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,
265
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
267
if e.errno in (errno.EPERM, errno.EACCES):
268
# If we try to rename a non-existant file onto cwd, we get EPERM
269
# instead of ENOENT, this will raise ENOENT if the old path
275
# Default is to just use the python builtins, but these can be rebound on
276
# particular platforms.
277
abspath = _posix_abspath
278
realpath = _posix_realpath
279
pathjoin = os.path.join
280
normpath = os.path.normpath
282
mkdtemp = tempfile.mkdtemp
284
dirname = os.path.dirname
285
basename = os.path.basename
286
rmtree = shutil.rmtree
288
MIN_ABS_PATHLENGTH = 1
291
if sys.platform == 'win32':
292
abspath = _win32_abspath
293
realpath = _win32_realpath
294
pathjoin = _win32_pathjoin
295
normpath = _win32_normpath
296
getcwd = _win32_getcwd
297
mkdtemp = _win32_mkdtemp
298
rename = _win32_rename
300
MIN_ABS_PATHLENGTH = 3
302
def _win32_delete_readonly(function, path, excinfo):
303
"""Error handler for shutil.rmtree function [for win32]
304
Helps to remove files and dirs marked as read-only.
306
type_, value = excinfo[:2]
307
if function in (os.remove, os.rmdir) \
308
and type_ == OSError \
309
and value.errno == errno.EACCES:
310
bzrlib.osutils.make_writable(path)
315
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
316
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
317
return shutil.rmtree(path, ignore_errors, onerror)
320
def get_terminal_encoding():
321
"""Find the best encoding for printing to the screen.
323
This attempts to check both sys.stdout and sys.stdin to see
324
what encoding they are in, and if that fails it falls back to
325
bzrlib.user_encoding.
326
The problem is that on Windows, locale.getpreferredencoding()
327
is not the same encoding as that used by the console:
328
http://mail.python.org/pipermail/python-list/2003-May/162357.html
330
On my standard US Windows XP, the preferred encoding is
331
cp1252, but the console is cp437
333
output_encoding = getattr(sys.stdout, 'encoding', None)
334
if not output_encoding:
335
input_encoding = getattr(sys.stdin, 'encoding', None)
336
if not input_encoding:
337
output_encoding = bzrlib.user_encoding
338
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
340
output_encoding = input_encoding
341
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
343
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
344
return output_encoding
347
def normalizepath(f):
348
if hasattr(os.path, 'realpath'):
352
[p,e] = os.path.split(f)
353
if e == "" or e == "." or e == "..":
356
return pathjoin(F(p), e)
360
"""Copy a file to a backup.
362
Backups are named in GNU-style, with a ~ suffix.
364
If the file is already a backup, it's not copied.
370
if has_symlinks() and os.path.islink(fn):
371
target = os.readlink(fn)
372
os.symlink(target, bfn)
380
outf = file(bfn, 'wb')
388
"""True if f is an accessible directory."""
390
return S_ISDIR(os.lstat(f)[ST_MODE])
396
"""True if f is a regular file."""
398
return S_ISREG(os.lstat(f)[ST_MODE])
403
"""True if f is a symlink."""
405
return S_ISLNK(os.lstat(f)[ST_MODE])
409
def is_inside(dir, fname):
410
"""True if fname is inside dir.
412
The parameters should typically be passed to osutils.normpath first, so
413
that . and .. and repeated slashes are eliminated, and the separators
414
are canonical for the platform.
416
The empty string as a dir name is taken as top-of-tree and matches
419
>>> is_inside('src', pathjoin('src', 'foo.c'))
421
>>> is_inside('src', 'srccontrol')
423
>>> is_inside('src', pathjoin('src', 'a', 'a', 'a', 'foo.c'))
425
>>> is_inside('foo.c', 'foo.c')
427
>>> is_inside('foo.c', '')
429
>>> is_inside('', 'foo.c')
432
# XXX: Most callers of this can actually do something smarter by
433
# looking at the inventory
443
return fname.startswith(dir)
446
def is_inside_any(dir_list, fname):
447
"""True if fname is inside any of given dirs."""
448
for dirname in dir_list:
449
if is_inside(dirname, fname):
455
def is_inside_or_parent_of_any(dir_list, fname):
456
"""True if fname is a child or a parent of any of the given files."""
457
for dirname in dir_list:
458
if is_inside(dirname, fname) or is_inside(fname, dirname):
464
def pumpfile(fromfile, tofile):
465
"""Copy contents of one file to another."""
468
b = fromfile.read(BUFSIZE)
474
def file_iterator(input_file, readsize=32768):
476
b = input_file.read(readsize)
483
if hasattr(f, 'tell'):
496
def sha_strings(strings):
497
"""Return the sha-1 of concatenation of strings"""
499
map(s.update, strings)
509
def fingerprint_file(f):
514
return {'size': size,
515
'sha1': s.hexdigest()}
518
def compare_files(a, b):
519
"""Returns true if equal in contents"""
530
def local_time_offset(t=None):
531
"""Return offset of local zone from GMT, either at present or at time t."""
532
# python2.3 localtime() can't take None
536
if time.localtime(t).tm_isdst and time.daylight:
539
return -time.timezone
542
def format_date(t, offset=0, timezone='original', date_fmt=None,
544
## TODO: Perhaps a global option to use either universal or local time?
545
## Or perhaps just let people set $TZ?
546
assert isinstance(t, float)
548
if timezone == 'utc':
551
elif timezone == 'original':
554
tt = time.gmtime(t + offset)
555
elif timezone == 'local':
556
tt = time.localtime(t)
557
offset = local_time_offset(t)
559
raise BzrError("unsupported timezone format %r" % timezone,
560
['options are "utc", "original", "local"'])
562
date_fmt = "%a %Y-%m-%d %H:%M:%S"
564
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
567
return (time.strftime(date_fmt, tt) + offset_str)
570
def compact_date(when):
571
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
576
"""Return size of given open file."""
577
return os.fstat(f.fileno())[ST_SIZE]
580
# Define rand_bytes based on platform.
582
# Python 2.4 and later have os.urandom,
583
# but it doesn't work on some arches
585
rand_bytes = os.urandom
586
except (NotImplementedError, AttributeError):
587
# If python doesn't have os.urandom, or it doesn't work,
588
# then try to first pull random data from /dev/urandom
589
if os.path.exists("/dev/urandom"):
590
rand_bytes = file('/dev/urandom', 'rb').read
591
# Otherwise, use this hack as a last resort
593
# not well seeded, but better than nothing
598
s += chr(random.randint(0, 255))
603
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
605
"""Return a random string of num alphanumeric characters
607
The result only contains lowercase chars because it may be used on
608
case-insensitive filesystems.
611
for raw_byte in rand_bytes(num):
612
s += ALNUM[ord(raw_byte) % 36]
616
## TODO: We could later have path objects that remember their list
617
## decomposition (might be too tricksy though.)
620
"""Turn string into list of parts.
626
>>> splitpath('a/./b')
628
>>> splitpath('a/.b')
630
>>> splitpath('a/../b')
631
Traceback (most recent call last):
633
BzrError: sorry, '..' not allowed in path
635
assert isinstance(p, types.StringTypes)
637
# split on either delimiter because people might use either on
639
ps = re.split(r'[\\/]', p)
644
raise BzrError("sorry, %r not allowed in path" % f)
645
elif (f == '.') or (f == ''):
652
assert isinstance(p, list)
654
if (f == '..') or (f == None) or (f == ''):
655
raise BzrError("sorry, %r not allowed in path" % f)
659
@deprecated_function(zero_nine)
660
def appendpath(p1, p2):
664
return pathjoin(p1, p2)
668
"""Split s into lines, but without removing the newline characters."""
669
lines = s.split('\n')
670
result = [line + '\n' for line in lines[:-1]]
672
result.append(lines[-1])
676
def hardlinks_good():
677
return sys.platform not in ('win32', 'cygwin', 'darwin')
680
def link_or_copy(src, dest):
681
"""Hardlink a file, or copy it if it can't be hardlinked."""
682
if not hardlinks_good():
687
except (OSError, IOError), e:
688
if e.errno != errno.EXDEV:
692
def delete_any(full_path):
693
"""Delete a file or directory."""
697
# We may be renaming a dangling inventory id
698
if e.errno not in (errno.EISDIR, errno.EACCES, errno.EPERM):
704
if hasattr(os, 'symlink'):
710
def contains_whitespace(s):
711
"""True if there are any whitespace characters in s."""
712
for ch in string.whitespace:
719
def contains_linebreaks(s):
720
"""True if there is any vertical whitespace in s."""
728
def relpath(base, path):
729
"""Return path relative to base, or raise exception.
731
The path may be either an absolute path or a path relative to the
732
current working directory.
734
os.path.commonprefix (python2.4) has a bad bug that it works just
735
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
739
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
740
' exceed the platform minimum length (which is %d)' %
747
while len(head) >= len(base):
750
head, tail = os.path.split(head)
754
raise PathNotChild(rp, base)
762
def safe_unicode(unicode_or_utf8_string):
763
"""Coerce unicode_or_utf8_string into unicode.
765
If it is unicode, it is returned.
766
Otherwise it is decoded from utf-8. If a decoding error
767
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
768
as a BzrBadParameter exception.
770
if isinstance(unicode_or_utf8_string, unicode):
771
return unicode_or_utf8_string
773
return unicode_or_utf8_string.decode('utf8')
774
except UnicodeDecodeError:
775
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
778
_platform_normalizes_filenames = False
779
if sys.platform == 'darwin':
780
_platform_normalizes_filenames = True
783
def normalizes_filenames():
784
"""Return True if this platform normalizes unicode filenames.
786
Mac OSX does, Windows/Linux do not.
788
return _platform_normalizes_filenames
791
if _platform_normalizes_filenames:
792
def unicode_filename(path):
793
"""Make sure 'path' is a properly normalized filename.
795
On platforms where the system normalizes filenames (Mac OSX),
796
you can access a file by any path which will normalize
798
Internally, bzr only supports NFC/NFKC normalization, since
799
that is the standard for XML documents.
800
So we return an normalized path, and indicate this has been
803
:return: (path, is_normalized) Return a path which can
804
access the file, and whether or not this path is
807
return unicodedata.normalize('NFKC', path), True
809
def unicode_filename(path):
810
"""Make sure 'path' is a properly normalized filename.
812
On platforms where the system does not normalize filenames
813
(Windows, Linux), you have to access a file by its exact path.
814
Internally, bzr only supports NFC/NFKC normalization, since
815
that is the standard for XML documents.
816
So we return the original path, and indicate if this is
819
:return: (path, is_normalized) Return a path which can
820
access the file, and whether or not this path is
823
return path, unicodedata.normalize('NFKC', path) == path
826
def terminal_width():
827
"""Return estimated terminal width."""
828
if sys.platform == 'win32':
829
import bzrlib.win32console
830
return bzrlib.win32console.get_console_size()[0]
833
import struct, fcntl, termios
834
s = struct.pack('HHHH', 0, 0, 0, 0)
835
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
836
width = struct.unpack('HHHH', x)[1]
841
width = int(os.environ['COLUMNS'])
849
def supports_executable():
850
return sys.platform != "win32"
853
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
856
def check_legal_path(path):
857
"""Check whether the supplied path is legal.
858
This is only required on Windows, so we don't test on other platforms
861
if sys.platform != "win32":
863
if _validWin32PathRE.match(path) is None:
864
raise IllegalPath(path)
867
def walkdirs(top, prefix=""):
868
"""Yield data about all the directories in a tree.
870
This yields all the data about the contents of a directory at a time.
871
After each directory has been yielded, if the caller has mutated the list
872
to exclude some directories, they are then not descended into.
874
The data yielded is of the form:
875
[(relpath, basename, kind, lstat, path_from_top), ...]
877
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
878
allows one to walk a subtree but get paths that are relative to a tree
880
:return: an iterator over the dirs.
884
_directory = _directory_kind
886
pending = [(prefix, "", _directory, None, top)]
889
currentdir = pending.pop()
890
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
893
relroot = currentdir[0] + '/'
896
for name in sorted(_listdir(top)):
897
abspath = top + '/' + name
898
statvalue = lstat(abspath)
899
dirblock.append ((relroot + name, name, file_kind_from_stat_mode(statvalue.st_mode), statvalue, abspath))
901
# push the user specified dirs from dirblock
902
for dir in reversed(dirblock):
903
if dir[2] == _directory:
907
def path_prefix_key(path):
908
"""Generate a prefix-order path key for path.
910
This can be used to sort paths in the same way that walkdirs does.
912
return (dirname(path) , path)
915
def compare_paths_prefix_order(path_a, path_b):
916
"""Compare path_a and path_b to generate the same order walkdirs uses."""
917
key_a = path_prefix_key(path_a)
918
key_b = path_prefix_key(path_b)
919
return cmp(key_a, key_b)
added
removed
bzrlib/osutils.py
