1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
from cStringIO import StringIO
21
from stat import (S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE,
22
S_ISCHR, S_ISBLK, S_ISFIFO, S_ISSOCK)
26
from bzrlib.lazy_import import lazy_import
27
lazy_import(globals(), """
29
from datetime import datetime
31
from ntpath import (abspath as _nt_abspath,
33
normpath as _nt_normpath,
34
realpath as _nt_realpath,
35
splitdrive as _nt_splitdrive,
44
from tempfile import (
58
from bzrlib import symbol_versioning
59
from bzrlib.symbol_versioning import (
62
from bzrlib.trace import mutter
65
# On win32, O_BINARY is used to indicate the file should
66
# be opened in binary mode, rather than text mode.
67
# On other platforms, O_BINARY doesn't exist, because
68
# they always open in binary mode, so it is okay to
69
# OR with 0 on those platforms
70
O_BINARY = getattr(os, 'O_BINARY', 0)
73
def make_readonly(filename):
74
"""Make a filename read-only."""
75
mod = os.lstat(filename).st_mode
76
if not stat.S_ISLNK(mod):
78
os.chmod(filename, mod)
81
def make_writable(filename):
82
mod = os.lstat(filename).st_mode
83
if not stat.S_ISLNK(mod):
85
os.chmod(filename, mod)
88
def minimum_path_selection(paths):
89
"""Return the smallset subset of paths which are outside paths.
91
:param paths: A container (and hence not None) of paths.
92
:return: A set of paths sufficient to include everything in paths via
93
is_inside_any, drawn from the paths parameter.
98
other_paths = paths.difference([path])
99
if not is_inside_any(other_paths, path):
100
# this is a top level path, we must check it.
101
search_paths.add(path)
109
"""Return a quoted filename filename
111
This previously used backslash quoting, but that works poorly on
113
# TODO: I'm not really sure this is the best format either.x
115
if _QUOTE_RE is None:
116
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
118
if _QUOTE_RE.search(f):
124
_directory_kind = 'directory'
127
"""Return the current umask"""
128
# Assume that people aren't messing with the umask while running
129
# XXX: This is not thread safe, but there is no way to get the
130
# umask without setting it
138
_directory_kind: "/",
140
'tree-reference': '+',
144
def kind_marker(kind):
146
return _kind_marker_map[kind]
148
raise errors.BzrError('invalid file kind %r' % kind)
151
lexists = getattr(os.path, 'lexists', None)
155
stat = getattr(os, 'lstat', os.stat)
159
if e.errno == errno.ENOENT:
162
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
165
def fancy_rename(old, new, rename_func, unlink_func):
166
"""A fancy rename, when you don't have atomic rename.
168
:param old: The old path, to rename from
169
:param new: The new path, to rename to
170
:param rename_func: The potentially non-atomic rename function
171
:param unlink_func: A way to delete the target file if the full rename succeeds
174
# sftp rename doesn't allow overwriting, so play tricks:
176
base = os.path.basename(new)
177
dirname = os.path.dirname(new)
178
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
179
tmp_name = pathjoin(dirname, tmp_name)
181
# Rename the file out of the way, but keep track if it didn't exist
182
# We don't want to grab just any exception
183
# something like EACCES should prevent us from continuing
184
# The downside is that the rename_func has to throw an exception
185
# with an errno = ENOENT, or NoSuchFile
188
rename_func(new, tmp_name)
189
except (errors.NoSuchFile,), e:
192
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
193
# function raises an IOError with errno is None when a rename fails.
194
# This then gets caught here.
195
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
198
if (getattr(e, 'errno', None) is None
199
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
207
# This may throw an exception, in which case success will
209
rename_func(old, new)
211
except (IOError, OSError), e:
212
# source and target may be aliases of each other (e.g. on a
213
# case-insensitive filesystem), so we may have accidentally renamed
214
# source by when we tried to rename target
215
if not (file_existed and e.errno in (None, errno.ENOENT)):
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 _win98_abspath(path):
263
"""Return the absolute version of a path.
264
Windows 98 safe implementation (python reimplementation
265
of Win32 API function GetFullPathNameW)
270
# \\HOST\path => //HOST/path
271
# //HOST/path => //HOST/path
272
# path => C:/cwd/path
275
# check for absolute path
276
drive = _nt_splitdrive(path)[0]
277
if drive == '' and path[:2] not in('//','\\\\'):
279
# we cannot simply os.path.join cwd and path
280
# because os.path.join('C:','/path') produce '/path'
281
# and this is incorrect
282
if path[:1] in ('/','\\'):
283
cwd = _nt_splitdrive(cwd)[0]
285
path = cwd + '\\' + path
286
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
288
if win32utils.winver == 'Windows 98':
289
_win32_abspath = _win98_abspath
292
def _win32_realpath(path):
293
# Real _nt_realpath doesn't have a problem with a unicode cwd
294
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
297
def _win32_pathjoin(*args):
298
return _nt_join(*args).replace('\\', '/')
301
def _win32_normpath(path):
302
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
306
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
309
def _win32_mkdtemp(*args, **kwargs):
310
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
313
def _win32_rename(old, new):
314
"""We expect to be able to atomically replace 'new' with old.
316
On win32, if new exists, it must be moved out of the way first,
320
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
322
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
323
# If we try to rename a non-existant file onto cwd, we get
324
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
325
# if the old path doesn't exist, sometimes we get EACCES
326
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
332
return unicodedata.normalize('NFC', os.getcwdu())
335
# Default is to just use the python builtins, but these can be rebound on
336
# particular platforms.
337
abspath = _posix_abspath
338
realpath = _posix_realpath
339
pathjoin = os.path.join
340
normpath = os.path.normpath
343
dirname = os.path.dirname
344
basename = os.path.basename
345
split = os.path.split
346
splitext = os.path.splitext
347
# These were already imported into local scope
348
# mkdtemp = tempfile.mkdtemp
349
# rmtree = shutil.rmtree
351
MIN_ABS_PATHLENGTH = 1
354
if sys.platform == 'win32':
355
abspath = _win32_abspath
356
realpath = _win32_realpath
357
pathjoin = _win32_pathjoin
358
normpath = _win32_normpath
359
getcwd = _win32_getcwd
360
mkdtemp = _win32_mkdtemp
361
rename = _win32_rename
363
MIN_ABS_PATHLENGTH = 3
365
def _win32_delete_readonly(function, path, excinfo):
366
"""Error handler for shutil.rmtree function [for win32]
367
Helps to remove files and dirs marked as read-only.
369
exception = excinfo[1]
370
if function in (os.remove, os.rmdir) \
371
and isinstance(exception, OSError) \
372
and exception.errno == errno.EACCES:
378
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
379
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
380
return shutil.rmtree(path, ignore_errors, onerror)
381
elif sys.platform == 'darwin':
385
def get_terminal_encoding():
386
"""Find the best encoding for printing to the screen.
388
This attempts to check both sys.stdout and sys.stdin to see
389
what encoding they are in, and if that fails it falls back to
390
bzrlib.user_encoding.
391
The problem is that on Windows, locale.getpreferredencoding()
392
is not the same encoding as that used by the console:
393
http://mail.python.org/pipermail/python-list/2003-May/162357.html
395
On my standard US Windows XP, the preferred encoding is
396
cp1252, but the console is cp437
398
output_encoding = getattr(sys.stdout, 'encoding', None)
399
if not output_encoding:
400
input_encoding = getattr(sys.stdin, 'encoding', None)
401
if not input_encoding:
402
output_encoding = bzrlib.user_encoding
403
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
405
output_encoding = input_encoding
406
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
408
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
409
if output_encoding == 'cp0':
410
# invalid encoding (cp0 means 'no codepage' on Windows)
411
output_encoding = bzrlib.user_encoding
412
mutter('cp0 is invalid encoding.'
413
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
416
codecs.lookup(output_encoding)
418
sys.stderr.write('bzr: warning:'
419
' unknown terminal encoding %s.\n'
420
' Using encoding %s instead.\n'
421
% (output_encoding, bzrlib.user_encoding)
423
output_encoding = bzrlib.user_encoding
425
return output_encoding
428
def normalizepath(f):
429
if getattr(os.path, 'realpath', None) is not None:
433
[p,e] = os.path.split(f)
434
if e == "" or e == "." or e == "..":
437
return pathjoin(F(p), e)
441
"""True if f is an accessible directory."""
443
return S_ISDIR(os.lstat(f)[ST_MODE])
449
"""True if f is a regular file."""
451
return S_ISREG(os.lstat(f)[ST_MODE])
456
"""True if f is a symlink."""
458
return S_ISLNK(os.lstat(f)[ST_MODE])
462
def is_inside(dir, fname):
463
"""True if fname is inside dir.
465
The parameters should typically be passed to osutils.normpath first, so
466
that . and .. and repeated slashes are eliminated, and the separators
467
are canonical for the platform.
469
The empty string as a dir name is taken as top-of-tree and matches
472
# XXX: Most callers of this can actually do something smarter by
473
# looking at the inventory
483
return fname.startswith(dir)
486
def is_inside_any(dir_list, fname):
487
"""True if fname is inside any of given dirs."""
488
for dirname in dir_list:
489
if is_inside(dirname, fname):
494
def is_inside_or_parent_of_any(dir_list, fname):
495
"""True if fname is a child or a parent of any of the given files."""
496
for dirname in dir_list:
497
if is_inside(dirname, fname) or is_inside(fname, dirname):
502
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768):
503
"""Copy contents of one file to another.
505
The read_length can either be -1 to read to end-of-file (EOF) or
506
it can specify the maximum number of bytes to read.
508
The buff_size represents the maximum size for each read operation
509
performed on from_file.
511
:return: The number of bytes copied.
515
# read specified number of bytes
517
while read_length > 0:
518
num_bytes_to_read = min(read_length, buff_size)
520
block = from_file.read(num_bytes_to_read)
526
actual_bytes_read = len(block)
527
read_length -= actual_bytes_read
528
length += actual_bytes_read
532
block = from_file.read(buff_size)
541
def pump_string_file(bytes, file_handle, segment_size=None):
542
"""Write bytes to file_handle in many smaller writes.
544
:param bytes: The string to write.
545
:param file_handle: The file to write to.
547
# Write data in chunks rather than all at once, because very large
548
# writes fail on some platforms (e.g. Windows with SMB mounted
551
segment_size = 5242880 # 5MB
552
segments = range(len(bytes) / segment_size + 1)
553
write = file_handle.write
554
for segment_index in segments:
555
segment = buffer(bytes, segment_index * segment_size, segment_size)
559
def file_iterator(input_file, readsize=32768):
561
b = input_file.read(readsize)
568
"""Calculate the hexdigest of an open file.
570
The file cursor should be already at the start.
582
def sha_file_by_name(fname):
583
"""Calculate the SHA1 of a file by reading the full text"""
585
f = os.open(fname, os.O_RDONLY | O_BINARY)
588
b = os.read(f, 1<<16)
596
def sha_strings(strings, _factory=sha.new):
597
"""Return the sha-1 of concatenation of strings"""
599
map(s.update, strings)
603
def sha_string(f, _factory=sha.new):
604
return _factory(f).hexdigest()
607
def fingerprint_file(f):
609
return {'size': len(b),
610
'sha1': sha.new(b).hexdigest()}
613
def compare_files(a, b):
614
"""Returns true if equal in contents"""
625
def local_time_offset(t=None):
626
"""Return offset of local zone from GMT, either at present or at time t."""
629
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
630
return offset.days * 86400 + offset.seconds
632
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
634
def format_date(t, offset=0, timezone='original', date_fmt=None,
636
"""Return a formatted date string.
638
:param t: Seconds since the epoch.
639
:param offset: Timezone offset in seconds east of utc.
640
:param timezone: How to display the time: 'utc', 'original' for the
641
timezone specified by offset, or 'local' for the process's current
643
:param show_offset: Whether to append the timezone.
644
:param date_fmt: strftime format.
646
if timezone == 'utc':
649
elif timezone == 'original':
652
tt = time.gmtime(t + offset)
653
elif timezone == 'local':
654
tt = time.localtime(t)
655
offset = local_time_offset(t)
657
raise errors.UnsupportedTimezoneFormat(timezone)
659
date_fmt = "%a %Y-%m-%d %H:%M:%S"
661
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
664
# day of week depends on locale, so we do this ourself
665
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
666
return (time.strftime(date_fmt, tt) + offset_str)
669
def compact_date(when):
670
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
673
def format_delta(delta):
674
"""Get a nice looking string for a time delta.
676
:param delta: The time difference in seconds, can be positive or negative.
677
positive indicates time in the past, negative indicates time in the
678
future. (usually time.time() - stored_time)
679
:return: String formatted to show approximate resolution
685
direction = 'in the future'
689
if seconds < 90: # print seconds up to 90 seconds
691
return '%d second %s' % (seconds, direction,)
693
return '%d seconds %s' % (seconds, direction)
695
minutes = int(seconds / 60)
696
seconds -= 60 * minutes
701
if minutes < 90: # print minutes, seconds up to 90 minutes
703
return '%d minute, %d second%s %s' % (
704
minutes, seconds, plural_seconds, direction)
706
return '%d minutes, %d second%s %s' % (
707
minutes, seconds, plural_seconds, direction)
709
hours = int(minutes / 60)
710
minutes -= 60 * hours
717
return '%d hour, %d minute%s %s' % (hours, minutes,
718
plural_minutes, direction)
719
return '%d hours, %d minute%s %s' % (hours, minutes,
720
plural_minutes, direction)
723
"""Return size of given open file."""
724
return os.fstat(f.fileno())[ST_SIZE]
727
# Define rand_bytes based on platform.
729
# Python 2.4 and later have os.urandom,
730
# but it doesn't work on some arches
732
rand_bytes = os.urandom
733
except (NotImplementedError, AttributeError):
734
# If python doesn't have os.urandom, or it doesn't work,
735
# then try to first pull random data from /dev/urandom
737
rand_bytes = file('/dev/urandom', 'rb').read
738
# Otherwise, use this hack as a last resort
739
except (IOError, OSError):
740
# not well seeded, but better than nothing
745
s += chr(random.randint(0, 255))
750
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
752
"""Return a random string of num alphanumeric characters
754
The result only contains lowercase chars because it may be used on
755
case-insensitive filesystems.
758
for raw_byte in rand_bytes(num):
759
s += ALNUM[ord(raw_byte) % 36]
763
## TODO: We could later have path objects that remember their list
764
## decomposition (might be too tricksy though.)
767
"""Turn string into list of parts."""
768
# split on either delimiter because people might use either on
770
ps = re.split(r'[\\/]', p)
775
raise errors.BzrError("sorry, %r not allowed in path" % f)
776
elif (f == '.') or (f == ''):
784
if (f == '..') or (f is None) or (f == ''):
785
raise errors.BzrError("sorry, %r not allowed in path" % f)
790
"""Split s into lines, but without removing the newline characters."""
791
lines = s.split('\n')
792
result = [line + '\n' for line in lines[:-1]]
794
result.append(lines[-1])
798
def hardlinks_good():
799
return sys.platform not in ('win32', 'cygwin', 'darwin')
802
def link_or_copy(src, dest):
803
"""Hardlink a file, or copy it if it can't be hardlinked."""
804
if not hardlinks_good():
805
shutil.copyfile(src, dest)
809
except (OSError, IOError), e:
810
if e.errno != errno.EXDEV:
812
shutil.copyfile(src, dest)
815
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
816
# Forgiveness than Permission (EAFP) because:
817
# - root can damage a solaris file system by using unlink,
818
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
819
# EACCES, OSX: EPERM) when invoked on a directory.
820
def delete_any(path):
821
"""Delete a file or directory."""
822
if isdir(path): # Takes care of symlinks
829
if getattr(os, 'symlink', None) is not None:
836
if getattr(os, 'link', None) is not None:
842
def host_os_dereferences_symlinks():
843
return (has_symlinks()
844
and sys.platform not in ('cygwin', 'win32'))
847
def contains_whitespace(s):
848
"""True if there are any whitespace characters in s."""
849
# string.whitespace can include '\xa0' in certain locales, because it is
850
# considered "non-breaking-space" as part of ISO-8859-1. But it
851
# 1) Isn't a breaking whitespace
852
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
854
# 3) '\xa0' isn't unicode safe since it is >128.
856
# This should *not* be a unicode set of characters in case the source
857
# string is not a Unicode string. We can auto-up-cast the characters since
858
# they are ascii, but we don't want to auto-up-cast the string in case it
860
for ch in ' \t\n\r\v\f':
867
def contains_linebreaks(s):
868
"""True if there is any vertical whitespace in s."""
876
def relpath(base, path):
877
"""Return path relative to base, or raise exception.
879
The path may be either an absolute path or a path relative to the
880
current working directory.
882
os.path.commonprefix (python2.4) has a bad bug that it works just
883
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
887
if len(base) < MIN_ABS_PATHLENGTH:
888
# must have space for e.g. a drive letter
889
raise ValueError('%r is too short to calculate a relative path'
896
while len(head) >= len(base):
899
head, tail = os.path.split(head)
903
raise errors.PathNotChild(rp, base)
911
def safe_unicode(unicode_or_utf8_string):
912
"""Coerce unicode_or_utf8_string into unicode.
914
If it is unicode, it is returned.
915
Otherwise it is decoded from utf-8. If a decoding error
916
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
917
as a BzrBadParameter exception.
919
if isinstance(unicode_or_utf8_string, unicode):
920
return unicode_or_utf8_string
922
return unicode_or_utf8_string.decode('utf8')
923
except UnicodeDecodeError:
924
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
927
def safe_utf8(unicode_or_utf8_string):
928
"""Coerce unicode_or_utf8_string to a utf8 string.
930
If it is a str, it is returned.
931
If it is Unicode, it is encoded into a utf-8 string.
933
if isinstance(unicode_or_utf8_string, str):
934
# TODO: jam 20070209 This is overkill, and probably has an impact on
935
# performance if we are dealing with lots of apis that want a
938
# Make sure it is a valid utf-8 string
939
unicode_or_utf8_string.decode('utf-8')
940
except UnicodeDecodeError:
941
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
942
return unicode_or_utf8_string
943
return unicode_or_utf8_string.encode('utf-8')
946
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
947
' Revision id generators should be creating utf8'
951
def safe_revision_id(unicode_or_utf8_string, warn=True):
952
"""Revision ids should now be utf8, but at one point they were unicode.
954
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
956
:param warn: Functions that are sanitizing user data can set warn=False
957
:return: None or a utf8 revision id.
959
if (unicode_or_utf8_string is None
960
or unicode_or_utf8_string.__class__ == str):
961
return unicode_or_utf8_string
963
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
965
return cache_utf8.encode(unicode_or_utf8_string)
968
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
969
' generators should be creating utf8 file ids.')
972
def safe_file_id(unicode_or_utf8_string, warn=True):
973
"""File ids should now be utf8, but at one point they were unicode.
975
This is the same as safe_utf8, except it uses the cached encode functions
976
to save a little bit of performance.
978
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
980
:param warn: Functions that are sanitizing user data can set warn=False
981
:return: None or a utf8 file id.
983
if (unicode_or_utf8_string is None
984
or unicode_or_utf8_string.__class__ == str):
985
return unicode_or_utf8_string
987
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
989
return cache_utf8.encode(unicode_or_utf8_string)
992
_platform_normalizes_filenames = False
993
if sys.platform == 'darwin':
994
_platform_normalizes_filenames = True
997
def normalizes_filenames():
998
"""Return True if this platform normalizes unicode filenames.
1000
Mac OSX does, Windows/Linux do not.
1002
return _platform_normalizes_filenames
1005
def _accessible_normalized_filename(path):
1006
"""Get the unicode normalized path, and if you can access the file.
1008
On platforms where the system normalizes filenames (Mac OSX),
1009
you can access a file by any path which will normalize correctly.
1010
On platforms where the system does not normalize filenames
1011
(Windows, Linux), you have to access a file by its exact path.
1013
Internally, bzr only supports NFC normalization, since that is
1014
the standard for XML documents.
1016
So return the normalized path, and a flag indicating if the file
1017
can be accessed by that path.
1020
return unicodedata.normalize('NFC', unicode(path)), True
1023
def _inaccessible_normalized_filename(path):
1024
__doc__ = _accessible_normalized_filename.__doc__
1026
normalized = unicodedata.normalize('NFC', unicode(path))
1027
return normalized, normalized == path
1030
if _platform_normalizes_filenames:
1031
normalized_filename = _accessible_normalized_filename
1033
normalized_filename = _inaccessible_normalized_filename
1036
def terminal_width():
1037
"""Return estimated terminal width."""
1038
if sys.platform == 'win32':
1039
return win32utils.get_console_size()[0]
1042
import struct, fcntl, termios
1043
s = struct.pack('HHHH', 0, 0, 0, 0)
1044
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1045
width = struct.unpack('HHHH', x)[1]
1050
width = int(os.environ['COLUMNS'])
1059
def supports_executable():
1060
return sys.platform != "win32"
1063
def supports_posix_readonly():
1064
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1066
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1067
directory controls creation/deletion, etc.
1069
And under win32, readonly means that the directory itself cannot be
1070
deleted. The contents of a readonly directory can be changed, unlike POSIX
1071
where files in readonly directories cannot be added, deleted or renamed.
1073
return sys.platform != "win32"
1076
def set_or_unset_env(env_variable, value):
1077
"""Modify the environment, setting or removing the env_variable.
1079
:param env_variable: The environment variable in question
1080
:param value: The value to set the environment to. If None, then
1081
the variable will be removed.
1082
:return: The original value of the environment variable.
1084
orig_val = os.environ.get(env_variable)
1086
if orig_val is not None:
1087
del os.environ[env_variable]
1089
if isinstance(value, unicode):
1090
value = value.encode(bzrlib.user_encoding)
1091
os.environ[env_variable] = value
1095
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1098
def check_legal_path(path):
1099
"""Check whether the supplied path is legal.
1100
This is only required on Windows, so we don't test on other platforms
1103
if sys.platform != "win32":
1105
if _validWin32PathRE.match(path) is None:
1106
raise errors.IllegalPath(path)
1109
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1111
def _is_error_enotdir(e):
1112
"""Check if this exception represents ENOTDIR.
1114
Unfortunately, python is very inconsistent about the exception
1115
here. The cases are:
1116
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1117
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1118
which is the windows error code.
1119
3) Windows, Python2.5 uses errno == EINVAL and
1120
winerror == ERROR_DIRECTORY
1122
:param e: An Exception object (expected to be OSError with an errno
1123
attribute, but we should be able to cope with anything)
1124
:return: True if this represents an ENOTDIR error. False otherwise.
1126
en = getattr(e, 'errno', None)
1127
if (en == errno.ENOTDIR
1128
or (sys.platform == 'win32'
1129
and (en == _WIN32_ERROR_DIRECTORY
1130
or (en == errno.EINVAL
1131
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1137
def walkdirs(top, prefix=""):
1138
"""Yield data about all the directories in a tree.
1140
This yields all the data about the contents of a directory at a time.
1141
After each directory has been yielded, if the caller has mutated the list
1142
to exclude some directories, they are then not descended into.
1144
The data yielded is of the form:
1145
((directory-relpath, directory-path-from-top),
1146
[(relpath, basename, kind, lstat, path-from-top), ...]),
1147
- directory-relpath is the relative path of the directory being returned
1148
with respect to top. prefix is prepended to this.
1149
- directory-path-from-root is the path including top for this directory.
1150
It is suitable for use with os functions.
1151
- relpath is the relative path within the subtree being walked.
1152
- basename is the basename of the path
1153
- kind is the kind of the file now. If unknown then the file is not
1154
present within the tree - but it may be recorded as versioned. See
1156
- lstat is the stat data *if* the file was statted.
1157
- planned, not implemented:
1158
path_from_tree_root is the path from the root of the tree.
1160
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1161
allows one to walk a subtree but get paths that are relative to a tree
1163
:return: an iterator over the dirs.
1165
#TODO there is a bit of a smell where the results of the directory-
1166
# summary in this, and the path from the root, may not agree
1167
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1168
# potentially confusing output. We should make this more robust - but
1169
# not at a speed cost. RBC 20060731
1171
_directory = _directory_kind
1172
_listdir = os.listdir
1173
_kind_from_mode = file_kind_from_stat_mode
1174
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1176
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1177
relroot, _, _, _, top = pending.pop()
1179
relprefix = relroot + u'/'
1182
top_slash = top + u'/'
1185
append = dirblock.append
1187
names = sorted(_listdir(top))
1189
if not _is_error_enotdir(e):
1193
abspath = top_slash + name
1194
statvalue = _lstat(abspath)
1195
kind = _kind_from_mode(statvalue.st_mode)
1196
append((relprefix + name, name, kind, statvalue, abspath))
1197
yield (relroot, top), dirblock
1199
# push the user specified dirs from dirblock
1200
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1203
class DirReader(object):
1204
"""An interface for reading directories."""
1206
def top_prefix_to_starting_dir(self, top, prefix=""):
1207
"""Converts top and prefix to a starting dir entry
1209
:param top: A utf8 path
1210
:param prefix: An optional utf8 path to prefix output relative paths
1212
:return: A tuple starting with prefix, and ending with the native
1215
raise NotImplementedError(self.top_prefix_to_starting_dir)
1217
def read_dir(self, prefix, top):
1218
"""Read a specific dir.
1220
:param prefix: A utf8 prefix to be preprended to the path basenames.
1221
:param top: A natively encoded path to read.
1222
:return: A list of the directories contents. Each item contains:
1223
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1225
raise NotImplementedError(self.read_dir)
1228
_selected_dir_reader = None
1231
def _walkdirs_utf8(top, prefix=""):
1232
"""Yield data about all the directories in a tree.
1234
This yields the same information as walkdirs() only each entry is yielded
1235
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1236
are returned as exact byte-strings.
1238
:return: yields a tuple of (dir_info, [file_info])
1239
dir_info is (utf8_relpath, path-from-top)
1240
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1241
if top is an absolute path, path-from-top is also an absolute path.
1242
path-from-top might be unicode or utf8, but it is the correct path to
1243
pass to os functions to affect the file in question. (such as os.lstat)
1245
global _selected_dir_reader
1246
if _selected_dir_reader is None:
1247
fs_encoding = _fs_enc.upper()
1248
if win32utils.winver == 'Windows NT':
1249
# Win98 doesn't have unicode apis like FindFirstFileW
1250
# TODO: We possibly could support Win98 by falling back to the
1251
# original FindFirstFile, and using TCHAR instead of WCHAR,
1252
# but that gets a bit tricky, and requires custom compiling
1255
from bzrlib._walkdirs_win32 import Win32ReadDir
1257
_selected_dir_reader = UnicodeDirReader()
1259
_selected_dir_reader = Win32ReadDir()
1260
elif fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1261
# ANSI_X3.4-1968 is a form of ASCII
1262
_selected_dir_reader = UnicodeDirReader()
1265
from bzrlib._readdir_pyx import UTF8DirReader
1267
# No optimised code path
1268
_selected_dir_reader = UnicodeDirReader()
1270
_selected_dir_reader = UTF8DirReader()
1271
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1272
# But we don't actually uses 1-3 in pending, so set them to None
1273
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1274
read_dir = _selected_dir_reader.read_dir
1275
_directory = _directory_kind
1277
relroot, _, _, _, top = pending[-1].pop()
1280
dirblock = sorted(read_dir(relroot, top))
1281
yield (relroot, top), dirblock
1282
# push the user specified dirs from dirblock
1283
next = [d for d in reversed(dirblock) if d[2] == _directory]
1285
pending.append(next)
1288
class UnicodeDirReader(DirReader):
1289
"""A dir reader for non-utf8 file systems, which transcodes."""
1291
__slots__ = ['_utf8_encode']
1294
self._utf8_encode = codecs.getencoder('utf8')
1296
def top_prefix_to_starting_dir(self, top, prefix=""):
1297
"""See DirReader.top_prefix_to_starting_dir."""
1298
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1300
def read_dir(self, prefix, top):
1301
"""Read a single directory from a non-utf8 file system.
1303
top, and the abspath element in the output are unicode, all other paths
1304
are utf8. Local disk IO is done via unicode calls to listdir etc.
1306
This is currently the fallback code path when the filesystem encoding is
1307
not UTF-8. It may be better to implement an alternative so that we can
1308
safely handle paths that are not properly decodable in the current
1311
See DirReader.read_dir for details.
1313
_utf8_encode = self._utf8_encode
1315
_listdir = os.listdir
1316
_kind_from_mode = file_kind_from_stat_mode
1319
relprefix = prefix + '/'
1322
top_slash = top + u'/'
1325
append = dirblock.append
1326
for name in sorted(_listdir(top)):
1328
name_utf8 = _utf8_encode(name)[0]
1329
except UnicodeDecodeError:
1330
raise errors.BadFilenameEncoding(
1331
_utf8_encode(relprefix)[0] + name, _fs_enc)
1332
abspath = top_slash + name
1333
statvalue = _lstat(abspath)
1334
kind = _kind_from_mode(statvalue.st_mode)
1335
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1339
def copy_tree(from_path, to_path, handlers={}):
1340
"""Copy all of the entries in from_path into to_path.
1342
:param from_path: The base directory to copy.
1343
:param to_path: The target directory. If it does not exist, it will
1345
:param handlers: A dictionary of functions, which takes a source and
1346
destinations for files, directories, etc.
1347
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1348
'file', 'directory', and 'symlink' should always exist.
1349
If they are missing, they will be replaced with 'os.mkdir()',
1350
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1352
# Now, just copy the existing cached tree to the new location
1353
# We use a cheap trick here.
1354
# Absolute paths are prefixed with the first parameter
1355
# relative paths are prefixed with the second.
1356
# So we can get both the source and target returned
1357
# without any extra work.
1359
def copy_dir(source, dest):
1362
def copy_link(source, dest):
1363
"""Copy the contents of a symlink"""
1364
link_to = os.readlink(source)
1365
os.symlink(link_to, dest)
1367
real_handlers = {'file':shutil.copy2,
1368
'symlink':copy_link,
1369
'directory':copy_dir,
1371
real_handlers.update(handlers)
1373
if not os.path.exists(to_path):
1374
real_handlers['directory'](from_path, to_path)
1376
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1377
for relpath, name, kind, st, abspath in entries:
1378
real_handlers[kind](abspath, relpath)
1381
def path_prefix_key(path):
1382
"""Generate a prefix-order path key for path.
1384
This can be used to sort paths in the same way that walkdirs does.
1386
return (dirname(path) , path)
1389
def compare_paths_prefix_order(path_a, path_b):
1390
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1391
key_a = path_prefix_key(path_a)
1392
key_b = path_prefix_key(path_b)
1393
return cmp(key_a, key_b)
1396
_cached_user_encoding = None
1399
def get_user_encoding(use_cache=True):
1400
"""Find out what the preferred user encoding is.
1402
This is generally the encoding that is used for command line parameters
1403
and file contents. This may be different from the terminal encoding
1404
or the filesystem encoding.
1406
:param use_cache: Enable cache for detected encoding.
1407
(This parameter is turned on by default,
1408
and required only for selftesting)
1410
:return: A string defining the preferred user encoding
1412
global _cached_user_encoding
1413
if _cached_user_encoding is not None and use_cache:
1414
return _cached_user_encoding
1416
if sys.platform == 'darwin':
1417
# python locale.getpreferredencoding() always return
1418
# 'mac-roman' on darwin. That's a lie.
1419
sys.platform = 'posix'
1421
if os.environ.get('LANG', None) is None:
1422
# If LANG is not set, we end up with 'ascii', which is bad
1423
# ('mac-roman' is more than ascii), so we set a default which
1424
# will give us UTF-8 (which appears to work in all cases on
1425
# OSX). Users are still free to override LANG of course, as
1426
# long as it give us something meaningful. This work-around
1427
# *may* not be needed with python 3k and/or OSX 10.5, but will
1428
# work with them too -- vila 20080908
1429
os.environ['LANG'] = 'en_US.UTF-8'
1432
sys.platform = 'darwin'
1437
user_encoding = locale.getpreferredencoding()
1438
except locale.Error, e:
1439
sys.stderr.write('bzr: warning: %s\n'
1440
' Could not determine what text encoding to use.\n'
1441
' This error usually means your Python interpreter\n'
1442
' doesn\'t support the locale set by $LANG (%s)\n'
1443
" Continuing with ascii encoding.\n"
1444
% (e, os.environ.get('LANG')))
1445
user_encoding = 'ascii'
1447
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1448
# treat that as ASCII, and not support printing unicode characters to the
1451
# For python scripts run under vim, we get '', so also treat that as ASCII
1452
if user_encoding in (None, 'cp0', ''):
1453
user_encoding = 'ascii'
1457
codecs.lookup(user_encoding)
1459
sys.stderr.write('bzr: warning:'
1460
' unknown encoding %s.'
1461
' Continuing with ascii encoding.\n'
1464
user_encoding = 'ascii'
1467
_cached_user_encoding = user_encoding
1469
return user_encoding
1472
def get_host_name():
1473
"""Return the current unicode host name.
1475
This is meant to be used in place of socket.gethostname() because that
1476
behaves inconsistently on different platforms.
1478
if sys.platform == "win32":
1480
return win32utils.get_host_name()
1483
return socket.gethostname().decode(get_user_encoding())
1486
def recv_all(socket, bytes):
1487
"""Receive an exact number of bytes.
1489
Regular Socket.recv() may return less than the requested number of bytes,
1490
dependning on what's in the OS buffer. MSG_WAITALL is not available
1491
on all platforms, but this should work everywhere. This will return
1492
less than the requested amount if the remote end closes.
1494
This isn't optimized and is intended mostly for use in testing.
1497
while len(b) < bytes:
1498
new = socket.recv(bytes - len(b))
1505
def send_all(socket, bytes):
1506
"""Send all bytes on a socket.
1508
Regular socket.sendall() can give socket error 10053 on Windows. This
1509
implementation sends no more than 64k at a time, which avoids this problem.
1512
for pos in xrange(0, len(bytes), chunk_size):
1513
socket.sendall(bytes[pos:pos+chunk_size])
1516
def dereference_path(path):
1517
"""Determine the real path to a file.
1519
All parent elements are dereferenced. But the file itself is not
1521
:param path: The original path. May be absolute or relative.
1522
:return: the real path *to* the file
1524
parent, base = os.path.split(path)
1525
# The pathjoin for '.' is a workaround for Python bug #1213894.
1526
# (initial path components aren't dereferenced)
1527
return pathjoin(realpath(pathjoin('.', parent)), base)
1530
def supports_mapi():
1531
"""Return True if we can use MAPI to launch a mail client."""
1532
return sys.platform == "win32"
1535
def resource_string(package, resource_name):
1536
"""Load a resource from a package and return it as a string.
1538
Note: Only packages that start with bzrlib are currently supported.
1540
This is designed to be a lightweight implementation of resource
1541
loading in a way which is API compatible with the same API from
1543
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1544
If and when pkg_resources becomes a standard library, this routine
1547
# Check package name is within bzrlib
1548
if package == "bzrlib":
1549
resource_relpath = resource_name
1550
elif package.startswith("bzrlib."):
1551
package = package[len("bzrlib."):].replace('.', os.sep)
1552
resource_relpath = pathjoin(package, resource_name)
1554
raise errors.BzrError('resource package %s not in bzrlib' % package)
1556
# Map the resource to a file and read its contents
1557
base = dirname(bzrlib.__file__)
1558
if getattr(sys, 'frozen', None): # bzr.exe
1559
base = abspath(pathjoin(base, '..', '..'))
1560
filename = pathjoin(base, resource_relpath)
1561
return open(filename, 'rU').read()
1564
def file_kind_from_stat_mode_thunk(mode):
1565
global file_kind_from_stat_mode
1566
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
1568
from bzrlib._readdir_pyx import UTF8DirReader
1569
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
1571
from bzrlib._readdir_py import (
1572
_kind_from_mode as file_kind_from_stat_mode
1574
return file_kind_from_stat_mode(mode)
1575
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
1578
def file_kind(f, _lstat=os.lstat):
1580
return file_kind_from_stat_mode(_lstat(f).st_mode)
1582
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
1583
raise errors.NoSuchFile(f)
added
removed
bzrlib/osutils.py
