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,
43
from tempfile import (
55
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
57
if sys.version_info < (2, 5):
70
from bzrlib import symbol_versioning
71
from bzrlib.symbol_versioning import (
74
from bzrlib.trace import mutter
77
# On win32, O_BINARY is used to indicate the file should
78
# be opened in binary mode, rather than text mode.
79
# On other platforms, O_BINARY doesn't exist, because
80
# they always open in binary mode, so it is okay to
81
# OR with 0 on those platforms
82
O_BINARY = getattr(os, 'O_BINARY', 0)
85
def make_readonly(filename):
86
"""Make a filename read-only."""
87
mod = os.lstat(filename).st_mode
88
if not stat.S_ISLNK(mod):
90
os.chmod(filename, mod)
93
def make_writable(filename):
94
mod = os.lstat(filename).st_mode
95
if not stat.S_ISLNK(mod):
97
os.chmod(filename, mod)
100
def minimum_path_selection(paths):
101
"""Return the smallset subset of paths which are outside paths.
103
:param paths: A container (and hence not None) of paths.
104
:return: A set of paths sufficient to include everything in paths via
105
is_inside_any, drawn from the paths parameter.
110
other_paths = paths.difference([path])
111
if not is_inside_any(other_paths, path):
112
# this is a top level path, we must check it.
113
search_paths.add(path)
121
"""Return a quoted filename filename
123
This previously used backslash quoting, but that works poorly on
125
# TODO: I'm not really sure this is the best format either.x
127
if _QUOTE_RE is None:
128
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
130
if _QUOTE_RE.search(f):
136
_directory_kind = 'directory'
139
"""Return the current umask"""
140
# Assume that people aren't messing with the umask while running
141
# XXX: This is not thread safe, but there is no way to get the
142
# umask without setting it
150
_directory_kind: "/",
152
'tree-reference': '+',
156
def kind_marker(kind):
158
return _kind_marker_map[kind]
160
raise errors.BzrError('invalid file kind %r' % kind)
163
lexists = getattr(os.path, 'lexists', None)
167
stat = getattr(os, 'lstat', os.stat)
171
if e.errno == errno.ENOENT:
174
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
177
def fancy_rename(old, new, rename_func, unlink_func):
178
"""A fancy rename, when you don't have atomic rename.
180
:param old: The old path, to rename from
181
:param new: The new path, to rename to
182
:param rename_func: The potentially non-atomic rename function
183
:param unlink_func: A way to delete the target file if the full rename succeeds
186
# sftp rename doesn't allow overwriting, so play tricks:
188
base = os.path.basename(new)
189
dirname = os.path.dirname(new)
190
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
191
tmp_name = pathjoin(dirname, tmp_name)
193
# Rename the file out of the way, but keep track if it didn't exist
194
# We don't want to grab just any exception
195
# something like EACCES should prevent us from continuing
196
# The downside is that the rename_func has to throw an exception
197
# with an errno = ENOENT, or NoSuchFile
200
rename_func(new, tmp_name)
201
except (errors.NoSuchFile,), e:
204
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
205
# function raises an IOError with errno is None when a rename fails.
206
# This then gets caught here.
207
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
210
if (getattr(e, 'errno', None) is None
211
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
219
# This may throw an exception, in which case success will
221
rename_func(old, new)
223
except (IOError, OSError), e:
224
# source and target may be aliases of each other (e.g. on a
225
# case-insensitive filesystem), so we may have accidentally renamed
226
# source by when we tried to rename target
227
if not (file_existed and e.errno in (None, errno.ENOENT)):
231
# If the file used to exist, rename it back into place
232
# otherwise just delete it from the tmp location
234
unlink_func(tmp_name)
236
rename_func(tmp_name, new)
239
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
240
# choke on a Unicode string containing a relative path if
241
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
243
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
244
def _posix_abspath(path):
245
# jam 20060426 rather than encoding to fsencoding
246
# copy posixpath.abspath, but use os.getcwdu instead
247
if not posixpath.isabs(path):
248
path = posixpath.join(getcwd(), path)
249
return posixpath.normpath(path)
252
def _posix_realpath(path):
253
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
256
def _win32_fixdrive(path):
257
"""Force drive letters to be consistent.
259
win32 is inconsistent whether it returns lower or upper case
260
and even if it was consistent the user might type the other
261
so we force it to uppercase
262
running python.exe under cmd.exe return capital C:\\
263
running win32 python inside a cygwin shell returns lowercase c:\\
265
drive, path = _nt_splitdrive(path)
266
return drive.upper() + path
269
def _win32_abspath(path):
270
# Real _nt_abspath doesn't have a problem with a unicode cwd
271
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
274
def _win98_abspath(path):
275
"""Return the absolute version of a path.
276
Windows 98 safe implementation (python reimplementation
277
of Win32 API function GetFullPathNameW)
282
# \\HOST\path => //HOST/path
283
# //HOST/path => //HOST/path
284
# path => C:/cwd/path
287
# check for absolute path
288
drive = _nt_splitdrive(path)[0]
289
if drive == '' and path[:2] not in('//','\\\\'):
291
# we cannot simply os.path.join cwd and path
292
# because os.path.join('C:','/path') produce '/path'
293
# and this is incorrect
294
if path[:1] in ('/','\\'):
295
cwd = _nt_splitdrive(cwd)[0]
297
path = cwd + '\\' + path
298
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
300
if win32utils.winver == 'Windows 98':
301
_win32_abspath = _win98_abspath
304
def _win32_realpath(path):
305
# Real _nt_realpath doesn't have a problem with a unicode cwd
306
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
309
def _win32_pathjoin(*args):
310
return _nt_join(*args).replace('\\', '/')
313
def _win32_normpath(path):
314
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
318
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
321
def _win32_mkdtemp(*args, **kwargs):
322
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
325
def _win32_rename(old, new):
326
"""We expect to be able to atomically replace 'new' with old.
328
On win32, if new exists, it must be moved out of the way first,
332
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
334
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
335
# If we try to rename a non-existant file onto cwd, we get
336
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
337
# if the old path doesn't exist, sometimes we get EACCES
338
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
344
return unicodedata.normalize('NFC', os.getcwdu())
347
# Default is to just use the python builtins, but these can be rebound on
348
# particular platforms.
349
abspath = _posix_abspath
350
realpath = _posix_realpath
351
pathjoin = os.path.join
352
normpath = os.path.normpath
355
dirname = os.path.dirname
356
basename = os.path.basename
357
split = os.path.split
358
splitext = os.path.splitext
359
# These were already imported into local scope
360
# mkdtemp = tempfile.mkdtemp
361
# rmtree = shutil.rmtree
363
MIN_ABS_PATHLENGTH = 1
366
if sys.platform == 'win32':
367
abspath = _win32_abspath
368
realpath = _win32_realpath
369
pathjoin = _win32_pathjoin
370
normpath = _win32_normpath
371
getcwd = _win32_getcwd
372
mkdtemp = _win32_mkdtemp
373
rename = _win32_rename
375
MIN_ABS_PATHLENGTH = 3
377
def _win32_delete_readonly(function, path, excinfo):
378
"""Error handler for shutil.rmtree function [for win32]
379
Helps to remove files and dirs marked as read-only.
381
exception = excinfo[1]
382
if function in (os.remove, os.rmdir) \
383
and isinstance(exception, OSError) \
384
and exception.errno == errno.EACCES:
390
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
391
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
392
return shutil.rmtree(path, ignore_errors, onerror)
393
elif sys.platform == 'darwin':
397
def get_terminal_encoding():
398
"""Find the best encoding for printing to the screen.
400
This attempts to check both sys.stdout and sys.stdin to see
401
what encoding they are in, and if that fails it falls back to
402
bzrlib.user_encoding.
403
The problem is that on Windows, locale.getpreferredencoding()
404
is not the same encoding as that used by the console:
405
http://mail.python.org/pipermail/python-list/2003-May/162357.html
407
On my standard US Windows XP, the preferred encoding is
408
cp1252, but the console is cp437
410
output_encoding = getattr(sys.stdout, 'encoding', None)
411
if not output_encoding:
412
input_encoding = getattr(sys.stdin, 'encoding', None)
413
if not input_encoding:
414
output_encoding = bzrlib.user_encoding
415
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
417
output_encoding = input_encoding
418
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
420
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
421
if output_encoding == 'cp0':
422
# invalid encoding (cp0 means 'no codepage' on Windows)
423
output_encoding = bzrlib.user_encoding
424
mutter('cp0 is invalid encoding.'
425
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
428
codecs.lookup(output_encoding)
430
sys.stderr.write('bzr: warning:'
431
' unknown terminal encoding %s.\n'
432
' Using encoding %s instead.\n'
433
% (output_encoding, bzrlib.user_encoding)
435
output_encoding = bzrlib.user_encoding
437
return output_encoding
440
def normalizepath(f):
441
if getattr(os.path, 'realpath', None) is not None:
445
[p,e] = os.path.split(f)
446
if e == "" or e == "." or e == "..":
449
return pathjoin(F(p), e)
453
"""True if f is an accessible directory."""
455
return S_ISDIR(os.lstat(f)[ST_MODE])
461
"""True if f is a regular file."""
463
return S_ISREG(os.lstat(f)[ST_MODE])
468
"""True if f is a symlink."""
470
return S_ISLNK(os.lstat(f)[ST_MODE])
474
def is_inside(dir, fname):
475
"""True if fname is inside dir.
477
The parameters should typically be passed to osutils.normpath first, so
478
that . and .. and repeated slashes are eliminated, and the separators
479
are canonical for the platform.
481
The empty string as a dir name is taken as top-of-tree and matches
484
# XXX: Most callers of this can actually do something smarter by
485
# looking at the inventory
495
return fname.startswith(dir)
498
def is_inside_any(dir_list, fname):
499
"""True if fname is inside any of given dirs."""
500
for dirname in dir_list:
501
if is_inside(dirname, fname):
506
def is_inside_or_parent_of_any(dir_list, fname):
507
"""True if fname is a child or a parent of any of the given files."""
508
for dirname in dir_list:
509
if is_inside(dirname, fname) or is_inside(fname, dirname):
514
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768):
515
"""Copy contents of one file to another.
517
The read_length can either be -1 to read to end-of-file (EOF) or
518
it can specify the maximum number of bytes to read.
520
The buff_size represents the maximum size for each read operation
521
performed on from_file.
523
:return: The number of bytes copied.
527
# read specified number of bytes
529
while read_length > 0:
530
num_bytes_to_read = min(read_length, buff_size)
532
block = from_file.read(num_bytes_to_read)
538
actual_bytes_read = len(block)
539
read_length -= actual_bytes_read
540
length += actual_bytes_read
544
block = from_file.read(buff_size)
553
def pump_string_file(bytes, file_handle, segment_size=None):
554
"""Write bytes to file_handle in many smaller writes.
556
:param bytes: The string to write.
557
:param file_handle: The file to write to.
559
# Write data in chunks rather than all at once, because very large
560
# writes fail on some platforms (e.g. Windows with SMB mounted
563
segment_size = 5242880 # 5MB
564
segments = range(len(bytes) / segment_size + 1)
565
write = file_handle.write
566
for segment_index in segments:
567
segment = buffer(bytes, segment_index * segment_size, segment_size)
571
def file_iterator(input_file, readsize=32768):
573
b = input_file.read(readsize)
580
"""Calculate the hexdigest of an open file.
582
The file cursor should be already at the start.
594
def sha_file_by_name(fname):
595
"""Calculate the SHA1 of a file by reading the full text"""
597
f = os.open(fname, os.O_RDONLY | O_BINARY)
600
b = os.read(f, 1<<16)
608
def sha_strings(strings, _factory=sha):
609
"""Return the sha-1 of concatenation of strings"""
611
map(s.update, strings)
615
def sha_string(f, _factory=sha):
616
return _factory(f).hexdigest()
619
def fingerprint_file(f):
621
return {'size': len(b),
622
'sha1': sha(b).hexdigest()}
625
def compare_files(a, b):
626
"""Returns true if equal in contents"""
637
def local_time_offset(t=None):
638
"""Return offset of local zone from GMT, either at present or at time t."""
641
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
642
return offset.days * 86400 + offset.seconds
644
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
646
def format_date(t, offset=0, timezone='original', date_fmt=None,
648
"""Return a formatted date string.
650
:param t: Seconds since the epoch.
651
:param offset: Timezone offset in seconds east of utc.
652
:param timezone: How to display the time: 'utc', 'original' for the
653
timezone specified by offset, or 'local' for the process's current
655
:param show_offset: Whether to append the timezone.
656
:param date_fmt: strftime format.
658
if timezone == 'utc':
661
elif timezone == 'original':
664
tt = time.gmtime(t + offset)
665
elif timezone == 'local':
666
tt = time.localtime(t)
667
offset = local_time_offset(t)
669
raise errors.UnsupportedTimezoneFormat(timezone)
671
date_fmt = "%a %Y-%m-%d %H:%M:%S"
673
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
676
# day of week depends on locale, so we do this ourself
677
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
678
return (time.strftime(date_fmt, tt) + offset_str)
681
def compact_date(when):
682
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
685
def format_delta(delta):
686
"""Get a nice looking string for a time delta.
688
:param delta: The time difference in seconds, can be positive or negative.
689
positive indicates time in the past, negative indicates time in the
690
future. (usually time.time() - stored_time)
691
:return: String formatted to show approximate resolution
697
direction = 'in the future'
701
if seconds < 90: # print seconds up to 90 seconds
703
return '%d second %s' % (seconds, direction,)
705
return '%d seconds %s' % (seconds, direction)
707
minutes = int(seconds / 60)
708
seconds -= 60 * minutes
713
if minutes < 90: # print minutes, seconds up to 90 minutes
715
return '%d minute, %d second%s %s' % (
716
minutes, seconds, plural_seconds, direction)
718
return '%d minutes, %d second%s %s' % (
719
minutes, seconds, plural_seconds, direction)
721
hours = int(minutes / 60)
722
minutes -= 60 * hours
729
return '%d hour, %d minute%s %s' % (hours, minutes,
730
plural_minutes, direction)
731
return '%d hours, %d minute%s %s' % (hours, minutes,
732
plural_minutes, direction)
735
"""Return size of given open file."""
736
return os.fstat(f.fileno())[ST_SIZE]
739
# Define rand_bytes based on platform.
741
# Python 2.4 and later have os.urandom,
742
# but it doesn't work on some arches
744
rand_bytes = os.urandom
745
except (NotImplementedError, AttributeError):
746
# If python doesn't have os.urandom, or it doesn't work,
747
# then try to first pull random data from /dev/urandom
749
rand_bytes = file('/dev/urandom', 'rb').read
750
# Otherwise, use this hack as a last resort
751
except (IOError, OSError):
752
# not well seeded, but better than nothing
757
s += chr(random.randint(0, 255))
762
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
764
"""Return a random string of num alphanumeric characters
766
The result only contains lowercase chars because it may be used on
767
case-insensitive filesystems.
770
for raw_byte in rand_bytes(num):
771
s += ALNUM[ord(raw_byte) % 36]
775
## TODO: We could later have path objects that remember their list
776
## decomposition (might be too tricksy though.)
779
"""Turn string into list of parts."""
780
# split on either delimiter because people might use either on
782
ps = re.split(r'[\\/]', p)
787
raise errors.BzrError("sorry, %r not allowed in path" % f)
788
elif (f == '.') or (f == ''):
796
if (f == '..') or (f is None) or (f == ''):
797
raise errors.BzrError("sorry, %r not allowed in path" % f)
802
"""Split s into lines, but without removing the newline characters."""
803
lines = s.split('\n')
804
result = [line + '\n' for line in lines[:-1]]
806
result.append(lines[-1])
810
def hardlinks_good():
811
return sys.platform not in ('win32', 'cygwin', 'darwin')
814
def link_or_copy(src, dest):
815
"""Hardlink a file, or copy it if it can't be hardlinked."""
816
if not hardlinks_good():
817
shutil.copyfile(src, dest)
821
except (OSError, IOError), e:
822
if e.errno != errno.EXDEV:
824
shutil.copyfile(src, dest)
827
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
828
# Forgiveness than Permission (EAFP) because:
829
# - root can damage a solaris file system by using unlink,
830
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
831
# EACCES, OSX: EPERM) when invoked on a directory.
832
def delete_any(path):
833
"""Delete a file or directory."""
834
if isdir(path): # Takes care of symlinks
841
if getattr(os, 'symlink', None) is not None:
848
if getattr(os, 'link', None) is not None:
854
def host_os_dereferences_symlinks():
855
return (has_symlinks()
856
and sys.platform not in ('cygwin', 'win32'))
859
def contains_whitespace(s):
860
"""True if there are any whitespace characters in s."""
861
# string.whitespace can include '\xa0' in certain locales, because it is
862
# considered "non-breaking-space" as part of ISO-8859-1. But it
863
# 1) Isn't a breaking whitespace
864
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
866
# 3) '\xa0' isn't unicode safe since it is >128.
868
# This should *not* be a unicode set of characters in case the source
869
# string is not a Unicode string. We can auto-up-cast the characters since
870
# they are ascii, but we don't want to auto-up-cast the string in case it
872
for ch in ' \t\n\r\v\f':
879
def contains_linebreaks(s):
880
"""True if there is any vertical whitespace in s."""
888
def relpath(base, path):
889
"""Return path relative to base, or raise exception.
891
The path may be either an absolute path or a path relative to the
892
current working directory.
894
os.path.commonprefix (python2.4) has a bad bug that it works just
895
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
899
if len(base) < MIN_ABS_PATHLENGTH:
900
# must have space for e.g. a drive letter
901
raise ValueError('%r is too short to calculate a relative path'
908
while len(head) >= len(base):
911
head, tail = os.path.split(head)
915
raise errors.PathNotChild(rp, base)
923
def safe_unicode(unicode_or_utf8_string):
924
"""Coerce unicode_or_utf8_string into unicode.
926
If it is unicode, it is returned.
927
Otherwise it is decoded from utf-8. If a decoding error
928
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
929
as a BzrBadParameter exception.
931
if isinstance(unicode_or_utf8_string, unicode):
932
return unicode_or_utf8_string
934
return unicode_or_utf8_string.decode('utf8')
935
except UnicodeDecodeError:
936
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
939
def safe_utf8(unicode_or_utf8_string):
940
"""Coerce unicode_or_utf8_string to a utf8 string.
942
If it is a str, it is returned.
943
If it is Unicode, it is encoded into a utf-8 string.
945
if isinstance(unicode_or_utf8_string, str):
946
# TODO: jam 20070209 This is overkill, and probably has an impact on
947
# performance if we are dealing with lots of apis that want a
950
# Make sure it is a valid utf-8 string
951
unicode_or_utf8_string.decode('utf-8')
952
except UnicodeDecodeError:
953
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
954
return unicode_or_utf8_string
955
return unicode_or_utf8_string.encode('utf-8')
958
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
959
' Revision id generators should be creating utf8'
963
def safe_revision_id(unicode_or_utf8_string, warn=True):
964
"""Revision ids should now be utf8, but at one point they were unicode.
966
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
968
:param warn: Functions that are sanitizing user data can set warn=False
969
:return: None or a utf8 revision id.
971
if (unicode_or_utf8_string is None
972
or unicode_or_utf8_string.__class__ == str):
973
return unicode_or_utf8_string
975
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
977
return cache_utf8.encode(unicode_or_utf8_string)
980
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
981
' generators should be creating utf8 file ids.')
984
def safe_file_id(unicode_or_utf8_string, warn=True):
985
"""File ids should now be utf8, but at one point they were unicode.
987
This is the same as safe_utf8, except it uses the cached encode functions
988
to save a little bit of performance.
990
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
992
:param warn: Functions that are sanitizing user data can set warn=False
993
:return: None or a utf8 file id.
995
if (unicode_or_utf8_string is None
996
or unicode_or_utf8_string.__class__ == str):
997
return unicode_or_utf8_string
999
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1001
return cache_utf8.encode(unicode_or_utf8_string)
1004
_platform_normalizes_filenames = False
1005
if sys.platform == 'darwin':
1006
_platform_normalizes_filenames = True
1009
def normalizes_filenames():
1010
"""Return True if this platform normalizes unicode filenames.
1012
Mac OSX does, Windows/Linux do not.
1014
return _platform_normalizes_filenames
1017
def _accessible_normalized_filename(path):
1018
"""Get the unicode normalized path, and if you can access the file.
1020
On platforms where the system normalizes filenames (Mac OSX),
1021
you can access a file by any path which will normalize correctly.
1022
On platforms where the system does not normalize filenames
1023
(Windows, Linux), you have to access a file by its exact path.
1025
Internally, bzr only supports NFC normalization, since that is
1026
the standard for XML documents.
1028
So return the normalized path, and a flag indicating if the file
1029
can be accessed by that path.
1032
return unicodedata.normalize('NFC', unicode(path)), True
1035
def _inaccessible_normalized_filename(path):
1036
__doc__ = _accessible_normalized_filename.__doc__
1038
normalized = unicodedata.normalize('NFC', unicode(path))
1039
return normalized, normalized == path
1042
if _platform_normalizes_filenames:
1043
normalized_filename = _accessible_normalized_filename
1045
normalized_filename = _inaccessible_normalized_filename
1048
def terminal_width():
1049
"""Return estimated terminal width."""
1050
if sys.platform == 'win32':
1051
return win32utils.get_console_size()[0]
1054
import struct, fcntl, termios
1055
s = struct.pack('HHHH', 0, 0, 0, 0)
1056
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1057
width = struct.unpack('HHHH', x)[1]
1062
width = int(os.environ['COLUMNS'])
1071
def supports_executable():
1072
return sys.platform != "win32"
1075
def supports_posix_readonly():
1076
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1078
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1079
directory controls creation/deletion, etc.
1081
And under win32, readonly means that the directory itself cannot be
1082
deleted. The contents of a readonly directory can be changed, unlike POSIX
1083
where files in readonly directories cannot be added, deleted or renamed.
1085
return sys.platform != "win32"
1088
def set_or_unset_env(env_variable, value):
1089
"""Modify the environment, setting or removing the env_variable.
1091
:param env_variable: The environment variable in question
1092
:param value: The value to set the environment to. If None, then
1093
the variable will be removed.
1094
:return: The original value of the environment variable.
1096
orig_val = os.environ.get(env_variable)
1098
if orig_val is not None:
1099
del os.environ[env_variable]
1101
if isinstance(value, unicode):
1102
value = value.encode(bzrlib.user_encoding)
1103
os.environ[env_variable] = value
1107
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1110
def check_legal_path(path):
1111
"""Check whether the supplied path is legal.
1112
This is only required on Windows, so we don't test on other platforms
1115
if sys.platform != "win32":
1117
if _validWin32PathRE.match(path) is None:
1118
raise errors.IllegalPath(path)
1121
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1123
def _is_error_enotdir(e):
1124
"""Check if this exception represents ENOTDIR.
1126
Unfortunately, python is very inconsistent about the exception
1127
here. The cases are:
1128
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1129
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1130
which is the windows error code.
1131
3) Windows, Python2.5 uses errno == EINVAL and
1132
winerror == ERROR_DIRECTORY
1134
:param e: An Exception object (expected to be OSError with an errno
1135
attribute, but we should be able to cope with anything)
1136
:return: True if this represents an ENOTDIR error. False otherwise.
1138
en = getattr(e, 'errno', None)
1139
if (en == errno.ENOTDIR
1140
or (sys.platform == 'win32'
1141
and (en == _WIN32_ERROR_DIRECTORY
1142
or (en == errno.EINVAL
1143
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1149
def walkdirs(top, prefix=""):
1150
"""Yield data about all the directories in a tree.
1152
This yields all the data about the contents of a directory at a time.
1153
After each directory has been yielded, if the caller has mutated the list
1154
to exclude some directories, they are then not descended into.
1156
The data yielded is of the form:
1157
((directory-relpath, directory-path-from-top),
1158
[(relpath, basename, kind, lstat, path-from-top), ...]),
1159
- directory-relpath is the relative path of the directory being returned
1160
with respect to top. prefix is prepended to this.
1161
- directory-path-from-root is the path including top for this directory.
1162
It is suitable for use with os functions.
1163
- relpath is the relative path within the subtree being walked.
1164
- basename is the basename of the path
1165
- kind is the kind of the file now. If unknown then the file is not
1166
present within the tree - but it may be recorded as versioned. See
1168
- lstat is the stat data *if* the file was statted.
1169
- planned, not implemented:
1170
path_from_tree_root is the path from the root of the tree.
1172
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1173
allows one to walk a subtree but get paths that are relative to a tree
1175
:return: an iterator over the dirs.
1177
#TODO there is a bit of a smell where the results of the directory-
1178
# summary in this, and the path from the root, may not agree
1179
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1180
# potentially confusing output. We should make this more robust - but
1181
# not at a speed cost. RBC 20060731
1183
_directory = _directory_kind
1184
_listdir = os.listdir
1185
_kind_from_mode = file_kind_from_stat_mode
1186
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1188
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1189
relroot, _, _, _, top = pending.pop()
1191
relprefix = relroot + u'/'
1194
top_slash = top + u'/'
1197
append = dirblock.append
1199
names = sorted(_listdir(top))
1201
if not _is_error_enotdir(e):
1205
abspath = top_slash + name
1206
statvalue = _lstat(abspath)
1207
kind = _kind_from_mode(statvalue.st_mode)
1208
append((relprefix + name, name, kind, statvalue, abspath))
1209
yield (relroot, top), dirblock
1211
# push the user specified dirs from dirblock
1212
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1215
class DirReader(object):
1216
"""An interface for reading directories."""
1218
def top_prefix_to_starting_dir(self, top, prefix=""):
1219
"""Converts top and prefix to a starting dir entry
1221
:param top: A utf8 path
1222
:param prefix: An optional utf8 path to prefix output relative paths
1224
:return: A tuple starting with prefix, and ending with the native
1227
raise NotImplementedError(self.top_prefix_to_starting_dir)
1229
def read_dir(self, prefix, top):
1230
"""Read a specific dir.
1232
:param prefix: A utf8 prefix to be preprended to the path basenames.
1233
:param top: A natively encoded path to read.
1234
:return: A list of the directories contents. Each item contains:
1235
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1237
raise NotImplementedError(self.read_dir)
1240
_selected_dir_reader = 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 _selected_dir_reader
1258
if _selected_dir_reader is None:
1259
fs_encoding = _fs_enc.upper()
1260
if 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 Win32ReadDir
1269
_selected_dir_reader = UnicodeDirReader()
1271
_selected_dir_reader = Win32ReadDir()
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
_selected_dir_reader = UnicodeDirReader()
1277
from bzrlib._readdir_pyx import UTF8DirReader
1279
# No optimised code path
1280
_selected_dir_reader = UnicodeDirReader()
1282
_selected_dir_reader = UTF8DirReader()
1283
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1284
# But we don't actually uses 1-3 in pending, so set them to None
1285
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1286
read_dir = _selected_dir_reader.read_dir
1287
_directory = _directory_kind
1289
relroot, _, _, _, top = pending[-1].pop()
1292
dirblock = sorted(read_dir(relroot, top))
1293
yield (relroot, top), dirblock
1294
# push the user specified dirs from dirblock
1295
next = [d for d in reversed(dirblock) if d[2] == _directory]
1297
pending.append(next)
1300
class UnicodeDirReader(DirReader):
1301
"""A dir reader for non-utf8 file systems, which transcodes."""
1303
__slots__ = ['_utf8_encode']
1306
self._utf8_encode = codecs.getencoder('utf8')
1308
def top_prefix_to_starting_dir(self, top, prefix=""):
1309
"""See DirReader.top_prefix_to_starting_dir."""
1310
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1312
def read_dir(self, prefix, top):
1313
"""Read a single directory from a non-utf8 file system.
1315
top, and the abspath element in the output are unicode, all other paths
1316
are utf8. Local disk IO is done via unicode calls to listdir etc.
1318
This is currently the fallback code path when the filesystem encoding is
1319
not UTF-8. It may be better to implement an alternative so that we can
1320
safely handle paths that are not properly decodable in the current
1323
See DirReader.read_dir for details.
1325
_utf8_encode = self._utf8_encode
1327
_listdir = os.listdir
1328
_kind_from_mode = file_kind_from_stat_mode
1331
relprefix = prefix + '/'
1334
top_slash = top + u'/'
1337
append = dirblock.append
1338
for name in sorted(_listdir(top)):
1340
name_utf8 = _utf8_encode(name)[0]
1341
except UnicodeDecodeError:
1342
raise errors.BadFilenameEncoding(
1343
_utf8_encode(relprefix)[0] + name, _fs_enc)
1344
abspath = top_slash + name
1345
statvalue = _lstat(abspath)
1346
kind = _kind_from_mode(statvalue.st_mode)
1347
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1351
def copy_tree(from_path, to_path, handlers={}):
1352
"""Copy all of the entries in from_path into to_path.
1354
:param from_path: The base directory to copy.
1355
:param to_path: The target directory. If it does not exist, it will
1357
:param handlers: A dictionary of functions, which takes a source and
1358
destinations for files, directories, etc.
1359
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1360
'file', 'directory', and 'symlink' should always exist.
1361
If they are missing, they will be replaced with 'os.mkdir()',
1362
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1364
# Now, just copy the existing cached tree to the new location
1365
# We use a cheap trick here.
1366
# Absolute paths are prefixed with the first parameter
1367
# relative paths are prefixed with the second.
1368
# So we can get both the source and target returned
1369
# without any extra work.
1371
def copy_dir(source, dest):
1374
def copy_link(source, dest):
1375
"""Copy the contents of a symlink"""
1376
link_to = os.readlink(source)
1377
os.symlink(link_to, dest)
1379
real_handlers = {'file':shutil.copy2,
1380
'symlink':copy_link,
1381
'directory':copy_dir,
1383
real_handlers.update(handlers)
1385
if not os.path.exists(to_path):
1386
real_handlers['directory'](from_path, to_path)
1388
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1389
for relpath, name, kind, st, abspath in entries:
1390
real_handlers[kind](abspath, relpath)
1393
def path_prefix_key(path):
1394
"""Generate a prefix-order path key for path.
1396
This can be used to sort paths in the same way that walkdirs does.
1398
return (dirname(path) , path)
1401
def compare_paths_prefix_order(path_a, path_b):
1402
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1403
key_a = path_prefix_key(path_a)
1404
key_b = path_prefix_key(path_b)
1405
return cmp(key_a, key_b)
1408
_cached_user_encoding = None
1411
def get_user_encoding(use_cache=True):
1412
"""Find out what the preferred user encoding is.
1414
This is generally the encoding that is used for command line parameters
1415
and file contents. This may be different from the terminal encoding
1416
or the filesystem encoding.
1418
:param use_cache: Enable cache for detected encoding.
1419
(This parameter is turned on by default,
1420
and required only for selftesting)
1422
:return: A string defining the preferred user encoding
1424
global _cached_user_encoding
1425
if _cached_user_encoding is not None and use_cache:
1426
return _cached_user_encoding
1428
if sys.platform == 'darwin':
1429
# python locale.getpreferredencoding() always return
1430
# 'mac-roman' on darwin. That's a lie.
1431
sys.platform = 'posix'
1433
if os.environ.get('LANG', None) is None:
1434
# If LANG is not set, we end up with 'ascii', which is bad
1435
# ('mac-roman' is more than ascii), so we set a default which
1436
# will give us UTF-8 (which appears to work in all cases on
1437
# OSX). Users are still free to override LANG of course, as
1438
# long as it give us something meaningful. This work-around
1439
# *may* not be needed with python 3k and/or OSX 10.5, but will
1440
# work with them too -- vila 20080908
1441
os.environ['LANG'] = 'en_US.UTF-8'
1444
sys.platform = 'darwin'
1449
user_encoding = locale.getpreferredencoding()
1450
except locale.Error, e:
1451
sys.stderr.write('bzr: warning: %s\n'
1452
' Could not determine what text encoding to use.\n'
1453
' This error usually means your Python interpreter\n'
1454
' doesn\'t support the locale set by $LANG (%s)\n'
1455
" Continuing with ascii encoding.\n"
1456
% (e, os.environ.get('LANG')))
1457
user_encoding = 'ascii'
1459
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1460
# treat that as ASCII, and not support printing unicode characters to the
1463
# For python scripts run under vim, we get '', so also treat that as ASCII
1464
if user_encoding in (None, 'cp0', ''):
1465
user_encoding = 'ascii'
1469
codecs.lookup(user_encoding)
1471
sys.stderr.write('bzr: warning:'
1472
' unknown encoding %s.'
1473
' Continuing with ascii encoding.\n'
1476
user_encoding = 'ascii'
1479
_cached_user_encoding = user_encoding
1481
return user_encoding
1484
def get_host_name():
1485
"""Return the current unicode host name.
1487
This is meant to be used in place of socket.gethostname() because that
1488
behaves inconsistently on different platforms.
1490
if sys.platform == "win32":
1492
return win32utils.get_host_name()
1495
return socket.gethostname().decode(get_user_encoding())
1498
def recv_all(socket, bytes):
1499
"""Receive an exact number of bytes.
1501
Regular Socket.recv() may return less than the requested number of bytes,
1502
dependning on what's in the OS buffer. MSG_WAITALL is not available
1503
on all platforms, but this should work everywhere. This will return
1504
less than the requested amount if the remote end closes.
1506
This isn't optimized and is intended mostly for use in testing.
1509
while len(b) < bytes:
1510
new = socket.recv(bytes - len(b))
1517
def send_all(socket, bytes):
1518
"""Send all bytes on a socket.
1520
Regular socket.sendall() can give socket error 10053 on Windows. This
1521
implementation sends no more than 64k at a time, which avoids this problem.
1524
for pos in xrange(0, len(bytes), chunk_size):
1525
socket.sendall(bytes[pos:pos+chunk_size])
1528
def dereference_path(path):
1529
"""Determine the real path to a file.
1531
All parent elements are dereferenced. But the file itself is not
1533
:param path: The original path. May be absolute or relative.
1534
:return: the real path *to* the file
1536
parent, base = os.path.split(path)
1537
# The pathjoin for '.' is a workaround for Python bug #1213894.
1538
# (initial path components aren't dereferenced)
1539
return pathjoin(realpath(pathjoin('.', parent)), base)
1542
def supports_mapi():
1543
"""Return True if we can use MAPI to launch a mail client."""
1544
return sys.platform == "win32"
1547
def resource_string(package, resource_name):
1548
"""Load a resource from a package and return it as a string.
1550
Note: Only packages that start with bzrlib are currently supported.
1552
This is designed to be a lightweight implementation of resource
1553
loading in a way which is API compatible with the same API from
1555
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1556
If and when pkg_resources becomes a standard library, this routine
1559
# Check package name is within bzrlib
1560
if package == "bzrlib":
1561
resource_relpath = resource_name
1562
elif package.startswith("bzrlib."):
1563
package = package[len("bzrlib."):].replace('.', os.sep)
1564
resource_relpath = pathjoin(package, resource_name)
1566
raise errors.BzrError('resource package %s not in bzrlib' % package)
1568
# Map the resource to a file and read its contents
1569
base = dirname(bzrlib.__file__)
1570
if getattr(sys, 'frozen', None): # bzr.exe
1571
base = abspath(pathjoin(base, '..', '..'))
1572
filename = pathjoin(base, resource_relpath)
1573
return open(filename, 'rU').read()
1576
def file_kind_from_stat_mode_thunk(mode):
1577
global file_kind_from_stat_mode
1578
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
1580
from bzrlib._readdir_pyx import UTF8DirReader
1581
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
1583
from bzrlib._readdir_py import (
1584
_kind_from_mode as file_kind_from_stat_mode
1586
return file_kind_from_stat_mode(mode)
1587
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
1590
def file_kind(f, _lstat=os.lstat):
1592
return file_kind_from_stat_mode(_lstat(f).st_mode)
1594
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
1595
raise errors.NoSuchFile(f)
added
removed
bzrlib/osutils.py
