49
from bzrlib.symbol_versioning import *
51
from bzrlib.symbol_versioning import (deprecated_function,
50
53
from bzrlib.trace import mutter
51
import bzrlib.win32console
56
# On win32, O_BINARY is used to indicate the file should
57
# be opened in binary mode, rather than text mode.
58
# On other platforms, O_BINARY doesn't exist, because
59
# they always open in binary mode, so it is okay to
60
# OR with 0 on those platforms
61
O_BINARY = getattr(os, 'O_BINARY', 0)
54
64
def make_readonly(filename):
204
224
_fs_enc = sys.getfilesystemencoding()
205
225
def _posix_abspath(path):
206
return os.path.abspath(path.encode(_fs_enc)).decode(_fs_enc)
207
# jam 20060426 This is another possibility which mimics
208
# os.path.abspath, only uses unicode characters instead
209
# if not os.path.isabs(path):
210
# return os.path.join(os.getcwdu(), path)
226
# jam 20060426 rather than encoding to fsencoding
227
# copy posixpath.abspath, but use os.getcwdu instead
228
if not posixpath.isabs(path):
229
path = posixpath.join(getcwd(), path)
230
return posixpath.normpath(path)
214
233
def _posix_realpath(path):
215
return os.path.realpath(path.encode(_fs_enc)).decode(_fs_enc)
234
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
237
def _win32_fixdrive(path):
238
"""Force drive letters to be consistent.
240
win32 is inconsistent whether it returns lower or upper case
241
and even if it was consistent the user might type the other
242
so we force it to uppercase
243
running python.exe under cmd.exe return capital C:\\
244
running win32 python inside a cygwin shell returns lowercase c:\\
246
drive, path = _nt_splitdrive(path)
247
return drive.upper() + path
218
250
def _win32_abspath(path):
219
return _nt_abspath(path.encode(_fs_enc)).decode(_fs_enc).replace('\\', '/')
251
# Real _nt_abspath doesn't have a problem with a unicode cwd
252
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
222
255
def _win32_realpath(path):
223
return _nt_realpath(path.encode(_fs_enc)).decode(_fs_enc).replace('\\', '/')
256
# Real _nt_realpath doesn't have a problem with a unicode cwd
257
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
226
260
def _win32_pathjoin(*args):
230
264
def _win32_normpath(path):
231
return _nt_normpath(path).replace('\\', '/')
265
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
234
268
def _win32_getcwd():
235
return os.getcwdu().replace('\\', '/')
269
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
238
272
def _win32_mkdtemp(*args, **kwargs):
239
return tempfile.mkdtemp(*args, **kwargs).replace('\\', '/')
273
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
242
276
def _win32_rename(old, new):
243
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
277
"""We expect to be able to atomically replace 'new' with old.
279
On win32, if new exists, it must be moved out of the way first,
283
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
285
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
286
# If we try to rename a non-existant file onto cwd, we get
287
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
288
# if the old path doesn't exist, sometimes we get EACCES
289
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
295
return unicodedata.normalize('NFKC', os.getcwdu())
246
298
# Default is to just use the python builtins, but these can be rebound on
286
338
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
287
339
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
288
340
return shutil.rmtree(path, ignore_errors, onerror)
341
elif sys.platform == 'darwin':
345
def get_terminal_encoding():
346
"""Find the best encoding for printing to the screen.
348
This attempts to check both sys.stdout and sys.stdin to see
349
what encoding they are in, and if that fails it falls back to
350
bzrlib.user_encoding.
351
The problem is that on Windows, locale.getpreferredencoding()
352
is not the same encoding as that used by the console:
353
http://mail.python.org/pipermail/python-list/2003-May/162357.html
355
On my standard US Windows XP, the preferred encoding is
356
cp1252, but the console is cp437
358
output_encoding = getattr(sys.stdout, 'encoding', None)
359
if not output_encoding:
360
input_encoding = getattr(sys.stdin, 'encoding', None)
361
if not input_encoding:
362
output_encoding = bzrlib.user_encoding
363
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
365
output_encoding = input_encoding
366
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
368
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
369
return output_encoding
291
372
def normalizepath(f):
292
if hasattr(os.path, 'realpath'):
373
if getattr(os.path, 'realpath', None) is not None:
732
813
return _platform_normalizes_filenames
816
def _accessible_normalized_filename(path):
817
"""Get the unicode normalized path, and if you can access the file.
819
On platforms where the system normalizes filenames (Mac OSX),
820
you can access a file by any path which will normalize correctly.
821
On platforms where the system does not normalize filenames
822
(Windows, Linux), you have to access a file by its exact path.
824
Internally, bzr only supports NFC/NFKC normalization, since that is
825
the standard for XML documents.
827
So return the normalized path, and a flag indicating if the file
828
can be accessed by that path.
831
return unicodedata.normalize('NFKC', unicode(path)), True
834
def _inaccessible_normalized_filename(path):
835
__doc__ = _accessible_normalized_filename.__doc__
837
normalized = unicodedata.normalize('NFKC', unicode(path))
838
return normalized, normalized == path
735
841
if _platform_normalizes_filenames:
736
def unicode_filename(path):
737
"""Make sure 'path' is a properly normalized filename.
739
On platforms where the system normalizes filenames (Mac OSX),
740
you can access a file by any path which will normalize
742
Internally, bzr only supports NFC/NFKC normalization, since
743
that is the standard for XML documents.
744
So we return an normalized path, and indicate this has been
747
:return: (path, is_normalized) Return a path which can
748
access the file, and whether or not this path is
751
return unicodedata.normalize('NFKC', path), True
842
normalized_filename = _accessible_normalized_filename
753
def unicode_filename(path):
754
"""Make sure 'path' is a properly normalized filename.
756
On platforms where the system does not normalize filenames
757
(Windows, Linux), you have to access a file by its exact path.
758
Internally, bzr only supports NFC/NFKC normalization, since
759
that is the standard for XML documents.
760
So we return the original path, and indicate if this is
763
:return: (path, is_normalized) Return a path which can
764
access the file, and whether or not this path is
767
return path, unicodedata.normalize('NFKC', path) == path
844
normalized_filename = _inaccessible_normalized_filename
770
847
def terminal_width():
793
871
def supports_executable():
794
872
return sys.platform != "win32"
875
def set_or_unset_env(env_variable, value):
876
"""Modify the environment, setting or removing the env_variable.
878
:param env_variable: The environment variable in question
879
:param value: The value to set the environment to. If None, then
880
the variable will be removed.
881
:return: The original value of the environment variable.
883
orig_val = os.environ.get(env_variable)
885
if orig_val is not None:
886
del os.environ[env_variable]
888
if isinstance(value, unicode):
889
value = value.encode(bzrlib.user_encoding)
890
os.environ[env_variable] = value
797
894
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
816
913
to exclude some directories, they are then not descended into.
818
915
The data yielded is of the form:
819
[(relpath, basename, kind, lstat, path_from_top), ...]
916
((directory-relpath, directory-path-from-top),
917
[(relpath, basename, kind, lstat), ...]),
918
- directory-relpath is the relative path of the directory being returned
919
with respect to top. prefix is prepended to this.
920
- directory-path-from-root is the path including top for this directory.
921
It is suitable for use with os functions.
922
- relpath is the relative path within the subtree being walked.
923
- basename is the basename of the path
924
- kind is the kind of the file now. If unknown then the file is not
925
present within the tree - but it may be recorded as versioned. See
927
- lstat is the stat data *if* the file was statted.
928
- planned, not implemented:
929
path_from_tree_root is the path from the root of the tree.
821
931
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
822
932
allows one to walk a subtree but get paths that are relative to a tree
823
933
rooted higher up.
824
934
:return: an iterator over the dirs.
936
#TODO there is a bit of a smell where the results of the directory-
937
# summary in this, and the path from the root, may not agree
938
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
939
# potentially confusing output. We should make this more robust - but
940
# not at a speed cost. RBC 20060731
828
943
_directory = _directory_kind
840
955
for name in sorted(_listdir(top)):
841
956
abspath = top + '/' + name
842
957
statvalue = lstat(abspath)
843
dirblock.append ((relroot + name, name, file_kind_from_stat_mode(statvalue.st_mode), statvalue, abspath))
958
dirblock.append((relroot + name, name,
959
file_kind_from_stat_mode(statvalue.st_mode),
961
yield (currentdir[0], top), dirblock
845
962
# push the user specified dirs from dirblock
846
963
for dir in reversed(dirblock):
847
964
if dir[2] == _directory:
848
965
pending.append(dir)
968
def copy_tree(from_path, to_path, handlers={}):
969
"""Copy all of the entries in from_path into to_path.
971
:param from_path: The base directory to copy.
972
:param to_path: The target directory. If it does not exist, it will
974
:param handlers: A dictionary of functions, which takes a source and
975
destinations for files, directories, etc.
976
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
977
'file', 'directory', and 'symlink' should always exist.
978
If they are missing, they will be replaced with 'os.mkdir()',
979
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
981
# Now, just copy the existing cached tree to the new location
982
# We use a cheap trick here.
983
# Absolute paths are prefixed with the first parameter
984
# relative paths are prefixed with the second.
985
# So we can get both the source and target returned
986
# without any extra work.
988
def copy_dir(source, dest):
991
def copy_link(source, dest):
992
"""Copy the contents of a symlink"""
993
link_to = os.readlink(source)
994
os.symlink(link_to, dest)
996
real_handlers = {'file':shutil.copy2,
998
'directory':copy_dir,
1000
real_handlers.update(handlers)
1002
if not os.path.exists(to_path):
1003
real_handlers['directory'](from_path, to_path)
1005
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1006
for relpath, name, kind, st, abspath in entries:
1007
real_handlers[kind](abspath, relpath)
1010
def path_prefix_key(path):
1011
"""Generate a prefix-order path key for path.
1013
This can be used to sort paths in the same way that walkdirs does.
1015
return (dirname(path) , path)
1018
def compare_paths_prefix_order(path_a, path_b):
1019
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1020
key_a = path_prefix_key(path_a)
1021
key_b = path_prefix_key(path_b)
1022
return cmp(key_a, key_b)
1025
_cached_user_encoding = None
1028
def get_user_encoding():
1029
"""Find out what the preferred user encoding is.
1031
This is generally the encoding that is used for command line parameters
1032
and file contents. This may be different from the terminal encoding
1033
or the filesystem encoding.
1035
:return: A string defining the preferred user encoding
1037
global _cached_user_encoding
1038
if _cached_user_encoding is not None:
1039
return _cached_user_encoding
1041
if sys.platform == 'darwin':
1042
# work around egregious python 2.4 bug
1043
sys.platform = 'posix'
1047
sys.platform = 'darwin'
1052
_cached_user_encoding = locale.getpreferredencoding()
1053
except locale.Error, e:
1054
sys.stderr.write('bzr: warning: %s\n'
1055
' Could not determine what text encoding to use.\n'
1056
' This error usually means your Python interpreter\n'
1057
' doesn\'t support the locale set by $LANG (%s)\n'
1058
" Continuing with ascii encoding.\n"
1059
% (e, os.environ.get('LANG')))
1061
if _cached_user_encoding is None:
1062
_cached_user_encoding = 'ascii'
1063
return _cached_user_encoding
added
removed
bzrlib/osutils.py
