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
69
mode = os.lstat(f)[ST_MODE]
146
88
def kind_marker(kind):
147
89
if kind == 'file':
149
elif kind == _directory_kind:
91
elif kind == 'directory':
151
93
elif kind == 'symlink':
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() or 'utf-8'
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
split = os.path.split
316
splitext = os.path.splitext
317
# These were already imported into local scope
318
# mkdtemp = tempfile.mkdtemp
319
# rmtree = shutil.rmtree
321
MIN_ABS_PATHLENGTH = 1
324
if sys.platform == 'win32':
325
abspath = _win32_abspath
326
realpath = _win32_realpath
327
pathjoin = _win32_pathjoin
328
normpath = _win32_normpath
329
getcwd = _win32_getcwd
330
mkdtemp = _win32_mkdtemp
331
rename = _win32_rename
333
MIN_ABS_PATHLENGTH = 3
335
def _win32_delete_readonly(function, path, excinfo):
336
"""Error handler for shutil.rmtree function [for win32]
337
Helps to remove files and dirs marked as read-only.
339
exception = excinfo[1]
340
if function in (os.remove, os.rmdir) \
341
and isinstance(exception, OSError) \
342
and exception.errno == errno.EACCES:
348
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
349
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
350
return shutil.rmtree(path, ignore_errors, onerror)
351
elif sys.platform == 'darwin':
355
def get_terminal_encoding():
356
"""Find the best encoding for printing to the screen.
358
This attempts to check both sys.stdout and sys.stdin to see
359
what encoding they are in, and if that fails it falls back to
360
bzrlib.user_encoding.
361
The problem is that on Windows, locale.getpreferredencoding()
362
is not the same encoding as that used by the console:
363
http://mail.python.org/pipermail/python-list/2003-May/162357.html
365
On my standard US Windows XP, the preferred encoding is
366
cp1252, but the console is cp437
368
output_encoding = getattr(sys.stdout, 'encoding', None)
369
if not output_encoding:
370
input_encoding = getattr(sys.stdin, 'encoding', None)
371
if not input_encoding:
372
output_encoding = bzrlib.user_encoding
373
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
375
output_encoding = input_encoding
376
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
378
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
379
if output_encoding == 'cp0':
380
# invalid encoding (cp0 means 'no codepage' on Windows)
381
output_encoding = bzrlib.user_encoding
382
mutter('cp0 is invalid encoding.'
383
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
386
codecs.lookup(output_encoding)
388
sys.stderr.write('bzr: warning:'
389
' unknown terminal encoding %s.\n'
390
' Using encoding %s instead.\n'
391
% (output_encoding, bzrlib.user_encoding)
393
output_encoding = bzrlib.user_encoding
395
return output_encoding
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))
398
111
def normalizepath(f):
399
if getattr(os.path, 'realpath', None) is not None:
112
if hasattr(os.path, 'realpath'):
403
116
[p,e] = os.path.split(f)
404
117
if e == "" or e == "." or e == "..":
407
return pathjoin(F(p), e)
120
return os.path.join(F(p), e)
410
123
def backup_file(fn):
411
124
"""Copy a file to a backup.
748
535
def link_or_copy(src, dest):
749
536
"""Hardlink a file, or copy it if it can't be hardlinked."""
750
537
if not hardlinks_good():
751
shutil.copyfile(src, dest)
754
541
os.link(src, dest)
755
542
except (OSError, IOError), e:
756
543
if e.errno != errno.EXDEV:
758
shutil.copyfile(src, dest)
760
def delete_any(full_path):
761
"""Delete a file or directory."""
765
# We may be renaming a dangling inventory id
766
if e.errno not in (errno.EISDIR, errno.EACCES, errno.EPERM):
771
548
def has_symlinks():
772
if getattr(os, 'symlink', None) is not None:
549
if hasattr(os, 'symlink'):
778
def contains_whitespace(s):
779
"""True if there are any whitespace characters in s."""
780
# string.whitespace can include '\xa0' in certain locales, because it is
781
# considered "non-breaking-space" as part of ISO-8859-1. But it
782
# 1) Isn't a breaking whitespace
783
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
785
# 3) '\xa0' isn't unicode safe since it is >128.
786
# So we are following textwrap's example and hard-coding our own.
787
# We probably could ignore \v and \f, too.
788
for ch in u' \t\n\r\v\f':
795
def contains_linebreaks(s):
796
"""True if there is any vertical whitespace in s."""
804
def relpath(base, path):
805
"""Return path relative to base, or raise exception.
807
The path may be either an absolute path or a path relative to the
808
current working directory.
810
os.path.commonprefix (python2.4) has a bad bug that it works just
811
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
815
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
816
' exceed the platform minimum length (which is %d)' %
823
while len(head) >= len(base):
826
head, tail = os.path.split(head)
830
raise errors.PathNotChild(rp, base)
838
def safe_unicode(unicode_or_utf8_string):
839
"""Coerce unicode_or_utf8_string into unicode.
841
If it is unicode, it is returned.
842
Otherwise it is decoded from utf-8. If a decoding error
843
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
844
as a BzrBadParameter exception.
846
if isinstance(unicode_or_utf8_string, unicode):
847
return unicode_or_utf8_string
849
return unicode_or_utf8_string.decode('utf8')
850
except UnicodeDecodeError:
851
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
854
_platform_normalizes_filenames = False
855
if sys.platform == 'darwin':
856
_platform_normalizes_filenames = True
859
def normalizes_filenames():
860
"""Return True if this platform normalizes unicode filenames.
862
Mac OSX does, Windows/Linux do not.
864
return _platform_normalizes_filenames
867
def _accessible_normalized_filename(path):
868
"""Get the unicode normalized path, and if you can access the file.
870
On platforms where the system normalizes filenames (Mac OSX),
871
you can access a file by any path which will normalize correctly.
872
On platforms where the system does not normalize filenames
873
(Windows, Linux), you have to access a file by its exact path.
875
Internally, bzr only supports NFC/NFKC normalization, since that is
876
the standard for XML documents.
878
So return the normalized path, and a flag indicating if the file
879
can be accessed by that path.
882
return unicodedata.normalize('NFKC', unicode(path)), True
885
def _inaccessible_normalized_filename(path):
886
__doc__ = _accessible_normalized_filename.__doc__
888
normalized = unicodedata.normalize('NFKC', unicode(path))
889
return normalized, normalized == path
892
if _platform_normalizes_filenames:
893
normalized_filename = _accessible_normalized_filename
895
normalized_filename = _inaccessible_normalized_filename
898
def terminal_width():
899
"""Return estimated terminal width."""
900
if sys.platform == 'win32':
901
return win32utils.get_console_size()[0]
904
import struct, fcntl, termios
905
s = struct.pack('HHHH', 0, 0, 0, 0)
906
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
907
width = struct.unpack('HHHH', x)[1]
912
width = int(os.environ['COLUMNS'])
921
def supports_executable():
922
return sys.platform != "win32"
925
def supports_posix_readonly():
926
"""Return True if 'readonly' has POSIX semantics, False otherwise.
928
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
929
directory controls creation/deletion, etc.
931
And under win32, readonly means that the directory itself cannot be
932
deleted. The contents of a readonly directory can be changed, unlike POSIX
933
where files in readonly directories cannot be added, deleted or renamed.
935
return sys.platform != "win32"
938
def set_or_unset_env(env_variable, value):
939
"""Modify the environment, setting or removing the env_variable.
941
:param env_variable: The environment variable in question
942
:param value: The value to set the environment to. If None, then
943
the variable will be removed.
944
:return: The original value of the environment variable.
946
orig_val = os.environ.get(env_variable)
948
if orig_val is not None:
949
del os.environ[env_variable]
951
if isinstance(value, unicode):
952
value = value.encode(bzrlib.user_encoding)
953
os.environ[env_variable] = value
957
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
960
def check_legal_path(path):
961
"""Check whether the supplied path is legal.
962
This is only required on Windows, so we don't test on other platforms
965
if sys.platform != "win32":
967
if _validWin32PathRE.match(path) is None:
968
raise errors.IllegalPath(path)
971
def walkdirs(top, prefix=""):
972
"""Yield data about all the directories in a tree.
974
This yields all the data about the contents of a directory at a time.
975
After each directory has been yielded, if the caller has mutated the list
976
to exclude some directories, they are then not descended into.
978
The data yielded is of the form:
979
((directory-relpath, directory-path-from-top),
980
[(relpath, basename, kind, lstat), ...]),
981
- directory-relpath is the relative path of the directory being returned
982
with respect to top. prefix is prepended to this.
983
- directory-path-from-root is the path including top for this directory.
984
It is suitable for use with os functions.
985
- relpath is the relative path within the subtree being walked.
986
- basename is the basename of the path
987
- kind is the kind of the file now. If unknown then the file is not
988
present within the tree - but it may be recorded as versioned. See
990
- lstat is the stat data *if* the file was statted.
991
- planned, not implemented:
992
path_from_tree_root is the path from the root of the tree.
994
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
995
allows one to walk a subtree but get paths that are relative to a tree
997
:return: an iterator over the dirs.
999
#TODO there is a bit of a smell where the results of the directory-
1000
# summary in this, and the path from the root, may not agree
1001
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1002
# potentially confusing output. We should make this more robust - but
1003
# not at a speed cost. RBC 20060731
1006
_directory = _directory_kind
1007
_listdir = os.listdir
1008
pending = [(prefix, "", _directory, None, top)]
1011
currentdir = pending.pop()
1012
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1015
relroot = currentdir[0] + '/'
1018
for name in sorted(_listdir(top)):
1019
abspath = top + '/' + name
1020
statvalue = lstat(abspath)
1021
dirblock.append((relroot + name, name,
1022
file_kind_from_stat_mode(statvalue.st_mode),
1023
statvalue, abspath))
1024
yield (currentdir[0], top), dirblock
1025
# push the user specified dirs from dirblock
1026
for dir in reversed(dirblock):
1027
if dir[2] == _directory:
1031
def copy_tree(from_path, to_path, handlers={}):
1032
"""Copy all of the entries in from_path into to_path.
1034
:param from_path: The base directory to copy.
1035
:param to_path: The target directory. If it does not exist, it will
1037
:param handlers: A dictionary of functions, which takes a source and
1038
destinations for files, directories, etc.
1039
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1040
'file', 'directory', and 'symlink' should always exist.
1041
If they are missing, they will be replaced with 'os.mkdir()',
1042
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1044
# Now, just copy the existing cached tree to the new location
1045
# We use a cheap trick here.
1046
# Absolute paths are prefixed with the first parameter
1047
# relative paths are prefixed with the second.
1048
# So we can get both the source and target returned
1049
# without any extra work.
1051
def copy_dir(source, dest):
1054
def copy_link(source, dest):
1055
"""Copy the contents of a symlink"""
1056
link_to = os.readlink(source)
1057
os.symlink(link_to, dest)
1059
real_handlers = {'file':shutil.copy2,
1060
'symlink':copy_link,
1061
'directory':copy_dir,
1063
real_handlers.update(handlers)
1065
if not os.path.exists(to_path):
1066
real_handlers['directory'](from_path, to_path)
1068
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1069
for relpath, name, kind, st, abspath in entries:
1070
real_handlers[kind](abspath, relpath)
1073
def path_prefix_key(path):
1074
"""Generate a prefix-order path key for path.
1076
This can be used to sort paths in the same way that walkdirs does.
1078
return (dirname(path) , path)
1081
def compare_paths_prefix_order(path_a, path_b):
1082
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1083
key_a = path_prefix_key(path_a)
1084
key_b = path_prefix_key(path_b)
1085
return cmp(key_a, key_b)
1088
_cached_user_encoding = None
1091
def get_user_encoding(use_cache=True):
1092
"""Find out what the preferred user encoding is.
1094
This is generally the encoding that is used for command line parameters
1095
and file contents. This may be different from the terminal encoding
1096
or the filesystem encoding.
1098
:param use_cache: Enable cache for detected encoding.
1099
(This parameter is turned on by default,
1100
and required only for selftesting)
1102
:return: A string defining the preferred user encoding
1104
global _cached_user_encoding
1105
if _cached_user_encoding is not None and use_cache:
1106
return _cached_user_encoding
1108
if sys.platform == 'darwin':
1109
# work around egregious python 2.4 bug
1110
sys.platform = 'posix'
1114
sys.platform = 'darwin'
1119
user_encoding = locale.getpreferredencoding()
1120
except locale.Error, e:
1121
sys.stderr.write('bzr: warning: %s\n'
1122
' Could not determine what text encoding to use.\n'
1123
' This error usually means your Python interpreter\n'
1124
' doesn\'t support the locale set by $LANG (%s)\n'
1125
" Continuing with ascii encoding.\n"
1126
% (e, os.environ.get('LANG')))
1127
user_encoding = 'ascii'
1129
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1130
# treat that as ASCII, and not support printing unicode characters to the
1132
if user_encoding in (None, 'cp0'):
1133
user_encoding = 'ascii'
1137
codecs.lookup(user_encoding)
1139
sys.stderr.write('bzr: warning:'
1140
' unknown encoding %s.'
1141
' Continuing with ascii encoding.\n'
1144
user_encoding = 'ascii'
1147
_cached_user_encoding = user_encoding
1149
return user_encoding
1152
def recv_all(socket, bytes):
1153
"""Receive an exact number of bytes.
1155
Regular Socket.recv() may return less than the requested number of bytes,
1156
dependning on what's in the OS buffer. MSG_WAITALL is not available
1157
on all platforms, but this should work everywhere. This will return
1158
less than the requested amount if the remote end closes.
1160
This isn't optimized and is intended mostly for use in testing.
1163
while len(b) < bytes:
1164
new = socket.recv(bytes - len(b))
1170
def dereference_path(path):
1171
"""Determine the real path to a file.
1173
All parent elements are dereferenced. But the file itself is not
1175
:param path: The original path. May be absolute or relative.
1176
:return: the real path *to* the file
1178
parent, base = os.path.split(path)
1179
# The pathjoin for '.' is a workaround for Python bug #1213894.
1180
# (initial path components aren't dereferenced)
1181
return pathjoin(realpath(pathjoin('.', parent)), base)
added
removed
bzrlib/osutils.py
