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, errno.EBUSY):
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
276
return unicodedata.normalize('NFKC', os.getcwdu())
279
# Default is to just use the python builtins, but these can be rebound on
280
# particular platforms.
281
abspath = _posix_abspath
282
realpath = _posix_realpath
283
pathjoin = os.path.join
284
normpath = os.path.normpath
286
mkdtemp = tempfile.mkdtemp
288
dirname = os.path.dirname
289
basename = os.path.basename
290
rmtree = shutil.rmtree
292
MIN_ABS_PATHLENGTH = 1
295
if sys.platform == 'win32':
296
abspath = _win32_abspath
297
realpath = _win32_realpath
298
pathjoin = _win32_pathjoin
299
normpath = _win32_normpath
300
getcwd = _win32_getcwd
301
mkdtemp = _win32_mkdtemp
302
rename = _win32_rename
304
MIN_ABS_PATHLENGTH = 3
306
def _win32_delete_readonly(function, path, excinfo):
307
"""Error handler for shutil.rmtree function [for win32]
308
Helps to remove files and dirs marked as read-only.
310
type_, value = excinfo[:2]
311
if function in (os.remove, os.rmdir) \
312
and type_ == OSError \
313
and value.errno == errno.EACCES:
314
bzrlib.osutils.make_writable(path)
319
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
320
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
321
return shutil.rmtree(path, ignore_errors, onerror)
322
elif sys.platform == 'darwin':
326
def get_terminal_encoding():
327
"""Find the best encoding for printing to the screen.
329
This attempts to check both sys.stdout and sys.stdin to see
330
what encoding they are in, and if that fails it falls back to
331
bzrlib.user_encoding.
332
The problem is that on Windows, locale.getpreferredencoding()
333
is not the same encoding as that used by the console:
334
http://mail.python.org/pipermail/python-list/2003-May/162357.html
336
On my standard US Windows XP, the preferred encoding is
337
cp1252, but the console is cp437
339
output_encoding = getattr(sys.stdout, 'encoding', None)
340
if not output_encoding:
341
input_encoding = getattr(sys.stdin, 'encoding', None)
342
if not input_encoding:
343
output_encoding = bzrlib.user_encoding
344
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
346
output_encoding = input_encoding
347
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
349
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
350
return output_encoding
353
def normalizepath(f):
354
if hasattr(os.path, 'realpath'):
358
[p,e] = os.path.split(f)
359
if e == "" or e == "." or e == "..":
362
return pathjoin(F(p), e)
366
"""Copy a file to a backup.
368
Backups are named in GNU-style, with a ~ suffix.
370
If the file is already a backup, it's not copied.
376
if has_symlinks() and os.path.islink(fn):
377
target = os.readlink(fn)
378
os.symlink(target, bfn)
386
outf = file(bfn, 'wb')
394
"""True if f is an accessible directory."""
396
return S_ISDIR(os.lstat(f)[ST_MODE])
402
"""True if f is a regular file."""
404
return S_ISREG(os.lstat(f)[ST_MODE])
409
"""True if f is a symlink."""
411
return S_ISLNK(os.lstat(f)[ST_MODE])
415
def is_inside(dir, fname):
416
"""True if fname is inside dir.
418
The parameters should typically be passed to osutils.normpath first, so
419
that . and .. and repeated slashes are eliminated, and the separators
420
are canonical for the platform.
422
The empty string as a dir name is taken as top-of-tree and matches
425
>>> is_inside('src', pathjoin('src', 'foo.c'))
427
>>> is_inside('src', 'srccontrol')
429
>>> is_inside('src', pathjoin('src', 'a', 'a', 'a', 'foo.c'))
431
>>> is_inside('foo.c', 'foo.c')
433
>>> is_inside('foo.c', '')
435
>>> is_inside('', 'foo.c')
438
# XXX: Most callers of this can actually do something smarter by
439
# looking at the inventory
449
return fname.startswith(dir)
452
def is_inside_any(dir_list, fname):
453
"""True if fname is inside any of given dirs."""
454
for dirname in dir_list:
455
if is_inside(dirname, fname):
461
def is_inside_or_parent_of_any(dir_list, fname):
462
"""True if fname is a child or a parent of any of the given files."""
463
for dirname in dir_list:
464
if is_inside(dirname, fname) or is_inside(fname, dirname):
470
def pumpfile(fromfile, tofile):
471
"""Copy contents of one file to another."""
474
b = fromfile.read(BUFSIZE)
480
def file_iterator(input_file, readsize=32768):
482
b = input_file.read(readsize)
489
if hasattr(f, 'tell'):
502
def sha_strings(strings):
503
"""Return the sha-1 of concatenation of strings"""
505
map(s.update, strings)
515
def fingerprint_file(f):
520
return {'size': size,
521
'sha1': s.hexdigest()}
524
def compare_files(a, b):
525
"""Returns true if equal in contents"""
536
def local_time_offset(t=None):
537
"""Return offset of local zone from GMT, either at present or at time t."""
538
# python2.3 localtime() can't take None
542
if time.localtime(t).tm_isdst and time.daylight:
545
return -time.timezone
548
def format_date(t, offset=0, timezone='original', date_fmt=None,
550
## TODO: Perhaps a global option to use either universal or local time?
551
## Or perhaps just let people set $TZ?
552
assert isinstance(t, float)
554
if timezone == 'utc':
557
elif timezone == 'original':
560
tt = time.gmtime(t + offset)
561
elif timezone == 'local':
562
tt = time.localtime(t)
563
offset = local_time_offset(t)
565
raise BzrError("unsupported timezone format %r" % timezone,
566
['options are "utc", "original", "local"'])
568
date_fmt = "%a %Y-%m-%d %H:%M:%S"
570
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
573
return (time.strftime(date_fmt, tt) + offset_str)
576
def compact_date(when):
577
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
582
"""Return size of given open file."""
583
return os.fstat(f.fileno())[ST_SIZE]
586
# Define rand_bytes based on platform.
588
# Python 2.4 and later have os.urandom,
589
# but it doesn't work on some arches
591
rand_bytes = os.urandom
592
except (NotImplementedError, AttributeError):
593
# If python doesn't have os.urandom, or it doesn't work,
594
# then try to first pull random data from /dev/urandom
595
if os.path.exists("/dev/urandom"):
596
rand_bytes = file('/dev/urandom', 'rb').read
597
# Otherwise, use this hack as a last resort
599
# not well seeded, but better than nothing
604
s += chr(random.randint(0, 255))
609
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
611
"""Return a random string of num alphanumeric characters
613
The result only contains lowercase chars because it may be used on
614
case-insensitive filesystems.
617
for raw_byte in rand_bytes(num):
618
s += ALNUM[ord(raw_byte) % 36]
622
## TODO: We could later have path objects that remember their list
623
## decomposition (might be too tricksy though.)
626
"""Turn string into list of parts.
632
>>> splitpath('a/./b')
634
>>> splitpath('a/.b')
636
>>> splitpath('a/../b')
637
Traceback (most recent call last):
639
BzrError: sorry, '..' not allowed in path
641
assert isinstance(p, types.StringTypes)
643
# split on either delimiter because people might use either on
645
ps = re.split(r'[\\/]', p)
650
raise BzrError("sorry, %r not allowed in path" % f)
651
elif (f == '.') or (f == ''):
658
assert isinstance(p, list)
660
if (f == '..') or (f == None) or (f == ''):
661
raise BzrError("sorry, %r not allowed in path" % f)
665
@deprecated_function(zero_nine)
666
def appendpath(p1, p2):
670
return pathjoin(p1, p2)
674
"""Split s into lines, but without removing the newline characters."""
675
lines = s.split('\n')
676
result = [line + '\n' for line in lines[:-1]]
678
result.append(lines[-1])
682
def hardlinks_good():
683
return sys.platform not in ('win32', 'cygwin', 'darwin')
686
def link_or_copy(src, dest):
687
"""Hardlink a file, or copy it if it can't be hardlinked."""
688
if not hardlinks_good():
693
except (OSError, IOError), e:
694
if e.errno != errno.EXDEV:
698
def delete_any(full_path):
699
"""Delete a file or directory."""
703
# We may be renaming a dangling inventory id
704
if e.errno not in (errno.EISDIR, errno.EACCES, errno.EPERM):
710
if hasattr(os, 'symlink'):
716
def contains_whitespace(s):
717
"""True if there are any whitespace characters in s."""
718
for ch in string.whitespace:
725
def contains_linebreaks(s):
726
"""True if there is any vertical whitespace in s."""
734
def relpath(base, path):
735
"""Return path relative to base, or raise exception.
737
The path may be either an absolute path or a path relative to the
738
current working directory.
740
os.path.commonprefix (python2.4) has a bad bug that it works just
741
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
745
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
746
' exceed the platform minimum length (which is %d)' %
753
while len(head) >= len(base):
756
head, tail = os.path.split(head)
760
raise PathNotChild(rp, base)
768
def safe_unicode(unicode_or_utf8_string):
769
"""Coerce unicode_or_utf8_string into unicode.
771
If it is unicode, it is returned.
772
Otherwise it is decoded from utf-8. If a decoding error
773
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
774
as a BzrBadParameter exception.
776
if isinstance(unicode_or_utf8_string, unicode):
777
return unicode_or_utf8_string
779
return unicode_or_utf8_string.decode('utf8')
780
except UnicodeDecodeError:
781
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
784
_platform_normalizes_filenames = False
785
if sys.platform == 'darwin':
786
_platform_normalizes_filenames = True
789
def normalizes_filenames():
790
"""Return True if this platform normalizes unicode filenames.
792
Mac OSX does, Windows/Linux do not.
794
return _platform_normalizes_filenames
797
def _accessible_normalized_filename(path):
798
"""Get the unicode normalized path, and if you can access the file.
800
On platforms where the system normalizes filenames (Mac OSX),
801
you can access a file by any path which will normalize correctly.
802
On platforms where the system does not normalize filenames
803
(Windows, Linux), you have to access a file by its exact path.
805
Internally, bzr only supports NFC/NFKC normalization, since that is
806
the standard for XML documents.
808
So return the normalized path, and a flag indicating if the file
809
can be accessed by that path.
812
return unicodedata.normalize('NFKC', unicode(path)), True
815
def _inaccessible_normalized_filename(path):
816
__doc__ = _accessible_normalized_filename.__doc__
818
normalized = unicodedata.normalize('NFKC', unicode(path))
819
return normalized, normalized == path
822
if _platform_normalizes_filenames:
823
normalized_filename = _accessible_normalized_filename
825
normalized_filename = _inaccessible_normalized_filename
828
def terminal_width():
829
"""Return estimated terminal width."""
830
if sys.platform == 'win32':
831
import bzrlib.win32console
832
return bzrlib.win32console.get_console_size()[0]
835
import struct, fcntl, termios
836
s = struct.pack('HHHH', 0, 0, 0, 0)
837
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
838
width = struct.unpack('HHHH', x)[1]
843
width = int(os.environ['COLUMNS'])
851
def supports_executable():
852
return sys.platform != "win32"
855
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
858
def check_legal_path(path):
859
"""Check whether the supplied path is legal.
860
This is only required on Windows, so we don't test on other platforms
863
if sys.platform != "win32":
865
if _validWin32PathRE.match(path) is None:
866
raise IllegalPath(path)
869
def walkdirs(top, prefix=""):
870
"""Yield data about all the directories in a tree.
872
This yields all the data about the contents of a directory at a time.
873
After each directory has been yielded, if the caller has mutated the list
874
to exclude some directories, they are then not descended into.
876
The data yielded is of the form:
877
[(relpath, basename, kind, lstat, path_from_top), ...]
879
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
880
allows one to walk a subtree but get paths that are relative to a tree
882
:return: an iterator over the dirs.
886
_directory = _directory_kind
888
pending = [(prefix, "", _directory, None, top)]
891
currentdir = pending.pop()
892
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
895
relroot = currentdir[0] + '/'
898
for name in sorted(_listdir(top)):
899
abspath = top + '/' + name
900
statvalue = lstat(abspath)
901
dirblock.append ((relroot + name, name, file_kind_from_stat_mode(statvalue.st_mode), statvalue, abspath))
903
# push the user specified dirs from dirblock
904
for dir in reversed(dirblock):
905
if dir[2] == _directory:
909
def path_prefix_key(path):
910
"""Generate a prefix-order path key for path.
912
This can be used to sort paths in the same way that walkdirs does.
914
return (dirname(path) , path)
917
def compare_paths_prefix_order(path_a, path_b):
918
"""Compare path_a and path_b to generate the same order walkdirs uses."""
919
key_a = path_prefix_key(path_a)
920
key_b = path_prefix_key(path_b)
921
return cmp(key_a, key_b)
added
removed
bzrlib/osutils.py
