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
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
262
# Default is to just use the python builtins, but these can be rebound on
263
# particular platforms.
264
abspath = _posix_abspath
265
realpath = _posix_realpath
266
pathjoin = os.path.join
267
normpath = os.path.normpath
269
mkdtemp = tempfile.mkdtemp
271
dirname = os.path.dirname
272
basename = os.path.basename
273
rmtree = shutil.rmtree
275
MIN_ABS_PATHLENGTH = 1
278
if sys.platform == 'win32':
279
abspath = _win32_abspath
280
realpath = _win32_realpath
281
pathjoin = _win32_pathjoin
282
normpath = _win32_normpath
283
getcwd = _win32_getcwd
284
mkdtemp = _win32_mkdtemp
285
rename = _win32_rename
287
MIN_ABS_PATHLENGTH = 3
289
def _win32_delete_readonly(function, path, excinfo):
290
"""Error handler for shutil.rmtree function [for win32]
291
Helps to remove files and dirs marked as read-only.
293
type_, value = excinfo[:2]
294
if function in (os.remove, os.rmdir) \
295
and type_ == OSError \
296
and value.errno == errno.EACCES:
297
bzrlib.osutils.make_writable(path)
302
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
303
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
304
return shutil.rmtree(path, ignore_errors, onerror)
307
def get_terminal_encoding():
308
"""Find the best encoding for printing to the screen.
310
This attempts to check both sys.stdout and sys.stdin to see
311
what encoding they are in, and if that fails it falls back to
312
bzrlib.user_encoding.
313
The problem is that on Windows, locale.getpreferredencoding()
314
is not the same encoding as that used by the console:
315
http://mail.python.org/pipermail/python-list/2003-May/162357.html
317
On my standard US Windows XP, the preferred encoding is
318
cp1252, but the console is cp437
320
output_encoding = getattr(sys.stdout, 'encoding', None)
321
if not output_encoding:
322
input_encoding = getattr(sys.stdin, 'encoding', None)
323
if not input_encoding:
324
output_encoding = bzrlib.user_encoding
325
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
327
output_encoding = input_encoding
328
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
330
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
331
return output_encoding
334
def normalizepath(f):
335
if hasattr(os.path, 'realpath'):
339
[p,e] = os.path.split(f)
340
if e == "" or e == "." or e == "..":
343
return pathjoin(F(p), e)
347
"""Copy a file to a backup.
349
Backups are named in GNU-style, with a ~ suffix.
351
If the file is already a backup, it's not copied.
357
if has_symlinks() and os.path.islink(fn):
358
target = os.readlink(fn)
359
os.symlink(target, bfn)
367
outf = file(bfn, 'wb')
375
"""True if f is an accessible directory."""
377
return S_ISDIR(os.lstat(f)[ST_MODE])
383
"""True if f is a regular file."""
385
return S_ISREG(os.lstat(f)[ST_MODE])
390
"""True if f is a symlink."""
392
return S_ISLNK(os.lstat(f)[ST_MODE])
396
def is_inside(dir, fname):
397
"""True if fname is inside dir.
399
The parameters should typically be passed to osutils.normpath first, so
400
that . and .. and repeated slashes are eliminated, and the separators
401
are canonical for the platform.
403
The empty string as a dir name is taken as top-of-tree and matches
406
>>> is_inside('src', pathjoin('src', 'foo.c'))
408
>>> is_inside('src', 'srccontrol')
410
>>> is_inside('src', pathjoin('src', 'a', 'a', 'a', 'foo.c'))
412
>>> is_inside('foo.c', 'foo.c')
414
>>> is_inside('foo.c', '')
416
>>> is_inside('', 'foo.c')
419
# XXX: Most callers of this can actually do something smarter by
420
# looking at the inventory
430
return fname.startswith(dir)
433
def is_inside_any(dir_list, fname):
434
"""True if fname is inside any of given dirs."""
435
for dirname in dir_list:
436
if is_inside(dirname, fname):
442
def is_inside_or_parent_of_any(dir_list, fname):
443
"""True if fname is a child or a parent of any of the given files."""
444
for dirname in dir_list:
445
if is_inside(dirname, fname) or is_inside(fname, dirname):
451
def pumpfile(fromfile, tofile):
452
"""Copy contents of one file to another."""
455
b = fromfile.read(BUFSIZE)
461
def file_iterator(input_file, readsize=32768):
463
b = input_file.read(readsize)
470
if hasattr(f, 'tell'):
483
def sha_strings(strings):
484
"""Return the sha-1 of concatenation of strings"""
486
map(s.update, strings)
496
def fingerprint_file(f):
501
return {'size': size,
502
'sha1': s.hexdigest()}
505
def compare_files(a, b):
506
"""Returns true if equal in contents"""
517
def local_time_offset(t=None):
518
"""Return offset of local zone from GMT, either at present or at time t."""
519
# python2.3 localtime() can't take None
523
if time.localtime(t).tm_isdst and time.daylight:
526
return -time.timezone
529
def format_date(t, offset=0, timezone='original', date_fmt=None,
531
## TODO: Perhaps a global option to use either universal or local time?
532
## Or perhaps just let people set $TZ?
533
assert isinstance(t, float)
535
if timezone == 'utc':
538
elif timezone == 'original':
541
tt = time.gmtime(t + offset)
542
elif timezone == 'local':
543
tt = time.localtime(t)
544
offset = local_time_offset(t)
546
raise BzrError("unsupported timezone format %r" % timezone,
547
['options are "utc", "original", "local"'])
549
date_fmt = "%a %Y-%m-%d %H:%M:%S"
551
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
554
return (time.strftime(date_fmt, tt) + offset_str)
557
def compact_date(when):
558
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
563
"""Return size of given open file."""
564
return os.fstat(f.fileno())[ST_SIZE]
567
# Define rand_bytes based on platform.
569
# Python 2.4 and later have os.urandom,
570
# but it doesn't work on some arches
572
rand_bytes = os.urandom
573
except (NotImplementedError, AttributeError):
574
# If python doesn't have os.urandom, or it doesn't work,
575
# then try to first pull random data from /dev/urandom
576
if os.path.exists("/dev/urandom"):
577
rand_bytes = file('/dev/urandom', 'rb').read
578
# Otherwise, use this hack as a last resort
580
# not well seeded, but better than nothing
585
s += chr(random.randint(0, 255))
590
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
592
"""Return a random string of num alphanumeric characters
594
The result only contains lowercase chars because it may be used on
595
case-insensitive filesystems.
598
for raw_byte in rand_bytes(num):
599
s += ALNUM[ord(raw_byte) % 36]
603
## TODO: We could later have path objects that remember their list
604
## decomposition (might be too tricksy though.)
607
"""Turn string into list of parts.
613
>>> splitpath('a/./b')
615
>>> splitpath('a/.b')
617
>>> splitpath('a/../b')
618
Traceback (most recent call last):
620
BzrError: sorry, '..' not allowed in path
622
assert isinstance(p, types.StringTypes)
624
# split on either delimiter because people might use either on
626
ps = re.split(r'[\\/]', p)
631
raise BzrError("sorry, %r not allowed in path" % f)
632
elif (f == '.') or (f == ''):
639
assert isinstance(p, list)
641
if (f == '..') or (f == None) or (f == ''):
642
raise BzrError("sorry, %r not allowed in path" % f)
646
@deprecated_function(zero_nine)
647
def appendpath(p1, p2):
651
return pathjoin(p1, p2)
655
"""Split s into lines, but without removing the newline characters."""
656
lines = s.split('\n')
657
result = [line + '\n' for line in lines[:-1]]
659
result.append(lines[-1])
663
def hardlinks_good():
664
return sys.platform not in ('win32', 'cygwin', 'darwin')
667
def link_or_copy(src, dest):
668
"""Hardlink a file, or copy it if it can't be hardlinked."""
669
if not hardlinks_good():
674
except (OSError, IOError), e:
675
if e.errno != errno.EXDEV:
679
def delete_any(full_path):
680
"""Delete a file or directory."""
684
# We may be renaming a dangling inventory id
685
if e.errno not in (errno.EISDIR, errno.EACCES, errno.EPERM):
691
if hasattr(os, 'symlink'):
697
def contains_whitespace(s):
698
"""True if there are any whitespace characters in s."""
699
for ch in string.whitespace:
706
def contains_linebreaks(s):
707
"""True if there is any vertical whitespace in s."""
715
def relpath(base, path):
716
"""Return path relative to base, or raise exception.
718
The path may be either an absolute path or a path relative to the
719
current working directory.
721
os.path.commonprefix (python2.4) has a bad bug that it works just
722
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
726
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
727
' exceed the platform minimum length (which is %d)' %
734
while len(head) >= len(base):
737
head, tail = os.path.split(head)
741
raise PathNotChild(rp, base)
749
def safe_unicode(unicode_or_utf8_string):
750
"""Coerce unicode_or_utf8_string into unicode.
752
If it is unicode, it is returned.
753
Otherwise it is decoded from utf-8. If a decoding error
754
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
755
as a BzrBadParameter exception.
757
if isinstance(unicode_or_utf8_string, unicode):
758
return unicode_or_utf8_string
760
return unicode_or_utf8_string.decode('utf8')
761
except UnicodeDecodeError:
762
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
765
_platform_normalizes_filenames = False
766
if sys.platform == 'darwin':
767
_platform_normalizes_filenames = True
770
def normalizes_filenames():
771
"""Return True if this platform normalizes unicode filenames.
773
Mac OSX does, Windows/Linux do not.
775
return _platform_normalizes_filenames
778
if _platform_normalizes_filenames:
779
def unicode_filename(path):
780
"""Make sure 'path' is a properly normalized filename.
782
On platforms where the system normalizes filenames (Mac OSX),
783
you can access a file by any path which will normalize
785
Internally, bzr only supports NFC/NFKC normalization, since
786
that is the standard for XML documents.
787
So we return an normalized path, and indicate this has been
790
:return: (path, is_normalized) Return a path which can
791
access the file, and whether or not this path is
794
return unicodedata.normalize('NFKC', path), True
796
def unicode_filename(path):
797
"""Make sure 'path' is a properly normalized filename.
799
On platforms where the system does not normalize filenames
800
(Windows, Linux), you have to access a file by its exact path.
801
Internally, bzr only supports NFC/NFKC normalization, since
802
that is the standard for XML documents.
803
So we return the original path, and indicate if this is
806
:return: (path, is_normalized) Return a path which can
807
access the file, and whether or not this path is
810
return path, unicodedata.normalize('NFKC', path) == path
813
def terminal_width():
814
"""Return estimated terminal width."""
815
if sys.platform == 'win32':
816
import bzrlib.win32console
817
return bzrlib.win32console.get_console_size()[0]
820
import struct, fcntl, termios
821
s = struct.pack('HHHH', 0, 0, 0, 0)
822
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
823
width = struct.unpack('HHHH', x)[1]
828
width = int(os.environ['COLUMNS'])
836
def supports_executable():
837
return sys.platform != "win32"
840
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
843
def check_legal_path(path):
844
"""Check whether the supplied path is legal.
845
This is only required on Windows, so we don't test on other platforms
848
if sys.platform != "win32":
850
if _validWin32PathRE.match(path) is None:
851
raise IllegalPath(path)
854
def walkdirs(top, prefix=""):
855
"""Yield data about all the directories in a tree.
857
This yields all the data about the contents of a directory at a time.
858
After each directory has been yielded, if the caller has mutated the list
859
to exclude some directories, they are then not descended into.
861
The data yielded is of the form:
862
[(relpath, basename, kind, lstat, path_from_top), ...]
864
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
865
allows one to walk a subtree but get paths that are relative to a tree
867
:return: an iterator over the dirs.
871
_directory = _directory_kind
873
pending = [(prefix, "", _directory, None, top)]
876
currentdir = pending.pop()
877
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
880
relroot = currentdir[0] + '/'
883
for name in sorted(_listdir(top)):
884
abspath = top + '/' + name
885
statvalue = lstat(abspath)
886
dirblock.append ((relroot + name, name, file_kind_from_stat_mode(statvalue.st_mode), statvalue, abspath))
888
# push the user specified dirs from dirblock
889
for dir in reversed(dirblock):
890
if dir[2] == _directory:
894
def path_prefix_key(path):
895
"""Generate a prefix-order path key for path.
897
This can be used to sort paths in the same way that walkdirs does.
899
return (dirname(path) , path)
902
def compare_paths_prefix_order(path_a, path_b):
903
"""Compare path_a and path_b to generate the same order walkdirs uses."""
904
key_a = path_prefix_key(path_a)
905
key_b = path_prefix_key(path_b)
906
return cmp(key_a, key_b)
added
removed
bzrlib/osutils.py
