
2479
new_details.append(self._inv_entry_to_details(entry))
2480
new_details.extend(new_location_suffix)
2481
by_path[new_entry_key] = new_details
2482
id_index[file_id].add(new_entry_key)
2483
# --- end generation of full tree mappings
2485
# sort and output all the entries
2486
new_entries = self._sort_entries(by_path.items())
2487
self._entries_to_current_state(new_entries)
2488
self._parents = [rev_id for rev_id, tree in trees]
2489
self._ghosts = list(ghosts)
2490
self._header_state = DirState.IN_MEMORY_MODIFIED
2491
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
2492
self._id_index = id_index
2494
def _sort_entries(self, entry_list):
2495
"""Given a list of entries, sort them into the right order.
2497
This is done when constructing a new dirstate from trees - normally we
2498
try to keep everything in sorted blocks all the time, but sometimes
2499
it's easier to sort after the fact.
2502
# sort by: directory parts, file name, file id
2503
return entry[0][0].split('/'), entry[0][1], entry[0][2]
2504
return sorted(entry_list, key=_key)
2506
def set_state_from_inventory(self, new_inv):
2507
"""Set new_inv as the current state.
2509
This API is called by tree transform, and will usually occur with
2510
existing parent trees.
2512
:param new_inv: The inventory object to set current state from.
2514
if 'evil' in debug.debug_flags:
2515
trace.mutter_callsite(1,
2516
"set_state_from_inventory called; please mutate the tree instead")
2517
self._read_dirblocks_if_needed()
2519
# Two iterators: current data and new data, both in dirblock order.
2520
# We zip them together, which tells about entries that are new in the
2521
# inventory, or removed in the inventory, or present in both and
2524
# You might think we could just synthesize a new dirstate directly
2525
# since we're processing it in the right order. However, we need to
2526
# also consider there may be any number of parent trees and relocation
2527
# pointers, and we don't want to duplicate that here.
2528
new_iterator = new_inv.iter_entries_by_dir()
2529
# we will be modifying the dirstate, so we need a stable iterator. In
2530
# future we might write one, for now we just clone the state into a
2531
# list - which is a shallow copy.
2532
old_iterator = iter(list(self._iter_entries()))
2533
# both must have roots so this is safe:
2534
current_new = new_iterator.next()
2535
current_old = old_iterator.next()
2536
def advance(iterator):
2538
return iterator.next()
2539
except StopIteration:
2541
while current_new or current_old:
2542
# skip entries in old that are not really there
2543
if current_old and current_old[1][0][0] in 'ar':
2544
# relocated or absent
2545
current_old = advance(old_iterator)
2548
# convert new into dirblock style
2549
new_path_utf8 = current_new[0].encode('utf8')
2550
new_dirname, new_basename = osutils.split(new_path_utf8)
2551
new_id = current_new[1].file_id
2552
new_entry_key = (new_dirname, new_basename, new_id)
2553
current_new_minikind = \
2554
DirState._kind_to_minikind[current_new[1].kind]
2555
if current_new_minikind == 't':
2556
fingerprint = current_new[1].reference_revision or ''
2558
# We normally only insert or remove records, or update
2559
# them when it has significantly changed. Then we want to
2560
# erase its fingerprint. Unaffected records should
2561
# normally not be updated at all.
2564
# for safety disable variables
2565
new_path_utf8 = new_dirname = new_basename = new_id = \
2566
new_entry_key = None
2567
# 5 cases, we dont have a value that is strictly greater than everything, so
2568
# we make both end conditions explicit
2570
# old is finished: insert current_new into the state.
2571
self.update_minimal(new_entry_key, current_new_minikind,
2572
executable=current_new[1].executable,
2573
path_utf8=new_path_utf8, fingerprint=fingerprint,
2575
current_new = advance(new_iterator)
2576
elif not current_new:
2578
self._make_absent(current_old)
2579
current_old = advance(old_iterator)
2580
elif new_entry_key == current_old[0]:
2581
# same - common case
2582
# We're looking at the same path and id in both the dirstate
2583
# and inventory, so just need to update the fields in the
2584
# dirstate from the one in the inventory.
2585
# TODO: update the record if anything significant has changed.
2586
# the minimal required trigger is if the execute bit or cached
2588
if (current_old[1][0][3] != current_new[1].executable or
2589
current_old[1][0][0] != current_new_minikind):
2590
self.update_minimal(current_old[0], current_new_minikind,
2591
executable=current_new[1].executable,
2592
path_utf8=new_path_utf8, fingerprint=fingerprint,
2594
# both sides are dealt with, move on
2595
current_old = advance(old_iterator)
2596
current_new = advance(new_iterator)
2597
elif (cmp_by_dirs(new_dirname, current_old[0][0]) < 0
2598
or (new_dirname == current_old[0][0]
2599
and new_entry_key[1:] < current_old[0][1:])):
2601
# add a entry for this and advance new
2602
self.update_minimal(new_entry_key, current_new_minikind,
2603
executable=current_new[1].executable,
2604
path_utf8=new_path_utf8, fingerprint=fingerprint,
2606
current_new = advance(new_iterator)
2608
# we've advanced past the place where the old key would be,
2609
# without seeing it in the new list. so it must be gone.
2610
self._make_absent(current_old)
2611
current_old = advance(old_iterator)
2612
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
2613
self._id_index = None
2614
self._packed_stat_index = None
2616
def _make_absent(self, current_old):
2617
"""Mark current_old - an entry - as absent for tree 0.
2619
:return: True if this was the last details entry for the entry key:
2620
that is, if the underlying block has had the entry removed, thus
2621
shrinking in length.
2623
# build up paths that this id will be left at after the change is made,
2624
# so we can update their cross references in tree 0
2625
all_remaining_keys = set()
2626
# Dont check the working tree, because it's going.
2627
for details in current_old[1][1:]:
2628
if details[0] not in 'ar': # absent, relocated
2629
all_remaining_keys.add(current_old[0])
2630
elif details[0] == 'r': # relocated
2631
# record the key for the real path.
2632
all_remaining_keys.add(tuple(osutils.split(details[1])) + (current_old[0][2],))
2633
# absent rows are not present at any path.
2634
last_reference = current_old[0] not in all_remaining_keys
2636
# the current row consists entire of the current item (being marked
2637
# absent), and relocated or absent entries for the other trees:
2638
# Remove it, its meaningless.
2639
block = self._find_block(current_old[0])
2640
entry_index, present = self._find_entry_index(current_old[0], block[1])
2642
raise AssertionError('could not find entry for %s' % (current_old,))
2643
block[1].pop(entry_index)
2644
# if we have an id_index in use, remove this key from it for this id.
2645
if self._id_index is not None:
2646
self._id_index[current_old[0][2]].remove(current_old[0])
2647
# update all remaining keys for this id to record it as absent. The
2648
# existing details may either be the record we are marking as deleted
2649
# (if there were other trees with the id present at this path), or may
2651
for update_key in all_remaining_keys:
2652
update_block_index, present = \
2653
self._find_block_index_from_key(update_key)
2655
raise AssertionError('could not find block for %s' % (update_key,))
2656
update_entry_index, present = \
2657
self._find_entry_index(update_key, self._dirblocks[update_block_index][1])
2659
raise AssertionError('could not find entry for %s' % (update_key,))
2660
update_tree_details = self._dirblocks[update_block_index][1][update_entry_index][1]
2661
# it must not be absent at the moment
2662
if update_tree_details[0][0] == 'a': # absent
2663
raise AssertionError('bad row %r' % (update_tree_details,))
2664
update_tree_details[0] = DirState.NULL_PARENT_DETAILS
2665
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
2666
return last_reference
2668
def update_minimal(self, key, minikind, executable=False, fingerprint='',
2669
packed_stat=None, size=0, path_utf8=None, fullscan=False):
2670
"""Update an entry to the state in tree 0.
2672
This will either create a new entry at 'key' or update an existing one.
2673
It also makes sure that any other records which might mention this are
2676
:param key: (dir, name, file_id) for the new entry
2677
:param minikind: The type for the entry ('f' == 'file', 'd' ==
2679
:param executable: Should the executable bit be set?
2680
:param fingerprint: Simple fingerprint for new entry: canonical-form
2681
sha1 for files, referenced revision id for subtrees, etc.
2682
:param packed_stat: Packed stat value for new entry.
2683
:param size: Size information for new entry
2684
:param path_utf8: key[0] + '/' + key[1], just passed in to avoid doing
2686
:param fullscan: If True then a complete scan of the dirstate is being
2687
done and checking for duplicate rows should not be done. This
2688
should only be set by set_state_from_inventory and similar methods.
2690
If packed_stat and fingerprint are not given, they're invalidated in
2693
block = self._find_block(key)[1]
2694
if packed_stat is None:
2695
packed_stat = DirState.NULLSTAT
2696
# XXX: Some callers pass '' as the packed_stat, and it seems to be
2697
# sometimes present in the dirstate - this seems oddly inconsistent.
2699
entry_index, present = self._find_entry_index(key, block)
2700
new_details = (minikind, fingerprint, size, executable, packed_stat)
2701
id_index = self._get_id_index()
2703
# New record. Check there isn't a entry at this path already.
2705
low_index, _ = self._find_entry_index(key[0:2] + ('',), block)
2706
while low_index < len(block):
2707
entry = block[low_index]
2708
if entry[0][0:2] == key[0:2]:
2709
if entry[1][0][0] not in 'ar':
2710
# This entry has the same path (but a different id) as
2711
# the new entry we're adding, and is present in ths
2713
raise errors.InconsistentDelta(
2714
("%s/%s" % key[0:2]).decode('utf8'), key[2],
2715
"Attempt to add item at path already occupied by "
2716
"id %r" % entry[0][2])
2719
# new entry, synthesis cross reference here,
2720
existing_keys = id_index.setdefault(key[2], set())
2721
if not existing_keys:
2722
# not currently in the state, simplest case
2723
new_entry = key, [new_details] + self._empty_parent_info()
2725
# present at one or more existing other paths.
2726
# grab one of them and use it to generate parent
2727
# relocation/absent entries.
2728
new_entry = key, [new_details]
2729
for other_key in existing_keys:
2730
# change the record at other to be a pointer to this new
2731
# record. The loop looks similar to the change to
2732
# relocations when updating an existing record but its not:
2733
# the test for existing kinds is different: this can be
2734
# factored out to a helper though.
2735
other_block_index, present = self._find_block_index_from_key(other_key)
2737
raise AssertionError('could not find block for %s' % (other_key,))
2738
other_entry_index, present = self._find_entry_index(other_key,
2739
self._dirblocks[other_block_index][1])
2741
raise AssertionError('could not find entry for %s' % (other_key,))
2742
if path_utf8 is None:
2743
raise AssertionError('no path')
2744
self._dirblocks[other_block_index][1][other_entry_index][1][0] = \
2745
('r', path_utf8, 0, False, '')
2747
num_present_parents = self._num_present_parents()
2748
for lookup_index in xrange(1, num_present_parents + 1):
2749
# grab any one entry, use it to find the right path.
2750
# TODO: optimise this to reduce memory use in highly
2751
# fragmented situations by reusing the relocation
2753
update_block_index, present = \
2754
self._find_block_index_from_key(other_key)
2756
raise AssertionError('could not find block for %s' % (other_key,))
2757
update_entry_index, present = \
2758
self._find_entry_index(other_key, self._dirblocks[update_block_index][1])
2760
raise AssertionError('could not find entry for %s' % (other_key,))
2761
update_details = self._dirblocks[update_block_index][1][update_entry_index][1][lookup_index]
2762
if update_details[0] in 'ar': # relocated, absent
2763
# its a pointer or absent in lookup_index's tree, use
2765
new_entry[1].append(update_details)
2767
# we have the right key, make a pointer to it.
2768
pointer_path = osutils.pathjoin(*other_key[0:2])
2769
new_entry[1].append(('r', pointer_path, 0, False, ''))
2770
block.insert(entry_index, new_entry)
2771
existing_keys.add(key)
2773
# Does the new state matter?
2774
block[entry_index][1][0] = new_details
2775
# parents cannot be affected by what we do.
2776
# other occurences of this id can be found
2777
# from the id index.
2779
# tree index consistency: All other paths for this id in this tree
2780
# index must point to the correct path. We have to loop here because
2781
# we may have passed entries in the state with this file id already
2782
# that were absent - where parent entries are - and they need to be
2783
# converted to relocated.
2784
if path_utf8 is None:
2785
raise AssertionError('no path')
2786
for entry_key in id_index.setdefault(key[2], set()):
2787
# TODO:PROFILING: It might be faster to just update
2788
# rather than checking if we need to, and then overwrite
2789
# the one we are located at.
2790
if entry_key != key:
2791
# this file id is at a different path in one of the
2792
# other trees, so put absent pointers there
2793
# This is the vertical axis in the matrix, all pointing
2795
block_index, present = self._find_block_index_from_key(entry_key)
2797
raise AssertionError('not present: %r', entry_key)
2798
entry_index, present = self._find_entry_index(entry_key, self._dirblocks[block_index][1])
2800
raise AssertionError('not present: %r', entry_key)
2801
self._dirblocks[block_index][1][entry_index][1][0] = \
2802
('r', path_utf8, 0, False, '')
2803
# add a containing dirblock if needed.
2804
if new_details[0] == 'd':
2805
subdir_key = (osutils.pathjoin(*key[0:2]), '', '')
2806
block_index, present = self._find_block_index_from_key(subdir_key)
2808
self._dirblocks.insert(block_index, (subdir_key[0], []))
2810
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
2812
def _validate(self):
2813
"""Check that invariants on the dirblock are correct.
2815
This can be useful in debugging; it shouldn't be necessary in
2818
This must be called with a lock held.
2820
# NOTE: This must always raise AssertionError not just assert,
2821
# otherwise it may not behave properly under python -O
2823
# TODO: All entries must have some content that's not 'a' or 'r',
2824
# otherwise it could just be removed.
2826
# TODO: All relocations must point directly to a real entry.
2828
# TODO: No repeated keys.
2831
from pprint import pformat
2832
self._read_dirblocks_if_needed()
2833
if len(self._dirblocks) > 0:
2834
if not self._dirblocks[0][0] == '':
2835
raise AssertionError(
2836
"dirblocks don't start with root block:\n" + \
2837
pformat(self._dirblocks))
2838
if len(self._dirblocks) > 1:
2839
if not self._dirblocks[1][0] == '':
2840
raise AssertionError(
2841
"dirblocks missing root directory:\n" + \
2842
pformat(self._dirblocks))
2843
# the dirblocks are sorted by their path components, name, and dir id
2844
dir_names = [d[0].split('/')
2845
for d in self._dirblocks[1:]]
2846
if dir_names != sorted(dir_names):
2847
raise AssertionError(
2848
"dir names are not in sorted order:\n" + \
2849
pformat(self._dirblocks) + \
2852
for dirblock in self._dirblocks:
2853
# within each dirblock, the entries are sorted by filename and
2855
for entry in dirblock[1]:
2856
if dirblock[0] != entry[0][0]:
2857
raise AssertionError(
2859
"doesn't match directory name in\n%r" %
2860
(entry, pformat(dirblock)))
2861
if dirblock[1] != sorted(dirblock[1]):
2862
raise AssertionError(
2863
"dirblock for %r is not sorted:\n%s" % \
2864
(dirblock[0], pformat(dirblock)))
2866
def check_valid_parent():
2867
"""Check that the current entry has a valid parent.
2869
This makes sure that the parent has a record,
2870
and that the parent isn't marked as "absent" in the
2871
current tree. (It is invalid to have a non-absent file in an absent
2874
if entry[0][0:2] == ('', ''):
2875
# There should be no parent for the root row
2877
parent_entry = self._get_entry(tree_index, path_utf8=entry[0][0])
2878
if parent_entry == (None, None):
2879
raise AssertionError(
2880
"no parent entry for: %s in tree %s"
2881
% (this_path, tree_index))
2882
if parent_entry[1][tree_index][0] != 'd':
2883
raise AssertionError(
2884
"Parent entry for %s is not marked as a valid"
2885
" directory. %s" % (this_path, parent_entry,))
2887
# For each file id, for each tree: either
2888
# the file id is not present at all; all rows with that id in the
2889
# key have it marked as 'absent'
2890
# OR the file id is present under exactly one name; any other entries
2891
# that mention that id point to the correct name.
2893
# We check this with a dict per tree pointing either to the present
2894
# name, or None if absent.
2895
tree_count = self._num_present_parents() + 1
2896
id_path_maps = [dict() for i in range(tree_count)]
2897
# Make sure that all renamed entries point to the correct location.
2898
for entry in self._iter_entries():
2899
file_id = entry[0][2]
2900
this_path = osutils.pathjoin(entry[0][0], entry[0][1])
2901
if len(entry[1]) != tree_count:
2902
raise AssertionError(
2903
"wrong number of entry details for row\n%s" \
2904
",\nexpected %d" % \
2905
(pformat(entry), tree_count))
2906
absent_positions = 0
2907
for tree_index, tree_state in enumerate(entry[1]):
2908
this_tree_map = id_path_maps[tree_index]
2909
minikind = tree_state[0]
2910
if minikind in 'ar':
2911
absent_positions += 1
2912
# have we seen this id before in this column?
2913
if file_id in this_tree_map:
2914
previous_path, previous_loc = this_tree_map[file_id]
2915
# any later mention of this file must be consistent with
2916
# what was said before
2918
if previous_path is not None:
2919
raise AssertionError(
2920
"file %s is absent in row %r but also present " \
2922
(file_id, entry, previous_path))
2923
elif minikind == 'r':
2924
target_location = tree_state[1]
2925
if previous_path != target_location:
2926
raise AssertionError(
2927
"file %s relocation in row %r but also at %r" \
2928
% (file_id, entry, previous_path))
2930
# a file, directory, etc - may have been previously
2931
# pointed to by a relocation, which must point here
2932
if previous_path != this_path:
2933
raise AssertionError(
2934
"entry %r inconsistent with previous path %r "
2936
(entry, previous_path, previous_loc))
2937
check_valid_parent()
2940
# absent; should not occur anywhere else
2941
this_tree_map[file_id] = None, this_path
2942
elif minikind == 'r':
2943
# relocation, must occur at expected location
2944
this_tree_map[file_id] = tree_state[1], this_path
2946
this_tree_map[file_id] = this_path, this_path
2947
check_valid_parent()
2948
if absent_positions == tree_count:
2949
raise AssertionError(
2950
"entry %r has no data for any tree." % (entry,))
2952
def _wipe_state(self):
2953
"""Forget all state information about the dirstate."""
2954
self._header_state = DirState.NOT_IN_MEMORY
2955
self._dirblock_state = DirState.NOT_IN_MEMORY
2956
self._changes_aborted = False
2959
self._dirblocks = []
2960
self._id_index = None
2961
self._packed_stat_index = None
2962
self._end_of_header = None
2963
self._cutoff_time = None
2964
self._split_path_cache = {}
2966
def lock_read(self):
2967
"""Acquire a read lock on the dirstate."""
2968
if self._lock_token is not None:
2969
raise errors.LockContention(self._lock_token)
2970
# TODO: jam 20070301 Rather than wiping completely, if the blocks are
2971
# already in memory, we could read just the header and check for
2972
# any modification. If not modified, we can just leave things
2974
self._lock_token = lock.ReadLock(self._filename)
2975
self._lock_state = 'r'
2976
self._state_file = self._lock_token.f
2979
def lock_write(self):
2980
"""Acquire a write lock on the dirstate."""
2981
if self._lock_token is not None:
2982
raise errors.LockContention(self._lock_token)
2983
# TODO: jam 20070301 Rather than wiping completely, if the blocks are
2984
# already in memory, we could read just the header and check for
2985
# any modification. If not modified, we can just leave things
2987
self._lock_token = lock.WriteLock(self._filename)
2988
self._lock_state = 'w'
2989
self._state_file = self._lock_token.f
2993
"""Drop any locks held on the dirstate."""
2994
if self._lock_token is None:
2995
raise errors.LockNotHeld(self)
2996
# TODO: jam 20070301 Rather than wiping completely, if the blocks are
2997
# already in memory, we could read just the header and check for
2998
# any modification. If not modified, we can just leave things
3000
self._state_file = None
3001
self._lock_state = None
3002
self._lock_token.unlock()
3003
self._lock_token = None
3004
self._split_path_cache = {}
3006
def _requires_lock(self):
3007
"""Check that a lock is currently held by someone on the dirstate."""
3008
if not self._lock_token:
3009
raise errors.ObjectNotLocked(self)
3012
def py_update_entry(state, entry, abspath, stat_value,
3013
_stat_to_minikind=DirState._stat_to_minikind,
3014
_pack_stat=pack_stat):
3015
"""Update the entry based on what is actually on disk.
3017
This function only calculates the sha if it needs to - if the entry is
3018
uncachable, or clearly different to the first parent's entry, no sha
3019
is calculated, and None is returned.
3021
:param state: The dirstate this entry is in.
3022
:param entry: This is the dirblock entry for the file in question.
3023
:param abspath: The path on disk for this file.
3024
:param stat_value: The stat value done on the path.
3025
:return: None, or The sha1 hexdigest of the file (40 bytes) or link
3026
target of a symlink.
3029
minikind = _stat_to_minikind[stat_value.st_mode & 0170000]
3033
packed_stat = _pack_stat(stat_value)
3034
(saved_minikind, saved_link_or_sha1, saved_file_size,
3035
saved_executable, saved_packed_stat) = entry[1][0]
3037
if minikind == 'd' and saved_minikind == 't':
3039
if (minikind == saved_minikind
3040
and packed_stat == saved_packed_stat):
3041
# The stat hasn't changed since we saved, so we can re-use the
3046
# size should also be in packed_stat
3047
if saved_file_size == stat_value.st_size:
3048
return saved_link_or_sha1
3050
# If we have gotten this far, that means that we need to actually
3051
# process this entry.
3054
executable = state._is_executable(stat_value.st_mode,
3056
if state._cutoff_time is None:
3057
state._sha_cutoff_time()
3058
if (stat_value.st_mtime < state._cutoff_time
3059
and stat_value.st_ctime < state._cutoff_time
3060
and len(entry[1]) > 1
3061
and entry[1][1][0] != 'a'):
3062
# Could check for size changes for further optimised
3063
# avoidance of sha1's. However the most prominent case of
3064
# over-shaing is during initial add, which this catches.
3065
# Besides, if content filtering happens, size and sha
3066
# are calculated at the same time, so checking just the size
3067
# gains nothing w.r.t. performance.
3068
link_or_sha1 = state._sha1_file(abspath)
3069
entry[1][0] = ('f', link_or_sha1, stat_value.st_size,
3070
executable, packed_stat)
3072
entry[1][0] = ('f', '', stat_value.st_size,
3073
executable, DirState.NULLSTAT)
3074
elif minikind == 'd':
3076
entry[1][0] = ('d', '', 0, False, packed_stat)
3077
if saved_minikind != 'd':
3078
# This changed from something into a directory. Make sure we
3079
# have a directory block for it. This doesn't happen very
3080
# often, so this doesn't have to be super fast.
3081
block_index, entry_index, dir_present, file_present = \
3082
state._get_block_entry_index(entry[0][0], entry[0][1], 0)
3083
state._ensure_block(block_index, entry_index,
3084
osutils.pathjoin(entry[0][0], entry[0][1]))
3085
elif minikind == 'l':
3086
link_or_sha1 = state._read_link(abspath, saved_link_or_sha1)
3087
if state._cutoff_time is None:
3088
state._sha_cutoff_time()
3089
if (stat_value.st_mtime < state._cutoff_time
3090
and stat_value.st_ctime < state._cutoff_time):
3091
entry[1][0] = ('l', link_or_sha1, stat_value.st_size,
3094
entry[1][0] = ('l', '', stat_value.st_size,
3095
False, DirState.NULLSTAT)
3096
state._dirblock_state = DirState.IN_MEMORY_MODIFIED
3100
class ProcessEntryPython(object):
3102
__slots__ = ["old_dirname_to_file_id", "new_dirname_to_file_id", "uninteresting",
3103
"last_source_parent", "last_target_parent", "include_unchanged",
3104
"use_filesystem_for_exec", "utf8_decode", "searched_specific_files",
3105
"search_specific_files", "state", "source_index", "target_index",
3106
"want_unversioned", "tree"]
3108
def __init__(self, include_unchanged, use_filesystem_for_exec,
3109
search_specific_files, state, source_index, target_index,
3110
want_unversioned, tree):
3111
self.old_dirname_to_file_id = {}
3112
self.new_dirname_to_file_id = {}
3113
# Just a sentry, so that _process_entry can say that this
3114
# record is handled, but isn't interesting to process (unchanged)
3115
self.uninteresting = object()
3116
# Using a list so that we can access the values and change them in
3117
# nested scope. Each one is [path, file_id, entry]
3118
self.last_source_parent = [None, None]
3119
self.last_target_parent = [None, None]
3120
self.include_unchanged = include_unchanged
3121
self.use_filesystem_for_exec = use_filesystem_for_exec
3122
self.utf8_decode = cache_utf8._utf8_decode
3123
# for all search_indexs in each path at or under each element of
3124
# search_specific_files, if the detail is relocated: add the id, and add the
3125
# relocated path as one to search if its not searched already. If the
3126
# detail is not relocated, add the id.
3127
self.searched_specific_files = set()
3128
self.search_specific_files = search_specific_files
3130
self.source_index = source_index
3131
self.target_index = target_index
3132
self.want_unversioned = want_unversioned
3135
def _process_entry(self, entry, path_info, pathjoin=osutils.pathjoin):
3136
"""Compare an entry and real disk to generate delta information.
3138
:param path_info: top_relpath, basename, kind, lstat, abspath for
3139
the path of entry. If None, then the path is considered absent.
3140
(Perhaps we should pass in a concrete entry for this ?)
3141
Basename is returned as a utf8 string because we expect this
3142
tuple will be ignored, and don't want to take the time to
3144
:return: None if these don't match
3145
A tuple of information about the change, or
3146
the object 'uninteresting' if these match, but are
3147
basically identical.
3149
if self.source_index is None:
3150
source_details = DirState.NULL_PARENT_DETAILS
3152
source_details = entry[1][self.source_index]
3153
target_details = entry[1][self.target_index]
3154
target_minikind = target_details[0]
3155
if path_info is not None and target_minikind in 'fdlt':
3156
if not (self.target_index == 0):
3157
raise AssertionError()
3158
link_or_sha1 = update_entry(self.state, entry,
3159
abspath=path_info[4], stat_value=path_info[3])
3160
# The entry may have been modified by update_entry
3161
target_details = entry[1][self.target_index]
3162
target_minikind = target_details[0]
3165
file_id = entry[0][2]
3166
source_minikind = source_details[0]
3167
if source_minikind in 'fdltr' and target_minikind in 'fdlt':
3168
# claimed content in both: diff
3169
# r | fdlt | | add source to search, add id path move and perform
3170
# | | | diff check on source-target
3171
# r | fdlt | a | dangling file that was present in the basis.
3173
if source_minikind in 'r':
3174
# add the source to the search path to find any children it
3175
# has. TODO ? : only add if it is a container ?
3176
if not osutils.is_inside_any(self.searched_specific_files,
3178
self.search_specific_files.add(source_details[1])
3179
# generate the old path; this is needed for stating later
3181
old_path = source_details[1]
3182
old_dirname, old_basename = os.path.split(old_path)
3183
path = pathjoin(entry[0][0], entry[0][1])
3184
old_entry = self.state._get_entry(self.source_index,
3186
# update the source details variable to be the real
3188
if old_entry == (None, None):
3189
raise errors.CorruptDirstate(self.state._filename,
3190
"entry '%s/%s' is considered renamed from %r"
3191
" but source does not exist\n"
3192
"entry: %s" % (entry[0][0], entry[0][1], old_path, entry))
3193
source_details = old_entry[1][self.source_index]
3194
source_minikind = source_details[0]
3196
old_dirname = entry[0][0]
3197
old_basename = entry[0][1]
3198
old_path = path = None
3199
if path_info is None:
3200
# the file is missing on disk, show as removed.
3201
content_change = True
3205
# source and target are both versioned and disk file is present.
3206
target_kind = path_info[2]
3207
if target_kind == 'directory':
3209
old_path = path = pathjoin(old_dirname, old_basename)
3210
self.new_dirname_to_file_id[path] = file_id
3211
if source_minikind != 'd':
3212
content_change = True
3214
# directories have no fingerprint
3215
content_change = False
3217
elif target_kind == 'file':
3218
if source_minikind != 'f':
3219
content_change = True
3221
# Check the sha. We can't just rely on the size as
3222
# content filtering may mean differ sizes actually
3223
# map to the same content
3224
if link_or_sha1 is None:
3226
statvalue, link_or_sha1 = \
3227
self.state._sha1_provider.stat_and_sha1(
3229
self.state._observed_sha1(entry, link_or_sha1,
3231
content_change = (link_or_sha1 != source_details[1])
3232
# Target details is updated at update_entry time
3233
if self.use_filesystem_for_exec:
3234
# We don't need S_ISREG here, because we are sure
3235
# we are dealing with a file.
3236
target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
3238
target_exec = target_details[3]
3239
elif target_kind == 'symlink':
3240
if source_minikind != 'l':
3241
content_change = True
3243
content_change = (link_or_sha1 != source_details[1])
3245
elif target_kind == 'tree-reference':
3246
if source_minikind != 't':
3247
content_change = True
3249
content_change = False
3252
raise Exception, "unknown kind %s" % path_info[2]
3253
if source_minikind == 'd':
3255
old_path = path = pathjoin(old_dirname, old_basename)
3256
self.old_dirname_to_file_id[old_path] = file_id
3257
# parent id is the entry for the path in the target tree
3258
if old_dirname == self.last_source_parent[0]:
3259
source_parent_id = self.last_source_parent[1]
3262
source_parent_id = self.old_dirname_to_file_id[old_dirname]
3264
source_parent_entry = self.state._get_entry(self.source_index,
3265
path_utf8=old_dirname)
3266
source_parent_id = source_parent_entry[0][2]
3267
if source_parent_id == entry[0][2]:
3268
# This is the root, so the parent is None
3269
source_parent_id = None
3271
self.last_source_parent[0] = old_dirname
3272
self.last_source_parent[1] = source_parent_id
3273
new_dirname = entry[0][0]
3274
if new_dirname == self.last_target_parent[0]:
3275
target_parent_id = self.last_target_parent[1]
3278
target_parent_id = self.new_dirname_to_file_id[new_dirname]
3280
# TODO: We don't always need to do the lookup, because the
3281
# parent entry will be the same as the source entry.
3282
target_parent_entry = self.state._get_entry(self.target_index,
3283
path_utf8=new_dirname)
3284
if target_parent_entry == (None, None):
3285
raise AssertionError(
3286
"Could not find target parent in wt: %s\nparent of: %s"
3287
% (new_dirname, entry))
3288
target_parent_id = target_parent_entry[0][2]
3289
if target_parent_id == entry[0][2]:
3290
# This is the root, so the parent is None
3291
target_parent_id = None
3293
self.last_target_parent[0] = new_dirname
3294
self.last_target_parent[1] = target_parent_id
3296
source_exec = source_details[3]
3297
if (self.include_unchanged
3299
or source_parent_id != target_parent_id
3300
or old_basename != entry[0][1]
3301
or source_exec != target_exec
3303
if old_path is None:
3304
old_path = path = pathjoin(old_dirname, old_basename)
3305
old_path_u = self.utf8_decode(old_path)[0]
3308
old_path_u = self.utf8_decode(old_path)[0]
3309
if old_path == path:
3312
path_u = self.utf8_decode(path)[0]
3313
source_kind = DirState._minikind_to_kind[source_minikind]
3314
return (entry[0][2],
3315
(old_path_u, path_u),
3318
(source_parent_id, target_parent_id),
3319
(self.utf8_decode(old_basename)[0], self.utf8_decode(entry[0][1])[0]),
3320
(source_kind, target_kind),
3321
(source_exec, target_exec))
3323
return self.uninteresting
3324
elif source_minikind in 'a' and target_minikind in 'fdlt':
3325
# looks like a new file
3326
path = pathjoin(entry[0][0], entry[0][1])
3327
# parent id is the entry for the path in the target tree
3328
# TODO: these are the same for an entire directory: cache em.
3329
parent_id = self.state._get_entry(self.target_index,
3330
path_utf8=entry[0][0])[0][2]
3331
if parent_id == entry[0][2]:
3333
if path_info is not None:
3335
if self.use_filesystem_for_exec:
3336
# We need S_ISREG here, because we aren't sure if this
3339
stat.S_ISREG(path_info[3].st_mode)
3340
and stat.S_IEXEC & path_info[3].st_mode)
3342
target_exec = target_details[3]
3343
return (entry[0][2],
3344
(None, self.utf8_decode(path)[0]),
3348
(None, self.utf8_decode(entry[0][1])[0]),
3349
(None, path_info[2]),
3350
(None, target_exec))
3352
# Its a missing file, report it as such.
3353
return (entry[0][2],
3354
(None, self.utf8_decode(path)[0]),
3358
(None, self.utf8_decode(entry[0][1])[0]),
3361
elif source_minikind in 'fdlt' and target_minikind in 'a':
3362
# unversioned, possibly, or possibly not deleted: we dont care.
3363
# if its still on disk, *and* theres no other entry at this
3364
# path [we dont know this in this routine at the moment -
3365
# perhaps we should change this - then it would be an unknown.
3366
old_path = pathjoin(entry[0][0], entry[0][1])
3367
# parent id is the entry for the path in the target tree
3368
parent_id = self.state._get_entry(self.source_index, path_utf8=entry[0][0])[0][2]
3369
if parent_id == entry[0][2]:
3371
return (entry[0][2],
3372
(self.utf8_decode(old_path)[0], None),
3376
(self.utf8_decode(entry[0][1])[0], None),
3377
(DirState._minikind_to_kind[source_minikind], None),
3378
(source_details[3], None))
3379
elif source_minikind in 'fdlt' and target_minikind in 'r':
3380
# a rename; could be a true rename, or a rename inherited from
3381
# a renamed parent. TODO: handle this efficiently. Its not
3382
# common case to rename dirs though, so a correct but slow
3383
# implementation will do.
3384
if not osutils.is_inside_any(self.searched_specific_files, target_details[1]):
3385
self.search_specific_files.add(target_details[1])
3386
elif source_minikind in 'ra' and target_minikind in 'ra':
3387
# neither of the selected trees contain this file,
3388
# so skip over it. This is not currently directly tested, but
3389
# is indirectly via test_too_much.TestCommands.test_conflicts.
3392
raise AssertionError("don't know how to compare "
3393
"source_minikind=%r, target_minikind=%r"
3394
% (source_minikind, target_minikind))
3395
## import pdb;pdb.set_trace()
3401
def iter_changes(self):
3402
"""Iterate over the changes."""
3403
utf8_decode = cache_utf8._utf8_decode
3404
_cmp_by_dirs = cmp_by_dirs
3405
_process_entry = self._process_entry
3406
uninteresting = self.uninteresting
3407
search_specific_files = self.search_specific_files
3408
searched_specific_files = self.searched_specific_files
3409
splitpath = osutils.splitpath
3411
# compare source_index and target_index at or under each element of search_specific_files.
3412
# follow the following comparison table. Note that we only want to do diff operations when
3413
# the target is fdl because thats when the walkdirs logic will have exposed the pathinfo
3417
# Source | Target | disk | action
3418
# r | fdlt | | add source to search, add id path move and perform
3419
# | | | diff check on source-target
3420
# r | fdlt | a | dangling file that was present in the basis.
3422
# r | a | | add source to search
3424
# r | r | | this path is present in a non-examined tree, skip.
3425
# r | r | a | this path is present in a non-examined tree, skip.
3426
# a | fdlt | | add new id
3427
# a | fdlt | a | dangling locally added file, skip
3428
# a | a | | not present in either tree, skip
3429
# a | a | a | not present in any tree, skip
3430
# a | r | | not present in either tree at this path, skip as it
3431
# | | | may not be selected by the users list of paths.
3432
# a | r | a | not present in either tree at this path, skip as it
3433
# | | | may not be selected by the users list of paths.
3434
# fdlt | fdlt | | content in both: diff them
3435
# fdlt | fdlt | a | deleted locally, but not unversioned - show as deleted ?
3436
# fdlt | a | | unversioned: output deleted id for now
3437
# fdlt | a | a | unversioned and deleted: output deleted id
3438
# fdlt | r | | relocated in this tree, so add target to search.
3439
# | | | Dont diff, we will see an r,fd; pair when we reach
3440
# | | | this id at the other path.
3441
# fdlt | r | a | relocated in this tree, so add target to search.
3442
# | | | Dont diff, we will see an r,fd; pair when we reach
3443
# | | | this id at the other path.
3445
# TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
3446
# keeping a cache of directories that we have seen.
3448
while search_specific_files:
3449
# TODO: the pending list should be lexically sorted? the
3450
# interface doesn't require it.
3451
current_root = search_specific_files.pop()
3452
current_root_unicode = current_root.decode('utf8')
3453
searched_specific_files.add(current_root)
3454
# process the entries for this containing directory: the rest will be
3455
# found by their parents recursively.
3456
root_entries = self.state._entries_for_path(current_root)
3457
root_abspath = self.tree.abspath(current_root_unicode)
3459
root_stat = os.lstat(root_abspath)
3461
if e.errno == errno.ENOENT:
3462
# the path does not exist: let _process_entry know that.
3463
root_dir_info = None
3465
# some other random error: hand it up.
3468
root_dir_info = ('', current_root,
3469
osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
3471
if root_dir_info[2] == 'directory':
3472
if self.tree._directory_is_tree_reference(
3473
current_root.decode('utf8')):
3474
root_dir_info = root_dir_info[:2] + \
3475
('tree-reference',) + root_dir_info[3:]
3477
if not root_entries and not root_dir_info:
3478
# this specified path is not present at all, skip it.
3480
path_handled = False
3481
for entry in root_entries:
3482
result = _process_entry(entry, root_dir_info)
3483
if result is not None:
3485
if result is not uninteresting:
3487
if self.want_unversioned and not path_handled and root_dir_info:
3488
new_executable = bool(
3489
stat.S_ISREG(root_dir_info[3].st_mode)
3490
and stat.S_IEXEC & root_dir_info[3].st_mode)
3492
(None, current_root_unicode),
3496
(None, splitpath(current_root_unicode)[-1]),
3497
(None, root_dir_info[2]),
3498
(None, new_executable)
3500
initial_key = (current_root, '', '')
3501
block_index, _ = self.state._find_block_index_from_key(initial_key)
3502
if block_index == 0:
3503
# we have processed the total root already, but because the
3504
# initial key matched it we should skip it here.
3506
if root_dir_info and root_dir_info[2] == 'tree-reference':
3507
current_dir_info = None
3509
dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
3511
current_dir_info = dir_iterator.next()
3513
# on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
3514
# python 2.5 has e.errno == EINVAL,
3515
# and e.winerror == ERROR_DIRECTORY
3516
e_winerror = getattr(e, 'winerror', None)
3517
win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
3518
# there may be directories in the inventory even though
3519
# this path is not a file on disk: so mark it as end of
3521
if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
3522
current_dir_info = None
3523
elif (sys.platform == 'win32'
3524
and (e.errno in win_errors
3525
or e_winerror in win_errors)):
3526
current_dir_info = None
3530
if current_dir_info[0][0] == '':
3531
# remove .bzr from iteration
3532
bzr_index = bisect.bisect_left(current_dir_info[1], ('.bzr',))
3533
if current_dir_info[1][bzr_index][0] != '.bzr':
3534
raise AssertionError()
3535
del current_dir_info[1][bzr_index]
3536
# walk until both the directory listing and the versioned metadata
3538
if (block_index < len(self.state._dirblocks) and
3539
osutils.is_inside(current_root, self.state._dirblocks[block_index][0])):
3540
current_block = self.state._dirblocks[block_index]
3542
current_block = None
3543
while (current_dir_info is not None or
3544
current_block is not None):
3545
if (current_dir_info and current_block
3546
and current_dir_info[0][0] != current_block[0]):
3547
if _cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
3548
# filesystem data refers to paths not covered by the dirblock.
3549
# this has two possibilities:
3550
# A) it is versioned but empty, so there is no block for it
3551
# B) it is not versioned.
3553
# if (A) then we need to recurse into it to check for
3554
# new unknown files or directories.
3555
# if (B) then we should ignore it, because we don't
3556
# recurse into unknown directories.
3558
while path_index < len(current_dir_info[1]):
3559
current_path_info = current_dir_info[1][path_index]
3560
if self.want_unversioned:
3561
if current_path_info[2] == 'directory':
3562
if self.tree._directory_is_tree_reference(
3563
current_path_info[0].decode('utf8')):
3564
current_path_info = current_path_info[:2] + \
3565
('tree-reference',) + current_path_info[3:]
3566
new_executable = bool(
3567
stat.S_ISREG(current_path_info[3].st_mode)
3568
and stat.S_IEXEC & current_path_info[3].st_mode)
3570
(None, utf8_decode(current_path_info[0])[0]),
3574
(None, utf8_decode(current_path_info[1])[0]),
3575
(None, current_path_info[2]),
3576
(None, new_executable))
3577
# dont descend into this unversioned path if it is
3579
if current_path_info[2] in ('directory',
3581
del current_dir_info[1][path_index]
3585
# This dir info has been handled, go to the next
3587
current_dir_info = dir_iterator.next()
3588
except StopIteration:
3589
current_dir_info = None
3591
# We have a dirblock entry for this location, but there
3592
# is no filesystem path for this. This is most likely
3593
# because a directory was removed from the disk.
3594
# We don't have to report the missing directory,
3595
# because that should have already been handled, but we
3596
# need to handle all of the files that are contained
3598
for current_entry in current_block[1]:
3599
# entry referring to file not present on disk.
3600
# advance the entry only, after processing.
3601
result = _process_entry(current_entry, None)
3602
if result is not None:
3603
if result is not uninteresting:
3606
if (block_index < len(self.state._dirblocks) and
3607
osutils.is_inside(current_root,
3608
self.state._dirblocks[block_index][0])):
3609
current_block = self.state._dirblocks[block_index]
3611
current_block = None
3614
if current_block and entry_index < len(current_block[1]):
3615
current_entry = current_block[1][entry_index]
3617
current_entry = None
3618
advance_entry = True
3620
if current_dir_info and path_index < len(current_dir_info[1]):
3621
current_path_info = current_dir_info[1][path_index]
3622
if current_path_info[2] == 'directory':
3623
if self.tree._directory_is_tree_reference(
3624
current_path_info[0].decode('utf8')):
3625
current_path_info = current_path_info[:2] + \
3626
('tree-reference',) + current_path_info[3:]
3628
current_path_info = None
3630
path_handled = False
3631
while (current_entry is not None or
3632
current_path_info is not None):
3633
if current_entry is None:
3634
# the check for path_handled when the path is advanced
3635
# will yield this path if needed.
3637
elif current_path_info is None:
3638
# no path is fine: the per entry code will handle it.
3639
result = _process_entry(current_entry, current_path_info)
3640
if result is not None:
3641
if result is not uninteresting:
3643
elif (current_entry[0][1] != current_path_info[1]
3644
or current_entry[1][self.target_index][0] in 'ar'):
3645
# The current path on disk doesn't match the dirblock
3646
# record. Either the dirblock is marked as absent, or
3647
# the file on disk is not present at all in the
3648
# dirblock. Either way, report about the dirblock
3649
# entry, and let other code handle the filesystem one.
3651
# Compare the basename for these files to determine
3653
if current_path_info[1] < current_entry[0][1]:
3654
# extra file on disk: pass for now, but only
3655
# increment the path, not the entry
3656
advance_entry = False
3658
# entry referring to file not present on disk.
3659
# advance the entry only, after processing.
3660
result = _process_entry(current_entry, None)
3661
if result is not None:
3662
if result is not uninteresting:
3664
advance_path = False
3666
result = _process_entry(current_entry, current_path_info)
3667
if result is not None:
3669
if result is not uninteresting:
3671
if advance_entry and current_entry is not None:
3673
if entry_index < len(current_block[1]):
3674
current_entry = current_block[1][entry_index]
3676
current_entry = None
3678
advance_entry = True # reset the advance flaga
3679
if advance_path and current_path_info is not None:
3680
if not path_handled:
3681
# unversioned in all regards
3682
if self.want_unversioned:
3683
new_executable = bool(
3684
stat.S_ISREG(current_path_info[3].st_mode)
3685
and stat.S_IEXEC & current_path_info[3].st_mode)
3687
relpath_unicode = utf8_decode(current_path_info[0])[0]
3688
except UnicodeDecodeError:
3689
raise errors.BadFilenameEncoding(
3690
current_path_info[0], osutils._fs_enc)
3692
(None, relpath_unicode),
3696
(None, utf8_decode(current_path_info[1])[0]),
3697
(None, current_path_info[2]),
3698
(None, new_executable))
3699
# dont descend into this unversioned path if it is
3701
if current_path_info[2] in ('directory'):
3702
del current_dir_info[1][path_index]
3704
# dont descend the disk iterator into any tree
3706
if current_path_info[2] == 'tree-reference':
3707
del current_dir_info[1][path_index]
3710
if path_index < len(current_dir_info[1]):
3711
current_path_info = current_dir_info[1][path_index]
3712
if current_path_info[2] == 'directory':
3713
if self.tree._directory_is_tree_reference(
3714
current_path_info[0].decode('utf8')):
3715
current_path_info = current_path_info[:2] + \
3716
('tree-reference',) + current_path_info[3:]
3718
current_path_info = None
3719
path_handled = False
3721
advance_path = True # reset the advance flagg.
3722
if current_block is not None:
3724
if (block_index < len(self.state._dirblocks) and
3725
osutils.is_inside(current_root, self.state._dirblocks[block_index][0])):
3726
current_block = self.state._dirblocks[block_index]
3728
current_block = None
3729
if current_dir_info is not None:
3731
current_dir_info = dir_iterator.next()
3732
except StopIteration:
3733
current_dir_info = None
3736
# Try to load the compiled form if possible
3738
from bzrlib._dirstate_helpers_pyx import (
3744
ProcessEntryC as _process_entry,
3745
update_entry as update_entry,
3748
from bzrlib._dirstate_helpers_py import (
3755
# FIXME: It would be nice to be able to track moved lines so that the
3756
# corresponding python code can be moved to the _dirstate_helpers_py
3757
# module. I don't want to break the history for this important piece of
3758
# code so I left the code here -- vila 20090622
3759
update_entry = py_update_entry
3760
_process_entry = ProcessEntryPython
added
removed
bzrlib/dirstate.py
