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 (
61
from bzrlib.trace import mutter
64
# On win32, O_BINARY is used to indicate the file should
65
# be opened in binary mode, rather than text mode.
66
# On other platforms, O_BINARY doesn't exist, because
67
# they always open in binary mode, so it is okay to
68
# OR with 0 on those platforms
69
O_BINARY = getattr(os, 'O_BINARY', 0)
72
def make_readonly(filename):
73
"""Make a filename read-only."""
74
mod = os.lstat(filename).st_mode
75
if not stat.S_ISLNK(mod):
77
os.chmod(filename, mod)
80
def make_writable(filename):
81
mod = os.lstat(filename).st_mode
82
if not stat.S_ISLNK(mod):
84
os.chmod(filename, mod)
87
def minimum_path_selection(paths):
88
"""Return the smallset subset of paths which are outside paths.
90
:param paths: A container (and hence not None) of paths.
91
:return: A set of paths sufficient to include everything in paths via
92
is_inside_any, drawn from the paths parameter.
97
other_paths = paths.difference([path])
98
if not is_inside_any(other_paths, path):
99
# this is a top level path, we must check it.
100
search_paths.add(path)
108
"""Return a quoted filename filename
110
This previously used backslash quoting, but that works poorly on
112
# TODO: I'm not really sure this is the best format either.x
114
if _QUOTE_RE is None:
115
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
117
if _QUOTE_RE.search(f):
123
_directory_kind = 'directory'
126
stat.S_IFDIR:_directory_kind,
127
stat.S_IFCHR:'chardev',
128
stat.S_IFBLK:'block',
131
stat.S_IFLNK:'symlink',
132
stat.S_IFSOCK:'socket',
136
def file_kind_from_stat_mode(stat_mode, _formats=_formats, _unknown='unknown'):
137
"""Generate a file kind from a stat mode. This is used in walkdirs.
139
Its performance is critical: Do not mutate without careful benchmarking.
142
return _formats[stat_mode & 0170000]
147
def file_kind(f, _lstat=os.lstat, _mapper=file_kind_from_stat_mode):
149
return _mapper(_lstat(f).st_mode)
151
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
152
raise errors.NoSuchFile(f)
157
"""Return the current umask"""
158
# Assume that people aren't messing with the umask while running
159
# XXX: This is not thread safe, but there is no way to get the
160
# umask without setting it
168
_directory_kind: "/",
170
'tree-reference': '+',
174
def kind_marker(kind):
176
return _kind_marker_map[kind]
178
raise errors.BzrError('invalid file kind %r' % kind)
181
lexists = getattr(os.path, 'lexists', None)
185
stat = getattr(os, 'lstat', os.stat)
189
if e.errno == errno.ENOENT:
192
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
195
def fancy_rename(old, new, rename_func, unlink_func):
196
"""A fancy rename, when you don't have atomic rename.
198
:param old: The old path, to rename from
199
:param new: The new path, to rename to
200
:param rename_func: The potentially non-atomic rename function
201
:param unlink_func: A way to delete the target file if the full rename succeeds
204
# sftp rename doesn't allow overwriting, so play tricks:
206
base = os.path.basename(new)
207
dirname = os.path.dirname(new)
208
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
209
tmp_name = pathjoin(dirname, tmp_name)
211
# Rename the file out of the way, but keep track if it didn't exist
212
# We don't want to grab just any exception
213
# something like EACCES should prevent us from continuing
214
# The downside is that the rename_func has to throw an exception
215
# with an errno = ENOENT, or NoSuchFile
218
rename_func(new, tmp_name)
219
except (errors.NoSuchFile,), e:
222
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
223
# function raises an IOError with errno is None when a rename fails.
224
# This then gets caught here.
225
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
228
if (getattr(e, 'errno', None) is None
229
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
237
# This may throw an exception, in which case success will
239
rename_func(old, new)
241
except (IOError, OSError), e:
242
# source and target may be aliases of each other (e.g. on a
243
# case-insensitive filesystem), so we may have accidentally renamed
244
# source by when we tried to rename target
245
if not (file_existed and e.errno in (None, errno.ENOENT)):
249
# If the file used to exist, rename it back into place
250
# otherwise just delete it from the tmp location
252
unlink_func(tmp_name)
254
rename_func(tmp_name, new)
257
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
258
# choke on a Unicode string containing a relative path if
259
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
261
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
262
def _posix_abspath(path):
263
# jam 20060426 rather than encoding to fsencoding
264
# copy posixpath.abspath, but use os.getcwdu instead
265
if not posixpath.isabs(path):
266
path = posixpath.join(getcwd(), path)
267
return posixpath.normpath(path)
270
def _posix_realpath(path):
271
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
274
def _win32_fixdrive(path):
275
"""Force drive letters to be consistent.
277
win32 is inconsistent whether it returns lower or upper case
278
and even if it was consistent the user might type the other
279
so we force it to uppercase
280
running python.exe under cmd.exe return capital C:\\
281
running win32 python inside a cygwin shell returns lowercase c:\\
283
drive, path = _nt_splitdrive(path)
284
return drive.upper() + path
287
def _win32_abspath(path):
288
# Real _nt_abspath doesn't have a problem with a unicode cwd
289
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
292
def _win98_abspath(path):
293
"""Return the absolute version of a path.
294
Windows 98 safe implementation (python reimplementation
295
of Win32 API function GetFullPathNameW)
300
# \\HOST\path => //HOST/path
301
# //HOST/path => //HOST/path
302
# path => C:/cwd/path
305
# check for absolute path
306
drive = _nt_splitdrive(path)[0]
307
if drive == '' and path[:2] not in('//','\\\\'):
309
# we cannot simply os.path.join cwd and path
310
# because os.path.join('C:','/path') produce '/path'
311
# and this is incorrect
312
if path[:1] in ('/','\\'):
313
cwd = _nt_splitdrive(cwd)[0]
315
path = cwd + '\\' + path
316
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
318
if win32utils.winver == 'Windows 98':
319
_win32_abspath = _win98_abspath
322
def _win32_realpath(path):
323
# Real _nt_realpath doesn't have a problem with a unicode cwd
324
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
327
def _win32_pathjoin(*args):
328
return _nt_join(*args).replace('\\', '/')
331
def _win32_normpath(path):
332
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
336
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
339
def _win32_mkdtemp(*args, **kwargs):
340
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
343
def _win32_rename(old, new):
344
"""We expect to be able to atomically replace 'new' with old.
346
On win32, if new exists, it must be moved out of the way first,
350
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
352
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
353
# If we try to rename a non-existant file onto cwd, we get
354
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
355
# if the old path doesn't exist, sometimes we get EACCES
356
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
362
return unicodedata.normalize('NFC', os.getcwdu())
365
# Default is to just use the python builtins, but these can be rebound on
366
# particular platforms.
367
abspath = _posix_abspath
368
realpath = _posix_realpath
369
pathjoin = os.path.join
370
normpath = os.path.normpath
373
dirname = os.path.dirname
374
basename = os.path.basename
375
split = os.path.split
376
splitext = os.path.splitext
377
# These were already imported into local scope
378
# mkdtemp = tempfile.mkdtemp
379
# rmtree = shutil.rmtree
381
MIN_ABS_PATHLENGTH = 1
384
if sys.platform == 'win32':
385
abspath = _win32_abspath
386
realpath = _win32_realpath
387
pathjoin = _win32_pathjoin
388
normpath = _win32_normpath
389
getcwd = _win32_getcwd
390
mkdtemp = _win32_mkdtemp
391
rename = _win32_rename
393
MIN_ABS_PATHLENGTH = 3
395
def _win32_delete_readonly(function, path, excinfo):
396
"""Error handler for shutil.rmtree function [for win32]
397
Helps to remove files and dirs marked as read-only.
399
exception = excinfo[1]
400
if function in (os.remove, os.rmdir) \
401
and isinstance(exception, OSError) \
402
and exception.errno == errno.EACCES:
408
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
409
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
410
return shutil.rmtree(path, ignore_errors, onerror)
411
elif sys.platform == 'darwin':
415
def get_terminal_encoding():
416
"""Find the best encoding for printing to the screen.
418
This attempts to check both sys.stdout and sys.stdin to see
419
what encoding they are in, and if that fails it falls back to
420
bzrlib.user_encoding.
421
The problem is that on Windows, locale.getpreferredencoding()
422
is not the same encoding as that used by the console:
423
http://mail.python.org/pipermail/python-list/2003-May/162357.html
425
On my standard US Windows XP, the preferred encoding is
426
cp1252, but the console is cp437
428
output_encoding = getattr(sys.stdout, 'encoding', None)
429
if not output_encoding:
430
input_encoding = getattr(sys.stdin, 'encoding', None)
431
if not input_encoding:
432
output_encoding = bzrlib.user_encoding
433
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
435
output_encoding = input_encoding
436
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
438
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
439
if output_encoding == 'cp0':
440
# invalid encoding (cp0 means 'no codepage' on Windows)
441
output_encoding = bzrlib.user_encoding
442
mutter('cp0 is invalid encoding.'
443
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
446
codecs.lookup(output_encoding)
448
sys.stderr.write('bzr: warning:'
449
' unknown terminal encoding %s.\n'
450
' Using encoding %s instead.\n'
451
% (output_encoding, bzrlib.user_encoding)
453
output_encoding = bzrlib.user_encoding
455
return output_encoding
458
def normalizepath(f):
459
if getattr(os.path, 'realpath', None) is not None:
463
[p,e] = os.path.split(f)
464
if e == "" or e == "." or e == "..":
467
return pathjoin(F(p), e)
471
"""True if f is an accessible directory."""
473
return S_ISDIR(os.lstat(f)[ST_MODE])
479
"""True if f is a regular file."""
481
return S_ISREG(os.lstat(f)[ST_MODE])
486
"""True if f is a symlink."""
488
return S_ISLNK(os.lstat(f)[ST_MODE])
492
def is_inside(dir, fname):
493
"""True if fname is inside dir.
495
The parameters should typically be passed to osutils.normpath first, so
496
that . and .. and repeated slashes are eliminated, and the separators
497
are canonical for the platform.
499
The empty string as a dir name is taken as top-of-tree and matches
502
# XXX: Most callers of this can actually do something smarter by
503
# looking at the inventory
513
return fname.startswith(dir)
516
def is_inside_any(dir_list, fname):
517
"""True if fname is inside any of given dirs."""
518
for dirname in dir_list:
519
if is_inside(dirname, fname):
524
def is_inside_or_parent_of_any(dir_list, fname):
525
"""True if fname is a child or a parent of any of the given files."""
526
for dirname in dir_list:
527
if is_inside(dirname, fname) or is_inside(fname, dirname):
532
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768):
533
"""Copy contents of one file to another.
535
The read_length can either be -1 to read to end-of-file (EOF) or
536
it can specify the maximum number of bytes to read.
538
The buff_size represents the maximum size for each read operation
539
performed on from_file.
541
:return: The number of bytes copied.
545
# read specified number of bytes
547
while read_length > 0:
548
num_bytes_to_read = min(read_length, buff_size)
550
block = from_file.read(num_bytes_to_read)
556
actual_bytes_read = len(block)
557
read_length -= actual_bytes_read
558
length += actual_bytes_read
562
block = from_file.read(buff_size)
571
def file_iterator(input_file, readsize=32768):
573
b = input_file.read(readsize)
580
"""Calculate the hexdigest of an open file.
582
The file cursor should be already at the start.
594
def sha_file_by_name(fname):
595
"""Calculate the SHA1 of a file by reading the full text"""
597
f = os.open(fname, os.O_RDONLY | O_BINARY)
600
b = os.read(f, 1<<16)
608
def sha_strings(strings, _factory=sha.new):
609
"""Return the sha-1 of concatenation of strings"""
611
map(s.update, strings)
615
def sha_string(f, _factory=sha.new):
616
return _factory(f).hexdigest()
619
def fingerprint_file(f):
621
return {'size': len(b),
622
'sha1': sha.new(b).hexdigest()}
625
def compare_files(a, b):
626
"""Returns true if equal in contents"""
637
def local_time_offset(t=None):
638
"""Return offset of local zone from GMT, either at present or at time t."""
641
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
642
return offset.days * 86400 + offset.seconds
645
def format_date(t, offset=0, timezone='original', date_fmt=None,
647
"""Return a formatted date string.
649
:param t: Seconds since the epoch.
650
:param offset: Timezone offset in seconds east of utc.
651
:param timezone: How to display the time: 'utc', 'original' for the
652
timezone specified by offset, or 'local' for the process's current
654
:param show_offset: Whether to append the timezone.
655
:param date_fmt: strftime format.
657
if timezone == 'utc':
660
elif timezone == 'original':
663
tt = time.gmtime(t + offset)
664
elif timezone == 'local':
665
tt = time.localtime(t)
666
offset = local_time_offset(t)
668
raise errors.UnsupportedTimezoneFormat(timezone)
670
date_fmt = "%a %Y-%m-%d %H:%M:%S"
672
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
675
return (time.strftime(date_fmt, tt) + offset_str)
678
def compact_date(when):
679
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
682
def format_delta(delta):
683
"""Get a nice looking string for a time delta.
685
:param delta: The time difference in seconds, can be positive or negative.
686
positive indicates time in the past, negative indicates time in the
687
future. (usually time.time() - stored_time)
688
:return: String formatted to show approximate resolution
694
direction = 'in the future'
698
if seconds < 90: # print seconds up to 90 seconds
700
return '%d second %s' % (seconds, direction,)
702
return '%d seconds %s' % (seconds, direction)
704
minutes = int(seconds / 60)
705
seconds -= 60 * minutes
710
if minutes < 90: # print minutes, seconds up to 90 minutes
712
return '%d minute, %d second%s %s' % (
713
minutes, seconds, plural_seconds, direction)
715
return '%d minutes, %d second%s %s' % (
716
minutes, seconds, plural_seconds, direction)
718
hours = int(minutes / 60)
719
minutes -= 60 * hours
726
return '%d hour, %d minute%s %s' % (hours, minutes,
727
plural_minutes, direction)
728
return '%d hours, %d minute%s %s' % (hours, minutes,
729
plural_minutes, direction)
732
"""Return size of given open file."""
733
return os.fstat(f.fileno())[ST_SIZE]
736
# Define rand_bytes based on platform.
738
# Python 2.4 and later have os.urandom,
739
# but it doesn't work on some arches
741
rand_bytes = os.urandom
742
except (NotImplementedError, AttributeError):
743
# If python doesn't have os.urandom, or it doesn't work,
744
# then try to first pull random data from /dev/urandom
746
rand_bytes = file('/dev/urandom', 'rb').read
747
# Otherwise, use this hack as a last resort
748
except (IOError, OSError):
749
# not well seeded, but better than nothing
754
s += chr(random.randint(0, 255))
759
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
761
"""Return a random string of num alphanumeric characters
763
The result only contains lowercase chars because it may be used on
764
case-insensitive filesystems.
767
for raw_byte in rand_bytes(num):
768
s += ALNUM[ord(raw_byte) % 36]
772
## TODO: We could later have path objects that remember their list
773
## decomposition (might be too tricksy though.)
776
"""Turn string into list of parts."""
777
# split on either delimiter because people might use either on
779
ps = re.split(r'[\\/]', p)
784
raise errors.BzrError("sorry, %r not allowed in path" % f)
785
elif (f == '.') or (f == ''):
793
if (f == '..') or (f is None) or (f == ''):
794
raise errors.BzrError("sorry, %r not allowed in path" % f)
799
"""Split s into lines, but without removing the newline characters."""
800
lines = s.split('\n')
801
result = [line + '\n' for line in lines[:-1]]
803
result.append(lines[-1])
807
def hardlinks_good():
808
return sys.platform not in ('win32', 'cygwin', 'darwin')
811
def link_or_copy(src, dest):
812
"""Hardlink a file, or copy it if it can't be hardlinked."""
813
if not hardlinks_good():
814
shutil.copyfile(src, dest)
818
except (OSError, IOError), e:
819
if e.errno != errno.EXDEV:
821
shutil.copyfile(src, dest)
824
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
825
# Forgiveness than Permission (EAFP) because:
826
# - root can damage a solaris file system by using unlink,
827
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
828
# EACCES, OSX: EPERM) when invoked on a directory.
829
def delete_any(path):
830
"""Delete a file or directory."""
831
if isdir(path): # Takes care of symlinks
838
if getattr(os, 'symlink', None) is not None:
845
if getattr(os, 'link', None) is not None:
851
def contains_whitespace(s):
852
"""True if there are any whitespace characters in s."""
853
# string.whitespace can include '\xa0' in certain locales, because it is
854
# considered "non-breaking-space" as part of ISO-8859-1. But it
855
# 1) Isn't a breaking whitespace
856
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
858
# 3) '\xa0' isn't unicode safe since it is >128.
860
# This should *not* be a unicode set of characters in case the source
861
# string is not a Unicode string. We can auto-up-cast the characters since
862
# they are ascii, but we don't want to auto-up-cast the string in case it
864
for ch in ' \t\n\r\v\f':
871
def contains_linebreaks(s):
872
"""True if there is any vertical whitespace in s."""
880
def relpath(base, path):
881
"""Return path relative to base, or raise exception.
883
The path may be either an absolute path or a path relative to the
884
current working directory.
886
os.path.commonprefix (python2.4) has a bad bug that it works just
887
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
891
if len(base) < MIN_ABS_PATHLENGTH:
892
# must have space for e.g. a drive letter
893
raise ValueError('%r is too short to calculate a relative path'
900
while len(head) >= len(base):
903
head, tail = os.path.split(head)
907
raise errors.PathNotChild(rp, base)
915
def safe_unicode(unicode_or_utf8_string):
916
"""Coerce unicode_or_utf8_string into unicode.
918
If it is unicode, it is returned.
919
Otherwise it is decoded from utf-8. If a decoding error
920
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
921
as a BzrBadParameter exception.
923
if isinstance(unicode_or_utf8_string, unicode):
924
return unicode_or_utf8_string
926
return unicode_or_utf8_string.decode('utf8')
927
except UnicodeDecodeError:
928
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
931
def safe_utf8(unicode_or_utf8_string):
932
"""Coerce unicode_or_utf8_string to a utf8 string.
934
If it is a str, it is returned.
935
If it is Unicode, it is encoded into a utf-8 string.
937
if isinstance(unicode_or_utf8_string, str):
938
# TODO: jam 20070209 This is overkill, and probably has an impact on
939
# performance if we are dealing with lots of apis that want a
942
# Make sure it is a valid utf-8 string
943
unicode_or_utf8_string.decode('utf-8')
944
except UnicodeDecodeError:
945
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
946
return unicode_or_utf8_string
947
return unicode_or_utf8_string.encode('utf-8')
950
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
951
' Revision id generators should be creating utf8'
955
def safe_revision_id(unicode_or_utf8_string, warn=True):
956
"""Revision ids should now be utf8, but at one point they were unicode.
958
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
960
:param warn: Functions that are sanitizing user data can set warn=False
961
:return: None or a utf8 revision id.
963
if (unicode_or_utf8_string is None
964
or unicode_or_utf8_string.__class__ == str):
965
return unicode_or_utf8_string
967
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
969
return cache_utf8.encode(unicode_or_utf8_string)
972
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
973
' generators should be creating utf8 file ids.')
976
def safe_file_id(unicode_or_utf8_string, warn=True):
977
"""File ids should now be utf8, but at one point they were unicode.
979
This is the same as safe_utf8, except it uses the cached encode functions
980
to save a little bit of performance.
982
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
984
:param warn: Functions that are sanitizing user data can set warn=False
985
:return: None or a utf8 file id.
987
if (unicode_or_utf8_string is None
988
or unicode_or_utf8_string.__class__ == str):
989
return unicode_or_utf8_string
991
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
993
return cache_utf8.encode(unicode_or_utf8_string)
996
_platform_normalizes_filenames = False
997
if sys.platform == 'darwin':
998
_platform_normalizes_filenames = True
1001
def normalizes_filenames():
1002
"""Return True if this platform normalizes unicode filenames.
1004
Mac OSX does, Windows/Linux do not.
1006
return _platform_normalizes_filenames
1009
def _accessible_normalized_filename(path):
1010
"""Get the unicode normalized path, and if you can access the file.
1012
On platforms where the system normalizes filenames (Mac OSX),
1013
you can access a file by any path which will normalize correctly.
1014
On platforms where the system does not normalize filenames
1015
(Windows, Linux), you have to access a file by its exact path.
1017
Internally, bzr only supports NFC normalization, since that is
1018
the standard for XML documents.
1020
So return the normalized path, and a flag indicating if the file
1021
can be accessed by that path.
1024
return unicodedata.normalize('NFC', unicode(path)), True
1027
def _inaccessible_normalized_filename(path):
1028
__doc__ = _accessible_normalized_filename.__doc__
1030
normalized = unicodedata.normalize('NFC', unicode(path))
1031
return normalized, normalized == path
1034
if _platform_normalizes_filenames:
1035
normalized_filename = _accessible_normalized_filename
1037
normalized_filename = _inaccessible_normalized_filename
1040
def terminal_width():
1041
"""Return estimated terminal width."""
1042
if sys.platform == 'win32':
1043
return win32utils.get_console_size()[0]
1046
import struct, fcntl, termios
1047
s = struct.pack('HHHH', 0, 0, 0, 0)
1048
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1049
width = struct.unpack('HHHH', x)[1]
1054
width = int(os.environ['COLUMNS'])
1063
def supports_executable():
1064
return sys.platform != "win32"
1067
def supports_posix_readonly():
1068
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1070
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1071
directory controls creation/deletion, etc.
1073
And under win32, readonly means that the directory itself cannot be
1074
deleted. The contents of a readonly directory can be changed, unlike POSIX
1075
where files in readonly directories cannot be added, deleted or renamed.
1077
return sys.platform != "win32"
1080
def set_or_unset_env(env_variable, value):
1081
"""Modify the environment, setting or removing the env_variable.
1083
:param env_variable: The environment variable in question
1084
:param value: The value to set the environment to. If None, then
1085
the variable will be removed.
1086
:return: The original value of the environment variable.
1088
orig_val = os.environ.get(env_variable)
1090
if orig_val is not None:
1091
del os.environ[env_variable]
1093
if isinstance(value, unicode):
1094
value = value.encode(bzrlib.user_encoding)
1095
os.environ[env_variable] = value
1099
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1102
def check_legal_path(path):
1103
"""Check whether the supplied path is legal.
1104
This is only required on Windows, so we don't test on other platforms
1107
if sys.platform != "win32":
1109
if _validWin32PathRE.match(path) is None:
1110
raise errors.IllegalPath(path)
1113
def walkdirs(top, prefix=""):
1114
"""Yield data about all the directories in a tree.
1116
This yields all the data about the contents of a directory at a time.
1117
After each directory has been yielded, if the caller has mutated the list
1118
to exclude some directories, they are then not descended into.
1120
The data yielded is of the form:
1121
((directory-relpath, directory-path-from-top),
1122
[(relpath, basename, kind, lstat, path-from-top), ...]),
1123
- directory-relpath is the relative path of the directory being returned
1124
with respect to top. prefix is prepended to this.
1125
- directory-path-from-root is the path including top for this directory.
1126
It is suitable for use with os functions.
1127
- relpath is the relative path within the subtree being walked.
1128
- basename is the basename of the path
1129
- kind is the kind of the file now. If unknown then the file is not
1130
present within the tree - but it may be recorded as versioned. See
1132
- lstat is the stat data *if* the file was statted.
1133
- planned, not implemented:
1134
path_from_tree_root is the path from the root of the tree.
1136
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1137
allows one to walk a subtree but get paths that are relative to a tree
1139
:return: an iterator over the dirs.
1141
#TODO there is a bit of a smell where the results of the directory-
1142
# summary in this, and the path from the root, may not agree
1143
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1144
# potentially confusing output. We should make this more robust - but
1145
# not at a speed cost. RBC 20060731
1147
_directory = _directory_kind
1148
_listdir = os.listdir
1149
_kind_from_mode = _formats.get
1150
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1152
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1153
relroot, _, _, _, top = pending.pop()
1155
relprefix = relroot + u'/'
1158
top_slash = top + u'/'
1161
append = dirblock.append
1162
for name in sorted(_listdir(top)):
1163
abspath = top_slash + name
1164
statvalue = _lstat(abspath)
1165
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1166
append((relprefix + name, name, kind, statvalue, abspath))
1167
yield (relroot, top), dirblock
1169
# push the user specified dirs from dirblock
1170
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1173
def _walkdirs_utf8(top, prefix=""):
1174
"""Yield data about all the directories in a tree.
1176
This yields the same information as walkdirs() only each entry is yielded
1177
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1178
are returned as exact byte-strings.
1180
:return: yields a tuple of (dir_info, [file_info])
1181
dir_info is (utf8_relpath, path-from-top)
1182
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1183
if top is an absolute path, path-from-top is also an absolute path.
1184
path-from-top might be unicode or utf8, but it is the correct path to
1185
pass to os functions to affect the file in question. (such as os.lstat)
1187
fs_encoding = _fs_enc.upper()
1188
if (sys.platform == 'win32' or
1189
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1190
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1192
return _walkdirs_fs_utf8(top, prefix=prefix)
1195
def _walkdirs_fs_utf8(top, prefix=""):
1196
"""See _walkdirs_utf8.
1198
This sub-function is called when we know the filesystem is already in utf8
1199
encoding. So we don't need to transcode filenames.
1202
_directory = _directory_kind
1203
_listdir = os.listdir
1204
_kind_from_mode = _formats.get
1206
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1207
# But we don't actually uses 1-3 in pending, so set them to None
1208
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1210
relroot, _, _, _, top = pending.pop()
1212
relprefix = relroot + '/'
1215
top_slash = top + '/'
1218
append = dirblock.append
1219
for name in sorted(_listdir(top)):
1220
abspath = top_slash + name
1221
statvalue = _lstat(abspath)
1222
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1223
append((relprefix + name, name, kind, statvalue, abspath))
1224
yield (relroot, top), dirblock
1226
# push the user specified dirs from dirblock
1227
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1230
def _walkdirs_unicode_to_utf8(top, prefix=""):
1231
"""See _walkdirs_utf8
1233
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1235
This is currently the fallback code path when the filesystem encoding is
1236
not UTF-8. It may be better to implement an alternative so that we can
1237
safely handle paths that are not properly decodable in the current
1240
_utf8_encode = codecs.getencoder('utf8')
1242
_directory = _directory_kind
1243
_listdir = os.listdir
1244
_kind_from_mode = _formats.get
1246
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1248
relroot, _, _, _, top = pending.pop()
1250
relprefix = relroot + '/'
1253
top_slash = top + u'/'
1256
append = dirblock.append
1257
for name in sorted(_listdir(top)):
1258
name_utf8 = _utf8_encode(name)[0]
1259
abspath = top_slash + name
1260
statvalue = _lstat(abspath)
1261
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1262
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1263
yield (relroot, top), dirblock
1265
# push the user specified dirs from dirblock
1266
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1269
def copy_tree(from_path, to_path, handlers={}):
1270
"""Copy all of the entries in from_path into to_path.
1272
:param from_path: The base directory to copy.
1273
:param to_path: The target directory. If it does not exist, it will
1275
:param handlers: A dictionary of functions, which takes a source and
1276
destinations for files, directories, etc.
1277
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1278
'file', 'directory', and 'symlink' should always exist.
1279
If they are missing, they will be replaced with 'os.mkdir()',
1280
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1282
# Now, just copy the existing cached tree to the new location
1283
# We use a cheap trick here.
1284
# Absolute paths are prefixed with the first parameter
1285
# relative paths are prefixed with the second.
1286
# So we can get both the source and target returned
1287
# without any extra work.
1289
def copy_dir(source, dest):
1292
def copy_link(source, dest):
1293
"""Copy the contents of a symlink"""
1294
link_to = os.readlink(source)
1295
os.symlink(link_to, dest)
1297
real_handlers = {'file':shutil.copy2,
1298
'symlink':copy_link,
1299
'directory':copy_dir,
1301
real_handlers.update(handlers)
1303
if not os.path.exists(to_path):
1304
real_handlers['directory'](from_path, to_path)
1306
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1307
for relpath, name, kind, st, abspath in entries:
1308
real_handlers[kind](abspath, relpath)
1311
def path_prefix_key(path):
1312
"""Generate a prefix-order path key for path.
1314
This can be used to sort paths in the same way that walkdirs does.
1316
return (dirname(path) , path)
1319
def compare_paths_prefix_order(path_a, path_b):
1320
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1321
key_a = path_prefix_key(path_a)
1322
key_b = path_prefix_key(path_b)
1323
return cmp(key_a, key_b)
1326
_cached_user_encoding = None
1329
def get_user_encoding(use_cache=True):
1330
"""Find out what the preferred user encoding is.
1332
This is generally the encoding that is used for command line parameters
1333
and file contents. This may be different from the terminal encoding
1334
or the filesystem encoding.
1336
:param use_cache: Enable cache for detected encoding.
1337
(This parameter is turned on by default,
1338
and required only for selftesting)
1340
:return: A string defining the preferred user encoding
1342
global _cached_user_encoding
1343
if _cached_user_encoding is not None and use_cache:
1344
return _cached_user_encoding
1346
if sys.platform == 'darwin':
1347
# work around egregious python 2.4 bug
1348
sys.platform = 'posix'
1352
sys.platform = 'darwin'
1357
user_encoding = locale.getpreferredencoding()
1358
except locale.Error, e:
1359
sys.stderr.write('bzr: warning: %s\n'
1360
' Could not determine what text encoding to use.\n'
1361
' This error usually means your Python interpreter\n'
1362
' doesn\'t support the locale set by $LANG (%s)\n'
1363
" Continuing with ascii encoding.\n"
1364
% (e, os.environ.get('LANG')))
1365
user_encoding = 'ascii'
1367
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1368
# treat that as ASCII, and not support printing unicode characters to the
1371
# For python scripts run under vim, we get '', so also treat that as ASCII
1372
if user_encoding in (None, 'cp0', ''):
1373
user_encoding = 'ascii'
1377
codecs.lookup(user_encoding)
1379
sys.stderr.write('bzr: warning:'
1380
' unknown encoding %s.'
1381
' Continuing with ascii encoding.\n'
1384
user_encoding = 'ascii'
1387
_cached_user_encoding = user_encoding
1389
return user_encoding
1392
def recv_all(socket, bytes):
1393
"""Receive an exact number of bytes.
1395
Regular Socket.recv() may return less than the requested number of bytes,
1396
dependning on what's in the OS buffer. MSG_WAITALL is not available
1397
on all platforms, but this should work everywhere. This will return
1398
less than the requested amount if the remote end closes.
1400
This isn't optimized and is intended mostly for use in testing.
1403
while len(b) < bytes:
1404
new = socket.recv(bytes - len(b))
1411
def send_all(socket, bytes):
1412
"""Send all bytes on a socket.
1414
Regular socket.sendall() can give socket error 10053 on Windows. This
1415
implementation sends no more than 64k at a time, which avoids this problem.
1418
for pos in xrange(0, len(bytes), chunk_size):
1419
socket.sendall(bytes[pos:pos+chunk_size])
1422
def dereference_path(path):
1423
"""Determine the real path to a file.
1425
All parent elements are dereferenced. But the file itself is not
1427
:param path: The original path. May be absolute or relative.
1428
:return: the real path *to* the file
1430
parent, base = os.path.split(path)
1431
# The pathjoin for '.' is a workaround for Python bug #1213894.
1432
# (initial path components aren't dereferenced)
1433
return pathjoin(realpath(pathjoin('.', parent)), base)
1436
def supports_mapi():
1437
"""Return True if we can use MAPI to launch a mail client."""
1438
return sys.platform == "win32"
1441
def resource_string(package, resource_name):
1442
"""Load a resource from a package and return it as a string.
1444
Note: Only packages that start with bzrlib are currently supported.
1446
This is designed to be a lightweight implementation of resource
1447
loading in a way which is API compatible with the same API from
1449
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1450
If and when pkg_resources becomes a standard library, this routine
1453
# Check package name is within bzrlib
1454
if package == "bzrlib":
1455
resource_relpath = resource_name
1456
elif package.startswith("bzrlib."):
1457
package = package[len("bzrlib."):].replace('.', os.sep)
1458
resource_relpath = pathjoin(package, resource_name)
1460
raise errors.BzrError('resource package %s not in bzrlib' % package)
1462
# Map the resource to a file and read its contents
1463
base = dirname(bzrlib.__file__)
1464
if getattr(sys, 'frozen', None): # bzr.exe
1465
base = abspath(pathjoin(base, '..', '..'))
1466
filename = pathjoin(base, resource_relpath)
1467
return open(filename, 'rU').read()
added
removed
bzrlib/osutils.py
