1
# Bazaar -- 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, errno.EINVAL):
268
# If we try to rename a non-existant file onto cwd, we get
269
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
270
# if the old path doesn't exist, sometimes we get EACCES
271
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
277
return unicodedata.normalize('NFKC', os.getcwdu())
280
# Default is to just use the python builtins, but these can be rebound on
281
# particular platforms.
282
abspath = _posix_abspath
283
realpath = _posix_realpath
284
pathjoin = os.path.join
285
normpath = os.path.normpath
287
mkdtemp = tempfile.mkdtemp
289
dirname = os.path.dirname
290
basename = os.path.basename
291
rmtree = shutil.rmtree
293
MIN_ABS_PATHLENGTH = 1
296
if sys.platform == 'win32':
297
abspath = _win32_abspath
298
realpath = _win32_realpath
299
pathjoin = _win32_pathjoin
300
normpath = _win32_normpath
301
getcwd = _win32_getcwd
302
mkdtemp = _win32_mkdtemp
303
rename = _win32_rename
305
MIN_ABS_PATHLENGTH = 3
307
def _win32_delete_readonly(function, path, excinfo):
308
"""Error handler for shutil.rmtree function [for win32]
309
Helps to remove files and dirs marked as read-only.
311
type_, value = excinfo[:2]
312
if function in (os.remove, os.rmdir) \
313
and type_ == OSError \
314
and value.errno == errno.EACCES:
315
bzrlib.osutils.make_writable(path)
320
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
321
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
322
return shutil.rmtree(path, ignore_errors, onerror)
323
elif sys.platform == 'darwin':
327
def get_terminal_encoding():
328
"""Find the best encoding for printing to the screen.
330
This attempts to check both sys.stdout and sys.stdin to see
331
what encoding they are in, and if that fails it falls back to
332
bzrlib.user_encoding.
333
The problem is that on Windows, locale.getpreferredencoding()
334
is not the same encoding as that used by the console:
335
http://mail.python.org/pipermail/python-list/2003-May/162357.html
337
On my standard US Windows XP, the preferred encoding is
338
cp1252, but the console is cp437
340
output_encoding = getattr(sys.stdout, 'encoding', None)
341
if not output_encoding:
342
input_encoding = getattr(sys.stdin, 'encoding', None)
343
if not input_encoding:
344
output_encoding = bzrlib.user_encoding
345
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
347
output_encoding = input_encoding
348
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
350
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
351
return output_encoding
354
def normalizepath(f):
355
if hasattr(os.path, 'realpath'):
359
[p,e] = os.path.split(f)
360
if e == "" or e == "." or e == "..":
363
return pathjoin(F(p), e)
367
"""Copy a file to a backup.
369
Backups are named in GNU-style, with a ~ suffix.
371
If the file is already a backup, it's not copied.
377
if has_symlinks() and os.path.islink(fn):
378
target = os.readlink(fn)
379
os.symlink(target, bfn)
387
outf = file(bfn, 'wb')
395
"""True if f is an accessible directory."""
397
return S_ISDIR(os.lstat(f)[ST_MODE])
403
"""True if f is a regular file."""
405
return S_ISREG(os.lstat(f)[ST_MODE])
410
"""True if f is a symlink."""
412
return S_ISLNK(os.lstat(f)[ST_MODE])
416
def is_inside(dir, fname):
417
"""True if fname is inside dir.
419
The parameters should typically be passed to osutils.normpath first, so
420
that . and .. and repeated slashes are eliminated, and the separators
421
are canonical for the platform.
423
The empty string as a dir name is taken as top-of-tree and matches
426
>>> is_inside('src', pathjoin('src', 'foo.c'))
428
>>> is_inside('src', 'srccontrol')
430
>>> is_inside('src', pathjoin('src', 'a', 'a', 'a', 'foo.c'))
432
>>> is_inside('foo.c', 'foo.c')
434
>>> is_inside('foo.c', '')
436
>>> is_inside('', 'foo.c')
439
# XXX: Most callers of this can actually do something smarter by
440
# looking at the inventory
450
return fname.startswith(dir)
453
def is_inside_any(dir_list, fname):
454
"""True if fname is inside any of given dirs."""
455
for dirname in dir_list:
456
if is_inside(dirname, fname):
462
def is_inside_or_parent_of_any(dir_list, fname):
463
"""True if fname is a child or a parent of any of the given files."""
464
for dirname in dir_list:
465
if is_inside(dirname, fname) or is_inside(fname, dirname):
471
def pumpfile(fromfile, tofile):
472
"""Copy contents of one file to another."""
475
b = fromfile.read(BUFSIZE)
481
def file_iterator(input_file, readsize=32768):
483
b = input_file.read(readsize)
490
if hasattr(f, 'tell'):
503
def sha_strings(strings):
504
"""Return the sha-1 of concatenation of strings"""
506
map(s.update, strings)
516
def fingerprint_file(f):
521
return {'size': size,
522
'sha1': s.hexdigest()}
525
def compare_files(a, b):
526
"""Returns true if equal in contents"""
537
def local_time_offset(t=None):
538
"""Return offset of local zone from GMT, either at present or at time t."""
539
# python2.3 localtime() can't take None
543
if time.localtime(t).tm_isdst and time.daylight:
546
return -time.timezone
549
def format_date(t, offset=0, timezone='original', date_fmt=None,
551
## TODO: Perhaps a global option to use either universal or local time?
552
## Or perhaps just let people set $TZ?
553
assert isinstance(t, float)
555
if timezone == 'utc':
558
elif timezone == 'original':
561
tt = time.gmtime(t + offset)
562
elif timezone == 'local':
563
tt = time.localtime(t)
564
offset = local_time_offset(t)
566
raise BzrError("unsupported timezone format %r" % timezone,
567
['options are "utc", "original", "local"'])
569
date_fmt = "%a %Y-%m-%d %H:%M:%S"
571
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
574
return (time.strftime(date_fmt, tt) + offset_str)
577
def compact_date(when):
578
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
583
"""Return size of given open file."""
584
return os.fstat(f.fileno())[ST_SIZE]
587
# Define rand_bytes based on platform.
589
# Python 2.4 and later have os.urandom,
590
# but it doesn't work on some arches
592
rand_bytes = os.urandom
593
except (NotImplementedError, AttributeError):
594
# If python doesn't have os.urandom, or it doesn't work,
595
# then try to first pull random data from /dev/urandom
596
if os.path.exists("/dev/urandom"):
597
rand_bytes = file('/dev/urandom', 'rb').read
598
# Otherwise, use this hack as a last resort
600
# not well seeded, but better than nothing
605
s += chr(random.randint(0, 255))
610
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
612
"""Return a random string of num alphanumeric characters
614
The result only contains lowercase chars because it may be used on
615
case-insensitive filesystems.
618
for raw_byte in rand_bytes(num):
619
s += ALNUM[ord(raw_byte) % 36]
623
## TODO: We could later have path objects that remember their list
624
## decomposition (might be too tricksy though.)
627
"""Turn string into list of parts.
633
>>> splitpath('a/./b')
635
>>> splitpath('a/.b')
637
>>> splitpath('a/../b')
638
Traceback (most recent call last):
640
BzrError: sorry, '..' not allowed in path
642
assert isinstance(p, types.StringTypes)
644
# split on either delimiter because people might use either on
646
ps = re.split(r'[\\/]', p)
651
raise BzrError("sorry, %r not allowed in path" % f)
652
elif (f == '.') or (f == ''):
659
assert isinstance(p, list)
661
if (f == '..') or (f == None) or (f == ''):
662
raise BzrError("sorry, %r not allowed in path" % f)
666
@deprecated_function(zero_nine)
667
def appendpath(p1, p2):
671
return pathjoin(p1, p2)
675
"""Split s into lines, but without removing the newline characters."""
676
lines = s.split('\n')
677
result = [line + '\n' for line in lines[:-1]]
679
result.append(lines[-1])
683
def hardlinks_good():
684
return sys.platform not in ('win32', 'cygwin', 'darwin')
687
def link_or_copy(src, dest):
688
"""Hardlink a file, or copy it if it can't be hardlinked."""
689
if not hardlinks_good():
694
except (OSError, IOError), e:
695
if e.errno != errno.EXDEV:
699
def delete_any(full_path):
700
"""Delete a file or directory."""
704
# We may be renaming a dangling inventory id
705
if e.errno not in (errno.EISDIR, errno.EACCES, errno.EPERM):
711
if hasattr(os, 'symlink'):
717
def contains_whitespace(s):
718
"""True if there are any whitespace characters in s."""
719
for ch in string.whitespace:
726
def contains_linebreaks(s):
727
"""True if there is any vertical whitespace in s."""
735
def relpath(base, path):
736
"""Return path relative to base, or raise exception.
738
The path may be either an absolute path or a path relative to the
739
current working directory.
741
os.path.commonprefix (python2.4) has a bad bug that it works just
742
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
746
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
747
' exceed the platform minimum length (which is %d)' %
754
while len(head) >= len(base):
757
head, tail = os.path.split(head)
761
raise PathNotChild(rp, base)
769
def safe_unicode(unicode_or_utf8_string):
770
"""Coerce unicode_or_utf8_string into unicode.
772
If it is unicode, it is returned.
773
Otherwise it is decoded from utf-8. If a decoding error
774
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
775
as a BzrBadParameter exception.
777
if isinstance(unicode_or_utf8_string, unicode):
778
return unicode_or_utf8_string
780
return unicode_or_utf8_string.decode('utf8')
781
except UnicodeDecodeError:
782
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
785
_platform_normalizes_filenames = False
786
if sys.platform == 'darwin':
787
_platform_normalizes_filenames = True
790
def normalizes_filenames():
791
"""Return True if this platform normalizes unicode filenames.
793
Mac OSX does, Windows/Linux do not.
795
return _platform_normalizes_filenames
798
def _accessible_normalized_filename(path):
799
"""Get the unicode normalized path, and if you can access the file.
801
On platforms where the system normalizes filenames (Mac OSX),
802
you can access a file by any path which will normalize correctly.
803
On platforms where the system does not normalize filenames
804
(Windows, Linux), you have to access a file by its exact path.
806
Internally, bzr only supports NFC/NFKC normalization, since that is
807
the standard for XML documents.
809
So return the normalized path, and a flag indicating if the file
810
can be accessed by that path.
813
return unicodedata.normalize('NFKC', unicode(path)), True
816
def _inaccessible_normalized_filename(path):
817
__doc__ = _accessible_normalized_filename.__doc__
819
normalized = unicodedata.normalize('NFKC', unicode(path))
820
return normalized, normalized == path
823
if _platform_normalizes_filenames:
824
normalized_filename = _accessible_normalized_filename
826
normalized_filename = _inaccessible_normalized_filename
829
def terminal_width():
830
"""Return estimated terminal width."""
831
if sys.platform == 'win32':
832
import bzrlib.win32console
833
return bzrlib.win32console.get_console_size()[0]
836
import struct, fcntl, termios
837
s = struct.pack('HHHH', 0, 0, 0, 0)
838
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
839
width = struct.unpack('HHHH', x)[1]
844
width = int(os.environ['COLUMNS'])
852
def supports_executable():
853
return sys.platform != "win32"
856
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
859
def check_legal_path(path):
860
"""Check whether the supplied path is legal.
861
This is only required on Windows, so we don't test on other platforms
864
if sys.platform != "win32":
866
if _validWin32PathRE.match(path) is None:
867
raise IllegalPath(path)
870
def walkdirs(top, prefix=""):
871
"""Yield data about all the directories in a tree.
873
This yields all the data about the contents of a directory at a time.
874
After each directory has been yielded, if the caller has mutated the list
875
to exclude some directories, they are then not descended into.
877
The data yielded is of the form:
878
[(relpath, basename, kind, lstat, path_from_top), ...]
880
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
881
allows one to walk a subtree but get paths that are relative to a tree
883
:return: an iterator over the dirs.
887
_directory = _directory_kind
889
pending = [(prefix, "", _directory, None, top)]
892
currentdir = pending.pop()
893
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
896
relroot = currentdir[0] + '/'
899
for name in sorted(_listdir(top)):
900
abspath = top + '/' + name
901
statvalue = lstat(abspath)
902
dirblock.append ((relroot + name, name, file_kind_from_stat_mode(statvalue.st_mode), statvalue, abspath))
904
# push the user specified dirs from dirblock
905
for dir in reversed(dirblock):
906
if dir[2] == _directory:
910
def path_prefix_key(path):
911
"""Generate a prefix-order path key for path.
913
This can be used to sort paths in the same way that walkdirs does.
915
return (dirname(path) , path)
918
def compare_paths_prefix_order(path_a, path_b):
919
"""Compare path_a and path_b to generate the same order walkdirs uses."""
920
key_a = path_prefix_key(path_a)
921
key_b = path_prefix_key(path_b)
922
return cmp(key_a, key_b)
added
removed
bzrlib/osutils.py
