69
mode = os.lstat(f)[ST_MODE]
103
_directory_kind = 'directory'
106
stat.S_IFDIR:_directory_kind,
107
stat.S_IFCHR:'chardev',
108
stat.S_IFBLK:'block',
111
stat.S_IFLNK:'symlink',
112
stat.S_IFSOCK:'socket',
116
def file_kind_from_stat_mode(stat_mode, _formats=_formats, _unknown='unknown'):
117
"""Generate a file kind from a stat mode. This is used in walkdirs.
119
Its performance is critical: Do not mutate without careful benchmarking.
122
return _formats[stat_mode & 0170000]
127
def file_kind(f, _lstat=os.lstat, _mapper=file_kind_from_stat_mode):
129
return _mapper(_lstat(f).st_mode)
131
if getattr(e, 'errno', None) == errno.ENOENT:
132
raise errors.NoSuchFile(f)
137
"""Return the current umask"""
138
# Assume that people aren't messing with the umask while running
139
# XXX: This is not thread safe, but there is no way to get the
140
# umask without setting it
88
146
def kind_marker(kind):
89
147
if kind == 'file':
91
elif kind == 'directory':
149
elif kind == _directory_kind:
93
151
elif kind == 'symlink':
96
raise BzrError('invalid file kind %r' % kind)
100
if hasattr(os, 'lstat'):
106
if e.errno == errno.ENOENT:
109
raise BzrError("lstat/stat of (%r): %r" % (f, e))
154
raise errors.BzrError('invalid file kind %r' % kind)
156
lexists = getattr(os.path, 'lexists', None)
160
if getattr(os, 'lstat') is not None:
166
if e.errno == errno.ENOENT:
169
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
172
def fancy_rename(old, new, rename_func, unlink_func):
173
"""A fancy rename, when you don't have atomic rename.
175
:param old: The old path, to rename from
176
:param new: The new path, to rename to
177
:param rename_func: The potentially non-atomic rename function
178
:param unlink_func: A way to delete the target file if the full rename succeeds
181
# sftp rename doesn't allow overwriting, so play tricks:
183
base = os.path.basename(new)
184
dirname = os.path.dirname(new)
185
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
186
tmp_name = pathjoin(dirname, tmp_name)
188
# Rename the file out of the way, but keep track if it didn't exist
189
# We don't want to grab just any exception
190
# something like EACCES should prevent us from continuing
191
# The downside is that the rename_func has to throw an exception
192
# with an errno = ENOENT, or NoSuchFile
195
rename_func(new, tmp_name)
196
except (errors.NoSuchFile,), e:
199
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
200
# function raises an IOError with errno is None when a rename fails.
201
# This then gets caught here.
202
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
205
if (getattr(e, 'errno', None) is None
206
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
213
# This may throw an exception, in which case success will
215
rename_func(old, new)
219
# If the file used to exist, rename it back into place
220
# otherwise just delete it from the tmp location
222
unlink_func(tmp_name)
224
rename_func(tmp_name, new)
227
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
228
# choke on a Unicode string containing a relative path if
229
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
231
_fs_enc = sys.getfilesystemencoding()
232
def _posix_abspath(path):
233
# jam 20060426 rather than encoding to fsencoding
234
# copy posixpath.abspath, but use os.getcwdu instead
235
if not posixpath.isabs(path):
236
path = posixpath.join(getcwd(), path)
237
return posixpath.normpath(path)
240
def _posix_realpath(path):
241
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
244
def _win32_fixdrive(path):
245
"""Force drive letters to be consistent.
247
win32 is inconsistent whether it returns lower or upper case
248
and even if it was consistent the user might type the other
249
so we force it to uppercase
250
running python.exe under cmd.exe return capital C:\\
251
running win32 python inside a cygwin shell returns lowercase c:\\
253
drive, path = _nt_splitdrive(path)
254
return drive.upper() + path
257
def _win32_abspath(path):
258
# Real _nt_abspath doesn't have a problem with a unicode cwd
259
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
262
def _win32_realpath(path):
263
# Real _nt_realpath doesn't have a problem with a unicode cwd
264
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
267
def _win32_pathjoin(*args):
268
return _nt_join(*args).replace('\\', '/')
271
def _win32_normpath(path):
272
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
276
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
279
def _win32_mkdtemp(*args, **kwargs):
280
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
283
def _win32_rename(old, new):
284
"""We expect to be able to atomically replace 'new' with old.
286
On win32, if new exists, it must be moved out of the way first,
290
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
292
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
293
# If we try to rename a non-existant file onto cwd, we get
294
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
295
# if the old path doesn't exist, sometimes we get EACCES
296
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
302
return unicodedata.normalize('NFKC', os.getcwdu())
305
# Default is to just use the python builtins, but these can be rebound on
306
# particular platforms.
307
abspath = _posix_abspath
308
realpath = _posix_realpath
309
pathjoin = os.path.join
310
normpath = os.path.normpath
313
dirname = os.path.dirname
314
basename = os.path.basename
315
# These were already imported into local scope
316
# mkdtemp = tempfile.mkdtemp
317
# rmtree = shutil.rmtree
319
MIN_ABS_PATHLENGTH = 1
322
if sys.platform == 'win32':
323
abspath = _win32_abspath
324
realpath = _win32_realpath
325
pathjoin = _win32_pathjoin
326
normpath = _win32_normpath
327
getcwd = _win32_getcwd
328
mkdtemp = _win32_mkdtemp
329
rename = _win32_rename
331
MIN_ABS_PATHLENGTH = 3
333
def _win32_delete_readonly(function, path, excinfo):
334
"""Error handler for shutil.rmtree function [for win32]
335
Helps to remove files and dirs marked as read-only.
337
type_, value = excinfo[:2]
338
if function in (os.remove, os.rmdir) \
339
and type_ == OSError \
340
and value.errno == errno.EACCES:
346
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
347
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
348
return shutil.rmtree(path, ignore_errors, onerror)
349
elif sys.platform == 'darwin':
353
def get_terminal_encoding():
354
"""Find the best encoding for printing to the screen.
356
This attempts to check both sys.stdout and sys.stdin to see
357
what encoding they are in, and if that fails it falls back to
358
bzrlib.user_encoding.
359
The problem is that on Windows, locale.getpreferredencoding()
360
is not the same encoding as that used by the console:
361
http://mail.python.org/pipermail/python-list/2003-May/162357.html
363
On my standard US Windows XP, the preferred encoding is
364
cp1252, but the console is cp437
366
output_encoding = getattr(sys.stdout, 'encoding', None)
367
if not output_encoding:
368
input_encoding = getattr(sys.stdin, 'encoding', None)
369
if not input_encoding:
370
output_encoding = bzrlib.user_encoding
371
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
373
output_encoding = input_encoding
374
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
376
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
377
return output_encoding
111
380
def normalizepath(f):
112
if hasattr(os.path, 'realpath'):
381
if getattr(os.path, 'realpath', None) is not None:
116
385
[p,e] = os.path.split(f)
117
386
if e == "" or e == "." or e == "..":
120
return os.path.join(F(p), e)
389
return pathjoin(F(p), e)
123
392
def backup_file(fn):
124
393
"""Copy a file to a backup.
259
535
'sha1': s.hexdigest()}
263
"""Return per-user configuration directory.
265
By default this is ~/.bzr.conf/
267
TODO: Global option --config-dir to override this.
269
return os.path.join(os.path.expanduser("~"), ".bzr.conf")
273
"""Calculate automatic user identification.
275
Returns (realname, email).
277
Only used when none is set in the environment or the id file.
279
This previously used the FQDN as the default domain, but that can
280
be very slow on machines where DNS is broken. So now we simply
285
# XXX: Any good way to get real user name on win32?
290
w = pwd.getpwuid(uid)
291
gecos = w.pw_gecos.decode(bzrlib.user_encoding)
292
username = w.pw_name.decode(bzrlib.user_encoding)
293
comma = gecos.find(',')
297
realname = gecos[:comma]
303
realname = username = getpass.getuser().decode(bzrlib.user_encoding)
305
return realname, (username + '@' + socket.gethostname())
308
def _get_user_id(branch):
309
"""Return the full user id from a file or environment variable.
311
e.g. "John Hacker <jhacker@foo.org>"
314
A branch to use for a per-branch configuration, or None.
316
The following are searched in order:
319
2. .bzr/email for this branch.
323
v = os.environ.get('BZREMAIL')
325
return v.decode(bzrlib.user_encoding)
329
return (branch.controlfile("email", "r")
331
.decode(bzrlib.user_encoding)
334
if e.errno != errno.ENOENT:
340
return (open(os.path.join(config_dir(), "email"))
342
.decode(bzrlib.user_encoding)
345
if e.errno != errno.ENOENT:
348
v = os.environ.get('EMAIL')
350
return v.decode(bzrlib.user_encoding)
355
def username(branch):
356
"""Return email-style username.
358
Something similar to 'Martin Pool <mbp@sourcefrog.net>'
360
TODO: Check it's reasonably well-formed.
362
v = _get_user_id(branch)
366
name, email = _auto_user_id()
368
return '%s <%s>' % (name, email)
373
def user_email(branch):
374
"""Return just the email component of a username."""
375
e = _get_user_id(branch)
377
m = re.search(r'[\w+.-]+@[\w+.-]+', e)
379
raise BzrError("%r doesn't seem to contain "
380
"a reasonable email address" % e)
383
return _auto_user_id()[1]
386
538
def compare_files(a, b):
387
539
"""Returns true if equal in contents"""
416
569
tt = time.gmtime(t)
418
571
elif timezone == 'original':
421
574
tt = time.gmtime(t + offset)
422
575
elif timezone == 'local':
423
576
tt = time.localtime(t)
424
577
offset = local_time_offset(t)
426
raise BzrError("unsupported timezone format %r" % timezone,
427
['options are "utc", "original", "local"'])
429
return (time.strftime("%a %Y-%m-%d %H:%M:%S", tt)
430
+ ' %+03d%02d' % (offset / 3600, (offset / 60) % 60))
579
raise errors.BzrError("unsupported timezone format %r" % timezone,
580
['options are "utc", "original", "local"'])
582
date_fmt = "%a %Y-%m-%d %H:%M:%S"
584
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
587
return (time.strftime(date_fmt, tt) + offset_str)
433
590
def compact_date(when):
434
591
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
594
def format_delta(delta):
595
"""Get a nice looking string for a time delta.
597
:param delta: The time difference in seconds, can be positive or negative.
598
positive indicates time in the past, negative indicates time in the
599
future. (usually time.time() - stored_time)
600
:return: String formatted to show approximate resolution
606
direction = 'in the future'
610
if seconds < 90: # print seconds up to 90 seconds
612
return '%d second %s' % (seconds, direction,)
614
return '%d seconds %s' % (seconds, direction)
616
minutes = int(seconds / 60)
617
seconds -= 60 * minutes
622
if minutes < 90: # print minutes, seconds up to 90 minutes
624
return '%d minute, %d second%s %s' % (
625
minutes, seconds, plural_seconds, direction)
627
return '%d minutes, %d second%s %s' % (
628
minutes, seconds, plural_seconds, direction)
630
hours = int(minutes / 60)
631
minutes -= 60 * hours
638
return '%d hour, %d minute%s %s' % (hours, minutes,
639
plural_minutes, direction)
640
return '%d hours, %d minute%s %s' % (hours, minutes,
641
plural_minutes, direction)
439
644
"""Return size of given open file."""
440
645
return os.fstat(f.fileno())[ST_SIZE]
442
648
# Define rand_bytes based on platform.
444
650
# Python 2.4 and later have os.urandom,
535
734
def link_or_copy(src, dest):
536
735
"""Hardlink a file, or copy it if it can't be hardlinked."""
537
736
if not hardlinks_good():
737
shutil.copyfile(src, dest)
541
740
os.link(src, dest)
542
741
except (OSError, IOError), e:
543
742
if e.errno != errno.EXDEV:
744
shutil.copyfile(src, dest)
746
def delete_any(full_path):
747
"""Delete a file or directory."""
751
# We may be renaming a dangling inventory id
752
if e.errno not in (errno.EISDIR, errno.EACCES, errno.EPERM):
548
757
def has_symlinks():
549
if hasattr(os, 'symlink'):
758
if getattr(os, 'symlink', None) is not None:
764
def contains_whitespace(s):
765
"""True if there are any whitespace characters in s."""
766
for ch in string.whitespace:
773
def contains_linebreaks(s):
774
"""True if there is any vertical whitespace in s."""
782
def relpath(base, path):
783
"""Return path relative to base, or raise exception.
785
The path may be either an absolute path or a path relative to the
786
current working directory.
788
os.path.commonprefix (python2.4) has a bad bug that it works just
789
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
793
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
794
' exceed the platform minimum length (which is %d)' %
801
while len(head) >= len(base):
804
head, tail = os.path.split(head)
808
raise errors.PathNotChild(rp, base)
816
def safe_unicode(unicode_or_utf8_string):
817
"""Coerce unicode_or_utf8_string into unicode.
819
If it is unicode, it is returned.
820
Otherwise it is decoded from utf-8. If a decoding error
821
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
822
as a BzrBadParameter exception.
824
if isinstance(unicode_or_utf8_string, unicode):
825
return unicode_or_utf8_string
827
return unicode_or_utf8_string.decode('utf8')
828
except UnicodeDecodeError:
829
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
832
_platform_normalizes_filenames = False
833
if sys.platform == 'darwin':
834
_platform_normalizes_filenames = True
837
def normalizes_filenames():
838
"""Return True if this platform normalizes unicode filenames.
840
Mac OSX does, Windows/Linux do not.
842
return _platform_normalizes_filenames
845
def _accessible_normalized_filename(path):
846
"""Get the unicode normalized path, and if you can access the file.
848
On platforms where the system normalizes filenames (Mac OSX),
849
you can access a file by any path which will normalize correctly.
850
On platforms where the system does not normalize filenames
851
(Windows, Linux), you have to access a file by its exact path.
853
Internally, bzr only supports NFC/NFKC normalization, since that is
854
the standard for XML documents.
856
So return the normalized path, and a flag indicating if the file
857
can be accessed by that path.
860
return unicodedata.normalize('NFKC', unicode(path)), True
863
def _inaccessible_normalized_filename(path):
864
__doc__ = _accessible_normalized_filename.__doc__
866
normalized = unicodedata.normalize('NFKC', unicode(path))
867
return normalized, normalized == path
870
if _platform_normalizes_filenames:
871
normalized_filename = _accessible_normalized_filename
873
normalized_filename = _inaccessible_normalized_filename
876
def terminal_width():
877
"""Return estimated terminal width."""
878
if sys.platform == 'win32':
879
import bzrlib.win32console
880
return bzrlib.win32console.get_console_size()[0]
883
import struct, fcntl, termios
884
s = struct.pack('HHHH', 0, 0, 0, 0)
885
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
886
width = struct.unpack('HHHH', x)[1]
891
width = int(os.environ['COLUMNS'])
900
def supports_executable():
901
return sys.platform != "win32"
904
def set_or_unset_env(env_variable, value):
905
"""Modify the environment, setting or removing the env_variable.
907
:param env_variable: The environment variable in question
908
:param value: The value to set the environment to. If None, then
909
the variable will be removed.
910
:return: The original value of the environment variable.
912
orig_val = os.environ.get(env_variable)
914
if orig_val is not None:
915
del os.environ[env_variable]
917
if isinstance(value, unicode):
918
value = value.encode(bzrlib.user_encoding)
919
os.environ[env_variable] = value
923
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
926
def check_legal_path(path):
927
"""Check whether the supplied path is legal.
928
This is only required on Windows, so we don't test on other platforms
931
if sys.platform != "win32":
933
if _validWin32PathRE.match(path) is None:
934
raise errors.IllegalPath(path)
937
def walkdirs(top, prefix=""):
938
"""Yield data about all the directories in a tree.
940
This yields all the data about the contents of a directory at a time.
941
After each directory has been yielded, if the caller has mutated the list
942
to exclude some directories, they are then not descended into.
944
The data yielded is of the form:
945
((directory-relpath, directory-path-from-top),
946
[(relpath, basename, kind, lstat), ...]),
947
- directory-relpath is the relative path of the directory being returned
948
with respect to top. prefix is prepended to this.
949
- directory-path-from-root is the path including top for this directory.
950
It is suitable for use with os functions.
951
- relpath is the relative path within the subtree being walked.
952
- basename is the basename of the path
953
- kind is the kind of the file now. If unknown then the file is not
954
present within the tree - but it may be recorded as versioned. See
956
- lstat is the stat data *if* the file was statted.
957
- planned, not implemented:
958
path_from_tree_root is the path from the root of the tree.
960
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
961
allows one to walk a subtree but get paths that are relative to a tree
963
:return: an iterator over the dirs.
965
#TODO there is a bit of a smell where the results of the directory-
966
# summary in this, and the path from the root, may not agree
967
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
968
# potentially confusing output. We should make this more robust - but
969
# not at a speed cost. RBC 20060731
972
_directory = _directory_kind
973
_listdir = os.listdir
974
pending = [(prefix, "", _directory, None, top)]
977
currentdir = pending.pop()
978
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
981
relroot = currentdir[0] + '/'
984
for name in sorted(_listdir(top)):
985
abspath = top + '/' + name
986
statvalue = lstat(abspath)
987
dirblock.append((relroot + name, name,
988
file_kind_from_stat_mode(statvalue.st_mode),
990
yield (currentdir[0], top), dirblock
991
# push the user specified dirs from dirblock
992
for dir in reversed(dirblock):
993
if dir[2] == _directory:
997
def copy_tree(from_path, to_path, handlers={}):
998
"""Copy all of the entries in from_path into to_path.
1000
:param from_path: The base directory to copy.
1001
:param to_path: The target directory. If it does not exist, it will
1003
:param handlers: A dictionary of functions, which takes a source and
1004
destinations for files, directories, etc.
1005
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1006
'file', 'directory', and 'symlink' should always exist.
1007
If they are missing, they will be replaced with 'os.mkdir()',
1008
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1010
# Now, just copy the existing cached tree to the new location
1011
# We use a cheap trick here.
1012
# Absolute paths are prefixed with the first parameter
1013
# relative paths are prefixed with the second.
1014
# So we can get both the source and target returned
1015
# without any extra work.
1017
def copy_dir(source, dest):
1020
def copy_link(source, dest):
1021
"""Copy the contents of a symlink"""
1022
link_to = os.readlink(source)
1023
os.symlink(link_to, dest)
1025
real_handlers = {'file':shutil.copy2,
1026
'symlink':copy_link,
1027
'directory':copy_dir,
1029
real_handlers.update(handlers)
1031
if not os.path.exists(to_path):
1032
real_handlers['directory'](from_path, to_path)
1034
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1035
for relpath, name, kind, st, abspath in entries:
1036
real_handlers[kind](abspath, relpath)
1039
def path_prefix_key(path):
1040
"""Generate a prefix-order path key for path.
1042
This can be used to sort paths in the same way that walkdirs does.
1044
return (dirname(path) , path)
1047
def compare_paths_prefix_order(path_a, path_b):
1048
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1049
key_a = path_prefix_key(path_a)
1050
key_b = path_prefix_key(path_b)
1051
return cmp(key_a, key_b)
1054
_cached_user_encoding = None
1057
def get_user_encoding():
1058
"""Find out what the preferred user encoding is.
1060
This is generally the encoding that is used for command line parameters
1061
and file contents. This may be different from the terminal encoding
1062
or the filesystem encoding.
1064
:return: A string defining the preferred user encoding
1066
global _cached_user_encoding
1067
if _cached_user_encoding is not None:
1068
return _cached_user_encoding
1070
if sys.platform == 'darwin':
1071
# work around egregious python 2.4 bug
1072
sys.platform = 'posix'
1076
sys.platform = 'darwin'
1081
_cached_user_encoding = locale.getpreferredencoding()
1082
except locale.Error, e:
1083
sys.stderr.write('bzr: warning: %s\n'
1084
' Could not determine what text encoding to use.\n'
1085
' This error usually means your Python interpreter\n'
1086
' doesn\'t support the locale set by $LANG (%s)\n'
1087
" Continuing with ascii encoding.\n"
1088
% (e, os.environ.get('LANG')))
1090
if _cached_user_encoding is None:
1091
_cached_user_encoding = 'ascii'
1092
return _cached_user_encoding
added
removed
bzrlib/osutils.py
