201
233
# choke on a Unicode string containing a relative path if
202
234
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
204
_fs_enc = sys.getfilesystemencoding()
236
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
205
237
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)
238
# jam 20060426 rather than encoding to fsencoding
239
# copy posixpath.abspath, but use os.getcwdu instead
240
if not posixpath.isabs(path):
241
path = posixpath.join(getcwd(), path)
242
return posixpath.normpath(path)
214
245
def _posix_realpath(path):
215
return os.path.realpath(path.encode(_fs_enc)).decode(_fs_enc)
246
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
249
def _win32_fixdrive(path):
250
"""Force drive letters to be consistent.
252
win32 is inconsistent whether it returns lower or upper case
253
and even if it was consistent the user might type the other
254
so we force it to uppercase
255
running python.exe under cmd.exe return capital C:\\
256
running win32 python inside a cygwin shell returns lowercase c:\\
258
drive, path = _nt_splitdrive(path)
259
return drive.upper() + path
218
262
def _win32_abspath(path):
219
return _nt_abspath(path.encode(_fs_enc)).decode(_fs_enc).replace('\\', '/')
263
# Real _nt_abspath doesn't have a problem with a unicode cwd
264
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
267
def _win98_abspath(path):
268
"""Return the absolute version of a path.
269
Windows 98 safe implementation (python reimplementation
270
of Win32 API function GetFullPathNameW)
275
# \\HOST\path => //HOST/path
276
# //HOST/path => //HOST/path
277
# path => C:/cwd/path
280
# check for absolute path
281
drive = _nt_splitdrive(path)[0]
282
if drive == '' and path[:2] not in('//','\\\\'):
284
# we cannot simply os.path.join cwd and path
285
# because os.path.join('C:','/path') produce '/path'
286
# and this is incorrect
287
if path[:1] in ('/','\\'):
288
cwd = _nt_splitdrive(cwd)[0]
290
path = cwd + '\\' + path
291
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
293
if win32utils.winver == 'Windows 98':
294
_win32_abspath = _win98_abspath
222
297
def _win32_realpath(path):
223
return _nt_realpath(path.encode(_fs_enc)).decode(_fs_enc).replace('\\', '/')
298
# Real _nt_realpath doesn't have a problem with a unicode cwd
299
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
226
302
def _win32_pathjoin(*args):
717
885
return unicode_or_utf8_string.decode('utf8')
718
886
except UnicodeDecodeError:
719
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
887
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
890
def safe_utf8(unicode_or_utf8_string):
891
"""Coerce unicode_or_utf8_string to a utf8 string.
893
If it is a str, it is returned.
894
If it is Unicode, it is encoded into a utf-8 string.
896
if isinstance(unicode_or_utf8_string, str):
897
# TODO: jam 20070209 This is overkill, and probably has an impact on
898
# performance if we are dealing with lots of apis that want a
901
# Make sure it is a valid utf-8 string
902
unicode_or_utf8_string.decode('utf-8')
903
except UnicodeDecodeError:
904
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
905
return unicode_or_utf8_string
906
return unicode_or_utf8_string.encode('utf-8')
909
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
910
' Revision id generators should be creating utf8'
914
def safe_revision_id(unicode_or_utf8_string, warn=True):
915
"""Revision ids should now be utf8, but at one point they were unicode.
917
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
919
:param warn: Functions that are sanitizing user data can set warn=False
920
:return: None or a utf8 revision id.
922
if (unicode_or_utf8_string is None
923
or unicode_or_utf8_string.__class__ == str):
924
return unicode_or_utf8_string
926
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
928
return cache_utf8.encode(unicode_or_utf8_string)
931
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
932
' generators should be creating utf8 file ids.')
935
def safe_file_id(unicode_or_utf8_string, warn=True):
936
"""File ids should now be utf8, but at one point they were unicode.
938
This is the same as safe_utf8, except it uses the cached encode functions
939
to save a little bit of performance.
941
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
943
:param warn: Functions that are sanitizing user data can set warn=False
944
:return: None or a utf8 file id.
946
if (unicode_or_utf8_string is None
947
or unicode_or_utf8_string.__class__ == str):
948
return unicode_or_utf8_string
950
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
952
return cache_utf8.encode(unicode_or_utf8_string)
722
955
_platform_normalizes_filenames = False
732
965
return _platform_normalizes_filenames
968
def _accessible_normalized_filename(path):
969
"""Get the unicode normalized path, and if you can access the file.
971
On platforms where the system normalizes filenames (Mac OSX),
972
you can access a file by any path which will normalize correctly.
973
On platforms where the system does not normalize filenames
974
(Windows, Linux), you have to access a file by its exact path.
976
Internally, bzr only supports NFC/NFKC normalization, since that is
977
the standard for XML documents.
979
So return the normalized path, and a flag indicating if the file
980
can be accessed by that path.
983
return unicodedata.normalize('NFKC', unicode(path)), True
986
def _inaccessible_normalized_filename(path):
987
__doc__ = _accessible_normalized_filename.__doc__
989
normalized = unicodedata.normalize('NFKC', unicode(path))
990
return normalized, normalized == path
735
993
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
994
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
996
normalized_filename = _inaccessible_normalized_filename
770
999
def terminal_width():
771
1000
"""Return estimated terminal width."""
772
1001
if sys.platform == 'win32':
773
import bzrlib.win32console
774
return bzrlib.win32console.get_console_size()[0]
1002
return win32utils.get_console_size()[0]
777
1005
import struct, fcntl, termios
816
1077
to exclude some directories, they are then not descended into.
818
1079
The data yielded is of the form:
819
[(relpath, basename, kind, lstat, path_from_top), ...]
1080
((directory-relpath, directory-path-from-top),
1081
[(directory-relpath, basename, kind, lstat, path-from-top), ...]),
1082
- directory-relpath is the relative path of the directory being returned
1083
with respect to top. prefix is prepended to this.
1084
- directory-path-from-root is the path including top for this directory.
1085
It is suitable for use with os functions.
1086
- relpath is the relative path within the subtree being walked.
1087
- basename is the basename of the path
1088
- kind is the kind of the file now. If unknown then the file is not
1089
present within the tree - but it may be recorded as versioned. See
1091
- lstat is the stat data *if* the file was statted.
1092
- planned, not implemented:
1093
path_from_tree_root is the path from the root of the tree.
821
1095
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
822
1096
allows one to walk a subtree but get paths that are relative to a tree
823
1097
rooted higher up.
824
1098
:return: an iterator over the dirs.
1100
#TODO there is a bit of a smell where the results of the directory-
1101
# summary in this, and the path from the root, may not agree
1102
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1103
# potentially confusing output. We should make this more robust - but
1104
# not at a speed cost. RBC 20060731
828
1106
_directory = _directory_kind
830
pending = [(prefix, "", _directory, None, top)]
1107
_listdir = os.listdir
1108
_kind_from_mode = _formats.get
1109
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
833
currentdir = pending.pop()
834
1111
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
837
relroot = currentdir[0] + '/'
840
for name in sorted(_listdir(top)):
841
abspath = top + '/' + name
842
statvalue = lstat(abspath)
843
dirblock.append ((relroot + name, name, file_kind_from_stat_mode(statvalue.st_mode), statvalue, abspath))
845
# push the user specified dirs from dirblock
846
for dir in reversed(dirblock):
847
if dir[2] == _directory:
1112
relroot, _, _, _, top = pending.pop()
1114
relprefix = relroot + u'/'
1117
top_slash = top + u'/'
1120
append = dirblock.append
1121
for name in sorted(_listdir(top)):
1122
abspath = top_slash + name
1123
statvalue = _lstat(abspath)
1124
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1125
append((relprefix + name, name, kind, statvalue, abspath))
1126
yield (relroot, top), dirblock
1128
# push the user specified dirs from dirblock
1129
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1132
def _walkdirs_utf8(top, prefix=""):
1133
"""Yield data about all the directories in a tree.
1135
This yields the same information as walkdirs() only each entry is yielded
1136
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1137
are returned as exact byte-strings.
1139
:return: yields a tuple of (dir_info, [file_info])
1140
dir_info is (utf8_relpath, path-from-top)
1141
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1142
if top is an absolute path, path-from-top is also an absolute path.
1143
path-from-top might be unicode or utf8, but it is the correct path to
1144
pass to os functions to affect the file in question. (such as os.lstat)
1146
fs_encoding = sys.getfilesystemencoding()
1147
if (sys.platform == 'win32' or
1148
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1149
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1151
return _walkdirs_fs_utf8(top, prefix=prefix)
1154
def _walkdirs_fs_utf8(top, prefix=""):
1155
"""See _walkdirs_utf8.
1157
This sub-function is called when we know the filesystem is already in utf8
1158
encoding. So we don't need to transcode filenames.
1161
_directory = _directory_kind
1162
_listdir = os.listdir
1163
_kind_from_mode = _formats.get
1165
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1166
# But we don't actually uses 1-3 in pending, so set them to None
1167
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1169
relroot, _, _, _, top = pending.pop()
1171
relprefix = relroot + '/'
1174
top_slash = top + '/'
1177
append = dirblock.append
1178
for name in sorted(_listdir(top)):
1179
abspath = top_slash + name
1180
statvalue = _lstat(abspath)
1181
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1182
append((relprefix + name, name, kind, statvalue, abspath))
1183
yield (relroot, top), dirblock
1185
# push the user specified dirs from dirblock
1186
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1189
def _walkdirs_unicode_to_utf8(top, prefix=""):
1190
"""See _walkdirs_utf8
1192
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1194
This is currently the fallback code path when the filesystem encoding is
1195
not UTF-8. It may be better to implement an alternative so that we can
1196
safely handle paths that are not properly decodable in the current
1199
_utf8_encode = codecs.getencoder('utf8')
1201
_directory = _directory_kind
1202
_listdir = os.listdir
1203
_kind_from_mode = _formats.get
1205
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1207
relroot, _, _, _, top = pending.pop()
1209
relprefix = relroot + '/'
1212
top_slash = top + u'/'
1215
append = dirblock.append
1216
for name in sorted(_listdir(top)):
1217
name_utf8 = _utf8_encode(name)[0]
1218
abspath = top_slash + name
1219
statvalue = _lstat(abspath)
1220
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1221
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1222
yield (relroot, top), dirblock
1224
# push the user specified dirs from dirblock
1225
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1228
def copy_tree(from_path, to_path, handlers={}):
1229
"""Copy all of the entries in from_path into to_path.
1231
:param from_path: The base directory to copy.
1232
:param to_path: The target directory. If it does not exist, it will
1234
:param handlers: A dictionary of functions, which takes a source and
1235
destinations for files, directories, etc.
1236
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1237
'file', 'directory', and 'symlink' should always exist.
1238
If they are missing, they will be replaced with 'os.mkdir()',
1239
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1241
# Now, just copy the existing cached tree to the new location
1242
# We use a cheap trick here.
1243
# Absolute paths are prefixed with the first parameter
1244
# relative paths are prefixed with the second.
1245
# So we can get both the source and target returned
1246
# without any extra work.
1248
def copy_dir(source, dest):
1251
def copy_link(source, dest):
1252
"""Copy the contents of a symlink"""
1253
link_to = os.readlink(source)
1254
os.symlink(link_to, dest)
1256
real_handlers = {'file':shutil.copy2,
1257
'symlink':copy_link,
1258
'directory':copy_dir,
1260
real_handlers.update(handlers)
1262
if not os.path.exists(to_path):
1263
real_handlers['directory'](from_path, to_path)
1265
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1266
for relpath, name, kind, st, abspath in entries:
1267
real_handlers[kind](abspath, relpath)
1270
def path_prefix_key(path):
1271
"""Generate a prefix-order path key for path.
1273
This can be used to sort paths in the same way that walkdirs does.
1275
return (dirname(path) , path)
1278
def compare_paths_prefix_order(path_a, path_b):
1279
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1280
key_a = path_prefix_key(path_a)
1281
key_b = path_prefix_key(path_b)
1282
return cmp(key_a, key_b)
1285
_cached_user_encoding = None
1288
def get_user_encoding(use_cache=True):
1289
"""Find out what the preferred user encoding is.
1291
This is generally the encoding that is used for command line parameters
1292
and file contents. This may be different from the terminal encoding
1293
or the filesystem encoding.
1295
:param use_cache: Enable cache for detected encoding.
1296
(This parameter is turned on by default,
1297
and required only for selftesting)
1299
:return: A string defining the preferred user encoding
1301
global _cached_user_encoding
1302
if _cached_user_encoding is not None and use_cache:
1303
return _cached_user_encoding
1305
if sys.platform == 'darwin':
1306
# work around egregious python 2.4 bug
1307
sys.platform = 'posix'
1311
sys.platform = 'darwin'
1316
user_encoding = locale.getpreferredencoding()
1317
except locale.Error, e:
1318
sys.stderr.write('bzr: warning: %s\n'
1319
' Could not determine what text encoding to use.\n'
1320
' This error usually means your Python interpreter\n'
1321
' doesn\'t support the locale set by $LANG (%s)\n'
1322
" Continuing with ascii encoding.\n"
1323
% (e, os.environ.get('LANG')))
1324
user_encoding = 'ascii'
1326
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1327
# treat that as ASCII, and not support printing unicode characters to the
1329
if user_encoding in (None, 'cp0'):
1330
user_encoding = 'ascii'
1334
codecs.lookup(user_encoding)
1336
sys.stderr.write('bzr: warning:'
1337
' unknown encoding %s.'
1338
' Continuing with ascii encoding.\n'
1341
user_encoding = 'ascii'
1344
_cached_user_encoding = user_encoding
1346
return user_encoding
1349
def recv_all(socket, bytes):
1350
"""Receive an exact number of bytes.
1352
Regular Socket.recv() may return less than the requested number of bytes,
1353
dependning on what's in the OS buffer. MSG_WAITALL is not available
1354
on all platforms, but this should work everywhere. This will return
1355
less than the requested amount if the remote end closes.
1357
This isn't optimized and is intended mostly for use in testing.
1360
while len(b) < bytes:
1361
new = socket.recv(bytes - len(b))
1367
def dereference_path(path):
1368
"""Determine the real path to a file.
1370
All parent elements are dereferenced. But the file itself is not
1372
:param path: The original path. May be absolute or relative.
1373
:return: the real path *to* the file
1375
parent, base = os.path.split(path)
1376
# The pathjoin for '.' is a workaround for Python bug #1213894.
1377
# (initial path components aren't dereferenced)
1378
return pathjoin(realpath(pathjoin('.', parent)), base)