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
20
from stat import (S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE,
21
S_ISCHR, S_ISBLK, S_ISFIFO, S_ISSOCK)
25
from bzrlib.lazy_import import lazy_import
26
lazy_import(globals(), """
28
from datetime import datetime
30
from ntpath import (abspath as _nt_abspath,
32
normpath as _nt_normpath,
33
realpath as _nt_realpath,
34
splitdrive as _nt_splitdrive,
43
from tempfile import (
56
from bzrlib import symbol_versioning
57
from bzrlib.symbol_versioning import (
63
# On win32, O_BINARY is used to indicate the file should
64
# be opened in binary mode, rather than text mode.
65
# On other platforms, O_BINARY doesn't exist, because
66
# they always open in binary mode, so it is okay to
67
# OR with 0 on those platforms
68
O_BINARY = getattr(os, 'O_BINARY', 0)
71
def make_readonly(filename):
72
"""Make a filename read-only."""
73
mod = os.lstat(filename).st_mode
74
if not stat.S_ISLNK(mod):
76
os.chmod(filename, mod)
79
def make_writable(filename):
80
mod = os.lstat(filename).st_mode
81
if not stat.S_ISLNK(mod):
83
os.chmod(filename, mod)
86
def minimum_path_selection(paths):
87
"""Return the smallset subset of paths which are outside paths.
89
:param paths: A container (and hence not None) of paths.
90
:return: A set of paths sufficient to include everything in paths via
91
is_inside_any, drawn from the paths parameter.
96
other_paths = paths.difference([path])
97
if not is_inside_any(other_paths, path):
98
# this is a top level path, we must check it.
99
search_paths.add(path)
107
"""Return a quoted filename filename
109
This previously used backslash quoting, but that works poorly on
111
# TODO: I'm not really sure this is the best format either.x
113
if _QUOTE_RE is None:
114
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
116
if _QUOTE_RE.search(f):
122
_directory_kind = 'directory'
125
stat.S_IFDIR:_directory_kind,
126
stat.S_IFCHR:'chardev',
127
stat.S_IFBLK:'block',
130
stat.S_IFLNK:'symlink',
131
stat.S_IFSOCK:'socket',
135
def file_kind_from_stat_mode(stat_mode, _formats=_formats, _unknown='unknown'):
136
"""Generate a file kind from a stat mode. This is used in walkdirs.
138
Its performance is critical: Do not mutate without careful benchmarking.
141
return _formats[stat_mode & 0170000]
146
def file_kind(f, _lstat=os.lstat, _mapper=file_kind_from_stat_mode):
148
return _mapper(_lstat(f).st_mode)
150
if getattr(e, 'errno', None) == errno.ENOENT:
151
raise errors.NoSuchFile(f)
156
"""Return the current umask"""
157
# Assume that people aren't messing with the umask while running
158
# XXX: This is not thread safe, but there is no way to get the
159
# umask without setting it
167
_directory_kind: "/",
169
'tree-reference': '+',
173
def kind_marker(kind):
175
return _kind_marker_map[kind]
177
raise errors.BzrError('invalid file kind %r' % kind)
180
lexists = getattr(os.path, 'lexists', None)
184
stat = getattr(os, 'lstat', os.stat)
188
if e.errno == errno.ENOENT:
191
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
194
def fancy_rename(old, new, rename_func, unlink_func):
195
"""A fancy rename, when you don't have atomic rename.
197
:param old: The old path, to rename from
198
:param new: The new path, to rename to
199
:param rename_func: The potentially non-atomic rename function
200
:param unlink_func: A way to delete the target file if the full rename succeeds
203
# sftp rename doesn't allow overwriting, so play tricks:
205
base = os.path.basename(new)
206
dirname = os.path.dirname(new)
207
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
208
tmp_name = pathjoin(dirname, tmp_name)
210
# Rename the file out of the way, but keep track if it didn't exist
211
# We don't want to grab just any exception
212
# something like EACCES should prevent us from continuing
213
# The downside is that the rename_func has to throw an exception
214
# with an errno = ENOENT, or NoSuchFile
217
rename_func(new, tmp_name)
218
except (errors.NoSuchFile,), e:
221
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
222
# function raises an IOError with errno is None when a rename fails.
223
# This then gets caught here.
224
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
227
if (getattr(e, 'errno', None) is None
228
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
236
# This may throw an exception, in which case success will
238
rename_func(old, new)
240
except (IOError, OSError), e:
241
# source and target may be aliases of each other (e.g. on a
242
# case-insensitive filesystem), so we may have accidentally renamed
243
# source by when we tried to rename target
244
if not (file_existed and e.errno in (None, errno.ENOENT)):
248
# If the file used to exist, rename it back into place
249
# otherwise just delete it from the tmp location
251
unlink_func(tmp_name)
253
rename_func(tmp_name, new)
256
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
257
# choke on a Unicode string containing a relative path if
258
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
260
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
261
def _posix_abspath(path):
262
# jam 20060426 rather than encoding to fsencoding
263
# copy posixpath.abspath, but use os.getcwdu instead
264
if not posixpath.isabs(path):
265
path = posixpath.join(getcwd(), path)
266
return posixpath.normpath(path)
269
def _posix_realpath(path):
270
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
273
def _win32_fixdrive(path):
274
"""Force drive letters to be consistent.
276
win32 is inconsistent whether it returns lower or upper case
277
and even if it was consistent the user might type the other
278
so we force it to uppercase
279
running python.exe under cmd.exe return capital C:\\
280
running win32 python inside a cygwin shell returns lowercase c:\\
282
drive, path = _nt_splitdrive(path)
283
return drive.upper() + path
286
def _win32_abspath(path):
287
global _win32_abspath
288
if win32utils.winver == 'Windows 98':
289
_win32_abspath = _win98_abspath
291
_win32_abspath = _real_win32_abspath
292
return _win32_abspath(path)
295
def _real_win32_abspath(path):
296
# Real _nt_abspath doesn't have a problem with a unicode cwd
297
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
300
def _win98_abspath(path):
301
"""Return the absolute version of a path.
302
Windows 98 safe implementation (python reimplementation
303
of Win32 API function GetFullPathNameW)
308
# \\HOST\path => //HOST/path
309
# //HOST/path => //HOST/path
310
# path => C:/cwd/path
313
# check for absolute path
314
drive = _nt_splitdrive(path)[0]
315
if drive == '' and path[:2] not in('//','\\\\'):
317
# we cannot simply os.path.join cwd and path
318
# because os.path.join('C:','/path') produce '/path'
319
# and this is incorrect
320
if path[:1] in ('/','\\'):
321
cwd = _nt_splitdrive(cwd)[0]
323
path = cwd + '\\' + path
324
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
327
def _win32_realpath(path):
328
# Real _nt_realpath doesn't have a problem with a unicode cwd
329
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
332
def _win32_pathjoin(*args):
333
return _nt_join(*args).replace('\\', '/')
336
def _win32_normpath(path):
337
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
341
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
344
def _win32_mkdtemp(*args, **kwargs):
345
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
348
def _win32_rename(old, new):
349
"""We expect to be able to atomically replace 'new' with old.
351
On win32, if new exists, it must be moved out of the way first,
355
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
357
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
358
# If we try to rename a non-existant file onto cwd, we get
359
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
360
# if the old path doesn't exist, sometimes we get EACCES
361
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
367
return unicodedata.normalize('NFC', os.getcwdu())
370
# Default is to just use the python builtins, but these can be rebound on
371
# particular platforms.
372
abspath = _posix_abspath
373
realpath = _posix_realpath
374
pathjoin = os.path.join
375
normpath = os.path.normpath
378
dirname = os.path.dirname
379
basename = os.path.basename
380
split = os.path.split
381
splitext = os.path.splitext
382
# These were already imported into local scope
383
# mkdtemp = tempfile.mkdtemp
384
# rmtree = shutil.rmtree
386
MIN_ABS_PATHLENGTH = 1
389
if sys.platform == 'win32':
390
abspath = _win32_abspath
391
realpath = _win32_realpath
392
pathjoin = _win32_pathjoin
393
normpath = _win32_normpath
394
getcwd = _win32_getcwd
395
mkdtemp = _win32_mkdtemp
396
rename = _win32_rename
398
MIN_ABS_PATHLENGTH = 3
400
def _win32_delete_readonly(function, path, excinfo):
401
"""Error handler for shutil.rmtree function [for win32]
402
Helps to remove files and dirs marked as read-only.
404
exception = excinfo[1]
405
if function in (os.remove, os.rmdir) \
406
and isinstance(exception, OSError) \
407
and exception.errno == errno.EACCES:
413
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
414
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
415
return shutil.rmtree(path, ignore_errors, onerror)
416
elif sys.platform == 'darwin':
420
def get_terminal_encoding():
421
"""Find the best encoding for printing to the screen.
423
This attempts to check both sys.stdout and sys.stdin to see
424
what encoding they are in, and if that fails it falls back to
425
osutils.get_user_encoding().
426
The problem is that on Windows, locale.getpreferredencoding()
427
is not the same encoding as that used by the console:
428
http://mail.python.org/pipermail/python-list/2003-May/162357.html
430
On my standard US Windows XP, the preferred encoding is
431
cp1252, but the console is cp437
433
from bzrlib.trace import mutter
434
output_encoding = getattr(sys.stdout, 'encoding', None)
435
if not output_encoding:
436
input_encoding = getattr(sys.stdin, 'encoding', None)
437
if not input_encoding:
438
output_encoding = get_user_encoding()
439
mutter('encoding stdout as osutils.get_user_encoding() %r',
442
output_encoding = input_encoding
443
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
445
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
446
if output_encoding == 'cp0':
447
# invalid encoding (cp0 means 'no codepage' on Windows)
448
output_encoding = get_user_encoding()
449
mutter('cp0 is invalid encoding.'
450
' encoding stdout as osutils.get_user_encoding() %r',
454
codecs.lookup(output_encoding)
456
sys.stderr.write('bzr: warning:'
457
' unknown terminal encoding %s.\n'
458
' Using encoding %s instead.\n'
459
% (output_encoding, get_user_encoding())
461
output_encoding = get_user_encoding()
463
return output_encoding
466
def normalizepath(f):
467
if getattr(os.path, 'realpath', None) is not None:
471
[p,e] = os.path.split(f)
472
if e == "" or e == "." or e == "..":
475
return pathjoin(F(p), e)
478
@deprecated_function(one_zero)
480
"""Copy a file to a backup.
482
Backups are named in GNU-style, with a ~ suffix.
484
If the file is already a backup, it's not copied.
490
if has_symlinks() and os.path.islink(fn):
491
target = os.readlink(fn)
492
os.symlink(target, bfn)
500
outf = file(bfn, 'wb')
508
"""True if f is an accessible directory."""
510
return S_ISDIR(os.lstat(f)[ST_MODE])
516
"""True if f is a regular file."""
518
return S_ISREG(os.lstat(f)[ST_MODE])
523
"""True if f is a symlink."""
525
return S_ISLNK(os.lstat(f)[ST_MODE])
529
def is_inside(dir, fname):
530
"""True if fname is inside dir.
532
The parameters should typically be passed to osutils.normpath first, so
533
that . and .. and repeated slashes are eliminated, and the separators
534
are canonical for the platform.
536
The empty string as a dir name is taken as top-of-tree and matches
539
# XXX: Most callers of this can actually do something smarter by
540
# looking at the inventory
550
return fname.startswith(dir)
553
def is_inside_any(dir_list, fname):
554
"""True if fname is inside any of given dirs."""
555
for dirname in dir_list:
556
if is_inside(dirname, fname):
561
def is_inside_or_parent_of_any(dir_list, fname):
562
"""True if fname is a child or a parent of any of the given files."""
563
for dirname in dir_list:
564
if is_inside(dirname, fname) or is_inside(fname, dirname):
569
def pumpfile(fromfile, tofile):
570
"""Copy contents of one file to another.
572
:return: The number of bytes copied.
577
b = fromfile.read(BUFSIZE)
585
def file_iterator(input_file, readsize=32768):
587
b = input_file.read(readsize)
594
if getattr(f, 'tell', None) is not None:
606
def sha_file_by_name(fname):
607
"""Calculate the SHA1 of a file by reading the full text"""
609
f = os.open(fname, os.O_RDONLY | O_BINARY)
612
b = os.read(f, 1<<16)
620
def sha_strings(strings, _factory=sha.new):
621
"""Return the sha-1 of concatenation of strings"""
623
map(s.update, strings)
627
def sha_string(f, _factory=sha.new):
628
return _factory(f).hexdigest()
631
def fingerprint_file(f):
633
return {'size': len(b),
634
'sha1': sha.new(b).hexdigest()}
637
def compare_files(a, b):
638
"""Returns true if equal in contents"""
649
def local_time_offset(t=None):
650
"""Return offset of local zone from GMT, either at present or at time t."""
653
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
654
return offset.days * 86400 + offset.seconds
657
def format_date(t, offset=0, timezone='original', date_fmt=None,
659
"""Return a formatted date string.
661
:param t: Seconds since the epoch.
662
:param offset: Timezone offset in seconds east of utc.
663
:param timezone: How to display the time: 'utc', 'original' for the
664
timezone specified by offset, or 'local' for the process's current
666
:param show_offset: Whether to append the timezone.
667
:param date_fmt: strftime format.
669
if timezone == 'utc':
672
elif timezone == 'original':
675
tt = time.gmtime(t + offset)
676
elif timezone == 'local':
677
tt = time.localtime(t)
678
offset = local_time_offset(t)
680
raise errors.UnsupportedTimezoneFormat(timezone)
682
date_fmt = "%a %Y-%m-%d %H:%M:%S"
684
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
687
return (time.strftime(date_fmt, tt) + offset_str)
690
def compact_date(when):
691
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
694
def format_delta(delta):
695
"""Get a nice looking string for a time delta.
697
:param delta: The time difference in seconds, can be positive or negative.
698
positive indicates time in the past, negative indicates time in the
699
future. (usually time.time() - stored_time)
700
:return: String formatted to show approximate resolution
706
direction = 'in the future'
710
if seconds < 90: # print seconds up to 90 seconds
712
return '%d second %s' % (seconds, direction,)
714
return '%d seconds %s' % (seconds, direction)
716
minutes = int(seconds / 60)
717
seconds -= 60 * minutes
722
if minutes < 90: # print minutes, seconds up to 90 minutes
724
return '%d minute, %d second%s %s' % (
725
minutes, seconds, plural_seconds, direction)
727
return '%d minutes, %d second%s %s' % (
728
minutes, seconds, plural_seconds, direction)
730
hours = int(minutes / 60)
731
minutes -= 60 * hours
738
return '%d hour, %d minute%s %s' % (hours, minutes,
739
plural_minutes, direction)
740
return '%d hours, %d minute%s %s' % (hours, minutes,
741
plural_minutes, direction)
744
"""Return size of given open file."""
745
return os.fstat(f.fileno())[ST_SIZE]
748
# Define rand_bytes based on platform.
750
# Python 2.4 and later have os.urandom,
751
# but it doesn't work on some arches
753
rand_bytes = os.urandom
754
except (NotImplementedError, AttributeError):
755
# If python doesn't have os.urandom, or it doesn't work,
756
# then try to first pull random data from /dev/urandom
758
rand_bytes = file('/dev/urandom', 'rb').read
759
# Otherwise, use this hack as a last resort
760
except (IOError, OSError):
761
# not well seeded, but better than nothing
766
s += chr(random.randint(0, 255))
771
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
773
"""Return a random string of num alphanumeric characters
775
The result only contains lowercase chars because it may be used on
776
case-insensitive filesystems.
779
for raw_byte in rand_bytes(num):
780
s += ALNUM[ord(raw_byte) % 36]
784
## TODO: We could later have path objects that remember their list
785
## decomposition (might be too tricksy though.)
788
"""Turn string into list of parts."""
789
assert isinstance(p, basestring)
791
# split on either delimiter because people might use either on
793
ps = re.split(r'[\\/]', p)
798
raise errors.BzrError("sorry, %r not allowed in path" % f)
799
elif (f == '.') or (f == ''):
806
assert isinstance(p, (list, tuple))
808
if (f == '..') or (f is None) or (f == ''):
809
raise errors.BzrError("sorry, %r not allowed in path" % f)
814
"""Split s into lines, but without removing the newline characters."""
815
lines = s.split('\n')
816
result = [line + '\n' for line in lines[:-1]]
818
result.append(lines[-1])
822
def hardlinks_good():
823
return sys.platform not in ('win32', 'cygwin', 'darwin')
826
def link_or_copy(src, dest):
827
"""Hardlink a file, or copy it if it can't be hardlinked."""
828
if not hardlinks_good():
829
shutil.copyfile(src, dest)
833
except (OSError, IOError), e:
834
if e.errno != errno.EXDEV:
836
shutil.copyfile(src, dest)
839
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
840
# Forgiveness than Permission (EAFP) because:
841
# - root can damage a solaris file system by using unlink,
842
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
843
# EACCES, OSX: EPERM) when invoked on a directory.
844
def delete_any(path):
845
"""Delete a file or directory."""
846
if isdir(path): # Takes care of symlinks
853
if getattr(os, 'symlink', None) is not None:
859
def contains_whitespace(s):
860
"""True if there are any whitespace characters in s."""
861
# string.whitespace can include '\xa0' in certain locales, because it is
862
# considered "non-breaking-space" as part of ISO-8859-1. But it
863
# 1) Isn't a breaking whitespace
864
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
866
# 3) '\xa0' isn't unicode safe since it is >128.
868
# This should *not* be a unicode set of characters in case the source
869
# string is not a Unicode string. We can auto-up-cast the characters since
870
# they are ascii, but we don't want to auto-up-cast the string in case it
872
for ch in ' \t\n\r\v\f':
879
def contains_linebreaks(s):
880
"""True if there is any vertical whitespace in s."""
888
def relpath(base, path):
889
"""Return path relative to base, or raise exception.
891
The path may be either an absolute path or a path relative to the
892
current working directory.
894
os.path.commonprefix (python2.4) has a bad bug that it works just
895
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
899
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
900
' exceed the platform minimum length (which is %d)' %
907
while len(head) >= len(base):
910
head, tail = os.path.split(head)
914
raise errors.PathNotChild(rp, base)
922
def safe_unicode(unicode_or_utf8_string):
923
"""Coerce unicode_or_utf8_string into unicode.
925
If it is unicode, it is returned.
926
Otherwise it is decoded from utf-8. If a decoding error
927
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
928
as a BzrBadParameter exception.
930
if isinstance(unicode_or_utf8_string, unicode):
931
return unicode_or_utf8_string
933
return unicode_or_utf8_string.decode('utf8')
934
except UnicodeDecodeError:
935
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
938
def safe_utf8(unicode_or_utf8_string):
939
"""Coerce unicode_or_utf8_string to a utf8 string.
941
If it is a str, it is returned.
942
If it is Unicode, it is encoded into a utf-8 string.
944
if isinstance(unicode_or_utf8_string, str):
945
# TODO: jam 20070209 This is overkill, and probably has an impact on
946
# performance if we are dealing with lots of apis that want a
949
# Make sure it is a valid utf-8 string
950
unicode_or_utf8_string.decode('utf-8')
951
except UnicodeDecodeError:
952
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
953
return unicode_or_utf8_string
954
return unicode_or_utf8_string.encode('utf-8')
957
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
958
' Revision id generators should be creating utf8'
962
def safe_revision_id(unicode_or_utf8_string, warn=True):
963
"""Revision ids should now be utf8, but at one point they were unicode.
965
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
967
:param warn: Functions that are sanitizing user data can set warn=False
968
:return: None or a utf8 revision id.
970
if (unicode_or_utf8_string is None
971
or unicode_or_utf8_string.__class__ == str):
972
return unicode_or_utf8_string
974
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
976
return cache_utf8.encode(unicode_or_utf8_string)
979
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
980
' generators should be creating utf8 file ids.')
983
def safe_file_id(unicode_or_utf8_string, warn=True):
984
"""File ids should now be utf8, but at one point they were unicode.
986
This is the same as safe_utf8, except it uses the cached encode functions
987
to save a little bit of performance.
989
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
991
:param warn: Functions that are sanitizing user data can set warn=False
992
:return: None or a utf8 file id.
994
if (unicode_or_utf8_string is None
995
or unicode_or_utf8_string.__class__ == str):
996
return unicode_or_utf8_string
998
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1000
return cache_utf8.encode(unicode_or_utf8_string)
1003
_platform_normalizes_filenames = False
1004
if sys.platform == 'darwin':
1005
_platform_normalizes_filenames = True
1008
def normalizes_filenames():
1009
"""Return True if this platform normalizes unicode filenames.
1011
Mac OSX does, Windows/Linux do not.
1013
return _platform_normalizes_filenames
1016
def _accessible_normalized_filename(path):
1017
"""Get the unicode normalized path, and if you can access the file.
1019
On platforms where the system normalizes filenames (Mac OSX),
1020
you can access a file by any path which will normalize correctly.
1021
On platforms where the system does not normalize filenames
1022
(Windows, Linux), you have to access a file by its exact path.
1024
Internally, bzr only supports NFC normalization, since that is
1025
the standard for XML documents.
1027
So return the normalized path, and a flag indicating if the file
1028
can be accessed by that path.
1031
return unicodedata.normalize('NFC', unicode(path)), True
1034
def _inaccessible_normalized_filename(path):
1035
__doc__ = _accessible_normalized_filename.__doc__
1037
normalized = unicodedata.normalize('NFC', unicode(path))
1038
return normalized, normalized == path
1041
if _platform_normalizes_filenames:
1042
normalized_filename = _accessible_normalized_filename
1044
normalized_filename = _inaccessible_normalized_filename
1047
def terminal_width():
1048
"""Return estimated terminal width."""
1049
if sys.platform == 'win32':
1050
return win32utils.get_console_size()[0]
1053
import struct, fcntl, termios
1054
s = struct.pack('HHHH', 0, 0, 0, 0)
1055
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1056
width = struct.unpack('HHHH', x)[1]
1061
width = int(os.environ['COLUMNS'])
1070
def supports_executable():
1071
return sys.platform != "win32"
1074
def supports_posix_readonly():
1075
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1077
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1078
directory controls creation/deletion, etc.
1080
And under win32, readonly means that the directory itself cannot be
1081
deleted. The contents of a readonly directory can be changed, unlike POSIX
1082
where files in readonly directories cannot be added, deleted or renamed.
1084
return sys.platform != "win32"
1087
def set_or_unset_env(env_variable, value):
1088
"""Modify the environment, setting or removing the env_variable.
1090
:param env_variable: The environment variable in question
1091
:param value: The value to set the environment to. If None, then
1092
the variable will be removed.
1093
:return: The original value of the environment variable.
1095
orig_val = os.environ.get(env_variable)
1097
if orig_val is not None:
1098
del os.environ[env_variable]
1100
if isinstance(value, unicode):
1101
value = value.encode(get_user_encoding())
1102
os.environ[env_variable] = value
1106
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1109
def check_legal_path(path):
1110
"""Check whether the supplied path is legal.
1111
This is only required on Windows, so we don't test on other platforms
1114
if sys.platform != "win32":
1116
if _validWin32PathRE.match(path) is None:
1117
raise errors.IllegalPath(path)
1120
def walkdirs(top, prefix=""):
1121
"""Yield data about all the directories in a tree.
1123
This yields all the data about the contents of a directory at a time.
1124
After each directory has been yielded, if the caller has mutated the list
1125
to exclude some directories, they are then not descended into.
1127
The data yielded is of the form:
1128
((directory-relpath, directory-path-from-top),
1129
[(relpath, basename, kind, lstat, path-from-top), ...]),
1130
- directory-relpath is the relative path of the directory being returned
1131
with respect to top. prefix is prepended to this.
1132
- directory-path-from-root is the path including top for this directory.
1133
It is suitable for use with os functions.
1134
- relpath is the relative path within the subtree being walked.
1135
- basename is the basename of the path
1136
- kind is the kind of the file now. If unknown then the file is not
1137
present within the tree - but it may be recorded as versioned. See
1139
- lstat is the stat data *if* the file was statted.
1140
- planned, not implemented:
1141
path_from_tree_root is the path from the root of the tree.
1143
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1144
allows one to walk a subtree but get paths that are relative to a tree
1146
:return: an iterator over the dirs.
1148
#TODO there is a bit of a smell where the results of the directory-
1149
# summary in this, and the path from the root, may not agree
1150
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1151
# potentially confusing output. We should make this more robust - but
1152
# not at a speed cost. RBC 20060731
1154
_directory = _directory_kind
1155
_listdir = os.listdir
1156
_kind_from_mode = _formats.get
1157
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1159
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1160
relroot, _, _, _, top = pending.pop()
1162
relprefix = relroot + u'/'
1165
top_slash = top + u'/'
1168
append = dirblock.append
1169
for name in sorted(_listdir(top)):
1170
abspath = top_slash + name
1171
statvalue = _lstat(abspath)
1172
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1173
append((relprefix + name, name, kind, statvalue, abspath))
1174
yield (relroot, top), dirblock
1176
# push the user specified dirs from dirblock
1177
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1180
def _walkdirs_utf8(top, prefix=""):
1181
"""Yield data about all the directories in a tree.
1183
This yields the same information as walkdirs() only each entry is yielded
1184
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1185
are returned as exact byte-strings.
1187
:return: yields a tuple of (dir_info, [file_info])
1188
dir_info is (utf8_relpath, path-from-top)
1189
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1190
if top is an absolute path, path-from-top is also an absolute path.
1191
path-from-top might be unicode or utf8, but it is the correct path to
1192
pass to os functions to affect the file in question. (such as os.lstat)
1194
fs_encoding = _fs_enc.upper()
1195
if (sys.platform == 'win32' or
1196
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1197
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1199
return _walkdirs_fs_utf8(top, prefix=prefix)
1202
def _walkdirs_fs_utf8(top, prefix=""):
1203
"""See _walkdirs_utf8.
1205
This sub-function is called when we know the filesystem is already in utf8
1206
encoding. So we don't need to transcode filenames.
1209
_directory = _directory_kind
1210
_listdir = os.listdir
1211
_kind_from_mode = _formats.get
1213
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1214
# But we don't actually uses 1-3 in pending, so set them to None
1215
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1217
relroot, _, _, _, top = pending.pop()
1219
relprefix = relroot + '/'
1222
top_slash = top + '/'
1225
append = dirblock.append
1226
for name in sorted(_listdir(top)):
1227
abspath = top_slash + name
1228
statvalue = _lstat(abspath)
1229
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1230
append((relprefix + name, name, kind, statvalue, abspath))
1231
yield (relroot, top), dirblock
1233
# push the user specified dirs from dirblock
1234
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1237
def _walkdirs_unicode_to_utf8(top, prefix=""):
1238
"""See _walkdirs_utf8
1240
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1242
This is currently the fallback code path when the filesystem encoding is
1243
not UTF-8. It may be better to implement an alternative so that we can
1244
safely handle paths that are not properly decodable in the current
1247
_utf8_encode = codecs.getencoder('utf8')
1249
_directory = _directory_kind
1250
_listdir = os.listdir
1251
_kind_from_mode = _formats.get
1253
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1255
relroot, _, _, _, top = pending.pop()
1257
relprefix = relroot + '/'
1260
top_slash = top + u'/'
1263
append = dirblock.append
1264
for name in sorted(_listdir(top)):
1265
name_utf8 = _utf8_encode(name)[0]
1266
abspath = top_slash + name
1267
statvalue = _lstat(abspath)
1268
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1269
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1270
yield (relroot, top), dirblock
1272
# push the user specified dirs from dirblock
1273
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1276
def copy_tree(from_path, to_path, handlers={}):
1277
"""Copy all of the entries in from_path into to_path.
1279
:param from_path: The base directory to copy.
1280
:param to_path: The target directory. If it does not exist, it will
1282
:param handlers: A dictionary of functions, which takes a source and
1283
destinations for files, directories, etc.
1284
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1285
'file', 'directory', and 'symlink' should always exist.
1286
If they are missing, they will be replaced with 'os.mkdir()',
1287
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1289
# Now, just copy the existing cached tree to the new location
1290
# We use a cheap trick here.
1291
# Absolute paths are prefixed with the first parameter
1292
# relative paths are prefixed with the second.
1293
# So we can get both the source and target returned
1294
# without any extra work.
1296
def copy_dir(source, dest):
1299
def copy_link(source, dest):
1300
"""Copy the contents of a symlink"""
1301
link_to = os.readlink(source)
1302
os.symlink(link_to, dest)
1304
real_handlers = {'file':shutil.copy2,
1305
'symlink':copy_link,
1306
'directory':copy_dir,
1308
real_handlers.update(handlers)
1310
if not os.path.exists(to_path):
1311
real_handlers['directory'](from_path, to_path)
1313
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1314
for relpath, name, kind, st, abspath in entries:
1315
real_handlers[kind](abspath, relpath)
1318
def path_prefix_key(path):
1319
"""Generate a prefix-order path key for path.
1321
This can be used to sort paths in the same way that walkdirs does.
1323
return (dirname(path) , path)
1326
def compare_paths_prefix_order(path_a, path_b):
1327
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1328
key_a = path_prefix_key(path_a)
1329
key_b = path_prefix_key(path_b)
1330
return cmp(key_a, key_b)
1333
_cached_user_encoding = None
1336
def get_user_encoding(use_cache=True):
1337
"""Find out what the preferred user encoding is.
1339
This is generally the encoding that is used for command line parameters
1340
and file contents. This may be different from the terminal encoding
1341
or the filesystem encoding.
1343
:param use_cache: Enable cache for detected encoding.
1344
(This parameter is turned on by default,
1345
and required only for selftesting)
1347
:return: A string defining the preferred user encoding
1349
global _cached_user_encoding
1350
if _cached_user_encoding is not None and use_cache:
1351
return _cached_user_encoding
1353
if sys.platform == 'darwin':
1354
# work around egregious python 2.4 bug
1355
sys.platform = 'posix'
1359
sys.platform = 'darwin'
1364
user_encoding = locale.getpreferredencoding()
1365
except locale.Error, e:
1366
sys.stderr.write('bzr: warning: %s\n'
1367
' Could not determine what text encoding to use.\n'
1368
' This error usually means your Python interpreter\n'
1369
' doesn\'t support the locale set by $LANG (%s)\n'
1370
" Continuing with ascii encoding.\n"
1371
% (e, os.environ.get('LANG')))
1372
user_encoding = 'ascii'
1374
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1375
# treat that as ASCII, and not support printing unicode characters to the
1377
if user_encoding in (None, 'cp0'):
1378
user_encoding = 'ascii'
1382
codecs.lookup(user_encoding)
1384
sys.stderr.write('bzr: warning:'
1385
' unknown encoding %s.'
1386
' Continuing with ascii encoding.\n'
1389
user_encoding = 'ascii'
1392
_cached_user_encoding = user_encoding
1394
return user_encoding
1397
def recv_all(socket, bytes):
1398
"""Receive an exact number of bytes.
1400
Regular Socket.recv() may return less than the requested number of bytes,
1401
dependning on what's in the OS buffer. MSG_WAITALL is not available
1402
on all platforms, but this should work everywhere. This will return
1403
less than the requested amount if the remote end closes.
1405
This isn't optimized and is intended mostly for use in testing.
1408
while len(b) < bytes:
1409
new = socket.recv(bytes - len(b))
1416
def send_all(socket, bytes):
1417
"""Send all bytes on a socket.
1419
Regular socket.sendall() can give socket error 10053 on Windows. This
1420
implementation sends no more than 64k at a time, which avoids this problem.
1423
for pos in xrange(0, len(bytes), chunk_size):
1424
socket.sendall(bytes[pos:pos+chunk_size])
1427
def dereference_path(path):
1428
"""Determine the real path to a file.
1430
All parent elements are dereferenced. But the file itself is not
1432
:param path: The original path. May be absolute or relative.
1433
:return: the real path *to* the file
1435
parent, base = os.path.split(path)
1436
# The pathjoin for '.' is a workaround for Python bug #1213894.
1437
# (initial path components aren't dereferenced)
1438
return pathjoin(realpath(pathjoin('.', parent)), base)
1441
def supports_mapi():
1442
"""Return True if we can use MAPI to launch a mail client."""
1443
return sys.platform == "win32"
1446
def resource_string(package, resource_name):
1447
"""Load a resource from a package and return it as a string.
1449
Note: Only packages that start with bzrlib are currently supported.
1451
This is designed to be a lightweight implementation of resource
1452
loading in a way which is API compatible with the same API from
1454
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1455
If and when pkg_resources becomes a standard library, this routine
1458
# Check package name is within bzrlib
1459
if package == "bzrlib":
1460
resource_relpath = resource_name
1461
elif package.startswith("bzrlib."):
1462
package = package[len("bzrlib."):].replace('.', os.sep)
1463
resource_relpath = pathjoin(package, resource_name)
1465
raise errors.BzrError('resource package %s not in bzrlib' % package)
1467
# Map the resource to a file and read its contents
1468
base = dirname(bzrlib.__file__)
1469
if getattr(sys, 'frozen', None): # bzr.exe
1470
base = abspath(pathjoin(base, '..', '..'))
1471
filename = pathjoin(base, resource_relpath)
1472
return open(filename, 'rU').read()
added
removed
bzrlib/osutils.py
