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)
73
def make_readonly(filename):
74
"""Make a filename read-only."""
75
mod = os.lstat(filename).st_mode
76
if not stat.S_ISLNK(mod):
78
os.chmod(filename, mod)
81
def make_writable(filename):
82
mod = os.lstat(filename).st_mode
83
if not stat.S_ISLNK(mod):
85
os.chmod(filename, mod)
88
def minimum_path_selection(paths):
89
"""Return the smallset subset of paths which are outside paths.
91
:param paths: A container (and hence not None) of paths.
92
:return: A set of paths sufficient to include everything in paths via
93
is_inside_any, drawn from the paths parameter.
98
other_paths = paths.difference([path])
99
if not is_inside_any(other_paths, path):
100
# this is a top level path, we must check it.
101
search_paths.add(path)
109
"""Return a quoted filename filename
111
This previously used backslash quoting, but that works poorly on
113
# TODO: I'm not really sure this is the best format either.x
115
if _QUOTE_RE is None:
116
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
118
if _QUOTE_RE.search(f):
124
_directory_kind = 'directory'
127
stat.S_IFDIR:_directory_kind,
128
stat.S_IFCHR:'chardev',
129
stat.S_IFBLK:'block',
132
stat.S_IFLNK:'symlink',
133
stat.S_IFSOCK:'socket',
137
def file_kind_from_stat_mode(stat_mode, _formats=_formats, _unknown='unknown'):
138
"""Generate a file kind from a stat mode. This is used in walkdirs.
140
Its performance is critical: Do not mutate without careful benchmarking.
143
return _formats[stat_mode & 0170000]
148
def file_kind(f, _lstat=os.lstat, _mapper=file_kind_from_stat_mode):
150
return _mapper(_lstat(f).st_mode)
152
if getattr(e, 'errno', None) == errno.ENOENT:
153
raise errors.NoSuchFile(f)
158
"""Return the current umask"""
159
# Assume that people aren't messing with the umask while running
160
# XXX: This is not thread safe, but there is no way to get the
161
# umask without setting it
169
_directory_kind: "/",
171
'tree-reference': '+',
175
def kind_marker(kind):
177
return _kind_marker_map[kind]
179
raise errors.BzrError('invalid file kind %r' % kind)
182
lexists = getattr(os.path, 'lexists', None)
186
stat = getattr(os, 'lstat', os.stat)
190
if e.errno == errno.ENOENT:
193
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
196
def fancy_rename(old, new, rename_func, unlink_func):
197
"""A fancy rename, when you don't have atomic rename.
199
:param old: The old path, to rename from
200
:param new: The new path, to rename to
201
:param rename_func: The potentially non-atomic rename function
202
:param unlink_func: A way to delete the target file if the full rename succeeds
205
# sftp rename doesn't allow overwriting, so play tricks:
207
base = os.path.basename(new)
208
dirname = os.path.dirname(new)
209
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
210
tmp_name = pathjoin(dirname, tmp_name)
212
# Rename the file out of the way, but keep track if it didn't exist
213
# We don't want to grab just any exception
214
# something like EACCES should prevent us from continuing
215
# The downside is that the rename_func has to throw an exception
216
# with an errno = ENOENT, or NoSuchFile
219
rename_func(new, tmp_name)
220
except (errors.NoSuchFile,), e:
223
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
224
# function raises an IOError with errno is None when a rename fails.
225
# This then gets caught here.
226
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
229
if (getattr(e, 'errno', None) is None
230
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
238
# This may throw an exception, in which case success will
240
rename_func(old, new)
242
except (IOError, OSError), e:
243
# source and target may be aliases of each other (e.g. on a
244
# case-insensitive filesystem), so we may have accidentally renamed
245
# source by when we tried to rename target
246
if not (file_existed and e.errno in (None, errno.ENOENT)):
250
# If the file used to exist, rename it back into place
251
# otherwise just delete it from the tmp location
253
unlink_func(tmp_name)
255
rename_func(tmp_name, new)
258
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
259
# choke on a Unicode string containing a relative path if
260
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
262
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
263
def _posix_abspath(path):
264
# jam 20060426 rather than encoding to fsencoding
265
# copy posixpath.abspath, but use os.getcwdu instead
266
if not posixpath.isabs(path):
267
path = posixpath.join(getcwd(), path)
268
return posixpath.normpath(path)
271
def _posix_realpath(path):
272
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
275
def _win32_fixdrive(path):
276
"""Force drive letters to be consistent.
278
win32 is inconsistent whether it returns lower or upper case
279
and even if it was consistent the user might type the other
280
so we force it to uppercase
281
running python.exe under cmd.exe return capital C:\\
282
running win32 python inside a cygwin shell returns lowercase c:\\
284
drive, path = _nt_splitdrive(path)
285
return drive.upper() + path
288
def _win32_abspath(path):
289
# Real _nt_abspath doesn't have a problem with a unicode cwd
290
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
293
def _win98_abspath(path):
294
"""Return the absolute version of a path.
295
Windows 98 safe implementation (python reimplementation
296
of Win32 API function GetFullPathNameW)
301
# \\HOST\path => //HOST/path
302
# //HOST/path => //HOST/path
303
# path => C:/cwd/path
306
# check for absolute path
307
drive = _nt_splitdrive(path)[0]
308
if drive == '' and path[:2] not in('//','\\\\'):
310
# we cannot simply os.path.join cwd and path
311
# because os.path.join('C:','/path') produce '/path'
312
# and this is incorrect
313
if path[:1] in ('/','\\'):
314
cwd = _nt_splitdrive(cwd)[0]
316
path = cwd + '\\' + path
317
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
319
if win32utils.winver == 'Windows 98':
320
_win32_abspath = _win98_abspath
323
def _win32_realpath(path):
324
# Real _nt_realpath doesn't have a problem with a unicode cwd
325
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
328
def _win32_pathjoin(*args):
329
return _nt_join(*args).replace('\\', '/')
332
def _win32_normpath(path):
333
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
337
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
340
def _win32_mkdtemp(*args, **kwargs):
341
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
344
def _win32_rename(old, new):
345
"""We expect to be able to atomically replace 'new' with old.
347
On win32, if new exists, it must be moved out of the way first,
351
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
353
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
354
# If we try to rename a non-existant file onto cwd, we get
355
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
356
# if the old path doesn't exist, sometimes we get EACCES
357
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
363
return unicodedata.normalize('NFKC', os.getcwdu())
366
# Default is to just use the python builtins, but these can be rebound on
367
# particular platforms.
368
abspath = _posix_abspath
369
realpath = _posix_realpath
370
pathjoin = os.path.join
371
normpath = os.path.normpath
374
dirname = os.path.dirname
375
basename = os.path.basename
376
split = os.path.split
377
splitext = os.path.splitext
378
# These were already imported into local scope
379
# mkdtemp = tempfile.mkdtemp
380
# rmtree = shutil.rmtree
382
MIN_ABS_PATHLENGTH = 1
385
if sys.platform == 'win32':
386
abspath = _win32_abspath
387
realpath = _win32_realpath
388
pathjoin = _win32_pathjoin
389
normpath = _win32_normpath
390
getcwd = _win32_getcwd
391
mkdtemp = _win32_mkdtemp
392
rename = _win32_rename
394
MIN_ABS_PATHLENGTH = 3
396
def _win32_delete_readonly(function, path, excinfo):
397
"""Error handler for shutil.rmtree function [for win32]
398
Helps to remove files and dirs marked as read-only.
400
exception = excinfo[1]
401
if function in (os.remove, os.rmdir) \
402
and isinstance(exception, OSError) \
403
and exception.errno == errno.EACCES:
409
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
410
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
411
return shutil.rmtree(path, ignore_errors, onerror)
412
elif sys.platform == 'darwin':
416
def get_terminal_encoding():
417
"""Find the best encoding for printing to the screen.
419
This attempts to check both sys.stdout and sys.stdin to see
420
what encoding they are in, and if that fails it falls back to
421
bzrlib.user_encoding.
422
The problem is that on Windows, locale.getpreferredencoding()
423
is not the same encoding as that used by the console:
424
http://mail.python.org/pipermail/python-list/2003-May/162357.html
426
On my standard US Windows XP, the preferred encoding is
427
cp1252, but the console is cp437
429
output_encoding = getattr(sys.stdout, 'encoding', None)
430
if not output_encoding:
431
input_encoding = getattr(sys.stdin, 'encoding', None)
432
if not input_encoding:
433
output_encoding = bzrlib.user_encoding
434
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
436
output_encoding = input_encoding
437
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
439
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
440
if output_encoding == 'cp0':
441
# invalid encoding (cp0 means 'no codepage' on Windows)
442
output_encoding = bzrlib.user_encoding
443
mutter('cp0 is invalid encoding.'
444
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
447
codecs.lookup(output_encoding)
449
sys.stderr.write('bzr: warning:'
450
' unknown terminal encoding %s.\n'
451
' Using encoding %s instead.\n'
452
% (output_encoding, bzrlib.user_encoding)
454
output_encoding = bzrlib.user_encoding
456
return output_encoding
459
def normalizepath(f):
460
if getattr(os.path, 'realpath', None) is not None:
464
[p,e] = os.path.split(f)
465
if e == "" or e == "." or e == "..":
468
return pathjoin(F(p), e)
471
@deprecated_function(one_zero)
473
"""Copy a file to a backup.
475
Backups are named in GNU-style, with a ~ suffix.
477
If the file is already a backup, it's not copied.
483
if has_symlinks() and os.path.islink(fn):
484
target = os.readlink(fn)
485
os.symlink(target, bfn)
493
outf = file(bfn, 'wb')
501
"""True if f is an accessible directory."""
503
return S_ISDIR(os.lstat(f)[ST_MODE])
509
"""True if f is a regular file."""
511
return S_ISREG(os.lstat(f)[ST_MODE])
516
"""True if f is a symlink."""
518
return S_ISLNK(os.lstat(f)[ST_MODE])
522
def is_inside(dir, fname):
523
"""True if fname is inside dir.
525
The parameters should typically be passed to osutils.normpath first, so
526
that . and .. and repeated slashes are eliminated, and the separators
527
are canonical for the platform.
529
The empty string as a dir name is taken as top-of-tree and matches
532
# XXX: Most callers of this can actually do something smarter by
533
# looking at the inventory
543
return fname.startswith(dir)
546
def is_inside_any(dir_list, fname):
547
"""True if fname is inside any of given dirs."""
548
for dirname in dir_list:
549
if is_inside(dirname, fname):
554
def is_inside_or_parent_of_any(dir_list, fname):
555
"""True if fname is a child or a parent of any of the given files."""
556
for dirname in dir_list:
557
if is_inside(dirname, fname) or is_inside(fname, dirname):
562
def pumpfile(fromfile, tofile):
563
"""Copy contents of one file to another.
565
:return: The number of bytes copied.
570
b = fromfile.read(BUFSIZE)
578
def file_iterator(input_file, readsize=32768):
580
b = input_file.read(readsize)
587
if getattr(f, 'tell', None) is not None:
599
def sha_file_by_name(fname):
600
"""Calculate the SHA1 of a file by reading the full text"""
602
f = os.open(fname, os.O_RDONLY | O_BINARY)
605
b = os.read(f, 1<<16)
613
def sha_strings(strings, _factory=sha.new):
614
"""Return the sha-1 of concatenation of strings"""
616
map(s.update, strings)
620
def sha_string(f, _factory=sha.new):
621
return _factory(f).hexdigest()
624
def fingerprint_file(f):
626
return {'size': len(b),
627
'sha1': sha.new(b).hexdigest()}
630
def compare_files(a, b):
631
"""Returns true if equal in contents"""
642
def local_time_offset(t=None):
643
"""Return offset of local zone from GMT, either at present or at time t."""
646
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
647
return offset.days * 86400 + offset.seconds
650
def format_date(t, offset=0, timezone='original', date_fmt=None,
652
"""Return a formatted date string.
654
:param t: Seconds since the epoch.
655
:param offset: Timezone offset in seconds east of utc.
656
:param timezone: How to display the time: 'utc', 'original' for the
657
timezone specified by offset, or 'local' for the process's current
659
:param show_offset: Whether to append the timezone.
660
:param date_fmt: strftime format.
662
if timezone == 'utc':
665
elif timezone == 'original':
668
tt = time.gmtime(t + offset)
669
elif timezone == 'local':
670
tt = time.localtime(t)
671
offset = local_time_offset(t)
673
raise errors.BzrError("unsupported timezone format %r" % timezone,
674
['options are "utc", "original", "local"'])
676
date_fmt = "%a %Y-%m-%d %H:%M:%S"
678
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
681
return (time.strftime(date_fmt, tt) + offset_str)
684
def compact_date(when):
685
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
688
def format_delta(delta):
689
"""Get a nice looking string for a time delta.
691
:param delta: The time difference in seconds, can be positive or negative.
692
positive indicates time in the past, negative indicates time in the
693
future. (usually time.time() - stored_time)
694
:return: String formatted to show approximate resolution
700
direction = 'in the future'
704
if seconds < 90: # print seconds up to 90 seconds
706
return '%d second %s' % (seconds, direction,)
708
return '%d seconds %s' % (seconds, direction)
710
minutes = int(seconds / 60)
711
seconds -= 60 * minutes
716
if minutes < 90: # print minutes, seconds up to 90 minutes
718
return '%d minute, %d second%s %s' % (
719
minutes, seconds, plural_seconds, direction)
721
return '%d minutes, %d second%s %s' % (
722
minutes, seconds, plural_seconds, direction)
724
hours = int(minutes / 60)
725
minutes -= 60 * hours
732
return '%d hour, %d minute%s %s' % (hours, minutes,
733
plural_minutes, direction)
734
return '%d hours, %d minute%s %s' % (hours, minutes,
735
plural_minutes, direction)
738
"""Return size of given open file."""
739
return os.fstat(f.fileno())[ST_SIZE]
742
# Define rand_bytes based on platform.
744
# Python 2.4 and later have os.urandom,
745
# but it doesn't work on some arches
747
rand_bytes = os.urandom
748
except (NotImplementedError, AttributeError):
749
# If python doesn't have os.urandom, or it doesn't work,
750
# then try to first pull random data from /dev/urandom
752
rand_bytes = file('/dev/urandom', 'rb').read
753
# Otherwise, use this hack as a last resort
754
except (IOError, OSError):
755
# not well seeded, but better than nothing
760
s += chr(random.randint(0, 255))
765
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
767
"""Return a random string of num alphanumeric characters
769
The result only contains lowercase chars because it may be used on
770
case-insensitive filesystems.
773
for raw_byte in rand_bytes(num):
774
s += ALNUM[ord(raw_byte) % 36]
778
## TODO: We could later have path objects that remember their list
779
## decomposition (might be too tricksy though.)
782
"""Turn string into list of parts."""
783
assert isinstance(p, basestring)
785
# split on either delimiter because people might use either on
787
ps = re.split(r'[\\/]', p)
792
raise errors.BzrError("sorry, %r not allowed in path" % f)
793
elif (f == '.') or (f == ''):
800
assert isinstance(p, (list, tuple))
802
if (f == '..') or (f is None) or (f == ''):
803
raise errors.BzrError("sorry, %r not allowed in path" % f)
808
"""Split s into lines, but without removing the newline characters."""
809
lines = s.split('\n')
810
result = [line + '\n' for line in lines[:-1]]
812
result.append(lines[-1])
816
def hardlinks_good():
817
return sys.platform not in ('win32', 'cygwin', 'darwin')
820
def link_or_copy(src, dest):
821
"""Hardlink a file, or copy it if it can't be hardlinked."""
822
if not hardlinks_good():
823
shutil.copyfile(src, dest)
827
except (OSError, IOError), e:
828
if e.errno != errno.EXDEV:
830
shutil.copyfile(src, dest)
833
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
834
# Forgiveness than Permission (EAFP) because:
835
# - root can damage a solaris file system by using unlink,
836
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
837
# EACCES, OSX: EPERM) when invoked on a directory.
838
def delete_any(path):
839
"""Delete a file or directory."""
840
if isdir(path): # Takes care of symlinks
847
if getattr(os, 'symlink', None) is not None:
854
if getattr(os, 'link', None) is not None:
860
def contains_whitespace(s):
861
"""True if there are any whitespace characters in s."""
862
# string.whitespace can include '\xa0' in certain locales, because it is
863
# considered "non-breaking-space" as part of ISO-8859-1. But it
864
# 1) Isn't a breaking whitespace
865
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
867
# 3) '\xa0' isn't unicode safe since it is >128.
869
# This should *not* be a unicode set of characters in case the source
870
# string is not a Unicode string. We can auto-up-cast the characters since
871
# they are ascii, but we don't want to auto-up-cast the string in case it
873
for ch in ' \t\n\r\v\f':
880
def contains_linebreaks(s):
881
"""True if there is any vertical whitespace in s."""
889
def relpath(base, path):
890
"""Return path relative to base, or raise exception.
892
The path may be either an absolute path or a path relative to the
893
current working directory.
895
os.path.commonprefix (python2.4) has a bad bug that it works just
896
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
900
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
901
' exceed the platform minimum length (which is %d)' %
908
while len(head) >= len(base):
911
head, tail = os.path.split(head)
915
raise errors.PathNotChild(rp, base)
923
def safe_unicode(unicode_or_utf8_string):
924
"""Coerce unicode_or_utf8_string into unicode.
926
If it is unicode, it is returned.
927
Otherwise it is decoded from utf-8. If a decoding error
928
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
929
as a BzrBadParameter exception.
931
if isinstance(unicode_or_utf8_string, unicode):
932
return unicode_or_utf8_string
934
return unicode_or_utf8_string.decode('utf8')
935
except UnicodeDecodeError:
936
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
939
def safe_utf8(unicode_or_utf8_string):
940
"""Coerce unicode_or_utf8_string to a utf8 string.
942
If it is a str, it is returned.
943
If it is Unicode, it is encoded into a utf-8 string.
945
if isinstance(unicode_or_utf8_string, str):
946
# TODO: jam 20070209 This is overkill, and probably has an impact on
947
# performance if we are dealing with lots of apis that want a
950
# Make sure it is a valid utf-8 string
951
unicode_or_utf8_string.decode('utf-8')
952
except UnicodeDecodeError:
953
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
954
return unicode_or_utf8_string
955
return unicode_or_utf8_string.encode('utf-8')
958
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
959
' Revision id generators should be creating utf8'
963
def safe_revision_id(unicode_or_utf8_string, warn=True):
964
"""Revision ids should now be utf8, but at one point they were unicode.
966
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
968
:param warn: Functions that are sanitizing user data can set warn=False
969
:return: None or a utf8 revision id.
971
if (unicode_or_utf8_string is None
972
or unicode_or_utf8_string.__class__ == str):
973
return unicode_or_utf8_string
975
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
977
return cache_utf8.encode(unicode_or_utf8_string)
980
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
981
' generators should be creating utf8 file ids.')
984
def safe_file_id(unicode_or_utf8_string, warn=True):
985
"""File ids should now be utf8, but at one point they were unicode.
987
This is the same as safe_utf8, except it uses the cached encode functions
988
to save a little bit of performance.
990
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
992
:param warn: Functions that are sanitizing user data can set warn=False
993
:return: None or a utf8 file id.
995
if (unicode_or_utf8_string is None
996
or unicode_or_utf8_string.__class__ == str):
997
return unicode_or_utf8_string
999
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1001
return cache_utf8.encode(unicode_or_utf8_string)
1004
_platform_normalizes_filenames = False
1005
if sys.platform == 'darwin':
1006
_platform_normalizes_filenames = True
1009
def normalizes_filenames():
1010
"""Return True if this platform normalizes unicode filenames.
1012
Mac OSX does, Windows/Linux do not.
1014
return _platform_normalizes_filenames
1017
def _accessible_normalized_filename(path):
1018
"""Get the unicode normalized path, and if you can access the file.
1020
On platforms where the system normalizes filenames (Mac OSX),
1021
you can access a file by any path which will normalize correctly.
1022
On platforms where the system does not normalize filenames
1023
(Windows, Linux), you have to access a file by its exact path.
1025
Internally, bzr only supports NFC/NFKC normalization, since that is
1026
the standard for XML documents.
1028
So return the normalized path, and a flag indicating if the file
1029
can be accessed by that path.
1032
return unicodedata.normalize('NFKC', unicode(path)), True
1035
def _inaccessible_normalized_filename(path):
1036
__doc__ = _accessible_normalized_filename.__doc__
1038
normalized = unicodedata.normalize('NFKC', unicode(path))
1039
return normalized, normalized == path
1042
if _platform_normalizes_filenames:
1043
normalized_filename = _accessible_normalized_filename
1045
normalized_filename = _inaccessible_normalized_filename
1048
def terminal_width():
1049
"""Return estimated terminal width."""
1050
if sys.platform == 'win32':
1051
return win32utils.get_console_size()[0]
1054
import struct, fcntl, termios
1055
s = struct.pack('HHHH', 0, 0, 0, 0)
1056
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1057
width = struct.unpack('HHHH', x)[1]
1062
width = int(os.environ['COLUMNS'])
1071
def supports_executable():
1072
return sys.platform != "win32"
1075
def supports_posix_readonly():
1076
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1078
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1079
directory controls creation/deletion, etc.
1081
And under win32, readonly means that the directory itself cannot be
1082
deleted. The contents of a readonly directory can be changed, unlike POSIX
1083
where files in readonly directories cannot be added, deleted or renamed.
1085
return sys.platform != "win32"
1088
def set_or_unset_env(env_variable, value):
1089
"""Modify the environment, setting or removing the env_variable.
1091
:param env_variable: The environment variable in question
1092
:param value: The value to set the environment to. If None, then
1093
the variable will be removed.
1094
:return: The original value of the environment variable.
1096
orig_val = os.environ.get(env_variable)
1098
if orig_val is not None:
1099
del os.environ[env_variable]
1101
if isinstance(value, unicode):
1102
value = value.encode(bzrlib.user_encoding)
1103
os.environ[env_variable] = value
1107
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1110
def check_legal_path(path):
1111
"""Check whether the supplied path is legal.
1112
This is only required on Windows, so we don't test on other platforms
1115
if sys.platform != "win32":
1117
if _validWin32PathRE.match(path) is None:
1118
raise errors.IllegalPath(path)
1121
def walkdirs(top, prefix=""):
1122
"""Yield data about all the directories in a tree.
1124
This yields all the data about the contents of a directory at a time.
1125
After each directory has been yielded, if the caller has mutated the list
1126
to exclude some directories, they are then not descended into.
1128
The data yielded is of the form:
1129
((directory-relpath, directory-path-from-top),
1130
[(relpath, basename, kind, lstat, path-from-top), ...]),
1131
- directory-relpath is the relative path of the directory being returned
1132
with respect to top. prefix is prepended to this.
1133
- directory-path-from-root is the path including top for this directory.
1134
It is suitable for use with os functions.
1135
- relpath is the relative path within the subtree being walked.
1136
- basename is the basename of the path
1137
- kind is the kind of the file now. If unknown then the file is not
1138
present within the tree - but it may be recorded as versioned. See
1140
- lstat is the stat data *if* the file was statted.
1141
- planned, not implemented:
1142
path_from_tree_root is the path from the root of the tree.
1144
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1145
allows one to walk a subtree but get paths that are relative to a tree
1147
:return: an iterator over the dirs.
1149
#TODO there is a bit of a smell where the results of the directory-
1150
# summary in this, and the path from the root, may not agree
1151
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1152
# potentially confusing output. We should make this more robust - but
1153
# not at a speed cost. RBC 20060731
1155
_directory = _directory_kind
1156
_listdir = os.listdir
1157
_kind_from_mode = _formats.get
1158
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1160
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1161
relroot, _, _, _, top = pending.pop()
1163
relprefix = relroot + u'/'
1166
top_slash = top + u'/'
1169
append = dirblock.append
1170
for name in sorted(_listdir(top)):
1171
abspath = top_slash + name
1172
statvalue = _lstat(abspath)
1173
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1174
append((relprefix + name, name, kind, statvalue, abspath))
1175
yield (relroot, top), dirblock
1177
# push the user specified dirs from dirblock
1178
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1181
def _walkdirs_utf8(top, prefix=""):
1182
"""Yield data about all the directories in a tree.
1184
This yields the same information as walkdirs() only each entry is yielded
1185
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1186
are returned as exact byte-strings.
1188
:return: yields a tuple of (dir_info, [file_info])
1189
dir_info is (utf8_relpath, path-from-top)
1190
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1191
if top is an absolute path, path-from-top is also an absolute path.
1192
path-from-top might be unicode or utf8, but it is the correct path to
1193
pass to os functions to affect the file in question. (such as os.lstat)
1195
fs_encoding = _fs_enc.upper()
1196
if (sys.platform == 'win32' or
1197
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1198
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1200
return _walkdirs_fs_utf8(top, prefix=prefix)
1203
def _walkdirs_fs_utf8(top, prefix=""):
1204
"""See _walkdirs_utf8.
1206
This sub-function is called when we know the filesystem is already in utf8
1207
encoding. So we don't need to transcode filenames.
1210
_directory = _directory_kind
1211
_listdir = os.listdir
1212
_kind_from_mode = _formats.get
1214
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1215
# But we don't actually uses 1-3 in pending, so set them to None
1216
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1218
relroot, _, _, _, top = pending.pop()
1220
relprefix = relroot + '/'
1223
top_slash = top + '/'
1226
append = dirblock.append
1227
for name in sorted(_listdir(top)):
1228
abspath = top_slash + name
1229
statvalue = _lstat(abspath)
1230
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1231
append((relprefix + name, name, kind, statvalue, abspath))
1232
yield (relroot, top), dirblock
1234
# push the user specified dirs from dirblock
1235
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1238
def _walkdirs_unicode_to_utf8(top, prefix=""):
1239
"""See _walkdirs_utf8
1241
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1243
This is currently the fallback code path when the filesystem encoding is
1244
not UTF-8. It may be better to implement an alternative so that we can
1245
safely handle paths that are not properly decodable in the current
1248
_utf8_encode = codecs.getencoder('utf8')
1250
_directory = _directory_kind
1251
_listdir = os.listdir
1252
_kind_from_mode = _formats.get
1254
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1256
relroot, _, _, _, top = pending.pop()
1258
relprefix = relroot + '/'
1261
top_slash = top + u'/'
1264
append = dirblock.append
1265
for name in sorted(_listdir(top)):
1266
name_utf8 = _utf8_encode(name)[0]
1267
abspath = top_slash + name
1268
statvalue = _lstat(abspath)
1269
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1270
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1271
yield (relroot, top), dirblock
1273
# push the user specified dirs from dirblock
1274
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1277
def copy_tree(from_path, to_path, handlers={}):
1278
"""Copy all of the entries in from_path into to_path.
1280
:param from_path: The base directory to copy.
1281
:param to_path: The target directory. If it does not exist, it will
1283
:param handlers: A dictionary of functions, which takes a source and
1284
destinations for files, directories, etc.
1285
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1286
'file', 'directory', and 'symlink' should always exist.
1287
If they are missing, they will be replaced with 'os.mkdir()',
1288
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1290
# Now, just copy the existing cached tree to the new location
1291
# We use a cheap trick here.
1292
# Absolute paths are prefixed with the first parameter
1293
# relative paths are prefixed with the second.
1294
# So we can get both the source and target returned
1295
# without any extra work.
1297
def copy_dir(source, dest):
1300
def copy_link(source, dest):
1301
"""Copy the contents of a symlink"""
1302
link_to = os.readlink(source)
1303
os.symlink(link_to, dest)
1305
real_handlers = {'file':shutil.copy2,
1306
'symlink':copy_link,
1307
'directory':copy_dir,
1309
real_handlers.update(handlers)
1311
if not os.path.exists(to_path):
1312
real_handlers['directory'](from_path, to_path)
1314
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1315
for relpath, name, kind, st, abspath in entries:
1316
real_handlers[kind](abspath, relpath)
1319
def path_prefix_key(path):
1320
"""Generate a prefix-order path key for path.
1322
This can be used to sort paths in the same way that walkdirs does.
1324
return (dirname(path) , path)
1327
def compare_paths_prefix_order(path_a, path_b):
1328
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1329
key_a = path_prefix_key(path_a)
1330
key_b = path_prefix_key(path_b)
1331
return cmp(key_a, key_b)
1334
_cached_user_encoding = None
1337
def get_user_encoding(use_cache=True):
1338
"""Find out what the preferred user encoding is.
1340
This is generally the encoding that is used for command line parameters
1341
and file contents. This may be different from the terminal encoding
1342
or the filesystem encoding.
1344
:param use_cache: Enable cache for detected encoding.
1345
(This parameter is turned on by default,
1346
and required only for selftesting)
1348
:return: A string defining the preferred user encoding
1350
global _cached_user_encoding
1351
if _cached_user_encoding is not None and use_cache:
1352
return _cached_user_encoding
1354
if sys.platform == 'darwin':
1355
# work around egregious python 2.4 bug
1356
sys.platform = 'posix'
1360
sys.platform = 'darwin'
1365
user_encoding = locale.getpreferredencoding()
1366
except locale.Error, e:
1367
sys.stderr.write('bzr: warning: %s\n'
1368
' Could not determine what text encoding to use.\n'
1369
' This error usually means your Python interpreter\n'
1370
' doesn\'t support the locale set by $LANG (%s)\n'
1371
" Continuing with ascii encoding.\n"
1372
% (e, os.environ.get('LANG')))
1373
user_encoding = 'ascii'
1375
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1376
# treat that as ASCII, and not support printing unicode characters to the
1378
if user_encoding in (None, 'cp0'):
1379
user_encoding = 'ascii'
1383
codecs.lookup(user_encoding)
1385
sys.stderr.write('bzr: warning:'
1386
' unknown encoding %s.'
1387
' Continuing with ascii encoding.\n'
1390
user_encoding = 'ascii'
1393
_cached_user_encoding = user_encoding
1395
return user_encoding
1398
def recv_all(socket, bytes):
1399
"""Receive an exact number of bytes.
1401
Regular Socket.recv() may return less than the requested number of bytes,
1402
dependning on what's in the OS buffer. MSG_WAITALL is not available
1403
on all platforms, but this should work everywhere. This will return
1404
less than the requested amount if the remote end closes.
1406
This isn't optimized and is intended mostly for use in testing.
1409
while len(b) < bytes:
1410
new = socket.recv(bytes - len(b))
1417
def send_all(socket, bytes):
1418
"""Send all bytes on a socket.
1420
Regular socket.sendall() can give socket error 10053 on Windows. This
1421
implementation sends no more than 64k at a time, which avoids this problem.
1424
for pos in xrange(0, len(bytes), chunk_size):
1425
socket.sendall(bytes[pos:pos+chunk_size])
1428
def dereference_path(path):
1429
"""Determine the real path to a file.
1431
All parent elements are dereferenced. But the file itself is not
1433
:param path: The original path. May be absolute or relative.
1434
:return: the real path *to* the file
1436
parent, base = os.path.split(path)
1437
# The pathjoin for '.' is a workaround for Python bug #1213894.
1438
# (initial path components aren't dereferenced)
1439
return pathjoin(realpath(pathjoin('.', parent)), base)
1442
def supports_mapi():
1443
"""Return True if we can use MAPI to launch a mail client."""
1444
return sys.platform == "win32"
1447
def resource_string(package, resource_name):
1448
"""Load a resource from a package and return it as a string.
1450
Note: Only packages that start with bzrlib are currently supported.
1452
This is designed to be a lightweight implementation of resource
1453
loading in a way which is API compatible with the same API from
1455
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1456
If and when pkg_resources becomes a standard library, this routine
1459
# Check package name is within bzrlib
1460
if package == "bzrlib":
1461
resource_relpath = resource_name
1462
elif package.startswith("bzrlib."):
1463
package = package[len("bzrlib."):].replace('.', os.sep)
1464
resource_relpath = pathjoin(package, resource_name)
1466
raise errors.BzrError('resource package %s not in bzrlib' % package)
1468
# Map the resource to a file and read its contents
1469
base = dirname(bzrlib.__file__)
1470
if getattr(sys, 'frozen', None): # bzr.exe
1471
base = abspath(pathjoin(base, '..', '..'))
1472
filename = pathjoin(base, resource_relpath)
1473
return open(filename, 'rU').read()
added
removed
bzrlib/osutils.py
