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
20
from stat import (S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE,
21
S_ISCHR, S_ISBLK, S_ISFIFO, S_ISSOCK)
25
from bzrlib.lazy_import import lazy_import
26
lazy_import(globals(), """
28
from datetime import datetime
30
from ntpath import (abspath as _nt_abspath,
32
normpath as _nt_normpath,
33
realpath as _nt_realpath,
34
splitdrive as _nt_splitdrive,
43
from tempfile import (
57
from bzrlib import symbol_versioning
58
from bzrlib.symbol_versioning import (
63
# On win32, O_BINARY is used to indicate the file should
64
# be opened in binary mode, rather than text mode.
65
# On other platforms, O_BINARY doesn't exist, because
66
# they always open in binary mode, so it is okay to
67
# OR with 0 on those platforms
68
O_BINARY = getattr(os, 'O_BINARY', 0)
71
def make_readonly(filename):
72
"""Make a filename read-only."""
73
mod = os.lstat(filename).st_mode
74
if not stat.S_ISLNK(mod):
76
os.chmod(filename, mod)
79
def make_writable(filename):
80
mod = os.lstat(filename).st_mode
81
if not stat.S_ISLNK(mod):
83
os.chmod(filename, mod)
86
def minimum_path_selection(paths):
87
"""Return the smallset subset of paths which are outside paths.
89
:param paths: A container (and hence not None) of paths.
90
:return: A set of paths sufficient to include everything in paths via
91
is_inside_any, drawn from the paths parameter.
96
other_paths = paths.difference([path])
97
if not is_inside_any(other_paths, path):
98
# this is a top level path, we must check it.
99
search_paths.add(path)
107
"""Return a quoted filename filename
109
This previously used backslash quoting, but that works poorly on
111
# TODO: I'm not really sure this is the best format either.x
113
if _QUOTE_RE is None:
114
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
116
if _QUOTE_RE.search(f):
122
_directory_kind = 'directory'
125
stat.S_IFDIR:_directory_kind,
126
stat.S_IFCHR:'chardev',
127
stat.S_IFBLK:'block',
130
stat.S_IFLNK:'symlink',
131
stat.S_IFSOCK:'socket',
135
def file_kind_from_stat_mode(stat_mode, _formats=_formats, _unknown='unknown'):
136
"""Generate a file kind from a stat mode. This is used in walkdirs.
138
Its performance is critical: Do not mutate without careful benchmarking.
141
return _formats[stat_mode & 0170000]
146
def file_kind(f, _lstat=os.lstat, _mapper=file_kind_from_stat_mode):
148
return _mapper(_lstat(f).st_mode)
150
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
151
raise errors.NoSuchFile(f)
156
"""Return the current umask"""
157
# Assume that people aren't messing with the umask while running
158
# XXX: This is not thread safe, but there is no way to get the
159
# umask without setting it
167
_directory_kind: "/",
169
'tree-reference': '+',
173
def kind_marker(kind):
175
return _kind_marker_map[kind]
177
raise errors.BzrError('invalid file kind %r' % kind)
180
lexists = getattr(os.path, 'lexists', None)
184
stat = getattr(os, 'lstat', os.stat)
188
if e.errno == errno.ENOENT:
191
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
194
def fancy_rename(old, new, rename_func, unlink_func):
195
"""A fancy rename, when you don't have atomic rename.
197
:param old: The old path, to rename from
198
:param new: The new path, to rename to
199
:param rename_func: The potentially non-atomic rename function
200
:param unlink_func: A way to delete the target file if the full rename succeeds
203
# sftp rename doesn't allow overwriting, so play tricks:
205
base = os.path.basename(new)
206
dirname = os.path.dirname(new)
207
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
208
tmp_name = pathjoin(dirname, tmp_name)
210
# Rename the file out of the way, but keep track if it didn't exist
211
# We don't want to grab just any exception
212
# something like EACCES should prevent us from continuing
213
# The downside is that the rename_func has to throw an exception
214
# with an errno = ENOENT, or NoSuchFile
217
rename_func(new, tmp_name)
218
except (errors.NoSuchFile,), e:
221
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
222
# function raises an IOError with errno is None when a rename fails.
223
# This then gets caught here.
224
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
227
if (getattr(e, 'errno', None) is None
228
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
236
# This may throw an exception, in which case success will
238
rename_func(old, new)
240
except (IOError, OSError), e:
241
# source and target may be aliases of each other (e.g. on a
242
# case-insensitive filesystem), so we may have accidentally renamed
243
# source by when we tried to rename target
244
if not (file_existed and e.errno in (None, errno.ENOENT)):
248
# If the file used to exist, rename it back into place
249
# otherwise just delete it from the tmp location
251
unlink_func(tmp_name)
253
rename_func(tmp_name, new)
256
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
257
# choke on a Unicode string containing a relative path if
258
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
260
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
261
def _posix_abspath(path):
262
# jam 20060426 rather than encoding to fsencoding
263
# copy posixpath.abspath, but use os.getcwdu instead
264
if not posixpath.isabs(path):
265
path = posixpath.join(getcwd(), path)
266
return posixpath.normpath(path)
269
def _posix_realpath(path):
270
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
273
def _win32_fixdrive(path):
274
"""Force drive letters to be consistent.
276
win32 is inconsistent whether it returns lower or upper case
277
and even if it was consistent the user might type the other
278
so we force it to uppercase
279
running python.exe under cmd.exe return capital C:\\
280
running win32 python inside a cygwin shell returns lowercase c:\\
282
drive, path = _nt_splitdrive(path)
283
return drive.upper() + path
286
def _win32_abspath(path):
287
global _win32_abspath
288
if win32utils.winver == 'Windows 98':
289
_win32_abspath = _win98_abspath
291
_win32_abspath = _real_win32_abspath
292
return _win32_abspath(path)
295
def _real_win32_abspath(path):
296
# Real _nt_abspath doesn't have a problem with a unicode cwd
297
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
300
def _win98_abspath(path):
301
"""Return the absolute version of a path.
302
Windows 98 safe implementation (python reimplementation
303
of Win32 API function GetFullPathNameW)
308
# \\HOST\path => //HOST/path
309
# //HOST/path => //HOST/path
310
# path => C:/cwd/path
313
# check for absolute path
314
drive = _nt_splitdrive(path)[0]
315
if drive == '' and path[:2] not in('//','\\\\'):
317
# we cannot simply os.path.join cwd and path
318
# because os.path.join('C:','/path') produce '/path'
319
# and this is incorrect
320
if path[:1] in ('/','\\'):
321
cwd = _nt_splitdrive(cwd)[0]
323
path = cwd + '\\' + path
324
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
327
def _win32_realpath(path):
328
# Real _nt_realpath doesn't have a problem with a unicode cwd
329
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
332
def _win32_pathjoin(*args):
333
return _nt_join(*args).replace('\\', '/')
336
def _win32_normpath(path):
337
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
341
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
344
def _win32_mkdtemp(*args, **kwargs):
345
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
348
def _win32_rename(old, new):
349
"""We expect to be able to atomically replace 'new' with old.
351
On win32, if new exists, it must be moved out of the way first,
355
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
357
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
358
# If we try to rename a non-existant file onto cwd, we get
359
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
360
# if the old path doesn't exist, sometimes we get EACCES
361
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
367
return unicodedata.normalize('NFC', os.getcwdu())
370
# Default is to just use the python builtins, but these can be rebound on
371
# particular platforms.
372
abspath = _posix_abspath
373
realpath = _posix_realpath
374
pathjoin = os.path.join
375
normpath = os.path.normpath
378
dirname = os.path.dirname
379
basename = os.path.basename
380
split = os.path.split
381
splitext = os.path.splitext
382
# These were already imported into local scope
383
# mkdtemp = tempfile.mkdtemp
384
# rmtree = shutil.rmtree
386
MIN_ABS_PATHLENGTH = 1
389
if sys.platform == 'win32':
390
abspath = _win32_abspath
391
realpath = _win32_realpath
392
pathjoin = _win32_pathjoin
393
normpath = _win32_normpath
394
getcwd = _win32_getcwd
395
mkdtemp = _win32_mkdtemp
396
rename = _win32_rename
398
MIN_ABS_PATHLENGTH = 3
400
def _win32_delete_readonly(function, path, excinfo):
401
"""Error handler for shutil.rmtree function [for win32]
402
Helps to remove files and dirs marked as read-only.
404
exception = excinfo[1]
405
if function in (os.remove, os.rmdir) \
406
and isinstance(exception, OSError) \
407
and exception.errno == errno.EACCES:
413
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
414
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
415
return shutil.rmtree(path, ignore_errors, onerror)
416
elif sys.platform == 'darwin':
420
def get_terminal_encoding():
421
"""Find the best encoding for printing to the screen.
423
This attempts to check both sys.stdout and sys.stdin to see
424
what encoding they are in, and if that fails it falls back to
425
osutils.get_user_encoding().
426
The problem is that on Windows, locale.getpreferredencoding()
427
is not the same encoding as that used by the console:
428
http://mail.python.org/pipermail/python-list/2003-May/162357.html
430
On my standard US Windows XP, the preferred encoding is
431
cp1252, but the console is cp437
433
from bzrlib.trace import mutter
434
output_encoding = getattr(sys.stdout, 'encoding', None)
435
if not output_encoding:
436
input_encoding = getattr(sys.stdin, 'encoding', None)
437
if not input_encoding:
438
output_encoding = get_user_encoding()
439
mutter('encoding stdout as osutils.get_user_encoding() %r',
442
output_encoding = input_encoding
443
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
445
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
446
if output_encoding == 'cp0':
447
# invalid encoding (cp0 means 'no codepage' on Windows)
448
output_encoding = get_user_encoding()
449
mutter('cp0 is invalid encoding.'
450
' encoding stdout as osutils.get_user_encoding() %r',
454
codecs.lookup(output_encoding)
456
sys.stderr.write('bzr: warning:'
457
' unknown terminal encoding %s.\n'
458
' Using encoding %s instead.\n'
459
% (output_encoding, get_user_encoding())
461
output_encoding = get_user_encoding()
463
return output_encoding
466
def normalizepath(f):
467
if getattr(os.path, 'realpath', None) is not None:
471
[p,e] = os.path.split(f)
472
if e == "" or e == "." or e == "..":
475
return pathjoin(F(p), e)
479
"""True if f is an accessible directory."""
481
return S_ISDIR(os.lstat(f)[ST_MODE])
487
"""True if f is a regular file."""
489
return S_ISREG(os.lstat(f)[ST_MODE])
494
"""True if f is a symlink."""
496
return S_ISLNK(os.lstat(f)[ST_MODE])
500
def is_inside(dir, fname):
501
"""True if fname is inside dir.
503
The parameters should typically be passed to osutils.normpath first, so
504
that . and .. and repeated slashes are eliminated, and the separators
505
are canonical for the platform.
507
The empty string as a dir name is taken as top-of-tree and matches
510
# XXX: Most callers of this can actually do something smarter by
511
# looking at the inventory
521
return fname.startswith(dir)
524
def is_inside_any(dir_list, fname):
525
"""True if fname is inside any of given dirs."""
526
for dirname in dir_list:
527
if is_inside(dirname, fname):
532
def is_inside_or_parent_of_any(dir_list, fname):
533
"""True if fname is a child or a parent of any of the given files."""
534
for dirname in dir_list:
535
if is_inside(dirname, fname) or is_inside(fname, dirname):
540
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768):
541
"""Copy contents of one file to another.
543
The read_length can either be -1 to read to end-of-file (EOF) or
544
it can specify the maximum number of bytes to read.
546
The buff_size represents the maximum size for each read operation
547
performed on from_file.
549
:return: The number of bytes copied.
553
# read specified number of bytes
555
while read_length > 0:
556
num_bytes_to_read = min(read_length, buff_size)
558
block = from_file.read(num_bytes_to_read)
564
actual_bytes_read = len(block)
565
read_length -= actual_bytes_read
566
length += actual_bytes_read
570
block = from_file.read(buff_size)
579
def pump_string_file(bytes, file_handle, segment_size=None):
580
"""Write bytes to file_handle in many smaller writes.
582
:param bytes: The string to write.
583
:param file_handle: The file to write to.
585
# Write data in chunks rather than all at once, because very large
586
# writes fail on some platforms (e.g. Windows with SMB mounted
589
segment_size = 5242880 # 5MB
590
segments = range(len(bytes) / segment_size + 1)
591
write = file_handle.write
592
for segment_index in segments:
593
segment = buffer(bytes, segment_index * segment_size, segment_size)
597
def file_iterator(input_file, readsize=32768):
599
b = input_file.read(readsize)
606
"""Calculate the hexdigest of an open file.
608
The file cursor should be already at the start.
620
def sha_file_by_name(fname):
621
"""Calculate the SHA1 of a file by reading the full text"""
623
f = os.open(fname, os.O_RDONLY | O_BINARY)
626
b = os.read(f, 1<<16)
634
def sha_strings(strings, _factory=sha.new):
635
"""Return the sha-1 of concatenation of strings"""
637
map(s.update, strings)
641
def sha_string(f, _factory=sha.new):
642
return _factory(f).hexdigest()
645
def fingerprint_file(f):
647
return {'size': len(b),
648
'sha1': sha.new(b).hexdigest()}
651
def compare_files(a, b):
652
"""Returns true if equal in contents"""
663
def local_time_offset(t=None):
664
"""Return offset of local zone from GMT, either at present or at time t."""
667
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
668
return offset.days * 86400 + offset.seconds
670
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
672
def format_date(t, offset=0, timezone='original', date_fmt=None,
674
"""Return a formatted date string.
676
:param t: Seconds since the epoch.
677
:param offset: Timezone offset in seconds east of utc.
678
:param timezone: How to display the time: 'utc', 'original' for the
679
timezone specified by offset, or 'local' for the process's current
681
:param show_offset: Whether to append the timezone.
682
:param date_fmt: strftime format.
684
if timezone == 'utc':
687
elif timezone == 'original':
690
tt = time.gmtime(t + offset)
691
elif timezone == 'local':
692
tt = time.localtime(t)
693
offset = local_time_offset(t)
695
raise errors.UnsupportedTimezoneFormat(timezone)
697
date_fmt = "%a %Y-%m-%d %H:%M:%S"
699
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
702
# day of week depends on locale, so we do this ourself
703
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
704
return (time.strftime(date_fmt, tt) + offset_str)
707
def compact_date(when):
708
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
711
def format_delta(delta):
712
"""Get a nice looking string for a time delta.
714
:param delta: The time difference in seconds, can be positive or negative.
715
positive indicates time in the past, negative indicates time in the
716
future. (usually time.time() - stored_time)
717
:return: String formatted to show approximate resolution
723
direction = 'in the future'
727
if seconds < 90: # print seconds up to 90 seconds
729
return '%d second %s' % (seconds, direction,)
731
return '%d seconds %s' % (seconds, direction)
733
minutes = int(seconds / 60)
734
seconds -= 60 * minutes
739
if minutes < 90: # print minutes, seconds up to 90 minutes
741
return '%d minute, %d second%s %s' % (
742
minutes, seconds, plural_seconds, direction)
744
return '%d minutes, %d second%s %s' % (
745
minutes, seconds, plural_seconds, direction)
747
hours = int(minutes / 60)
748
minutes -= 60 * hours
755
return '%d hour, %d minute%s %s' % (hours, minutes,
756
plural_minutes, direction)
757
return '%d hours, %d minute%s %s' % (hours, minutes,
758
plural_minutes, direction)
761
"""Return size of given open file."""
762
return os.fstat(f.fileno())[ST_SIZE]
765
# Define rand_bytes based on platform.
767
# Python 2.4 and later have os.urandom,
768
# but it doesn't work on some arches
770
rand_bytes = os.urandom
771
except (NotImplementedError, AttributeError):
772
# If python doesn't have os.urandom, or it doesn't work,
773
# then try to first pull random data from /dev/urandom
775
rand_bytes = file('/dev/urandom', 'rb').read
776
# Otherwise, use this hack as a last resort
777
except (IOError, OSError):
778
# not well seeded, but better than nothing
783
s += chr(random.randint(0, 255))
788
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
790
"""Return a random string of num alphanumeric characters
792
The result only contains lowercase chars because it may be used on
793
case-insensitive filesystems.
796
for raw_byte in rand_bytes(num):
797
s += ALNUM[ord(raw_byte) % 36]
801
## TODO: We could later have path objects that remember their list
802
## decomposition (might be too tricksy though.)
805
"""Turn string into list of parts."""
806
# split on either delimiter because people might use either on
808
ps = re.split(r'[\\/]', p)
813
raise errors.BzrError("sorry, %r not allowed in path" % f)
814
elif (f == '.') or (f == ''):
822
if (f == '..') or (f is None) or (f == ''):
823
raise errors.BzrError("sorry, %r not allowed in path" % f)
828
"""Split s into lines, but without removing the newline characters."""
829
lines = s.split('\n')
830
result = [line + '\n' for line in lines[:-1]]
832
result.append(lines[-1])
836
def hardlinks_good():
837
return sys.platform not in ('win32', 'cygwin', 'darwin')
840
def link_or_copy(src, dest):
841
"""Hardlink a file, or copy it if it can't be hardlinked."""
842
if not hardlinks_good():
843
shutil.copyfile(src, dest)
847
except (OSError, IOError), e:
848
if e.errno != errno.EXDEV:
850
shutil.copyfile(src, dest)
853
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
854
# Forgiveness than Permission (EAFP) because:
855
# - root can damage a solaris file system by using unlink,
856
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
857
# EACCES, OSX: EPERM) when invoked on a directory.
858
def delete_any(path):
859
"""Delete a file or directory."""
860
if isdir(path): # Takes care of symlinks
867
if getattr(os, 'symlink', None) is not None:
874
if getattr(os, 'link', None) is not None:
880
def host_os_dereferences_symlinks():
881
return (has_symlinks()
882
and sys.platform not in ('cygwin', 'win32'))
885
def contains_whitespace(s):
886
"""True if there are any whitespace characters in s."""
887
# string.whitespace can include '\xa0' in certain locales, because it is
888
# considered "non-breaking-space" as part of ISO-8859-1. But it
889
# 1) Isn't a breaking whitespace
890
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
892
# 3) '\xa0' isn't unicode safe since it is >128.
894
# This should *not* be a unicode set of characters in case the source
895
# string is not a Unicode string. We can auto-up-cast the characters since
896
# they are ascii, but we don't want to auto-up-cast the string in case it
898
for ch in ' \t\n\r\v\f':
905
def contains_linebreaks(s):
906
"""True if there is any vertical whitespace in s."""
914
def relpath(base, path):
915
"""Return path relative to base, or raise exception.
917
The path may be either an absolute path or a path relative to the
918
current working directory.
920
os.path.commonprefix (python2.4) has a bad bug that it works just
921
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
925
if len(base) < MIN_ABS_PATHLENGTH:
926
# must have space for e.g. a drive letter
927
raise ValueError('%r is too short to calculate a relative path'
934
while len(head) >= len(base):
937
head, tail = os.path.split(head)
941
raise errors.PathNotChild(rp, base)
949
def safe_unicode(unicode_or_utf8_string):
950
"""Coerce unicode_or_utf8_string into unicode.
952
If it is unicode, it is returned.
953
Otherwise it is decoded from utf-8. If a decoding error
954
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
955
as a BzrBadParameter exception.
957
if isinstance(unicode_or_utf8_string, unicode):
958
return unicode_or_utf8_string
960
return unicode_or_utf8_string.decode('utf8')
961
except UnicodeDecodeError:
962
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
965
def safe_utf8(unicode_or_utf8_string):
966
"""Coerce unicode_or_utf8_string to a utf8 string.
968
If it is a str, it is returned.
969
If it is Unicode, it is encoded into a utf-8 string.
971
if isinstance(unicode_or_utf8_string, str):
972
# TODO: jam 20070209 This is overkill, and probably has an impact on
973
# performance if we are dealing with lots of apis that want a
976
# Make sure it is a valid utf-8 string
977
unicode_or_utf8_string.decode('utf-8')
978
except UnicodeDecodeError:
979
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
980
return unicode_or_utf8_string
981
return unicode_or_utf8_string.encode('utf-8')
984
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
985
' Revision id generators should be creating utf8'
989
def safe_revision_id(unicode_or_utf8_string, warn=True):
990
"""Revision ids should now be utf8, but at one point they were unicode.
992
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
994
:param warn: Functions that are sanitizing user data can set warn=False
995
:return: None or a utf8 revision id.
997
if (unicode_or_utf8_string is None
998
or unicode_or_utf8_string.__class__ == str):
999
return unicode_or_utf8_string
1001
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1003
return cache_utf8.encode(unicode_or_utf8_string)
1006
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1007
' generators should be creating utf8 file ids.')
1010
def safe_file_id(unicode_or_utf8_string, warn=True):
1011
"""File ids should now be utf8, but at one point they were unicode.
1013
This is the same as safe_utf8, except it uses the cached encode functions
1014
to save a little bit of performance.
1016
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1018
:param warn: Functions that are sanitizing user data can set warn=False
1019
:return: None or a utf8 file id.
1021
if (unicode_or_utf8_string is None
1022
or unicode_or_utf8_string.__class__ == str):
1023
return unicode_or_utf8_string
1025
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1027
return cache_utf8.encode(unicode_or_utf8_string)
1030
_platform_normalizes_filenames = False
1031
if sys.platform == 'darwin':
1032
_platform_normalizes_filenames = True
1035
def normalizes_filenames():
1036
"""Return True if this platform normalizes unicode filenames.
1038
Mac OSX does, Windows/Linux do not.
1040
return _platform_normalizes_filenames
1043
def _accessible_normalized_filename(path):
1044
"""Get the unicode normalized path, and if you can access the file.
1046
On platforms where the system normalizes filenames (Mac OSX),
1047
you can access a file by any path which will normalize correctly.
1048
On platforms where the system does not normalize filenames
1049
(Windows, Linux), you have to access a file by its exact path.
1051
Internally, bzr only supports NFC normalization, since that is
1052
the standard for XML documents.
1054
So return the normalized path, and a flag indicating if the file
1055
can be accessed by that path.
1058
return unicodedata.normalize('NFC', unicode(path)), True
1061
def _inaccessible_normalized_filename(path):
1062
__doc__ = _accessible_normalized_filename.__doc__
1064
normalized = unicodedata.normalize('NFC', unicode(path))
1065
return normalized, normalized == path
1068
if _platform_normalizes_filenames:
1069
normalized_filename = _accessible_normalized_filename
1071
normalized_filename = _inaccessible_normalized_filename
1074
def terminal_width():
1075
"""Return estimated terminal width."""
1076
if sys.platform == 'win32':
1077
return win32utils.get_console_size()[0]
1080
import struct, fcntl, termios
1081
s = struct.pack('HHHH', 0, 0, 0, 0)
1082
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1083
width = struct.unpack('HHHH', x)[1]
1088
width = int(os.environ['COLUMNS'])
1097
def supports_executable():
1098
return sys.platform != "win32"
1101
def supports_posix_readonly():
1102
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1104
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1105
directory controls creation/deletion, etc.
1107
And under win32, readonly means that the directory itself cannot be
1108
deleted. The contents of a readonly directory can be changed, unlike POSIX
1109
where files in readonly directories cannot be added, deleted or renamed.
1111
return sys.platform != "win32"
1114
def set_or_unset_env(env_variable, value):
1115
"""Modify the environment, setting or removing the env_variable.
1117
:param env_variable: The environment variable in question
1118
:param value: The value to set the environment to. If None, then
1119
the variable will be removed.
1120
:return: The original value of the environment variable.
1122
orig_val = os.environ.get(env_variable)
1124
if orig_val is not None:
1125
del os.environ[env_variable]
1127
if isinstance(value, unicode):
1128
value = value.encode(get_user_encoding())
1129
os.environ[env_variable] = value
1133
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1136
def check_legal_path(path):
1137
"""Check whether the supplied path is legal.
1138
This is only required on Windows, so we don't test on other platforms
1141
if sys.platform != "win32":
1143
if _validWin32PathRE.match(path) is None:
1144
raise errors.IllegalPath(path)
1147
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1149
def _is_error_enotdir(e):
1150
"""Check if this exception represents ENOTDIR.
1152
Unfortunately, python is very inconsistent about the exception
1153
here. The cases are:
1154
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1155
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1156
which is the windows error code.
1157
3) Windows, Python2.5 uses errno == EINVAL and
1158
winerror == ERROR_DIRECTORY
1160
:param e: An Exception object (expected to be OSError with an errno
1161
attribute, but we should be able to cope with anything)
1162
:return: True if this represents an ENOTDIR error. False otherwise.
1164
en = getattr(e, 'errno', None)
1165
if (en == errno.ENOTDIR
1166
or (sys.platform == 'win32'
1167
and (en == _WIN32_ERROR_DIRECTORY
1168
or (en == errno.EINVAL
1169
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1175
def walkdirs(top, prefix=""):
1176
"""Yield data about all the directories in a tree.
1178
This yields all the data about the contents of a directory at a time.
1179
After each directory has been yielded, if the caller has mutated the list
1180
to exclude some directories, they are then not descended into.
1182
The data yielded is of the form:
1183
((directory-relpath, directory-path-from-top),
1184
[(relpath, basename, kind, lstat, path-from-top), ...]),
1185
- directory-relpath is the relative path of the directory being returned
1186
with respect to top. prefix is prepended to this.
1187
- directory-path-from-root is the path including top for this directory.
1188
It is suitable for use with os functions.
1189
- relpath is the relative path within the subtree being walked.
1190
- basename is the basename of the path
1191
- kind is the kind of the file now. If unknown then the file is not
1192
present within the tree - but it may be recorded as versioned. See
1194
- lstat is the stat data *if* the file was statted.
1195
- planned, not implemented:
1196
path_from_tree_root is the path from the root of the tree.
1198
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1199
allows one to walk a subtree but get paths that are relative to a tree
1201
:return: an iterator over the dirs.
1203
#TODO there is a bit of a smell where the results of the directory-
1204
# summary in this, and the path from the root, may not agree
1205
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1206
# potentially confusing output. We should make this more robust - but
1207
# not at a speed cost. RBC 20060731
1209
_directory = _directory_kind
1210
_listdir = os.listdir
1211
_kind_from_mode = _formats.get
1212
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1214
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1215
relroot, _, _, _, top = pending.pop()
1217
relprefix = relroot + u'/'
1220
top_slash = top + u'/'
1223
append = dirblock.append
1225
names = sorted(_listdir(top))
1227
if not _is_error_enotdir(e):
1231
abspath = top_slash + name
1232
statvalue = _lstat(abspath)
1233
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1234
append((relprefix + name, name, kind, statvalue, abspath))
1235
yield (relroot, top), dirblock
1237
# push the user specified dirs from dirblock
1238
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1241
_real_walkdirs_utf8 = None
1243
def _walkdirs_utf8(top, prefix=""):
1244
"""Yield data about all the directories in a tree.
1246
This yields the same information as walkdirs() only each entry is yielded
1247
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1248
are returned as exact byte-strings.
1250
:return: yields a tuple of (dir_info, [file_info])
1251
dir_info is (utf8_relpath, path-from-top)
1252
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1253
if top is an absolute path, path-from-top is also an absolute path.
1254
path-from-top might be unicode or utf8, but it is the correct path to
1255
pass to os functions to affect the file in question. (such as os.lstat)
1257
global _real_walkdirs_utf8
1258
if _real_walkdirs_utf8 is None:
1259
fs_encoding = _fs_enc.upper()
1260
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1261
# Win98 doesn't have unicode apis like FindFirstFileW
1262
# TODO: We possibly could support Win98 by falling back to the
1263
# original FindFirstFile, and using TCHAR instead of WCHAR,
1264
# but that gets a bit tricky, and requires custom compiling
1267
from bzrlib._walkdirs_win32 import _walkdirs_utf8_win32_find_file
1269
_real_walkdirs_utf8 = _walkdirs_unicode_to_utf8
1271
_real_walkdirs_utf8 = _walkdirs_utf8_win32_find_file
1272
elif fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1273
# ANSI_X3.4-1968 is a form of ASCII
1274
_real_walkdirs_utf8 = _walkdirs_unicode_to_utf8
1276
_real_walkdirs_utf8 = _walkdirs_fs_utf8
1277
return _real_walkdirs_utf8(top, prefix=prefix)
1280
def _walkdirs_fs_utf8(top, prefix=""):
1281
"""See _walkdirs_utf8.
1283
This sub-function is called when we know the filesystem is already in utf8
1284
encoding. So we don't need to transcode filenames.
1287
_directory = _directory_kind
1288
# Use C accelerated directory listing.
1289
_listdir = _read_dir
1290
_kind_from_mode = _formats.get
1292
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1293
# But we don't actually uses 1-3 in pending, so set them to None
1294
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1296
relroot, _, _, _, top = pending.pop()
1298
relprefix = relroot + '/'
1301
top_slash = top + '/'
1304
append = dirblock.append
1305
# read_dir supplies in should-stat order.
1306
for _, name in sorted(_listdir(top)):
1307
abspath = top_slash + name
1308
statvalue = _lstat(abspath)
1309
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1310
append((relprefix + name, name, kind, statvalue, abspath))
1312
yield (relroot, top), dirblock
1314
# push the user specified dirs from dirblock
1315
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1318
def _walkdirs_unicode_to_utf8(top, prefix=""):
1319
"""See _walkdirs_utf8
1321
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1323
This is currently the fallback code path when the filesystem encoding is
1324
not UTF-8. It may be better to implement an alternative so that we can
1325
safely handle paths that are not properly decodable in the current
1328
_utf8_encode = codecs.getencoder('utf8')
1330
_directory = _directory_kind
1331
_listdir = os.listdir
1332
_kind_from_mode = _formats.get
1334
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1336
relroot, _, _, _, top = pending.pop()
1338
relprefix = relroot + '/'
1341
top_slash = top + u'/'
1344
append = dirblock.append
1345
for name in sorted(_listdir(top)):
1346
name_utf8 = _utf8_encode(name)[0]
1347
abspath = top_slash + name
1348
statvalue = _lstat(abspath)
1349
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1350
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1351
yield (relroot, top), dirblock
1353
# push the user specified dirs from dirblock
1354
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1357
def copy_tree(from_path, to_path, handlers={}):
1358
"""Copy all of the entries in from_path into to_path.
1360
:param from_path: The base directory to copy.
1361
:param to_path: The target directory. If it does not exist, it will
1363
:param handlers: A dictionary of functions, which takes a source and
1364
destinations for files, directories, etc.
1365
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1366
'file', 'directory', and 'symlink' should always exist.
1367
If they are missing, they will be replaced with 'os.mkdir()',
1368
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1370
# Now, just copy the existing cached tree to the new location
1371
# We use a cheap trick here.
1372
# Absolute paths are prefixed with the first parameter
1373
# relative paths are prefixed with the second.
1374
# So we can get both the source and target returned
1375
# without any extra work.
1377
def copy_dir(source, dest):
1380
def copy_link(source, dest):
1381
"""Copy the contents of a symlink"""
1382
link_to = os.readlink(source)
1383
os.symlink(link_to, dest)
1385
real_handlers = {'file':shutil.copy2,
1386
'symlink':copy_link,
1387
'directory':copy_dir,
1389
real_handlers.update(handlers)
1391
if not os.path.exists(to_path):
1392
real_handlers['directory'](from_path, to_path)
1394
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1395
for relpath, name, kind, st, abspath in entries:
1396
real_handlers[kind](abspath, relpath)
1399
def path_prefix_key(path):
1400
"""Generate a prefix-order path key for path.
1402
This can be used to sort paths in the same way that walkdirs does.
1404
return (dirname(path) , path)
1407
def compare_paths_prefix_order(path_a, path_b):
1408
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1409
key_a = path_prefix_key(path_a)
1410
key_b = path_prefix_key(path_b)
1411
return cmp(key_a, key_b)
1414
_cached_user_encoding = None
1417
def get_user_encoding(use_cache=True):
1418
"""Find out what the preferred user encoding is.
1420
This is generally the encoding that is used for command line parameters
1421
and file contents. This may be different from the terminal encoding
1422
or the filesystem encoding.
1424
:param use_cache: Enable cache for detected encoding.
1425
(This parameter is turned on by default,
1426
and required only for selftesting)
1428
:return: A string defining the preferred user encoding
1430
global _cached_user_encoding
1431
if _cached_user_encoding is not None and use_cache:
1432
return _cached_user_encoding
1434
if sys.platform == 'darwin':
1435
# work around egregious python 2.4 bug
1436
sys.platform = 'posix'
1440
sys.platform = 'darwin'
1445
user_encoding = locale.getpreferredencoding()
1446
except locale.Error, e:
1447
sys.stderr.write('bzr: warning: %s\n'
1448
' Could not determine what text encoding to use.\n'
1449
' This error usually means your Python interpreter\n'
1450
' doesn\'t support the locale set by $LANG (%s)\n'
1451
" Continuing with ascii encoding.\n"
1452
% (e, os.environ.get('LANG')))
1453
user_encoding = 'ascii'
1455
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1456
# treat that as ASCII, and not support printing unicode characters to the
1459
# For python scripts run under vim, we get '', so also treat that as ASCII
1460
if user_encoding in (None, 'cp0', ''):
1461
user_encoding = 'ascii'
1465
codecs.lookup(user_encoding)
1467
sys.stderr.write('bzr: warning:'
1468
' unknown encoding %s.'
1469
' Continuing with ascii encoding.\n'
1472
user_encoding = 'ascii'
1475
_cached_user_encoding = user_encoding
1477
return user_encoding
1480
def get_host_name():
1481
"""Return the current unicode host name.
1483
This is meant to be used in place of socket.gethostname() because that
1484
behaves inconsistently on different platforms.
1486
if sys.platform == "win32":
1488
return win32utils.get_host_name()
1491
return socket.gethostname().decode(get_user_encoding())
1494
def recv_all(socket, bytes):
1495
"""Receive an exact number of bytes.
1497
Regular Socket.recv() may return less than the requested number of bytes,
1498
dependning on what's in the OS buffer. MSG_WAITALL is not available
1499
on all platforms, but this should work everywhere. This will return
1500
less than the requested amount if the remote end closes.
1502
This isn't optimized and is intended mostly for use in testing.
1505
while len(b) < bytes:
1506
new = socket.recv(bytes - len(b))
1513
def send_all(socket, bytes):
1514
"""Send all bytes on a socket.
1516
Regular socket.sendall() can give socket error 10053 on Windows. This
1517
implementation sends no more than 64k at a time, which avoids this problem.
1520
for pos in xrange(0, len(bytes), chunk_size):
1521
socket.sendall(bytes[pos:pos+chunk_size])
1524
def dereference_path(path):
1525
"""Determine the real path to a file.
1527
All parent elements are dereferenced. But the file itself is not
1529
:param path: The original path. May be absolute or relative.
1530
:return: the real path *to* the file
1532
parent, base = os.path.split(path)
1533
# The pathjoin for '.' is a workaround for Python bug #1213894.
1534
# (initial path components aren't dereferenced)
1535
return pathjoin(realpath(pathjoin('.', parent)), base)
1538
def supports_mapi():
1539
"""Return True if we can use MAPI to launch a mail client."""
1540
return sys.platform == "win32"
1543
def resource_string(package, resource_name):
1544
"""Load a resource from a package and return it as a string.
1546
Note: Only packages that start with bzrlib are currently supported.
1548
This is designed to be a lightweight implementation of resource
1549
loading in a way which is API compatible with the same API from
1551
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1552
If and when pkg_resources becomes a standard library, this routine
1555
# Check package name is within bzrlib
1556
if package == "bzrlib":
1557
resource_relpath = resource_name
1558
elif package.startswith("bzrlib."):
1559
package = package[len("bzrlib."):].replace('.', os.sep)
1560
resource_relpath = pathjoin(package, resource_name)
1562
raise errors.BzrError('resource package %s not in bzrlib' % package)
1564
# Map the resource to a file and read its contents
1565
base = dirname(bzrlib.__file__)
1566
if getattr(sys, 'frozen', None): # bzr.exe
1567
base = abspath(pathjoin(base, '..', '..'))
1568
filename = pathjoin(base, resource_relpath)
1569
return open(filename, 'rU').read()
1573
from bzrlib._readdir_pyx import read_dir as _read_dir
1575
from bzrlib._readdir_py import read_dir as _read_dir
added
removed
bzrlib/osutils.py
