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) in (errno.ENOENT, errno.ENOTDIR):
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('NFC', 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)
472
"""True if f is an accessible directory."""
474
return S_ISDIR(os.lstat(f)[ST_MODE])
480
"""True if f is a regular file."""
482
return S_ISREG(os.lstat(f)[ST_MODE])
487
"""True if f is a symlink."""
489
return S_ISLNK(os.lstat(f)[ST_MODE])
493
def is_inside(dir, fname):
494
"""True if fname is inside dir.
496
The parameters should typically be passed to osutils.normpath first, so
497
that . and .. and repeated slashes are eliminated, and the separators
498
are canonical for the platform.
500
The empty string as a dir name is taken as top-of-tree and matches
503
# XXX: Most callers of this can actually do something smarter by
504
# looking at the inventory
514
return fname.startswith(dir)
517
def is_inside_any(dir_list, fname):
518
"""True if fname is inside any of given dirs."""
519
for dirname in dir_list:
520
if is_inside(dirname, fname):
525
def is_inside_or_parent_of_any(dir_list, fname):
526
"""True if fname is a child or a parent of any of the given files."""
527
for dirname in dir_list:
528
if is_inside(dirname, fname) or is_inside(fname, dirname):
533
def pumpfile(fromfile, tofile):
534
"""Copy contents of one file to another.
536
:return: The number of bytes copied.
541
b = fromfile.read(BUFSIZE)
549
def file_iterator(input_file, readsize=32768):
551
b = input_file.read(readsize)
558
if getattr(f, 'tell', None) is not None:
570
def sha_file_by_name(fname):
571
"""Calculate the SHA1 of a file by reading the full text"""
573
f = os.open(fname, os.O_RDONLY | O_BINARY)
576
b = os.read(f, 1<<16)
584
def sha_strings(strings, _factory=sha.new):
585
"""Return the sha-1 of concatenation of strings"""
587
map(s.update, strings)
591
def sha_string(f, _factory=sha.new):
592
return _factory(f).hexdigest()
595
def fingerprint_file(f):
597
return {'size': len(b),
598
'sha1': sha.new(b).hexdigest()}
601
def compare_files(a, b):
602
"""Returns true if equal in contents"""
613
def local_time_offset(t=None):
614
"""Return offset of local zone from GMT, either at present or at time t."""
617
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
618
return offset.days * 86400 + offset.seconds
621
def format_date(t, offset=0, timezone='original', date_fmt=None,
623
"""Return a formatted date string.
625
:param t: Seconds since the epoch.
626
:param offset: Timezone offset in seconds east of utc.
627
:param timezone: How to display the time: 'utc', 'original' for the
628
timezone specified by offset, or 'local' for the process's current
630
:param show_offset: Whether to append the timezone.
631
:param date_fmt: strftime format.
633
if timezone == 'utc':
636
elif timezone == 'original':
639
tt = time.gmtime(t + offset)
640
elif timezone == 'local':
641
tt = time.localtime(t)
642
offset = local_time_offset(t)
644
raise errors.UnsupportedTimezoneFormat(timezone)
646
date_fmt = "%a %Y-%m-%d %H:%M:%S"
648
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
651
return (time.strftime(date_fmt, tt) + offset_str)
654
def compact_date(when):
655
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
658
def format_delta(delta):
659
"""Get a nice looking string for a time delta.
661
:param delta: The time difference in seconds, can be positive or negative.
662
positive indicates time in the past, negative indicates time in the
663
future. (usually time.time() - stored_time)
664
:return: String formatted to show approximate resolution
670
direction = 'in the future'
674
if seconds < 90: # print seconds up to 90 seconds
676
return '%d second %s' % (seconds, direction,)
678
return '%d seconds %s' % (seconds, direction)
680
minutes = int(seconds / 60)
681
seconds -= 60 * minutes
686
if minutes < 90: # print minutes, seconds up to 90 minutes
688
return '%d minute, %d second%s %s' % (
689
minutes, seconds, plural_seconds, direction)
691
return '%d minutes, %d second%s %s' % (
692
minutes, seconds, plural_seconds, direction)
694
hours = int(minutes / 60)
695
minutes -= 60 * hours
702
return '%d hour, %d minute%s %s' % (hours, minutes,
703
plural_minutes, direction)
704
return '%d hours, %d minute%s %s' % (hours, minutes,
705
plural_minutes, direction)
708
"""Return size of given open file."""
709
return os.fstat(f.fileno())[ST_SIZE]
712
# Define rand_bytes based on platform.
714
# Python 2.4 and later have os.urandom,
715
# but it doesn't work on some arches
717
rand_bytes = os.urandom
718
except (NotImplementedError, AttributeError):
719
# If python doesn't have os.urandom, or it doesn't work,
720
# then try to first pull random data from /dev/urandom
722
rand_bytes = file('/dev/urandom', 'rb').read
723
# Otherwise, use this hack as a last resort
724
except (IOError, OSError):
725
# not well seeded, but better than nothing
730
s += chr(random.randint(0, 255))
735
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
737
"""Return a random string of num alphanumeric characters
739
The result only contains lowercase chars because it may be used on
740
case-insensitive filesystems.
743
for raw_byte in rand_bytes(num):
744
s += ALNUM[ord(raw_byte) % 36]
748
## TODO: We could later have path objects that remember their list
749
## decomposition (might be too tricksy though.)
752
"""Turn string into list of parts."""
753
assert isinstance(p, basestring)
755
# split on either delimiter because people might use either on
757
ps = re.split(r'[\\/]', p)
762
raise errors.BzrError("sorry, %r not allowed in path" % f)
763
elif (f == '.') or (f == ''):
770
assert isinstance(p, (list, tuple))
772
if (f == '..') or (f is None) or (f == ''):
773
raise errors.BzrError("sorry, %r not allowed in path" % f)
778
"""Split s into lines, but without removing the newline characters."""
779
lines = s.split('\n')
780
result = [line + '\n' for line in lines[:-1]]
782
result.append(lines[-1])
786
def hardlinks_good():
787
return sys.platform not in ('win32', 'cygwin', 'darwin')
790
def link_or_copy(src, dest):
791
"""Hardlink a file, or copy it if it can't be hardlinked."""
792
if not hardlinks_good():
793
shutil.copyfile(src, dest)
797
except (OSError, IOError), e:
798
if e.errno != errno.EXDEV:
800
shutil.copyfile(src, dest)
803
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
804
# Forgiveness than Permission (EAFP) because:
805
# - root can damage a solaris file system by using unlink,
806
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
807
# EACCES, OSX: EPERM) when invoked on a directory.
808
def delete_any(path):
809
"""Delete a file or directory."""
810
if isdir(path): # Takes care of symlinks
817
if getattr(os, 'symlink', None) is not None:
824
if getattr(os, 'link', None) is not None:
830
def contains_whitespace(s):
831
"""True if there are any whitespace characters in s."""
832
# string.whitespace can include '\xa0' in certain locales, because it is
833
# considered "non-breaking-space" as part of ISO-8859-1. But it
834
# 1) Isn't a breaking whitespace
835
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
837
# 3) '\xa0' isn't unicode safe since it is >128.
839
# This should *not* be a unicode set of characters in case the source
840
# string is not a Unicode string. We can auto-up-cast the characters since
841
# they are ascii, but we don't want to auto-up-cast the string in case it
843
for ch in ' \t\n\r\v\f':
850
def contains_linebreaks(s):
851
"""True if there is any vertical whitespace in s."""
859
def relpath(base, path):
860
"""Return path relative to base, or raise exception.
862
The path may be either an absolute path or a path relative to the
863
current working directory.
865
os.path.commonprefix (python2.4) has a bad bug that it works just
866
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
870
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
871
' exceed the platform minimum length (which is %d)' %
878
while len(head) >= len(base):
881
head, tail = os.path.split(head)
885
raise errors.PathNotChild(rp, base)
893
def safe_unicode(unicode_or_utf8_string):
894
"""Coerce unicode_or_utf8_string into unicode.
896
If it is unicode, it is returned.
897
Otherwise it is decoded from utf-8. If a decoding error
898
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
899
as a BzrBadParameter exception.
901
if isinstance(unicode_or_utf8_string, unicode):
902
return unicode_or_utf8_string
904
return unicode_or_utf8_string.decode('utf8')
905
except UnicodeDecodeError:
906
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
909
def safe_utf8(unicode_or_utf8_string):
910
"""Coerce unicode_or_utf8_string to a utf8 string.
912
If it is a str, it is returned.
913
If it is Unicode, it is encoded into a utf-8 string.
915
if isinstance(unicode_or_utf8_string, str):
916
# TODO: jam 20070209 This is overkill, and probably has an impact on
917
# performance if we are dealing with lots of apis that want a
920
# Make sure it is a valid utf-8 string
921
unicode_or_utf8_string.decode('utf-8')
922
except UnicodeDecodeError:
923
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
924
return unicode_or_utf8_string
925
return unicode_or_utf8_string.encode('utf-8')
928
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
929
' Revision id generators should be creating utf8'
933
def safe_revision_id(unicode_or_utf8_string, warn=True):
934
"""Revision ids should now be utf8, but at one point they were unicode.
936
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
938
:param warn: Functions that are sanitizing user data can set warn=False
939
:return: None or a utf8 revision id.
941
if (unicode_or_utf8_string is None
942
or unicode_or_utf8_string.__class__ == str):
943
return unicode_or_utf8_string
945
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
947
return cache_utf8.encode(unicode_or_utf8_string)
950
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
951
' generators should be creating utf8 file ids.')
954
def safe_file_id(unicode_or_utf8_string, warn=True):
955
"""File ids should now be utf8, but at one point they were unicode.
957
This is the same as safe_utf8, except it uses the cached encode functions
958
to save a little bit of performance.
960
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
962
:param warn: Functions that are sanitizing user data can set warn=False
963
:return: None or a utf8 file id.
965
if (unicode_or_utf8_string is None
966
or unicode_or_utf8_string.__class__ == str):
967
return unicode_or_utf8_string
969
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
971
return cache_utf8.encode(unicode_or_utf8_string)
974
_platform_normalizes_filenames = False
975
if sys.platform == 'darwin':
976
_platform_normalizes_filenames = True
979
def normalizes_filenames():
980
"""Return True if this platform normalizes unicode filenames.
982
Mac OSX does, Windows/Linux do not.
984
return _platform_normalizes_filenames
987
def _accessible_normalized_filename(path):
988
"""Get the unicode normalized path, and if you can access the file.
990
On platforms where the system normalizes filenames (Mac OSX),
991
you can access a file by any path which will normalize correctly.
992
On platforms where the system does not normalize filenames
993
(Windows, Linux), you have to access a file by its exact path.
995
Internally, bzr only supports NFC normalization, since that is
996
the standard for XML documents.
998
So return the normalized path, and a flag indicating if the file
999
can be accessed by that path.
1002
return unicodedata.normalize('NFC', unicode(path)), True
1005
def _inaccessible_normalized_filename(path):
1006
__doc__ = _accessible_normalized_filename.__doc__
1008
normalized = unicodedata.normalize('NFC', unicode(path))
1009
return normalized, normalized == path
1012
if _platform_normalizes_filenames:
1013
normalized_filename = _accessible_normalized_filename
1015
normalized_filename = _inaccessible_normalized_filename
1018
def terminal_width():
1019
"""Return estimated terminal width."""
1020
if sys.platform == 'win32':
1021
return win32utils.get_console_size()[0]
1024
import struct, fcntl, termios
1025
s = struct.pack('HHHH', 0, 0, 0, 0)
1026
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1027
width = struct.unpack('HHHH', x)[1]
1032
width = int(os.environ['COLUMNS'])
1041
def supports_executable():
1042
return sys.platform != "win32"
1045
def supports_posix_readonly():
1046
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1048
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1049
directory controls creation/deletion, etc.
1051
And under win32, readonly means that the directory itself cannot be
1052
deleted. The contents of a readonly directory can be changed, unlike POSIX
1053
where files in readonly directories cannot be added, deleted or renamed.
1055
return sys.platform != "win32"
1058
def set_or_unset_env(env_variable, value):
1059
"""Modify the environment, setting or removing the env_variable.
1061
:param env_variable: The environment variable in question
1062
:param value: The value to set the environment to. If None, then
1063
the variable will be removed.
1064
:return: The original value of the environment variable.
1066
orig_val = os.environ.get(env_variable)
1068
if orig_val is not None:
1069
del os.environ[env_variable]
1071
if isinstance(value, unicode):
1072
value = value.encode(bzrlib.user_encoding)
1073
os.environ[env_variable] = value
1077
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1080
def check_legal_path(path):
1081
"""Check whether the supplied path is legal.
1082
This is only required on Windows, so we don't test on other platforms
1085
if sys.platform != "win32":
1087
if _validWin32PathRE.match(path) is None:
1088
raise errors.IllegalPath(path)
1091
def walkdirs(top, prefix=""):
1092
"""Yield data about all the directories in a tree.
1094
This yields all the data about the contents of a directory at a time.
1095
After each directory has been yielded, if the caller has mutated the list
1096
to exclude some directories, they are then not descended into.
1098
The data yielded is of the form:
1099
((directory-relpath, directory-path-from-top),
1100
[(relpath, basename, kind, lstat, path-from-top), ...]),
1101
- directory-relpath is the relative path of the directory being returned
1102
with respect to top. prefix is prepended to this.
1103
- directory-path-from-root is the path including top for this directory.
1104
It is suitable for use with os functions.
1105
- relpath is the relative path within the subtree being walked.
1106
- basename is the basename of the path
1107
- kind is the kind of the file now. If unknown then the file is not
1108
present within the tree - but it may be recorded as versioned. See
1110
- lstat is the stat data *if* the file was statted.
1111
- planned, not implemented:
1112
path_from_tree_root is the path from the root of the tree.
1114
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1115
allows one to walk a subtree but get paths that are relative to a tree
1117
:return: an iterator over the dirs.
1119
#TODO there is a bit of a smell where the results of the directory-
1120
# summary in this, and the path from the root, may not agree
1121
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1122
# potentially confusing output. We should make this more robust - but
1123
# not at a speed cost. RBC 20060731
1125
_directory = _directory_kind
1126
_listdir = os.listdir
1127
_kind_from_mode = _formats.get
1128
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1130
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1131
relroot, _, _, _, top = pending.pop()
1133
relprefix = relroot + u'/'
1136
top_slash = top + u'/'
1139
append = dirblock.append
1140
for name in sorted(_listdir(top)):
1141
abspath = top_slash + name
1142
statvalue = _lstat(abspath)
1143
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1144
append((relprefix + name, name, kind, statvalue, abspath))
1145
yield (relroot, top), dirblock
1147
# push the user specified dirs from dirblock
1148
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1151
def _walkdirs_utf8(top, prefix=""):
1152
"""Yield data about all the directories in a tree.
1154
This yields the same information as walkdirs() only each entry is yielded
1155
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1156
are returned as exact byte-strings.
1158
:return: yields a tuple of (dir_info, [file_info])
1159
dir_info is (utf8_relpath, path-from-top)
1160
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1161
if top is an absolute path, path-from-top is also an absolute path.
1162
path-from-top might be unicode or utf8, but it is the correct path to
1163
pass to os functions to affect the file in question. (such as os.lstat)
1165
fs_encoding = _fs_enc.upper()
1166
if (sys.platform == 'win32' or
1167
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1168
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1170
return _walkdirs_fs_utf8(top, prefix=prefix)
1173
def _walkdirs_fs_utf8(top, prefix=""):
1174
"""See _walkdirs_utf8.
1176
This sub-function is called when we know the filesystem is already in utf8
1177
encoding. So we don't need to transcode filenames.
1180
_directory = _directory_kind
1181
_listdir = os.listdir
1182
_kind_from_mode = _formats.get
1184
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1185
# But we don't actually uses 1-3 in pending, so set them to None
1186
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1188
relroot, _, _, _, top = pending.pop()
1190
relprefix = relroot + '/'
1193
top_slash = top + '/'
1196
append = dirblock.append
1197
for name in sorted(_listdir(top)):
1198
abspath = top_slash + name
1199
statvalue = _lstat(abspath)
1200
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1201
append((relprefix + name, name, kind, statvalue, abspath))
1202
yield (relroot, top), dirblock
1204
# push the user specified dirs from dirblock
1205
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1208
def _walkdirs_unicode_to_utf8(top, prefix=""):
1209
"""See _walkdirs_utf8
1211
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1213
This is currently the fallback code path when the filesystem encoding is
1214
not UTF-8. It may be better to implement an alternative so that we can
1215
safely handle paths that are not properly decodable in the current
1218
_utf8_encode = codecs.getencoder('utf8')
1220
_directory = _directory_kind
1221
_listdir = os.listdir
1222
_kind_from_mode = _formats.get
1224
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1226
relroot, _, _, _, top = pending.pop()
1228
relprefix = relroot + '/'
1231
top_slash = top + u'/'
1234
append = dirblock.append
1235
for name in sorted(_listdir(top)):
1236
name_utf8 = _utf8_encode(name)[0]
1237
abspath = top_slash + name
1238
statvalue = _lstat(abspath)
1239
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1240
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1241
yield (relroot, top), dirblock
1243
# push the user specified dirs from dirblock
1244
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1247
def copy_tree(from_path, to_path, handlers={}):
1248
"""Copy all of the entries in from_path into to_path.
1250
:param from_path: The base directory to copy.
1251
:param to_path: The target directory. If it does not exist, it will
1253
:param handlers: A dictionary of functions, which takes a source and
1254
destinations for files, directories, etc.
1255
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1256
'file', 'directory', and 'symlink' should always exist.
1257
If they are missing, they will be replaced with 'os.mkdir()',
1258
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1260
# Now, just copy the existing cached tree to the new location
1261
# We use a cheap trick here.
1262
# Absolute paths are prefixed with the first parameter
1263
# relative paths are prefixed with the second.
1264
# So we can get both the source and target returned
1265
# without any extra work.
1267
def copy_dir(source, dest):
1270
def copy_link(source, dest):
1271
"""Copy the contents of a symlink"""
1272
link_to = os.readlink(source)
1273
os.symlink(link_to, dest)
1275
real_handlers = {'file':shutil.copy2,
1276
'symlink':copy_link,
1277
'directory':copy_dir,
1279
real_handlers.update(handlers)
1281
if not os.path.exists(to_path):
1282
real_handlers['directory'](from_path, to_path)
1284
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1285
for relpath, name, kind, st, abspath in entries:
1286
real_handlers[kind](abspath, relpath)
1289
def path_prefix_key(path):
1290
"""Generate a prefix-order path key for path.
1292
This can be used to sort paths in the same way that walkdirs does.
1294
return (dirname(path) , path)
1297
def compare_paths_prefix_order(path_a, path_b):
1298
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1299
key_a = path_prefix_key(path_a)
1300
key_b = path_prefix_key(path_b)
1301
return cmp(key_a, key_b)
1304
_cached_user_encoding = None
1307
def get_user_encoding(use_cache=True):
1308
"""Find out what the preferred user encoding is.
1310
This is generally the encoding that is used for command line parameters
1311
and file contents. This may be different from the terminal encoding
1312
or the filesystem encoding.
1314
:param use_cache: Enable cache for detected encoding.
1315
(This parameter is turned on by default,
1316
and required only for selftesting)
1318
:return: A string defining the preferred user encoding
1320
global _cached_user_encoding
1321
if _cached_user_encoding is not None and use_cache:
1322
return _cached_user_encoding
1324
if sys.platform == 'darwin':
1325
# work around egregious python 2.4 bug
1326
sys.platform = 'posix'
1330
sys.platform = 'darwin'
1335
user_encoding = locale.getpreferredencoding()
1336
except locale.Error, e:
1337
sys.stderr.write('bzr: warning: %s\n'
1338
' Could not determine what text encoding to use.\n'
1339
' This error usually means your Python interpreter\n'
1340
' doesn\'t support the locale set by $LANG (%s)\n'
1341
" Continuing with ascii encoding.\n"
1342
% (e, os.environ.get('LANG')))
1343
user_encoding = 'ascii'
1345
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1346
# treat that as ASCII, and not support printing unicode characters to the
1348
if user_encoding in (None, 'cp0'):
1349
user_encoding = 'ascii'
1353
codecs.lookup(user_encoding)
1355
sys.stderr.write('bzr: warning:'
1356
' unknown encoding %s.'
1357
' Continuing with ascii encoding.\n'
1360
user_encoding = 'ascii'
1363
_cached_user_encoding = user_encoding
1365
return user_encoding
1368
def recv_all(socket, bytes):
1369
"""Receive an exact number of bytes.
1371
Regular Socket.recv() may return less than the requested number of bytes,
1372
dependning on what's in the OS buffer. MSG_WAITALL is not available
1373
on all platforms, but this should work everywhere. This will return
1374
less than the requested amount if the remote end closes.
1376
This isn't optimized and is intended mostly for use in testing.
1379
while len(b) < bytes:
1380
new = socket.recv(bytes - len(b))
1387
def send_all(socket, bytes):
1388
"""Send all bytes on a socket.
1390
Regular socket.sendall() can give socket error 10053 on Windows. This
1391
implementation sends no more than 64k at a time, which avoids this problem.
1394
for pos in xrange(0, len(bytes), chunk_size):
1395
socket.sendall(bytes[pos:pos+chunk_size])
1398
def dereference_path(path):
1399
"""Determine the real path to a file.
1401
All parent elements are dereferenced. But the file itself is not
1403
:param path: The original path. May be absolute or relative.
1404
:return: the real path *to* the file
1406
parent, base = os.path.split(path)
1407
# The pathjoin for '.' is a workaround for Python bug #1213894.
1408
# (initial path components aren't dereferenced)
1409
return pathjoin(realpath(pathjoin('.', parent)), base)
1412
def supports_mapi():
1413
"""Return True if we can use MAPI to launch a mail client."""
1414
return sys.platform == "win32"
1417
def resource_string(package, resource_name):
1418
"""Load a resource from a package and return it as a string.
1420
Note: Only packages that start with bzrlib are currently supported.
1422
This is designed to be a lightweight implementation of resource
1423
loading in a way which is API compatible with the same API from
1425
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1426
If and when pkg_resources becomes a standard library, this routine
1429
# Check package name is within bzrlib
1430
if package == "bzrlib":
1431
resource_relpath = resource_name
1432
elif package.startswith("bzrlib."):
1433
package = package[len("bzrlib."):].replace('.', os.sep)
1434
resource_relpath = pathjoin(package, resource_name)
1436
raise errors.BzrError('resource package %s not in bzrlib' % package)
1438
# Map the resource to a file and read its contents
1439
base = dirname(bzrlib.__file__)
1440
if getattr(sys, 'frozen', None): # bzr.exe
1441
base = abspath(pathjoin(base, '..', '..'))
1442
filename = pathjoin(base, resource_relpath)
1443
return open(filename, 'rU').read()
added
removed
bzrlib/osutils.py
