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)
100
unsigned long PyInt_AsUnsignedLongMask(object number) except? -1
102
char *PyString_AsString(object p)
103
char *PyString_AsString_obj "PyString_AsString" (PyObject *string)
104
char *PyString_AS_STRING_void "PyString_AS_STRING" (void *p)
105
int PyString_AsStringAndSize(object str, char **buffer, Py_ssize_t *length) except -1
106
object PyString_FromString(char *)
107
object PyString_FromStringAndSize(char *, Py_ssize_t)
108
int PyString_Size(object p)
109
int PyString_GET_SIZE_void "PyString_GET_SIZE" (void *p)
110
int PyString_CheckExact(object p)
111
void Py_INCREF(object o)
112
void Py_DECREF(object o)
115
cdef extern from "string.h":
116
int strncmp(char *s1, char *s2, int len)
117
void *memchr(void *s, int c, size_t len)
118
int memcmp(void *b1, void *b2, size_t len)
119
# ??? memrchr is a GNU extension :(
120
# void *memrchr(void *s, int c, size_t len)
122
# cimport all of the definitions we will need to access
123
from _static_tuple_c cimport import_static_tuple_c, StaticTuple, \
124
StaticTuple_New, StaticTuple_SET_ITEM
126
import_static_tuple_c()
128
cdef void* _my_memrchr(void *s, int c, size_t n): # cannot_raise
129
# memrchr seems to be a GNU extension, so we have to implement it ourselves
142
def _py_memrchr(s, c):
143
"""Just to expose _my_memrchr for testing.
145
:param s: The Python string to search
146
:param c: The character to search for
147
:return: The offset to the last instance of 'c' in s
154
_s = PyString_AsString(s)
155
length = PyString_Size(s)
157
_c = PyString_AsString(c)
158
assert PyString_Size(c) == 1,\
159
'Must be a single character string, not %s' % (c,)
160
found = _my_memrchr(_s, _c[0], length)
163
return <char*>found - <char*>_s
166
cdef object safe_string_from_size(char *s, Py_ssize_t size):
168
raise AssertionError(
169
'tried to create a string with an invalid size: %d'
171
return PyString_FromStringAndSize(s, size)
174
cdef int _is_aligned(void *ptr): # cannot_raise
175
"""Is this pointer aligned to an integer size offset?
177
:return: 1 if this pointer is aligned, 0 otherwise.
179
return ((<intptr_t>ptr) & ((sizeof(int))-1)) == 0
182
cdef int _cmp_by_dirs(char *path1, int size1, char *path2, int size2): # cannot_raise
183
cdef unsigned char *cur1
184
cdef unsigned char *cur2
185
cdef unsigned char *end1
186
cdef unsigned char *end2
192
if path1 == path2 and size1 == size2:
195
end1 = <unsigned char*>path1+size1
196
end2 = <unsigned char*>path2+size2
198
# Use 32-bit comparisons for the matching portion of the string.
199
# Almost all CPU's are faster at loading and comparing 32-bit integers,
200
# than they are at 8-bit integers.
201
# 99% of the time, these will be aligned, but in case they aren't just skip
203
if _is_aligned(path1) and _is_aligned(path2):
204
cur_int1 = <int*>path1
205
cur_int2 = <int*>path2
206
end_int1 = <int*>(path1 + size1 - (size1 % sizeof(int)))
207
end_int2 = <int*>(path2 + size2 - (size2 % sizeof(int)))
209
while cur_int1 < end_int1 and cur_int2 < end_int2:
210
if cur_int1[0] != cur_int2[0]:
212
cur_int1 = cur_int1 + 1
213
cur_int2 = cur_int2 + 1
215
cur1 = <unsigned char*>cur_int1
216
cur2 = <unsigned char*>cur_int2
218
cur1 = <unsigned char*>path1
219
cur2 = <unsigned char*>path2
221
while cur1 < end1 and cur2 < end2:
222
if cur1[0] == cur2[0]:
223
# This character matches, just go to the next one
227
# The current characters do not match
229
return -1 # Reached the end of path1 segment first
230
elif cur2[0] == c'/':
231
return 1 # Reached the end of path2 segment first
232
elif cur1[0] < cur2[0]:
237
# We reached the end of at least one of the strings
239
return 1 # Not at the end of cur1, must be at the end of cur2
241
return -1 # At the end of cur1, but not at cur2
242
# We reached the end of both strings
246
def cmp_by_dirs(path1, path2):
247
"""Compare two paths directory by directory.
249
This is equivalent to doing::
251
cmp(path1.split('/'), path2.split('/'))
253
The idea is that you should compare path components separately. This
254
differs from plain ``cmp(path1, path2)`` for paths like ``'a-b'`` and
255
``a/b``. "a-b" comes after "a" but would come before "a/b" lexically.
257
:param path1: first path
258
:param path2: second path
259
:return: negative number if ``path1`` comes first,
260
0 if paths are equal,
261
and positive number if ``path2`` sorts first
263
if not PyString_CheckExact(path1):
264
raise TypeError("'path1' must be a plain string, not %s: %r"
265
% (type(path1), path1))
266
if not PyString_CheckExact(path2):
267
raise TypeError("'path2' must be a plain string, not %s: %r"
268
% (type(path2), path2))
269
return _cmp_by_dirs(PyString_AsString(path1),
270
PyString_Size(path1),
271
PyString_AsString(path2),
272
PyString_Size(path2))
275
def _cmp_path_by_dirblock(path1, path2):
276
"""Compare two paths based on what directory they are in.
278
This generates a sort order, such that all children of a directory are
279
sorted together, and grandchildren are in the same order as the
280
children appear. But all grandchildren come after all children.
282
In other words, all entries in a directory are sorted together, and
283
directorys are sorted in cmp_by_dirs order.
285
:param path1: first path
286
:param path2: the second path
287
:return: negative number if ``path1`` comes first,
289
and a positive number if ``path2`` sorts first
291
if not PyString_CheckExact(path1):
292
raise TypeError("'path1' must be a plain string, not %s: %r"
293
% (type(path1), path1))
294
if not PyString_CheckExact(path2):
295
raise TypeError("'path2' must be a plain string, not %s: %r"
296
% (type(path2), path2))
297
return _cmp_path_by_dirblock_intern(PyString_AsString(path1),
298
PyString_Size(path1),
299
PyString_AsString(path2),
300
PyString_Size(path2))
303
cdef int _cmp_path_by_dirblock_intern(char *path1, int path1_len,
304
char *path2, int path2_len): # cannot_raise
305
"""Compare two paths by what directory they are in.
307
see ``_cmp_path_by_dirblock`` for details.
310
cdef int dirname1_len
312
cdef int dirname2_len
314
cdef int basename1_len
316
cdef int basename2_len
320
if path1_len == 0 and path2_len == 0:
323
if path1 == path2 and path1_len == path2_len:
332
basename1 = <char*>_my_memrchr(path1, c'/', path1_len)
334
if basename1 == NULL:
336
basename1_len = path1_len
341
dirname1_len = basename1 - path1
342
basename1 = basename1 + 1
343
basename1_len = path1_len - dirname1_len - 1
345
basename2 = <char*>_my_memrchr(path2, c'/', path2_len)
347
if basename2 == NULL:
349
basename2_len = path2_len
354
dirname2_len = basename2 - path2
355
basename2 = basename2 + 1
356
basename2_len = path2_len - dirname2_len - 1
358
cmp_val = _cmp_by_dirs(dirname1, dirname1_len,
359
dirname2, dirname2_len)
363
cur_len = basename1_len
364
if basename2_len < basename1_len:
365
cur_len = basename2_len
367
cmp_val = memcmp(basename1, basename2, cur_len)
370
if basename1_len == basename2_len:
372
if basename1_len < basename2_len:
377
def _bisect_path_left(paths, path):
378
"""Return the index where to insert path into paths.
380
This uses a path-wise comparison so we get::
390
:param paths: A list of paths to search through
391
:param path: A single path to insert
392
:return: An offset where 'path' can be inserted.
393
:seealso: bisect.bisect_left
404
if not PyList_CheckExact(paths):
405
raise TypeError("you must pass a python list for 'paths' not: %s %r"
406
% (type(paths), paths))
407
if not PyString_CheckExact(path):
408
raise TypeError("you must pass a string for 'path' not: %s %r"
409
% (type(path), path))
414
path_cstr = PyString_AsString(path)
415
path_size = PyString_Size(path)
418
_mid = (_lo + _hi) / 2
419
cur = PyList_GetItem_object_void(paths, _mid)
420
cur_cstr = PyString_AS_STRING_void(cur)
421
cur_size = PyString_GET_SIZE_void(cur)
422
if _cmp_path_by_dirblock_intern(cur_cstr, cur_size,
423
path_cstr, path_size) < 0:
430
def _bisect_path_right(paths, path):
431
"""Return the index where to insert path into paths.
433
This uses a path-wise comparison so we get::
443
:param paths: A list of paths to search through
444
:param path: A single path to insert
445
:return: An offset where 'path' can be inserted.
446
:seealso: bisect.bisect_right
457
if not PyList_CheckExact(paths):
458
raise TypeError("you must pass a python list for 'paths' not: %s %r"
459
% (type(paths), paths))
460
if not PyString_CheckExact(path):
461
raise TypeError("you must pass a string for 'path' not: %s %r"
462
% (type(path), path))
467
path_cstr = PyString_AsString(path)
468
path_size = PyString_Size(path)
471
_mid = (_lo + _hi) / 2
472
cur = PyList_GetItem_object_void(paths, _mid)
473
cur_cstr = PyString_AS_STRING_void(cur)
474
cur_size = PyString_GET_SIZE_void(cur)
475
if _cmp_path_by_dirblock_intern(path_cstr, path_size,
476
cur_cstr, cur_size) < 0:
483
def bisect_dirblock(dirblocks, dirname, lo=0, hi=None, cache=None):
484
"""Return the index where to insert dirname into the dirblocks.
486
The return value idx is such that all directories blocks in dirblock[:idx]
487
have names < dirname, and all blocks in dirblock[idx:] have names >=
490
Optional args lo (default 0) and hi (default len(dirblocks)) bound the
491
slice of a to be searched.
496
cdef char *dirname_cstr
497
cdef int dirname_size
502
if not PyList_CheckExact(dirblocks):
503
raise TypeError("you must pass a python list for 'dirblocks' not: %s %r"
504
% (type(dirblocks), dirblocks))
505
if not PyString_CheckExact(dirname):
506
raise TypeError("you must pass a string for dirname not: %s %r"
507
% (type(dirname), dirname))
514
dirname_cstr = PyString_AsString(dirname)
515
dirname_size = PyString_Size(dirname)
518
_mid = (_lo + _hi) / 2
519
# Grab the dirname for the current dirblock
520
# cur = dirblocks[_mid][0]
521
cur = PyTuple_GetItem_void_void(
522
PyList_GetItem_object_void(dirblocks, _mid), 0)
523
cur_cstr = PyString_AS_STRING_void(cur)
524
cur_size = PyString_GET_SIZE_void(cur)
525
if _cmp_by_dirs(cur_cstr, cur_size, dirname_cstr, dirname_size) < 0:
533
"""Maintain the current location, and return fields as you parse them."""
535
cdef object state # The DirState object
536
cdef object text # The overall string object
537
cdef char *text_cstr # Pointer to the beginning of text
538
cdef int text_size # Length of text
540
cdef char *end_cstr # End of text
541
cdef char *cur_cstr # Pointer to the current record
542
cdef char *next # Pointer to the end of this record
544
def __init__(self, text, state):
547
self.text_cstr = PyString_AsString(text)
548
self.text_size = PyString_Size(text)
549
self.end_cstr = self.text_cstr + self.text_size
550
self.cur_cstr = self.text_cstr
552
cdef char *get_next(self, int *size) except NULL:
553
"""Return a pointer to the start of the next field."""
555
cdef Py_ssize_t extra_len
557
if self.cur_cstr == NULL:
558
raise AssertionError('get_next() called when cur_str is NULL')
559
elif self.cur_cstr >= self.end_cstr:
560
raise AssertionError('get_next() called when there are no chars'
563
self.cur_cstr = <char*>memchr(next, c'\0', self.end_cstr - next)
564
if self.cur_cstr == NULL:
565
extra_len = self.end_cstr - next
566
raise errors.DirstateCorrupt(self.state,
567
'failed to find trailing NULL (\\0).'
568
' Trailing garbage: %r'
569
% safe_string_from_size(next, extra_len))
570
size[0] = self.cur_cstr - next
571
self.cur_cstr = self.cur_cstr + 1
574
cdef object get_next_str(self):
575
"""Get the next field as a Python string."""
578
next = self.get_next(&size)
579
return safe_string_from_size(next, size)
581
cdef int _init(self) except -1:
582
"""Get the pointer ready.
584
This assumes that the dirstate header has already been read, and we
585
already have the dirblock string loaded into memory.
586
This just initializes our memory pointers, etc for parsing of the
591
# The first field should be an empty string left over from the Header
592
first = self.get_next(&size)
593
if first[0] != c'\0' and size == 0:
594
raise AssertionError('First character should be null not: %s'
598
cdef object _get_entry(self, int num_trees, void **p_current_dirname,
600
"""Extract the next entry.
602
This parses the next entry based on the current location in
604
Each entry can be considered a "row" in the total table. And each row
605
has a fixed number of columns. It is generally broken up into "key"
606
columns, then "current" columns, and then "parent" columns.
608
:param num_trees: How many parent trees need to be parsed
609
:param p_current_dirname: A pointer to the current PyString
610
representing the directory name.
611
We pass this in as a void * so that pyrex doesn't have to
612
increment/decrement the PyObject reference counter for each
614
We use a pointer so that _get_entry can update it with the new
616
:param new_block: This is to let the caller know that it needs to
617
create a new directory block to store the next entry.
619
cdef StaticTuple path_name_file_id_key
621
cdef char *entry_size_cstr
622
cdef unsigned long int entry_size
623
cdef char* executable_cstr
624
cdef int is_executable
625
cdef char* dirname_cstr
630
cdef object fingerprint
633
# Read the 'key' information (dirname, name, file_id)
634
dirname_cstr = self.get_next(&cur_size)
635
# Check to see if we have started a new directory block.
636
# If so, then we need to create a new dirname PyString, so that it can
637
# be used in all of the tuples. This saves time and memory, by re-using
638
# the same object repeatedly.
640
# Do the cheap 'length of string' check first. If the string is a
641
# different length, then we *have* to be a different directory.
642
if (cur_size != PyString_GET_SIZE_void(p_current_dirname[0])
643
or strncmp(dirname_cstr,
644
# Extract the char* from our current dirname string. We
645
# know it is a PyString, so we can use
646
# PyString_AS_STRING, we use the _void version because
647
# we are tricking Pyrex by using a void* rather than an
649
PyString_AS_STRING_void(p_current_dirname[0]),
651
dirname = safe_string_from_size(dirname_cstr, cur_size)
652
p_current_dirname[0] = <void*>dirname
657
# Build up the key that will be used.
658
# By using <object>(void *) Pyrex will automatically handle the
659
# Py_INCREF that we need.
660
cur_dirname = <object>p_current_dirname[0]
661
# Use StaticTuple_New to pre-allocate, rather than creating a regular
662
# tuple and passing it to the StaticTuple constructor.
663
# path_name_file_id_key = StaticTuple(<object>p_current_dirname[0],
664
# self.get_next_str(),
665
# self.get_next_str(),
667
tmp = StaticTuple_New(3)
668
Py_INCREF(cur_dirname); StaticTuple_SET_ITEM(tmp, 0, cur_dirname)
669
cur_basename = self.get_next_str()
670
cur_file_id = self.get_next_str()
671
Py_INCREF(cur_basename); StaticTuple_SET_ITEM(tmp, 1, cur_basename)
672
Py_INCREF(cur_file_id); StaticTuple_SET_ITEM(tmp, 2, cur_file_id)
673
path_name_file_id_key = tmp
675
# Parse all of the per-tree information. current has the information in
676
# the same location as parent trees. The only difference is that 'info'
677
# is a 'packed_stat' for current, while it is a 'revision_id' for
679
# minikind, fingerprint, and info will be returned as regular python
681
# entry_size and is_executable will be parsed into a python Long and
682
# python Boolean, respectively.
683
# TODO: jam 20070718 Consider changin the entry_size conversion to
684
# prefer python Int when possible. They are generally faster to
685
# work with, and it will be rare that we have a file >2GB.
686
# Especially since this code is pretty much fixed at a max of
689
for i from 0 <= i < num_trees:
690
minikind = self.get_next_str()
691
fingerprint = self.get_next_str()
692
entry_size_cstr = self.get_next(&cur_size)
693
entry_size = strtoul(entry_size_cstr, NULL, 10)
694
executable_cstr = self.get_next(&cur_size)
695
is_executable = (executable_cstr[0] == c'y')
696
info = self.get_next_str()
697
# TODO: If we want to use StaticTuple_New here we need to be pretty
698
# careful. We are relying on a bit of Pyrex
699
# automatic-conversion from 'int' to PyInt, and that doesn't
700
# play well with the StaticTuple_SET_ITEM macro.
701
# Timing doesn't (yet) show a worthwile improvement in speed
702
# versus complexity and maintainability.
703
# tmp = StaticTuple_New(5)
704
# Py_INCREF(minikind); StaticTuple_SET_ITEM(tmp, 0, minikind)
705
# Py_INCREF(fingerprint); StaticTuple_SET_ITEM(tmp, 1, fingerprint)
706
# Py_INCREF(entry_size); StaticTuple_SET_ITEM(tmp, 2, entry_size)
707
# Py_INCREF(is_executable); StaticTuple_SET_ITEM(tmp, 3, is_executable)
708
# Py_INCREF(info); StaticTuple_SET_ITEM(tmp, 4, info)
709
# PyList_Append(trees, tmp)
710
PyList_Append(trees, StaticTuple(
712
fingerprint, # fingerprint
714
is_executable,# executable
715
info, # packed_stat or revision_id
718
# The returned tuple is (key, [trees])
719
ret = (path_name_file_id_key, trees)
720
# Ignore the trailing newline, but assert that it does exist, this
721
# ensures that we always finish parsing a line on an end-of-entry
723
trailing = self.get_next(&cur_size)
724
if cur_size != 1 or trailing[0] != c'\n':
725
raise errors.DirstateCorrupt(self.state,
726
'Bad parse, we expected to end on \\n, not: %d %s: %s'
727
% (cur_size, safe_string_from_size(trailing, cur_size),
731
def _parse_dirblocks(self):
732
"""Parse all dirblocks in the state file."""
734
cdef object current_block
736
cdef void * current_dirname
738
cdef int expected_entry_count
741
num_trees = self.state._num_present_parents() + 1
742
expected_entry_count = self.state._num_entries
744
# Ignore the first record
748
dirblocks = [('', current_block), ('', [])]
749
self.state._dirblocks = dirblocks
751
current_dirname = <void*>obj
755
# TODO: jam 2007-05-07 Consider pre-allocating some space for the
756
# members, and then growing and shrinking from there. If most
757
# directories have close to 10 entries in them, it would save a
758
# few mallocs if we default our list size to something
759
# reasonable. Or we could malloc it to something large (100 or
760
# so), and then truncate. That would give us a malloc + realloc,
761
# rather than lots of reallocs.
762
while self.cur_cstr < self.end_cstr:
763
entry = self._get_entry(num_trees, ¤t_dirname, &new_block)
765
# new block - different dirname
767
PyList_Append(dirblocks,
768
(<object>current_dirname, current_block))
769
PyList_Append(current_block, entry)
770
entry_count = entry_count + 1
771
if entry_count != expected_entry_count:
772
raise errors.DirstateCorrupt(self.state,
773
'We read the wrong number of entries.'
774
' We expected to read %s, but read %s'
775
% (expected_entry_count, entry_count))
776
self.state._split_root_dirblock_into_contents()
779
def _read_dirblocks(state):
780
"""Read in the dirblocks for the given DirState object.
782
This is tightly bound to the DirState internal representation. It should be
783
thought of as a member function, which is only separated out so that we can
784
re-write it in pyrex.
786
:param state: A DirState object.
788
:postcondition: The dirblocks will be loaded into the appropriate fields in
791
state._state_file.seek(state._end_of_header)
792
text = state._state_file.read()
793
# TODO: check the crc checksums. crc_measured = zlib.crc32(text)
795
reader = Reader(text, state)
797
reader._parse_dirblocks()
798
state._dirblock_state = DirState.IN_MEMORY_UNMODIFIED
801
cdef int minikind_from_mode(int mode): # cannot_raise
802
# in order of frequency:
812
_encode = binascii.b2a_base64
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(PyInt_AsUnsignedLongMask(stat_value.st_size))
825
# mtime and ctime will often be floats but get converted to PyInt within
826
aliased[1] = htonl(PyInt_AsUnsignedLongMask(stat_value.st_mtime))
827
aliased[2] = htonl(PyInt_AsUnsignedLongMask(stat_value.st_ctime))
828
aliased[3] = htonl(PyInt_AsUnsignedLongMask(stat_value.st_dev))
829
aliased[4] = htonl(PyInt_AsUnsignedLongMask(stat_value.st_ino))
830
aliased[5] = htonl(PyInt_AsUnsignedLongMask(stat_value.st_mode))
831
packed = PyString_FromStringAndSize(result, 6*4)
832
return _encode(packed)[:-1]
835
def pack_stat(stat_value):
836
"""Convert stat value into a packed representation quickly with pyrex"""
837
return _pack_stat(stat_value)
840
def update_entry(self, entry, abspath, stat_value):
841
"""Update the entry based on what is actually on disk.
843
This function only calculates the sha if it needs to - if the entry is
844
uncachable, or clearly different to the first parent's entry, no sha
845
is calculated, and None is returned.
847
:param entry: This is the dirblock entry for the file in question.
848
:param abspath: The path on disk for this file.
849
:param stat_value: (optional) if we already have done a stat on the
851
:return: None, or The sha1 hexdigest of the file (40 bytes) or link
854
return _update_entry(self, entry, abspath, stat_value)
857
cdef _update_entry(self, entry, abspath, stat_value):
858
"""Update the entry based on what is actually on disk.
860
This function only calculates the sha if it needs to - if the entry is
861
uncachable, or clearly different to the first parent's entry, no sha
862
is calculated, and None is returned.
864
:param self: The dirstate object this is operating on.
865
:param entry: This is the dirblock entry for the file in question.
866
:param abspath: The path on disk for this file.
867
:param stat_value: The stat value done on the path.
868
:return: None, or The sha1 hexdigest of the file (40 bytes) or link
871
# TODO - require pyrex 0.9.8, then use a pyd file to define access to the
872
# _st mode of the compiled stat objects.
873
cdef int minikind, saved_minikind
875
cdef int worth_saving
876
minikind = minikind_from_mode(stat_value.st_mode)
879
packed_stat = _pack_stat(stat_value)
880
details = PyList_GetItem_void_void(PyTuple_GetItem_void_void(<void *>entry, 1), 0)
881
saved_minikind = PyString_AsString_obj(<PyObject *>PyTuple_GetItem_void_void(details, 0))[0]
882
if minikind == c'd' and saved_minikind == c't':
884
saved_link_or_sha1 = PyTuple_GetItem_void_object(details, 1)
885
saved_file_size = PyTuple_GetItem_void_object(details, 2)
886
saved_executable = PyTuple_GetItem_void_object(details, 3)
887
saved_packed_stat = PyTuple_GetItem_void_object(details, 4)
888
# Deal with pyrex decrefing the objects
889
Py_INCREF(saved_link_or_sha1)
890
Py_INCREF(saved_file_size)
891
Py_INCREF(saved_executable)
892
Py_INCREF(saved_packed_stat)
893
#(saved_minikind, saved_link_or_sha1, saved_file_size,
894
# saved_executable, saved_packed_stat) = entry[1][0]
896
if (minikind == saved_minikind
897
and packed_stat == saved_packed_stat):
898
# The stat hasn't changed since we saved, so we can re-use the
903
# size should also be in packed_stat
904
if saved_file_size == stat_value.st_size:
905
return saved_link_or_sha1
907
# If we have gotten this far, that means that we need to actually
908
# process this entry.
912
executable = self._is_executable(stat_value.st_mode,
914
if self._cutoff_time is None:
915
self._sha_cutoff_time()
916
if (stat_value.st_mtime < self._cutoff_time
917
and stat_value.st_ctime < self._cutoff_time
918
and len(entry[1]) > 1
919
and entry[1][1][0] != 'a'):
920
# Could check for size changes for further optimised
921
# avoidance of sha1's. However the most prominent case of
922
# over-shaing is during initial add, which this catches.
923
link_or_sha1 = self._sha1_file(abspath)
924
entry[1][0] = ('f', link_or_sha1, stat_value.st_size,
925
executable, packed_stat)
927
# This file is not worth caching the sha1. Either it is too new, or
928
# it is newly added. Regardless, the only things we are changing
929
# are derived from the stat, and so are not worth caching. So we do
930
# *not* set the IN_MEMORY_MODIFIED flag. (But we'll save the
931
# updated values if there is *other* data worth saving.)
932
entry[1][0] = ('f', '', stat_value.st_size, executable,
935
elif minikind == c'd':
936
entry[1][0] = ('d', '', 0, False, packed_stat)
937
if saved_minikind != c'd':
938
# This changed from something into a directory. Make sure we
939
# have a directory block for it. This doesn't happen very
940
# often, so this doesn't have to be super fast.
941
block_index, entry_index, dir_present, file_present = \
942
self._get_block_entry_index(entry[0][0], entry[0][1], 0)
943
self._ensure_block(block_index, entry_index,
944
pathjoin(entry[0][0], entry[0][1]))
946
# Any changes are derived trivially from the stat object, not worth
947
# re-writing a dirstate for just this
949
elif minikind == c'l':
950
if saved_minikind == c'l':
951
# If the object hasn't changed kind, it isn't worth saving the
952
# dirstate just for a symlink. The default is 'fast symlinks' which
953
# save the target in the inode entry, rather than separately. So to
954
# stat, we've already read everything off disk.
956
link_or_sha1 = self._read_link(abspath, saved_link_or_sha1)
957
if self._cutoff_time is None:
958
self._sha_cutoff_time()
959
if (stat_value.st_mtime < self._cutoff_time
960
and stat_value.st_ctime < self._cutoff_time):
961
entry[1][0] = ('l', link_or_sha1, stat_value.st_size,
964
entry[1][0] = ('l', '', stat_value.st_size,
965
False, DirState.NULLSTAT)
967
# Note, even though _mark_modified will only set
968
# IN_MEMORY_HASH_MODIFIED, it still isn't worth
969
self._mark_modified([entry])
973
# TODO: Do we want to worry about exceptions here?
974
cdef char _minikind_from_string(object string) except? -1:
975
"""Convert a python string to a char."""
976
return PyString_AsString(string)[0]
979
cdef object _kind_absent
980
cdef object _kind_file
981
cdef object _kind_directory
982
cdef object _kind_symlink
983
cdef object _kind_relocated
984
cdef object _kind_tree_reference
985
_kind_absent = "absent"
987
_kind_directory = "directory"
988
_kind_symlink = "symlink"
989
_kind_relocated = "relocated"
990
_kind_tree_reference = "tree-reference"
993
cdef object _minikind_to_kind(char minikind):
994
"""Create a string kind for minikind."""
995
cdef char _minikind[1]
998
elif minikind == c'd':
999
return _kind_directory
1000
elif minikind == c'a':
1002
elif minikind == c'r':
1003
return _kind_relocated
1004
elif minikind == c'l':
1005
return _kind_symlink
1006
elif minikind == c't':
1007
return _kind_tree_reference
1008
_minikind[0] = minikind
1009
raise KeyError(PyString_FromStringAndSize(_minikind, 1))
1012
cdef int _versioned_minikind(char minikind): # cannot_raise
1013
"""Return non-zero if minikind is in fltd"""
1014
return (minikind == c'f' or
1020
cdef class ProcessEntryC:
1022
cdef int doing_consistency_expansion
1023
cdef object old_dirname_to_file_id # dict
1024
cdef object new_dirname_to_file_id # dict
1025
cdef object last_source_parent
1026
cdef object last_target_parent
1027
cdef int include_unchanged
1029
cdef object use_filesystem_for_exec
1030
cdef object utf8_decode
1031
cdef readonly object searched_specific_files
1032
cdef readonly object searched_exact_paths
1033
cdef object search_specific_files
1034
# The parents up to the root of the paths we are searching.
1035
# After all normal paths are returned, these specific items are returned.
1036
cdef object search_specific_file_parents
1038
# Current iteration variables:
1039
cdef object current_root
1040
cdef object current_root_unicode
1041
cdef object root_entries
1042
cdef int root_entries_pos, root_entries_len
1043
cdef object root_abspath
1044
cdef int source_index, target_index
1045
cdef int want_unversioned
1047
cdef object dir_iterator
1048
cdef int block_index
1049
cdef object current_block
1050
cdef int current_block_pos
1051
cdef object current_block_list
1052
cdef object current_dir_info
1053
cdef object current_dir_list
1054
cdef object _pending_consistent_entries # list
1056
cdef object root_dir_info
1057
cdef object bisect_left
1058
cdef object pathjoin
1060
# A set of the ids we've output when doing partial output.
1061
cdef object seen_ids
1062
cdef object sha_file
1064
def __init__(self, include_unchanged, use_filesystem_for_exec,
1065
search_specific_files, state, source_index, target_index,
1066
want_unversioned, tree):
1067
self.doing_consistency_expansion = 0
1068
self.old_dirname_to_file_id = {}
1069
self.new_dirname_to_file_id = {}
1070
# Are we doing a partial iter_changes?
1071
self.partial = set(['']).__ne__(search_specific_files)
1072
# Using a list so that we can access the values and change them in
1073
# nested scope. Each one is [path, file_id, entry]
1074
self.last_source_parent = [None, None]
1075
self.last_target_parent = [None, None]
1076
if include_unchanged is None:
1077
self.include_unchanged = False
1079
self.include_unchanged = int(include_unchanged)
1080
self.use_filesystem_for_exec = use_filesystem_for_exec
1081
self.utf8_decode = cache_utf8._utf8_decode
1082
# for all search_indexs in each path at or under each element of
1083
# search_specific_files, if the detail is relocated: add the id, and
1084
# add the relocated path as one to search if its not searched already.
1085
# If the detail is not relocated, add the id.
1086
self.searched_specific_files = set()
1087
# When we search exact paths without expanding downwards, we record
1089
self.searched_exact_paths = set()
1090
self.search_specific_files = search_specific_files
1091
# The parents up to the root of the paths we are searching.
1092
# After all normal paths are returned, these specific items are returned.
1093
self.search_specific_file_parents = set()
1094
# The ids we've sent out in the delta.
1095
self.seen_ids = set()
1097
self.current_root = None
1098
self.current_root_unicode = None
1099
self.root_entries = None
1100
self.root_entries_pos = 0
1101
self.root_entries_len = 0
1102
self.root_abspath = None
1103
if source_index is None:
1104
self.source_index = -1
1106
self.source_index = source_index
1107
self.target_index = target_index
1108
self.want_unversioned = want_unversioned
1110
self.dir_iterator = None
1111
self.block_index = -1
1112
self.current_block = None
1113
self.current_block_list = None
1114
self.current_block_pos = -1
1115
self.current_dir_info = None
1116
self.current_dir_list = None
1117
self._pending_consistent_entries = []
1119
self.root_dir_info = None
1120
self.bisect_left = bisect.bisect_left
1121
self.pathjoin = osutils.pathjoin
1122
self.fstat = os.fstat
1123
self.sha_file = osutils.sha_file
1124
if target_index != 0:
1125
# A lot of code in here depends on target_index == 0
1126
raise errors.BzrError('unsupported target index')
1128
cdef _process_entry(self, entry, path_info):
1129
"""Compare an entry and real disk to generate delta information.
1131
:param path_info: top_relpath, basename, kind, lstat, abspath for
1132
the path of entry. If None, then the path is considered absent in
1133
the target (Perhaps we should pass in a concrete entry for this ?)
1134
Basename is returned as a utf8 string because we expect this
1135
tuple will be ignored, and don't want to take the time to
1137
:return: (iter_changes_result, changed). If the entry has not been
1138
handled then changed is None. Otherwise it is False if no content
1139
or metadata changes have occured, and True if any content or
1140
metadata change has occurred. If self.include_unchanged is True then
1141
if changed is not None, iter_changes_result will always be a result
1142
tuple. Otherwise, iter_changes_result is None unless changed is
1145
cdef char target_minikind
1146
cdef char source_minikind
1148
cdef int content_change
1149
cdef object details_list
1151
details_list = entry[1]
1152
if -1 == self.source_index:
1153
source_details = DirState.NULL_PARENT_DETAILS
1155
source_details = details_list[self.source_index]
1156
target_details = details_list[self.target_index]
1157
target_minikind = _minikind_from_string(target_details[0])
1158
if path_info is not None and _versioned_minikind(target_minikind):
1159
if self.target_index != 0:
1160
raise AssertionError("Unsupported target index %d" %
1162
link_or_sha1 = _update_entry(self.state, entry, path_info[4], path_info[3])
1163
# The entry may have been modified by update_entry
1164
target_details = details_list[self.target_index]
1165
target_minikind = _minikind_from_string(target_details[0])
1168
# the rest of this function is 0.3 seconds on 50K paths, or
1169
# 0.000006 seconds per call.
1170
source_minikind = _minikind_from_string(source_details[0])
1171
if ((_versioned_minikind(source_minikind) or source_minikind == c'r')
1172
and _versioned_minikind(target_minikind)):
1173
# claimed content in both: diff
1174
# r | fdlt | | add source to search, add id path move and perform
1175
# | | | diff check on source-target
1176
# r | fdlt | a | dangling file that was present in the basis.
1178
if source_minikind != c'r':
1179
old_dirname = entry[0][0]
1180
old_basename = entry[0][1]
1181
old_path = path = None
1183
# add the source to the search path to find any children it
1184
# has. TODO ? : only add if it is a container ?
1185
if (not self.doing_consistency_expansion and
1186
not osutils.is_inside_any(self.searched_specific_files,
1187
source_details[1])):
1188
self.search_specific_files.add(source_details[1])
1189
# expanding from a user requested path, parent expansion
1190
# for delta consistency happens later.
1191
# generate the old path; this is needed for stating later
1193
old_path = source_details[1]
1194
old_dirname, old_basename = os.path.split(old_path)
1195
path = self.pathjoin(entry[0][0], entry[0][1])
1196
old_entry = self.state._get_entry(self.source_index,
1198
# update the source details variable to be the real
1200
if old_entry == (None, None):
1201
raise errors.CorruptDirstate(self.state._filename,
1202
"entry '%s/%s' is considered renamed from %r"
1203
" but source does not exist\n"
1204
"entry: %s" % (entry[0][0], entry[0][1], old_path, entry))
1205
source_details = old_entry[1][self.source_index]
1206
source_minikind = _minikind_from_string(source_details[0])
1207
if path_info is None:
1208
# the file is missing on disk, show as removed.
1213
# source and target are both versioned and disk file is present.
1214
target_kind = path_info[2]
1215
if target_kind == 'directory':
1217
old_path = path = self.pathjoin(old_dirname, old_basename)
1218
file_id = entry[0][2]
1219
self.new_dirname_to_file_id[path] = file_id
1220
if source_minikind != c'd':
1223
# directories have no fingerprint
1226
elif target_kind == 'file':
1227
if source_minikind != c'f':
1230
# Check the sha. We can't just rely on the size as
1231
# content filtering may mean differ sizes actually
1232
# map to the same content
1233
if link_or_sha1 is None:
1235
statvalue, link_or_sha1 = \
1236
self.state._sha1_provider.stat_and_sha1(
1238
self.state._observed_sha1(entry, link_or_sha1,
1240
content_change = (link_or_sha1 != source_details[1])
1241
# Target details is updated at update_entry time
1242
if self.use_filesystem_for_exec:
1243
# We don't need S_ISREG here, because we are sure
1244
# we are dealing with a file.
1245
target_exec = bool(S_IXUSR & path_info[3].st_mode)
1247
target_exec = target_details[3]
1248
elif target_kind == 'symlink':
1249
if source_minikind != c'l':
1252
content_change = (link_or_sha1 != source_details[1])
1254
elif target_kind == 'tree-reference':
1255
if source_minikind != c't':
1262
path = self.pathjoin(old_dirname, old_basename)
1263
raise errors.BadFileKindError(path, path_info[2])
1264
if source_minikind == c'd':
1266
old_path = path = self.pathjoin(old_dirname, old_basename)
1268
file_id = entry[0][2]
1269
self.old_dirname_to_file_id[old_path] = file_id
1270
# parent id is the entry for the path in the target tree
1271
if old_basename and old_dirname == self.last_source_parent[0]:
1272
# use a cached hit for non-root source entries.
1273
source_parent_id = self.last_source_parent[1]
1276
source_parent_id = self.old_dirname_to_file_id[old_dirname]
1278
source_parent_entry = self.state._get_entry(self.source_index,
1279
path_utf8=old_dirname)
1280
source_parent_id = source_parent_entry[0][2]
1281
if source_parent_id == entry[0][2]:
1282
# This is the root, so the parent is None
1283
source_parent_id = None
1285
self.last_source_parent[0] = old_dirname
1286
self.last_source_parent[1] = source_parent_id
1287
new_dirname = entry[0][0]
1288
if entry[0][1] and new_dirname == self.last_target_parent[0]:
1289
# use a cached hit for non-root target entries.
1290
target_parent_id = self.last_target_parent[1]
1293
target_parent_id = self.new_dirname_to_file_id[new_dirname]
1295
# TODO: We don't always need to do the lookup, because the
1296
# parent entry will be the same as the source entry.
1297
target_parent_entry = self.state._get_entry(self.target_index,
1298
path_utf8=new_dirname)
1299
if target_parent_entry == (None, None):
1300
raise AssertionError(
1301
"Could not find target parent in wt: %s\nparent of: %s"
1302
% (new_dirname, entry))
1303
target_parent_id = target_parent_entry[0][2]
1304
if target_parent_id == entry[0][2]:
1305
# This is the root, so the parent is None
1306
target_parent_id = None
1308
self.last_target_parent[0] = new_dirname
1309
self.last_target_parent[1] = target_parent_id
1311
source_exec = source_details[3]
1312
changed = (content_change
1313
or source_parent_id != target_parent_id
1314
or old_basename != entry[0][1]
1315
or source_exec != target_exec
1317
if not changed and not self.include_unchanged:
1320
if old_path is None:
1321
path = self.pathjoin(old_dirname, old_basename)
1323
old_path_u = self.utf8_decode(old_path)[0]
1326
old_path_u = self.utf8_decode(old_path)[0]
1327
if old_path == path:
1330
path_u = self.utf8_decode(path)[0]
1331
source_kind = _minikind_to_kind(source_minikind)
1332
return (entry[0][2],
1333
(old_path_u, path_u),
1336
(source_parent_id, target_parent_id),
1337
(self.utf8_decode(old_basename)[0], self.utf8_decode(entry[0][1])[0]),
1338
(source_kind, target_kind),
1339
(source_exec, target_exec)), changed
1340
elif source_minikind == c'a' and _versioned_minikind(target_minikind):
1341
# looks like a new file
1342
path = self.pathjoin(entry[0][0], entry[0][1])
1343
# parent id is the entry for the path in the target tree
1344
# TODO: these are the same for an entire directory: cache em.
1345
parent_entry = self.state._get_entry(self.target_index,
1346
path_utf8=entry[0][0])
1347
if parent_entry is None:
1348
raise errors.DirstateCorrupt(self.state,
1349
"We could not find the parent entry in index %d"
1350
" for the entry: %s"
1351
% (self.target_index, entry[0]))
1352
parent_id = parent_entry[0][2]
1353
if parent_id == entry[0][2]:
1355
if path_info is not None:
1357
if self.use_filesystem_for_exec:
1358
# We need S_ISREG here, because we aren't sure if this
1361
S_ISREG(path_info[3].st_mode)
1362
and S_IXUSR & path_info[3].st_mode)
1364
target_exec = target_details[3]
1365
return (entry[0][2],
1366
(None, self.utf8_decode(path)[0]),
1370
(None, self.utf8_decode(entry[0][1])[0]),
1371
(None, path_info[2]),
1372
(None, target_exec)), True
1374
# Its a missing file, report it as such.
1375
return (entry[0][2],
1376
(None, self.utf8_decode(path)[0]),
1380
(None, self.utf8_decode(entry[0][1])[0]),
1382
(None, False)), True
1383
elif _versioned_minikind(source_minikind) and target_minikind == c'a':
1384
# unversioned, possibly, or possibly not deleted: we dont care.
1385
# if its still on disk, *and* theres no other entry at this
1386
# path [we dont know this in this routine at the moment -
1387
# perhaps we should change this - then it would be an unknown.
1388
old_path = self.pathjoin(entry[0][0], entry[0][1])
1389
# parent id is the entry for the path in the target tree
1390
parent_id = self.state._get_entry(self.source_index, path_utf8=entry[0][0])[0][2]
1391
if parent_id == entry[0][2]:
1393
return (entry[0][2],
1394
(self.utf8_decode(old_path)[0], None),
1398
(self.utf8_decode(entry[0][1])[0], None),
1399
(_minikind_to_kind(source_minikind), None),
1400
(source_details[3], None)), True
1401
elif _versioned_minikind(source_minikind) and target_minikind == c'r':
1402
# a rename; could be a true rename, or a rename inherited from
1403
# a renamed parent. TODO: handle this efficiently. Its not
1404
# common case to rename dirs though, so a correct but slow
1405
# implementation will do.
1406
if (not self.doing_consistency_expansion and
1407
not osutils.is_inside_any(self.searched_specific_files,
1408
target_details[1])):
1409
self.search_specific_files.add(target_details[1])
1410
# We don't expand the specific files parents list here as
1411
# the path is absent in target and won't create a delta with
1413
elif ((source_minikind == c'r' or source_minikind == c'a') and
1414
(target_minikind == c'r' or target_minikind == c'a')):
1415
# neither of the selected trees contain this path,
1416
# so skip over it. This is not currently directly tested, but
1417
# is indirectly via test_too_much.TestCommands.test_conflicts.
1420
raise AssertionError("don't know how to compare "
1421
"source_minikind=%r, target_minikind=%r"
1422
% (source_minikind, target_minikind))
1423
## import pdb;pdb.set_trace()
1429
def iter_changes(self):
1432
cdef int _gather_result_for_consistency(self, result) except -1:
1433
"""Check a result we will yield to make sure we are consistent later.
1435
This gathers result's parents into a set to output later.
1437
:param result: A result tuple.
1439
if not self.partial or not result[0]:
1441
self.seen_ids.add(result[0])
1442
new_path = result[1][1]
1444
# Not the root and not a delete: queue up the parents of the path.
1445
self.search_specific_file_parents.update(
1446
osutils.parent_directories(new_path.encode('utf8')))
1447
# Add the root directory which parent_directories does not
1449
self.search_specific_file_parents.add('')
1452
cdef int _update_current_block(self) except -1:
1453
if (self.block_index < len(self.state._dirblocks) and
1454
osutils.is_inside(self.current_root, self.state._dirblocks[self.block_index][0])):
1455
self.current_block = self.state._dirblocks[self.block_index]
1456
self.current_block_list = self.current_block[1]
1457
self.current_block_pos = 0
1459
self.current_block = None
1460
self.current_block_list = None
1464
# Simple thunk to allow tail recursion without pyrex confusion
1465
return self._iter_next()
1467
cdef _iter_next(self):
1468
"""Iterate over the changes."""
1469
# This function single steps through an iterator. As such while loops
1470
# are often exited by 'return' - the code is structured so that the
1471
# next call into the function will return to the same while loop. Note
1472
# that all flow control needed to re-reach that step is reexecuted,
1473
# which can be a performance problem. It has not yet been tuned to
1474
# minimise this; a state machine is probably the simplest restructuring
1475
# to both minimise this overhead and make the code considerably more
1479
# compare source_index and target_index at or under each element of search_specific_files.
1480
# follow the following comparison table. Note that we only want to do diff operations when
1481
# the target is fdl because thats when the walkdirs logic will have exposed the pathinfo
1485
# Source | Target | disk | action
1486
# r | fdlt | | add source to search, add id path move and perform
1487
# | | | diff check on source-target
1488
# r | fdlt | a | dangling file that was present in the basis.
1490
# r | a | | add source to search
1492
# r | r | | this path is present in a non-examined tree, skip.
1493
# r | r | a | this path is present in a non-examined tree, skip.
1494
# a | fdlt | | add new id
1495
# a | fdlt | a | dangling locally added file, skip
1496
# a | a | | not present in either tree, skip
1497
# a | a | a | not present in any tree, skip
1498
# a | r | | not present in either tree at this path, skip as it
1499
# | | | may not be selected by the users list of paths.
1500
# a | r | a | not present in either tree at this path, skip as it
1501
# | | | may not be selected by the users list of paths.
1502
# fdlt | fdlt | | content in both: diff them
1503
# fdlt | fdlt | a | deleted locally, but not unversioned - show as deleted ?
1504
# fdlt | a | | unversioned: output deleted id for now
1505
# fdlt | a | a | unversioned and deleted: output deleted id
1506
# fdlt | r | | relocated in this tree, so add target to search.
1507
# | | | Dont diff, we will see an r,fd; pair when we reach
1508
# | | | this id at the other path.
1509
# fdlt | r | a | relocated in this tree, so add target to search.
1510
# | | | Dont diff, we will see an r,fd; pair when we reach
1511
# | | | this id at the other path.
1513
# TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
1514
# keeping a cache of directories that we have seen.
1515
cdef object current_dirname, current_blockname
1516
cdef char * current_dirname_c, * current_blockname_c
1517
cdef int advance_entry, advance_path
1518
cdef int path_handled
1519
searched_specific_files = self.searched_specific_files
1520
# Are we walking a root?
1521
while self.root_entries_pos < self.root_entries_len:
1522
entry = self.root_entries[self.root_entries_pos]
1523
self.root_entries_pos = self.root_entries_pos + 1
1524
result, changed = self._process_entry(entry, self.root_dir_info)
1525
if changed is not None:
1527
self._gather_result_for_consistency(result)
1528
if changed or self.include_unchanged:
1530
# Have we finished the prior root, or never started one ?
1531
if self.current_root is None:
1532
# TODO: the pending list should be lexically sorted? the
1533
# interface doesn't require it.
1535
self.current_root = self.search_specific_files.pop()
1537
raise StopIteration()
1538
self.searched_specific_files.add(self.current_root)
1539
# process the entries for this containing directory: the rest will be
1540
# found by their parents recursively.
1541
self.root_entries = self.state._entries_for_path(self.current_root)
1542
self.root_entries_len = len(self.root_entries)
1543
self.current_root_unicode = self.current_root.decode('utf8')
1544
self.root_abspath = self.tree.abspath(self.current_root_unicode)
1546
root_stat = os.lstat(self.root_abspath)
1548
if e.errno == errno.ENOENT:
1549
# the path does not exist: let _process_entry know that.
1550
self.root_dir_info = None
1552
# some other random error: hand it up.
1555
self.root_dir_info = ('', self.current_root,
1556
osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
1558
if self.root_dir_info[2] == 'directory':
1559
if self.tree._directory_is_tree_reference(
1560
self.current_root_unicode):
1561
self.root_dir_info = self.root_dir_info[:2] + \
1562
('tree-reference',) + self.root_dir_info[3:]
1563
if not self.root_entries and not self.root_dir_info:
1564
# this specified path is not present at all, skip it.
1565
# (tail recursion, can do a loop once the full structure is
1567
return self._iter_next()
1569
self.root_entries_pos = 0
1570
# XXX Clarity: This loop is duplicated a out the self.current_root
1571
# is None guard above: if we return from it, it completes there
1572
# (and the following if block cannot trigger because
1573
# path_handled must be true, so the if block is not # duplicated.
1574
while self.root_entries_pos < self.root_entries_len:
1575
entry = self.root_entries[self.root_entries_pos]
1576
self.root_entries_pos = self.root_entries_pos + 1
1577
result, changed = self._process_entry(entry, self.root_dir_info)
1578
if changed is not None:
1581
self._gather_result_for_consistency(result)
1582
if changed or self.include_unchanged:
1584
# handle unversioned specified paths:
1585
if self.want_unversioned and not path_handled and self.root_dir_info:
1586
new_executable = bool(
1587
stat.S_ISREG(self.root_dir_info[3].st_mode)
1588
and stat.S_IEXEC & self.root_dir_info[3].st_mode)
1590
(None, self.current_root_unicode),
1594
(None, splitpath(self.current_root_unicode)[-1]),
1595
(None, self.root_dir_info[2]),
1596
(None, new_executable)
1598
# If we reach here, the outer flow continues, which enters into the
1599
# per-root setup logic.
1600
if (self.current_dir_info is None and self.current_block is None and not
1601
self.doing_consistency_expansion):
1602
# setup iteration of this root:
1603
self.current_dir_list = None
1604
if self.root_dir_info and self.root_dir_info[2] == 'tree-reference':
1605
self.current_dir_info = None
1607
self.dir_iterator = osutils._walkdirs_utf8(self.root_abspath,
1608
prefix=self.current_root)
1611
self.current_dir_info = self.dir_iterator.next()
1612
self.current_dir_list = self.current_dir_info[1]
1614
# there may be directories in the inventory even though
1615
# this path is not a file on disk: so mark it as end of
1617
if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
1618
self.current_dir_info = None
1619
elif sys.platform == 'win32':
1620
# on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
1621
# python 2.5 has e.errno == EINVAL,
1622
# and e.winerror == ERROR_DIRECTORY
1624
e_winerror = e.winerror
1625
except AttributeError, _:
1627
win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
1628
if (e.errno in win_errors or e_winerror in win_errors):
1629
self.current_dir_info = None
1631
# Will this really raise the right exception ?
1636
if self.current_dir_info[0][0] == '':
1637
# remove .bzr from iteration
1638
bzr_index = self.bisect_left(self.current_dir_list, ('.bzr',))
1639
if self.current_dir_list[bzr_index][0] != '.bzr':
1640
raise AssertionError()
1641
del self.current_dir_list[bzr_index]
1642
initial_key = (self.current_root, '', '')
1643
self.block_index, _ = self.state._find_block_index_from_key(initial_key)
1644
if self.block_index == 0:
1645
# we have processed the total root already, but because the
1646
# initial key matched it we should skip it here.
1647
self.block_index = self.block_index + 1
1648
self._update_current_block()
1649
# walk until both the directory listing and the versioned metadata
1651
while (self.current_dir_info is not None
1652
or self.current_block is not None):
1653
# Uncommon case - a missing directory or an unversioned directory:
1654
if (self.current_dir_info and self.current_block
1655
and self.current_dir_info[0][0] != self.current_block[0]):
1656
# Work around pyrex broken heuristic - current_dirname has
1657
# the same scope as current_dirname_c
1658
current_dirname = self.current_dir_info[0][0]
1659
current_dirname_c = PyString_AS_STRING_void(
1660
<void *>current_dirname)
1661
current_blockname = self.current_block[0]
1662
current_blockname_c = PyString_AS_STRING_void(
1663
<void *>current_blockname)
1664
# In the python generator we evaluate this if block once per
1665
# dir+block; because we reenter in the pyrex version its being
1666
# evaluated once per path: we could cache the result before
1667
# doing the while loop and probably save time.
1668
if _cmp_by_dirs(current_dirname_c,
1669
PyString_Size(current_dirname),
1670
current_blockname_c,
1671
PyString_Size(current_blockname)) < 0:
1672
# filesystem data refers to paths not covered by the
1673
# dirblock. this has two possibilities:
1674
# A) it is versioned but empty, so there is no block for it
1675
# B) it is not versioned.
1677
# if (A) then we need to recurse into it to check for
1678
# new unknown files or directories.
1679
# if (B) then we should ignore it, because we don't
1680
# recurse into unknown directories.
1681
# We are doing a loop
1682
while self.path_index < len(self.current_dir_list):
1683
current_path_info = self.current_dir_list[self.path_index]
1684
# dont descend into this unversioned path if it is
1686
if current_path_info[2] in ('directory',
1688
del self.current_dir_list[self.path_index]
1689
self.path_index = self.path_index - 1
1690
self.path_index = self.path_index + 1
1691
if self.want_unversioned:
1692
if current_path_info[2] == 'directory':
1693
if self.tree._directory_is_tree_reference(
1694
self.utf8_decode(current_path_info[0])[0]):
1695
current_path_info = current_path_info[:2] + \
1696
('tree-reference',) + current_path_info[3:]
1697
new_executable = bool(
1698
stat.S_ISREG(current_path_info[3].st_mode)
1699
and stat.S_IEXEC & current_path_info[3].st_mode)
1701
(None, self.utf8_decode(current_path_info[0])[0]),
1705
(None, self.utf8_decode(current_path_info[1])[0]),
1706
(None, current_path_info[2]),
1707
(None, new_executable))
1708
# This dir info has been handled, go to the next
1710
self.current_dir_list = None
1712
self.current_dir_info = self.dir_iterator.next()
1713
self.current_dir_list = self.current_dir_info[1]
1714
except StopIteration, _:
1715
self.current_dir_info = None
1717
# We have a dirblock entry for this location, but there
1718
# is no filesystem path for this. This is most likely
1719
# because a directory was removed from the disk.
1720
# We don't have to report the missing directory,
1721
# because that should have already been handled, but we
1722
# need to handle all of the files that are contained
1724
while self.current_block_pos < len(self.current_block_list):
1725
current_entry = self.current_block_list[self.current_block_pos]
1726
self.current_block_pos = self.current_block_pos + 1
1727
# entry referring to file not present on disk.
1728
# advance the entry only, after processing.
1729
result, changed = self._process_entry(current_entry, None)
1730
if changed is not None:
1732
self._gather_result_for_consistency(result)
1733
if changed or self.include_unchanged:
1735
self.block_index = self.block_index + 1
1736
self._update_current_block()
1737
continue # next loop-on-block/dir
1738
result = self._loop_one_block()
1739
if result is not None:
1741
if len(self.search_specific_files):
1742
# More supplied paths to process
1743
self.current_root = None
1744
return self._iter_next()
1745
# Start expanding more conservatively, adding paths the user may not
1746
# have intended but required for consistent deltas.
1747
self.doing_consistency_expansion = 1
1748
if not self._pending_consistent_entries:
1749
self._pending_consistent_entries = self._next_consistent_entries()
1750
while self._pending_consistent_entries:
1751
result, changed = self._pending_consistent_entries.pop()
1752
if changed is not None:
1754
raise StopIteration()
1756
cdef object _maybe_tree_ref(self, current_path_info):
1757
if self.tree._directory_is_tree_reference(
1758
self.utf8_decode(current_path_info[0])[0]):
1759
return current_path_info[:2] + \
1760
('tree-reference',) + current_path_info[3:]
1762
return current_path_info
1764
cdef object _loop_one_block(self):
1765
# current_dir_info and current_block refer to the same directory -
1766
# this is the common case code.
1767
# Assign local variables for current path and entry:
1768
cdef object current_entry
1769
cdef object current_path_info
1770
cdef int path_handled
1773
# cdef char * temp_str
1774
# cdef Py_ssize_t temp_str_length
1775
# PyString_AsStringAndSize(disk_kind, &temp_str, &temp_str_length)
1776
# if not strncmp(temp_str, "directory", temp_str_length):
1777
if (self.current_block is not None and
1778
self.current_block_pos < PyList_GET_SIZE(self.current_block_list)):
1779
current_entry = PyList_GET_ITEM(self.current_block_list,
1780
self.current_block_pos)
1782
Py_INCREF(current_entry)
1784
current_entry = None
1785
if (self.current_dir_info is not None and
1786
self.path_index < PyList_GET_SIZE(self.current_dir_list)):
1787
current_path_info = PyList_GET_ITEM(self.current_dir_list,
1790
Py_INCREF(current_path_info)
1791
disk_kind = PyTuple_GET_ITEM(current_path_info, 2)
1793
Py_INCREF(disk_kind)
1794
if disk_kind == "directory":
1795
current_path_info = self._maybe_tree_ref(current_path_info)
1797
current_path_info = None
1798
while (current_entry is not None or current_path_info is not None):
1804
if current_entry is None:
1805
# unversioned - the check for path_handled when the path
1806
# is advanced will yield this path if needed.
1808
elif current_path_info is None:
1809
# no path is fine: the per entry code will handle it.
1810
result, changed = self._process_entry(current_entry,
1813
minikind = _minikind_from_string(
1814
current_entry[1][self.target_index][0])
1815
cmp_result = cmp(current_path_info[1], current_entry[0][1])
1816
if (cmp_result or minikind == c'a' or minikind == c'r'):
1817
# The current path on disk doesn't match the dirblock
1818
# record. Either the dirblock record is marked as
1819
# absent/renamed, or the file on disk is not present at all
1820
# in the dirblock. Either way, report about the dirblock
1821
# entry, and let other code handle the filesystem one.
1823
# Compare the basename for these files to determine
1826
# extra file on disk: pass for now, but only
1827
# increment the path, not the entry
1830
# entry referring to file not present on disk.
1831
# advance the entry only, after processing.
1832
result, changed = self._process_entry(current_entry,
1836
# paths are the same,and the dirstate entry is not
1837
# absent or renamed.
1838
result, changed = self._process_entry(current_entry,
1840
if changed is not None:
1842
if not changed and not self.include_unchanged:
1844
# >- loop control starts here:
1846
if advance_entry and current_entry is not None:
1847
self.current_block_pos = self.current_block_pos + 1
1848
if self.current_block_pos < PyList_GET_SIZE(self.current_block_list):
1849
current_entry = self.current_block_list[self.current_block_pos]
1851
current_entry = None
1853
if advance_path and current_path_info is not None:
1854
if not path_handled:
1855
# unversioned in all regards
1856
if self.want_unversioned:
1857
new_executable = bool(
1858
stat.S_ISREG(current_path_info[3].st_mode)
1859
and stat.S_IEXEC & current_path_info[3].st_mode)
1861
relpath_unicode = self.utf8_decode(current_path_info[0])[0]
1862
except UnicodeDecodeError, _:
1863
raise errors.BadFilenameEncoding(
1864
current_path_info[0], osutils._fs_enc)
1865
if changed is not None:
1866
raise AssertionError(
1867
"result is not None: %r" % result)
1869
(None, relpath_unicode),
1873
(None, self.utf8_decode(current_path_info[1])[0]),
1874
(None, current_path_info[2]),
1875
(None, new_executable))
1877
# dont descend into this unversioned path if it is
1879
if current_path_info[2] in ('directory'):
1880
del self.current_dir_list[self.path_index]
1881
self.path_index = self.path_index - 1
1882
# dont descend the disk iterator into any tree
1884
if current_path_info[2] == 'tree-reference':
1885
del self.current_dir_list[self.path_index]
1886
self.path_index = self.path_index - 1
1887
self.path_index = self.path_index + 1
1888
if self.path_index < len(self.current_dir_list):
1889
current_path_info = self.current_dir_list[self.path_index]
1890
if current_path_info[2] == 'directory':
1891
current_path_info = self._maybe_tree_ref(
1894
current_path_info = None
1895
if changed is not None:
1896
# Found a result on this pass, yield it
1898
self._gather_result_for_consistency(result)
1899
if changed or self.include_unchanged:
1901
if self.current_block is not None:
1902
self.block_index = self.block_index + 1
1903
self._update_current_block()
1904
if self.current_dir_info is not None:
1906
self.current_dir_list = None
1908
self.current_dir_info = self.dir_iterator.next()
1909
self.current_dir_list = self.current_dir_info[1]
1910
except StopIteration, _:
1911
self.current_dir_info = None
1913
cdef object _next_consistent_entries(self):
1914
"""Grabs the next specific file parent case to consider.
1916
:return: A list of the results, each of which is as for _process_entry.
1919
while self.search_specific_file_parents:
1920
# Process the parent directories for the paths we were iterating.
1921
# Even in extremely large trees this should be modest, so currently
1922
# no attempt is made to optimise.
1923
path_utf8 = self.search_specific_file_parents.pop()
1924
if path_utf8 in self.searched_exact_paths:
1925
# We've examined this path.
1927
if osutils.is_inside_any(self.searched_specific_files, path_utf8):
1928
# We've examined this path.
1930
path_entries = self.state._entries_for_path(path_utf8)
1931
# We need either one or two entries. If the path in
1932
# self.target_index has moved (so the entry in source_index is in
1933
# 'ar') then we need to also look for the entry for this path in
1934
# self.source_index, to output the appropriate delete-or-rename.
1935
selected_entries = []
1937
for candidate_entry in path_entries:
1938
# Find entries present in target at this path:
1939
if candidate_entry[1][self.target_index][0] not in 'ar':
1941
selected_entries.append(candidate_entry)
1942
# Find entries present in source at this path:
1943
elif (self.source_index is not None and
1944
candidate_entry[1][self.source_index][0] not in 'ar'):
1946
if candidate_entry[1][self.target_index][0] == 'a':
1947
# Deleted, emit it here.
1948
selected_entries.append(candidate_entry)
1950
# renamed, emit it when we process the directory it
1952
self.search_specific_file_parents.add(
1953
candidate_entry[1][self.target_index][1])
1955
raise AssertionError(
1956
"Missing entry for specific path parent %r, %r" % (
1957
path_utf8, path_entries))
1958
path_info = self._path_info(path_utf8, path_utf8.decode('utf8'))
1959
for entry in selected_entries:
1960
if entry[0][2] in self.seen_ids:
1962
result, changed = self._process_entry(entry, path_info)
1964
raise AssertionError(
1965
"Got entry<->path mismatch for specific path "
1966
"%r entry %r path_info %r " % (
1967
path_utf8, entry, path_info))
1968
# Only include changes - we're outside the users requested
1971
self._gather_result_for_consistency(result)
1972
if (result[6][0] == 'directory' and
1973
result[6][1] != 'directory'):
1974
# This stopped being a directory, the old children have
1976
if entry[1][self.source_index][0] == 'r':
1977
# renamed, take the source path
1978
entry_path_utf8 = entry[1][self.source_index][1]
1980
entry_path_utf8 = path_utf8
1981
initial_key = (entry_path_utf8, '', '')
1982
block_index, _ = self.state._find_block_index_from_key(
1984
if block_index == 0:
1985
# The children of the root are in block index 1.
1986
block_index = block_index + 1
1987
current_block = None
1988
if block_index < len(self.state._dirblocks):
1989
current_block = self.state._dirblocks[block_index]
1990
if not osutils.is_inside(
1991
entry_path_utf8, current_block[0]):
1992
# No entries for this directory at all.
1993
current_block = None
1994
if current_block is not None:
1995
for entry in current_block[1]:
1996
if entry[1][self.source_index][0] in 'ar':
1997
# Not in the source tree, so doesn't have to be
2000
# Path of the entry itself.
2001
self.search_specific_file_parents.add(
2002
self.pathjoin(*entry[0][:2]))
2003
if changed or self.include_unchanged:
2004
results.append((result, changed))
2005
self.searched_exact_paths.add(path_utf8)
2008
cdef object _path_info(self, utf8_path, unicode_path):
2009
"""Generate path_info for unicode_path.
2011
:return: None if unicode_path does not exist, or a path_info tuple.
2013
abspath = self.tree.abspath(unicode_path)
2015
stat = os.lstat(abspath)
2017
if e.errno == errno.ENOENT:
2018
# the path does not exist.
2022
utf8_basename = utf8_path.rsplit('/', 1)[-1]
2023
dir_info = (utf8_path, utf8_basename,
2024
osutils.file_kind_from_stat_mode(stat.st_mode), stat,
2026
if dir_info[2] == 'directory':
2027
if self.tree._directory_is_tree_reference(
2029
self.root_dir_info = self.root_dir_info[:2] + \
2030
('tree-reference',) + self.root_dir_info[3:]
added
removed
bzrlib/_dirstate_helpers_pyx.pyx
