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,
43
from tempfile import (
55
# sha is deprecated in python2.6 but hashlib is available as of 2.5
56
if sys.version_info < (2, 5):
69
from bzrlib import symbol_versioning
70
from bzrlib.symbol_versioning import (
74
from bzrlib.trace import mutter
77
# On win32, O_BINARY is used to indicate the file should
78
# be opened in binary mode, rather than text mode.
79
# On other platforms, O_BINARY doesn't exist, because
80
# they always open in binary mode, so it is okay to
81
# OR with 0 on those platforms
82
O_BINARY = getattr(os, 'O_BINARY', 0)
85
def make_readonly(filename):
86
"""Make a filename read-only."""
87
mod = os.lstat(filename).st_mode
88
if not stat.S_ISLNK(mod):
90
os.chmod(filename, mod)
93
def make_writable(filename):
94
mod = os.lstat(filename).st_mode
95
if not stat.S_ISLNK(mod):
97
os.chmod(filename, mod)
100
def minimum_path_selection(paths):
101
"""Return the smallset subset of paths which are outside paths.
103
:param paths: A container (and hence not None) of paths.
104
:return: A set of paths sufficient to include everything in paths via
105
is_inside_any, drawn from the paths parameter.
110
other_paths = paths.difference([path])
111
if not is_inside_any(other_paths, path):
112
# this is a top level path, we must check it.
113
search_paths.add(path)
121
"""Return a quoted filename filename
123
This previously used backslash quoting, but that works poorly on
125
# TODO: I'm not really sure this is the best format either.x
127
if _QUOTE_RE is None:
128
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
130
if _QUOTE_RE.search(f):
136
_directory_kind = 'directory'
139
stat.S_IFDIR:_directory_kind,
140
stat.S_IFCHR:'chardev',
141
stat.S_IFBLK:'block',
144
stat.S_IFLNK:'symlink',
145
stat.S_IFSOCK:'socket',
149
def file_kind_from_stat_mode(stat_mode, _formats=_formats, _unknown='unknown'):
150
"""Generate a file kind from a stat mode. This is used in walkdirs.
152
Its performance is critical: Do not mutate without careful benchmarking.
155
return _formats[stat_mode & 0170000]
160
def file_kind(f, _lstat=os.lstat, _mapper=file_kind_from_stat_mode):
162
return _mapper(_lstat(f).st_mode)
164
if getattr(e, 'errno', None) == errno.ENOENT:
165
raise errors.NoSuchFile(f)
170
"""Return the current umask"""
171
# Assume that people aren't messing with the umask while running
172
# XXX: This is not thread safe, but there is no way to get the
173
# umask without setting it
181
_directory_kind: "/",
183
'tree-reference': '+',
187
def kind_marker(kind):
189
return _kind_marker_map[kind]
191
raise errors.BzrError('invalid file kind %r' % kind)
194
lexists = getattr(os.path, 'lexists', None)
198
stat = getattr(os, 'lstat', os.stat)
202
if e.errno == errno.ENOENT:
205
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
208
def fancy_rename(old, new, rename_func, unlink_func):
209
"""A fancy rename, when you don't have atomic rename.
211
:param old: The old path, to rename from
212
:param new: The new path, to rename to
213
:param rename_func: The potentially non-atomic rename function
214
:param unlink_func: A way to delete the target file if the full rename succeeds
217
# sftp rename doesn't allow overwriting, so play tricks:
219
base = os.path.basename(new)
220
dirname = os.path.dirname(new)
221
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
222
tmp_name = pathjoin(dirname, tmp_name)
224
# Rename the file out of the way, but keep track if it didn't exist
225
# We don't want to grab just any exception
226
# something like EACCES should prevent us from continuing
227
# The downside is that the rename_func has to throw an exception
228
# with an errno = ENOENT, or NoSuchFile
231
rename_func(new, tmp_name)
232
except (errors.NoSuchFile,), e:
235
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
236
# function raises an IOError with errno is None when a rename fails.
237
# This then gets caught here.
238
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
241
if (getattr(e, 'errno', None) is None
242
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
250
# This may throw an exception, in which case success will
252
rename_func(old, new)
254
except (IOError, OSError), e:
255
# source and target may be aliases of each other (e.g. on a
256
# case-insensitive filesystem), so we may have accidentally renamed
257
# source by when we tried to rename target
258
if not (file_existed and e.errno in (None, errno.ENOENT)):
262
# If the file used to exist, rename it back into place
263
# otherwise just delete it from the tmp location
265
unlink_func(tmp_name)
267
rename_func(tmp_name, new)
270
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
271
# choke on a Unicode string containing a relative path if
272
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
274
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
275
def _posix_abspath(path):
276
# jam 20060426 rather than encoding to fsencoding
277
# copy posixpath.abspath, but use os.getcwdu instead
278
if not posixpath.isabs(path):
279
path = posixpath.join(getcwd(), path)
280
return posixpath.normpath(path)
283
def _posix_realpath(path):
284
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
287
def _win32_fixdrive(path):
288
"""Force drive letters to be consistent.
290
win32 is inconsistent whether it returns lower or upper case
291
and even if it was consistent the user might type the other
292
so we force it to uppercase
293
running python.exe under cmd.exe return capital C:\\
294
running win32 python inside a cygwin shell returns lowercase c:\\
296
drive, path = _nt_splitdrive(path)
297
return drive.upper() + path
300
def _win32_abspath(path):
301
# Real _nt_abspath doesn't have a problem with a unicode cwd
302
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
305
def _win98_abspath(path):
306
"""Return the absolute version of a path.
307
Windows 98 safe implementation (python reimplementation
308
of Win32 API function GetFullPathNameW)
313
# \\HOST\path => //HOST/path
314
# //HOST/path => //HOST/path
315
# path => C:/cwd/path
318
# check for absolute path
319
drive = _nt_splitdrive(path)[0]
320
if drive == '' and path[:2] not in('//','\\\\'):
322
# we cannot simply os.path.join cwd and path
323
# because os.path.join('C:','/path') produce '/path'
324
# and this is incorrect
325
if path[:1] in ('/','\\'):
326
cwd = _nt_splitdrive(cwd)[0]
328
path = cwd + '\\' + path
329
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
331
if win32utils.winver == 'Windows 98':
332
_win32_abspath = _win98_abspath
335
def _win32_realpath(path):
336
# Real _nt_realpath doesn't have a problem with a unicode cwd
337
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
340
def _win32_pathjoin(*args):
341
return _nt_join(*args).replace('\\', '/')
344
def _win32_normpath(path):
345
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
349
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
352
def _win32_mkdtemp(*args, **kwargs):
353
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
356
def _win32_rename(old, new):
357
"""We expect to be able to atomically replace 'new' with old.
359
On win32, if new exists, it must be moved out of the way first,
363
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
365
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
366
# If we try to rename a non-existant file onto cwd, we get
367
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
368
# if the old path doesn't exist, sometimes we get EACCES
369
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
375
return unicodedata.normalize('NFKC', os.getcwdu())
378
# Default is to just use the python builtins, but these can be rebound on
379
# particular platforms.
380
abspath = _posix_abspath
381
realpath = _posix_realpath
382
pathjoin = os.path.join
383
normpath = os.path.normpath
386
dirname = os.path.dirname
387
basename = os.path.basename
388
split = os.path.split
389
splitext = os.path.splitext
390
# These were already imported into local scope
391
# mkdtemp = tempfile.mkdtemp
392
# rmtree = shutil.rmtree
394
MIN_ABS_PATHLENGTH = 1
397
if sys.platform == 'win32':
398
abspath = _win32_abspath
399
realpath = _win32_realpath
400
pathjoin = _win32_pathjoin
401
normpath = _win32_normpath
402
getcwd = _win32_getcwd
403
mkdtemp = _win32_mkdtemp
404
rename = _win32_rename
406
MIN_ABS_PATHLENGTH = 3
408
def _win32_delete_readonly(function, path, excinfo):
409
"""Error handler for shutil.rmtree function [for win32]
410
Helps to remove files and dirs marked as read-only.
412
exception = excinfo[1]
413
if function in (os.remove, os.rmdir) \
414
and isinstance(exception, OSError) \
415
and exception.errno == errno.EACCES:
421
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
422
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
423
return shutil.rmtree(path, ignore_errors, onerror)
424
elif sys.platform == 'darwin':
428
def get_terminal_encoding():
429
"""Find the best encoding for printing to the screen.
431
This attempts to check both sys.stdout and sys.stdin to see
432
what encoding they are in, and if that fails it falls back to
433
bzrlib.user_encoding.
434
The problem is that on Windows, locale.getpreferredencoding()
435
is not the same encoding as that used by the console:
436
http://mail.python.org/pipermail/python-list/2003-May/162357.html
438
On my standard US Windows XP, the preferred encoding is
439
cp1252, but the console is cp437
441
output_encoding = getattr(sys.stdout, 'encoding', None)
442
if not output_encoding:
443
input_encoding = getattr(sys.stdin, 'encoding', None)
444
if not input_encoding:
445
output_encoding = bzrlib.user_encoding
446
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
448
output_encoding = input_encoding
449
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
451
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
452
if output_encoding == 'cp0':
453
# invalid encoding (cp0 means 'no codepage' on Windows)
454
output_encoding = bzrlib.user_encoding
455
mutter('cp0 is invalid encoding.'
456
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
459
codecs.lookup(output_encoding)
461
sys.stderr.write('bzr: warning:'
462
' unknown terminal encoding %s.\n'
463
' Using encoding %s instead.\n'
464
% (output_encoding, bzrlib.user_encoding)
466
output_encoding = bzrlib.user_encoding
468
return output_encoding
471
def normalizepath(f):
472
if getattr(os.path, 'realpath', None) is not None:
476
[p,e] = os.path.split(f)
477
if e == "" or e == "." or e == "..":
480
return pathjoin(F(p), e)
483
@deprecated_function(one_zero)
485
"""Copy a file to a backup.
487
Backups are named in GNU-style, with a ~ suffix.
489
If the file is already a backup, it's not copied.
495
if has_symlinks() and os.path.islink(fn):
496
target = os.readlink(fn)
497
os.symlink(target, bfn)
505
outf = file(bfn, 'wb')
513
"""True if f is an accessible directory."""
515
return S_ISDIR(os.lstat(f)[ST_MODE])
521
"""True if f is a regular file."""
523
return S_ISREG(os.lstat(f)[ST_MODE])
528
"""True if f is a symlink."""
530
return S_ISLNK(os.lstat(f)[ST_MODE])
534
def is_inside(dir, fname):
535
"""True if fname is inside dir.
537
The parameters should typically be passed to osutils.normpath first, so
538
that . and .. and repeated slashes are eliminated, and the separators
539
are canonical for the platform.
541
The empty string as a dir name is taken as top-of-tree and matches
544
# XXX: Most callers of this can actually do something smarter by
545
# looking at the inventory
555
return fname.startswith(dir)
558
def is_inside_any(dir_list, fname):
559
"""True if fname is inside any of given dirs."""
560
for dirname in dir_list:
561
if is_inside(dirname, fname):
566
def is_inside_or_parent_of_any(dir_list, fname):
567
"""True if fname is a child or a parent of any of the given files."""
568
for dirname in dir_list:
569
if is_inside(dirname, fname) or is_inside(fname, dirname):
574
def pumpfile(fromfile, tofile):
575
"""Copy contents of one file to another.
577
:return: The number of bytes copied.
582
b = fromfile.read(BUFSIZE)
590
def file_iterator(input_file, readsize=32768):
592
b = input_file.read(readsize)
599
if getattr(f, 'tell', None) is not None:
611
def sha_file_by_name(fname):
612
"""Calculate the SHA1 of a file by reading the full text"""
614
f = os.open(fname, os.O_RDONLY | O_BINARY)
617
b = os.read(f, 1<<16)
625
def sha_strings(strings, _factory=sha):
626
"""Return the sha-1 of concatenation of strings"""
628
map(s.update, strings)
632
def sha_string(f, _factory=sha):
633
return _factory(f).hexdigest()
636
def fingerprint_file(f):
638
return {'size': len(b),
639
'sha1': sha(b).hexdigest()}
642
def compare_files(a, b):
643
"""Returns true if equal in contents"""
654
def local_time_offset(t=None):
655
"""Return offset of local zone from GMT, either at present or at time t."""
658
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
659
return offset.days * 86400 + offset.seconds
662
def format_date(t, offset=0, timezone='original', date_fmt=None,
664
"""Return a formatted date string.
666
:param t: Seconds since the epoch.
667
:param offset: Timezone offset in seconds east of utc.
668
:param timezone: How to display the time: 'utc', 'original' for the
669
timezone specified by offset, or 'local' for the process's current
671
:param show_offset: Whether to append the timezone.
672
:param date_fmt: strftime format.
674
if timezone == 'utc':
677
elif timezone == 'original':
680
tt = time.gmtime(t + offset)
681
elif timezone == 'local':
682
tt = time.localtime(t)
683
offset = local_time_offset(t)
685
raise errors.UnsupportedTimezoneFormat(timezone)
687
date_fmt = "%a %Y-%m-%d %H:%M:%S"
689
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
692
return (time.strftime(date_fmt, tt) + offset_str)
695
def compact_date(when):
696
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
699
def format_delta(delta):
700
"""Get a nice looking string for a time delta.
702
:param delta: The time difference in seconds, can be positive or negative.
703
positive indicates time in the past, negative indicates time in the
704
future. (usually time.time() - stored_time)
705
:return: String formatted to show approximate resolution
711
direction = 'in the future'
715
if seconds < 90: # print seconds up to 90 seconds
717
return '%d second %s' % (seconds, direction,)
719
return '%d seconds %s' % (seconds, direction)
721
minutes = int(seconds / 60)
722
seconds -= 60 * minutes
727
if minutes < 90: # print minutes, seconds up to 90 minutes
729
return '%d minute, %d second%s %s' % (
730
minutes, seconds, plural_seconds, direction)
732
return '%d minutes, %d second%s %s' % (
733
minutes, seconds, plural_seconds, direction)
735
hours = int(minutes / 60)
736
minutes -= 60 * hours
743
return '%d hour, %d minute%s %s' % (hours, minutes,
744
plural_minutes, direction)
745
return '%d hours, %d minute%s %s' % (hours, minutes,
746
plural_minutes, direction)
749
"""Return size of given open file."""
750
return os.fstat(f.fileno())[ST_SIZE]
753
# Define rand_bytes based on platform.
755
# Python 2.4 and later have os.urandom,
756
# but it doesn't work on some arches
758
rand_bytes = os.urandom
759
except (NotImplementedError, AttributeError):
760
# If python doesn't have os.urandom, or it doesn't work,
761
# then try to first pull random data from /dev/urandom
763
rand_bytes = file('/dev/urandom', 'rb').read
764
# Otherwise, use this hack as a last resort
765
except (IOError, OSError):
766
# not well seeded, but better than nothing
771
s += chr(random.randint(0, 255))
776
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
778
"""Return a random string of num alphanumeric characters
780
The result only contains lowercase chars because it may be used on
781
case-insensitive filesystems.
784
for raw_byte in rand_bytes(num):
785
s += ALNUM[ord(raw_byte) % 36]
789
## TODO: We could later have path objects that remember their list
790
## decomposition (might be too tricksy though.)
793
"""Turn string into list of parts."""
794
assert isinstance(p, basestring)
796
# split on either delimiter because people might use either on
798
ps = re.split(r'[\\/]', p)
803
raise errors.BzrError("sorry, %r not allowed in path" % f)
804
elif (f == '.') or (f == ''):
811
assert isinstance(p, (list, tuple))
813
if (f == '..') or (f is None) or (f == ''):
814
raise errors.BzrError("sorry, %r not allowed in path" % f)
819
"""Split s into lines, but without removing the newline characters."""
820
lines = s.split('\n')
821
result = [line + '\n' for line in lines[:-1]]
823
result.append(lines[-1])
827
def hardlinks_good():
828
return sys.platform not in ('win32', 'cygwin', 'darwin')
831
def link_or_copy(src, dest):
832
"""Hardlink a file, or copy it if it can't be hardlinked."""
833
if not hardlinks_good():
834
shutil.copyfile(src, dest)
838
except (OSError, IOError), e:
839
if e.errno != errno.EXDEV:
841
shutil.copyfile(src, dest)
844
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
845
# Forgiveness than Permission (EAFP) because:
846
# - root can damage a solaris file system by using unlink,
847
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
848
# EACCES, OSX: EPERM) when invoked on a directory.
849
def delete_any(path):
850
"""Delete a file or directory."""
851
if isdir(path): # Takes care of symlinks
858
if getattr(os, 'symlink', None) is not None:
864
def contains_whitespace(s):
865
"""True if there are any whitespace characters in s."""
866
# string.whitespace can include '\xa0' in certain locales, because it is
867
# considered "non-breaking-space" as part of ISO-8859-1. But it
868
# 1) Isn't a breaking whitespace
869
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
871
# 3) '\xa0' isn't unicode safe since it is >128.
873
# This should *not* be a unicode set of characters in case the source
874
# string is not a Unicode string. We can auto-up-cast the characters since
875
# they are ascii, but we don't want to auto-up-cast the string in case it
877
for ch in ' \t\n\r\v\f':
884
def contains_linebreaks(s):
885
"""True if there is any vertical whitespace in s."""
893
def relpath(base, path):
894
"""Return path relative to base, or raise exception.
896
The path may be either an absolute path or a path relative to the
897
current working directory.
899
os.path.commonprefix (python2.4) has a bad bug that it works just
900
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
904
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
905
' exceed the platform minimum length (which is %d)' %
912
while len(head) >= len(base):
915
head, tail = os.path.split(head)
919
raise errors.PathNotChild(rp, base)
927
def safe_unicode(unicode_or_utf8_string):
928
"""Coerce unicode_or_utf8_string into unicode.
930
If it is unicode, it is returned.
931
Otherwise it is decoded from utf-8. If a decoding error
932
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
933
as a BzrBadParameter exception.
935
if isinstance(unicode_or_utf8_string, unicode):
936
return unicode_or_utf8_string
938
return unicode_or_utf8_string.decode('utf8')
939
except UnicodeDecodeError:
940
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
943
def safe_utf8(unicode_or_utf8_string):
944
"""Coerce unicode_or_utf8_string to a utf8 string.
946
If it is a str, it is returned.
947
If it is Unicode, it is encoded into a utf-8 string.
949
if isinstance(unicode_or_utf8_string, str):
950
# TODO: jam 20070209 This is overkill, and probably has an impact on
951
# performance if we are dealing with lots of apis that want a
954
# Make sure it is a valid utf-8 string
955
unicode_or_utf8_string.decode('utf-8')
956
except UnicodeDecodeError:
957
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
958
return unicode_or_utf8_string
959
return unicode_or_utf8_string.encode('utf-8')
962
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
963
' Revision id generators should be creating utf8'
967
def safe_revision_id(unicode_or_utf8_string, warn=True):
968
"""Revision ids should now be utf8, but at one point they were unicode.
970
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
972
:param warn: Functions that are sanitizing user data can set warn=False
973
:return: None or a utf8 revision id.
975
if (unicode_or_utf8_string is None
976
or unicode_or_utf8_string.__class__ == str):
977
return unicode_or_utf8_string
979
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
981
return cache_utf8.encode(unicode_or_utf8_string)
984
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
985
' generators should be creating utf8 file ids.')
988
def safe_file_id(unicode_or_utf8_string, warn=True):
989
"""File ids should now be utf8, but at one point they were unicode.
991
This is the same as safe_utf8, except it uses the cached encode functions
992
to save a little bit of performance.
994
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
996
:param warn: Functions that are sanitizing user data can set warn=False
997
:return: None or a utf8 file id.
999
if (unicode_or_utf8_string is None
1000
or unicode_or_utf8_string.__class__ == str):
1001
return unicode_or_utf8_string
1003
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1005
return cache_utf8.encode(unicode_or_utf8_string)
1008
_platform_normalizes_filenames = False
1009
if sys.platform == 'darwin':
1010
_platform_normalizes_filenames = True
1013
def normalizes_filenames():
1014
"""Return True if this platform normalizes unicode filenames.
1016
Mac OSX does, Windows/Linux do not.
1018
return _platform_normalizes_filenames
1021
def _accessible_normalized_filename(path):
1022
"""Get the unicode normalized path, and if you can access the file.
1024
On platforms where the system normalizes filenames (Mac OSX),
1025
you can access a file by any path which will normalize correctly.
1026
On platforms where the system does not normalize filenames
1027
(Windows, Linux), you have to access a file by its exact path.
1029
Internally, bzr only supports NFC/NFKC normalization, since that is
1030
the standard for XML documents.
1032
So return the normalized path, and a flag indicating if the file
1033
can be accessed by that path.
1036
return unicodedata.normalize('NFKC', unicode(path)), True
1039
def _inaccessible_normalized_filename(path):
1040
__doc__ = _accessible_normalized_filename.__doc__
1042
normalized = unicodedata.normalize('NFKC', unicode(path))
1043
return normalized, normalized == path
1046
if _platform_normalizes_filenames:
1047
normalized_filename = _accessible_normalized_filename
1049
normalized_filename = _inaccessible_normalized_filename
1052
def terminal_width():
1053
"""Return estimated terminal width."""
1054
if sys.platform == 'win32':
1055
return win32utils.get_console_size()[0]
1058
import struct, fcntl, termios
1059
s = struct.pack('HHHH', 0, 0, 0, 0)
1060
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1061
width = struct.unpack('HHHH', x)[1]
1066
width = int(os.environ['COLUMNS'])
1075
def supports_executable():
1076
return sys.platform != "win32"
1079
def supports_posix_readonly():
1080
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1082
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1083
directory controls creation/deletion, etc.
1085
And under win32, readonly means that the directory itself cannot be
1086
deleted. The contents of a readonly directory can be changed, unlike POSIX
1087
where files in readonly directories cannot be added, deleted or renamed.
1089
return sys.platform != "win32"
1092
def set_or_unset_env(env_variable, value):
1093
"""Modify the environment, setting or removing the env_variable.
1095
:param env_variable: The environment variable in question
1096
:param value: The value to set the environment to. If None, then
1097
the variable will be removed.
1098
:return: The original value of the environment variable.
1100
orig_val = os.environ.get(env_variable)
1102
if orig_val is not None:
1103
del os.environ[env_variable]
1105
if isinstance(value, unicode):
1106
value = value.encode(bzrlib.user_encoding)
1107
os.environ[env_variable] = value
1111
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1114
def check_legal_path(path):
1115
"""Check whether the supplied path is legal.
1116
This is only required on Windows, so we don't test on other platforms
1119
if sys.platform != "win32":
1121
if _validWin32PathRE.match(path) is None:
1122
raise errors.IllegalPath(path)
1125
def walkdirs(top, prefix=""):
1126
"""Yield data about all the directories in a tree.
1128
This yields all the data about the contents of a directory at a time.
1129
After each directory has been yielded, if the caller has mutated the list
1130
to exclude some directories, they are then not descended into.
1132
The data yielded is of the form:
1133
((directory-relpath, directory-path-from-top),
1134
[(relpath, basename, kind, lstat, path-from-top), ...]),
1135
- directory-relpath is the relative path of the directory being returned
1136
with respect to top. prefix is prepended to this.
1137
- directory-path-from-root is the path including top for this directory.
1138
It is suitable for use with os functions.
1139
- relpath is the relative path within the subtree being walked.
1140
- basename is the basename of the path
1141
- kind is the kind of the file now. If unknown then the file is not
1142
present within the tree - but it may be recorded as versioned. See
1144
- lstat is the stat data *if* the file was statted.
1145
- planned, not implemented:
1146
path_from_tree_root is the path from the root of the tree.
1148
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1149
allows one to walk a subtree but get paths that are relative to a tree
1151
:return: an iterator over the dirs.
1153
#TODO there is a bit of a smell where the results of the directory-
1154
# summary in this, and the path from the root, may not agree
1155
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1156
# potentially confusing output. We should make this more robust - but
1157
# not at a speed cost. RBC 20060731
1159
_directory = _directory_kind
1160
_listdir = os.listdir
1161
_kind_from_mode = _formats.get
1162
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1164
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1165
relroot, _, _, _, top = pending.pop()
1167
relprefix = relroot + u'/'
1170
top_slash = top + u'/'
1173
append = dirblock.append
1174
for name in sorted(_listdir(top)):
1175
abspath = top_slash + name
1176
statvalue = _lstat(abspath)
1177
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1178
append((relprefix + name, name, kind, statvalue, abspath))
1179
yield (relroot, top), dirblock
1181
# push the user specified dirs from dirblock
1182
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1185
def _walkdirs_utf8(top, prefix=""):
1186
"""Yield data about all the directories in a tree.
1188
This yields the same information as walkdirs() only each entry is yielded
1189
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1190
are returned as exact byte-strings.
1192
:return: yields a tuple of (dir_info, [file_info])
1193
dir_info is (utf8_relpath, path-from-top)
1194
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1195
if top is an absolute path, path-from-top is also an absolute path.
1196
path-from-top might be unicode or utf8, but it is the correct path to
1197
pass to os functions to affect the file in question. (such as os.lstat)
1199
fs_encoding = _fs_enc.upper()
1200
if (sys.platform == 'win32' or
1201
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1202
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1204
return _walkdirs_fs_utf8(top, prefix=prefix)
1207
def _walkdirs_fs_utf8(top, prefix=""):
1208
"""See _walkdirs_utf8.
1210
This sub-function is called when we know the filesystem is already in utf8
1211
encoding. So we don't need to transcode filenames.
1214
_directory = _directory_kind
1215
_listdir = os.listdir
1216
_kind_from_mode = _formats.get
1218
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1219
# But we don't actually uses 1-3 in pending, so set them to None
1220
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1222
relroot, _, _, _, top = pending.pop()
1224
relprefix = relroot + '/'
1227
top_slash = top + '/'
1230
append = dirblock.append
1231
for name in sorted(_listdir(top)):
1232
abspath = top_slash + name
1233
statvalue = _lstat(abspath)
1234
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1235
append((relprefix + name, name, kind, statvalue, abspath))
1236
yield (relroot, top), dirblock
1238
# push the user specified dirs from dirblock
1239
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1242
def _walkdirs_unicode_to_utf8(top, prefix=""):
1243
"""See _walkdirs_utf8
1245
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1247
This is currently the fallback code path when the filesystem encoding is
1248
not UTF-8. It may be better to implement an alternative so that we can
1249
safely handle paths that are not properly decodable in the current
1252
_utf8_encode = codecs.getencoder('utf8')
1254
_directory = _directory_kind
1255
_listdir = os.listdir
1256
_kind_from_mode = _formats.get
1258
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1260
relroot, _, _, _, top = pending.pop()
1262
relprefix = relroot + '/'
1265
top_slash = top + u'/'
1268
append = dirblock.append
1269
for name in sorted(_listdir(top)):
1270
name_utf8 = _utf8_encode(name)[0]
1271
abspath = top_slash + name
1272
statvalue = _lstat(abspath)
1273
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1274
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1275
yield (relroot, top), dirblock
1277
# push the user specified dirs from dirblock
1278
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1281
def copy_tree(from_path, to_path, handlers={}):
1282
"""Copy all of the entries in from_path into to_path.
1284
:param from_path: The base directory to copy.
1285
:param to_path: The target directory. If it does not exist, it will
1287
:param handlers: A dictionary of functions, which takes a source and
1288
destinations for files, directories, etc.
1289
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1290
'file', 'directory', and 'symlink' should always exist.
1291
If they are missing, they will be replaced with 'os.mkdir()',
1292
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1294
# Now, just copy the existing cached tree to the new location
1295
# We use a cheap trick here.
1296
# Absolute paths are prefixed with the first parameter
1297
# relative paths are prefixed with the second.
1298
# So we can get both the source and target returned
1299
# without any extra work.
1301
def copy_dir(source, dest):
1304
def copy_link(source, dest):
1305
"""Copy the contents of a symlink"""
1306
link_to = os.readlink(source)
1307
os.symlink(link_to, dest)
1309
real_handlers = {'file':shutil.copy2,
1310
'symlink':copy_link,
1311
'directory':copy_dir,
1313
real_handlers.update(handlers)
1315
if not os.path.exists(to_path):
1316
real_handlers['directory'](from_path, to_path)
1318
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1319
for relpath, name, kind, st, abspath in entries:
1320
real_handlers[kind](abspath, relpath)
1323
def path_prefix_key(path):
1324
"""Generate a prefix-order path key for path.
1326
This can be used to sort paths in the same way that walkdirs does.
1328
return (dirname(path) , path)
1331
def compare_paths_prefix_order(path_a, path_b):
1332
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1333
key_a = path_prefix_key(path_a)
1334
key_b = path_prefix_key(path_b)
1335
return cmp(key_a, key_b)
1338
_cached_user_encoding = None
1341
def get_user_encoding(use_cache=True):
1342
"""Find out what the preferred user encoding is.
1344
This is generally the encoding that is used for command line parameters
1345
and file contents. This may be different from the terminal encoding
1346
or the filesystem encoding.
1348
:param use_cache: Enable cache for detected encoding.
1349
(This parameter is turned on by default,
1350
and required only for selftesting)
1352
:return: A string defining the preferred user encoding
1354
global _cached_user_encoding
1355
if _cached_user_encoding is not None and use_cache:
1356
return _cached_user_encoding
1358
if sys.platform == 'darwin':
1359
# work around egregious python 2.4 bug
1360
sys.platform = 'posix'
1364
sys.platform = 'darwin'
1369
user_encoding = locale.getpreferredencoding()
1370
except locale.Error, e:
1371
sys.stderr.write('bzr: warning: %s\n'
1372
' Could not determine what text encoding to use.\n'
1373
' This error usually means your Python interpreter\n'
1374
' doesn\'t support the locale set by $LANG (%s)\n'
1375
" Continuing with ascii encoding.\n"
1376
% (e, os.environ.get('LANG')))
1377
user_encoding = 'ascii'
1379
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1380
# treat that as ASCII, and not support printing unicode characters to the
1382
if user_encoding in (None, 'cp0'):
1383
user_encoding = 'ascii'
1387
codecs.lookup(user_encoding)
1389
sys.stderr.write('bzr: warning:'
1390
' unknown encoding %s.'
1391
' Continuing with ascii encoding.\n'
1394
user_encoding = 'ascii'
1397
_cached_user_encoding = user_encoding
1399
return user_encoding
1402
def recv_all(socket, bytes):
1403
"""Receive an exact number of bytes.
1405
Regular Socket.recv() may return less than the requested number of bytes,
1406
dependning on what's in the OS buffer. MSG_WAITALL is not available
1407
on all platforms, but this should work everywhere. This will return
1408
less than the requested amount if the remote end closes.
1410
This isn't optimized and is intended mostly for use in testing.
1413
while len(b) < bytes:
1414
new = socket.recv(bytes - len(b))
1421
def send_all(socket, bytes):
1422
"""Send all bytes on a socket.
1424
Regular socket.sendall() can give socket error 10053 on Windows. This
1425
implementation sends no more than 64k at a time, which avoids this problem.
1428
for pos in xrange(0, len(bytes), chunk_size):
1429
socket.sendall(bytes[pos:pos+chunk_size])
1432
def dereference_path(path):
1433
"""Determine the real path to a file.
1435
All parent elements are dereferenced. But the file itself is not
1437
:param path: The original path. May be absolute or relative.
1438
:return: the real path *to* the file
1440
parent, base = os.path.split(path)
1441
# The pathjoin for '.' is a workaround for Python bug #1213894.
1442
# (initial path components aren't dereferenced)
1443
return pathjoin(realpath(pathjoin('.', parent)), base)
1446
def supports_mapi():
1447
"""Return True if we can use MAPI to launch a mail client."""
1448
return sys.platform == "win32"
1451
def resource_string(package, resource_name):
1452
"""Load a resource from a package and return it as a string.
1454
Note: Only packages that start with bzrlib are currently supported.
1456
This is designed to be a lightweight implementation of resource
1457
loading in a way which is API compatible with the same API from
1459
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1460
If and when pkg_resources becomes a standard library, this routine
1463
# Check package name is within bzrlib
1464
if package == "bzrlib":
1465
resource_relpath = resource_name
1466
elif package.startswith("bzrlib."):
1467
package = package[len("bzrlib."):].replace('.', os.sep)
1468
resource_relpath = pathjoin(package, resource_name)
1470
raise errors.BzrError('resource package %s not in bzrlib' % package)
1472
# Map the resource to a file and read its contents
1473
base = dirname(bzrlib.__file__)
1474
if getattr(sys, 'frozen', None): # bzr.exe
1475
base = abspath(pathjoin(base, '..', '..'))
1476
filename = pathjoin(base, resource_relpath)
1477
return open(filename, 'rU').read()
added
removed
bzrlib/osutils.py
