1
# Copyright (C) 2007-2010 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
"""Helper functions for DirState.
19
This is the python implementation for DirState functions.
29
from bzrlib import cache_utf8, errors, osutils
30
from bzrlib.dirstate import DirState
31
from bzrlib.osutils import parent_directories, pathjoin, splitpath
34
# This is the Windows equivalent of ENOTDIR
35
# It is defined in pywin32.winerror, but we don't want a strong dependency for
37
# XXX: Perhaps we could get it from a windows header ?
38
cdef int ERROR_PATH_NOT_FOUND
39
ERROR_PATH_NOT_FOUND = 3
40
cdef int ERROR_DIRECTORY
43
#python2.4 support, and other platform-dependent includes
44
cdef extern from "python-compat.h":
45
unsigned long htonl(unsigned long)
47
# Give Pyrex some function definitions for it to understand.
48
# All of these are just hints to Pyrex, so that it can try to convert python
49
# objects into similar C objects. (such as PyInt => int).
50
# In anything defined 'cdef extern from XXX' the real C header will be
51
# imported, and the real definition will be used from there. So these are just
52
# hints, and do not need to match exactly to the C definitions.
55
ctypedef unsigned long size_t
57
cdef extern from "_dirstate_helpers_pyx.h":
62
cdef extern from "stdlib.h":
63
unsigned long int strtoul(char *nptr, char **endptr, int base)
66
cdef extern from 'sys/stat.h':
69
# On win32, this actually comes from "python-compat.h"
73
# These functions allow us access to a bit of the 'bare metal' of python
74
# objects, rather than going through the object abstraction. (For example,
75
# PyList_Append, rather than getting the 'append' attribute of the object, and
76
# creating a tuple, and then using PyCallObject).
77
# Functions that return (or take) a void* are meant to grab a C PyObject*. This
78
# differs from the Pyrex 'object'. If you declare a variable as 'object' Pyrex
79
# will automatically Py_INCREF and Py_DECREF when appropriate. But for some
80
# inner loops, we don't need to do that at all, as the reference only lasts for
82
# Note that the C API GetItem calls borrow references, so pyrex does the wrong
83
# thing if you declare e.g. object PyList_GetItem(object lst, int index) - you
84
# need to manually Py_INCREF yourself.
85
cdef extern from "Python.h":
86
ctypedef int Py_ssize_t
87
ctypedef struct PyObject:
89
int PyList_Append(object lst, object item) except -1
90
void *PyList_GetItem_object_void "PyList_GET_ITEM" (object lst, int index)
91
void *PyList_GetItem_void_void "PyList_GET_ITEM" (void * lst, int index)
92
object PyList_GET_ITEM(object lst, Py_ssize_t index)
93
int PyList_CheckExact(object)
94
Py_ssize_t PyList_GET_SIZE (object p)
96
void *PyTuple_GetItem_void_void "PyTuple_GET_ITEM" (void* tpl, int index)
97
object PyTuple_GetItem_void_object "PyTuple_GET_ITEM" (void* tpl, int index)
98
object PyTuple_GET_ITEM(object tpl, Py_ssize_t index)
101
char *PyString_AsString(object p)
102
char *PyString_AsString_obj "PyString_AsString" (PyObject *string)
103
char *PyString_AS_STRING_void "PyString_AS_STRING" (void *p)
104
int PyString_AsStringAndSize(object str, char **buffer, Py_ssize_t *length) except -1
105
object PyString_FromString(char *)
106
object PyString_FromStringAndSize(char *, Py_ssize_t)
107
int PyString_Size(object p)
108
int PyString_GET_SIZE_void "PyString_GET_SIZE" (void *p)
109
int PyString_CheckExact(object p)
110
void Py_INCREF(object o)
111
void Py_DECREF(object o)
114
cdef extern from "string.h":
115
int strncmp(char *s1, char *s2, int len)
116
void *memchr(void *s, int c, size_t len)
117
int memcmp(void *b1, void *b2, size_t len)
118
# ??? memrchr is a GNU extension :(
119
# void *memrchr(void *s, int c, size_t len)
121
# cimport all of the definitions we will need to access
122
from _static_tuple_c cimport import_static_tuple_c, StaticTuple, \
123
StaticTuple_New, StaticTuple_SET_ITEM
125
import_static_tuple_c()
127
cdef void* _my_memrchr(void *s, int c, size_t n): # cannot_raise
128
# memrchr seems to be a GNU extension, so we have to implement it ourselves
141
def _py_memrchr(s, c):
142
"""Just to expose _my_memrchr for testing.
144
:param s: The Python string to search
145
:param c: The character to search for
146
:return: The offset to the last instance of 'c' in s
153
_s = PyString_AsString(s)
154
length = PyString_Size(s)
156
_c = PyString_AsString(c)
157
assert PyString_Size(c) == 1,\
158
'Must be a single character string, not %s' % (c,)
159
found = _my_memrchr(_s, _c[0], length)
162
return <char*>found - <char*>_s
165
cdef object safe_string_from_size(char *s, Py_ssize_t size):
167
raise AssertionError(
168
'tried to create a string with an invalid size: %d'
170
return PyString_FromStringAndSize(s, size)
173
cdef int _is_aligned(void *ptr): # cannot_raise
174
"""Is this pointer aligned to an integer size offset?
176
:return: 1 if this pointer is aligned, 0 otherwise.
178
return ((<intptr_t>ptr) & ((sizeof(int))-1)) == 0
181
cdef int _cmp_by_dirs(char *path1, int size1, char *path2, int size2): # cannot_raise
182
cdef unsigned char *cur1
183
cdef unsigned char *cur2
184
cdef unsigned char *end1
185
cdef unsigned char *end2
191
if path1 == path2 and size1 == size2:
194
end1 = <unsigned char*>path1+size1
195
end2 = <unsigned char*>path2+size2
197
# Use 32-bit comparisons for the matching portion of the string.
198
# Almost all CPU's are faster at loading and comparing 32-bit integers,
199
# than they are at 8-bit integers.
200
# 99% of the time, these will be aligned, but in case they aren't just skip
202
if _is_aligned(path1) and _is_aligned(path2):
203
cur_int1 = <int*>path1
204
cur_int2 = <int*>path2
205
end_int1 = <int*>(path1 + size1 - (size1 % sizeof(int)))
206
end_int2 = <int*>(path2 + size2 - (size2 % sizeof(int)))
208
while cur_int1 < end_int1 and cur_int2 < end_int2:
209
if cur_int1[0] != cur_int2[0]:
211
cur_int1 = cur_int1 + 1
212
cur_int2 = cur_int2 + 1
214
cur1 = <unsigned char*>cur_int1
215
cur2 = <unsigned char*>cur_int2
217
cur1 = <unsigned char*>path1
218
cur2 = <unsigned char*>path2
220
while cur1 < end1 and cur2 < end2:
221
if cur1[0] == cur2[0]:
222
# This character matches, just go to the next one
226
# The current characters do not match
228
return -1 # Reached the end of path1 segment first
229
elif cur2[0] == c'/':
230
return 1 # Reached the end of path2 segment first
231
elif cur1[0] < cur2[0]:
236
# We reached the end of at least one of the strings
238
return 1 # Not at the end of cur1, must be at the end of cur2
240
return -1 # At the end of cur1, but not at cur2
241
# We reached the end of both strings
245
def cmp_by_dirs(path1, path2):
246
"""Compare two paths directory by directory.
248
This is equivalent to doing::
250
cmp(path1.split('/'), path2.split('/'))
252
The idea is that you should compare path components separately. This
253
differs from plain ``cmp(path1, path2)`` for paths like ``'a-b'`` and
254
``a/b``. "a-b" comes after "a" but would come before "a/b" lexically.
256
:param path1: first path
257
:param path2: second path
258
:return: negative number if ``path1`` comes first,
259
0 if paths are equal,
260
and positive number if ``path2`` sorts first
262
if not PyString_CheckExact(path1):
263
raise TypeError("'path1' must be a plain string, not %s: %r"
264
% (type(path1), path1))
265
if not PyString_CheckExact(path2):
266
raise TypeError("'path2' must be a plain string, not %s: %r"
267
% (type(path2), path2))
268
return _cmp_by_dirs(PyString_AsString(path1),
269
PyString_Size(path1),
270
PyString_AsString(path2),
271
PyString_Size(path2))
274
def _cmp_path_by_dirblock(path1, path2):
275
"""Compare two paths based on what directory they are in.
277
This generates a sort order, such that all children of a directory are
278
sorted together, and grandchildren are in the same order as the
279
children appear. But all grandchildren come after all children.
281
In other words, all entries in a directory are sorted together, and
282
directorys are sorted in cmp_by_dirs order.
284
:param path1: first path
285
:param path2: the second path
286
:return: negative number if ``path1`` comes first,
288
and a positive number if ``path2`` sorts first
290
if not PyString_CheckExact(path1):
291
raise TypeError("'path1' must be a plain string, not %s: %r"
292
% (type(path1), path1))
293
if not PyString_CheckExact(path2):
294
raise TypeError("'path2' must be a plain string, not %s: %r"
295
% (type(path2), path2))
296
return _cmp_path_by_dirblock_intern(PyString_AsString(path1),
297
PyString_Size(path1),
298
PyString_AsString(path2),
299
PyString_Size(path2))
302
cdef int _cmp_path_by_dirblock_intern(char *path1, int path1_len,
303
char *path2, int path2_len): # cannot_raise
304
"""Compare two paths by what directory they are in.
306
see ``_cmp_path_by_dirblock`` for details.
309
cdef int dirname1_len
311
cdef int dirname2_len
313
cdef int basename1_len
315
cdef int basename2_len
319
if path1_len == 0 and path2_len == 0:
322
if path1 == path2 and path1_len == path2_len:
331
basename1 = <char*>_my_memrchr(path1, c'/', path1_len)
333
if basename1 == NULL:
335
basename1_len = path1_len
340
dirname1_len = basename1 - path1
341
basename1 = basename1 + 1
342
basename1_len = path1_len - dirname1_len - 1
344
basename2 = <char*>_my_memrchr(path2, c'/', path2_len)
346
if basename2 == NULL:
348
basename2_len = path2_len
353
dirname2_len = basename2 - path2
354
basename2 = basename2 + 1
355
basename2_len = path2_len - dirname2_len - 1
357
cmp_val = _cmp_by_dirs(dirname1, dirname1_len,
358
dirname2, dirname2_len)
362
cur_len = basename1_len
363
if basename2_len < basename1_len:
364
cur_len = basename2_len
366
cmp_val = memcmp(basename1, basename2, cur_len)
369
if basename1_len == basename2_len:
371
if basename1_len < basename2_len:
376
def _bisect_path_left(paths, path):
377
"""Return the index where to insert path into paths.
379
This uses a path-wise comparison so we get::
389
:param paths: A list of paths to search through
390
:param path: A single path to insert
391
:return: An offset where 'path' can be inserted.
392
:seealso: bisect.bisect_left
403
if not PyList_CheckExact(paths):
404
raise TypeError("you must pass a python list for 'paths' not: %s %r"
405
% (type(paths), paths))
406
if not PyString_CheckExact(path):
407
raise TypeError("you must pass a string for 'path' not: %s %r"
408
% (type(path), path))
413
path_cstr = PyString_AsString(path)
414
path_size = PyString_Size(path)
417
_mid = (_lo + _hi) / 2
418
cur = PyList_GetItem_object_void(paths, _mid)
419
cur_cstr = PyString_AS_STRING_void(cur)
420
cur_size = PyString_GET_SIZE_void(cur)
421
if _cmp_path_by_dirblock_intern(cur_cstr, cur_size,
422
path_cstr, path_size) < 0:
429
def _bisect_path_right(paths, path):
430
"""Return the index where to insert path into paths.
432
This uses a path-wise comparison so we get::
442
:param paths: A list of paths to search through
443
:param path: A single path to insert
444
:return: An offset where 'path' can be inserted.
445
:seealso: bisect.bisect_right
456
if not PyList_CheckExact(paths):
457
raise TypeError("you must pass a python list for 'paths' not: %s %r"
458
% (type(paths), paths))
459
if not PyString_CheckExact(path):
460
raise TypeError("you must pass a string for 'path' not: %s %r"
461
% (type(path), path))
466
path_cstr = PyString_AsString(path)
467
path_size = PyString_Size(path)
470
_mid = (_lo + _hi) / 2
471
cur = PyList_GetItem_object_void(paths, _mid)
472
cur_cstr = PyString_AS_STRING_void(cur)
473
cur_size = PyString_GET_SIZE_void(cur)
474
if _cmp_path_by_dirblock_intern(path_cstr, path_size,
475
cur_cstr, cur_size) < 0:
482
def bisect_dirblock(dirblocks, dirname, lo=0, hi=None, cache=None):
483
"""Return the index where to insert dirname into the dirblocks.
485
The return value idx is such that all directories blocks in dirblock[:idx]
486
have names < dirname, and all blocks in dirblock[idx:] have names >=
489
Optional args lo (default 0) and hi (default len(dirblocks)) bound the
490
slice of a to be searched.
495
cdef char *dirname_cstr
496
cdef int dirname_size
501
if not PyList_CheckExact(dirblocks):
502
raise TypeError("you must pass a python list for 'dirblocks' not: %s %r"
503
% (type(dirblocks), dirblocks))
504
if not PyString_CheckExact(dirname):
505
raise TypeError("you must pass a string for dirname not: %s %r"
506
% (type(dirname), dirname))
513
dirname_cstr = PyString_AsString(dirname)
514
dirname_size = PyString_Size(dirname)
517
_mid = (_lo + _hi) / 2
518
# Grab the dirname for the current dirblock
519
# cur = dirblocks[_mid][0]
520
cur = PyTuple_GetItem_void_void(
521
PyList_GetItem_object_void(dirblocks, _mid), 0)
522
cur_cstr = PyString_AS_STRING_void(cur)
523
cur_size = PyString_GET_SIZE_void(cur)
524
if _cmp_by_dirs(cur_cstr, cur_size, dirname_cstr, dirname_size) < 0:
532
"""Maintain the current location, and return fields as you parse them."""
534
cdef object state # The DirState object
535
cdef object text # The overall string object
536
cdef char *text_cstr # Pointer to the beginning of text
537
cdef int text_size # Length of text
539
cdef char *end_cstr # End of text
540
cdef char *cur_cstr # Pointer to the current record
541
cdef char *next # Pointer to the end of this record
543
def __init__(self, text, state):
546
self.text_cstr = PyString_AsString(text)
547
self.text_size = PyString_Size(text)
548
self.end_cstr = self.text_cstr + self.text_size
549
self.cur_cstr = self.text_cstr
551
cdef char *get_next(self, int *size) except NULL:
552
"""Return a pointer to the start of the next field."""
554
cdef Py_ssize_t extra_len
556
if self.cur_cstr == NULL:
557
raise AssertionError('get_next() called when cur_str is NULL')
558
elif self.cur_cstr >= self.end_cstr:
559
raise AssertionError('get_next() called when there are no chars'
562
self.cur_cstr = <char*>memchr(next, c'\0', self.end_cstr - next)
563
if self.cur_cstr == NULL:
564
extra_len = self.end_cstr - next
565
raise errors.DirstateCorrupt(self.state,
566
'failed to find trailing NULL (\\0).'
567
' Trailing garbage: %r'
568
% safe_string_from_size(next, extra_len))
569
size[0] = self.cur_cstr - next
570
self.cur_cstr = self.cur_cstr + 1
573
cdef object get_next_str(self):
574
"""Get the next field as a Python string."""
577
next = self.get_next(&size)
578
return safe_string_from_size(next, size)
580
cdef int _init(self) except -1:
581
"""Get the pointer ready.
583
This assumes that the dirstate header has already been read, and we
584
already have the dirblock string loaded into memory.
585
This just initializes our memory pointers, etc for parsing of the
590
# The first field should be an empty string left over from the Header
591
first = self.get_next(&size)
592
if first[0] != c'\0' and size == 0:
593
raise AssertionError('First character should be null not: %s'
597
cdef object _get_entry(self, int num_trees, void **p_current_dirname,
599
"""Extract the next entry.
601
This parses the next entry based on the current location in
603
Each entry can be considered a "row" in the total table. And each row
604
has a fixed number of columns. It is generally broken up into "key"
605
columns, then "current" columns, and then "parent" columns.
607
:param num_trees: How many parent trees need to be parsed
608
:param p_current_dirname: A pointer to the current PyString
609
representing the directory name.
610
We pass this in as a void * so that pyrex doesn't have to
611
increment/decrement the PyObject reference counter for each
613
We use a pointer so that _get_entry can update it with the new
615
:param new_block: This is to let the caller know that it needs to
616
create a new directory block to store the next entry.
618
cdef StaticTuple path_name_file_id_key
620
cdef char *entry_size_cstr
621
cdef unsigned long int entry_size
622
cdef char* executable_cstr
623
cdef int is_executable
624
cdef char* dirname_cstr
629
cdef object fingerprint
632
# Read the 'key' information (dirname, name, file_id)
633
dirname_cstr = self.get_next(&cur_size)
634
# Check to see if we have started a new directory block.
635
# If so, then we need to create a new dirname PyString, so that it can
636
# be used in all of the tuples. This saves time and memory, by re-using
637
# the same object repeatedly.
639
# Do the cheap 'length of string' check first. If the string is a
640
# different length, then we *have* to be a different directory.
641
if (cur_size != PyString_GET_SIZE_void(p_current_dirname[0])
642
or strncmp(dirname_cstr,
643
# Extract the char* from our current dirname string. We
644
# know it is a PyString, so we can use
645
# PyString_AS_STRING, we use the _void version because
646
# we are tricking Pyrex by using a void* rather than an
648
PyString_AS_STRING_void(p_current_dirname[0]),
650
dirname = safe_string_from_size(dirname_cstr, cur_size)
651
p_current_dirname[0] = <void*>dirname
656
# Build up the key that will be used.
657
# By using <object>(void *) Pyrex will automatically handle the
658
# Py_INCREF that we need.
659
cur_dirname = <object>p_current_dirname[0]
660
# Use StaticTuple_New to pre-allocate, rather than creating a regular
661
# tuple and passing it to the StaticTuple constructor.
662
# path_name_file_id_key = StaticTuple(<object>p_current_dirname[0],
663
# self.get_next_str(),
664
# self.get_next_str(),
666
tmp = StaticTuple_New(3)
667
Py_INCREF(cur_dirname); StaticTuple_SET_ITEM(tmp, 0, cur_dirname)
668
cur_basename = self.get_next_str()
669
cur_file_id = self.get_next_str()
670
Py_INCREF(cur_basename); StaticTuple_SET_ITEM(tmp, 1, cur_basename)
671
Py_INCREF(cur_file_id); StaticTuple_SET_ITEM(tmp, 2, cur_file_id)
672
path_name_file_id_key = tmp
674
# Parse all of the per-tree information. current has the information in
675
# the same location as parent trees. The only difference is that 'info'
676
# is a 'packed_stat' for current, while it is a 'revision_id' for
678
# minikind, fingerprint, and info will be returned as regular python
680
# entry_size and is_executable will be parsed into a python Long and
681
# python Boolean, respectively.
682
# TODO: jam 20070718 Consider changin the entry_size conversion to
683
# prefer python Int when possible. They are generally faster to
684
# work with, and it will be rare that we have a file >2GB.
685
# Especially since this code is pretty much fixed at a max of
688
for i from 0 <= i < num_trees:
689
minikind = self.get_next_str()
690
fingerprint = self.get_next_str()
691
entry_size_cstr = self.get_next(&cur_size)
692
entry_size = strtoul(entry_size_cstr, NULL, 10)
693
executable_cstr = self.get_next(&cur_size)
694
is_executable = (executable_cstr[0] == c'y')
695
info = self.get_next_str()
696
# TODO: If we want to use StaticTuple_New here we need to be pretty
697
# careful. We are relying on a bit of Pyrex
698
# automatic-conversion from 'int' to PyInt, and that doesn't
699
# play well with the StaticTuple_SET_ITEM macro.
700
# Timing doesn't (yet) show a worthwile improvement in speed
701
# versus complexity and maintainability.
702
# tmp = StaticTuple_New(5)
703
# Py_INCREF(minikind); StaticTuple_SET_ITEM(tmp, 0, minikind)
704
# Py_INCREF(fingerprint); StaticTuple_SET_ITEM(tmp, 1, fingerprint)
705
# Py_INCREF(entry_size); StaticTuple_SET_ITEM(tmp, 2, entry_size)
706
# Py_INCREF(is_executable); StaticTuple_SET_ITEM(tmp, 3, is_executable)
707
# Py_INCREF(info); StaticTuple_SET_ITEM(tmp, 4, info)
708
# PyList_Append(trees, tmp)
709
PyList_Append(trees, StaticTuple(
711
fingerprint, # fingerprint
713
is_executable,# executable
714
info, # packed_stat or revision_id
717
# The returned tuple is (key, [trees])
718
ret = (path_name_file_id_key, trees)
719
# Ignore the trailing newline, but assert that it does exist, this
720
# ensures that we always finish parsing a line on an end-of-entry
722
trailing = self.get_next(&cur_size)
723
if cur_size != 1 or trailing[0] != c'\n':
724
raise errors.DirstateCorrupt(self.state,
725
'Bad parse, we expected to end on \\n, not: %d %s: %s'
726
% (cur_size, safe_string_from_size(trailing, cur_size),
730
def _parse_dirblocks(self):
731
"""Parse all dirblocks in the state file."""
733
cdef object current_block
735
cdef void * current_dirname
737
cdef int expected_entry_count
740
num_trees = self.state._num_present_parents() + 1
741
expected_entry_count = self.state._num_entries
743
# Ignore the first record
747
dirblocks = [('', current_block), ('', [])]
748
self.state._dirblocks = dirblocks
750
current_dirname = <void*>obj
754
# TODO: jam 2007-05-07 Consider pre-allocating some space for the
755
# members, and then growing and shrinking from there. If most
756
# directories have close to 10 entries in them, it would save a
757
# few mallocs if we default our list size to something
758
# reasonable. Or we could malloc it to something large (100 or
759
# so), and then truncate. That would give us a malloc + realloc,
760
# rather than lots of reallocs.
761
while self.cur_cstr < self.end_cstr:
762
entry = self._get_entry(num_trees, ¤t_dirname, &new_block)
764
# new block - different dirname
766
PyList_Append(dirblocks,
767
(<object>current_dirname, current_block))
768
PyList_Append(current_block, entry)
769
entry_count = entry_count + 1
770
if entry_count != expected_entry_count:
771
raise errors.DirstateCorrupt(self.state,
772
'We read the wrong number of entries.'
773
' We expected to read %s, but read %s'
774
% (expected_entry_count, entry_count))
775
self.state._split_root_dirblock_into_contents()
778
def _read_dirblocks(state):
779
"""Read in the dirblocks for the given DirState object.
781
This is tightly bound to the DirState internal representation. It should be
782
thought of as a member function, which is only separated out so that we can
783
re-write it in pyrex.
785
:param state: A DirState object.
787
:postcondition: The dirblocks will be loaded into the appropriate fields in
790
state._state_file.seek(state._end_of_header)
791
text = state._state_file.read()
792
# TODO: check the crc checksums. crc_measured = zlib.crc32(text)
794
reader = Reader(text, state)
796
reader._parse_dirblocks()
797
state._dirblock_state = DirState.IN_MEMORY_UNMODIFIED
800
cdef int minikind_from_mode(int mode): # cannot_raise
801
# in order of frequency:
811
_encode = binascii.b2a_base64
814
from struct import pack
815
cdef _pack_stat(stat_value):
816
"""return a string representing the stat value's key fields.
818
:param stat_value: A stat oject with st_size, st_mtime, st_ctime, st_dev,
819
st_ino and st_mode fields.
821
cdef char result[6*4] # 6 long ints
823
aliased = <int *>result
824
aliased[0] = htonl(stat_value.st_size)
825
aliased[1] = htonl(int(stat_value.st_mtime))
826
aliased[2] = htonl(int(stat_value.st_ctime))
827
aliased[3] = htonl(stat_value.st_dev)
828
aliased[4] = htonl(stat_value.st_ino & 0xFFFFFFFF)
829
aliased[5] = htonl(stat_value.st_mode)
830
packed = PyString_FromStringAndSize(result, 6*4)
831
return _encode(packed)[:-1]
834
def update_entry(self, entry, abspath, stat_value):
835
"""Update the entry based on what is actually on disk.
837
This function only calculates the sha if it needs to - if the entry is
838
uncachable, or clearly different to the first parent's entry, no sha
839
is calculated, and None is returned.
841
:param entry: This is the dirblock entry for the file in question.
842
:param abspath: The path on disk for this file.
843
:param stat_value: (optional) if we already have done a stat on the
845
:return: None, or The sha1 hexdigest of the file (40 bytes) or link
848
return _update_entry(self, entry, abspath, stat_value)
851
cdef _update_entry(self, entry, abspath, stat_value):
852
"""Update the entry based on what is actually on disk.
854
This function only calculates the sha if it needs to - if the entry is
855
uncachable, or clearly different to the first parent's entry, no sha
856
is calculated, and None is returned.
858
:param self: The dirstate object this is operating on.
859
:param entry: This is the dirblock entry for the file in question.
860
:param abspath: The path on disk for this file.
861
:param stat_value: The stat value done on the path.
862
:return: None, or The sha1 hexdigest of the file (40 bytes) or link
865
# TODO - require pyrex 0.9.8, then use a pyd file to define access to the
866
# _st mode of the compiled stat objects.
867
cdef int minikind, saved_minikind
869
cdef int worth_saving
870
minikind = minikind_from_mode(stat_value.st_mode)
873
packed_stat = _pack_stat(stat_value)
874
details = PyList_GetItem_void_void(PyTuple_GetItem_void_void(<void *>entry, 1), 0)
875
saved_minikind = PyString_AsString_obj(<PyObject *>PyTuple_GetItem_void_void(details, 0))[0]
876
if minikind == c'd' and saved_minikind == c't':
878
saved_link_or_sha1 = PyTuple_GetItem_void_object(details, 1)
879
saved_file_size = PyTuple_GetItem_void_object(details, 2)
880
saved_executable = PyTuple_GetItem_void_object(details, 3)
881
saved_packed_stat = PyTuple_GetItem_void_object(details, 4)
882
# Deal with pyrex decrefing the objects
883
Py_INCREF(saved_link_or_sha1)
884
Py_INCREF(saved_file_size)
885
Py_INCREF(saved_executable)
886
Py_INCREF(saved_packed_stat)
887
#(saved_minikind, saved_link_or_sha1, saved_file_size,
888
# saved_executable, saved_packed_stat) = entry[1][0]
890
if (minikind == saved_minikind
891
and packed_stat == saved_packed_stat):
892
# The stat hasn't changed since we saved, so we can re-use the
897
# size should also be in packed_stat
898
if saved_file_size == stat_value.st_size:
899
return saved_link_or_sha1
901
# If we have gotten this far, that means that we need to actually
902
# process this entry.
906
executable = self._is_executable(stat_value.st_mode,
908
if self._cutoff_time is None:
909
self._sha_cutoff_time()
910
if (stat_value.st_mtime < self._cutoff_time
911
and stat_value.st_ctime < self._cutoff_time
912
and len(entry[1]) > 1
913
and entry[1][1][0] != 'a'):
914
# Could check for size changes for further optimised
915
# avoidance of sha1's. However the most prominent case of
916
# over-shaing is during initial add, which this catches.
917
link_or_sha1 = self._sha1_file(abspath)
918
entry[1][0] = ('f', link_or_sha1, stat_value.st_size,
919
executable, packed_stat)
921
# This file is not worth caching the sha1. Either it is too new, or
922
# it is newly added. Regardless, the only things we are changing
923
# are derived from the stat, and so are not worth caching. So we do
924
# *not* set the IN_MEMORY_MODIFIED flag. (But we'll save the
925
# updated values if there is *other* data worth saving.)
926
entry[1][0] = ('f', '', stat_value.st_size, executable,
929
elif minikind == c'd':
930
entry[1][0] = ('d', '', 0, False, packed_stat)
931
if saved_minikind != c'd':
932
# This changed from something into a directory. Make sure we
933
# have a directory block for it. This doesn't happen very
934
# often, so this doesn't have to be super fast.
935
block_index, entry_index, dir_present, file_present = \
936
self._get_block_entry_index(entry[0][0], entry[0][1], 0)
937
self._ensure_block(block_index, entry_index,
938
pathjoin(entry[0][0], entry[0][1]))
940
# Any changes are derived trivially from the stat object, not worth
941
# re-writing a dirstate for just this
943
elif minikind == c'l':
944
link_or_sha1 = self._read_link(abspath, saved_link_or_sha1)
945
if self._cutoff_time is None:
946
self._sha_cutoff_time()
947
if (stat_value.st_mtime < self._cutoff_time
948
and stat_value.st_ctime < self._cutoff_time):
949
entry[1][0] = ('l', link_or_sha1, stat_value.st_size,
952
entry[1][0] = ('l', '', stat_value.st_size,
953
False, DirState.NULLSTAT)
955
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
959
# TODO: Do we want to worry about exceptions here?
960
cdef char _minikind_from_string(object string) except? -1:
961
"""Convert a python string to a char."""
962
return PyString_AsString(string)[0]
965
cdef object _kind_absent
966
cdef object _kind_file
967
cdef object _kind_directory
968
cdef object _kind_symlink
969
cdef object _kind_relocated
970
cdef object _kind_tree_reference
971
_kind_absent = "absent"
973
_kind_directory = "directory"
974
_kind_symlink = "symlink"
975
_kind_relocated = "relocated"
976
_kind_tree_reference = "tree-reference"
979
cdef object _minikind_to_kind(char minikind):
980
"""Create a string kind for minikind."""
981
cdef char _minikind[1]
984
elif minikind == c'd':
985
return _kind_directory
986
elif minikind == c'a':
988
elif minikind == c'r':
989
return _kind_relocated
990
elif minikind == c'l':
992
elif minikind == c't':
993
return _kind_tree_reference
994
_minikind[0] = minikind
995
raise KeyError(PyString_FromStringAndSize(_minikind, 1))
998
cdef int _versioned_minikind(char minikind): # cannot_raise
999
"""Return non-zero if minikind is in fltd"""
1000
return (minikind == c'f' or
1006
cdef class ProcessEntryC:
1008
cdef int doing_consistency_expansion
1009
cdef object old_dirname_to_file_id # dict
1010
cdef object new_dirname_to_file_id # dict
1011
cdef object last_source_parent
1012
cdef object last_target_parent
1013
cdef int include_unchanged
1015
cdef object use_filesystem_for_exec
1016
cdef object utf8_decode
1017
cdef readonly object searched_specific_files
1018
cdef readonly object searched_exact_paths
1019
cdef object search_specific_files
1020
# The parents up to the root of the paths we are searching.
1021
# After all normal paths are returned, these specific items are returned.
1022
cdef object search_specific_file_parents
1024
# Current iteration variables:
1025
cdef object current_root
1026
cdef object current_root_unicode
1027
cdef object root_entries
1028
cdef int root_entries_pos, root_entries_len
1029
cdef object root_abspath
1030
cdef int source_index, target_index
1031
cdef int want_unversioned
1033
cdef object dir_iterator
1034
cdef int block_index
1035
cdef object current_block
1036
cdef int current_block_pos
1037
cdef object current_block_list
1038
cdef object current_dir_info
1039
cdef object current_dir_list
1040
cdef object _pending_consistent_entries # list
1042
cdef object root_dir_info
1043
cdef object bisect_left
1044
cdef object pathjoin
1046
# A set of the ids we've output when doing partial output.
1047
cdef object seen_ids
1048
cdef object sha_file
1050
def __init__(self, include_unchanged, use_filesystem_for_exec,
1051
search_specific_files, state, source_index, target_index,
1052
want_unversioned, tree):
1053
self.doing_consistency_expansion = 0
1054
self.old_dirname_to_file_id = {}
1055
self.new_dirname_to_file_id = {}
1056
# Are we doing a partial iter_changes?
1057
self.partial = set(['']).__ne__(search_specific_files)
1058
# Using a list so that we can access the values and change them in
1059
# nested scope. Each one is [path, file_id, entry]
1060
self.last_source_parent = [None, None]
1061
self.last_target_parent = [None, None]
1062
if include_unchanged is None:
1063
self.include_unchanged = False
1065
self.include_unchanged = int(include_unchanged)
1066
self.use_filesystem_for_exec = use_filesystem_for_exec
1067
self.utf8_decode = cache_utf8._utf8_decode
1068
# for all search_indexs in each path at or under each element of
1069
# search_specific_files, if the detail is relocated: add the id, and
1070
# add the relocated path as one to search if its not searched already.
1071
# If the detail is not relocated, add the id.
1072
self.searched_specific_files = set()
1073
# When we search exact paths without expanding downwards, we record
1075
self.searched_exact_paths = set()
1076
self.search_specific_files = search_specific_files
1077
# The parents up to the root of the paths we are searching.
1078
# After all normal paths are returned, these specific items are returned.
1079
self.search_specific_file_parents = set()
1080
# The ids we've sent out in the delta.
1081
self.seen_ids = set()
1083
self.current_root = None
1084
self.current_root_unicode = None
1085
self.root_entries = None
1086
self.root_entries_pos = 0
1087
self.root_entries_len = 0
1088
self.root_abspath = None
1089
if source_index is None:
1090
self.source_index = -1
1092
self.source_index = source_index
1093
self.target_index = target_index
1094
self.want_unversioned = want_unversioned
1096
self.dir_iterator = None
1097
self.block_index = -1
1098
self.current_block = None
1099
self.current_block_list = None
1100
self.current_block_pos = -1
1101
self.current_dir_info = None
1102
self.current_dir_list = None
1103
self._pending_consistent_entries = []
1105
self.root_dir_info = None
1106
self.bisect_left = bisect.bisect_left
1107
self.pathjoin = osutils.pathjoin
1108
self.fstat = os.fstat
1109
self.sha_file = osutils.sha_file
1110
if target_index != 0:
1111
# A lot of code in here depends on target_index == 0
1112
raise errors.BzrError('unsupported target index')
1114
cdef _process_entry(self, entry, path_info):
1115
"""Compare an entry and real disk to generate delta information.
1117
:param path_info: top_relpath, basename, kind, lstat, abspath for
1118
the path of entry. If None, then the path is considered absent in
1119
the target (Perhaps we should pass in a concrete entry for this ?)
1120
Basename is returned as a utf8 string because we expect this
1121
tuple will be ignored, and don't want to take the time to
1123
:return: (iter_changes_result, changed). If the entry has not been
1124
handled then changed is None. Otherwise it is False if no content
1125
or metadata changes have occured, and True if any content or
1126
metadata change has occurred. If self.include_unchanged is True then
1127
if changed is not None, iter_changes_result will always be a result
1128
tuple. Otherwise, iter_changes_result is None unless changed is
1131
cdef char target_minikind
1132
cdef char source_minikind
1134
cdef int content_change
1135
cdef object details_list
1137
details_list = entry[1]
1138
if -1 == self.source_index:
1139
source_details = DirState.NULL_PARENT_DETAILS
1141
source_details = details_list[self.source_index]
1142
target_details = details_list[self.target_index]
1143
target_minikind = _minikind_from_string(target_details[0])
1144
if path_info is not None and _versioned_minikind(target_minikind):
1145
if self.target_index != 0:
1146
raise AssertionError("Unsupported target index %d" %
1148
link_or_sha1 = _update_entry(self.state, entry, path_info[4], path_info[3])
1149
# The entry may have been modified by update_entry
1150
target_details = details_list[self.target_index]
1151
target_minikind = _minikind_from_string(target_details[0])
1154
# the rest of this function is 0.3 seconds on 50K paths, or
1155
# 0.000006 seconds per call.
1156
source_minikind = _minikind_from_string(source_details[0])
1157
if ((_versioned_minikind(source_minikind) or source_minikind == c'r')
1158
and _versioned_minikind(target_minikind)):
1159
# claimed content in both: diff
1160
# r | fdlt | | add source to search, add id path move and perform
1161
# | | | diff check on source-target
1162
# r | fdlt | a | dangling file that was present in the basis.
1164
if source_minikind != c'r':
1165
old_dirname = entry[0][0]
1166
old_basename = entry[0][1]
1167
old_path = path = None
1169
# add the source to the search path to find any children it
1170
# has. TODO ? : only add if it is a container ?
1171
if (not self.doing_consistency_expansion and
1172
not osutils.is_inside_any(self.searched_specific_files,
1173
source_details[1])):
1174
self.search_specific_files.add(source_details[1])
1175
# expanding from a user requested path, parent expansion
1176
# for delta consistency happens later.
1177
# generate the old path; this is needed for stating later
1179
old_path = source_details[1]
1180
old_dirname, old_basename = os.path.split(old_path)
1181
path = self.pathjoin(entry[0][0], entry[0][1])
1182
old_entry = self.state._get_entry(self.source_index,
1184
# update the source details variable to be the real
1186
if old_entry == (None, None):
1187
raise errors.CorruptDirstate(self.state._filename,
1188
"entry '%s/%s' is considered renamed from %r"
1189
" but source does not exist\n"
1190
"entry: %s" % (entry[0][0], entry[0][1], old_path, entry))
1191
source_details = old_entry[1][self.source_index]
1192
source_minikind = _minikind_from_string(source_details[0])
1193
if path_info is None:
1194
# the file is missing on disk, show as removed.
1199
# source and target are both versioned and disk file is present.
1200
target_kind = path_info[2]
1201
if target_kind == 'directory':
1203
old_path = path = self.pathjoin(old_dirname, old_basename)
1204
file_id = entry[0][2]
1205
self.new_dirname_to_file_id[path] = file_id
1206
if source_minikind != c'd':
1209
# directories have no fingerprint
1212
elif target_kind == 'file':
1213
if source_minikind != c'f':
1216
# Check the sha. We can't just rely on the size as
1217
# content filtering may mean differ sizes actually
1218
# map to the same content
1219
if link_or_sha1 is None:
1221
statvalue, link_or_sha1 = \
1222
self.state._sha1_provider.stat_and_sha1(
1224
self.state._observed_sha1(entry, link_or_sha1,
1226
content_change = (link_or_sha1 != source_details[1])
1227
# Target details is updated at update_entry time
1228
if self.use_filesystem_for_exec:
1229
# We don't need S_ISREG here, because we are sure
1230
# we are dealing with a file.
1231
target_exec = bool(S_IXUSR & path_info[3].st_mode)
1233
target_exec = target_details[3]
1234
elif target_kind == 'symlink':
1235
if source_minikind != c'l':
1238
content_change = (link_or_sha1 != source_details[1])
1240
elif target_kind == 'tree-reference':
1241
if source_minikind != c't':
1248
path = self.pathjoin(old_dirname, old_basename)
1249
raise errors.BadFileKindError(path, path_info[2])
1250
if source_minikind == c'd':
1252
old_path = path = self.pathjoin(old_dirname, old_basename)
1254
file_id = entry[0][2]
1255
self.old_dirname_to_file_id[old_path] = file_id
1256
# parent id is the entry for the path in the target tree
1257
if old_basename and old_dirname == self.last_source_parent[0]:
1258
# use a cached hit for non-root source entries.
1259
source_parent_id = self.last_source_parent[1]
1262
source_parent_id = self.old_dirname_to_file_id[old_dirname]
1264
source_parent_entry = self.state._get_entry(self.source_index,
1265
path_utf8=old_dirname)
1266
source_parent_id = source_parent_entry[0][2]
1267
if source_parent_id == entry[0][2]:
1268
# This is the root, so the parent is None
1269
source_parent_id = None
1271
self.last_source_parent[0] = old_dirname
1272
self.last_source_parent[1] = source_parent_id
1273
new_dirname = entry[0][0]
1274
if entry[0][1] and new_dirname == self.last_target_parent[0]:
1275
# use a cached hit for non-root target entries.
1276
target_parent_id = self.last_target_parent[1]
1279
target_parent_id = self.new_dirname_to_file_id[new_dirname]
1281
# TODO: We don't always need to do the lookup, because the
1282
# parent entry will be the same as the source entry.
1283
target_parent_entry = self.state._get_entry(self.target_index,
1284
path_utf8=new_dirname)
1285
if target_parent_entry == (None, None):
1286
raise AssertionError(
1287
"Could not find target parent in wt: %s\nparent of: %s"
1288
% (new_dirname, entry))
1289
target_parent_id = target_parent_entry[0][2]
1290
if target_parent_id == entry[0][2]:
1291
# This is the root, so the parent is None
1292
target_parent_id = None
1294
self.last_target_parent[0] = new_dirname
1295
self.last_target_parent[1] = target_parent_id
1297
source_exec = source_details[3]
1298
changed = (content_change
1299
or source_parent_id != target_parent_id
1300
or old_basename != entry[0][1]
1301
or source_exec != target_exec
1303
if not changed and not self.include_unchanged:
1306
if old_path is None:
1307
path = self.pathjoin(old_dirname, old_basename)
1309
old_path_u = self.utf8_decode(old_path)[0]
1312
old_path_u = self.utf8_decode(old_path)[0]
1313
if old_path == path:
1316
path_u = self.utf8_decode(path)[0]
1317
source_kind = _minikind_to_kind(source_minikind)
1318
return (entry[0][2],
1319
(old_path_u, path_u),
1322
(source_parent_id, target_parent_id),
1323
(self.utf8_decode(old_basename)[0], self.utf8_decode(entry[0][1])[0]),
1324
(source_kind, target_kind),
1325
(source_exec, target_exec)), changed
1326
elif source_minikind == c'a' and _versioned_minikind(target_minikind):
1327
# looks like a new file
1328
path = self.pathjoin(entry[0][0], entry[0][1])
1329
# parent id is the entry for the path in the target tree
1330
# TODO: these are the same for an entire directory: cache em.
1331
parent_entry = self.state._get_entry(self.target_index,
1332
path_utf8=entry[0][0])
1333
if parent_entry is None:
1334
raise errors.DirstateCorrupt(self.state,
1335
"We could not find the parent entry in index %d"
1336
" for the entry: %s"
1337
% (self.target_index, entry[0]))
1338
parent_id = parent_entry[0][2]
1339
if parent_id == entry[0][2]:
1341
if path_info is not None:
1343
if self.use_filesystem_for_exec:
1344
# We need S_ISREG here, because we aren't sure if this
1347
S_ISREG(path_info[3].st_mode)
1348
and S_IXUSR & path_info[3].st_mode)
1350
target_exec = target_details[3]
1351
return (entry[0][2],
1352
(None, self.utf8_decode(path)[0]),
1356
(None, self.utf8_decode(entry[0][1])[0]),
1357
(None, path_info[2]),
1358
(None, target_exec)), True
1360
# Its a missing file, report it as such.
1361
return (entry[0][2],
1362
(None, self.utf8_decode(path)[0]),
1366
(None, self.utf8_decode(entry[0][1])[0]),
1368
(None, False)), True
1369
elif _versioned_minikind(source_minikind) and target_minikind == c'a':
1370
# unversioned, possibly, or possibly not deleted: we dont care.
1371
# if its still on disk, *and* theres no other entry at this
1372
# path [we dont know this in this routine at the moment -
1373
# perhaps we should change this - then it would be an unknown.
1374
old_path = self.pathjoin(entry[0][0], entry[0][1])
1375
# parent id is the entry for the path in the target tree
1376
parent_id = self.state._get_entry(self.source_index, path_utf8=entry[0][0])[0][2]
1377
if parent_id == entry[0][2]:
1379
return (entry[0][2],
1380
(self.utf8_decode(old_path)[0], None),
1384
(self.utf8_decode(entry[0][1])[0], None),
1385
(_minikind_to_kind(source_minikind), None),
1386
(source_details[3], None)), True
1387
elif _versioned_minikind(source_minikind) and target_minikind == c'r':
1388
# a rename; could be a true rename, or a rename inherited from
1389
# a renamed parent. TODO: handle this efficiently. Its not
1390
# common case to rename dirs though, so a correct but slow
1391
# implementation will do.
1392
if (not self.doing_consistency_expansion and
1393
not osutils.is_inside_any(self.searched_specific_files,
1394
target_details[1])):
1395
self.search_specific_files.add(target_details[1])
1396
# We don't expand the specific files parents list here as
1397
# the path is absent in target and won't create a delta with
1399
elif ((source_minikind == c'r' or source_minikind == c'a') and
1400
(target_minikind == c'r' or target_minikind == c'a')):
1401
# neither of the selected trees contain this path,
1402
# so skip over it. This is not currently directly tested, but
1403
# is indirectly via test_too_much.TestCommands.test_conflicts.
1406
raise AssertionError("don't know how to compare "
1407
"source_minikind=%r, target_minikind=%r"
1408
% (source_minikind, target_minikind))
1409
## import pdb;pdb.set_trace()
1415
def iter_changes(self):
1418
cdef int _gather_result_for_consistency(self, result) except -1:
1419
"""Check a result we will yield to make sure we are consistent later.
1421
This gathers result's parents into a set to output later.
1423
:param result: A result tuple.
1425
if not self.partial or not result[0]:
1427
self.seen_ids.add(result[0])
1428
new_path = result[1][1]
1430
# Not the root and not a delete: queue up the parents of the path.
1431
self.search_specific_file_parents.update(
1432
osutils.parent_directories(new_path.encode('utf8')))
1433
# Add the root directory which parent_directories does not
1435
self.search_specific_file_parents.add('')
1438
cdef int _update_current_block(self) except -1:
1439
if (self.block_index < len(self.state._dirblocks) and
1440
osutils.is_inside(self.current_root, self.state._dirblocks[self.block_index][0])):
1441
self.current_block = self.state._dirblocks[self.block_index]
1442
self.current_block_list = self.current_block[1]
1443
self.current_block_pos = 0
1445
self.current_block = None
1446
self.current_block_list = None
1450
# Simple thunk to allow tail recursion without pyrex confusion
1451
return self._iter_next()
1453
cdef _iter_next(self):
1454
"""Iterate over the changes."""
1455
# This function single steps through an iterator. As such while loops
1456
# are often exited by 'return' - the code is structured so that the
1457
# next call into the function will return to the same while loop. Note
1458
# that all flow control needed to re-reach that step is reexecuted,
1459
# which can be a performance problem. It has not yet been tuned to
1460
# minimise this; a state machine is probably the simplest restructuring
1461
# to both minimise this overhead and make the code considerably more
1465
# compare source_index and target_index at or under each element of search_specific_files.
1466
# follow the following comparison table. Note that we only want to do diff operations when
1467
# the target is fdl because thats when the walkdirs logic will have exposed the pathinfo
1471
# Source | Target | disk | action
1472
# r | fdlt | | add source to search, add id path move and perform
1473
# | | | diff check on source-target
1474
# r | fdlt | a | dangling file that was present in the basis.
1476
# r | a | | add source to search
1478
# r | r | | this path is present in a non-examined tree, skip.
1479
# r | r | a | this path is present in a non-examined tree, skip.
1480
# a | fdlt | | add new id
1481
# a | fdlt | a | dangling locally added file, skip
1482
# a | a | | not present in either tree, skip
1483
# a | a | a | not present in any tree, skip
1484
# a | r | | not present in either tree at this path, skip as it
1485
# | | | may not be selected by the users list of paths.
1486
# a | r | a | not present in either tree at this path, skip as it
1487
# | | | may not be selected by the users list of paths.
1488
# fdlt | fdlt | | content in both: diff them
1489
# fdlt | fdlt | a | deleted locally, but not unversioned - show as deleted ?
1490
# fdlt | a | | unversioned: output deleted id for now
1491
# fdlt | a | a | unversioned and deleted: output deleted id
1492
# fdlt | r | | relocated in this tree, so add target to search.
1493
# | | | Dont diff, we will see an r,fd; pair when we reach
1494
# | | | this id at the other path.
1495
# fdlt | r | a | relocated in this tree, so add target to search.
1496
# | | | Dont diff, we will see an r,fd; pair when we reach
1497
# | | | this id at the other path.
1499
# TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
1500
# keeping a cache of directories that we have seen.
1501
cdef object current_dirname, current_blockname
1502
cdef char * current_dirname_c, * current_blockname_c
1503
cdef int advance_entry, advance_path
1504
cdef int path_handled
1505
searched_specific_files = self.searched_specific_files
1506
# Are we walking a root?
1507
while self.root_entries_pos < self.root_entries_len:
1508
entry = self.root_entries[self.root_entries_pos]
1509
self.root_entries_pos = self.root_entries_pos + 1
1510
result, changed = self._process_entry(entry, self.root_dir_info)
1511
if changed is not None:
1513
self._gather_result_for_consistency(result)
1514
if changed or self.include_unchanged:
1516
# Have we finished the prior root, or never started one ?
1517
if self.current_root is None:
1518
# TODO: the pending list should be lexically sorted? the
1519
# interface doesn't require it.
1521
self.current_root = self.search_specific_files.pop()
1523
raise StopIteration()
1524
self.searched_specific_files.add(self.current_root)
1525
# process the entries for this containing directory: the rest will be
1526
# found by their parents recursively.
1527
self.root_entries = self.state._entries_for_path(self.current_root)
1528
self.root_entries_len = len(self.root_entries)
1529
self.current_root_unicode = self.current_root.decode('utf8')
1530
self.root_abspath = self.tree.abspath(self.current_root_unicode)
1532
root_stat = os.lstat(self.root_abspath)
1534
if e.errno == errno.ENOENT:
1535
# the path does not exist: let _process_entry know that.
1536
self.root_dir_info = None
1538
# some other random error: hand it up.
1541
self.root_dir_info = ('', self.current_root,
1542
osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
1544
if self.root_dir_info[2] == 'directory':
1545
if self.tree._directory_is_tree_reference(
1546
self.current_root_unicode):
1547
self.root_dir_info = self.root_dir_info[:2] + \
1548
('tree-reference',) + self.root_dir_info[3:]
1549
if not self.root_entries and not self.root_dir_info:
1550
# this specified path is not present at all, skip it.
1551
# (tail recursion, can do a loop once the full structure is
1553
return self._iter_next()
1555
self.root_entries_pos = 0
1556
# XXX Clarity: This loop is duplicated a out the self.current_root
1557
# is None guard above: if we return from it, it completes there
1558
# (and the following if block cannot trigger because
1559
# path_handled must be true, so the if block is not # duplicated.
1560
while self.root_entries_pos < self.root_entries_len:
1561
entry = self.root_entries[self.root_entries_pos]
1562
self.root_entries_pos = self.root_entries_pos + 1
1563
result, changed = self._process_entry(entry, self.root_dir_info)
1564
if changed is not None:
1567
self._gather_result_for_consistency(result)
1568
if changed or self.include_unchanged:
1570
# handle unversioned specified paths:
1571
if self.want_unversioned and not path_handled and self.root_dir_info:
1572
new_executable = bool(
1573
stat.S_ISREG(self.root_dir_info[3].st_mode)
1574
and stat.S_IEXEC & self.root_dir_info[3].st_mode)
1576
(None, self.current_root_unicode),
1580
(None, splitpath(self.current_root_unicode)[-1]),
1581
(None, self.root_dir_info[2]),
1582
(None, new_executable)
1584
# If we reach here, the outer flow continues, which enters into the
1585
# per-root setup logic.
1586
if (self.current_dir_info is None and self.current_block is None and not
1587
self.doing_consistency_expansion):
1588
# setup iteration of this root:
1589
self.current_dir_list = None
1590
if self.root_dir_info and self.root_dir_info[2] == 'tree-reference':
1591
self.current_dir_info = None
1593
self.dir_iterator = osutils._walkdirs_utf8(self.root_abspath,
1594
prefix=self.current_root)
1597
self.current_dir_info = self.dir_iterator.next()
1598
self.current_dir_list = self.current_dir_info[1]
1600
# there may be directories in the inventory even though
1601
# this path is not a file on disk: so mark it as end of
1603
if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
1604
self.current_dir_info = None
1605
elif sys.platform == 'win32':
1606
# on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
1607
# python 2.5 has e.errno == EINVAL,
1608
# and e.winerror == ERROR_DIRECTORY
1610
e_winerror = e.winerror
1611
except AttributeError, _:
1613
win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
1614
if (e.errno in win_errors or e_winerror in win_errors):
1615
self.current_dir_info = None
1617
# Will this really raise the right exception ?
1622
if self.current_dir_info[0][0] == '':
1623
# remove .bzr from iteration
1624
bzr_index = self.bisect_left(self.current_dir_list, ('.bzr',))
1625
if self.current_dir_list[bzr_index][0] != '.bzr':
1626
raise AssertionError()
1627
del self.current_dir_list[bzr_index]
1628
initial_key = (self.current_root, '', '')
1629
self.block_index, _ = self.state._find_block_index_from_key(initial_key)
1630
if self.block_index == 0:
1631
# we have processed the total root already, but because the
1632
# initial key matched it we should skip it here.
1633
self.block_index = self.block_index + 1
1634
self._update_current_block()
1635
# walk until both the directory listing and the versioned metadata
1637
while (self.current_dir_info is not None
1638
or self.current_block is not None):
1639
# Uncommon case - a missing directory or an unversioned directory:
1640
if (self.current_dir_info and self.current_block
1641
and self.current_dir_info[0][0] != self.current_block[0]):
1642
# Work around pyrex broken heuristic - current_dirname has
1643
# the same scope as current_dirname_c
1644
current_dirname = self.current_dir_info[0][0]
1645
current_dirname_c = PyString_AS_STRING_void(
1646
<void *>current_dirname)
1647
current_blockname = self.current_block[0]
1648
current_blockname_c = PyString_AS_STRING_void(
1649
<void *>current_blockname)
1650
# In the python generator we evaluate this if block once per
1651
# dir+block; because we reenter in the pyrex version its being
1652
# evaluated once per path: we could cache the result before
1653
# doing the while loop and probably save time.
1654
if _cmp_by_dirs(current_dirname_c,
1655
PyString_Size(current_dirname),
1656
current_blockname_c,
1657
PyString_Size(current_blockname)) < 0:
1658
# filesystem data refers to paths not covered by the
1659
# dirblock. this has two possibilities:
1660
# A) it is versioned but empty, so there is no block for it
1661
# B) it is not versioned.
1663
# if (A) then we need to recurse into it to check for
1664
# new unknown files or directories.
1665
# if (B) then we should ignore it, because we don't
1666
# recurse into unknown directories.
1667
# We are doing a loop
1668
while self.path_index < len(self.current_dir_list):
1669
current_path_info = self.current_dir_list[self.path_index]
1670
# dont descend into this unversioned path if it is
1672
if current_path_info[2] in ('directory',
1674
del self.current_dir_list[self.path_index]
1675
self.path_index = self.path_index - 1
1676
self.path_index = self.path_index + 1
1677
if self.want_unversioned:
1678
if current_path_info[2] == 'directory':
1679
if self.tree._directory_is_tree_reference(
1680
self.utf8_decode(current_path_info[0])[0]):
1681
current_path_info = current_path_info[:2] + \
1682
('tree-reference',) + current_path_info[3:]
1683
new_executable = bool(
1684
stat.S_ISREG(current_path_info[3].st_mode)
1685
and stat.S_IEXEC & current_path_info[3].st_mode)
1687
(None, self.utf8_decode(current_path_info[0])[0]),
1691
(None, self.utf8_decode(current_path_info[1])[0]),
1692
(None, current_path_info[2]),
1693
(None, new_executable))
1694
# This dir info has been handled, go to the next
1696
self.current_dir_list = None
1698
self.current_dir_info = self.dir_iterator.next()
1699
self.current_dir_list = self.current_dir_info[1]
1700
except StopIteration, _:
1701
self.current_dir_info = None
1703
# We have a dirblock entry for this location, but there
1704
# is no filesystem path for this. This is most likely
1705
# because a directory was removed from the disk.
1706
# We don't have to report the missing directory,
1707
# because that should have already been handled, but we
1708
# need to handle all of the files that are contained
1710
while self.current_block_pos < len(self.current_block_list):
1711
current_entry = self.current_block_list[self.current_block_pos]
1712
self.current_block_pos = self.current_block_pos + 1
1713
# entry referring to file not present on disk.
1714
# advance the entry only, after processing.
1715
result, changed = self._process_entry(current_entry, None)
1716
if changed is not None:
1718
self._gather_result_for_consistency(result)
1719
if changed or self.include_unchanged:
1721
self.block_index = self.block_index + 1
1722
self._update_current_block()
1723
continue # next loop-on-block/dir
1724
result = self._loop_one_block()
1725
if result is not None:
1727
if len(self.search_specific_files):
1728
# More supplied paths to process
1729
self.current_root = None
1730
return self._iter_next()
1731
# Start expanding more conservatively, adding paths the user may not
1732
# have intended but required for consistent deltas.
1733
self.doing_consistency_expansion = 1
1734
if not self._pending_consistent_entries:
1735
self._pending_consistent_entries = self._next_consistent_entries()
1736
while self._pending_consistent_entries:
1737
result, changed = self._pending_consistent_entries.pop()
1738
if changed is not None:
1740
raise StopIteration()
1742
cdef object _maybe_tree_ref(self, current_path_info):
1743
if self.tree._directory_is_tree_reference(
1744
self.utf8_decode(current_path_info[0])[0]):
1745
return current_path_info[:2] + \
1746
('tree-reference',) + current_path_info[3:]
1748
return current_path_info
1750
cdef object _loop_one_block(self):
1751
# current_dir_info and current_block refer to the same directory -
1752
# this is the common case code.
1753
# Assign local variables for current path and entry:
1754
cdef object current_entry
1755
cdef object current_path_info
1756
cdef int path_handled
1759
# cdef char * temp_str
1760
# cdef Py_ssize_t temp_str_length
1761
# PyString_AsStringAndSize(disk_kind, &temp_str, &temp_str_length)
1762
# if not strncmp(temp_str, "directory", temp_str_length):
1763
if (self.current_block is not None and
1764
self.current_block_pos < PyList_GET_SIZE(self.current_block_list)):
1765
current_entry = PyList_GET_ITEM(self.current_block_list,
1766
self.current_block_pos)
1768
Py_INCREF(current_entry)
1770
current_entry = None
1771
if (self.current_dir_info is not None and
1772
self.path_index < PyList_GET_SIZE(self.current_dir_list)):
1773
current_path_info = PyList_GET_ITEM(self.current_dir_list,
1776
Py_INCREF(current_path_info)
1777
disk_kind = PyTuple_GET_ITEM(current_path_info, 2)
1779
Py_INCREF(disk_kind)
1780
if disk_kind == "directory":
1781
current_path_info = self._maybe_tree_ref(current_path_info)
1783
current_path_info = None
1784
while (current_entry is not None or current_path_info is not None):
1789
if current_entry is None:
1790
# unversioned - the check for path_handled when the path
1791
# is advanced will yield this path if needed.
1793
elif current_path_info is None:
1794
# no path is fine: the per entry code will handle it.
1795
result, changed = self._process_entry(current_entry,
1798
minikind = _minikind_from_string(
1799
current_entry[1][self.target_index][0])
1800
cmp_result = cmp(current_path_info[1], current_entry[0][1])
1801
if (cmp_result or minikind == c'a' or minikind == c'r'):
1802
# The current path on disk doesn't match the dirblock
1803
# record. Either the dirblock record is marked as
1804
# absent/renamed, or the file on disk is not present at all
1805
# in the dirblock. Either way, report about the dirblock
1806
# entry, and let other code handle the filesystem one.
1808
# Compare the basename for these files to determine
1811
# extra file on disk: pass for now, but only
1812
# increment the path, not the entry
1815
# entry referring to file not present on disk.
1816
# advance the entry only, after processing.
1817
result, changed = self._process_entry(current_entry,
1821
# paths are the same,and the dirstate entry is not
1822
# absent or renamed.
1823
result, changed = self._process_entry(current_entry,
1825
if changed is not None:
1827
if not changed and not self.include_unchanged:
1829
# >- loop control starts here:
1831
if advance_entry and current_entry is not None:
1832
self.current_block_pos = self.current_block_pos + 1
1833
if self.current_block_pos < PyList_GET_SIZE(self.current_block_list):
1834
current_entry = self.current_block_list[self.current_block_pos]
1836
current_entry = None
1838
if advance_path and current_path_info is not None:
1839
if not path_handled:
1840
# unversioned in all regards
1841
if self.want_unversioned:
1842
new_executable = bool(
1843
stat.S_ISREG(current_path_info[3].st_mode)
1844
and stat.S_IEXEC & current_path_info[3].st_mode)
1846
relpath_unicode = self.utf8_decode(current_path_info[0])[0]
1847
except UnicodeDecodeError, _:
1848
raise errors.BadFilenameEncoding(
1849
current_path_info[0], osutils._fs_enc)
1850
if changed is not None:
1851
raise AssertionError(
1852
"result is not None: %r" % result)
1854
(None, relpath_unicode),
1858
(None, self.utf8_decode(current_path_info[1])[0]),
1859
(None, current_path_info[2]),
1860
(None, new_executable))
1862
# dont descend into this unversioned path if it is
1864
if current_path_info[2] in ('directory'):
1865
del self.current_dir_list[self.path_index]
1866
self.path_index = self.path_index - 1
1867
# dont descend the disk iterator into any tree
1869
if current_path_info[2] == 'tree-reference':
1870
del self.current_dir_list[self.path_index]
1871
self.path_index = self.path_index - 1
1872
self.path_index = self.path_index + 1
1873
if self.path_index < len(self.current_dir_list):
1874
current_path_info = self.current_dir_list[self.path_index]
1875
if current_path_info[2] == 'directory':
1876
current_path_info = self._maybe_tree_ref(
1879
current_path_info = None
1880
if changed is not None:
1881
# Found a result on this pass, yield it
1883
self._gather_result_for_consistency(result)
1884
if changed or self.include_unchanged:
1886
if self.current_block is not None:
1887
self.block_index = self.block_index + 1
1888
self._update_current_block()
1889
if self.current_dir_info is not None:
1891
self.current_dir_list = None
1893
self.current_dir_info = self.dir_iterator.next()
1894
self.current_dir_list = self.current_dir_info[1]
1895
except StopIteration, _:
1896
self.current_dir_info = None
1898
cdef object _next_consistent_entries(self):
1899
"""Grabs the next specific file parent case to consider.
1901
:return: A list of the results, each of which is as for _process_entry.
1904
while self.search_specific_file_parents:
1905
# Process the parent directories for the paths we were iterating.
1906
# Even in extremely large trees this should be modest, so currently
1907
# no attempt is made to optimise.
1908
path_utf8 = self.search_specific_file_parents.pop()
1909
if path_utf8 in self.searched_exact_paths:
1910
# We've examined this path.
1912
if osutils.is_inside_any(self.searched_specific_files, path_utf8):
1913
# We've examined this path.
1915
path_entries = self.state._entries_for_path(path_utf8)
1916
# We need either one or two entries. If the path in
1917
# self.target_index has moved (so the entry in source_index is in
1918
# 'ar') then we need to also look for the entry for this path in
1919
# self.source_index, to output the appropriate delete-or-rename.
1920
selected_entries = []
1922
for candidate_entry in path_entries:
1923
# Find entries present in target at this path:
1924
if candidate_entry[1][self.target_index][0] not in 'ar':
1926
selected_entries.append(candidate_entry)
1927
# Find entries present in source at this path:
1928
elif (self.source_index is not None and
1929
candidate_entry[1][self.source_index][0] not in 'ar'):
1931
if candidate_entry[1][self.target_index][0] == 'a':
1932
# Deleted, emit it here.
1933
selected_entries.append(candidate_entry)
1935
# renamed, emit it when we process the directory it
1937
self.search_specific_file_parents.add(
1938
candidate_entry[1][self.target_index][1])
1940
raise AssertionError(
1941
"Missing entry for specific path parent %r, %r" % (
1942
path_utf8, path_entries))
1943
path_info = self._path_info(path_utf8, path_utf8.decode('utf8'))
1944
for entry in selected_entries:
1945
if entry[0][2] in self.seen_ids:
1947
result, changed = self._process_entry(entry, path_info)
1949
raise AssertionError(
1950
"Got entry<->path mismatch for specific path "
1951
"%r entry %r path_info %r " % (
1952
path_utf8, entry, path_info))
1953
# Only include changes - we're outside the users requested
1956
self._gather_result_for_consistency(result)
1957
if (result[6][0] == 'directory' and
1958
result[6][1] != 'directory'):
1959
# This stopped being a directory, the old children have
1961
if entry[1][self.source_index][0] == 'r':
1962
# renamed, take the source path
1963
entry_path_utf8 = entry[1][self.source_index][1]
1965
entry_path_utf8 = path_utf8
1966
initial_key = (entry_path_utf8, '', '')
1967
block_index, _ = self.state._find_block_index_from_key(
1969
if block_index == 0:
1970
# The children of the root are in block index 1.
1971
block_index = block_index + 1
1972
current_block = None
1973
if block_index < len(self.state._dirblocks):
1974
current_block = self.state._dirblocks[block_index]
1975
if not osutils.is_inside(
1976
entry_path_utf8, current_block[0]):
1977
# No entries for this directory at all.
1978
current_block = None
1979
if current_block is not None:
1980
for entry in current_block[1]:
1981
if entry[1][self.source_index][0] in 'ar':
1982
# Not in the source tree, so doesn't have to be
1985
# Path of the entry itself.
1986
self.search_specific_file_parents.add(
1987
self.pathjoin(*entry[0][:2]))
1988
if changed or self.include_unchanged:
1989
results.append((result, changed))
1990
self.searched_exact_paths.add(path_utf8)
1993
cdef object _path_info(self, utf8_path, unicode_path):
1994
"""Generate path_info for unicode_path.
1996
:return: None if unicode_path does not exist, or a path_info tuple.
1998
abspath = self.tree.abspath(unicode_path)
2000
stat = os.lstat(abspath)
2002
if e.errno == errno.ENOENT:
2003
# the path does not exist.
2007
utf8_basename = utf8_path.rsplit('/', 1)[-1]
2008
dir_info = (utf8_path, utf8_basename,
2009
osutils.file_kind_from_stat_mode(stat.st_mode), stat,
2011
if dir_info[2] == 'directory':
2012
if self.tree._directory_is_tree_reference(
2014
self.root_dir_info = self.root_dir_info[:2] + \
2015
('tree-reference',) + self.root_dir_info[3:]
added
removed
bzrlib/_dirstate_helpers_pyx.pyx
