1
# Copyright (C) 2006, 2007 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
"""DirState objects record the state of a directory and its bzr metadata.
19
Pseudo EBNF grammar for the state file. Fields are separated by NULLs, and
20
lines by NL. The field delimiters are ommitted in the grammar, line delimiters
21
are not - this is done for clarity of reading. All string data is in utf8.
23
MINIKIND = "f" | "d" | "l" | "a" | "r";
26
WHOLE_NUMBER = {digit}, digit;
28
REVISION_ID = a non-empty utf8 string;
30
dirstate format = header line, full checksum, row count, parent details,
31
ghost_details, entries;
32
header line = "#bazaar dirstate flat format 2", NL;
33
full checksum = "adler32: ", ["-"], WHOLE_NUMBER, NL;
34
row count = "num_entries: ", digit, NL;
35
parent_details = WHOLE NUMBER, {REVISION_ID}* NL;
36
ghost_details = WHOLE NUMBER, {REVISION_ID}*, NL;
38
entry = entry_key, current_entry_details, {parent_entry_details};
39
entry_key = dirname, basename, fileid;
40
current_entry_details = common_entry_details, working_entry_details;
41
parent_entry_details = common_entry_details, history_entry_details;
42
common_entry_details = MINIKIND, fingerprint, size, executable
43
working_entry_details = packed_stat
44
history_entry_details = REVISION_ID;
47
fingerprint = a nonempty utf8 sequence with meaning defined by minikind.
49
Given this definition, the following is useful to know:
50
entry (aka row) - all the data for a given key.
51
entry[0]: The key (dirname, basename, fileid)
55
entry[1]: The tree(s) data for this path and id combination.
56
entry[1][0]: The current tree
57
entry[1][1]: The second tree
59
For an entry for a tree, we have (using tree 0 - current tree) to demonstrate:
60
entry[1][0][0]: minikind
61
entry[1][0][1]: fingerprint
63
entry[1][0][3]: executable
64
entry[1][0][4]: packed_stat
66
entry[1][1][4]: revision_id
68
There may be multiple rows at the root, one per id present in the root, so the
69
in memory root row is now:
70
self._dirblocks[0] -> ('', [entry ...]),
71
and the entries in there are
74
entries[0][2]: file_id
75
entries[1][0]: The tree data for the current tree for this fileid at /
79
'r' is a relocated entry: This path is not present in this tree with this id,
80
but the id can be found at another location. The fingerprint is used to
81
point to the target location.
82
'a' is an absent entry: In that tree the id is not present at this path.
83
'd' is a directory entry: This path in this tree is a directory with the
84
current file id. There is no fingerprint for directories.
85
'f' is a file entry: As for directory, but its a file. The fingerprint is a
87
'l' is a symlink entry: As for directory, but a symlink. The fingerprint is the
91
--- Format 1 had the following different definition: ---
92
rows = dirname, NULL, basename, NULL, MINIKIND, NULL, fileid_utf8, NULL,
93
WHOLE NUMBER (* size *), NULL, packed stat, NULL, sha1|symlink target,
95
PARENT ROW = NULL, revision_utf8, NULL, MINIKIND, NULL, dirname, NULL,
96
basename, NULL, WHOLE NUMBER (* size *), NULL, "y" | "n", NULL,
99
PARENT ROW's are emitted for every parent that is not in the ghosts details
100
line. That is, if the parents are foo, bar, baz, and the ghosts are bar, then
101
each row will have a PARENT ROW for foo and baz, but not for bar.
104
In any tree, a kind of 'moved' indicates that the fingerprint field
105
(which we treat as opaque data specific to the 'kind' anyway) has the
106
details for the id of this row in that tree.
108
I'm strongly tempted to add a id->path index as well, but I think that
109
where we need id->path mapping; we also usually read the whole file, so
110
I'm going to skip that for the moment, as we have the ability to locate
111
via bisect any path in any tree, and if we lookup things by path, we can
112
accumulate a id->path mapping as we go, which will tend to match what we
115
I plan to implement this asap, so please speak up now to alter/tweak the
116
design - and once we stabilise on this, I'll update the wiki page for
119
The rationale for all this is that we want fast operations for the
120
common case (diff/status/commit/merge on all files) and extremely fast
121
operations for the less common but still occurs a lot status/diff/commit
122
on specific files). Operations on specific files involve a scan for all
123
the children of a path, *in every involved tree*, which the current
124
format did not accommodate.
128
1) Fast end to end use for bzr's top 5 uses cases. (commmit/diff/status/merge/???)
129
2) fall back current object model as needed.
130
3) scale usably to the largest trees known today - say 50K entries. (mozilla
131
is an example of this)
135
Eventually reuse dirstate objects across locks IFF the dirstate file has not
136
been modified, but will require that we flush/ignore cached stat-hit data
137
because we wont want to restat all files on disk just because a lock was
138
acquired, yet we cannot trust the data after the previous lock was released.
140
Memory representation:
141
vector of all directories, and vector of the childen ?
143
root_entrie = (direntry for root, [parent_direntries_for_root]),
145
('', ['data for achild', 'data for bchild', 'data for cchild'])
146
('dir', ['achild', 'cchild', 'echild'])
148
- single bisect to find N subtrees from a path spec
149
- in-order for serialisation - this is 'dirblock' grouping.
150
- insertion of a file '/a' affects only the '/' child-vector, that is, to
151
insert 10K elements from scratch does not generates O(N^2) memoves of a
152
single vector, rather each individual, which tends to be limited to a
153
manageable number. Will scale badly on trees with 10K entries in a
154
single directory. compare with Inventory.InventoryDirectory which has
155
a dictionary for the children. No bisect capability, can only probe for
156
exact matches, or grab all elements and sorta.
157
- Whats the risk of error here? Once we have the base format being processed
158
we should have a net win regardless of optimality. So we are going to
159
go with what seems reasonably.
162
maybe we should do a test profile of these core structure - 10K simulated searches/lookups/etc?
164
Objects for each row?
165
The lifetime of Dirstate objects is current per lock, but see above for
166
possible extensions. The lifetime of a row from a dirstate is expected to be
167
very short in the optimistic case: which we are optimising for. For instance,
168
subtree status will determine from analysis of the disk data what rows need to
169
be examined at all, and will be able to determine from a single row whether
170
that file has altered or not, so we are aiming to process tens of thousands of
171
entries each second within the dirstate context, before exposing anything to
172
the larger codebase. This suggests we want the time for a single file
173
comparison to be < 0.1 milliseconds. That would give us 10000 paths per second
174
processed, and to scale to 100 thousand we'll another order of magnitude to do
175
that. Now, as the lifetime for all unchanged entries is the time to parse, stat
176
the file on disk, and then immediately discard, the overhead of object creation
177
becomes a significant cost.
179
Figures: Creating a tuple from from 3 elements was profiled at 0.0625
180
microseconds, whereas creating a object which is subclassed from tuple was
181
0.500 microseconds, and creating an object with 3 elements and slots was 3
182
microseconds long. 0.1 milliseconds is 100 microseconds, and ideally we'll get
183
down to 10 microseconds for the total processing - having 33% of that be object
184
creation is a huge overhead. There is a potential cost in using tuples within
185
each row which is that the conditional code to do comparisons may be slower
186
than method invocation, but method invocation is known to be slow due to stack
187
frame creation, so avoiding methods in these tight inner loops in unfortunately
188
desirable. We can consider a pyrex version of this with objects in future if
207
import bzrlib.inventory
208
from bzrlib import osutils
209
from bzrlib.osutils import (
217
class _Bisector(object):
218
"""This just keeps track of information as we are bisecting."""
221
class DirState(object):
222
"""Record directory and metadata state for fast access.
224
A dirstate is a specialised data structure for managing local working
225
tree state information. Its not yet well defined whether it is platform
226
specific, and if it is how we detect/parameterise that.
229
_kind_to_minikind = {'absent':'a', 'file':'f', 'directory':'d', 'relocated':'r', 'symlink':'l'}
230
_minikind_to_kind = {'a':'absent', 'f':'file', 'd':'directory', 'l':'symlink', 'r':'relocated'}
231
_to_yesno = {True:'y', False: 'n'} # TODO profile the performance gain
232
# of using int conversion rather than a dict here. AND BLAME ANDREW IF
235
# TODO: jam 20070221 Make sure we handle when there are duplicated records
236
# (like when we remove + add the same path, or we have a rename)
237
# TODO: jam 20070221 Figure out what to do if we have a record that exceeds
238
# the BISECT_PAGE_SIZE. For now, we just have to make it large enough
239
# that we are sure a single record will always fit.
240
BISECT_PAGE_SIZE = 4096
243
IN_MEMORY_UNMODIFIED = 1
244
IN_MEMORY_MODIFIED = 2
246
# A pack_stat (the x's) that is just noise and will never match the output
249
NULL_PARENT_DETAILS = ('a', '', 0, False, '')
251
def __init__(self, path):
252
"""Create a DirState object.
256
:attr _root_entrie: The root row of the directory/file information,
257
- contains the path to / - '', ''
258
- kind of 'directory',
259
- the file id of the root in utf8
262
- and no sha information.
263
:param path: The path at which the dirstate file on disk should live.
265
# _header_state and _dirblock_state represent the current state
266
# of the dirstate metadata and the per-row data respectiely.
267
# NOT_IN_MEMORY indicates that no data is in memory
268
# IN_MEMORY_UNMODIFIED indicates that what we have in memory
269
# is the same as is on disk
270
# IN_MEMORY_MODIFIED indicates that we have a modified version
271
# of what is on disk.
272
# In future we will add more granularity, for instance _dirblock_state
273
# will probably support partially-in-memory as a separate variable,
274
# allowing for partially-in-memory unmodified and partially-in-memory
276
self._header_state = DirState.NOT_IN_MEMORY
277
self._dirblock_state = DirState.NOT_IN_MEMORY
281
self._state_file = None
282
self._filename = path
283
self._lock_token = None
284
self._id_index = None
285
self._end_of_header = None
286
self._bisect_page_size = DirState.BISECT_PAGE_SIZE
288
def add(self, path, file_id, kind, stat, link_or_sha1):
289
"""Add a path to be tracked.
291
:param path: The path within the dirstate - '' is the root, 'foo' is the
292
path foo within the root, 'foo/bar' is the path bar within foo
294
:param file_id: The file id of the path being added.
295
:param kind: The kind of the path.
296
:param stat: The output of os.lstate for the path.
297
:param link_or_sha1: The sha value of the file, or the target of a
298
symlink. '' for directories.
301
# find the block its in.
302
# find the location in the block.
303
# check its not there
305
#------- copied from bzrlib.inventory.make_entry
306
# --- normalized_filename wants a unicode basename only, so get one.
307
dirname, basename = osutils.split(path)
308
# we dont import normalized_filename directly because we want to be
309
# able to change the implementation at runtime for tests.
310
norm_name, can_access = osutils.normalized_filename(basename)
311
if norm_name != basename:
315
raise errors.InvalidNormalization(path)
316
# now that we've normalised, we need the correct utf8 path and
317
# dirname and basename elements. This single encode and split should be
318
# faster than three separate encodes.
319
utf8path = (dirname + '/' + basename).strip('/').encode('utf8')
320
dirname, basename = osutils.split(utf8path)
321
assert file_id.__class__ == str, \
322
"must be a utf8 file_id not %s" % (type(file_id))
323
entry_key = (dirname, basename, file_id)
324
self._read_dirblocks_if_needed()
325
block_index, present = self._find_block_index_from_key(entry_key)
327
# The block where we want to put the file is not present. But it
328
# might be because the directory was empty, or not loaded yet. Look
329
# for a parent entry, if not found, raise NotVersionedError
330
parent_dir, parent_base = osutils.split(dirname)
331
parent_block_idx, parent_entry_idx, _, parent_present = \
332
self._get_block_entry_index(parent_dir, parent_base, 0)
333
if not parent_present:
334
raise errors.NotVersionedError(path, str(self))
335
self._ensure_block(parent_block_idx, parent_entry_idx, dirname)
336
block = self._dirblocks[block_index][1]
339
packed_stat = DirState.NULLSTAT
342
packed_stat = pack_stat(stat)
343
parent_info = self._empty_parent_info()
344
minikind = DirState._kind_to_minikind[kind]
346
entry_data = entry_key, [
347
(minikind, link_or_sha1, size, False, packed_stat),
349
elif kind == 'directory':
350
entry_data = entry_key, [
351
(minikind, '', 0, False, packed_stat),
353
elif kind == 'symlink':
354
entry_data = entry_key, [
355
(minikind, link_or_sha1, size, False, packed_stat),
358
raise errors.BzrError('unknown kind %r' % kind)
359
entry_index, present = self._find_entry_index(entry_key, block)
360
assert not present, "basename %r already added" % basename
361
block.insert(entry_index, entry_data)
363
if kind == 'directory':
364
# insert a new dirblock
365
self._ensure_block(block_index, entry_index, utf8path)
366
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
368
self._id_index.setdefault(entry_key[2], set()).add(entry_key)
370
def _bisect(self, dir_name_list):
371
"""Bisect through the disk structure for specific rows.
373
:param dir_name_list: A list of (dir, name) pairs.
374
:return: A dict mapping (dir, name) => entry for found entries. Missing
375
entries will not be in the map.
377
self._requires_lock()
378
# We need the file pointer to be right after the initial header block
379
self._read_header_if_needed()
380
# If _dirblock_state was in memory, we should just return info from
381
# there, this function is only meant to handle when we want to read
383
assert self._dirblock_state == DirState.NOT_IN_MEMORY
385
# The disk representation is generally info + '\0\n\0' at the end. But
386
# for bisecting, it is easier to treat this as '\0' + info + '\0\n'
387
# Because it means we can sync on the '\n'
388
state_file = self._state_file
389
file_size = os.fstat(state_file.fileno()).st_size
390
# We end up with 2 extra fields, we should have a trailing '\n' to
391
# ensure that we read the whole record, and we should have a precursur
392
# '' which ensures that we start after the previous '\n'
393
entry_field_count = self._fields_per_entry() + 1
395
low = self._end_of_header
396
high = file_size - 1 # Ignore the final '\0'
397
# Map from (dir, name) => entry
400
# Avoid infinite seeking
401
max_count = 30*len(dir_name_list)
403
# pending is a list of places to look.
404
# each entry is a tuple of low, high, dir_names
405
# low -> the first byte offset to read (inclusive)
406
# high -> the last byte offset (inclusive)
407
# dir_names -> The list of (dir, name) pairs that should be found in
408
# the [low, high] range
409
pending = [(low, high, dir_name_list)]
411
page_size = self._bisect_page_size
413
fields_to_entry = self._get_fields_to_entry()
416
low, high, cur_files = pending.pop()
418
if not cur_files or low >= high:
423
if count > max_count:
424
raise errors.BzrError('Too many seeks, most likely a bug.')
426
mid = max(low, (low+high-page_size)/2)
429
# limit the read size, so we don't end up reading data that we have
431
read_size = min(page_size, (high-mid)+1)
432
block = state_file.read(read_size)
435
entries = block.split('\n')
438
# We didn't find a '\n', so we cannot have found any records.
439
# So put this range back and try again. But we know we have to
440
# increase the page size, because a single read did not contain
441
# a record break (so records must be larger than page_size)
443
pending.append((low, high, cur_files))
446
# Check the first and last entries, in case they are partial, or if
447
# we don't care about the rest of this page
449
first_fields = entries[0].split('\0')
450
if len(first_fields) < entry_field_count:
451
# We didn't get the complete first entry
452
# so move start, and grab the next, which
453
# should be a full entry
454
start += len(entries[0])+1
455
first_fields = entries[1].split('\0')
458
if len(first_fields) <= 2:
459
# We didn't even get a filename here... what do we do?
460
# Try a large page size and repeat this query
462
pending.append((low, high, cur_files))
465
# Find what entries we are looking for, which occur before and
466
# after this first record.
468
first_dir_name = (first_fields[1], first_fields[2])
469
first_loc = bisect.bisect_left(cur_files, first_dir_name)
471
# These exist before the current location
472
pre = cur_files[:first_loc]
473
# These occur after the current location, which may be in the
474
# data we read, or might be after the last entry
475
post = cur_files[first_loc:]
477
if post and len(first_fields) >= entry_field_count:
478
# We have files after the first entry
480
# Parse the last entry
481
last_entry_num = len(entries)-1
482
last_fields = entries[last_entry_num].split('\0')
483
if len(last_fields) < entry_field_count:
484
# The very last hunk was not complete,
485
# read the previous hunk
486
after = mid + len(block) - len(entries[-1])
488
last_fields = entries[last_entry_num].split('\0')
490
after = mid + len(block)
492
last_dir_name = (last_fields[1], last_fields[2])
493
last_loc = bisect.bisect_right(post, last_dir_name)
495
middle_files = post[:last_loc]
496
post = post[last_loc:]
499
# We have files that should occur in this block
500
# (>= first, <= last)
501
# Either we will find them here, or we can mark them as
504
if middle_files[0] == first_dir_name:
505
# We might need to go before this location
506
pre.append(first_dir_name)
507
if middle_files[-1] == last_dir_name:
508
post.insert(0, last_dir_name)
510
# Find out what paths we have
511
paths = {first_dir_name:[first_fields]}
512
# last_dir_name might == first_dir_name so we need to be
513
# careful if we should append rather than overwrite
514
if last_entry_num != first_entry_num:
515
paths.setdefault(last_dir_name, []).append(last_fields)
516
for num in xrange(first_entry_num+1, last_entry_num):
517
# TODO: jam 20070223 We are already splitting here, so
518
# shouldn't we just split the whole thing rather
519
# than doing the split again in add_one_record?
520
fields = entries[num].split('\0')
521
dir_name = (fields[1], fields[2])
522
paths.setdefault(dir_name, []).append(fields)
524
for dir_name in middle_files:
525
for fields in paths.get(dir_name, []):
526
# offset by 1 because of the opening '\0'
527
# consider changing fields_to_entry to avoid the
529
entry = fields_to_entry(fields[1:])
530
found.setdefault(dir_name, []).append(entry)
532
# Now we have split up everything into pre, middle, and post, and
533
# we have handled everything that fell in 'middle'.
534
# We add 'post' first, so that we prefer to seek towards the
535
# beginning, so that we will tend to go as early as we need, and
536
# then only seek forward after that.
538
pending.append((after, high, post))
540
pending.append((low, start-1, pre))
542
# Consider that we may want to return the directory entries in sorted
543
# order. For now, we just return them in whatever order we found them,
544
# and leave it up to the caller if they care if it is ordered or not.
547
def _bisect_dirblocks(self, dir_list):
548
"""Bisect through the disk structure to find entries in given dirs.
550
_bisect_dirblocks is meant to find the contents of directories, which
551
differs from _bisect, which only finds individual entries.
553
:param dir_list: An sorted list of directory names ['', 'dir', 'foo'].
554
:return: A map from dir => entries_for_dir
556
# TODO: jam 20070223 A lot of the bisecting logic could be shared
557
# between this and _bisect. It would require parameterizing the
558
# inner loop with a function, though. We should evaluate the
559
# performance difference.
560
self._requires_lock()
561
# We need the file pointer to be right after the initial header block
562
self._read_header_if_needed()
563
# If _dirblock_state was in memory, we should just return info from
564
# there, this function is only meant to handle when we want to read
566
assert self._dirblock_state == DirState.NOT_IN_MEMORY
568
# The disk representation is generally info + '\0\n\0' at the end. But
569
# for bisecting, it is easier to treat this as '\0' + info + '\0\n'
570
# Because it means we can sync on the '\n'
571
state_file = self._state_file
572
file_size = os.fstat(state_file.fileno()).st_size
573
# We end up with 2 extra fields, we should have a trailing '\n' to
574
# ensure that we read the whole record, and we should have a precursur
575
# '' which ensures that we start after the previous '\n'
576
entry_field_count = self._fields_per_entry() + 1
578
low = self._end_of_header
579
high = file_size - 1 # Ignore the final '\0'
580
# Map from dir => entry
583
# Avoid infinite seeking
584
max_count = 30*len(dir_list)
586
# pending is a list of places to look.
587
# each entry is a tuple of low, high, dir_names
588
# low -> the first byte offset to read (inclusive)
589
# high -> the last byte offset (inclusive)
590
# dirs -> The list of directories that should be found in
591
# the [low, high] range
592
pending = [(low, high, dir_list)]
594
page_size = self._bisect_page_size
596
fields_to_entry = self._get_fields_to_entry()
599
low, high, cur_dirs = pending.pop()
601
if not cur_dirs or low >= high:
606
if count > max_count:
607
raise errors.BzrError('Too many seeks, most likely a bug.')
609
mid = max(low, (low+high-page_size)/2)
612
# limit the read size, so we don't end up reading data that we have
614
read_size = min(page_size, (high-mid)+1)
615
block = state_file.read(read_size)
618
entries = block.split('\n')
621
# We didn't find a '\n', so we cannot have found any records.
622
# So put this range back and try again. But we know we have to
623
# increase the page size, because a single read did not contain
624
# a record break (so records must be larger than page_size)
626
pending.append((low, high, cur_dirs))
629
# Check the first and last entries, in case they are partial, or if
630
# we don't care about the rest of this page
632
first_fields = entries[0].split('\0')
633
if len(first_fields) < entry_field_count:
634
# We didn't get the complete first entry
635
# so move start, and grab the next, which
636
# should be a full entry
637
start += len(entries[0])+1
638
first_fields = entries[1].split('\0')
641
if len(first_fields) <= 1:
642
# We didn't even get a dirname here... what do we do?
643
# Try a large page size and repeat this query
645
pending.append((low, high, cur_dirs))
648
# Find what entries we are looking for, which occur before and
649
# after this first record.
651
first_dir = first_fields[1]
652
first_loc = bisect.bisect_left(cur_dirs, first_dir)
654
# These exist before the current location
655
pre = cur_dirs[:first_loc]
656
# These occur after the current location, which may be in the
657
# data we read, or might be after the last entry
658
post = cur_dirs[first_loc:]
660
if post and len(first_fields) >= entry_field_count:
661
# We have records to look at after the first entry
663
# Parse the last entry
664
last_entry_num = len(entries)-1
665
last_fields = entries[last_entry_num].split('\0')
666
if len(last_fields) < entry_field_count:
667
# The very last hunk was not complete,
668
# read the previous hunk
669
after = mid + len(block) - len(entries[-1])
671
last_fields = entries[last_entry_num].split('\0')
673
after = mid + len(block)
675
last_dir = last_fields[1]
676
last_loc = bisect.bisect_right(post, last_dir)
678
middle_files = post[:last_loc]
679
post = post[last_loc:]
682
# We have files that should occur in this block
683
# (>= first, <= last)
684
# Either we will find them here, or we can mark them as
687
if middle_files[0] == first_dir:
688
# We might need to go before this location
689
pre.append(first_dir)
690
if middle_files[-1] == last_dir:
691
post.insert(0, last_dir)
693
# Find out what paths we have
694
paths = {first_dir:[first_fields]}
695
# last_dir might == first_dir so we need to be
696
# careful if we should append rather than overwrite
697
if last_entry_num != first_entry_num:
698
paths.setdefault(last_dir, []).append(last_fields)
699
for num in xrange(first_entry_num+1, last_entry_num):
700
# TODO: jam 20070223 We are already splitting here, so
701
# shouldn't we just split the whole thing rather
702
# than doing the split again in add_one_record?
703
fields = entries[num].split('\0')
704
paths.setdefault(fields[1], []).append(fields)
706
for cur_dir in middle_files:
707
for fields in paths.get(cur_dir, []):
708
# offset by 1 because of the opening '\0'
709
# consider changing fields_to_entry to avoid the
711
entry = fields_to_entry(fields[1:])
712
found.setdefault(cur_dir, []).append(entry)
714
# Now we have split up everything into pre, middle, and post, and
715
# we have handled everything that fell in 'middle'.
716
# We add 'post' first, so that we prefer to seek towards the
717
# beginning, so that we will tend to go as early as we need, and
718
# then only seek forward after that.
720
pending.append((after, high, post))
722
pending.append((low, start-1, pre))
726
def _bisect_recursive(self, dir_name_list):
727
"""Bisect for entries for all paths and their children.
729
This will use bisect to find all records for the supplied paths. It
730
will then continue to bisect for any records which are marked as
731
directories. (and renames?)
733
:param paths: A sorted list of (dir, name) pairs
734
eg: [('', 'a'), ('', 'f'), ('a/b', 'c')]
735
:return: A dictionary mapping (dir, name, file_id) => [tree_info]
737
# Map from (dir, name, file_id) => [tree_info]
740
found_dir_names = set()
742
# Directories that have been read
743
processed_dirs = set()
744
# Get the ball rolling with the first bisect for all entries.
745
newly_found = self._bisect(dir_name_list)
748
# Directories that need to be read
750
paths_to_search = set()
751
for entry_list in newly_found.itervalues():
752
for dir_name_id, trees_info in entry_list:
753
found[dir_name_id] = trees_info
754
found_dir_names.add(dir_name_id[:2])
756
for tree_info in trees_info:
757
minikind = tree_info[0]
760
# We already processed this one as a directory,
761
# we don't need to do the extra work again.
763
subdir, name, file_id = dir_name_id
764
path = osutils.pathjoin(subdir, name)
766
if path not in processed_dirs:
767
pending_dirs.add(path)
768
elif minikind == 'r':
769
# Rename, we need to directly search the target
770
# which is contained in the fingerprint column
771
dir_name = osutils.split(tree_info[1])
772
if dir_name[0] in pending_dirs:
773
# This entry will be found in the dir search
775
# TODO: We need to check if this entry has
776
# already been found. Otherwise we might be
777
# hitting infinite recursion.
778
if dir_name not in found_dir_names:
779
paths_to_search.add(dir_name)
780
# Now we have a list of paths to look for directly, and
781
# directory blocks that need to be read.
782
# newly_found is mixing the keys between (dir, name) and path
783
# entries, but that is okay, because we only really care about the
785
newly_found = self._bisect(sorted(paths_to_search))
786
newly_found.update(self._bisect_dirblocks(sorted(pending_dirs)))
787
processed_dirs.update(pending_dirs)
790
def _empty_parent_info(self):
791
return [DirState.NULL_PARENT_DETAILS] * (len(self._parents) -
794
def _ensure_block(self, parent_block_index, parent_row_index, dirname):
795
"""Ensure a block for dirname exists.
797
This function exists to let callers which know that there is a
798
directory dirname ensure that the block for it exists. This block can
799
fail to exist because of demand loading, or because a directory had no
800
children. In either case it is not an error. It is however an error to
801
call this if there is no parent entry for the directory, and thus the
802
function requires the coordinates of such an entry to be provided.
804
The root row is special cased and can be indicated with a parent block
807
:param parent_block_index: The index of the block in which dirname's row
809
:param parent_row_index: The index in the parent block where the row
811
:param dirname: The utf8 dirname to ensure there is a block for.
812
:return: The index for the block.
814
if dirname == '' and parent_row_index == 0 and parent_block_index == 0:
815
# This is the signature of the root row, and the
816
# contents-of-root row is always index 1
818
# the basename of the directory must be the end of its full name.
819
if not (parent_block_index == -1 and
820
parent_block_index == -1 and dirname == ''):
821
assert dirname.endswith(
822
self._dirblocks[parent_block_index][1][parent_row_index][0][1])
823
block_index, present = self._find_block_index_from_key((dirname, '', ''))
825
## In future, when doing partial parsing, this should load and
826
# populate the entire block.
827
self._dirblocks.insert(block_index, (dirname, []))
830
def _entries_to_current_state(self, new_entries):
831
"""Load new_entries into self.dirblocks.
833
Process new_entries into the current state object, making them the active
836
:param new_entries: A sorted list of entries. This function does not sort
837
to prevent unneeded overhead when callers have a sorted list already.
840
assert new_entries[0][0][0:2] == ('', ''), \
841
"Missing root row %r" % new_entries[0][0]
842
# The two blocks here are deliberate: the root block and the
843
# contents-of-root block.
844
self._dirblocks = [('', []), ('', [])]
845
current_block = self._dirblocks[0][1]
848
append_entry = current_block.append
849
for entry in new_entries:
850
if entry[0][0] != current_dirname:
851
# new block - different dirname
853
current_dirname = entry[0][0]
854
self._dirblocks.append((current_dirname, current_block))
855
append_entry = current_block.append
856
# append the entry to the current block
858
self._split_root_dirblock_into_contents()
860
def _split_root_dirblock_into_contents(self):
861
"""Split the root dirblocks into root and contents-of-root.
863
After parsing by path, we end up with root entries and contents-of-root
864
entries in the same block. This loop splits them out again.
866
# The above loop leaves the "root block" entries mixed with the
867
# "contents-of-root block". But we don't want an if check on
868
# all entries, so instead we just fix it up here.
869
assert self._dirblocks[1] == ('', [])
871
contents_of_root_block = []
872
for entry in self._dirblocks[0][1]:
873
if not entry[0][1]: # This is a root entry
874
root_block.append(entry)
876
contents_of_root_block.append(entry)
877
self._dirblocks[0] = ('', root_block)
878
self._dirblocks[1] = ('', contents_of_root_block)
880
def _entry_to_line(self, entry):
881
"""Serialize entry to a NULL delimited line ready for _get_output_lines.
883
:param entry: An entry_tuple as defined in the module docstring.
885
entire_entry = list(entry[0])
886
for tree_number, tree_data in enumerate(entry[1]):
887
# (minikind, fingerprint, size, executable, tree_specific_string)
888
entire_entry.extend(tree_data)
889
# 3 for the key, 5 for the fields per tree.
890
tree_offset = 3 + tree_number * 5
892
entire_entry[tree_offset + 0] = tree_data[0]
894
entire_entry[tree_offset + 2] = str(tree_data[2])
896
entire_entry[tree_offset + 3] = DirState._to_yesno[tree_data[3]]
897
return '\0'.join(entire_entry)
899
def _fields_per_entry(self):
900
"""How many null separated fields should be in each entry row.
902
Each line now has an extra '\n' field which is not used
903
so we just skip over it
906
+ number of fields per tree_data (5) * tree count
909
tree_count = 1 + self._num_present_parents()
910
return 3 + 5 * tree_count + 1
912
def _find_block(self, key, add_if_missing=False):
913
"""Return the block that key should be present in.
915
:param key: A dirstate entry key.
916
:return: The block tuple.
918
block_index, present = self._find_block_index_from_key(key)
920
if not add_if_missing:
921
# check to see if key is versioned itself - we might want to
922
# add it anyway, because dirs with no entries dont get a
923
# dirblock at parse time.
924
# This is an uncommon branch to take: most dirs have children,
925
# and most code works with versioned paths.
926
parent_base, parent_name = osutils.split(key[0])
927
if not self._get_block_entry_index(parent_base, parent_name, 0)[3]:
928
# some parent path has not been added - its an error to add
930
raise errors.NotVersionedError(key[0:2], str(self))
931
self._dirblocks.insert(block_index, (key[0], []))
932
return self._dirblocks[block_index]
934
def _find_block_index_from_key(self, key):
935
"""Find the dirblock index for a key.
937
:return: The block index, True if the block for the key is present.
939
if key[0:2] == ('', ''):
941
block_index = bisect.bisect_left(self._dirblocks, (key[0], []), 1)
942
# _right returns one-past-where-key is so we have to subtract
943
# one to use it. we use _right here because there are two
944
# '' blocks - the root, and the contents of root
945
# we always have a minimum of 2 in self._dirblocks: root and
946
# root-contents, and for '', we get 2 back, so this is
947
# simple and correct:
948
present = (block_index < len(self._dirblocks) and
949
self._dirblocks[block_index][0] == key[0])
950
return block_index, present
952
def _find_entry_index(self, key, block):
953
"""Find the entry index for a key in a block.
955
:return: The entry index, True if the entry for the key is present.
957
entry_index = bisect.bisect_left(block, (key, []))
958
present = (entry_index < len(block) and
959
block[entry_index][0] == key)
960
return entry_index, present
963
def from_tree(tree, dir_state_filename):
964
"""Create a dirstate from a bzr Tree.
966
:param tree: The tree which should provide parent information and
968
:return: a DirState object which is currently locked for writing.
969
(it was locked by DirState.initialize)
971
result = DirState.initialize(dir_state_filename)
975
parent_ids = tree.get_parent_ids()
976
num_parents = len(parent_ids)
978
for parent_id in parent_ids:
979
parent_tree = tree.branch.repository.revision_tree(parent_id)
980
parent_trees.append((parent_id, parent_tree))
981
parent_tree.lock_read()
982
result.set_parent_trees(parent_trees, [])
983
result.set_state_from_inventory(tree.inventory)
985
for revid, parent_tree in parent_trees:
989
# The caller won't have a chance to unlock this, so make sure we
995
def get_ghosts(self):
996
"""Return a list of the parent tree revision ids that are ghosts."""
997
self._read_header_if_needed()
1000
def get_lines(self):
1001
"""Serialise the entire dirstate to a sequence of lines."""
1002
if (self._header_state == DirState.IN_MEMORY_UNMODIFIED and
1003
self._dirblock_state == DirState.IN_MEMORY_UNMODIFIED):
1004
# read whats on disk.
1005
self._state_file.seek(0)
1006
return self._state_file.readlines()
1008
lines.append(self._get_parents_line(self.get_parent_ids()))
1009
lines.append(self._get_ghosts_line(self._ghosts))
1010
# append the root line which is special cased
1011
lines.extend(map(self._entry_to_line, self._iter_entries()))
1012
return self._get_output_lines(lines)
1014
def _get_ghosts_line(self, ghost_ids):
1015
"""Create a line for the state file for ghost information."""
1016
return '\0'.join([str(len(ghost_ids))] + ghost_ids)
1018
def _get_parents_line(self, parent_ids):
1019
"""Create a line for the state file for parents information."""
1020
return '\0'.join([str(len(parent_ids))] + parent_ids)
1022
def _get_fields_to_entry(self):
1023
"""Get a function which converts entry fields into a entry record.
1025
This handles size and executable, as well as parent records.
1027
:return: A function which takes a list of fields, and returns an
1028
appropriate record for storing in memory.
1030
# This is intentionally unrolled for performance
1031
num_present_parents = self._num_present_parents()
1032
if num_present_parents == 0:
1033
def fields_to_entry_0_parents(fields, _int=int):
1034
path_name_file_id_key = (fields[0], fields[1], fields[2])
1035
return (path_name_file_id_key, [
1037
fields[3], # minikind
1038
fields[4], # fingerprint
1039
_int(fields[5]), # size
1040
fields[6] == 'y', # executable
1041
fields[7], # packed_stat or revision_id
1043
return fields_to_entry_0_parents
1044
elif num_present_parents == 1:
1045
def fields_to_entry_1_parent(fields, _int=int):
1046
path_name_file_id_key = (fields[0], fields[1], fields[2])
1047
return (path_name_file_id_key, [
1049
fields[3], # minikind
1050
fields[4], # fingerprint
1051
_int(fields[5]), # size
1052
fields[6] == 'y', # executable
1053
fields[7], # packed_stat or revision_id
1056
fields[8], # minikind
1057
fields[9], # fingerprint
1058
_int(fields[10]), # size
1059
fields[11] == 'y', # executable
1060
fields[12], # packed_stat or revision_id
1063
return fields_to_entry_1_parent
1064
elif num_present_parents == 2:
1065
def fields_to_entry_2_parents(fields, _int=int):
1066
path_name_file_id_key = (fields[0], fields[1], fields[2])
1067
return (path_name_file_id_key, [
1069
fields[3], # minikind
1070
fields[4], # fingerprint
1071
_int(fields[5]), # size
1072
fields[6] == 'y', # executable
1073
fields[7], # packed_stat or revision_id
1076
fields[8], # minikind
1077
fields[9], # fingerprint
1078
_int(fields[10]), # size
1079
fields[11] == 'y', # executable
1080
fields[12], # packed_stat or revision_id
1083
fields[13], # minikind
1084
fields[14], # fingerprint
1085
_int(fields[15]), # size
1086
fields[16] == 'y', # executable
1087
fields[17], # packed_stat or revision_id
1090
return fields_to_entry_2_parents
1092
def fields_to_entry_n_parents(fields, _int=int):
1093
path_name_file_id_key = (fields[0], fields[1], fields[2])
1094
trees = [(fields[cur], # minikind
1095
fields[cur+1], # fingerprint
1096
_int(fields[cur+2]), # size
1097
fields[cur+3] == 'y', # executable
1098
fields[cur+4], # stat or revision_id
1099
) for cur in xrange(3, len(fields)-1, 5)]
1100
return path_name_file_id_key, trees
1101
return fields_to_entry_n_parents
1103
def get_parent_ids(self):
1104
"""Return a list of the parent tree ids for the directory state."""
1105
self._read_header_if_needed()
1106
return list(self._parents)
1108
def _get_block_entry_index(self, dirname, basename, tree_index):
1109
"""Get the coordinates for a path in the state structure.
1111
:param dirname: The utf8 dirname to lookup.
1112
:param basename: The utf8 basename to lookup.
1113
:param tree_index: The index of the tree for which this lookup should
1115
:return: A tuple describing where the path is located, or should be
1116
inserted. The tuple contains four fields: the block index, the row
1117
index, anda two booleans are True when the directory is present, and
1118
when the entire path is present. There is no guarantee that either
1119
coordinate is currently reachable unless the found field for it is
1120
True. For instance, a directory not present in the searched tree
1121
may be returned with a value one greater than the current highest
1122
block offset. The directory present field will always be True when
1123
the path present field is True. The directory present field does
1124
NOT indicate that the directory is present in the searched tree,
1125
rather it indicates that there are at least some files in some
1128
self._read_dirblocks_if_needed()
1129
key = dirname, basename, ''
1130
block_index, present = self._find_block_index_from_key(key)
1132
# no such directory - return the dir index and 0 for the row.
1133
return block_index, 0, False, False
1134
block = self._dirblocks[block_index][1] # access the entries only
1135
entry_index, present = self._find_entry_index(key, block)
1136
# linear search through present entries at this path to find the one
1138
while entry_index < len(block) and block[entry_index][0][1] == basename:
1139
if block[entry_index][1][tree_index][0] not in \
1140
('a', 'r'): # absent, relocated
1141
return block_index, entry_index, True, True
1143
return block_index, entry_index, True, False
1145
def _get_entry(self, tree_index, fileid_utf8=None, path_utf8=None):
1146
"""Get the dirstate entry for path in tree tree_index
1148
If either file_id or path is supplied, it is used as the key to lookup.
1149
If both are supplied, the fastest lookup is used, and an error is
1150
raised if they do not both point at the same row.
1152
:param tree_index: The index of the tree we wish to locate this path
1153
in. If the path is present in that tree, the entry containing its
1154
details is returned, otherwise (None, None) is returned
1155
:param fileid_utf8: A utf8 file_id to look up.
1156
:param path_utf8: An utf8 path to be looked up.
1157
:return: The dirstate entry tuple for path, or (None, None)
1159
self._read_dirblocks_if_needed()
1160
if path_utf8 is not None:
1161
assert path_utf8.__class__ == str, 'path_utf8 is not a str: %s %s' % (type(path_utf8), path_utf8)
1162
# path lookups are faster
1163
dirname, basename = osutils.split(path_utf8)
1164
block_index, entry_index, dir_present, file_present = \
1165
self._get_block_entry_index(dirname, basename, tree_index)
1166
if not file_present:
1168
entry = self._dirblocks[block_index][1][entry_index]
1169
assert entry[0][2] and entry[1][tree_index][0] not in ('a', 'r'), 'unversioned entry?!?!'
1171
if entry[0][2] != fileid_utf8:
1172
raise errors.BzrError('integrity error ? : mismatching'
1173
' tree_index, file_id and path')
1176
assert fileid_utf8 is not None
1177
possible_keys = self._get_id_index().get(fileid_utf8, None)
1178
if not possible_keys:
1180
for key in possible_keys:
1181
(block_index, entry_index, dir_present,
1182
file_present) = self._get_block_entry_index(key[0], key[1],
1185
entry = self._dirblocks[block_index][1][entry_index]
1186
# _get_block_entry_index only returns entries that are not
1187
# absent in the current tree. _get_id_index will return
1188
# both locations for a renamed file. It is possible that a
1189
# new file was added at the same location that the old file
1190
# was renamed away. So _get_block_entry_index will actually
1191
# match the new file, skipping the fact that the real entry
1192
# we want is the rename. By just continuing here, we should
1193
# find the record at the target location, because
1194
# _get_id_index should return all locations.
1195
if entry[0][2] != fileid_utf8:
1197
assert entry[1][tree_index][0] not in ('a', 'r')
1198
assert key == entry[0], ('We were told that %s would be at'
1199
' %s, %s, but we found %s' % (key, block_index,
1200
entry_index, entry))
1205
def initialize(path):
1206
"""Create a new dirstate on path.
1208
The new dirstate will be an empty tree - that is it has no parents,
1209
and only a root node - which has id ROOT_ID.
1211
The object will be write locked when returned to the caller,
1212
unless there was an exception in the writing, in which case it
1215
:param path: The name of the file for the dirstate.
1216
:return: A DirState object.
1218
# This constructs a new DirState object on a path, sets the _state_file
1219
# to a new empty file for that path. It then calls _set_data() with our
1220
# stock empty dirstate information - a root with ROOT_ID, no children,
1221
# and no parents. Finally it calls save() to ensure that this data will
1223
result = DirState(path)
1224
# root dir and root dir contents with no children.
1225
empty_tree_dirblocks = [('', []), ('', [])]
1226
# a new root directory, with a NULLSTAT.
1227
empty_tree_dirblocks[0][1].append(
1228
(('', '', bzrlib.inventory.ROOT_ID), [
1229
('d', '', 0, False, DirState.NULLSTAT),
1233
result._set_data([], empty_tree_dirblocks)
1240
def _inv_entry_to_details(self, inv_entry):
1241
"""Convert an inventory entry (from a revision tree) to state details.
1243
:param inv_entry: An inventory entry whose sha1 and link targets can be
1244
relied upon, and which has a revision set.
1245
:return: A details tuple - the details for a single tree at a path +
1248
kind = inv_entry.kind
1249
minikind = DirState._kind_to_minikind[kind]
1250
tree_data = inv_entry.revision
1251
assert len(tree_data) > 0, 'empty revision for the inv_entry.'
1252
if kind == 'directory':
1256
elif kind == 'symlink':
1257
fingerprint = inv_entry.symlink_target or ''
1260
elif kind == 'file':
1261
fingerprint = inv_entry.text_sha1 or ''
1262
size = inv_entry.text_size or 0
1263
executable = inv_entry.executable
1266
return (minikind, fingerprint, size, executable, tree_data)
1268
def _iter_entries(self):
1269
"""Iterate over all the entries in the dirstate.
1271
Each yelt item is an entry in the standard format described in the
1272
docstring of bzrlib.dirstate.
1274
self._read_dirblocks_if_needed()
1275
for directory in self._dirblocks:
1276
for entry in directory[1]:
1279
def _get_id_index(self):
1280
"""Get an id index of self._dirblocks."""
1281
if self._id_index is None:
1283
for key, tree_details in self._iter_entries():
1284
id_index.setdefault(key[2], set()).add(key)
1285
self._id_index = id_index
1286
return self._id_index
1288
def _get_output_lines(self, lines):
1289
"""format lines for final output.
1291
:param lines: A sequece of lines containing the parents list and the
1294
output_lines = ['#bazaar dirstate flat format 2\n']
1295
lines.append('') # a final newline
1296
inventory_text = '\0\n\0'.join(lines)
1297
output_lines.append('adler32: %s\n' % (zlib.adler32(inventory_text),))
1298
# -3, 1 for num parents, 1 for ghosts, 1 for final newline
1299
num_entries = len(lines)-3
1300
output_lines.append('num_entries: %s\n' % (num_entries,))
1301
output_lines.append(inventory_text)
1304
def _make_deleted_row(self, fileid_utf8, parents):
1305
"""Return a deleted for for fileid_utf8."""
1306
return ('/', 'RECYCLED.BIN', 'file', fileid_utf8, 0, DirState.NULLSTAT,
1309
def _num_present_parents(self):
1310
"""The number of parent entries in each record row."""
1311
return len(self._parents) - len(self._ghosts)
1315
"""Construct a DirState on the file at path path.
1317
:return: An unlocked DirState object, associated with the given path.
1319
result = DirState(path)
1322
def _read_dirblocks_if_needed(self):
1323
"""Read in all the dirblocks from the file if they are not in memory.
1325
This populates self._dirblocks, and sets self._dirblock_state to
1326
IN_MEMORY_UNMODIFIED. It is not currently ready for incremental block
1329
self._read_header_if_needed()
1330
if self._dirblock_state == DirState.NOT_IN_MEMORY:
1331
# move the _state_file pointer to after the header (in case bisect
1332
# has been called in the mean time)
1333
self._state_file.seek(self._end_of_header)
1334
text = self._state_file.read()
1335
# TODO: check the adler checksums. adler_measured = zlib.adler32(text)
1337
fields = text.split('\0')
1338
# Remove the last blank entry
1339
trailing = fields.pop()
1340
assert trailing == ''
1341
# consider turning fields into a tuple.
1343
# skip the first field which is the trailing null from the header.
1345
# Each line now has an extra '\n' field which is not used
1346
# so we just skip over it
1348
# 3 fields for the key
1349
# + number of fields per tree_data (5) * tree count
1351
num_present_parents = self._num_present_parents()
1352
tree_count = 1 + num_present_parents
1353
entry_size = self._fields_per_entry()
1354
expected_field_count = entry_size * self._num_entries
1355
if len(fields) - cur > expected_field_count:
1356
fields = fields[:expected_field_count + cur]
1357
trace.mutter('Unexpectedly long dirstate field count!')
1358
print "XXX: incorrectly truncated dirstate file bug triggered."
1359
field_count = len(fields)
1360
# this checks our adjustment, and also catches file too short.
1361
assert field_count - cur == expected_field_count, \
1362
'field count incorrect %s != %s, entry_size=%s, '\
1363
'num_entries=%s fields=%r' % (
1364
field_count - cur, expected_field_count, entry_size,
1365
self._num_entries, fields)
1367
if num_present_parents == 1:
1368
# Bind external functions to local names
1370
# We access all fields in order, so we can just iterate over
1371
# them. Grab an straight iterator over the fields. (We use an
1372
# iterator because we don't want to do a lot of additions, nor
1373
# do we want to do a lot of slicing)
1374
next = iter(fields).next
1375
# Move the iterator to the current position
1376
for x in xrange(cur):
1378
# The two blocks here are deliberate: the root block and the
1379
# contents-of-root block.
1380
self._dirblocks = [('', []), ('', [])]
1381
current_block = self._dirblocks[0][1]
1382
current_dirname = ''
1383
append_entry = current_block.append
1384
for count in xrange(self._num_entries):
1388
if dirname != current_dirname:
1389
# new block - different dirname
1391
current_dirname = dirname
1392
self._dirblocks.append((current_dirname, current_block))
1393
append_entry = current_block.append
1394
# we know current_dirname == dirname, so re-use it to avoid
1395
# creating new strings
1396
entry = ((current_dirname, name, file_id),
1399
next(), # fingerprint
1400
_int(next()), # size
1401
next() == 'y', # executable
1402
next(), # packed_stat or revision_id
1406
next(), # fingerprint
1407
_int(next()), # size
1408
next() == 'y', # executable
1409
next(), # packed_stat or revision_id
1413
assert trailing == '\n'
1414
# append the entry to the current block
1416
self._split_root_dirblock_into_contents()
1418
fields_to_entry = self._get_fields_to_entry()
1419
entries = [fields_to_entry(fields[pos:pos+entry_size])
1420
for pos in xrange(cur, field_count, entry_size)]
1421
self._entries_to_current_state(entries)
1422
self._dirblock_state = DirState.IN_MEMORY_UNMODIFIED
1424
def _read_header(self):
1425
"""This reads in the metadata header, and the parent ids.
1427
After reading in, the file should be positioned at the null
1428
just before the start of the first record in the file.
1430
:return: (expected adler checksum, number of entries, parent list)
1432
self._read_prelude()
1433
parent_line = self._state_file.readline()
1434
info = parent_line.split('\0')
1435
num_parents = int(info[0])
1436
assert num_parents == len(info)-2, 'incorrect parent info line'
1437
self._parents = info[1:-1]
1439
ghost_line = self._state_file.readline()
1440
info = ghost_line.split('\0')
1441
num_ghosts = int(info[1])
1442
assert num_ghosts == len(info)-3, 'incorrect ghost info line'
1443
self._ghosts = info[2:-1]
1444
self._header_state = DirState.IN_MEMORY_UNMODIFIED
1445
self._end_of_header = self._state_file.tell()
1447
def _read_header_if_needed(self):
1448
"""Read the header of the dirstate file if needed."""
1449
# inline this as it will be called a lot
1450
if not self._lock_token:
1451
raise errors.ObjectNotLocked(self)
1452
if self._header_state == DirState.NOT_IN_MEMORY:
1455
def _read_prelude(self):
1456
"""Read in the prelude header of the dirstate file
1458
This only reads in the stuff that is not connected to the adler
1459
checksum. The position will be correct to read in the rest of
1460
the file and check the checksum after this point.
1461
The next entry in the file should be the number of parents,
1462
and their ids. Followed by a newline.
1464
header = self._state_file.readline()
1465
assert header == '#bazaar dirstate flat format 2\n', \
1466
'invalid header line: %r' % (header,)
1467
adler_line = self._state_file.readline()
1468
assert adler_line.startswith('adler32: '), 'missing adler32 checksum'
1469
self.adler_expected = int(adler_line[len('adler32: '):-1])
1470
num_entries_line = self._state_file.readline()
1471
assert num_entries_line.startswith('num_entries: '), 'missing num_entries line'
1472
self._num_entries = int(num_entries_line[len('num_entries: '):-1])
1475
"""Save any pending changes created during this session.
1477
We reuse the existing file, because that prevents race conditions with
1478
file creation, and we expect to be using oslocks on it in the near
1479
future to prevent concurrent modification and reads - because dirstates
1480
incremental data aggretation is not compatible with reading a modified
1481
file, and replacing a file in use by another process is impossible on
1484
A dirstate in read only mode should be smart enough though to validate
1485
that the file has not changed, and otherwise discard its cache and
1486
start over, to allow for fine grained read lock duration, so 'status'
1487
wont block 'commit' - for example.
1489
if (self._header_state == DirState.IN_MEMORY_MODIFIED or
1490
self._dirblock_state == DirState.IN_MEMORY_MODIFIED):
1491
self._state_file.seek(0)
1492
self._state_file.writelines(self.get_lines())
1493
self._state_file.truncate()
1494
self._state_file.flush()
1495
self._header_state = DirState.IN_MEMORY_UNMODIFIED
1496
self._dirblock_state = DirState.IN_MEMORY_UNMODIFIED
1498
def _set_data(self, parent_ids, dirblocks):
1499
"""Set the full dirstate data in memory.
1501
This is an internal function used to completely replace the objects
1502
in memory state. It puts the dirstate into state 'full-dirty'.
1504
:param parent_ids: A list of parent tree revision ids.
1505
:param dirblocks: A list containing one tuple for each directory in the
1506
tree. Each tuple contains the directory path and a list of entries
1507
found in that directory.
1509
# our memory copy is now authoritative.
1510
self._dirblocks = dirblocks
1511
self._header_state = DirState.IN_MEMORY_MODIFIED
1512
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
1513
self._parents = list(parent_ids)
1514
self._id_index = None
1516
def set_path_id(self, path, new_id):
1517
"""Change the id of path to new_id in the current working tree.
1519
:param path: The path inside the tree to set - '' is the root, 'foo'
1520
is the path foo in the root.
1521
:param new_id: The new id to assign to the path. This must be a utf8
1522
file id (not unicode, and not None).
1524
# TODO: start warning here.
1525
assert new_id.__class__ == str
1526
self._read_dirblocks_if_needed()
1528
import pdb;pdb.set_trace()
1530
raise NotImplementedError(self.set_path_id)
1531
# TODO: check new id is unique
1532
entry = self._get_entry(0, path_utf8=path)
1533
if entry[0][2] == new_id:
1534
# Nothing to change.
1536
# mark the old path absent, and insert a new root path
1537
self._make_absent(entry)
1538
self.update_minimal(('', '', new_id), 'd',
1539
path_utf8='', packed_stat=entry[1][0][4])
1540
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
1541
if self._id_index is not None:
1542
self._id_index.setdefault(new_id, set()).add(entry[0])
1544
def set_parent_trees(self, trees, ghosts):
1545
"""Set the parent trees for the dirstate.
1547
:param trees: A list of revision_id, tree tuples. tree must be provided
1548
even if the revision_id refers to a ghost: supply an empty tree in
1550
:param ghosts: A list of the revision_ids that are ghosts at the time
1553
# TODO: generate a list of parent indexes to preserve to save
1554
# processing specific parent trees. In the common case one tree will
1555
# be preserved - the left most parent.
1556
# TODO: if the parent tree is a dirstate, we might want to walk them
1557
# all by path in parallel for 'optimal' common-case performance.
1558
# generate new root row.
1559
self._read_dirblocks_if_needed()
1560
# TODO future sketch: Examine the existing parents to generate a change
1561
# map and then walk the new parent trees only, mapping them into the
1562
# dirstate. Walk the dirstate at the same time to remove unreferenced
1565
# sketch: loop over all entries in the dirstate, cherry picking
1566
# entries from the parent trees, if they are not ghost trees.
1567
# after we finish walking the dirstate, all entries not in the dirstate
1568
# are deletes, so we want to append them to the end as per the design
1569
# discussions. So do a set difference on ids with the parents to
1570
# get deletes, and add them to the end.
1571
# During the update process we need to answer the following questions:
1572
# - find other keys containing a fileid in order to create cross-path
1573
# links. We dont't trivially use the inventory from other trees
1574
# because this leads to either double touching, or to accessing
1576
# - find other keys containing a path
1577
# We accumulate each entry via this dictionary, including the root
1580
# we could do parallel iterators, but because file id data may be
1581
# scattered throughout, we dont save on index overhead: we have to look
1582
# at everything anyway. We can probably save cycles by reusing parent
1583
# data and doing an incremental update when adding an additional
1584
# parent, but for now the common cases are adding a new parent (merge),
1585
# and replacing completely (commit), and commit is more common: so
1586
# optimise merge later.
1588
# ---- start generation of full tree mapping data
1589
# what trees should we use?
1590
parent_trees = [tree for rev_id, tree in trees if rev_id not in ghosts]
1591
# how many trees do we end up with
1592
parent_count = len(parent_trees)
1594
# one: the current tree
1595
for entry in self._iter_entries():
1596
# skip entries not in the current tree
1597
if entry[1][0][0] in ('a', 'r'): # absent, relocated
1599
by_path[entry[0]] = [entry[1][0]] + \
1600
[DirState.NULL_PARENT_DETAILS] * parent_count
1601
id_index[entry[0][2]] = set([entry[0]])
1603
# now the parent trees:
1604
for tree_index, tree in enumerate(parent_trees):
1605
# the index is off by one, adjust it.
1606
tree_index = tree_index + 1
1607
# when we add new locations for a fileid we need these ranges for
1608
# any fileid in this tree as we set the by_path[id] to:
1609
# already_processed_tree_details + new_details + new_location_suffix
1610
# the suffix is from tree_index+1:parent_count+1.
1611
new_location_suffix = [DirState.NULL_PARENT_DETAILS] * (parent_count - tree_index)
1612
# now stitch in all the entries from this tree
1613
for path, entry in tree.inventory.iter_entries_by_dir():
1614
# here we process each trees details for each item in the tree.
1615
# we first update any existing entries for the id at other paths,
1616
# then we either create or update the entry for the id at the
1617
# right path, and finally we add (if needed) a mapping from
1618
# file_id to this path. We do it in this order to allow us to
1619
# avoid checking all known paths for the id when generating a
1620
# new entry at this path: by adding the id->path mapping last,
1621
# all the mappings are valid and have correct relocation
1622
# records where needed.
1623
file_id = entry.file_id
1624
path_utf8 = path.encode('utf8')
1625
dirname, basename = osutils.split(path_utf8)
1626
new_entry_key = (dirname, basename, file_id)
1627
# tree index consistency: All other paths for this id in this tree
1628
# index must point to the correct path.
1629
for entry_key in id_index.setdefault(file_id, set()):
1630
# TODO:PROFILING: It might be faster to just update
1631
# rather than checking if we need to, and then overwrite
1632
# the one we are located at.
1633
if entry_key != new_entry_key:
1634
# this file id is at a different path in one of the
1635
# other trees, so put absent pointers there
1636
# This is the vertical axis in the matrix, all pointing
1638
by_path[entry_key][tree_index] = ('r', path_utf8, 0, False, '')
1639
# by path consistency: Insert into an existing path record (trivial), or
1640
# add a new one with relocation pointers for the other tree indexes.
1641
if new_entry_key in id_index[file_id]:
1642
# there is already an entry where this data belongs, just insert it.
1643
by_path[new_entry_key][tree_index] = \
1644
self._inv_entry_to_details(entry)
1646
# add relocated entries to the horizontal axis - this row
1647
# mapping from path,id. We need to look up the correct path
1648
# for the indexes from 0 to tree_index -1
1650
for lookup_index in xrange(tree_index):
1651
# boundary case: this is the first occurence of file_id
1652
# so there are no id_indexs, possibly take this out of
1654
if not len(id_index[file_id]):
1655
new_details.append(DirState.NULL_PARENT_DETAILS)
1657
# grab any one entry, use it to find the right path.
1658
# TODO: optimise this to reduce memory use in highly
1659
# fragmented situations by reusing the relocation
1661
a_key = iter(id_index[file_id]).next()
1662
if by_path[a_key][lookup_index][0] in ('r', 'a'):
1663
# its a pointer or missing statement, use it as is.
1664
new_details.append(by_path[a_key][lookup_index])
1666
# we have the right key, make a pointer to it.
1667
real_path = ('/'.join(a_key[0:2])).strip('/')
1668
new_details.append(('r', real_path, 0, False, ''))
1669
new_details.append(self._inv_entry_to_details(entry))
1670
new_details.extend(new_location_suffix)
1671
by_path[new_entry_key] = new_details
1672
id_index[file_id].add(new_entry_key)
1673
# --- end generation of full tree mappings
1675
# sort and output all the entries
1676
new_entries = sorted(by_path.items())
1677
self._entries_to_current_state(new_entries)
1678
self._parents = [rev_id for rev_id, tree in trees]
1679
self._ghosts = list(ghosts)
1680
self._header_state = DirState.IN_MEMORY_MODIFIED
1681
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
1682
self._id_index = id_index
1684
def set_state_from_inventory(self, new_inv):
1685
"""Set new_inv as the current state.
1687
This API is called by tree transform, and will usually occur with
1688
existing parent trees.
1690
:param new_inv: The inventory object to set current state from.
1692
self._read_dirblocks_if_needed()
1694
# incremental algorithm:
1695
# two iterators: current data and new data, both in dirblock order.
1696
new_iterator = new_inv.iter_entries_by_dir()
1697
# we will be modifying the dirstate, so we need a stable iterator. In
1698
# future we might write one, for now we just clone the state into a
1699
# list - which is a shallow copy, so each
1700
old_iterator = iter(list(self._iter_entries()))
1701
# both must have roots so this is safe:
1702
current_new = new_iterator.next()
1703
current_old = old_iterator.next()
1704
def advance(iterator):
1706
return iterator.next()
1707
except StopIteration:
1709
while current_new or current_old:
1710
# skip entries in old that are not really there
1711
if current_old and current_old[1][0][0] in ('r', 'a'):
1712
# relocated or absent
1713
current_old = advance(old_iterator)
1716
# convert new into dirblock style
1717
new_path_utf8 = current_new[0].encode('utf8')
1718
new_dirname, new_basename = osutils.split(new_path_utf8)
1719
new_id = current_new[1].file_id
1720
new_entry_key = (new_dirname, new_basename, new_id)
1721
current_new_minikind = \
1722
DirState._kind_to_minikind[current_new[1].kind]
1724
# for safety disable variables
1725
new_path_utf8 = new_dirname = new_basename = new_id = new_entry_key = None
1726
# 5 cases, we dont have a value that is strictly greater than everything, so
1727
# we make both end conditions explicit
1729
# old is finished: insert current_new into the state.
1730
self.update_minimal(new_entry_key, current_new_minikind,
1731
executable=current_new[1].executable,
1732
path_utf8=new_path_utf8)
1733
current_new = advance(new_iterator)
1734
elif not current_new:
1736
self._make_absent(current_old)
1737
current_old = advance(old_iterator)
1738
elif new_entry_key == current_old[0]:
1739
# same - common case
1740
# TODO: update the record if anything significant has changed.
1741
# the minimal required trigger is if the execute bit or cached
1743
if (current_old[1][0][3] != current_new[1].executable or
1744
current_old[1][0][0] != current_new_minikind):
1745
self.update_minimal(current_old[0], current_new_minikind,
1746
executable=current_new[1].executable,
1747
path_utf8=new_path_utf8)
1748
# both sides are dealt with, move on
1749
current_old = advance(old_iterator)
1750
current_new = advance(new_iterator)
1751
elif new_entry_key < current_old[0]:
1753
# add a entry for this and advance new
1754
self.update_minimal(new_entry_key, current_new_minikind,
1755
executable=current_new[1].executable,
1756
path_utf8=new_path_utf8)
1757
current_new = advance(new_iterator)
1760
self._make_absent(current_old)
1761
current_old = advance(old_iterator)
1762
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
1763
self._id_index = None
1765
def _make_absent(self, current_old):
1766
"""Mark current_old - an entry - as absent for tree 0.
1768
:return: True if this was the last details entry for they entry key:
1769
that is, if the underlying block has had the entry removed, thus
1770
shrinking in length.
1772
# build up paths that this id will be left at after the change is made,
1773
# so we can update their cross references in tree 0
1774
all_remaining_keys = set()
1775
# Dont check the working tree, because its going.
1776
for details in current_old[1][1:]:
1777
if details[0] not in ('a', 'r'): # absent, relocated
1778
all_remaining_keys.add(current_old[0])
1779
elif details[0] == 'r': # relocated
1780
# record the key for the real path.
1781
all_remaining_keys.add(tuple(osutils.split(details[1])) + (current_old[0][2],))
1782
# absent rows are not present at any path.
1783
last_reference = current_old[0] not in all_remaining_keys
1785
# the current row consists entire of the current item (being marked
1786
# absent), and relocated or absent entries for the other trees:
1787
# Remove it, its meaningless.
1788
block = self._find_block(current_old[0])
1789
entry_index, present = self._find_entry_index(current_old[0], block[1])
1791
block[1].pop(entry_index)
1792
# if we have an id_index in use, remove this key from it for this id.
1793
if self._id_index is not None:
1794
self._id_index[current_old[0][2]].remove(current_old[0])
1795
# update all remaining keys for this id to record it as absent. The
1796
# existing details may either be the record we are making as deleted
1797
# (if there were other trees with the id present at this path), or may
1799
for update_key in all_remaining_keys:
1800
update_block_index, present = \
1801
self._find_block_index_from_key(update_key)
1803
update_entry_index, present = \
1804
self._find_entry_index(update_key, self._dirblocks[update_block_index][1])
1806
update_tree_details = self._dirblocks[update_block_index][1][update_entry_index][1]
1807
# it must not be absent at the moment
1808
assert update_tree_details[0][0] != 'a' # absent
1809
update_tree_details[0] = DirState.NULL_PARENT_DETAILS
1810
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
1811
return last_reference
1813
def update_minimal(self, key, minikind, executable=False, fingerprint='',
1814
packed_stat=None, size=0, path_utf8=None):
1815
"""Update an entry to the state in tree 0.
1817
This will either create a new entry at 'key' or update an existing one.
1818
It also makes sure that any other records which might mention this are
1821
:param key: (dir, name, file_id) for the new entry
1822
:param minikind: The type for the entry ('f' == 'file', 'd' ==
1824
:param executable: Should the executable bit be set?
1825
:param fingerprint: Simple fingerprint for new entry.
1826
:param packed_stat: packed stat value for new entry.
1827
:param size: Size information for new entry
1828
:param path_utf8: key[0] + '/' + key[1], just passed in to avoid doing
1831
block = self._find_block(key)[1]
1832
if packed_stat is None:
1833
packed_stat = DirState.NULLSTAT
1834
entry_index, present = self._find_entry_index(key, block)
1835
new_details = (minikind, fingerprint, size, executable, packed_stat)
1836
id_index = self._get_id_index()
1838
# new entry, synthesis cross reference here,
1839
existing_keys = id_index.setdefault(key[2], set())
1840
if not existing_keys:
1841
# not currently in the state, simplest case
1842
new_entry = key, [new_details] + self._empty_parent_info()
1844
# present at one or more existing other paths.
1845
# grab one of them and use it to generate parent
1846
# relocation/absent entries.
1847
new_entry = key, [new_details]
1848
for other_key in existing_keys:
1849
# change the record at other to be a pointer to this new
1850
# record. The loop looks similar to the change to
1851
# relocations when updating an existing record but its not:
1852
# the test for existing kinds is different: this can be
1853
# factored out to a helper though.
1854
other_block_index, present = self._find_block_index_from_key(other_key)
1856
other_entry_index, present = self._find_entry_index(other_key, self._dirblocks[other_block_index][1])
1858
assert path_utf8 is not None
1859
self._dirblocks[other_block_index][1][other_entry_index][1][0] = \
1860
('r', path_utf8, 0, False, '')
1862
num_present_parents = self._num_present_parents()
1863
for lookup_index in xrange(1, num_present_parents + 1):
1864
# grab any one entry, use it to find the right path.
1865
# TODO: optimise this to reduce memory use in highly
1866
# fragmented situations by reusing the relocation
1868
update_block_index, present = \
1869
self._find_block_index_from_key(other_key)
1871
update_entry_index, present = \
1872
self._find_entry_index(other_key, self._dirblocks[update_block_index][1])
1874
update_details = self._dirblocks[update_block_index][1][update_entry_index][1][lookup_index]
1875
if update_details[0] in ('r', 'a'): # relocated, absent
1876
# its a pointer or absent in lookup_index's tree, use
1878
new_entry[1].append(update_details)
1880
# we have the right key, make a pointer to it.
1881
pointer_path = osutils.pathjoin(*other_key[0:2])
1882
new_entry[1].append(('r', pointer_path, 0, False, ''))
1883
block.insert(entry_index, new_entry)
1884
existing_keys.add(key)
1886
# Does the new state matter?
1887
block[entry_index][1][0] = new_details
1888
# parents cannot be affected by what we do.
1889
# other occurences of this id can be found
1890
# from the id index.
1892
# tree index consistency: All other paths for this id in this tree
1893
# index must point to the correct path. We have to loop here because
1894
# we may have passed entries in the state with this file id already
1895
# that were absent - where parent entries are - and they need to be
1896
# converted to relocated.
1897
assert path_utf8 is not None
1898
for entry_key in id_index.setdefault(key[2], set()):
1899
# TODO:PROFILING: It might be faster to just update
1900
# rather than checking if we need to, and then overwrite
1901
# the one we are located at.
1902
if entry_key != key:
1903
# this file id is at a different path in one of the
1904
# other trees, so put absent pointers there
1905
# This is the vertical axis in the matrix, all pointing
1907
block_index, present = self._find_block_index_from_key(entry_key)
1909
entry_index, present = self._find_entry_index(entry_key, self._dirblocks[block_index][1])
1911
self._dirblocks[block_index][1][entry_index][1][0] = \
1912
('r', path_utf8, 0, False, '')
1913
# add a containing dirblock if needed.
1914
if new_details[0] == 'd':
1915
subdir_key = (osutils.pathjoin(*key[0:2]), '', '')
1916
block_index, present = self._find_block_index_from_key(subdir_key)
1918
self._dirblocks.insert(block_index, (subdir_key[0], []))
1920
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
1923
def _wipe_state(self):
1924
"""Forget all state information about the dirstate."""
1925
self._header_state = DirState.NOT_IN_MEMORY
1926
self._dirblock_state = DirState.NOT_IN_MEMORY
1929
self._dirblocks = []
1931
def lock_read(self):
1932
"""Acquire a read lock on the dirstate"""
1933
if self._lock_token is not None:
1934
raise errors.LockContention(self._lock_token)
1935
self._lock_token = lock.ReadLock(self._filename)
1936
self._state_file = self._lock_token.f
1939
def lock_write(self):
1940
"""Acquire a write lock on the dirstate"""
1941
if self._lock_token is not None:
1942
raise errors.LockContention(self._lock_token)
1943
self._lock_token = lock.WriteLock(self._filename)
1944
self._state_file = self._lock_token.f
1948
"""Drop any locks held on the dirstate"""
1949
if self._lock_token is None:
1950
raise errors.LockNotHeld(self)
1951
self._state_file = None
1952
self._lock_token.unlock()
1953
self._lock_token = None
1955
def _requires_lock(self):
1956
"""Checks that a lock is currently held by someone on the dirstate"""
1957
if not self._lock_token:
1958
raise errors.ObjectNotLocked(self)
1960
def pack_stat(st, _encode=base64.encodestring, _pack=struct.pack):
1961
"""Convert stat values into a packed representation."""
1962
# jam 20060614 it isn't really worth removing more entries if we
1963
# are going to leave it in packed form.
1964
# With only st_mtime and st_mode filesize is 5.5M and read time is 275ms
1965
# With all entries filesize is 5.9M and read time is mabye 280ms
1966
# well within the noise margin
1968
# base64.encode always adds a final newline, so strip it off
1969
return _encode(_pack('>llllll'
1970
, st.st_size, st.st_mtime, st.st_ctime
1971
, st.st_dev, st.st_ino, st.st_mode))[:-1]
added
removed
bzrlib/dirstate.py
