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)):
237
# This may throw an exception, in which case success will
239
rename_func(old, new)
243
# If the file used to exist, rename it back into place
244
# otherwise just delete it from the tmp location
246
unlink_func(tmp_name)
248
rename_func(tmp_name, new)
251
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
252
# choke on a Unicode string containing a relative path if
253
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
255
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
256
def _posix_abspath(path):
257
# jam 20060426 rather than encoding to fsencoding
258
# copy posixpath.abspath, but use os.getcwdu instead
259
if not posixpath.isabs(path):
260
path = posixpath.join(getcwd(), path)
261
return posixpath.normpath(path)
264
def _posix_realpath(path):
265
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
268
def _win32_fixdrive(path):
269
"""Force drive letters to be consistent.
271
win32 is inconsistent whether it returns lower or upper case
272
and even if it was consistent the user might type the other
273
so we force it to uppercase
274
running python.exe under cmd.exe return capital C:\\
275
running win32 python inside a cygwin shell returns lowercase c:\\
277
drive, path = _nt_splitdrive(path)
278
return drive.upper() + path
281
def _win32_abspath(path):
282
# Real _nt_abspath doesn't have a problem with a unicode cwd
283
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
286
def _win98_abspath(path):
287
"""Return the absolute version of a path.
288
Windows 98 safe implementation (python reimplementation
289
of Win32 API function GetFullPathNameW)
294
# \\HOST\path => //HOST/path
295
# //HOST/path => //HOST/path
296
# path => C:/cwd/path
299
# check for absolute path
300
drive = _nt_splitdrive(path)[0]
301
if drive == '' and path[:2] not in('//','\\\\'):
303
# we cannot simply os.path.join cwd and path
304
# because os.path.join('C:','/path') produce '/path'
305
# and this is incorrect
306
if path[:1] in ('/','\\'):
307
cwd = _nt_splitdrive(cwd)[0]
309
path = cwd + '\\' + path
310
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
312
if win32utils.winver == 'Windows 98':
313
_win32_abspath = _win98_abspath
316
def _win32_realpath(path):
317
# Real _nt_realpath doesn't have a problem with a unicode cwd
318
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
321
def _win32_pathjoin(*args):
322
return _nt_join(*args).replace('\\', '/')
325
def _win32_normpath(path):
326
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
330
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
333
def _win32_mkdtemp(*args, **kwargs):
334
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
337
def _win32_rename(old, new):
338
"""We expect to be able to atomically replace 'new' with old.
340
On win32, if new exists, it must be moved out of the way first,
344
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
346
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
347
# If we try to rename a non-existant file onto cwd, we get
348
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
349
# if the old path doesn't exist, sometimes we get EACCES
350
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
356
return unicodedata.normalize('NFKC', os.getcwdu())
359
# Default is to just use the python builtins, but these can be rebound on
360
# particular platforms.
361
abspath = _posix_abspath
362
realpath = _posix_realpath
363
pathjoin = os.path.join
364
normpath = os.path.normpath
367
dirname = os.path.dirname
368
basename = os.path.basename
369
split = os.path.split
370
splitext = os.path.splitext
371
# These were already imported into local scope
372
# mkdtemp = tempfile.mkdtemp
373
# rmtree = shutil.rmtree
375
MIN_ABS_PATHLENGTH = 1
378
if sys.platform == 'win32':
379
abspath = _win32_abspath
380
realpath = _win32_realpath
381
pathjoin = _win32_pathjoin
382
normpath = _win32_normpath
383
getcwd = _win32_getcwd
384
mkdtemp = _win32_mkdtemp
385
rename = _win32_rename
387
MIN_ABS_PATHLENGTH = 3
389
def _win32_delete_readonly(function, path, excinfo):
390
"""Error handler for shutil.rmtree function [for win32]
391
Helps to remove files and dirs marked as read-only.
393
exception = excinfo[1]
394
if function in (os.remove, os.rmdir) \
395
and isinstance(exception, OSError) \
396
and exception.errno == errno.EACCES:
402
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
403
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
404
return shutil.rmtree(path, ignore_errors, onerror)
405
elif sys.platform == 'darwin':
409
def get_terminal_encoding():
410
"""Find the best encoding for printing to the screen.
412
This attempts to check both sys.stdout and sys.stdin to see
413
what encoding they are in, and if that fails it falls back to
414
bzrlib.user_encoding.
415
The problem is that on Windows, locale.getpreferredencoding()
416
is not the same encoding as that used by the console:
417
http://mail.python.org/pipermail/python-list/2003-May/162357.html
419
On my standard US Windows XP, the preferred encoding is
420
cp1252, but the console is cp437
422
output_encoding = getattr(sys.stdout, 'encoding', None)
423
if not output_encoding:
424
input_encoding = getattr(sys.stdin, 'encoding', None)
425
if not input_encoding:
426
output_encoding = bzrlib.user_encoding
427
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
429
output_encoding = input_encoding
430
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
432
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
433
if output_encoding == 'cp0':
434
# invalid encoding (cp0 means 'no codepage' on Windows)
435
output_encoding = bzrlib.user_encoding
436
mutter('cp0 is invalid encoding.'
437
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
440
codecs.lookup(output_encoding)
442
sys.stderr.write('bzr: warning:'
443
' unknown terminal encoding %s.\n'
444
' Using encoding %s instead.\n'
445
% (output_encoding, bzrlib.user_encoding)
447
output_encoding = bzrlib.user_encoding
449
return output_encoding
452
def normalizepath(f):
453
if getattr(os.path, 'realpath', None) is not None:
457
[p,e] = os.path.split(f)
458
if e == "" or e == "." or e == "..":
461
return pathjoin(F(p), e)
464
@deprecated_function(zero_ninetythree)
466
"""Copy a file to a backup.
468
Backups are named in GNU-style, with a ~ suffix.
470
If the file is already a backup, it's not copied.
476
if has_symlinks() and os.path.islink(fn):
477
target = os.readlink(fn)
478
os.symlink(target, bfn)
486
outf = file(bfn, 'wb')
494
"""True if f is an accessible directory."""
496
return S_ISDIR(os.lstat(f)[ST_MODE])
502
"""True if f is a regular file."""
504
return S_ISREG(os.lstat(f)[ST_MODE])
509
"""True if f is a symlink."""
511
return S_ISLNK(os.lstat(f)[ST_MODE])
515
def is_inside(dir, fname):
516
"""True if fname is inside dir.
518
The parameters should typically be passed to osutils.normpath first, so
519
that . and .. and repeated slashes are eliminated, and the separators
520
are canonical for the platform.
522
The empty string as a dir name is taken as top-of-tree and matches
525
# XXX: Most callers of this can actually do something smarter by
526
# looking at the inventory
536
return fname.startswith(dir)
539
def is_inside_any(dir_list, fname):
540
"""True if fname is inside any of given dirs."""
541
for dirname in dir_list:
542
if is_inside(dirname, fname):
547
def is_inside_or_parent_of_any(dir_list, fname):
548
"""True if fname is a child or a parent of any of the given files."""
549
for dirname in dir_list:
550
if is_inside(dirname, fname) or is_inside(fname, dirname):
555
def pumpfile(fromfile, tofile):
556
"""Copy contents of one file to another.
558
:return: The number of bytes copied.
563
b = fromfile.read(BUFSIZE)
571
def file_iterator(input_file, readsize=32768):
573
b = input_file.read(readsize)
580
if getattr(f, 'tell', None) is not None:
592
def sha_file_by_name(fname):
593
"""Calculate the SHA1 of a file by reading the full text"""
595
f = os.open(fname, os.O_RDONLY | O_BINARY)
598
b = os.read(f, 1<<16)
606
def sha_strings(strings, _factory=sha.new):
607
"""Return the sha-1 of concatenation of strings"""
609
map(s.update, strings)
613
def sha_string(f, _factory=sha.new):
614
return _factory(f).hexdigest()
617
def fingerprint_file(f):
619
return {'size': len(b),
620
'sha1': sha.new(b).hexdigest()}
623
def compare_files(a, b):
624
"""Returns true if equal in contents"""
635
def local_time_offset(t=None):
636
"""Return offset of local zone from GMT, either at present or at time t."""
639
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
640
return offset.days * 86400 + offset.seconds
643
def format_date(t, offset=0, timezone='original', date_fmt=None,
645
"""Return a formatted date string.
647
:param t: Seconds since the epoch.
648
:param offset: Timezone offset in seconds east of utc.
649
:param timezone: How to display the time: 'utc', 'original' for the
650
timezone specified by offset, or 'local' for the process's current
652
:param show_offset: Whether to append the timezone.
653
:param date_fmt: strftime format.
655
if timezone == 'utc':
658
elif timezone == 'original':
661
tt = time.gmtime(t + offset)
662
elif timezone == 'local':
663
tt = time.localtime(t)
664
offset = local_time_offset(t)
666
raise errors.BzrError("unsupported timezone format %r" % timezone,
667
['options are "utc", "original", "local"'])
669
date_fmt = "%a %Y-%m-%d %H:%M:%S"
671
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
674
return (time.strftime(date_fmt, tt) + offset_str)
677
def compact_date(when):
678
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
681
def format_delta(delta):
682
"""Get a nice looking string for a time delta.
684
:param delta: The time difference in seconds, can be positive or negative.
685
positive indicates time in the past, negative indicates time in the
686
future. (usually time.time() - stored_time)
687
:return: String formatted to show approximate resolution
693
direction = 'in the future'
697
if seconds < 90: # print seconds up to 90 seconds
699
return '%d second %s' % (seconds, direction,)
701
return '%d seconds %s' % (seconds, direction)
703
minutes = int(seconds / 60)
704
seconds -= 60 * minutes
709
if minutes < 90: # print minutes, seconds up to 90 minutes
711
return '%d minute, %d second%s %s' % (
712
minutes, seconds, plural_seconds, direction)
714
return '%d minutes, %d second%s %s' % (
715
minutes, seconds, plural_seconds, direction)
717
hours = int(minutes / 60)
718
minutes -= 60 * hours
725
return '%d hour, %d minute%s %s' % (hours, minutes,
726
plural_minutes, direction)
727
return '%d hours, %d minute%s %s' % (hours, minutes,
728
plural_minutes, direction)
731
"""Return size of given open file."""
732
return os.fstat(f.fileno())[ST_SIZE]
735
# Define rand_bytes based on platform.
737
# Python 2.4 and later have os.urandom,
738
# but it doesn't work on some arches
740
rand_bytes = os.urandom
741
except (NotImplementedError, AttributeError):
742
# If python doesn't have os.urandom, or it doesn't work,
743
# then try to first pull random data from /dev/urandom
745
rand_bytes = file('/dev/urandom', 'rb').read
746
# Otherwise, use this hack as a last resort
747
except (IOError, OSError):
748
# not well seeded, but better than nothing
753
s += chr(random.randint(0, 255))
758
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
760
"""Return a random string of num alphanumeric characters
762
The result only contains lowercase chars because it may be used on
763
case-insensitive filesystems.
766
for raw_byte in rand_bytes(num):
767
s += ALNUM[ord(raw_byte) % 36]
771
## TODO: We could later have path objects that remember their list
772
## decomposition (might be too tricksy though.)
775
"""Turn string into list of parts."""
776
assert isinstance(p, basestring)
778
# split on either delimiter because people might use either on
780
ps = re.split(r'[\\/]', p)
785
raise errors.BzrError("sorry, %r not allowed in path" % f)
786
elif (f == '.') or (f == ''):
793
assert isinstance(p, (list, tuple))
795
if (f == '..') or (f is None) or (f == ''):
796
raise errors.BzrError("sorry, %r not allowed in path" % f)
801
"""Split s into lines, but without removing the newline characters."""
802
lines = s.split('\n')
803
result = [line + '\n' for line in lines[:-1]]
805
result.append(lines[-1])
809
def hardlinks_good():
810
return sys.platform not in ('win32', 'cygwin', 'darwin')
813
def link_or_copy(src, dest):
814
"""Hardlink a file, or copy it if it can't be hardlinked."""
815
if not hardlinks_good():
816
shutil.copyfile(src, dest)
820
except (OSError, IOError), e:
821
if e.errno != errno.EXDEV:
823
shutil.copyfile(src, dest)
826
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
827
# Forgiveness than Permission (EAFP) because:
828
# - root can damage a solaris file system by using unlink,
829
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
830
# EACCES, OSX: EPERM) when invoked on a directory.
831
def delete_any(path):
832
"""Delete a file or directory."""
833
if isdir(path): # Takes care of symlinks
840
if getattr(os, 'symlink', None) is not None:
846
def contains_whitespace(s):
847
"""True if there are any whitespace characters in s."""
848
# string.whitespace can include '\xa0' in certain locales, because it is
849
# considered "non-breaking-space" as part of ISO-8859-1. But it
850
# 1) Isn't a breaking whitespace
851
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
853
# 3) '\xa0' isn't unicode safe since it is >128.
855
# This should *not* be a unicode set of characters in case the source
856
# string is not a Unicode string. We can auto-up-cast the characters since
857
# they are ascii, but we don't want to auto-up-cast the string in case it
859
for ch in ' \t\n\r\v\f':
866
def contains_linebreaks(s):
867
"""True if there is any vertical whitespace in s."""
875
def relpath(base, path):
876
"""Return path relative to base, or raise exception.
878
The path may be either an absolute path or a path relative to the
879
current working directory.
881
os.path.commonprefix (python2.4) has a bad bug that it works just
882
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
886
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
887
' exceed the platform minimum length (which is %d)' %
894
while len(head) >= len(base):
897
head, tail = os.path.split(head)
901
raise errors.PathNotChild(rp, base)
909
def safe_unicode(unicode_or_utf8_string):
910
"""Coerce unicode_or_utf8_string into unicode.
912
If it is unicode, it is returned.
913
Otherwise it is decoded from utf-8. If a decoding error
914
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
915
as a BzrBadParameter exception.
917
if isinstance(unicode_or_utf8_string, unicode):
918
return unicode_or_utf8_string
920
return unicode_or_utf8_string.decode('utf8')
921
except UnicodeDecodeError:
922
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
925
def safe_utf8(unicode_or_utf8_string):
926
"""Coerce unicode_or_utf8_string to a utf8 string.
928
If it is a str, it is returned.
929
If it is Unicode, it is encoded into a utf-8 string.
931
if isinstance(unicode_or_utf8_string, str):
932
# TODO: jam 20070209 This is overkill, and probably has an impact on
933
# performance if we are dealing with lots of apis that want a
936
# Make sure it is a valid utf-8 string
937
unicode_or_utf8_string.decode('utf-8')
938
except UnicodeDecodeError:
939
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
940
return unicode_or_utf8_string
941
return unicode_or_utf8_string.encode('utf-8')
944
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
945
' Revision id generators should be creating utf8'
949
def safe_revision_id(unicode_or_utf8_string, warn=True):
950
"""Revision ids should now be utf8, but at one point they were unicode.
952
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
954
:param warn: Functions that are sanitizing user data can set warn=False
955
:return: None or a utf8 revision 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(_revision_id_warning, DeprecationWarning,
963
return cache_utf8.encode(unicode_or_utf8_string)
966
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
967
' generators should be creating utf8 file ids.')
970
def safe_file_id(unicode_or_utf8_string, warn=True):
971
"""File ids should now be utf8, but at one point they were unicode.
973
This is the same as safe_utf8, except it uses the cached encode functions
974
to save a little bit of performance.
976
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
978
:param warn: Functions that are sanitizing user data can set warn=False
979
:return: None or a utf8 file id.
981
if (unicode_or_utf8_string is None
982
or unicode_or_utf8_string.__class__ == str):
983
return unicode_or_utf8_string
985
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
987
return cache_utf8.encode(unicode_or_utf8_string)
990
_platform_normalizes_filenames = False
991
if sys.platform == 'darwin':
992
_platform_normalizes_filenames = True
995
def normalizes_filenames():
996
"""Return True if this platform normalizes unicode filenames.
998
Mac OSX does, Windows/Linux do not.
1000
return _platform_normalizes_filenames
1003
def _accessible_normalized_filename(path):
1004
"""Get the unicode normalized path, and if you can access the file.
1006
On platforms where the system normalizes filenames (Mac OSX),
1007
you can access a file by any path which will normalize correctly.
1008
On platforms where the system does not normalize filenames
1009
(Windows, Linux), you have to access a file by its exact path.
1011
Internally, bzr only supports NFC/NFKC normalization, since that is
1012
the standard for XML documents.
1014
So return the normalized path, and a flag indicating if the file
1015
can be accessed by that path.
1018
return unicodedata.normalize('NFKC', unicode(path)), True
1021
def _inaccessible_normalized_filename(path):
1022
__doc__ = _accessible_normalized_filename.__doc__
1024
normalized = unicodedata.normalize('NFKC', unicode(path))
1025
return normalized, normalized == path
1028
if _platform_normalizes_filenames:
1029
normalized_filename = _accessible_normalized_filename
1031
normalized_filename = _inaccessible_normalized_filename
1034
def terminal_width():
1035
"""Return estimated terminal width."""
1036
if sys.platform == 'win32':
1037
return win32utils.get_console_size()[0]
1040
import struct, fcntl, termios
1041
s = struct.pack('HHHH', 0, 0, 0, 0)
1042
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1043
width = struct.unpack('HHHH', x)[1]
1048
width = int(os.environ['COLUMNS'])
1057
def supports_executable():
1058
return sys.platform != "win32"
1061
def supports_posix_readonly():
1062
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1064
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1065
directory controls creation/deletion, etc.
1067
And under win32, readonly means that the directory itself cannot be
1068
deleted. The contents of a readonly directory can be changed, unlike POSIX
1069
where files in readonly directories cannot be added, deleted or renamed.
1071
return sys.platform != "win32"
1074
def set_or_unset_env(env_variable, value):
1075
"""Modify the environment, setting or removing the env_variable.
1077
:param env_variable: The environment variable in question
1078
:param value: The value to set the environment to. If None, then
1079
the variable will be removed.
1080
:return: The original value of the environment variable.
1082
orig_val = os.environ.get(env_variable)
1084
if orig_val is not None:
1085
del os.environ[env_variable]
1087
if isinstance(value, unicode):
1088
value = value.encode(bzrlib.user_encoding)
1089
os.environ[env_variable] = value
1093
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1096
def check_legal_path(path):
1097
"""Check whether the supplied path is legal.
1098
This is only required on Windows, so we don't test on other platforms
1101
if sys.platform != "win32":
1103
if _validWin32PathRE.match(path) is None:
1104
raise errors.IllegalPath(path)
1107
def walkdirs(top, prefix=""):
1108
"""Yield data about all the directories in a tree.
1110
This yields all the data about the contents of a directory at a time.
1111
After each directory has been yielded, if the caller has mutated the list
1112
to exclude some directories, they are then not descended into.
1114
The data yielded is of the form:
1115
((directory-relpath, directory-path-from-top),
1116
[(relpath, basename, kind, lstat, path-from-top), ...]),
1117
- directory-relpath is the relative path of the directory being returned
1118
with respect to top. prefix is prepended to this.
1119
- directory-path-from-root is the path including top for this directory.
1120
It is suitable for use with os functions.
1121
- relpath is the relative path within the subtree being walked.
1122
- basename is the basename of the path
1123
- kind is the kind of the file now. If unknown then the file is not
1124
present within the tree - but it may be recorded as versioned. See
1126
- lstat is the stat data *if* the file was statted.
1127
- planned, not implemented:
1128
path_from_tree_root is the path from the root of the tree.
1130
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1131
allows one to walk a subtree but get paths that are relative to a tree
1133
:return: an iterator over the dirs.
1135
#TODO there is a bit of a smell where the results of the directory-
1136
# summary in this, and the path from the root, may not agree
1137
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1138
# potentially confusing output. We should make this more robust - but
1139
# not at a speed cost. RBC 20060731
1141
_directory = _directory_kind
1142
_listdir = os.listdir
1143
_kind_from_mode = _formats.get
1144
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1146
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1147
relroot, _, _, _, top = pending.pop()
1149
relprefix = relroot + u'/'
1152
top_slash = top + u'/'
1155
append = dirblock.append
1156
for name in sorted(_listdir(top)):
1157
abspath = top_slash + name
1158
statvalue = _lstat(abspath)
1159
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1160
append((relprefix + name, name, kind, statvalue, abspath))
1161
yield (relroot, top), dirblock
1163
# push the user specified dirs from dirblock
1164
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1167
def _walkdirs_utf8(top, prefix=""):
1168
"""Yield data about all the directories in a tree.
1170
This yields the same information as walkdirs() only each entry is yielded
1171
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1172
are returned as exact byte-strings.
1174
:return: yields a tuple of (dir_info, [file_info])
1175
dir_info is (utf8_relpath, path-from-top)
1176
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1177
if top is an absolute path, path-from-top is also an absolute path.
1178
path-from-top might be unicode or utf8, but it is the correct path to
1179
pass to os functions to affect the file in question. (such as os.lstat)
1181
fs_encoding = _fs_enc.upper()
1182
if (sys.platform == 'win32' or
1183
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1184
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1186
return _walkdirs_fs_utf8(top, prefix=prefix)
1189
def _walkdirs_fs_utf8(top, prefix=""):
1190
"""See _walkdirs_utf8.
1192
This sub-function is called when we know the filesystem is already in utf8
1193
encoding. So we don't need to transcode filenames.
1196
_directory = _directory_kind
1197
_listdir = os.listdir
1198
_kind_from_mode = _formats.get
1200
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1201
# But we don't actually uses 1-3 in pending, so set them to None
1202
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1204
relroot, _, _, _, top = pending.pop()
1206
relprefix = relroot + '/'
1209
top_slash = top + '/'
1212
append = dirblock.append
1213
for name in sorted(_listdir(top)):
1214
abspath = top_slash + name
1215
statvalue = _lstat(abspath)
1216
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1217
append((relprefix + name, name, kind, statvalue, abspath))
1218
yield (relroot, top), dirblock
1220
# push the user specified dirs from dirblock
1221
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1224
def _walkdirs_unicode_to_utf8(top, prefix=""):
1225
"""See _walkdirs_utf8
1227
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1229
This is currently the fallback code path when the filesystem encoding is
1230
not UTF-8. It may be better to implement an alternative so that we can
1231
safely handle paths that are not properly decodable in the current
1234
_utf8_encode = codecs.getencoder('utf8')
1236
_directory = _directory_kind
1237
_listdir = os.listdir
1238
_kind_from_mode = _formats.get
1240
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1242
relroot, _, _, _, top = pending.pop()
1244
relprefix = relroot + '/'
1247
top_slash = top + u'/'
1250
append = dirblock.append
1251
for name in sorted(_listdir(top)):
1252
name_utf8 = _utf8_encode(name)[0]
1253
abspath = top_slash + name
1254
statvalue = _lstat(abspath)
1255
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1256
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1257
yield (relroot, top), dirblock
1259
# push the user specified dirs from dirblock
1260
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1263
def copy_tree(from_path, to_path, handlers={}):
1264
"""Copy all of the entries in from_path into to_path.
1266
:param from_path: The base directory to copy.
1267
:param to_path: The target directory. If it does not exist, it will
1269
:param handlers: A dictionary of functions, which takes a source and
1270
destinations for files, directories, etc.
1271
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1272
'file', 'directory', and 'symlink' should always exist.
1273
If they are missing, they will be replaced with 'os.mkdir()',
1274
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1276
# Now, just copy the existing cached tree to the new location
1277
# We use a cheap trick here.
1278
# Absolute paths are prefixed with the first parameter
1279
# relative paths are prefixed with the second.
1280
# So we can get both the source and target returned
1281
# without any extra work.
1283
def copy_dir(source, dest):
1286
def copy_link(source, dest):
1287
"""Copy the contents of a symlink"""
1288
link_to = os.readlink(source)
1289
os.symlink(link_to, dest)
1291
real_handlers = {'file':shutil.copy2,
1292
'symlink':copy_link,
1293
'directory':copy_dir,
1295
real_handlers.update(handlers)
1297
if not os.path.exists(to_path):
1298
real_handlers['directory'](from_path, to_path)
1300
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1301
for relpath, name, kind, st, abspath in entries:
1302
real_handlers[kind](abspath, relpath)
1305
def path_prefix_key(path):
1306
"""Generate a prefix-order path key for path.
1308
This can be used to sort paths in the same way that walkdirs does.
1310
return (dirname(path) , path)
1313
def compare_paths_prefix_order(path_a, path_b):
1314
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1315
key_a = path_prefix_key(path_a)
1316
key_b = path_prefix_key(path_b)
1317
return cmp(key_a, key_b)
1320
_cached_user_encoding = None
1323
def get_user_encoding(use_cache=True):
1324
"""Find out what the preferred user encoding is.
1326
This is generally the encoding that is used for command line parameters
1327
and file contents. This may be different from the terminal encoding
1328
or the filesystem encoding.
1330
:param use_cache: Enable cache for detected encoding.
1331
(This parameter is turned on by default,
1332
and required only for selftesting)
1334
:return: A string defining the preferred user encoding
1336
global _cached_user_encoding
1337
if _cached_user_encoding is not None and use_cache:
1338
return _cached_user_encoding
1340
if sys.platform == 'darwin':
1341
# work around egregious python 2.4 bug
1342
sys.platform = 'posix'
1346
sys.platform = 'darwin'
1351
user_encoding = locale.getpreferredencoding()
1352
except locale.Error, e:
1353
sys.stderr.write('bzr: warning: %s\n'
1354
' Could not determine what text encoding to use.\n'
1355
' This error usually means your Python interpreter\n'
1356
' doesn\'t support the locale set by $LANG (%s)\n'
1357
" Continuing with ascii encoding.\n"
1358
% (e, os.environ.get('LANG')))
1359
user_encoding = 'ascii'
1361
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1362
# treat that as ASCII, and not support printing unicode characters to the
1364
if user_encoding in (None, 'cp0'):
1365
user_encoding = 'ascii'
1369
codecs.lookup(user_encoding)
1371
sys.stderr.write('bzr: warning:'
1372
' unknown encoding %s.'
1373
' Continuing with ascii encoding.\n'
1376
user_encoding = 'ascii'
1379
_cached_user_encoding = user_encoding
1381
return user_encoding
1384
def recv_all(socket, bytes):
1385
"""Receive an exact number of bytes.
1387
Regular Socket.recv() may return less than the requested number of bytes,
1388
dependning on what's in the OS buffer. MSG_WAITALL is not available
1389
on all platforms, but this should work everywhere. This will return
1390
less than the requested amount if the remote end closes.
1392
This isn't optimized and is intended mostly for use in testing.
1395
while len(b) < bytes:
1396
new = socket.recv(bytes - len(b))
1402
def dereference_path(path):
1403
"""Determine the real path to a file.
1405
All parent elements are dereferenced. But the file itself is not
1407
:param path: The original path. May be absolute or relative.
1408
:return: the real path *to* the file
1410
parent, base = os.path.split(path)
1411
# The pathjoin for '.' is a workaround for Python bug #1213894.
1412
# (initial path components aren't dereferenced)
1413
return pathjoin(realpath(pathjoin('.', parent)), base)
1416
def supports_mapi():
1417
"""Return True if we can use MAPI to launch a mail client."""
1418
return sys.platform == "win32"
added
removed
bzrlib/osutils.py
