1
# Bazaar-NG -- distributed version control
3
# Copyright (C) 2005 by Canonical Ltd
5
# This program is free software; you can redistribute it and/or modify
6
# it under the terms of the GNU General Public License as published by
7
# the Free Software Foundation; either version 2 of the License, or
8
# (at your option) any later version.
10
# This program is distributed in the hope that it will be useful,
11
# but WITHOUT ANY WARRANTY; without even the implied warranty of
12
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13
# GNU General Public License for more details.
15
# You should have received a copy of the GNU General Public License
16
# along with this program; if not, write to the Free Software
17
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
19
from cStringIO import StringIO
21
from ntpath import (abspath as _nt_abspath,
23
normpath as _nt_normpath,
24
realpath as _nt_realpath,
25
splitdrive as _nt_splitdrive,
28
from os import listdir
33
from shutil import copyfile
35
from stat import (S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE,
36
S_ISCHR, S_ISBLK, S_ISFIFO, S_ISSOCK)
45
from bzrlib.errors import (BzrError,
46
BzrBadParameterNotUnicode,
51
from bzrlib.symbol_versioning import (deprecated_function,
53
from bzrlib.trace import mutter
56
def make_readonly(filename):
57
"""Make a filename read-only."""
58
mod = os.stat(filename).st_mode
60
os.chmod(filename, mod)
63
def make_writable(filename):
64
mod = os.stat(filename).st_mode
66
os.chmod(filename, mod)
73
"""Return a quoted filename filename
75
This previously used backslash quoting, but that works poorly on
77
# TODO: I'm not really sure this is the best format either.x
80
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
82
if _QUOTE_RE.search(f):
88
_directory_kind = 'directory'
91
stat.S_IFDIR:_directory_kind,
92
stat.S_IFCHR:'chardev',
96
stat.S_IFLNK:'symlink',
97
stat.S_IFSOCK:'socket',
101
def file_kind_from_stat_mode(stat_mode, _formats=_formats, _unknown='unknown'):
102
"""Generate a file kind from a stat mode. This is used in walkdirs.
104
Its performance is critical: Do not mutate without careful benchmarking.
107
return _formats[stat_mode & 0170000]
112
def file_kind(f, _lstat=os.lstat, _mapper=file_kind_from_stat_mode):
114
return _mapper(_lstat(f).st_mode)
116
if getattr(e, 'errno', None) == errno.ENOENT:
117
raise bzrlib.errors.NoSuchFile(f)
121
def kind_marker(kind):
124
elif kind == _directory_kind:
126
elif kind == 'symlink':
129
raise BzrError('invalid file kind %r' % kind)
131
lexists = getattr(os.path, 'lexists', None)
135
if hasattr(os, 'lstat'):
141
if e.errno == errno.ENOENT:
144
raise BzrError("lstat/stat of (%r): %r" % (f, e))
147
def fancy_rename(old, new, rename_func, unlink_func):
148
"""A fancy rename, when you don't have atomic rename.
150
:param old: The old path, to rename from
151
:param new: The new path, to rename to
152
:param rename_func: The potentially non-atomic rename function
153
:param unlink_func: A way to delete the target file if the full rename succeeds
156
# sftp rename doesn't allow overwriting, so play tricks:
158
base = os.path.basename(new)
159
dirname = os.path.dirname(new)
160
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
161
tmp_name = pathjoin(dirname, tmp_name)
163
# Rename the file out of the way, but keep track if it didn't exist
164
# We don't want to grab just any exception
165
# something like EACCES should prevent us from continuing
166
# The downside is that the rename_func has to throw an exception
167
# with an errno = ENOENT, or NoSuchFile
170
rename_func(new, tmp_name)
171
except (NoSuchFile,), e:
174
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
175
# function raises an IOError with errno == None when a rename fails.
176
# This then gets caught here.
177
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
180
if (not hasattr(e, 'errno')
181
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
188
# This may throw an exception, in which case success will
190
rename_func(old, new)
194
# If the file used to exist, rename it back into place
195
# otherwise just delete it from the tmp location
197
unlink_func(tmp_name)
199
rename_func(tmp_name, new)
202
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
203
# choke on a Unicode string containing a relative path if
204
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
206
_fs_enc = sys.getfilesystemencoding()
207
def _posix_abspath(path):
208
# jam 20060426 rather than encoding to fsencoding
209
# copy posixpath.abspath, but use os.getcwdu instead
210
if not posixpath.isabs(path):
211
path = posixpath.join(getcwd(), path)
212
return posixpath.normpath(path)
215
def _posix_realpath(path):
216
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
219
def _win32_fixdrive(path):
220
"""Force drive letters to be consistent.
222
win32 is inconsistent whether it returns lower or upper case
223
and even if it was consistent the user might type the other
224
so we force it to uppercase
225
running python.exe under cmd.exe return capital C:\\
226
running win32 python inside a cygwin shell returns lowercase c:\\
228
drive, path = _nt_splitdrive(path)
229
return drive.upper() + path
232
def _win32_abspath(path):
233
# Real _nt_abspath doesn't have a problem with a unicode cwd
234
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
237
def _win32_realpath(path):
238
# Real _nt_realpath doesn't have a problem with a unicode cwd
239
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
242
def _win32_pathjoin(*args):
243
return _nt_join(*args).replace('\\', '/')
246
def _win32_normpath(path):
247
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
251
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
254
def _win32_mkdtemp(*args, **kwargs):
255
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
258
def _win32_rename(old, new):
259
"""We expect to be able to atomically replace 'new' with old.
261
On win32, if new exists, it must be moved out of the way first, and then
262
deleted. However, if this function is called and 'old' doesn't exist
263
we should get a ENOENT error.
265
# Force ENOENT in the case that 'old' doesn't exist, before we do
266
# all the work of trying to move the file around
268
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
271
# Default is to just use the python builtins, but these can be rebound on
272
# particular platforms.
273
abspath = _posix_abspath
274
realpath = _posix_realpath
275
pathjoin = os.path.join
276
normpath = os.path.normpath
278
mkdtemp = tempfile.mkdtemp
280
dirname = os.path.dirname
281
basename = os.path.basename
282
rmtree = shutil.rmtree
284
MIN_ABS_PATHLENGTH = 1
287
if sys.platform == 'win32':
288
abspath = _win32_abspath
289
realpath = _win32_realpath
290
pathjoin = _win32_pathjoin
291
normpath = _win32_normpath
292
getcwd = _win32_getcwd
293
mkdtemp = _win32_mkdtemp
294
rename = _win32_rename
296
MIN_ABS_PATHLENGTH = 3
298
def _win32_delete_readonly(function, path, excinfo):
299
"""Error handler for shutil.rmtree function [for win32]
300
Helps to remove files and dirs marked as read-only.
302
type_, value = excinfo[:2]
303
if function in (os.remove, os.rmdir) \
304
and type_ == OSError \
305
and value.errno == errno.EACCES:
306
bzrlib.osutils.make_writable(path)
311
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
312
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
313
return shutil.rmtree(path, ignore_errors, onerror)
316
def get_terminal_encoding():
317
"""Find the best encoding for printing to the screen.
319
This attempts to check both sys.stdout and sys.stdin to see
320
what encoding they are in, and if that fails it falls back to
321
bzrlib.user_encoding.
322
The problem is that on Windows, locale.getpreferredencoding()
323
is not the same encoding as that used by the console:
324
http://mail.python.org/pipermail/python-list/2003-May/162357.html
326
On my standard US Windows XP, the preferred encoding is
327
cp1252, but the console is cp437
329
output_encoding = getattr(sys.stdout, 'encoding', None)
330
if not output_encoding:
331
input_encoding = getattr(sys.stdin, 'encoding', None)
332
if not input_encoding:
333
output_encoding = bzrlib.user_encoding
334
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
336
output_encoding = input_encoding
337
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
339
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
340
return output_encoding
343
def normalizepath(f):
344
if hasattr(os.path, 'realpath'):
348
[p,e] = os.path.split(f)
349
if e == "" or e == "." or e == "..":
352
return pathjoin(F(p), e)
356
"""Copy a file to a backup.
358
Backups are named in GNU-style, with a ~ suffix.
360
If the file is already a backup, it's not copied.
366
if has_symlinks() and os.path.islink(fn):
367
target = os.readlink(fn)
368
os.symlink(target, bfn)
376
outf = file(bfn, 'wb')
384
"""True if f is an accessible directory."""
386
return S_ISDIR(os.lstat(f)[ST_MODE])
392
"""True if f is a regular file."""
394
return S_ISREG(os.lstat(f)[ST_MODE])
399
"""True if f is a symlink."""
401
return S_ISLNK(os.lstat(f)[ST_MODE])
405
def is_inside(dir, fname):
406
"""True if fname is inside dir.
408
The parameters should typically be passed to osutils.normpath first, so
409
that . and .. and repeated slashes are eliminated, and the separators
410
are canonical for the platform.
412
The empty string as a dir name is taken as top-of-tree and matches
415
>>> is_inside('src', pathjoin('src', 'foo.c'))
417
>>> is_inside('src', 'srccontrol')
419
>>> is_inside('src', pathjoin('src', 'a', 'a', 'a', 'foo.c'))
421
>>> is_inside('foo.c', 'foo.c')
423
>>> is_inside('foo.c', '')
425
>>> is_inside('', 'foo.c')
428
# XXX: Most callers of this can actually do something smarter by
429
# looking at the inventory
439
return fname.startswith(dir)
442
def is_inside_any(dir_list, fname):
443
"""True if fname is inside any of given dirs."""
444
for dirname in dir_list:
445
if is_inside(dirname, fname):
451
def is_inside_or_parent_of_any(dir_list, fname):
452
"""True if fname is a child or a parent of any of the given files."""
453
for dirname in dir_list:
454
if is_inside(dirname, fname) or is_inside(fname, dirname):
460
def pumpfile(fromfile, tofile):
461
"""Copy contents of one file to another."""
464
b = fromfile.read(BUFSIZE)
470
def file_iterator(input_file, readsize=32768):
472
b = input_file.read(readsize)
479
if hasattr(f, 'tell'):
492
def sha_strings(strings):
493
"""Return the sha-1 of concatenation of strings"""
495
map(s.update, strings)
505
def fingerprint_file(f):
510
return {'size': size,
511
'sha1': s.hexdigest()}
514
def compare_files(a, b):
515
"""Returns true if equal in contents"""
526
def local_time_offset(t=None):
527
"""Return offset of local zone from GMT, either at present or at time t."""
528
# python2.3 localtime() can't take None
532
if time.localtime(t).tm_isdst and time.daylight:
535
return -time.timezone
538
def format_date(t, offset=0, timezone='original', date_fmt=None,
540
## TODO: Perhaps a global option to use either universal or local time?
541
## Or perhaps just let people set $TZ?
542
assert isinstance(t, float)
544
if timezone == 'utc':
547
elif timezone == 'original':
550
tt = time.gmtime(t + offset)
551
elif timezone == 'local':
552
tt = time.localtime(t)
553
offset = local_time_offset(t)
555
raise BzrError("unsupported timezone format %r" % timezone,
556
['options are "utc", "original", "local"'])
558
date_fmt = "%a %Y-%m-%d %H:%M:%S"
560
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
563
return (time.strftime(date_fmt, tt) + offset_str)
566
def compact_date(when):
567
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
572
"""Return size of given open file."""
573
return os.fstat(f.fileno())[ST_SIZE]
576
# Define rand_bytes based on platform.
578
# Python 2.4 and later have os.urandom,
579
# but it doesn't work on some arches
581
rand_bytes = os.urandom
582
except (NotImplementedError, AttributeError):
583
# If python doesn't have os.urandom, or it doesn't work,
584
# then try to first pull random data from /dev/urandom
585
if os.path.exists("/dev/urandom"):
586
rand_bytes = file('/dev/urandom', 'rb').read
587
# Otherwise, use this hack as a last resort
589
# not well seeded, but better than nothing
594
s += chr(random.randint(0, 255))
599
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
601
"""Return a random string of num alphanumeric characters
603
The result only contains lowercase chars because it may be used on
604
case-insensitive filesystems.
607
for raw_byte in rand_bytes(num):
608
s += ALNUM[ord(raw_byte) % 36]
612
## TODO: We could later have path objects that remember their list
613
## decomposition (might be too tricksy though.)
616
"""Turn string into list of parts.
622
>>> splitpath('a/./b')
624
>>> splitpath('a/.b')
626
>>> splitpath('a/../b')
627
Traceback (most recent call last):
629
BzrError: sorry, '..' not allowed in path
631
assert isinstance(p, types.StringTypes)
633
# split on either delimiter because people might use either on
635
ps = re.split(r'[\\/]', p)
640
raise BzrError("sorry, %r not allowed in path" % f)
641
elif (f == '.') or (f == ''):
648
assert isinstance(p, list)
650
if (f == '..') or (f == None) or (f == ''):
651
raise BzrError("sorry, %r not allowed in path" % f)
655
@deprecated_function(zero_nine)
656
def appendpath(p1, p2):
660
return pathjoin(p1, p2)
664
"""Split s into lines, but without removing the newline characters."""
665
lines = s.split('\n')
666
result = [line + '\n' for line in lines[:-1]]
668
result.append(lines[-1])
672
def hardlinks_good():
673
return sys.platform not in ('win32', 'cygwin', 'darwin')
676
def link_or_copy(src, dest):
677
"""Hardlink a file, or copy it if it can't be hardlinked."""
678
if not hardlinks_good():
683
except (OSError, IOError), e:
684
if e.errno != errno.EXDEV:
688
def delete_any(full_path):
689
"""Delete a file or directory."""
693
# We may be renaming a dangling inventory id
694
if e.errno not in (errno.EISDIR, errno.EACCES, errno.EPERM):
700
if hasattr(os, 'symlink'):
706
def contains_whitespace(s):
707
"""True if there are any whitespace characters in s."""
708
for ch in string.whitespace:
715
def contains_linebreaks(s):
716
"""True if there is any vertical whitespace in s."""
724
def relpath(base, path):
725
"""Return path relative to base, or raise exception.
727
The path may be either an absolute path or a path relative to the
728
current working directory.
730
os.path.commonprefix (python2.4) has a bad bug that it works just
731
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
735
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
736
' exceed the platform minimum length (which is %d)' %
743
while len(head) >= len(base):
746
head, tail = os.path.split(head)
750
raise PathNotChild(rp, base)
758
def safe_unicode(unicode_or_utf8_string):
759
"""Coerce unicode_or_utf8_string into unicode.
761
If it is unicode, it is returned.
762
Otherwise it is decoded from utf-8. If a decoding error
763
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
764
as a BzrBadParameter exception.
766
if isinstance(unicode_or_utf8_string, unicode):
767
return unicode_or_utf8_string
769
return unicode_or_utf8_string.decode('utf8')
770
except UnicodeDecodeError:
771
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
774
_platform_normalizes_filenames = False
775
if sys.platform == 'darwin':
776
_platform_normalizes_filenames = True
779
def normalizes_filenames():
780
"""Return True if this platform normalizes unicode filenames.
782
Mac OSX does, Windows/Linux do not.
784
return _platform_normalizes_filenames
787
if _platform_normalizes_filenames:
788
def unicode_filename(path):
789
"""Make sure 'path' is a properly normalized filename.
791
On platforms where the system normalizes filenames (Mac OSX),
792
you can access a file by any path which will normalize
794
Internally, bzr only supports NFC/NFKC normalization, since
795
that is the standard for XML documents.
796
So we return an normalized path, and indicate this has been
799
:return: (path, is_normalized) Return a path which can
800
access the file, and whether or not this path is
803
return unicodedata.normalize('NFKC', path), True
805
def unicode_filename(path):
806
"""Make sure 'path' is a properly normalized filename.
808
On platforms where the system does not normalize filenames
809
(Windows, Linux), you have to access a file by its exact path.
810
Internally, bzr only supports NFC/NFKC normalization, since
811
that is the standard for XML documents.
812
So we return the original path, and indicate if this is
815
:return: (path, is_normalized) Return a path which can
816
access the file, and whether or not this path is
819
return path, unicodedata.normalize('NFKC', path) == path
822
def terminal_width():
823
"""Return estimated terminal width."""
824
if sys.platform == 'win32':
825
import bzrlib.win32console
826
return bzrlib.win32console.get_console_size()[0]
829
import struct, fcntl, termios
830
s = struct.pack('HHHH', 0, 0, 0, 0)
831
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
832
width = struct.unpack('HHHH', x)[1]
837
width = int(os.environ['COLUMNS'])
845
def supports_executable():
846
return sys.platform != "win32"
849
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
852
def check_legal_path(path):
853
"""Check whether the supplied path is legal.
854
This is only required on Windows, so we don't test on other platforms
857
if sys.platform != "win32":
859
if _validWin32PathRE.match(path) is None:
860
raise IllegalPath(path)
863
def walkdirs(top, prefix=""):
864
"""Yield data about all the directories in a tree.
866
This yields all the data about the contents of a directory at a time.
867
After each directory has been yielded, if the caller has mutated the list
868
to exclude some directories, they are then not descended into.
870
The data yielded is of the form:
871
[(relpath, basename, kind, lstat, path_from_top), ...]
873
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
874
allows one to walk a subtree but get paths that are relative to a tree
876
:return: an iterator over the dirs.
880
_directory = _directory_kind
882
pending = [(prefix, "", _directory, None, top)]
885
currentdir = pending.pop()
886
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
889
relroot = currentdir[0] + '/'
892
for name in sorted(_listdir(top)):
893
abspath = top + '/' + name
894
statvalue = lstat(abspath)
895
dirblock.append ((relroot + name, name, file_kind_from_stat_mode(statvalue.st_mode), statvalue, abspath))
897
# push the user specified dirs from dirblock
898
for dir in reversed(dirblock):
899
if dir[2] == _directory:
903
def path_prefix_key(path):
904
"""Generate a prefix-order path key for path.
906
This can be used to sort paths in the same way that walkdirs does.
908
return (dirname(path) , path)
911
def compare_paths_prefix_order(path_a, path_b):
912
"""Compare path_a and path_b to generate the same order walkdirs uses."""
913
key_a = path_prefix_key(path_a)
914
key_b = path_prefix_key(path_b)
915
return cmp(key_a, key_b)
added
removed
bzrlib/osutils.py
