1
# Bazaar-NG -- distributed version control
3
# Copyright (C) 2005 by Canonical Ltd
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
5
3
# This program is free software; you can redistribute it and/or modify
6
4
# it under the terms of the GNU General Public License as published by
7
5
# the Free Software Foundation; either version 2 of the License, or
8
6
# (at your option) any later version.
10
8
# This program is distributed in the hope that it will be useful,
11
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
12
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13
11
# GNU General Public License for more details.
15
13
# You should have received a copy of the GNU General Public License
16
14
# along with this program; if not, write to the Free Software
17
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
19
import os, types, re, time, errno
20
from stat import S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE
22
from errors import bailout, BzrError
23
from trace import mutter
17
from cStringIO import StringIO
21
from stat import (S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE,
22
S_ISCHR, S_ISBLK, S_ISFIFO, S_ISSOCK)
26
from bzrlib.lazy_import import lazy_import
27
lazy_import(globals(), """
29
from datetime import datetime
31
from ntpath import (abspath as _nt_abspath,
33
normpath as _nt_normpath,
34
realpath as _nt_realpath,
35
splitdrive as _nt_splitdrive,
43
from tempfile import (
55
# sha is deprecated in python2.6 but haslib is available as of 2.5
56
if sys.version_info < (2, 5):
67
from bzrlib import symbol_versioning
68
from bzrlib.symbol_versioning import (
71
from bzrlib.trace import mutter
74
# On win32, O_BINARY is used to indicate the file should
75
# be opened in binary mode, rather than text mode.
76
# On other platforms, O_BINARY doesn't exist, because
77
# they always open in binary mode, so it is okay to
78
# OR with 0 on those platforms
79
O_BINARY = getattr(os, 'O_BINARY', 0)
81
# On posix, use lstat instead of stat so that we can
82
# operate on broken symlinks. On Windows revert to stat.
83
lstat = getattr(os, 'lstat', os.stat)
26
85
def make_readonly(filename):
27
86
"""Make a filename read-only."""
28
# TODO: probably needs to be fixed for windows
29
mod = os.stat(filename).st_mode
31
os.chmod(filename, mod)
87
mod = lstat(filename).st_mode
88
if not stat.S_ISLNK(mod):
90
os.chmod(filename, mod)
34
93
def make_writable(filename):
35
mod = os.stat(filename).st_mode
37
os.chmod(filename, mod)
40
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/_~-])')
94
mod = lstat(filename).st_mode
95
if not stat.S_ISLNK(mod):
97
os.chmod(filename, mod)
100
def minimum_path_selection(paths):
101
"""Return the smallset subset of paths which are outside paths.
103
:param paths: A container (and hence not None) of paths.
104
:return: A set of paths sufficient to include everything in paths via
105
is_inside_any, drawn from the paths parameter.
110
other_paths = paths.difference([path])
111
if not is_inside_any(other_paths, path):
112
# this is a top level path, we must check it.
113
search_paths.add(path)
42
"""Return shell-quoted filename"""
43
## We could be a bit more terse by using double-quotes etc
44
f = _QUOTE_RE.sub(r'\\\1', f)
51
mode = os.lstat(f)[ST_MODE]
59
raise BzrError("can't handle file kind with mode %o of %r" % (mode, f))
121
"""Return a quoted filename filename
123
This previously used backslash quoting, but that works poorly on
125
# TODO: I'm not really sure this is the best format either.x
127
if _QUOTE_RE is None:
128
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
130
if _QUOTE_RE.search(f):
136
_directory_kind = 'directory'
139
stat.S_IFDIR:_directory_kind,
140
stat.S_IFCHR:'chardev',
141
stat.S_IFBLK:'block',
144
stat.S_IFLNK:'symlink',
145
stat.S_IFSOCK:'socket',
149
def file_kind_from_stat_mode(stat_mode, _formats=_formats, _unknown='unknown'):
150
"""Generate a file kind from a stat mode. This is used in walkdirs.
152
Its performance is critical: Do not mutate without careful benchmarking.
155
return _formats[stat_mode & 0170000]
160
def file_kind(f, _lstat=os.lstat, _mapper=file_kind_from_stat_mode):
162
return _mapper(_lstat(f).st_mode)
164
if getattr(e, 'errno', None) == errno.ENOENT:
165
raise errors.NoSuchFile(f)
170
"""Return the current umask"""
171
# Assume that people aren't messing with the umask while running
172
# XXX: This is not thread safe, but there is no way to get the
173
# umask without setting it
181
_directory_kind: "/",
183
'tree-reference': '+',
187
def kind_marker(kind):
189
return _kind_marker_map[kind]
191
raise errors.BzrError('invalid file kind %r' % kind)
194
lexists = getattr(os.path, 'lexists', None)
198
stat = getattr(os, 'lstat', os.stat)
202
if e.errno == errno.ENOENT:
205
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
208
def fancy_rename(old, new, rename_func, unlink_func):
209
"""A fancy rename, when you don't have atomic rename.
211
:param old: The old path, to rename from
212
:param new: The new path, to rename to
213
:param rename_func: The potentially non-atomic rename function
214
:param unlink_func: A way to delete the target file if the full rename succeeds
217
# sftp rename doesn't allow overwriting, so play tricks:
219
base = os.path.basename(new)
220
dirname = os.path.dirname(new)
221
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
222
tmp_name = pathjoin(dirname, tmp_name)
224
# Rename the file out of the way, but keep track if it didn't exist
225
# We don't want to grab just any exception
226
# something like EACCES should prevent us from continuing
227
# The downside is that the rename_func has to throw an exception
228
# with an errno = ENOENT, or NoSuchFile
231
rename_func(new, tmp_name)
232
except (errors.NoSuchFile,), e:
235
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
236
# function raises an IOError with errno is None when a rename fails.
237
# This then gets caught here.
238
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
241
if (getattr(e, 'errno', None) is None
242
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
249
# This may throw an exception, in which case success will
251
rename_func(old, new)
255
# If the file used to exist, rename it back into place
256
# otherwise just delete it from the tmp location
258
unlink_func(tmp_name)
260
rename_func(tmp_name, new)
263
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
264
# choke on a Unicode string containing a relative path if
265
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
267
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
268
def _posix_abspath(path):
269
# jam 20060426 rather than encoding to fsencoding
270
# copy posixpath.abspath, but use os.getcwdu instead
271
if not posixpath.isabs(path):
272
path = posixpath.join(getcwd(), path)
273
return posixpath.normpath(path)
276
def _posix_realpath(path):
277
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
280
def _win32_fixdrive(path):
281
"""Force drive letters to be consistent.
283
win32 is inconsistent whether it returns lower or upper case
284
and even if it was consistent the user might type the other
285
so we force it to uppercase
286
running python.exe under cmd.exe return capital C:\\
287
running win32 python inside a cygwin shell returns lowercase c:\\
289
drive, path = _nt_splitdrive(path)
290
return drive.upper() + path
293
def _win32_abspath(path):
294
# Real _nt_abspath doesn't have a problem with a unicode cwd
295
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
298
def _win98_abspath(path):
299
"""Return the absolute version of a path.
300
Windows 98 safe implementation (python reimplementation
301
of Win32 API function GetFullPathNameW)
306
# \\HOST\path => //HOST/path
307
# //HOST/path => //HOST/path
308
# path => C:/cwd/path
311
# check for absolute path
312
drive = _nt_splitdrive(path)[0]
313
if drive == '' and path[:2] not in('//','\\\\'):
315
# we cannot simply os.path.join cwd and path
316
# because os.path.join('C:','/path') produce '/path'
317
# and this is incorrect
318
if path[:1] in ('/','\\'):
319
cwd = _nt_splitdrive(cwd)[0]
321
path = cwd + '\\' + path
322
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
324
if win32utils.winver == 'Windows 98':
325
_win32_abspath = _win98_abspath
328
def _win32_realpath(path):
329
# Real _nt_realpath doesn't have a problem with a unicode cwd
330
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
333
def _win32_pathjoin(*args):
334
return _nt_join(*args).replace('\\', '/')
337
def _win32_normpath(path):
338
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
342
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
345
def _win32_mkdtemp(*args, **kwargs):
346
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
349
def _win32_rename(old, new):
350
"""We expect to be able to atomically replace 'new' with old.
352
On win32, if new exists, it must be moved out of the way first,
356
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
358
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
359
# If we try to rename a non-existant file onto cwd, we get
360
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
361
# if the old path doesn't exist, sometimes we get EACCES
362
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
368
return unicodedata.normalize('NFKC', os.getcwdu())
371
# Default is to just use the python builtins, but these can be rebound on
372
# particular platforms.
373
abspath = _posix_abspath
374
realpath = _posix_realpath
375
pathjoin = os.path.join
376
normpath = os.path.normpath
379
dirname = os.path.dirname
380
basename = os.path.basename
381
split = os.path.split
382
splitext = os.path.splitext
383
# These were already imported into local scope
384
# mkdtemp = tempfile.mkdtemp
385
# rmtree = shutil.rmtree
387
MIN_ABS_PATHLENGTH = 1
390
if sys.platform == 'win32':
391
abspath = _win32_abspath
392
realpath = _win32_realpath
393
pathjoin = _win32_pathjoin
394
normpath = _win32_normpath
395
getcwd = _win32_getcwd
396
mkdtemp = _win32_mkdtemp
397
rename = _win32_rename
399
MIN_ABS_PATHLENGTH = 3
401
def _win32_delete_readonly(function, path, excinfo):
402
"""Error handler for shutil.rmtree function [for win32]
403
Helps to remove files and dirs marked as read-only.
405
exception = excinfo[1]
406
if function in (os.remove, os.rmdir) \
407
and isinstance(exception, OSError) \
408
and exception.errno == errno.EACCES:
414
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
415
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
416
return shutil.rmtree(path, ignore_errors, onerror)
417
elif sys.platform == 'darwin':
421
def get_terminal_encoding():
422
"""Find the best encoding for printing to the screen.
424
This attempts to check both sys.stdout and sys.stdin to see
425
what encoding they are in, and if that fails it falls back to
426
bzrlib.user_encoding.
427
The problem is that on Windows, locale.getpreferredencoding()
428
is not the same encoding as that used by the console:
429
http://mail.python.org/pipermail/python-list/2003-May/162357.html
431
On my standard US Windows XP, the preferred encoding is
432
cp1252, but the console is cp437
434
output_encoding = getattr(sys.stdout, 'encoding', None)
435
if not output_encoding:
436
input_encoding = getattr(sys.stdin, 'encoding', None)
437
if not input_encoding:
438
output_encoding = bzrlib.user_encoding
439
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
441
output_encoding = input_encoding
442
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
444
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
445
if output_encoding == 'cp0':
446
# invalid encoding (cp0 means 'no codepage' on Windows)
447
output_encoding = bzrlib.user_encoding
448
mutter('cp0 is invalid encoding.'
449
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
452
codecs.lookup(output_encoding)
454
sys.stderr.write('bzr: warning:'
455
' unknown terminal encoding %s.\n'
456
' Using encoding %s instead.\n'
457
% (output_encoding, bzrlib.user_encoding)
459
output_encoding = bzrlib.user_encoding
461
return output_encoding
464
def normalizepath(f):
465
if getattr(os.path, 'realpath', None) is not None:
469
[p,e] = os.path.split(f)
470
if e == "" or e == "." or e == "..":
473
return pathjoin(F(p), e)
477
"""Copy a file to a backup.
479
Backups are named in GNU-style, with a ~ suffix.
481
If the file is already a backup, it's not copied.
487
if has_symlinks() and os.path.islink(fn):
488
target = os.readlink(fn)
489
os.symlink(target, bfn)
497
outf = file(bfn, 'wb')
241
646
def local_time_offset(t=None):
242
647
"""Return offset of local zone from GMT, either at present or at time t."""
243
# python2.3 localtime() can't take None
247
if time.localtime(t).tm_isdst and time.daylight:
250
return -time.timezone
253
def format_date(t, offset=0, timezone='original'):
254
## TODO: Perhaps a global option to use either universal or local time?
255
## Or perhaps just let people set $TZ?
256
assert isinstance(t, float)
650
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
651
return offset.days * 86400 + offset.seconds
654
def format_date(t, offset=0, timezone='original', date_fmt=None,
656
"""Return a formatted date string.
658
:param t: Seconds since the epoch.
659
:param offset: Timezone offset in seconds east of utc.
660
:param timezone: How to display the time: 'utc', 'original' for the
661
timezone specified by offset, or 'local' for the process's current
663
:param show_offset: Whether to append the timezone.
664
:param date_fmt: strftime format.
258
666
if timezone == 'utc':
259
667
tt = time.gmtime(t)
261
669
elif timezone == 'original':
264
672
tt = time.gmtime(t + offset)
265
673
elif timezone == 'local':
266
674
tt = time.localtime(t)
267
675
offset = local_time_offset(t)
269
bailout("unsupported timezone format %r",
270
['options are "utc", "original", "local"'])
272
return (time.strftime("%a %Y-%m-%d %H:%M:%S", tt)
273
+ ' %+03d%02d' % (offset / 3600, (offset / 60) % 60))
677
raise errors.BzrError("unsupported timezone format %r" % timezone,
678
['options are "utc", "original", "local"'])
680
date_fmt = "%a %Y-%m-%d %H:%M:%S"
682
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
685
return (time.strftime(date_fmt, tt) + offset_str)
276
688
def compact_date(when):
277
689
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
692
def format_delta(delta):
693
"""Get a nice looking string for a time delta.
695
:param delta: The time difference in seconds, can be positive or negative.
696
positive indicates time in the past, negative indicates time in the
697
future. (usually time.time() - stored_time)
698
:return: String formatted to show approximate resolution
704
direction = 'in the future'
708
if seconds < 90: # print seconds up to 90 seconds
710
return '%d second %s' % (seconds, direction,)
712
return '%d seconds %s' % (seconds, direction)
714
minutes = int(seconds / 60)
715
seconds -= 60 * minutes
720
if minutes < 90: # print minutes, seconds up to 90 minutes
722
return '%d minute, %d second%s %s' % (
723
minutes, seconds, plural_seconds, direction)
725
return '%d minutes, %d second%s %s' % (
726
minutes, seconds, plural_seconds, direction)
728
hours = int(minutes / 60)
729
minutes -= 60 * hours
736
return '%d hour, %d minute%s %s' % (hours, minutes,
737
plural_minutes, direction)
738
return '%d hours, %d minute%s %s' % (hours, minutes,
739
plural_minutes, direction)
282
742
"""Return size of given open file."""
283
743
return os.fstat(f.fileno())[ST_SIZE]
286
if hasattr(os, 'urandom'): # python 2.4 and later
746
# Define rand_bytes based on platform.
748
# Python 2.4 and later have os.urandom,
749
# but it doesn't work on some arches
287
751
rand_bytes = os.urandom
289
# FIXME: No good on non-Linux
290
_rand_file = file('/dev/urandom', 'rb')
291
rand_bytes = _rand_file.read
752
except (NotImplementedError, AttributeError):
753
# If python doesn't have os.urandom, or it doesn't work,
754
# then try to first pull random data from /dev/urandom
756
rand_bytes = file('/dev/urandom', 'rb').read
757
# Otherwise, use this hack as a last resort
758
except (IOError, OSError):
759
# not well seeded, but better than nothing
764
s += chr(random.randint(0, 255))
769
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
771
"""Return a random string of num alphanumeric characters
773
The result only contains lowercase chars because it may be used on
774
case-insensitive filesystems.
777
for raw_byte in rand_bytes(num):
778
s += ALNUM[ord(raw_byte) % 36]
294
782
## TODO: We could later have path objects that remember their list
295
783
## decomposition (might be too tricksy though.)
297
785
def splitpath(p):
298
"""Turn string into list of parts.
304
>>> splitpath('a/./b')
306
>>> splitpath('a/.b')
308
>>> splitpath('a/../b')
309
Traceback (most recent call last):
311
BzrError: ("sorry, '..' not allowed in path", [])
313
assert isinstance(p, types.StringTypes)
786
"""Turn string into list of parts."""
787
assert isinstance(p, basestring)
315
789
# split on either delimiter because people might use either on
330
assert isinstance(p, list)
804
assert isinstance(p, (list, tuple))
332
if (f == '..') or (f == None) or (f == ''):
333
bailout("sorry, %r not allowed in path" % f)
334
return os.path.join(*p)
337
def appendpath(p1, p2):
341
return os.path.join(p1, p2)
344
def extern_command(cmd, ignore_errors = False):
345
mutter('external command: %s' % `cmd`)
347
if not ignore_errors:
348
bailout('command failed')
806
if (f == '..') or (f is None) or (f == ''):
807
raise errors.BzrError("sorry, %r not allowed in path" % f)
812
"""Split s into lines, but without removing the newline characters."""
813
lines = s.split('\n')
814
result = [line + '\n' for line in lines[:-1]]
816
result.append(lines[-1])
820
def hardlinks_good():
821
return sys.platform not in ('win32', 'cygwin', 'darwin')
824
def link_or_copy(src, dest):
825
"""Hardlink a file, or copy it if it can't be hardlinked."""
826
if not hardlinks_good():
827
shutil.copyfile(src, dest)
831
except (OSError, IOError), e:
832
if e.errno != errno.EXDEV:
834
shutil.copyfile(src, dest)
837
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
838
# Forgiveness than Permission (EAFP) because:
839
# - root can damage a solaris file system by using unlink,
840
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
841
# EACCES, OSX: EPERM) when invoked on a directory.
842
def delete_any(path):
843
"""Delete a file or directory."""
844
if isdir(path): # Takes care of symlinks
851
if getattr(os, 'symlink', None) is not None:
857
def contains_whitespace(s):
858
"""True if there are any whitespace characters in s."""
859
# string.whitespace can include '\xa0' in certain locales, because it is
860
# considered "non-breaking-space" as part of ISO-8859-1. But it
861
# 1) Isn't a breaking whitespace
862
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
864
# 3) '\xa0' isn't unicode safe since it is >128.
866
# This should *not* be a unicode set of characters in case the source
867
# string is not a Unicode string. We can auto-up-cast the characters since
868
# they are ascii, but we don't want to auto-up-cast the string in case it
870
for ch in ' \t\n\r\v\f':
877
def contains_linebreaks(s):
878
"""True if there is any vertical whitespace in s."""
886
def relpath(base, path):
887
"""Return path relative to base, or raise exception.
889
The path may be either an absolute path or a path relative to the
890
current working directory.
892
os.path.commonprefix (python2.4) has a bad bug that it works just
893
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
897
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
898
' exceed the platform minimum length (which is %d)' %
905
while len(head) >= len(base):
908
head, tail = os.path.split(head)
912
raise errors.PathNotChild(rp, base)
920
def safe_unicode(unicode_or_utf8_string):
921
"""Coerce unicode_or_utf8_string into unicode.
923
If it is unicode, it is returned.
924
Otherwise it is decoded from utf-8. If a decoding error
925
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
926
as a BzrBadParameter exception.
928
if isinstance(unicode_or_utf8_string, unicode):
929
return unicode_or_utf8_string
931
return unicode_or_utf8_string.decode('utf8')
932
except UnicodeDecodeError:
933
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
936
def safe_utf8(unicode_or_utf8_string):
937
"""Coerce unicode_or_utf8_string to a utf8 string.
939
If it is a str, it is returned.
940
If it is Unicode, it is encoded into a utf-8 string.
942
if isinstance(unicode_or_utf8_string, str):
943
# TODO: jam 20070209 This is overkill, and probably has an impact on
944
# performance if we are dealing with lots of apis that want a
947
# Make sure it is a valid utf-8 string
948
unicode_or_utf8_string.decode('utf-8')
949
except UnicodeDecodeError:
950
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
951
return unicode_or_utf8_string
952
return unicode_or_utf8_string.encode('utf-8')
955
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
956
' Revision id generators should be creating utf8'
960
def safe_revision_id(unicode_or_utf8_string, warn=True):
961
"""Revision ids should now be utf8, but at one point they were unicode.
963
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
965
:param warn: Functions that are sanitizing user data can set warn=False
966
:return: None or a utf8 revision id.
968
if (unicode_or_utf8_string is None
969
or unicode_or_utf8_string.__class__ == str):
970
return unicode_or_utf8_string
972
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
974
return cache_utf8.encode(unicode_or_utf8_string)
977
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
978
' generators should be creating utf8 file ids.')
981
def safe_file_id(unicode_or_utf8_string, warn=True):
982
"""File ids should now be utf8, but at one point they were unicode.
984
This is the same as safe_utf8, except it uses the cached encode functions
985
to save a little bit of performance.
987
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
989
:param warn: Functions that are sanitizing user data can set warn=False
990
:return: None or a utf8 file id.
992
if (unicode_or_utf8_string is None
993
or unicode_or_utf8_string.__class__ == str):
994
return unicode_or_utf8_string
996
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
998
return cache_utf8.encode(unicode_or_utf8_string)
1001
_platform_normalizes_filenames = False
1002
if sys.platform == 'darwin':
1003
_platform_normalizes_filenames = True
1006
def normalizes_filenames():
1007
"""Return True if this platform normalizes unicode filenames.
1009
Mac OSX does, Windows/Linux do not.
1011
return _platform_normalizes_filenames
1014
def _accessible_normalized_filename(path):
1015
"""Get the unicode normalized path, and if you can access the file.
1017
On platforms where the system normalizes filenames (Mac OSX),
1018
you can access a file by any path which will normalize correctly.
1019
On platforms where the system does not normalize filenames
1020
(Windows, Linux), you have to access a file by its exact path.
1022
Internally, bzr only supports NFC/NFKC normalization, since that is
1023
the standard for XML documents.
1025
So return the normalized path, and a flag indicating if the file
1026
can be accessed by that path.
1029
return unicodedata.normalize('NFKC', unicode(path)), True
1032
def _inaccessible_normalized_filename(path):
1033
__doc__ = _accessible_normalized_filename.__doc__
1035
normalized = unicodedata.normalize('NFKC', unicode(path))
1036
return normalized, normalized == path
1039
if _platform_normalizes_filenames:
1040
normalized_filename = _accessible_normalized_filename
1042
normalized_filename = _inaccessible_normalized_filename
1045
def terminal_width():
1046
"""Return estimated terminal width."""
1047
if sys.platform == 'win32':
1048
return win32utils.get_console_size()[0]
1051
import struct, fcntl, termios
1052
s = struct.pack('HHHH', 0, 0, 0, 0)
1053
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1054
width = struct.unpack('HHHH', x)[1]
1059
width = int(os.environ['COLUMNS'])
1068
def supports_executable():
1069
return sys.platform != "win32"
1072
def supports_posix_readonly():
1073
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1075
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1076
directory controls creation/deletion, etc.
1078
And under win32, readonly means that the directory itself cannot be
1079
deleted. The contents of a readonly directory can be changed, unlike POSIX
1080
where files in readonly directories cannot be added, deleted or renamed.
1082
return sys.platform != "win32"
1085
def set_or_unset_env(env_variable, value):
1086
"""Modify the environment, setting or removing the env_variable.
1088
:param env_variable: The environment variable in question
1089
:param value: The value to set the environment to. If None, then
1090
the variable will be removed.
1091
:return: The original value of the environment variable.
1093
orig_val = os.environ.get(env_variable)
1095
if orig_val is not None:
1096
del os.environ[env_variable]
1098
if isinstance(value, unicode):
1099
value = value.encode(bzrlib.user_encoding)
1100
os.environ[env_variable] = value
1104
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1107
def check_legal_path(path):
1108
"""Check whether the supplied path is legal.
1109
This is only required on Windows, so we don't test on other platforms
1112
if sys.platform != "win32":
1114
if _validWin32PathRE.match(path) is None:
1115
raise errors.IllegalPath(path)
1118
def walkdirs(top, prefix=""):
1119
"""Yield data about all the directories in a tree.
1121
This yields all the data about the contents of a directory at a time.
1122
After each directory has been yielded, if the caller has mutated the list
1123
to exclude some directories, they are then not descended into.
1125
The data yielded is of the form:
1126
((directory-relpath, directory-path-from-top),
1127
[(relpath, basename, kind, lstat, path-from-top), ...]),
1128
- directory-relpath is the relative path of the directory being returned
1129
with respect to top. prefix is prepended to this.
1130
- directory-path-from-root is the path including top for this directory.
1131
It is suitable for use with os functions.
1132
- relpath is the relative path within the subtree being walked.
1133
- basename is the basename of the path
1134
- kind is the kind of the file now. If unknown then the file is not
1135
present within the tree - but it may be recorded as versioned. See
1137
- lstat is the stat data *if* the file was statted.
1138
- planned, not implemented:
1139
path_from_tree_root is the path from the root of the tree.
1141
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1142
allows one to walk a subtree but get paths that are relative to a tree
1144
:return: an iterator over the dirs.
1146
#TODO there is a bit of a smell where the results of the directory-
1147
# summary in this, and the path from the root, may not agree
1148
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1149
# potentially confusing output. We should make this more robust - but
1150
# not at a speed cost. RBC 20060731
1152
_directory = _directory_kind
1153
_listdir = os.listdir
1154
_kind_from_mode = _formats.get
1155
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1157
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1158
relroot, _, _, _, top = pending.pop()
1160
relprefix = relroot + u'/'
1163
top_slash = top + u'/'
1166
append = dirblock.append
1167
for name in sorted(_listdir(top)):
1168
abspath = top_slash + name
1169
statvalue = _lstat(abspath)
1170
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1171
append((relprefix + name, name, kind, statvalue, abspath))
1172
yield (relroot, top), dirblock
1174
# push the user specified dirs from dirblock
1175
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1178
def _walkdirs_utf8(top, prefix=""):
1179
"""Yield data about all the directories in a tree.
1181
This yields the same information as walkdirs() only each entry is yielded
1182
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1183
are returned as exact byte-strings.
1185
:return: yields a tuple of (dir_info, [file_info])
1186
dir_info is (utf8_relpath, path-from-top)
1187
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1188
if top is an absolute path, path-from-top is also an absolute path.
1189
path-from-top might be unicode or utf8, but it is the correct path to
1190
pass to os functions to affect the file in question. (such as os.lstat)
1192
fs_encoding = _fs_enc.upper()
1193
if (sys.platform == 'win32' or
1194
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1195
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1197
return _walkdirs_fs_utf8(top, prefix=prefix)
1200
def _walkdirs_fs_utf8(top, prefix=""):
1201
"""See _walkdirs_utf8.
1203
This sub-function is called when we know the filesystem is already in utf8
1204
encoding. So we don't need to transcode filenames.
1207
_directory = _directory_kind
1208
_listdir = os.listdir
1209
_kind_from_mode = _formats.get
1211
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1212
# But we don't actually uses 1-3 in pending, so set them to None
1213
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1215
relroot, _, _, _, top = pending.pop()
1217
relprefix = relroot + '/'
1220
top_slash = top + '/'
1223
append = dirblock.append
1224
for name in sorted(_listdir(top)):
1225
abspath = top_slash + name
1226
statvalue = _lstat(abspath)
1227
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1228
append((relprefix + name, name, kind, statvalue, abspath))
1229
yield (relroot, top), dirblock
1231
# push the user specified dirs from dirblock
1232
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1235
def _walkdirs_unicode_to_utf8(top, prefix=""):
1236
"""See _walkdirs_utf8
1238
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1240
This is currently the fallback code path when the filesystem encoding is
1241
not UTF-8. It may be better to implement an alternative so that we can
1242
safely handle paths that are not properly decodable in the current
1245
_utf8_encode = codecs.getencoder('utf8')
1247
_directory = _directory_kind
1248
_listdir = os.listdir
1249
_kind_from_mode = _formats.get
1251
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1253
relroot, _, _, _, top = pending.pop()
1255
relprefix = relroot + '/'
1258
top_slash = top + u'/'
1261
append = dirblock.append
1262
for name in sorted(_listdir(top)):
1263
name_utf8 = _utf8_encode(name)[0]
1264
abspath = top_slash + name
1265
statvalue = _lstat(abspath)
1266
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1267
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1268
yield (relroot, top), dirblock
1270
# push the user specified dirs from dirblock
1271
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1274
def copy_tree(from_path, to_path, handlers={}):
1275
"""Copy all of the entries in from_path into to_path.
1277
:param from_path: The base directory to copy.
1278
:param to_path: The target directory. If it does not exist, it will
1280
:param handlers: A dictionary of functions, which takes a source and
1281
destinations for files, directories, etc.
1282
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1283
'file', 'directory', and 'symlink' should always exist.
1284
If they are missing, they will be replaced with 'os.mkdir()',
1285
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1287
# Now, just copy the existing cached tree to the new location
1288
# We use a cheap trick here.
1289
# Absolute paths are prefixed with the first parameter
1290
# relative paths are prefixed with the second.
1291
# So we can get both the source and target returned
1292
# without any extra work.
1294
def copy_dir(source, dest):
1297
def copy_link(source, dest):
1298
"""Copy the contents of a symlink"""
1299
link_to = os.readlink(source)
1300
os.symlink(link_to, dest)
1302
real_handlers = {'file':shutil.copy2,
1303
'symlink':copy_link,
1304
'directory':copy_dir,
1306
real_handlers.update(handlers)
1308
if not os.path.exists(to_path):
1309
real_handlers['directory'](from_path, to_path)
1311
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1312
for relpath, name, kind, st, abspath in entries:
1313
real_handlers[kind](abspath, relpath)
1316
def path_prefix_key(path):
1317
"""Generate a prefix-order path key for path.
1319
This can be used to sort paths in the same way that walkdirs does.
1321
return (dirname(path) , path)
1324
def compare_paths_prefix_order(path_a, path_b):
1325
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1326
key_a = path_prefix_key(path_a)
1327
key_b = path_prefix_key(path_b)
1328
return cmp(key_a, key_b)
1331
_cached_user_encoding = None
1334
def get_user_encoding(use_cache=True):
1335
"""Find out what the preferred user encoding is.
1337
This is generally the encoding that is used for command line parameters
1338
and file contents. This may be different from the terminal encoding
1339
or the filesystem encoding.
1341
:param use_cache: Enable cache for detected encoding.
1342
(This parameter is turned on by default,
1343
and required only for selftesting)
1345
:return: A string defining the preferred user encoding
1347
global _cached_user_encoding
1348
if _cached_user_encoding is not None and use_cache:
1349
return _cached_user_encoding
1351
if sys.platform == 'darwin':
1352
# work around egregious python 2.4 bug
1353
sys.platform = 'posix'
1357
sys.platform = 'darwin'
1362
user_encoding = locale.getpreferredencoding()
1363
except locale.Error, e:
1364
sys.stderr.write('bzr: warning: %s\n'
1365
' Could not determine what text encoding to use.\n'
1366
' This error usually means your Python interpreter\n'
1367
' doesn\'t support the locale set by $LANG (%s)\n'
1368
" Continuing with ascii encoding.\n"
1369
% (e, os.environ.get('LANG')))
1370
user_encoding = 'ascii'
1372
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1373
# treat that as ASCII, and not support printing unicode characters to the
1375
if user_encoding in (None, 'cp0'):
1376
user_encoding = 'ascii'
1380
codecs.lookup(user_encoding)
1382
sys.stderr.write('bzr: warning:'
1383
' unknown encoding %s.'
1384
' Continuing with ascii encoding.\n'
1387
user_encoding = 'ascii'
1390
_cached_user_encoding = user_encoding
1392
return user_encoding
1395
def recv_all(socket, bytes):
1396
"""Receive an exact number of bytes.
1398
Regular Socket.recv() may return less than the requested number of bytes,
1399
dependning on what's in the OS buffer. MSG_WAITALL is not available
1400
on all platforms, but this should work everywhere. This will return
1401
less than the requested amount if the remote end closes.
1403
This isn't optimized and is intended mostly for use in testing.
1406
while len(b) < bytes:
1407
new = socket.recv(bytes - len(b))
1413
def dereference_path(path):
1414
"""Determine the real path to a file.
1416
All parent elements are dereferenced. But the file itself is not
1418
:param path: The original path. May be absolute or relative.
1419
:return: the real path *to* the file
1421
parent, base = os.path.split(path)
1422
# The pathjoin for '.' is a workaround for Python bug #1213894.
1423
# (initial path components aren't dereferenced)
1424
return pathjoin(realpath(pathjoin('.', parent)), base)
1427
def supports_mapi():
1428
"""Return True if we can use MAPI to launch a mail client."""
1429
return sys.platform == "win32"
added
removed
bzrlib/osutils.py
