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)
69
mode = os.lstat(f)[ST_MODE]
121
88
def kind_marker(kind):
122
89
if kind == 'file':
124
elif kind == _directory_kind:
91
elif kind == 'directory':
126
93
elif kind == 'symlink':
129
96
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
100
if hasattr(os, 'lstat'):
106
if e.errno == errno.ENOENT:
109
raise BzrError("lstat/stat of (%r): %r" % (f, e))
354
111
def normalizepath(f):
355
112
if hasattr(os.path, 'realpath'):
359
116
[p,e] = os.path.split(f)
360
117
if e == "" or e == "." or e == "..":
363
return pathjoin(F(p), e)
120
return os.path.join(F(p), e)
366
123
def backup_file(fn):
367
124
"""Copy a file to a backup.
522
259
'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]
525
386
def compare_files(a, b):
526
387
"""Returns true if equal in contents"""
697
545
copyfile(src, dest)
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):
710
548
def has_symlinks():
711
549
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
((directory-relpath, directory-path-from-top),
879
[(relpath, basename, kind, lstat), ...]),
880
- directory-relpath is the relative path of the directory being returned
881
with respect to top. prefix is prepended to this.
882
- directory-path-from-root is the path including top for this directory.
883
It is suitable for use with os functions.
884
- relpath is the relative path within the subtree being walked.
885
- basename is the basename of the path
886
- kind is the kind of the file now. If unknown then the file is not
887
present within the tree - but it may be recorded as versioned. See
889
- lstat is the stat data *if* the file was statted.
890
- planned, not implemented:
891
path_from_tree_root is the path from the root of the tree.
893
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
894
allows one to walk a subtree but get paths that are relative to a tree
896
:return: an iterator over the dirs.
898
#TODO there is a bit of a smell where the results of the directory-
899
# summary in this, and the path from the root, may not agree
900
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
901
# potentially confusing output. We should make this more robust - but
902
# not at a speed cost. RBC 20060731
905
_directory = _directory_kind
907
pending = [(prefix, "", _directory, None, top)]
910
currentdir = pending.pop()
911
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
914
relroot = currentdir[0] + '/'
917
for name in sorted(_listdir(top)):
918
abspath = top + '/' + name
919
statvalue = lstat(abspath)
920
dirblock.append((relroot + name, name,
921
file_kind_from_stat_mode(statvalue.st_mode),
923
yield (currentdir[0], top), dirblock
924
# push the user specified dirs from dirblock
925
for dir in reversed(dirblock):
926
if dir[2] == _directory:
930
def path_prefix_key(path):
931
"""Generate a prefix-order path key for path.
933
This can be used to sort paths in the same way that walkdirs does.
935
return (dirname(path) , path)
938
def compare_paths_prefix_order(path_a, path_b):
939
"""Compare path_a and path_b to generate the same order walkdirs uses."""
940
key_a = path_prefix_key(path_a)
941
key_b = path_prefix_key(path_b)
942
return cmp(key_a, key_b)
added
removed
bzrlib/osutils.py
