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, sys
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,
44
from tempfile import (
57
from bzrlib import symbol_versioning
58
from bzrlib.symbol_versioning import (
62
from bzrlib.trace import mutter
65
# On win32, O_BINARY is used to indicate the file should
66
# be opened in binary mode, rather than text mode.
67
# On other platforms, O_BINARY doesn't exist, because
68
# they always open in binary mode, so it is okay to
69
# OR with 0 on those platforms
70
O_BINARY = getattr(os, 'O_BINARY', 0)
72
# On posix, use lstat instead of stat so that we can
73
# operate on broken symlinks. On Windows revert to stat.
74
lstat = getattr(os, 'lstat', os.stat)
26
76
def make_readonly(filename):
27
77
"""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)
78
mod = lstat(filename).st_mode
79
if not stat.S_ISLNK(mod):
81
os.chmod(filename, mod)
34
84
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.,:/_~-])')
85
mod = lstat(filename).st_mode
86
if not stat.S_ISLNK(mod):
88
os.chmod(filename, mod)
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]
95
"""Return a quoted filename filename
97
This previously used backslash quoting, but that works poorly on
99
# TODO: I'm not really sure this is the best format either.x
101
if _QUOTE_RE is None:
102
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
104
if _QUOTE_RE.search(f):
59
raise BzrError("can't handle file kind with mode %o of %r" % (mode, f))
110
_directory_kind = 'directory'
113
stat.S_IFDIR:_directory_kind,
114
stat.S_IFCHR:'chardev',
115
stat.S_IFBLK:'block',
118
stat.S_IFLNK:'symlink',
119
stat.S_IFSOCK:'socket',
123
def file_kind_from_stat_mode(stat_mode, _formats=_formats, _unknown='unknown'):
124
"""Generate a file kind from a stat mode. This is used in walkdirs.
126
Its performance is critical: Do not mutate without careful benchmarking.
129
return _formats[stat_mode & 0170000]
134
def file_kind(f, _lstat=os.lstat, _mapper=file_kind_from_stat_mode):
136
return _mapper(_lstat(f).st_mode)
138
if getattr(e, 'errno', None) == errno.ENOENT:
139
raise errors.NoSuchFile(f)
144
"""Return the current umask"""
145
# Assume that people aren't messing with the umask while running
146
# XXX: This is not thread safe, but there is no way to get the
147
# umask without setting it
155
_directory_kind: "/",
157
'tree-reference': '+',
62
161
def kind_marker(kind):
65
elif kind == 'directory':
67
elif kind == 'symlink':
70
raise BzrError('invalid file kind %r' % kind)
163
return _kind_marker_map[kind]
165
raise errors.BzrError('invalid file kind %r' % kind)
168
lexists = getattr(os.path, 'lexists', None)
172
stat = getattr(os, 'lstat', os.stat)
176
if e.errno == errno.ENOENT:
179
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
182
def fancy_rename(old, new, rename_func, unlink_func):
183
"""A fancy rename, when you don't have atomic rename.
185
:param old: The old path, to rename from
186
:param new: The new path, to rename to
187
:param rename_func: The potentially non-atomic rename function
188
:param unlink_func: A way to delete the target file if the full rename succeeds
191
# sftp rename doesn't allow overwriting, so play tricks:
193
base = os.path.basename(new)
194
dirname = os.path.dirname(new)
195
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
196
tmp_name = pathjoin(dirname, tmp_name)
198
# Rename the file out of the way, but keep track if it didn't exist
199
# We don't want to grab just any exception
200
# something like EACCES should prevent us from continuing
201
# The downside is that the rename_func has to throw an exception
202
# with an errno = ENOENT, or NoSuchFile
205
rename_func(new, tmp_name)
206
except (errors.NoSuchFile,), e:
209
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
210
# function raises an IOError with errno is None when a rename fails.
211
# This then gets caught here.
212
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
215
if (getattr(e, 'errno', None) is None
216
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
223
# This may throw an exception, in which case success will
225
rename_func(old, new)
229
# If the file used to exist, rename it back into place
230
# otherwise just delete it from the tmp location
232
unlink_func(tmp_name)
234
rename_func(tmp_name, new)
237
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
238
# choke on a Unicode string containing a relative path if
239
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
241
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
242
def _posix_abspath(path):
243
# jam 20060426 rather than encoding to fsencoding
244
# copy posixpath.abspath, but use os.getcwdu instead
245
if not posixpath.isabs(path):
246
path = posixpath.join(getcwd(), path)
247
return posixpath.normpath(path)
250
def _posix_realpath(path):
251
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
254
def _win32_fixdrive(path):
255
"""Force drive letters to be consistent.
257
win32 is inconsistent whether it returns lower or upper case
258
and even if it was consistent the user might type the other
259
so we force it to uppercase
260
running python.exe under cmd.exe return capital C:\\
261
running win32 python inside a cygwin shell returns lowercase c:\\
263
drive, path = _nt_splitdrive(path)
264
return drive.upper() + path
267
def _win32_abspath(path):
268
# Real _nt_abspath doesn't have a problem with a unicode cwd
269
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
272
def _win98_abspath(path):
273
"""Return the absolute version of a path.
274
Windows 98 safe implementation (python reimplementation
275
of Win32 API function GetFullPathNameW)
280
# \\HOST\path => //HOST/path
281
# //HOST/path => //HOST/path
282
# path => C:/cwd/path
285
# check for absolute path
286
drive = _nt_splitdrive(path)[0]
287
if drive == '' and path[:2] not in('//','\\\\'):
289
# we cannot simply os.path.join cwd and path
290
# because os.path.join('C:','/path') produce '/path'
291
# and this is incorrect
292
if path[:1] in ('/','\\'):
293
cwd = _nt_splitdrive(cwd)[0]
295
path = cwd + '\\' + path
296
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
298
if win32utils.winver == 'Windows 98':
299
_win32_abspath = _win98_abspath
302
def _win32_realpath(path):
303
# Real _nt_realpath doesn't have a problem with a unicode cwd
304
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
307
def _win32_pathjoin(*args):
308
return _nt_join(*args).replace('\\', '/')
311
def _win32_normpath(path):
312
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
316
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
319
def _win32_mkdtemp(*args, **kwargs):
320
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
323
def _win32_rename(old, new):
324
"""We expect to be able to atomically replace 'new' with old.
326
On win32, if new exists, it must be moved out of the way first,
330
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
332
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
333
# If we try to rename a non-existant file onto cwd, we get
334
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
335
# if the old path doesn't exist, sometimes we get EACCES
336
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
342
return unicodedata.normalize('NFKC', os.getcwdu())
345
# Default is to just use the python builtins, but these can be rebound on
346
# particular platforms.
347
abspath = _posix_abspath
348
realpath = _posix_realpath
349
pathjoin = os.path.join
350
normpath = os.path.normpath
353
dirname = os.path.dirname
354
basename = os.path.basename
355
split = os.path.split
356
splitext = os.path.splitext
357
# These were already imported into local scope
358
# mkdtemp = tempfile.mkdtemp
359
# rmtree = shutil.rmtree
361
MIN_ABS_PATHLENGTH = 1
364
if sys.platform == 'win32':
365
abspath = _win32_abspath
366
realpath = _win32_realpath
367
pathjoin = _win32_pathjoin
368
normpath = _win32_normpath
369
getcwd = _win32_getcwd
370
mkdtemp = _win32_mkdtemp
371
rename = _win32_rename
373
MIN_ABS_PATHLENGTH = 3
375
def _win32_delete_readonly(function, path, excinfo):
376
"""Error handler for shutil.rmtree function [for win32]
377
Helps to remove files and dirs marked as read-only.
379
exception = excinfo[1]
380
if function in (os.remove, os.rmdir) \
381
and isinstance(exception, OSError) \
382
and exception.errno == errno.EACCES:
388
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
389
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
390
return shutil.rmtree(path, ignore_errors, onerror)
391
elif sys.platform == 'darwin':
395
def get_terminal_encoding():
396
"""Find the best encoding for printing to the screen.
398
This attempts to check both sys.stdout and sys.stdin to see
399
what encoding they are in, and if that fails it falls back to
400
bzrlib.user_encoding.
401
The problem is that on Windows, locale.getpreferredencoding()
402
is not the same encoding as that used by the console:
403
http://mail.python.org/pipermail/python-list/2003-May/162357.html
405
On my standard US Windows XP, the preferred encoding is
406
cp1252, but the console is cp437
408
output_encoding = getattr(sys.stdout, 'encoding', None)
409
if not output_encoding:
410
input_encoding = getattr(sys.stdin, 'encoding', None)
411
if not input_encoding:
412
output_encoding = bzrlib.user_encoding
413
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
415
output_encoding = input_encoding
416
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
418
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
419
if output_encoding == 'cp0':
420
# invalid encoding (cp0 means 'no codepage' on Windows)
421
output_encoding = bzrlib.user_encoding
422
mutter('cp0 is invalid encoding.'
423
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
426
codecs.lookup(output_encoding)
428
sys.stderr.write('bzr: warning:'
429
' unknown terminal encoding %s.\n'
430
' Using encoding %s instead.\n'
431
% (output_encoding, bzrlib.user_encoding)
433
output_encoding = bzrlib.user_encoding
435
return output_encoding
438
def normalizepath(f):
439
if getattr(os.path, 'realpath', None) is not None:
443
[p,e] = os.path.split(f)
444
if e == "" or e == "." or e == "..":
447
return pathjoin(F(p), e)
451
"""Copy a file to a backup.
453
Backups are named in GNU-style, with a ~ suffix.
455
If the file is already a backup, it's not copied.
461
if has_symlinks() and os.path.islink(fn):
462
target = os.readlink(fn)
463
os.symlink(target, bfn)
471
outf = file(bfn, 'wb')
271
606
def local_time_offset(t=None):
272
607
"""Return offset of local zone from GMT, either at present or at time t."""
273
# python2.3 localtime() can't take None
277
if time.localtime(t).tm_isdst and time.daylight:
280
return -time.timezone
283
def format_date(t, offset=0, timezone='original'):
284
## TODO: Perhaps a global option to use either universal or local time?
285
## Or perhaps just let people set $TZ?
286
assert isinstance(t, float)
610
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
611
return offset.days * 86400 + offset.seconds
614
def format_date(t, offset=0, timezone='original', date_fmt=None,
616
"""Return a formatted date string.
618
:param t: Seconds since the epoch.
619
:param offset: Timezone offset in seconds east of utc.
620
:param timezone: How to display the time: 'utc', 'original' for the
621
timezone specified by offset, or 'local' for the process's current
623
:param show_offset: Whether to append the timezone.
624
:param date_fmt: strftime format.
288
626
if timezone == 'utc':
289
627
tt = time.gmtime(t)
291
629
elif timezone == 'original':
294
632
tt = time.gmtime(t + offset)
295
633
elif timezone == 'local':
296
634
tt = time.localtime(t)
297
635
offset = local_time_offset(t)
299
bailout("unsupported timezone format %r",
300
['options are "utc", "original", "local"'])
302
return (time.strftime("%a %Y-%m-%d %H:%M:%S", tt)
303
+ ' %+03d%02d' % (offset / 3600, (offset / 60) % 60))
637
raise errors.BzrError("unsupported timezone format %r" % timezone,
638
['options are "utc", "original", "local"'])
640
date_fmt = "%a %Y-%m-%d %H:%M:%S"
642
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
645
return (time.strftime(date_fmt, tt) + offset_str)
306
648
def compact_date(when):
307
649
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
652
def format_delta(delta):
653
"""Get a nice looking string for a time delta.
655
:param delta: The time difference in seconds, can be positive or negative.
656
positive indicates time in the past, negative indicates time in the
657
future. (usually time.time() - stored_time)
658
:return: String formatted to show approximate resolution
664
direction = 'in the future'
668
if seconds < 90: # print seconds up to 90 seconds
670
return '%d second %s' % (seconds, direction,)
672
return '%d seconds %s' % (seconds, direction)
674
minutes = int(seconds / 60)
675
seconds -= 60 * minutes
680
if minutes < 90: # print minutes, seconds up to 90 minutes
682
return '%d minute, %d second%s %s' % (
683
minutes, seconds, plural_seconds, direction)
685
return '%d minutes, %d second%s %s' % (
686
minutes, seconds, plural_seconds, direction)
688
hours = int(minutes / 60)
689
minutes -= 60 * hours
696
return '%d hour, %d minute%s %s' % (hours, minutes,
697
plural_minutes, direction)
698
return '%d hours, %d minute%s %s' % (hours, minutes,
699
plural_minutes, direction)
312
702
"""Return size of given open file."""
313
703
return os.fstat(f.fileno())[ST_SIZE]
316
if hasattr(os, 'urandom'): # python 2.4 and later
706
# Define rand_bytes based on platform.
708
# Python 2.4 and later have os.urandom,
709
# but it doesn't work on some arches
317
711
rand_bytes = os.urandom
318
elif sys.platform == 'linux2':
319
rand_bytes = file('/dev/urandom', 'rb').read
321
# not well seeded, but better than nothing
326
s += chr(random.randint(0, 255))
712
except (NotImplementedError, AttributeError):
713
# If python doesn't have os.urandom, or it doesn't work,
714
# then try to first pull random data from /dev/urandom
716
rand_bytes = file('/dev/urandom', 'rb').read
717
# Otherwise, use this hack as a last resort
718
except (IOError, OSError):
719
# not well seeded, but better than nothing
724
s += chr(random.randint(0, 255))
729
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
731
"""Return a random string of num alphanumeric characters
733
The result only contains lowercase chars because it may be used on
734
case-insensitive filesystems.
737
for raw_byte in rand_bytes(num):
738
s += ALNUM[ord(raw_byte) % 36]
331
742
## TODO: We could later have path objects that remember their list
332
743
## decomposition (might be too tricksy though.)
334
745
def splitpath(p):
335
"""Turn string into list of parts.
341
>>> splitpath('a/./b')
343
>>> splitpath('a/.b')
345
>>> splitpath('a/../b')
346
Traceback (most recent call last):
348
BzrError: ("sorry, '..' not allowed in path", [])
350
assert isinstance(p, types.StringTypes)
746
"""Turn string into list of parts."""
747
assert isinstance(p, basestring)
352
749
# split on either delimiter because people might use either on
367
assert isinstance(p, list)
764
assert isinstance(p, (list, tuple))
369
if (f == '..') or (f == None) or (f == ''):
370
bailout("sorry, %r not allowed in path" % f)
371
return os.path.join(*p)
766
if (f == '..') or (f is None) or (f == ''):
767
raise errors.BzrError("sorry, %r not allowed in path" % f)
771
@deprecated_function(zero_nine)
374
772
def appendpath(p1, p2):
378
return os.path.join(p1, p2)
381
def extern_command(cmd, ignore_errors = False):
382
mutter('external command: %s' % `cmd`)
384
if not ignore_errors:
385
bailout('command failed')
776
return pathjoin(p1, p2)
780
"""Split s into lines, but without removing the newline characters."""
781
lines = s.split('\n')
782
result = [line + '\n' for line in lines[:-1]]
784
result.append(lines[-1])
788
def hardlinks_good():
789
return sys.platform not in ('win32', 'cygwin', 'darwin')
792
def link_or_copy(src, dest):
793
"""Hardlink a file, or copy it if it can't be hardlinked."""
794
if not hardlinks_good():
795
shutil.copyfile(src, dest)
799
except (OSError, IOError), e:
800
if e.errno != errno.EXDEV:
802
shutil.copyfile(src, dest)
804
def delete_any(full_path):
805
"""Delete a file or directory."""
809
# We may be renaming a dangling inventory id
810
if e.errno not in (errno.EISDIR, errno.EACCES, errno.EPERM):
816
if getattr(os, 'symlink', None) is not None:
822
def contains_whitespace(s):
823
"""True if there are any whitespace characters in s."""
824
# string.whitespace can include '\xa0' in certain locales, because it is
825
# considered "non-breaking-space" as part of ISO-8859-1. But it
826
# 1) Isn't a breaking whitespace
827
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
829
# 3) '\xa0' isn't unicode safe since it is >128.
831
# This should *not* be a unicode set of characters in case the source
832
# string is not a Unicode string. We can auto-up-cast the characters since
833
# they are ascii, but we don't want to auto-up-cast the string in case it
835
for ch in ' \t\n\r\v\f':
842
def contains_linebreaks(s):
843
"""True if there is any vertical whitespace in s."""
851
def relpath(base, path):
852
"""Return path relative to base, or raise exception.
854
The path may be either an absolute path or a path relative to the
855
current working directory.
857
os.path.commonprefix (python2.4) has a bad bug that it works just
858
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
862
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
863
' exceed the platform minimum length (which is %d)' %
870
while len(head) >= len(base):
873
head, tail = os.path.split(head)
877
raise errors.PathNotChild(rp, base)
885
def safe_unicode(unicode_or_utf8_string):
886
"""Coerce unicode_or_utf8_string into unicode.
888
If it is unicode, it is returned.
889
Otherwise it is decoded from utf-8. If a decoding error
890
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
891
as a BzrBadParameter exception.
893
if isinstance(unicode_or_utf8_string, unicode):
894
return unicode_or_utf8_string
896
return unicode_or_utf8_string.decode('utf8')
897
except UnicodeDecodeError:
898
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
901
def safe_utf8(unicode_or_utf8_string):
902
"""Coerce unicode_or_utf8_string to a utf8 string.
904
If it is a str, it is returned.
905
If it is Unicode, it is encoded into a utf-8 string.
907
if isinstance(unicode_or_utf8_string, str):
908
# TODO: jam 20070209 This is overkill, and probably has an impact on
909
# performance if we are dealing with lots of apis that want a
912
# Make sure it is a valid utf-8 string
913
unicode_or_utf8_string.decode('utf-8')
914
except UnicodeDecodeError:
915
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
916
return unicode_or_utf8_string
917
return unicode_or_utf8_string.encode('utf-8')
920
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
921
' Revision id generators should be creating utf8'
925
def safe_revision_id(unicode_or_utf8_string, warn=True):
926
"""Revision ids should now be utf8, but at one point they were unicode.
928
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
930
:param warn: Functions that are sanitizing user data can set warn=False
931
:return: None or a utf8 revision id.
933
if (unicode_or_utf8_string is None
934
or unicode_or_utf8_string.__class__ == str):
935
return unicode_or_utf8_string
937
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
939
return cache_utf8.encode(unicode_or_utf8_string)
942
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
943
' generators should be creating utf8 file ids.')
946
def safe_file_id(unicode_or_utf8_string, warn=True):
947
"""File ids should now be utf8, but at one point they were unicode.
949
This is the same as safe_utf8, except it uses the cached encode functions
950
to save a little bit of performance.
952
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
954
:param warn: Functions that are sanitizing user data can set warn=False
955
:return: None or a utf8 file id.
957
if (unicode_or_utf8_string is None
958
or unicode_or_utf8_string.__class__ == str):
959
return unicode_or_utf8_string
961
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
963
return cache_utf8.encode(unicode_or_utf8_string)
966
_platform_normalizes_filenames = False
967
if sys.platform == 'darwin':
968
_platform_normalizes_filenames = True
971
def normalizes_filenames():
972
"""Return True if this platform normalizes unicode filenames.
974
Mac OSX does, Windows/Linux do not.
976
return _platform_normalizes_filenames
979
def _accessible_normalized_filename(path):
980
"""Get the unicode normalized path, and if you can access the file.
982
On platforms where the system normalizes filenames (Mac OSX),
983
you can access a file by any path which will normalize correctly.
984
On platforms where the system does not normalize filenames
985
(Windows, Linux), you have to access a file by its exact path.
987
Internally, bzr only supports NFC/NFKC normalization, since that is
988
the standard for XML documents.
990
So return the normalized path, and a flag indicating if the file
991
can be accessed by that path.
994
return unicodedata.normalize('NFKC', unicode(path)), True
997
def _inaccessible_normalized_filename(path):
998
__doc__ = _accessible_normalized_filename.__doc__
1000
normalized = unicodedata.normalize('NFKC', unicode(path))
1001
return normalized, normalized == path
1004
if _platform_normalizes_filenames:
1005
normalized_filename = _accessible_normalized_filename
1007
normalized_filename = _inaccessible_normalized_filename
1010
def terminal_width():
1011
"""Return estimated terminal width."""
1012
if sys.platform == 'win32':
1013
return win32utils.get_console_size()[0]
1016
import struct, fcntl, termios
1017
s = struct.pack('HHHH', 0, 0, 0, 0)
1018
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1019
width = struct.unpack('HHHH', x)[1]
1024
width = int(os.environ['COLUMNS'])
1033
def supports_executable():
1034
return sys.platform != "win32"
1037
def supports_posix_readonly():
1038
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1040
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1041
directory controls creation/deletion, etc.
1043
And under win32, readonly means that the directory itself cannot be
1044
deleted. The contents of a readonly directory can be changed, unlike POSIX
1045
where files in readonly directories cannot be added, deleted or renamed.
1047
return sys.platform != "win32"
1050
def set_or_unset_env(env_variable, value):
1051
"""Modify the environment, setting or removing the env_variable.
1053
:param env_variable: The environment variable in question
1054
:param value: The value to set the environment to. If None, then
1055
the variable will be removed.
1056
:return: The original value of the environment variable.
1058
orig_val = os.environ.get(env_variable)
1060
if orig_val is not None:
1061
del os.environ[env_variable]
1063
if isinstance(value, unicode):
1064
value = value.encode(bzrlib.user_encoding)
1065
os.environ[env_variable] = value
1069
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1072
def check_legal_path(path):
1073
"""Check whether the supplied path is legal.
1074
This is only required on Windows, so we don't test on other platforms
1077
if sys.platform != "win32":
1079
if _validWin32PathRE.match(path) is None:
1080
raise errors.IllegalPath(path)
1083
def walkdirs(top, prefix=""):
1084
"""Yield data about all the directories in a tree.
1086
This yields all the data about the contents of a directory at a time.
1087
After each directory has been yielded, if the caller has mutated the list
1088
to exclude some directories, they are then not descended into.
1090
The data yielded is of the form:
1091
((directory-relpath, directory-path-from-top),
1092
[(directory-relpath, basename, kind, lstat, path-from-top), ...]),
1093
- directory-relpath is the relative path of the directory being returned
1094
with respect to top. prefix is prepended to this.
1095
- directory-path-from-root is the path including top for this directory.
1096
It is suitable for use with os functions.
1097
- relpath is the relative path within the subtree being walked.
1098
- basename is the basename of the path
1099
- kind is the kind of the file now. If unknown then the file is not
1100
present within the tree - but it may be recorded as versioned. See
1102
- lstat is the stat data *if* the file was statted.
1103
- planned, not implemented:
1104
path_from_tree_root is the path from the root of the tree.
1106
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1107
allows one to walk a subtree but get paths that are relative to a tree
1109
:return: an iterator over the dirs.
1111
#TODO there is a bit of a smell where the results of the directory-
1112
# summary in this, and the path from the root, may not agree
1113
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1114
# potentially confusing output. We should make this more robust - but
1115
# not at a speed cost. RBC 20060731
1117
_directory = _directory_kind
1118
_listdir = os.listdir
1119
_kind_from_mode = _formats.get
1120
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1122
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1123
relroot, _, _, _, top = pending.pop()
1125
relprefix = relroot + u'/'
1128
top_slash = top + u'/'
1131
append = dirblock.append
1132
for name in sorted(_listdir(top)):
1133
abspath = top_slash + name
1134
statvalue = _lstat(abspath)
1135
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1136
append((relprefix + name, name, kind, statvalue, abspath))
1137
yield (relroot, top), dirblock
1139
# push the user specified dirs from dirblock
1140
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1143
def _walkdirs_utf8(top, prefix=""):
1144
"""Yield data about all the directories in a tree.
1146
This yields the same information as walkdirs() only each entry is yielded
1147
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1148
are returned as exact byte-strings.
1150
:return: yields a tuple of (dir_info, [file_info])
1151
dir_info is (utf8_relpath, path-from-top)
1152
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1153
if top is an absolute path, path-from-top is also an absolute path.
1154
path-from-top might be unicode or utf8, but it is the correct path to
1155
pass to os functions to affect the file in question. (such as os.lstat)
1157
fs_encoding = _fs_enc.upper()
1158
if (sys.platform == 'win32' or
1159
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1160
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1162
return _walkdirs_fs_utf8(top, prefix=prefix)
1165
def _walkdirs_fs_utf8(top, prefix=""):
1166
"""See _walkdirs_utf8.
1168
This sub-function is called when we know the filesystem is already in utf8
1169
encoding. So we don't need to transcode filenames.
1172
_directory = _directory_kind
1173
_listdir = os.listdir
1174
_kind_from_mode = _formats.get
1176
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1177
# But we don't actually uses 1-3 in pending, so set them to None
1178
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1180
relroot, _, _, _, top = pending.pop()
1182
relprefix = relroot + '/'
1185
top_slash = top + '/'
1188
append = dirblock.append
1189
for name in sorted(_listdir(top)):
1190
abspath = top_slash + name
1191
statvalue = _lstat(abspath)
1192
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1193
append((relprefix + name, name, kind, statvalue, abspath))
1194
yield (relroot, top), dirblock
1196
# push the user specified dirs from dirblock
1197
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1200
def _walkdirs_unicode_to_utf8(top, prefix=""):
1201
"""See _walkdirs_utf8
1203
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1205
This is currently the fallback code path when the filesystem encoding is
1206
not UTF-8. It may be better to implement an alternative so that we can
1207
safely handle paths that are not properly decodable in the current
1210
_utf8_encode = codecs.getencoder('utf8')
1212
_directory = _directory_kind
1213
_listdir = os.listdir
1214
_kind_from_mode = _formats.get
1216
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1218
relroot, _, _, _, top = pending.pop()
1220
relprefix = relroot + '/'
1223
top_slash = top + u'/'
1226
append = dirblock.append
1227
for name in sorted(_listdir(top)):
1228
name_utf8 = _utf8_encode(name)[0]
1229
abspath = top_slash + name
1230
statvalue = _lstat(abspath)
1231
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1232
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1233
yield (relroot, top), dirblock
1235
# push the user specified dirs from dirblock
1236
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1239
def copy_tree(from_path, to_path, handlers={}):
1240
"""Copy all of the entries in from_path into to_path.
1242
:param from_path: The base directory to copy.
1243
:param to_path: The target directory. If it does not exist, it will
1245
:param handlers: A dictionary of functions, which takes a source and
1246
destinations for files, directories, etc.
1247
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1248
'file', 'directory', and 'symlink' should always exist.
1249
If they are missing, they will be replaced with 'os.mkdir()',
1250
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1252
# Now, just copy the existing cached tree to the new location
1253
# We use a cheap trick here.
1254
# Absolute paths are prefixed with the first parameter
1255
# relative paths are prefixed with the second.
1256
# So we can get both the source and target returned
1257
# without any extra work.
1259
def copy_dir(source, dest):
1262
def copy_link(source, dest):
1263
"""Copy the contents of a symlink"""
1264
link_to = os.readlink(source)
1265
os.symlink(link_to, dest)
1267
real_handlers = {'file':shutil.copy2,
1268
'symlink':copy_link,
1269
'directory':copy_dir,
1271
real_handlers.update(handlers)
1273
if not os.path.exists(to_path):
1274
real_handlers['directory'](from_path, to_path)
1276
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1277
for relpath, name, kind, st, abspath in entries:
1278
real_handlers[kind](abspath, relpath)
1281
def path_prefix_key(path):
1282
"""Generate a prefix-order path key for path.
1284
This can be used to sort paths in the same way that walkdirs does.
1286
return (dirname(path) , path)
1289
def compare_paths_prefix_order(path_a, path_b):
1290
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1291
key_a = path_prefix_key(path_a)
1292
key_b = path_prefix_key(path_b)
1293
return cmp(key_a, key_b)
1296
_cached_user_encoding = None
1299
def get_user_encoding(use_cache=True):
1300
"""Find out what the preferred user encoding is.
1302
This is generally the encoding that is used for command line parameters
1303
and file contents. This may be different from the terminal encoding
1304
or the filesystem encoding.
1306
:param use_cache: Enable cache for detected encoding.
1307
(This parameter is turned on by default,
1308
and required only for selftesting)
1310
:return: A string defining the preferred user encoding
1312
global _cached_user_encoding
1313
if _cached_user_encoding is not None and use_cache:
1314
return _cached_user_encoding
1316
if sys.platform == 'darwin':
1317
# work around egregious python 2.4 bug
1318
sys.platform = 'posix'
1322
sys.platform = 'darwin'
1327
user_encoding = locale.getpreferredencoding()
1328
except locale.Error, e:
1329
sys.stderr.write('bzr: warning: %s\n'
1330
' Could not determine what text encoding to use.\n'
1331
' This error usually means your Python interpreter\n'
1332
' doesn\'t support the locale set by $LANG (%s)\n'
1333
" Continuing with ascii encoding.\n"
1334
% (e, os.environ.get('LANG')))
1335
user_encoding = 'ascii'
1337
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1338
# treat that as ASCII, and not support printing unicode characters to the
1340
if user_encoding in (None, 'cp0'):
1341
user_encoding = 'ascii'
1345
codecs.lookup(user_encoding)
1347
sys.stderr.write('bzr: warning:'
1348
' unknown encoding %s.'
1349
' Continuing with ascii encoding.\n'
1352
user_encoding = 'ascii'
1355
_cached_user_encoding = user_encoding
1357
return user_encoding
1360
def recv_all(socket, bytes):
1361
"""Receive an exact number of bytes.
1363
Regular Socket.recv() may return less than the requested number of bytes,
1364
dependning on what's in the OS buffer. MSG_WAITALL is not available
1365
on all platforms, but this should work everywhere. This will return
1366
less than the requested amount if the remote end closes.
1368
This isn't optimized and is intended mostly for use in testing.
1371
while len(b) < bytes:
1372
new = socket.recv(bytes - len(b))
1378
def dereference_path(path):
1379
"""Determine the real path to a file.
1381
All parent elements are dereferenced. But the file itself is not
1383
:param path: The original path. May be absolute or relative.
1384
:return: the real path *to* the file
1386
parent, base = os.path.split(path)
1387
# The pathjoin for '.' is a workaround for Python bug #1213894.
1388
# (initial path components aren't dereferenced)
1389
return pathjoin(realpath(pathjoin('.', parent)), base)
added
removed
bzrlib/osutils.py
