1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
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)
76
def make_readonly(filename):
77
"""Make a filename read-only."""
78
mod = lstat(filename).st_mode
79
if not stat.S_ISLNK(mod):
81
os.chmod(filename, mod)
84
def make_writable(filename):
85
mod = lstat(filename).st_mode
86
if not stat.S_ISLNK(mod):
88
os.chmod(filename, mod)
91
def minimum_path_selection(paths):
92
"""Return the smallset subset of paths which are outside paths.
94
:param paths: A container (and hence not None) of paths.
95
:return: A set of paths sufficient to include everything in paths via
96
is_inside_any, drawn from the paths parameter.
101
other_paths = paths.difference([path])
102
if not is_inside_any(other_paths, path):
103
# this is a top level path, we must check it.
104
search_paths.add(path)
112
"""Return a quoted filename filename
114
This previously used backslash quoting, but that works poorly on
116
# TODO: I'm not really sure this is the best format either.x
118
if _QUOTE_RE is None:
119
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
121
if _QUOTE_RE.search(f):
127
_directory_kind = 'directory'
130
stat.S_IFDIR:_directory_kind,
131
stat.S_IFCHR:'chardev',
132
stat.S_IFBLK:'block',
135
stat.S_IFLNK:'symlink',
136
stat.S_IFSOCK:'socket',
140
def file_kind_from_stat_mode(stat_mode, _formats=_formats, _unknown='unknown'):
141
"""Generate a file kind from a stat mode. This is used in walkdirs.
143
Its performance is critical: Do not mutate without careful benchmarking.
146
return _formats[stat_mode & 0170000]
151
def file_kind(f, _lstat=os.lstat, _mapper=file_kind_from_stat_mode):
153
return _mapper(_lstat(f).st_mode)
155
if getattr(e, 'errno', None) == errno.ENOENT:
156
raise errors.NoSuchFile(f)
161
"""Return the current umask"""
162
# Assume that people aren't messing with the umask while running
163
# XXX: This is not thread safe, but there is no way to get the
164
# umask without setting it
172
_directory_kind: "/",
174
'tree-reference': '+',
178
def kind_marker(kind):
180
return _kind_marker_map[kind]
182
raise errors.BzrError('invalid file kind %r' % kind)
185
lexists = getattr(os.path, 'lexists', None)
189
stat = getattr(os, 'lstat', os.stat)
193
if e.errno == errno.ENOENT:
196
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
199
def fancy_rename(old, new, rename_func, unlink_func):
200
"""A fancy rename, when you don't have atomic rename.
202
:param old: The old path, to rename from
203
:param new: The new path, to rename to
204
:param rename_func: The potentially non-atomic rename function
205
:param unlink_func: A way to delete the target file if the full rename succeeds
208
# sftp rename doesn't allow overwriting, so play tricks:
210
base = os.path.basename(new)
211
dirname = os.path.dirname(new)
212
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
213
tmp_name = pathjoin(dirname, tmp_name)
215
# Rename the file out of the way, but keep track if it didn't exist
216
# We don't want to grab just any exception
217
# something like EACCES should prevent us from continuing
218
# The downside is that the rename_func has to throw an exception
219
# with an errno = ENOENT, or NoSuchFile
222
rename_func(new, tmp_name)
223
except (errors.NoSuchFile,), e:
226
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
227
# function raises an IOError with errno is None when a rename fails.
228
# This then gets caught here.
229
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
232
if (getattr(e, 'errno', None) is None
233
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
240
# This may throw an exception, in which case success will
242
rename_func(old, new)
246
# If the file used to exist, rename it back into place
247
# otherwise just delete it from the tmp location
249
unlink_func(tmp_name)
251
rename_func(tmp_name, new)
254
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
255
# choke on a Unicode string containing a relative path if
256
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
258
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
259
def _posix_abspath(path):
260
# jam 20060426 rather than encoding to fsencoding
261
# copy posixpath.abspath, but use os.getcwdu instead
262
if not posixpath.isabs(path):
263
path = posixpath.join(getcwd(), path)
264
return posixpath.normpath(path)
267
def _posix_realpath(path):
268
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
271
def _win32_fixdrive(path):
272
"""Force drive letters to be consistent.
274
win32 is inconsistent whether it returns lower or upper case
275
and even if it was consistent the user might type the other
276
so we force it to uppercase
277
running python.exe under cmd.exe return capital C:\\
278
running win32 python inside a cygwin shell returns lowercase c:\\
280
drive, path = _nt_splitdrive(path)
281
return drive.upper() + path
284
def _win32_abspath(path):
285
# Real _nt_abspath doesn't have a problem with a unicode cwd
286
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
289
def _win98_abspath(path):
290
"""Return the absolute version of a path.
291
Windows 98 safe implementation (python reimplementation
292
of Win32 API function GetFullPathNameW)
297
# \\HOST\path => //HOST/path
298
# //HOST/path => //HOST/path
299
# path => C:/cwd/path
302
# check for absolute path
303
drive = _nt_splitdrive(path)[0]
304
if drive == '' and path[:2] not in('//','\\\\'):
306
# we cannot simply os.path.join cwd and path
307
# because os.path.join('C:','/path') produce '/path'
308
# and this is incorrect
309
if path[:1] in ('/','\\'):
310
cwd = _nt_splitdrive(cwd)[0]
312
path = cwd + '\\' + path
313
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
315
if win32utils.winver == 'Windows 98':
316
_win32_abspath = _win98_abspath
319
def _win32_realpath(path):
320
# Real _nt_realpath doesn't have a problem with a unicode cwd
321
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
324
def _win32_pathjoin(*args):
325
return _nt_join(*args).replace('\\', '/')
328
def _win32_normpath(path):
329
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
333
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
336
def _win32_mkdtemp(*args, **kwargs):
337
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
340
def _win32_rename(old, new):
341
"""We expect to be able to atomically replace 'new' with old.
343
On win32, if new exists, it must be moved out of the way first,
347
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
349
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
350
# If we try to rename a non-existant file onto cwd, we get
351
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
352
# if the old path doesn't exist, sometimes we get EACCES
353
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
359
return unicodedata.normalize('NFKC', os.getcwdu())
362
# Default is to just use the python builtins, but these can be rebound on
363
# particular platforms.
364
abspath = _posix_abspath
365
realpath = _posix_realpath
366
pathjoin = os.path.join
367
normpath = os.path.normpath
370
dirname = os.path.dirname
371
basename = os.path.basename
372
split = os.path.split
373
splitext = os.path.splitext
374
# These were already imported into local scope
375
# mkdtemp = tempfile.mkdtemp
376
# rmtree = shutil.rmtree
378
MIN_ABS_PATHLENGTH = 1
381
if sys.platform == 'win32':
382
abspath = _win32_abspath
383
realpath = _win32_realpath
384
pathjoin = _win32_pathjoin
385
normpath = _win32_normpath
386
getcwd = _win32_getcwd
387
mkdtemp = _win32_mkdtemp
388
rename = _win32_rename
390
MIN_ABS_PATHLENGTH = 3
392
def _win32_delete_readonly(function, path, excinfo):
393
"""Error handler for shutil.rmtree function [for win32]
394
Helps to remove files and dirs marked as read-only.
396
exception = excinfo[1]
397
if function in (os.remove, os.rmdir) \
398
and isinstance(exception, OSError) \
399
and exception.errno == errno.EACCES:
405
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
406
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
407
return shutil.rmtree(path, ignore_errors, onerror)
408
elif sys.platform == 'darwin':
412
def get_terminal_encoding():
413
"""Find the best encoding for printing to the screen.
415
This attempts to check both sys.stdout and sys.stdin to see
416
what encoding they are in, and if that fails it falls back to
417
bzrlib.user_encoding.
418
The problem is that on Windows, locale.getpreferredencoding()
419
is not the same encoding as that used by the console:
420
http://mail.python.org/pipermail/python-list/2003-May/162357.html
422
On my standard US Windows XP, the preferred encoding is
423
cp1252, but the console is cp437
425
output_encoding = getattr(sys.stdout, 'encoding', None)
426
if not output_encoding:
427
input_encoding = getattr(sys.stdin, 'encoding', None)
428
if not input_encoding:
429
output_encoding = bzrlib.user_encoding
430
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
432
output_encoding = input_encoding
433
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
435
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
436
if output_encoding == 'cp0':
437
# invalid encoding (cp0 means 'no codepage' on Windows)
438
output_encoding = bzrlib.user_encoding
439
mutter('cp0 is invalid encoding.'
440
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
443
codecs.lookup(output_encoding)
445
sys.stderr.write('bzr: warning:'
446
' unknown terminal encoding %s.\n'
447
' Using encoding %s instead.\n'
448
% (output_encoding, bzrlib.user_encoding)
450
output_encoding = bzrlib.user_encoding
452
return output_encoding
455
def normalizepath(f):
456
if getattr(os.path, 'realpath', None) is not None:
460
[p,e] = os.path.split(f)
461
if e == "" or e == "." or e == "..":
464
return pathjoin(F(p), e)
467
@deprecated_function(zero_ninetythree)
469
"""Copy a file to a backup.
471
Backups are named in GNU-style, with a ~ suffix.
473
If the file is already a backup, it's not copied.
479
if has_symlinks() and os.path.islink(fn):
480
target = os.readlink(fn)
481
os.symlink(target, bfn)
489
outf = file(bfn, 'wb')
497
"""True if f is an accessible directory."""
499
return S_ISDIR(os.lstat(f)[ST_MODE])
505
"""True if f is a regular file."""
507
return S_ISREG(os.lstat(f)[ST_MODE])
512
"""True if f is a symlink."""
514
return S_ISLNK(os.lstat(f)[ST_MODE])
518
def is_inside(dir, fname):
519
"""True if fname is inside dir.
521
The parameters should typically be passed to osutils.normpath first, so
522
that . and .. and repeated slashes are eliminated, and the separators
523
are canonical for the platform.
525
The empty string as a dir name is taken as top-of-tree and matches
528
# XXX: Most callers of this can actually do something smarter by
529
# looking at the inventory
539
return fname.startswith(dir)
542
def is_inside_any(dir_list, fname):
543
"""True if fname is inside any of given dirs."""
544
for dirname in dir_list:
545
if is_inside(dirname, fname):
550
def is_inside_or_parent_of_any(dir_list, fname):
551
"""True if fname is a child or a parent of any of the given files."""
552
for dirname in dir_list:
553
if is_inside(dirname, fname) or is_inside(fname, dirname):
558
def pumpfile(fromfile, tofile):
559
"""Copy contents of one file to another.
561
:return: The number of bytes copied.
566
b = fromfile.read(BUFSIZE)
574
def file_iterator(input_file, readsize=32768):
576
b = input_file.read(readsize)
583
if getattr(f, 'tell', None) is not None:
595
def sha_file_by_name(fname):
596
"""Calculate the SHA1 of a file by reading the full text"""
598
f = os.open(fname, os.O_RDONLY | O_BINARY)
601
b = os.read(f, 1<<16)
609
def sha_strings(strings, _factory=sha.new):
610
"""Return the sha-1 of concatenation of strings"""
612
map(s.update, strings)
616
def sha_string(f, _factory=sha.new):
617
return _factory(f).hexdigest()
620
def fingerprint_file(f):
622
return {'size': len(b),
623
'sha1': sha.new(b).hexdigest()}
626
def compare_files(a, b):
627
"""Returns true if equal in contents"""
638
def local_time_offset(t=None):
639
"""Return offset of local zone from GMT, either at present or at time t."""
642
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
643
return offset.days * 86400 + offset.seconds
646
def format_date(t, offset=0, timezone='original', date_fmt=None,
648
"""Return a formatted date string.
650
:param t: Seconds since the epoch.
651
:param offset: Timezone offset in seconds east of utc.
652
:param timezone: How to display the time: 'utc', 'original' for the
653
timezone specified by offset, or 'local' for the process's current
655
:param show_offset: Whether to append the timezone.
656
:param date_fmt: strftime format.
658
if timezone == 'utc':
661
elif timezone == 'original':
664
tt = time.gmtime(t + offset)
665
elif timezone == 'local':
666
tt = time.localtime(t)
667
offset = local_time_offset(t)
669
raise errors.BzrError("unsupported timezone format %r" % timezone,
670
['options are "utc", "original", "local"'])
672
date_fmt = "%a %Y-%m-%d %H:%M:%S"
674
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
677
return (time.strftime(date_fmt, tt) + offset_str)
680
def compact_date(when):
681
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
684
def format_delta(delta):
685
"""Get a nice looking string for a time delta.
687
:param delta: The time difference in seconds, can be positive or negative.
688
positive indicates time in the past, negative indicates time in the
689
future. (usually time.time() - stored_time)
690
:return: String formatted to show approximate resolution
696
direction = 'in the future'
700
if seconds < 90: # print seconds up to 90 seconds
702
return '%d second %s' % (seconds, direction,)
704
return '%d seconds %s' % (seconds, direction)
706
minutes = int(seconds / 60)
707
seconds -= 60 * minutes
712
if minutes < 90: # print minutes, seconds up to 90 minutes
714
return '%d minute, %d second%s %s' % (
715
minutes, seconds, plural_seconds, direction)
717
return '%d minutes, %d second%s %s' % (
718
minutes, seconds, plural_seconds, direction)
720
hours = int(minutes / 60)
721
minutes -= 60 * hours
728
return '%d hour, %d minute%s %s' % (hours, minutes,
729
plural_minutes, direction)
730
return '%d hours, %d minute%s %s' % (hours, minutes,
731
plural_minutes, direction)
734
"""Return size of given open file."""
735
return os.fstat(f.fileno())[ST_SIZE]
738
# Define rand_bytes based on platform.
740
# Python 2.4 and later have os.urandom,
741
# but it doesn't work on some arches
743
rand_bytes = os.urandom
744
except (NotImplementedError, AttributeError):
745
# If python doesn't have os.urandom, or it doesn't work,
746
# then try to first pull random data from /dev/urandom
748
rand_bytes = file('/dev/urandom', 'rb').read
749
# Otherwise, use this hack as a last resort
750
except (IOError, OSError):
751
# not well seeded, but better than nothing
756
s += chr(random.randint(0, 255))
761
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
763
"""Return a random string of num alphanumeric characters
765
The result only contains lowercase chars because it may be used on
766
case-insensitive filesystems.
769
for raw_byte in rand_bytes(num):
770
s += ALNUM[ord(raw_byte) % 36]
774
## TODO: We could later have path objects that remember their list
775
## decomposition (might be too tricksy though.)
778
"""Turn string into list of parts."""
779
assert isinstance(p, basestring)
781
# split on either delimiter because people might use either on
783
ps = re.split(r'[\\/]', p)
788
raise errors.BzrError("sorry, %r not allowed in path" % f)
789
elif (f == '.') or (f == ''):
796
assert isinstance(p, (list, tuple))
798
if (f == '..') or (f is None) or (f == ''):
799
raise errors.BzrError("sorry, %r not allowed in path" % f)
804
"""Split s into lines, but without removing the newline characters."""
805
lines = s.split('\n')
806
result = [line + '\n' for line in lines[:-1]]
808
result.append(lines[-1])
812
def hardlinks_good():
813
return sys.platform not in ('win32', 'cygwin', 'darwin')
816
def link_or_copy(src, dest):
817
"""Hardlink a file, or copy it if it can't be hardlinked."""
818
if not hardlinks_good():
819
shutil.copyfile(src, dest)
823
except (OSError, IOError), e:
824
if e.errno != errno.EXDEV:
826
shutil.copyfile(src, dest)
829
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
830
# Forgiveness than Permission (EAFP) because:
831
# - root can damage a solaris file system by using unlink,
832
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
833
# EACCES, OSX: EPERM) when invoked on a directory.
834
def delete_any(path):
835
"""Delete a file or directory."""
836
if isdir(path): # Takes care of symlinks
843
if getattr(os, 'symlink', None) is not None:
849
def contains_whitespace(s):
850
"""True if there are any whitespace characters in s."""
851
# string.whitespace can include '\xa0' in certain locales, because it is
852
# considered "non-breaking-space" as part of ISO-8859-1. But it
853
# 1) Isn't a breaking whitespace
854
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
856
# 3) '\xa0' isn't unicode safe since it is >128.
858
# This should *not* be a unicode set of characters in case the source
859
# string is not a Unicode string. We can auto-up-cast the characters since
860
# they are ascii, but we don't want to auto-up-cast the string in case it
862
for ch in ' \t\n\r\v\f':
869
def contains_linebreaks(s):
870
"""True if there is any vertical whitespace in s."""
878
def relpath(base, path):
879
"""Return path relative to base, or raise exception.
881
The path may be either an absolute path or a path relative to the
882
current working directory.
884
os.path.commonprefix (python2.4) has a bad bug that it works just
885
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
889
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
890
' exceed the platform minimum length (which is %d)' %
897
while len(head) >= len(base):
900
head, tail = os.path.split(head)
904
raise errors.PathNotChild(rp, base)
912
def safe_unicode(unicode_or_utf8_string):
913
"""Coerce unicode_or_utf8_string into unicode.
915
If it is unicode, it is returned.
916
Otherwise it is decoded from utf-8. If a decoding error
917
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
918
as a BzrBadParameter exception.
920
if isinstance(unicode_or_utf8_string, unicode):
921
return unicode_or_utf8_string
923
return unicode_or_utf8_string.decode('utf8')
924
except UnicodeDecodeError:
925
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
928
def safe_utf8(unicode_or_utf8_string):
929
"""Coerce unicode_or_utf8_string to a utf8 string.
931
If it is a str, it is returned.
932
If it is Unicode, it is encoded into a utf-8 string.
934
if isinstance(unicode_or_utf8_string, str):
935
# TODO: jam 20070209 This is overkill, and probably has an impact on
936
# performance if we are dealing with lots of apis that want a
939
# Make sure it is a valid utf-8 string
940
unicode_or_utf8_string.decode('utf-8')
941
except UnicodeDecodeError:
942
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
943
return unicode_or_utf8_string
944
return unicode_or_utf8_string.encode('utf-8')
947
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
948
' Revision id generators should be creating utf8'
952
def safe_revision_id(unicode_or_utf8_string, warn=True):
953
"""Revision ids should now be utf8, but at one point they were unicode.
955
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
957
:param warn: Functions that are sanitizing user data can set warn=False
958
:return: None or a utf8 revision id.
960
if (unicode_or_utf8_string is None
961
or unicode_or_utf8_string.__class__ == str):
962
return unicode_or_utf8_string
964
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
966
return cache_utf8.encode(unicode_or_utf8_string)
969
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
970
' generators should be creating utf8 file ids.')
973
def safe_file_id(unicode_or_utf8_string, warn=True):
974
"""File ids should now be utf8, but at one point they were unicode.
976
This is the same as safe_utf8, except it uses the cached encode functions
977
to save a little bit of performance.
979
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
981
:param warn: Functions that are sanitizing user data can set warn=False
982
:return: None or a utf8 file id.
984
if (unicode_or_utf8_string is None
985
or unicode_or_utf8_string.__class__ == str):
986
return unicode_or_utf8_string
988
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
990
return cache_utf8.encode(unicode_or_utf8_string)
993
_platform_normalizes_filenames = False
994
if sys.platform == 'darwin':
995
_platform_normalizes_filenames = True
998
def normalizes_filenames():
999
"""Return True if this platform normalizes unicode filenames.
1001
Mac OSX does, Windows/Linux do not.
1003
return _platform_normalizes_filenames
1006
def _accessible_normalized_filename(path):
1007
"""Get the unicode normalized path, and if you can access the file.
1009
On platforms where the system normalizes filenames (Mac OSX),
1010
you can access a file by any path which will normalize correctly.
1011
On platforms where the system does not normalize filenames
1012
(Windows, Linux), you have to access a file by its exact path.
1014
Internally, bzr only supports NFC/NFKC normalization, since that is
1015
the standard for XML documents.
1017
So return the normalized path, and a flag indicating if the file
1018
can be accessed by that path.
1021
return unicodedata.normalize('NFKC', unicode(path)), True
1024
def _inaccessible_normalized_filename(path):
1025
__doc__ = _accessible_normalized_filename.__doc__
1027
normalized = unicodedata.normalize('NFKC', unicode(path))
1028
return normalized, normalized == path
1031
if _platform_normalizes_filenames:
1032
normalized_filename = _accessible_normalized_filename
1034
normalized_filename = _inaccessible_normalized_filename
1037
def terminal_width():
1038
"""Return estimated terminal width."""
1039
if sys.platform == 'win32':
1040
return win32utils.get_console_size()[0]
1043
import struct, fcntl, termios
1044
s = struct.pack('HHHH', 0, 0, 0, 0)
1045
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1046
width = struct.unpack('HHHH', x)[1]
1051
width = int(os.environ['COLUMNS'])
1060
def supports_executable():
1061
return sys.platform != "win32"
1064
def supports_posix_readonly():
1065
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1067
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1068
directory controls creation/deletion, etc.
1070
And under win32, readonly means that the directory itself cannot be
1071
deleted. The contents of a readonly directory can be changed, unlike POSIX
1072
where files in readonly directories cannot be added, deleted or renamed.
1074
return sys.platform != "win32"
1077
def set_or_unset_env(env_variable, value):
1078
"""Modify the environment, setting or removing the env_variable.
1080
:param env_variable: The environment variable in question
1081
:param value: The value to set the environment to. If None, then
1082
the variable will be removed.
1083
:return: The original value of the environment variable.
1085
orig_val = os.environ.get(env_variable)
1087
if orig_val is not None:
1088
del os.environ[env_variable]
1090
if isinstance(value, unicode):
1091
value = value.encode(bzrlib.user_encoding)
1092
os.environ[env_variable] = value
1096
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1099
def check_legal_path(path):
1100
"""Check whether the supplied path is legal.
1101
This is only required on Windows, so we don't test on other platforms
1104
if sys.platform != "win32":
1106
if _validWin32PathRE.match(path) is None:
1107
raise errors.IllegalPath(path)
1110
def walkdirs(top, prefix=""):
1111
"""Yield data about all the directories in a tree.
1113
This yields all the data about the contents of a directory at a time.
1114
After each directory has been yielded, if the caller has mutated the list
1115
to exclude some directories, they are then not descended into.
1117
The data yielded is of the form:
1118
((directory-relpath, directory-path-from-top),
1119
[(relpath, basename, kind, lstat, path-from-top), ...]),
1120
- directory-relpath is the relative path of the directory being returned
1121
with respect to top. prefix is prepended to this.
1122
- directory-path-from-root is the path including top for this directory.
1123
It is suitable for use with os functions.
1124
- relpath is the relative path within the subtree being walked.
1125
- basename is the basename of the path
1126
- kind is the kind of the file now. If unknown then the file is not
1127
present within the tree - but it may be recorded as versioned. See
1129
- lstat is the stat data *if* the file was statted.
1130
- planned, not implemented:
1131
path_from_tree_root is the path from the root of the tree.
1133
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1134
allows one to walk a subtree but get paths that are relative to a tree
1136
:return: an iterator over the dirs.
1138
#TODO there is a bit of a smell where the results of the directory-
1139
# summary in this, and the path from the root, may not agree
1140
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1141
# potentially confusing output. We should make this more robust - but
1142
# not at a speed cost. RBC 20060731
1144
_directory = _directory_kind
1145
_listdir = os.listdir
1146
_kind_from_mode = _formats.get
1147
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1149
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1150
relroot, _, _, _, top = pending.pop()
1152
relprefix = relroot + u'/'
1155
top_slash = top + u'/'
1158
append = dirblock.append
1159
for name in sorted(_listdir(top)):
1160
abspath = top_slash + name
1161
statvalue = _lstat(abspath)
1162
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1163
append((relprefix + name, name, kind, statvalue, abspath))
1164
yield (relroot, top), dirblock
1166
# push the user specified dirs from dirblock
1167
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1170
def _walkdirs_utf8(top, prefix=""):
1171
"""Yield data about all the directories in a tree.
1173
This yields the same information as walkdirs() only each entry is yielded
1174
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1175
are returned as exact byte-strings.
1177
:return: yields a tuple of (dir_info, [file_info])
1178
dir_info is (utf8_relpath, path-from-top)
1179
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1180
if top is an absolute path, path-from-top is also an absolute path.
1181
path-from-top might be unicode or utf8, but it is the correct path to
1182
pass to os functions to affect the file in question. (such as os.lstat)
1184
fs_encoding = _fs_enc.upper()
1185
if (sys.platform == 'win32' or
1186
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1187
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1189
return _walkdirs_fs_utf8(top, prefix=prefix)
1192
def _walkdirs_fs_utf8(top, prefix=""):
1193
"""See _walkdirs_utf8.
1195
This sub-function is called when we know the filesystem is already in utf8
1196
encoding. So we don't need to transcode filenames.
1199
_directory = _directory_kind
1200
_listdir = os.listdir
1201
_kind_from_mode = _formats.get
1203
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1204
# But we don't actually uses 1-3 in pending, so set them to None
1205
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1207
relroot, _, _, _, top = pending.pop()
1209
relprefix = relroot + '/'
1212
top_slash = top + '/'
1215
append = dirblock.append
1216
for name in sorted(_listdir(top)):
1217
abspath = top_slash + name
1218
statvalue = _lstat(abspath)
1219
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1220
append((relprefix + name, name, kind, statvalue, abspath))
1221
yield (relroot, top), dirblock
1223
# push the user specified dirs from dirblock
1224
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1227
def _walkdirs_unicode_to_utf8(top, prefix=""):
1228
"""See _walkdirs_utf8
1230
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1232
This is currently the fallback code path when the filesystem encoding is
1233
not UTF-8. It may be better to implement an alternative so that we can
1234
safely handle paths that are not properly decodable in the current
1237
_utf8_encode = codecs.getencoder('utf8')
1239
_directory = _directory_kind
1240
_listdir = os.listdir
1241
_kind_from_mode = _formats.get
1243
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1245
relroot, _, _, _, top = pending.pop()
1247
relprefix = relroot + '/'
1250
top_slash = top + u'/'
1253
append = dirblock.append
1254
for name in sorted(_listdir(top)):
1255
name_utf8 = _utf8_encode(name)[0]
1256
abspath = top_slash + name
1257
statvalue = _lstat(abspath)
1258
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1259
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1260
yield (relroot, top), dirblock
1262
# push the user specified dirs from dirblock
1263
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1266
def copy_tree(from_path, to_path, handlers={}):
1267
"""Copy all of the entries in from_path into to_path.
1269
:param from_path: The base directory to copy.
1270
:param to_path: The target directory. If it does not exist, it will
1272
:param handlers: A dictionary of functions, which takes a source and
1273
destinations for files, directories, etc.
1274
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1275
'file', 'directory', and 'symlink' should always exist.
1276
If they are missing, they will be replaced with 'os.mkdir()',
1277
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1279
# Now, just copy the existing cached tree to the new location
1280
# We use a cheap trick here.
1281
# Absolute paths are prefixed with the first parameter
1282
# relative paths are prefixed with the second.
1283
# So we can get both the source and target returned
1284
# without any extra work.
1286
def copy_dir(source, dest):
1289
def copy_link(source, dest):
1290
"""Copy the contents of a symlink"""
1291
link_to = os.readlink(source)
1292
os.symlink(link_to, dest)
1294
real_handlers = {'file':shutil.copy2,
1295
'symlink':copy_link,
1296
'directory':copy_dir,
1298
real_handlers.update(handlers)
1300
if not os.path.exists(to_path):
1301
real_handlers['directory'](from_path, to_path)
1303
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1304
for relpath, name, kind, st, abspath in entries:
1305
real_handlers[kind](abspath, relpath)
1308
def path_prefix_key(path):
1309
"""Generate a prefix-order path key for path.
1311
This can be used to sort paths in the same way that walkdirs does.
1313
return (dirname(path) , path)
1316
def compare_paths_prefix_order(path_a, path_b):
1317
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1318
key_a = path_prefix_key(path_a)
1319
key_b = path_prefix_key(path_b)
1320
return cmp(key_a, key_b)
1323
_cached_user_encoding = None
1326
def get_user_encoding(use_cache=True):
1327
"""Find out what the preferred user encoding is.
1329
This is generally the encoding that is used for command line parameters
1330
and file contents. This may be different from the terminal encoding
1331
or the filesystem encoding.
1333
:param use_cache: Enable cache for detected encoding.
1334
(This parameter is turned on by default,
1335
and required only for selftesting)
1337
:return: A string defining the preferred user encoding
1339
global _cached_user_encoding
1340
if _cached_user_encoding is not None and use_cache:
1341
return _cached_user_encoding
1343
if sys.platform == 'darwin':
1344
# work around egregious python 2.4 bug
1345
sys.platform = 'posix'
1349
sys.platform = 'darwin'
1354
user_encoding = locale.getpreferredencoding()
1355
except locale.Error, e:
1356
sys.stderr.write('bzr: warning: %s\n'
1357
' Could not determine what text encoding to use.\n'
1358
' This error usually means your Python interpreter\n'
1359
' doesn\'t support the locale set by $LANG (%s)\n'
1360
" Continuing with ascii encoding.\n"
1361
% (e, os.environ.get('LANG')))
1362
user_encoding = 'ascii'
1364
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1365
# treat that as ASCII, and not support printing unicode characters to the
1367
if user_encoding in (None, 'cp0'):
1368
user_encoding = 'ascii'
1372
codecs.lookup(user_encoding)
1374
sys.stderr.write('bzr: warning:'
1375
' unknown encoding %s.'
1376
' Continuing with ascii encoding.\n'
1379
user_encoding = 'ascii'
1382
_cached_user_encoding = user_encoding
1384
return user_encoding
1387
def recv_all(socket, bytes):
1388
"""Receive an exact number of bytes.
1390
Regular Socket.recv() may return less than the requested number of bytes,
1391
dependning on what's in the OS buffer. MSG_WAITALL is not available
1392
on all platforms, but this should work everywhere. This will return
1393
less than the requested amount if the remote end closes.
1395
This isn't optimized and is intended mostly for use in testing.
1398
while len(b) < bytes:
1399
new = socket.recv(bytes - len(b))
1405
def dereference_path(path):
1406
"""Determine the real path to a file.
1408
All parent elements are dereferenced. But the file itself is not
1410
:param path: The original path. May be absolute or relative.
1411
:return: the real path *to* the file
1413
parent, base = os.path.split(path)
1414
# The pathjoin for '.' is a workaround for Python bug #1213894.
1415
# (initial path components aren't dereferenced)
1416
return pathjoin(realpath(pathjoin('.', parent)), base)
1419
def supports_mapi():
1420
"""Return True if we can use MAPI to launch a mail client."""
1421
return sys.platform == "win32"
added
removed
bzrlib/osutils.py
