1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
from cStringIO import StringIO
21
from stat import (S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE,
22
S_ISCHR, S_ISBLK, S_ISFIFO, S_ISSOCK)
26
from bzrlib.lazy_import import lazy_import
27
lazy_import(globals(), """
29
from datetime import datetime
31
from ntpath import (abspath as _nt_abspath,
33
normpath as _nt_normpath,
34
realpath as _nt_realpath,
35
splitdrive as _nt_splitdrive,
44
from tempfile import (
57
from bzrlib import symbol_versioning
58
from bzrlib.symbol_versioning import (
62
from bzrlib.trace import mutter
65
# On win32, O_BINARY is used to indicate the file should
66
# be opened in binary mode, rather than text mode.
67
# On other platforms, O_BINARY doesn't exist, because
68
# they always open in binary mode, so it is okay to
69
# OR with 0 on those platforms
70
O_BINARY = getattr(os, 'O_BINARY', 0)
73
def make_readonly(filename):
74
"""Make a filename read-only."""
75
mod = os.lstat(filename).st_mode
76
if not stat.S_ISLNK(mod):
78
os.chmod(filename, mod)
81
def make_writable(filename):
82
mod = os.lstat(filename).st_mode
83
if not stat.S_ISLNK(mod):
85
os.chmod(filename, mod)
88
def minimum_path_selection(paths):
89
"""Return the smallset subset of paths which are outside paths.
91
:param paths: A container (and hence not None) of paths.
92
:return: A set of paths sufficient to include everything in paths via
93
is_inside_any, drawn from the paths parameter.
98
other_paths = paths.difference([path])
99
if not is_inside_any(other_paths, path):
100
# this is a top level path, we must check it.
101
search_paths.add(path)
109
"""Return a quoted filename filename
111
This previously used backslash quoting, but that works poorly on
113
# TODO: I'm not really sure this is the best format either.x
115
if _QUOTE_RE is None:
116
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
118
if _QUOTE_RE.search(f):
124
_directory_kind = 'directory'
127
stat.S_IFDIR:_directory_kind,
128
stat.S_IFCHR:'chardev',
129
stat.S_IFBLK:'block',
132
stat.S_IFLNK:'symlink',
133
stat.S_IFSOCK:'socket',
137
def file_kind_from_stat_mode(stat_mode, _formats=_formats, _unknown='unknown'):
138
"""Generate a file kind from a stat mode. This is used in walkdirs.
140
Its performance is critical: Do not mutate without careful benchmarking.
143
return _formats[stat_mode & 0170000]
148
def file_kind(f, _lstat=os.lstat, _mapper=file_kind_from_stat_mode):
150
return _mapper(_lstat(f).st_mode)
152
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
153
raise errors.NoSuchFile(f)
158
"""Return the current umask"""
159
# Assume that people aren't messing with the umask while running
160
# XXX: This is not thread safe, but there is no way to get the
161
# umask without setting it
169
_directory_kind: "/",
171
'tree-reference': '+',
175
def kind_marker(kind):
177
return _kind_marker_map[kind]
179
raise errors.BzrError('invalid file kind %r' % kind)
182
lexists = getattr(os.path, 'lexists', None)
186
stat = getattr(os, 'lstat', os.stat)
190
if e.errno == errno.ENOENT:
193
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
196
def fancy_rename(old, new, rename_func, unlink_func):
197
"""A fancy rename, when you don't have atomic rename.
199
:param old: The old path, to rename from
200
:param new: The new path, to rename to
201
:param rename_func: The potentially non-atomic rename function
202
:param unlink_func: A way to delete the target file if the full rename succeeds
205
# sftp rename doesn't allow overwriting, so play tricks:
207
base = os.path.basename(new)
208
dirname = os.path.dirname(new)
209
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
210
tmp_name = pathjoin(dirname, tmp_name)
212
# Rename the file out of the way, but keep track if it didn't exist
213
# We don't want to grab just any exception
214
# something like EACCES should prevent us from continuing
215
# The downside is that the rename_func has to throw an exception
216
# with an errno = ENOENT, or NoSuchFile
219
rename_func(new, tmp_name)
220
except (errors.NoSuchFile,), e:
223
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
224
# function raises an IOError with errno is None when a rename fails.
225
# This then gets caught here.
226
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
229
if (getattr(e, 'errno', None) is None
230
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
238
# This may throw an exception, in which case success will
240
rename_func(old, new)
242
except (IOError, OSError), e:
243
# source and target may be aliases of each other (e.g. on a
244
# case-insensitive filesystem), so we may have accidentally renamed
245
# source by when we tried to rename target
246
if not (file_existed and e.errno in (None, errno.ENOENT)):
250
# If the file used to exist, rename it back into place
251
# otherwise just delete it from the tmp location
253
unlink_func(tmp_name)
255
rename_func(tmp_name, new)
258
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
259
# choke on a Unicode string containing a relative path if
260
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
262
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
263
def _posix_abspath(path):
264
# jam 20060426 rather than encoding to fsencoding
265
# copy posixpath.abspath, but use os.getcwdu instead
266
if not posixpath.isabs(path):
267
path = posixpath.join(getcwd(), path)
268
return posixpath.normpath(path)
271
def _posix_realpath(path):
272
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
275
def _win32_fixdrive(path):
276
"""Force drive letters to be consistent.
278
win32 is inconsistent whether it returns lower or upper case
279
and even if it was consistent the user might type the other
280
so we force it to uppercase
281
running python.exe under cmd.exe return capital C:\\
282
running win32 python inside a cygwin shell returns lowercase c:\\
284
drive, path = _nt_splitdrive(path)
285
return drive.upper() + path
288
def _win32_abspath(path):
289
# Real _nt_abspath doesn't have a problem with a unicode cwd
290
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
293
def _win98_abspath(path):
294
"""Return the absolute version of a path.
295
Windows 98 safe implementation (python reimplementation
296
of Win32 API function GetFullPathNameW)
301
# \\HOST\path => //HOST/path
302
# //HOST/path => //HOST/path
303
# path => C:/cwd/path
306
# check for absolute path
307
drive = _nt_splitdrive(path)[0]
308
if drive == '' and path[:2] not in('//','\\\\'):
310
# we cannot simply os.path.join cwd and path
311
# because os.path.join('C:','/path') produce '/path'
312
# and this is incorrect
313
if path[:1] in ('/','\\'):
314
cwd = _nt_splitdrive(cwd)[0]
316
path = cwd + '\\' + path
317
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
319
if win32utils.winver == 'Windows 98':
320
_win32_abspath = _win98_abspath
323
def _win32_realpath(path):
324
# Real _nt_realpath doesn't have a problem with a unicode cwd
325
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
328
def _win32_pathjoin(*args):
329
return _nt_join(*args).replace('\\', '/')
332
def _win32_normpath(path):
333
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
337
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
340
def _win32_mkdtemp(*args, **kwargs):
341
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
344
def _win32_rename(old, new):
345
"""We expect to be able to atomically replace 'new' with old.
347
On win32, if new exists, it must be moved out of the way first,
351
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
353
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
354
# If we try to rename a non-existant file onto cwd, we get
355
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
356
# if the old path doesn't exist, sometimes we get EACCES
357
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
363
return unicodedata.normalize('NFC', os.getcwdu())
366
# Default is to just use the python builtins, but these can be rebound on
367
# particular platforms.
368
abspath = _posix_abspath
369
realpath = _posix_realpath
370
pathjoin = os.path.join
371
normpath = os.path.normpath
374
dirname = os.path.dirname
375
basename = os.path.basename
376
split = os.path.split
377
splitext = os.path.splitext
378
# These were already imported into local scope
379
# mkdtemp = tempfile.mkdtemp
380
# rmtree = shutil.rmtree
382
MIN_ABS_PATHLENGTH = 1
385
if sys.platform == 'win32':
386
abspath = _win32_abspath
387
realpath = _win32_realpath
388
pathjoin = _win32_pathjoin
389
normpath = _win32_normpath
390
getcwd = _win32_getcwd
391
mkdtemp = _win32_mkdtemp
392
rename = _win32_rename
394
MIN_ABS_PATHLENGTH = 3
396
def _win32_delete_readonly(function, path, excinfo):
397
"""Error handler for shutil.rmtree function [for win32]
398
Helps to remove files and dirs marked as read-only.
400
exception = excinfo[1]
401
if function in (os.remove, os.rmdir) \
402
and isinstance(exception, OSError) \
403
and exception.errno == errno.EACCES:
409
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
410
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
411
return shutil.rmtree(path, ignore_errors, onerror)
412
elif sys.platform == 'darwin':
416
def get_terminal_encoding():
417
"""Find the best encoding for printing to the screen.
419
This attempts to check both sys.stdout and sys.stdin to see
420
what encoding they are in, and if that fails it falls back to
421
bzrlib.user_encoding.
422
The problem is that on Windows, locale.getpreferredencoding()
423
is not the same encoding as that used by the console:
424
http://mail.python.org/pipermail/python-list/2003-May/162357.html
426
On my standard US Windows XP, the preferred encoding is
427
cp1252, but the console is cp437
429
output_encoding = getattr(sys.stdout, 'encoding', None)
430
if not output_encoding:
431
input_encoding = getattr(sys.stdin, 'encoding', None)
432
if not input_encoding:
433
output_encoding = bzrlib.user_encoding
434
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
436
output_encoding = input_encoding
437
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
439
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
440
if output_encoding == 'cp0':
441
# invalid encoding (cp0 means 'no codepage' on Windows)
442
output_encoding = bzrlib.user_encoding
443
mutter('cp0 is invalid encoding.'
444
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
447
codecs.lookup(output_encoding)
449
sys.stderr.write('bzr: warning:'
450
' unknown terminal encoding %s.\n'
451
' Using encoding %s instead.\n'
452
% (output_encoding, bzrlib.user_encoding)
454
output_encoding = bzrlib.user_encoding
456
return output_encoding
459
def normalizepath(f):
460
if getattr(os.path, 'realpath', None) is not None:
464
[p,e] = os.path.split(f)
465
if e == "" or e == "." or e == "..":
468
return pathjoin(F(p), e)
472
"""True if f is an accessible directory."""
474
return S_ISDIR(os.lstat(f)[ST_MODE])
480
"""True if f is a regular file."""
482
return S_ISREG(os.lstat(f)[ST_MODE])
487
"""True if f is a symlink."""
489
return S_ISLNK(os.lstat(f)[ST_MODE])
493
def is_inside(dir, fname):
494
"""True if fname is inside dir.
496
The parameters should typically be passed to osutils.normpath first, so
497
that . and .. and repeated slashes are eliminated, and the separators
498
are canonical for the platform.
500
The empty string as a dir name is taken as top-of-tree and matches
503
# XXX: Most callers of this can actually do something smarter by
504
# looking at the inventory
514
return fname.startswith(dir)
517
def is_inside_any(dir_list, fname):
518
"""True if fname is inside any of given dirs."""
519
for dirname in dir_list:
520
if is_inside(dirname, fname):
525
def is_inside_or_parent_of_any(dir_list, fname):
526
"""True if fname is a child or a parent of any of the given files."""
527
for dirname in dir_list:
528
if is_inside(dirname, fname) or is_inside(fname, dirname):
533
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768):
534
"""Copy contents of one file to another.
536
The read_length can either be -1 to read to end-of-file (EOF) or
537
it can specify the maximum number of bytes to read.
539
The buff_size represents the maximum size for each read operation
540
performed on from_file.
542
:return: The number of bytes copied.
546
# read specified number of bytes
548
while read_length > 0:
549
num_bytes_to_read = min(read_length, buff_size)
551
block = from_file.read(num_bytes_to_read)
557
actual_bytes_read = len(block)
558
read_length -= actual_bytes_read
559
length += actual_bytes_read
563
block = from_file.read(buff_size)
572
def file_iterator(input_file, readsize=32768):
574
b = input_file.read(readsize)
581
if getattr(f, 'tell', None) is not None:
593
def sha_file_by_name(fname):
594
"""Calculate the SHA1 of a file by reading the full text"""
596
f = os.open(fname, os.O_RDONLY | O_BINARY)
599
b = os.read(f, 1<<16)
607
def sha_strings(strings, _factory=sha.new):
608
"""Return the sha-1 of concatenation of strings"""
610
map(s.update, strings)
614
def sha_string(f, _factory=sha.new):
615
return _factory(f).hexdigest()
618
def fingerprint_file(f):
620
return {'size': len(b),
621
'sha1': sha.new(b).hexdigest()}
624
def compare_files(a, b):
625
"""Returns true if equal in contents"""
636
def local_time_offset(t=None):
637
"""Return offset of local zone from GMT, either at present or at time t."""
640
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
641
return offset.days * 86400 + offset.seconds
644
def format_date(t, offset=0, timezone='original', date_fmt=None,
646
"""Return a formatted date string.
648
:param t: Seconds since the epoch.
649
:param offset: Timezone offset in seconds east of utc.
650
:param timezone: How to display the time: 'utc', 'original' for the
651
timezone specified by offset, or 'local' for the process's current
653
:param show_offset: Whether to append the timezone.
654
:param date_fmt: strftime format.
656
if timezone == 'utc':
659
elif timezone == 'original':
662
tt = time.gmtime(t + offset)
663
elif timezone == 'local':
664
tt = time.localtime(t)
665
offset = local_time_offset(t)
667
raise errors.UnsupportedTimezoneFormat(timezone)
669
date_fmt = "%a %Y-%m-%d %H:%M:%S"
671
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
674
return (time.strftime(date_fmt, tt) + offset_str)
677
def compact_date(when):
678
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
681
def format_delta(delta):
682
"""Get a nice looking string for a time delta.
684
:param delta: The time difference in seconds, can be positive or negative.
685
positive indicates time in the past, negative indicates time in the
686
future. (usually time.time() - stored_time)
687
:return: String formatted to show approximate resolution
693
direction = 'in the future'
697
if seconds < 90: # print seconds up to 90 seconds
699
return '%d second %s' % (seconds, direction,)
701
return '%d seconds %s' % (seconds, direction)
703
minutes = int(seconds / 60)
704
seconds -= 60 * minutes
709
if minutes < 90: # print minutes, seconds up to 90 minutes
711
return '%d minute, %d second%s %s' % (
712
minutes, seconds, plural_seconds, direction)
714
return '%d minutes, %d second%s %s' % (
715
minutes, seconds, plural_seconds, direction)
717
hours = int(minutes / 60)
718
minutes -= 60 * hours
725
return '%d hour, %d minute%s %s' % (hours, minutes,
726
plural_minutes, direction)
727
return '%d hours, %d minute%s %s' % (hours, minutes,
728
plural_minutes, direction)
731
"""Return size of given open file."""
732
return os.fstat(f.fileno())[ST_SIZE]
735
# Define rand_bytes based on platform.
737
# Python 2.4 and later have os.urandom,
738
# but it doesn't work on some arches
740
rand_bytes = os.urandom
741
except (NotImplementedError, AttributeError):
742
# If python doesn't have os.urandom, or it doesn't work,
743
# then try to first pull random data from /dev/urandom
745
rand_bytes = file('/dev/urandom', 'rb').read
746
# Otherwise, use this hack as a last resort
747
except (IOError, OSError):
748
# not well seeded, but better than nothing
753
s += chr(random.randint(0, 255))
758
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
760
"""Return a random string of num alphanumeric characters
762
The result only contains lowercase chars because it may be used on
763
case-insensitive filesystems.
766
for raw_byte in rand_bytes(num):
767
s += ALNUM[ord(raw_byte) % 36]
771
## TODO: We could later have path objects that remember their list
772
## decomposition (might be too tricksy though.)
775
"""Turn string into list of parts."""
776
assert isinstance(p, basestring)
778
# split on either delimiter because people might use either on
780
ps = re.split(r'[\\/]', p)
785
raise errors.BzrError("sorry, %r not allowed in path" % f)
786
elif (f == '.') or (f == ''):
793
assert isinstance(p, (list, tuple))
795
if (f == '..') or (f is None) or (f == ''):
796
raise errors.BzrError("sorry, %r not allowed in path" % f)
801
"""Split s into lines, but without removing the newline characters."""
802
lines = s.split('\n')
803
result = [line + '\n' for line in lines[:-1]]
805
result.append(lines[-1])
809
def hardlinks_good():
810
return sys.platform not in ('win32', 'cygwin', 'darwin')
813
def link_or_copy(src, dest):
814
"""Hardlink a file, or copy it if it can't be hardlinked."""
815
if not hardlinks_good():
816
shutil.copyfile(src, dest)
820
except (OSError, IOError), e:
821
if e.errno != errno.EXDEV:
823
shutil.copyfile(src, dest)
826
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
827
# Forgiveness than Permission (EAFP) because:
828
# - root can damage a solaris file system by using unlink,
829
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
830
# EACCES, OSX: EPERM) when invoked on a directory.
831
def delete_any(path):
832
"""Delete a file or directory."""
833
if isdir(path): # Takes care of symlinks
840
if getattr(os, 'symlink', None) is not None:
847
if getattr(os, 'link', None) is not None:
853
def contains_whitespace(s):
854
"""True if there are any whitespace characters in s."""
855
# string.whitespace can include '\xa0' in certain locales, because it is
856
# considered "non-breaking-space" as part of ISO-8859-1. But it
857
# 1) Isn't a breaking whitespace
858
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
860
# 3) '\xa0' isn't unicode safe since it is >128.
862
# This should *not* be a unicode set of characters in case the source
863
# string is not a Unicode string. We can auto-up-cast the characters since
864
# they are ascii, but we don't want to auto-up-cast the string in case it
866
for ch in ' \t\n\r\v\f':
873
def contains_linebreaks(s):
874
"""True if there is any vertical whitespace in s."""
882
def relpath(base, path):
883
"""Return path relative to base, or raise exception.
885
The path may be either an absolute path or a path relative to the
886
current working directory.
888
os.path.commonprefix (python2.4) has a bad bug that it works just
889
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
893
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
894
' exceed the platform minimum length (which is %d)' %
901
while len(head) >= len(base):
904
head, tail = os.path.split(head)
908
raise errors.PathNotChild(rp, base)
916
def safe_unicode(unicode_or_utf8_string):
917
"""Coerce unicode_or_utf8_string into unicode.
919
If it is unicode, it is returned.
920
Otherwise it is decoded from utf-8. If a decoding error
921
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
922
as a BzrBadParameter exception.
924
if isinstance(unicode_or_utf8_string, unicode):
925
return unicode_or_utf8_string
927
return unicode_or_utf8_string.decode('utf8')
928
except UnicodeDecodeError:
929
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
932
def safe_utf8(unicode_or_utf8_string):
933
"""Coerce unicode_or_utf8_string to a utf8 string.
935
If it is a str, it is returned.
936
If it is Unicode, it is encoded into a utf-8 string.
938
if isinstance(unicode_or_utf8_string, str):
939
# TODO: jam 20070209 This is overkill, and probably has an impact on
940
# performance if we are dealing with lots of apis that want a
943
# Make sure it is a valid utf-8 string
944
unicode_or_utf8_string.decode('utf-8')
945
except UnicodeDecodeError:
946
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
947
return unicode_or_utf8_string
948
return unicode_or_utf8_string.encode('utf-8')
951
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
952
' Revision id generators should be creating utf8'
956
def safe_revision_id(unicode_or_utf8_string, warn=True):
957
"""Revision ids should now be utf8, but at one point they were unicode.
959
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
961
:param warn: Functions that are sanitizing user data can set warn=False
962
:return: None or a utf8 revision id.
964
if (unicode_or_utf8_string is None
965
or unicode_or_utf8_string.__class__ == str):
966
return unicode_or_utf8_string
968
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
970
return cache_utf8.encode(unicode_or_utf8_string)
973
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
974
' generators should be creating utf8 file ids.')
977
def safe_file_id(unicode_or_utf8_string, warn=True):
978
"""File ids should now be utf8, but at one point they were unicode.
980
This is the same as safe_utf8, except it uses the cached encode functions
981
to save a little bit of performance.
983
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
985
:param warn: Functions that are sanitizing user data can set warn=False
986
:return: None or a utf8 file id.
988
if (unicode_or_utf8_string is None
989
or unicode_or_utf8_string.__class__ == str):
990
return unicode_or_utf8_string
992
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
994
return cache_utf8.encode(unicode_or_utf8_string)
997
_platform_normalizes_filenames = False
998
if sys.platform == 'darwin':
999
_platform_normalizes_filenames = True
1002
def normalizes_filenames():
1003
"""Return True if this platform normalizes unicode filenames.
1005
Mac OSX does, Windows/Linux do not.
1007
return _platform_normalizes_filenames
1010
def _accessible_normalized_filename(path):
1011
"""Get the unicode normalized path, and if you can access the file.
1013
On platforms where the system normalizes filenames (Mac OSX),
1014
you can access a file by any path which will normalize correctly.
1015
On platforms where the system does not normalize filenames
1016
(Windows, Linux), you have to access a file by its exact path.
1018
Internally, bzr only supports NFC normalization, since that is
1019
the standard for XML documents.
1021
So return the normalized path, and a flag indicating if the file
1022
can be accessed by that path.
1025
return unicodedata.normalize('NFC', unicode(path)), True
1028
def _inaccessible_normalized_filename(path):
1029
__doc__ = _accessible_normalized_filename.__doc__
1031
normalized = unicodedata.normalize('NFC', unicode(path))
1032
return normalized, normalized == path
1035
if _platform_normalizes_filenames:
1036
normalized_filename = _accessible_normalized_filename
1038
normalized_filename = _inaccessible_normalized_filename
1041
def terminal_width():
1042
"""Return estimated terminal width."""
1043
if sys.platform == 'win32':
1044
return win32utils.get_console_size()[0]
1047
import struct, fcntl, termios
1048
s = struct.pack('HHHH', 0, 0, 0, 0)
1049
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1050
width = struct.unpack('HHHH', x)[1]
1055
width = int(os.environ['COLUMNS'])
1064
def supports_executable():
1065
return sys.platform != "win32"
1068
def supports_posix_readonly():
1069
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1071
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1072
directory controls creation/deletion, etc.
1074
And under win32, readonly means that the directory itself cannot be
1075
deleted. The contents of a readonly directory can be changed, unlike POSIX
1076
where files in readonly directories cannot be added, deleted or renamed.
1078
return sys.platform != "win32"
1081
def set_or_unset_env(env_variable, value):
1082
"""Modify the environment, setting or removing the env_variable.
1084
:param env_variable: The environment variable in question
1085
:param value: The value to set the environment to. If None, then
1086
the variable will be removed.
1087
:return: The original value of the environment variable.
1089
orig_val = os.environ.get(env_variable)
1091
if orig_val is not None:
1092
del os.environ[env_variable]
1094
if isinstance(value, unicode):
1095
value = value.encode(bzrlib.user_encoding)
1096
os.environ[env_variable] = value
1100
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1103
def check_legal_path(path):
1104
"""Check whether the supplied path is legal.
1105
This is only required on Windows, so we don't test on other platforms
1108
if sys.platform != "win32":
1110
if _validWin32PathRE.match(path) is None:
1111
raise errors.IllegalPath(path)
1114
def walkdirs(top, prefix=""):
1115
"""Yield data about all the directories in a tree.
1117
This yields all the data about the contents of a directory at a time.
1118
After each directory has been yielded, if the caller has mutated the list
1119
to exclude some directories, they are then not descended into.
1121
The data yielded is of the form:
1122
((directory-relpath, directory-path-from-top),
1123
[(relpath, basename, kind, lstat, path-from-top), ...]),
1124
- directory-relpath is the relative path of the directory being returned
1125
with respect to top. prefix is prepended to this.
1126
- directory-path-from-root is the path including top for this directory.
1127
It is suitable for use with os functions.
1128
- relpath is the relative path within the subtree being walked.
1129
- basename is the basename of the path
1130
- kind is the kind of the file now. If unknown then the file is not
1131
present within the tree - but it may be recorded as versioned. See
1133
- lstat is the stat data *if* the file was statted.
1134
- planned, not implemented:
1135
path_from_tree_root is the path from the root of the tree.
1137
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1138
allows one to walk a subtree but get paths that are relative to a tree
1140
:return: an iterator over the dirs.
1142
#TODO there is a bit of a smell where the results of the directory-
1143
# summary in this, and the path from the root, may not agree
1144
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1145
# potentially confusing output. We should make this more robust - but
1146
# not at a speed cost. RBC 20060731
1148
_directory = _directory_kind
1149
_listdir = os.listdir
1150
_kind_from_mode = _formats.get
1151
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1153
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1154
relroot, _, _, _, top = pending.pop()
1156
relprefix = relroot + u'/'
1159
top_slash = top + u'/'
1162
append = dirblock.append
1163
for name in sorted(_listdir(top)):
1164
abspath = top_slash + name
1165
statvalue = _lstat(abspath)
1166
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1167
append((relprefix + name, name, kind, statvalue, abspath))
1168
yield (relroot, top), dirblock
1170
# push the user specified dirs from dirblock
1171
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1174
def _walkdirs_utf8(top, prefix=""):
1175
"""Yield data about all the directories in a tree.
1177
This yields the same information as walkdirs() only each entry is yielded
1178
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1179
are returned as exact byte-strings.
1181
:return: yields a tuple of (dir_info, [file_info])
1182
dir_info is (utf8_relpath, path-from-top)
1183
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1184
if top is an absolute path, path-from-top is also an absolute path.
1185
path-from-top might be unicode or utf8, but it is the correct path to
1186
pass to os functions to affect the file in question. (such as os.lstat)
1188
fs_encoding = _fs_enc.upper()
1189
if (sys.platform == 'win32' or
1190
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1191
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1193
return _walkdirs_fs_utf8(top, prefix=prefix)
1196
def _walkdirs_fs_utf8(top, prefix=""):
1197
"""See _walkdirs_utf8.
1199
This sub-function is called when we know the filesystem is already in utf8
1200
encoding. So we don't need to transcode filenames.
1203
_directory = _directory_kind
1204
_listdir = os.listdir
1205
_kind_from_mode = _formats.get
1207
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1208
# But we don't actually uses 1-3 in pending, so set them to None
1209
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1211
relroot, _, _, _, top = pending.pop()
1213
relprefix = relroot + '/'
1216
top_slash = top + '/'
1219
append = dirblock.append
1220
for name in sorted(_listdir(top)):
1221
abspath = top_slash + name
1222
statvalue = _lstat(abspath)
1223
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1224
append((relprefix + name, name, kind, statvalue, abspath))
1225
yield (relroot, top), dirblock
1227
# push the user specified dirs from dirblock
1228
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1231
def _walkdirs_unicode_to_utf8(top, prefix=""):
1232
"""See _walkdirs_utf8
1234
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1236
This is currently the fallback code path when the filesystem encoding is
1237
not UTF-8. It may be better to implement an alternative so that we can
1238
safely handle paths that are not properly decodable in the current
1241
_utf8_encode = codecs.getencoder('utf8')
1243
_directory = _directory_kind
1244
_listdir = os.listdir
1245
_kind_from_mode = _formats.get
1247
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1249
relroot, _, _, _, top = pending.pop()
1251
relprefix = relroot + '/'
1254
top_slash = top + u'/'
1257
append = dirblock.append
1258
for name in sorted(_listdir(top)):
1259
name_utf8 = _utf8_encode(name)[0]
1260
abspath = top_slash + name
1261
statvalue = _lstat(abspath)
1262
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1263
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1264
yield (relroot, top), dirblock
1266
# push the user specified dirs from dirblock
1267
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1270
def copy_tree(from_path, to_path, handlers={}):
1271
"""Copy all of the entries in from_path into to_path.
1273
:param from_path: The base directory to copy.
1274
:param to_path: The target directory. If it does not exist, it will
1276
:param handlers: A dictionary of functions, which takes a source and
1277
destinations for files, directories, etc.
1278
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1279
'file', 'directory', and 'symlink' should always exist.
1280
If they are missing, they will be replaced with 'os.mkdir()',
1281
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1283
# Now, just copy the existing cached tree to the new location
1284
# We use a cheap trick here.
1285
# Absolute paths are prefixed with the first parameter
1286
# relative paths are prefixed with the second.
1287
# So we can get both the source and target returned
1288
# without any extra work.
1290
def copy_dir(source, dest):
1293
def copy_link(source, dest):
1294
"""Copy the contents of a symlink"""
1295
link_to = os.readlink(source)
1296
os.symlink(link_to, dest)
1298
real_handlers = {'file':shutil.copy2,
1299
'symlink':copy_link,
1300
'directory':copy_dir,
1302
real_handlers.update(handlers)
1304
if not os.path.exists(to_path):
1305
real_handlers['directory'](from_path, to_path)
1307
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1308
for relpath, name, kind, st, abspath in entries:
1309
real_handlers[kind](abspath, relpath)
1312
def path_prefix_key(path):
1313
"""Generate a prefix-order path key for path.
1315
This can be used to sort paths in the same way that walkdirs does.
1317
return (dirname(path) , path)
1320
def compare_paths_prefix_order(path_a, path_b):
1321
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1322
key_a = path_prefix_key(path_a)
1323
key_b = path_prefix_key(path_b)
1324
return cmp(key_a, key_b)
1327
_cached_user_encoding = None
1330
def get_user_encoding(use_cache=True):
1331
"""Find out what the preferred user encoding is.
1333
This is generally the encoding that is used for command line parameters
1334
and file contents. This may be different from the terminal encoding
1335
or the filesystem encoding.
1337
:param use_cache: Enable cache for detected encoding.
1338
(This parameter is turned on by default,
1339
and required only for selftesting)
1341
:return: A string defining the preferred user encoding
1343
global _cached_user_encoding
1344
if _cached_user_encoding is not None and use_cache:
1345
return _cached_user_encoding
1347
if sys.platform == 'darwin':
1348
# work around egregious python 2.4 bug
1349
sys.platform = 'posix'
1353
sys.platform = 'darwin'
1358
user_encoding = locale.getpreferredencoding()
1359
except locale.Error, e:
1360
sys.stderr.write('bzr: warning: %s\n'
1361
' Could not determine what text encoding to use.\n'
1362
' This error usually means your Python interpreter\n'
1363
' doesn\'t support the locale set by $LANG (%s)\n'
1364
" Continuing with ascii encoding.\n"
1365
% (e, os.environ.get('LANG')))
1366
user_encoding = 'ascii'
1368
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1369
# treat that as ASCII, and not support printing unicode characters to the
1371
if user_encoding in (None, 'cp0'):
1372
user_encoding = 'ascii'
1376
codecs.lookup(user_encoding)
1378
sys.stderr.write('bzr: warning:'
1379
' unknown encoding %s.'
1380
' Continuing with ascii encoding.\n'
1383
user_encoding = 'ascii'
1386
_cached_user_encoding = user_encoding
1388
return user_encoding
1391
def recv_all(socket, bytes):
1392
"""Receive an exact number of bytes.
1394
Regular Socket.recv() may return less than the requested number of bytes,
1395
dependning on what's in the OS buffer. MSG_WAITALL is not available
1396
on all platforms, but this should work everywhere. This will return
1397
less than the requested amount if the remote end closes.
1399
This isn't optimized and is intended mostly for use in testing.
1402
while len(b) < bytes:
1403
new = socket.recv(bytes - len(b))
1410
def send_all(socket, bytes):
1411
"""Send all bytes on a socket.
1413
Regular socket.sendall() can give socket error 10053 on Windows. This
1414
implementation sends no more than 64k at a time, which avoids this problem.
1417
for pos in xrange(0, len(bytes), chunk_size):
1418
socket.sendall(bytes[pos:pos+chunk_size])
1421
def dereference_path(path):
1422
"""Determine the real path to a file.
1424
All parent elements are dereferenced. But the file itself is not
1426
:param path: The original path. May be absolute or relative.
1427
:return: the real path *to* the file
1429
parent, base = os.path.split(path)
1430
# The pathjoin for '.' is a workaround for Python bug #1213894.
1431
# (initial path components aren't dereferenced)
1432
return pathjoin(realpath(pathjoin('.', parent)), base)
1435
def supports_mapi():
1436
"""Return True if we can use MAPI to launch a mail client."""
1437
return sys.platform == "win32"
1440
def resource_string(package, resource_name):
1441
"""Load a resource from a package and return it as a string.
1443
Note: Only packages that start with bzrlib are currently supported.
1445
This is designed to be a lightweight implementation of resource
1446
loading in a way which is API compatible with the same API from
1448
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1449
If and when pkg_resources becomes a standard library, this routine
1452
# Check package name is within bzrlib
1453
if package == "bzrlib":
1454
resource_relpath = resource_name
1455
elif package.startswith("bzrlib."):
1456
package = package[len("bzrlib."):].replace('.', os.sep)
1457
resource_relpath = pathjoin(package, resource_name)
1459
raise errors.BzrError('resource package %s not in bzrlib' % package)
1461
# Map the resource to a file and read its contents
1462
base = dirname(bzrlib.__file__)
1463
if getattr(sys, 'frozen', None): # bzr.exe
1464
base = abspath(pathjoin(base, '..', '..'))
1465
filename = pathjoin(base, resource_relpath)
1466
return open(filename, 'rU').read()
added
removed
bzrlib/osutils.py
