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,
42
from tempfile import (
54
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
56
if sys.version_info < (2, 5):
57
import md5 as _mod_md5
59
import sha as _mod_sha
69
from bzrlib import symbol_versioning
72
# On win32, O_BINARY is used to indicate the file should
73
# be opened in binary mode, rather than text mode.
74
# On other platforms, O_BINARY doesn't exist, because
75
# they always open in binary mode, so it is okay to
76
# OR with 0 on those platforms
77
O_BINARY = getattr(os, 'O_BINARY', 0)
80
def make_readonly(filename):
81
"""Make a filename read-only."""
82
mod = os.lstat(filename).st_mode
83
if not stat.S_ISLNK(mod):
85
os.chmod(filename, mod)
88
def make_writable(filename):
89
mod = os.lstat(filename).st_mode
90
if not stat.S_ISLNK(mod):
92
os.chmod(filename, mod)
95
def minimum_path_selection(paths):
96
"""Return the smallset subset of paths which are outside paths.
98
:param paths: A container (and hence not None) of paths.
99
:return: A set of paths sufficient to include everything in paths via
100
is_inside_any, drawn from the paths parameter.
105
other_paths = paths.difference([path])
106
if not is_inside_any(other_paths, path):
107
# this is a top level path, we must check it.
108
search_paths.add(path)
116
"""Return a quoted filename filename
118
This previously used backslash quoting, but that works poorly on
120
# TODO: I'm not really sure this is the best format either.x
122
if _QUOTE_RE is None:
123
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
125
if _QUOTE_RE.search(f):
131
_directory_kind = 'directory'
134
"""Return the current umask"""
135
# Assume that people aren't messing with the umask while running
136
# XXX: This is not thread safe, but there is no way to get the
137
# umask without setting it
145
_directory_kind: "/",
147
'tree-reference': '+',
151
def kind_marker(kind):
153
return _kind_marker_map[kind]
155
raise errors.BzrError('invalid file kind %r' % kind)
158
lexists = getattr(os.path, 'lexists', None)
162
stat = getattr(os, 'lstat', os.stat)
166
if e.errno == errno.ENOENT:
169
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
172
def fancy_rename(old, new, rename_func, unlink_func):
173
"""A fancy rename, when you don't have atomic rename.
175
:param old: The old path, to rename from
176
:param new: The new path, to rename to
177
:param rename_func: The potentially non-atomic rename function
178
:param unlink_func: A way to delete the target file if the full rename succeeds
181
# sftp rename doesn't allow overwriting, so play tricks:
182
base = os.path.basename(new)
183
dirname = os.path.dirname(new)
184
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
185
tmp_name = pathjoin(dirname, tmp_name)
187
# Rename the file out of the way, but keep track if it didn't exist
188
# We don't want to grab just any exception
189
# something like EACCES should prevent us from continuing
190
# The downside is that the rename_func has to throw an exception
191
# with an errno = ENOENT, or NoSuchFile
194
rename_func(new, tmp_name)
195
except (errors.NoSuchFile,), e:
198
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
199
# function raises an IOError with errno is None when a rename fails.
200
# This then gets caught here.
201
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
204
if (getattr(e, 'errno', None) is None
205
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
213
# This may throw an exception, in which case success will
215
rename_func(old, new)
217
except (IOError, OSError), e:
218
# source and target may be aliases of each other (e.g. on a
219
# case-insensitive filesystem), so we may have accidentally renamed
220
# source by when we tried to rename target
221
if not (file_existed and e.errno in (None, errno.ENOENT)):
225
# If the file used to exist, rename it back into place
226
# otherwise just delete it from the tmp location
228
unlink_func(tmp_name)
230
rename_func(tmp_name, new)
233
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
234
# choke on a Unicode string containing a relative path if
235
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
237
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
238
def _posix_abspath(path):
239
# jam 20060426 rather than encoding to fsencoding
240
# copy posixpath.abspath, but use os.getcwdu instead
241
if not posixpath.isabs(path):
242
path = posixpath.join(getcwd(), path)
243
return posixpath.normpath(path)
246
def _posix_realpath(path):
247
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
250
def _win32_fixdrive(path):
251
"""Force drive letters to be consistent.
253
win32 is inconsistent whether it returns lower or upper case
254
and even if it was consistent the user might type the other
255
so we force it to uppercase
256
running python.exe under cmd.exe return capital C:\\
257
running win32 python inside a cygwin shell returns lowercase c:\\
259
drive, path = _nt_splitdrive(path)
260
return drive.upper() + path
263
def _win32_abspath(path):
264
# Real _nt_abspath doesn't have a problem with a unicode cwd
265
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
268
def _win98_abspath(path):
269
"""Return the absolute version of a path.
270
Windows 98 safe implementation (python reimplementation
271
of Win32 API function GetFullPathNameW)
276
# \\HOST\path => //HOST/path
277
# //HOST/path => //HOST/path
278
# path => C:/cwd/path
281
# check for absolute path
282
drive = _nt_splitdrive(path)[0]
283
if drive == '' and path[:2] not in('//','\\\\'):
285
# we cannot simply os.path.join cwd and path
286
# because os.path.join('C:','/path') produce '/path'
287
# and this is incorrect
288
if path[:1] in ('/','\\'):
289
cwd = _nt_splitdrive(cwd)[0]
291
path = cwd + '\\' + path
292
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
295
def _win32_realpath(path):
296
# Real _nt_realpath doesn't have a problem with a unicode cwd
297
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
300
def _win32_pathjoin(*args):
301
return _nt_join(*args).replace('\\', '/')
304
def _win32_normpath(path):
305
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
309
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
312
def _win32_mkdtemp(*args, **kwargs):
313
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
316
def _win32_rename(old, new):
317
"""We expect to be able to atomically replace 'new' with old.
319
On win32, if new exists, it must be moved out of the way first,
323
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
325
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
326
# If we try to rename a non-existant file onto cwd, we get
327
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
328
# if the old path doesn't exist, sometimes we get EACCES
329
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
335
return unicodedata.normalize('NFC', os.getcwdu())
338
# Default is to just use the python builtins, but these can be rebound on
339
# particular platforms.
340
abspath = _posix_abspath
341
realpath = _posix_realpath
342
pathjoin = os.path.join
343
normpath = os.path.normpath
346
dirname = os.path.dirname
347
basename = os.path.basename
348
split = os.path.split
349
splitext = os.path.splitext
350
# These were already imported into local scope
351
# mkdtemp = tempfile.mkdtemp
352
# rmtree = shutil.rmtree
354
MIN_ABS_PATHLENGTH = 1
357
if sys.platform == 'win32':
358
if win32utils.winver == 'Windows 98':
359
abspath = _win98_abspath
361
abspath = _win32_abspath
362
realpath = _win32_realpath
363
pathjoin = _win32_pathjoin
364
normpath = _win32_normpath
365
getcwd = _win32_getcwd
366
mkdtemp = _win32_mkdtemp
367
rename = _win32_rename
369
MIN_ABS_PATHLENGTH = 3
371
def _win32_delete_readonly(function, path, excinfo):
372
"""Error handler for shutil.rmtree function [for win32]
373
Helps to remove files and dirs marked as read-only.
375
exception = excinfo[1]
376
if function in (os.remove, os.rmdir) \
377
and isinstance(exception, OSError) \
378
and exception.errno == errno.EACCES:
384
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
385
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
386
return shutil.rmtree(path, ignore_errors, onerror)
387
elif sys.platform == 'darwin':
391
def get_terminal_encoding():
392
"""Find the best encoding for printing to the screen.
394
This attempts to check both sys.stdout and sys.stdin to see
395
what encoding they are in, and if that fails it falls back to
396
osutils.get_user_encoding().
397
The problem is that on Windows, locale.getpreferredencoding()
398
is not the same encoding as that used by the console:
399
http://mail.python.org/pipermail/python-list/2003-May/162357.html
401
On my standard US Windows XP, the preferred encoding is
402
cp1252, but the console is cp437
404
from bzrlib.trace import mutter
405
output_encoding = getattr(sys.stdout, 'encoding', None)
406
if not output_encoding:
407
input_encoding = getattr(sys.stdin, 'encoding', None)
408
if not input_encoding:
409
output_encoding = get_user_encoding()
410
mutter('encoding stdout as osutils.get_user_encoding() %r',
413
output_encoding = input_encoding
414
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
416
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
417
if output_encoding == 'cp0':
418
# invalid encoding (cp0 means 'no codepage' on Windows)
419
output_encoding = get_user_encoding()
420
mutter('cp0 is invalid encoding.'
421
' encoding stdout as osutils.get_user_encoding() %r',
425
codecs.lookup(output_encoding)
427
sys.stderr.write('bzr: warning:'
428
' unknown terminal encoding %s.\n'
429
' Using encoding %s instead.\n'
430
% (output_encoding, get_user_encoding())
432
output_encoding = get_user_encoding()
434
return output_encoding
437
def normalizepath(f):
438
if getattr(os.path, 'realpath', None) is not None:
442
[p,e] = os.path.split(f)
443
if e == "" or e == "." or e == "..":
446
return pathjoin(F(p), e)
450
"""True if f is an accessible directory."""
452
return S_ISDIR(os.lstat(f)[ST_MODE])
458
"""True if f is a regular file."""
460
return S_ISREG(os.lstat(f)[ST_MODE])
465
"""True if f is a symlink."""
467
return S_ISLNK(os.lstat(f)[ST_MODE])
471
def is_inside(dir, fname):
472
"""True if fname is inside dir.
474
The parameters should typically be passed to osutils.normpath first, so
475
that . and .. and repeated slashes are eliminated, and the separators
476
are canonical for the platform.
478
The empty string as a dir name is taken as top-of-tree and matches
481
# XXX: Most callers of this can actually do something smarter by
482
# looking at the inventory
492
return fname.startswith(dir)
495
def is_inside_any(dir_list, fname):
496
"""True if fname is inside any of given dirs."""
497
for dirname in dir_list:
498
if is_inside(dirname, fname):
503
def is_inside_or_parent_of_any(dir_list, fname):
504
"""True if fname is a child or a parent of any of the given files."""
505
for dirname in dir_list:
506
if is_inside(dirname, fname) or is_inside(fname, dirname):
511
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768):
512
"""Copy contents of one file to another.
514
The read_length can either be -1 to read to end-of-file (EOF) or
515
it can specify the maximum number of bytes to read.
517
The buff_size represents the maximum size for each read operation
518
performed on from_file.
520
:return: The number of bytes copied.
524
# read specified number of bytes
526
while read_length > 0:
527
num_bytes_to_read = min(read_length, buff_size)
529
block = from_file.read(num_bytes_to_read)
535
actual_bytes_read = len(block)
536
read_length -= actual_bytes_read
537
length += actual_bytes_read
541
block = from_file.read(buff_size)
550
def pump_string_file(bytes, file_handle, segment_size=None):
551
"""Write bytes to file_handle in many smaller writes.
553
:param bytes: The string to write.
554
:param file_handle: The file to write to.
556
# Write data in chunks rather than all at once, because very large
557
# writes fail on some platforms (e.g. Windows with SMB mounted
560
segment_size = 5242880 # 5MB
561
segments = range(len(bytes) / segment_size + 1)
562
write = file_handle.write
563
for segment_index in segments:
564
segment = buffer(bytes, segment_index * segment_size, segment_size)
568
def file_iterator(input_file, readsize=32768):
570
b = input_file.read(readsize)
577
"""Calculate the hexdigest of an open file.
579
The file cursor should be already at the start.
591
def sha_file_by_name(fname):
592
"""Calculate the SHA1 of a file by reading the full text"""
594
f = os.open(fname, os.O_RDONLY | O_BINARY)
597
b = os.read(f, 1<<16)
605
def sha_strings(strings, _factory=sha):
606
"""Return the sha-1 of concatenation of strings"""
608
map(s.update, strings)
612
def sha_string(f, _factory=sha):
613
return _factory(f).hexdigest()
616
def fingerprint_file(f):
618
return {'size': len(b),
619
'sha1': sha(b).hexdigest()}
622
def compare_files(a, b):
623
"""Returns true if equal in contents"""
634
def local_time_offset(t=None):
635
"""Return offset of local zone from GMT, either at present or at time t."""
638
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
639
return offset.days * 86400 + offset.seconds
641
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
643
def format_date(t, offset=0, timezone='original', date_fmt=None,
645
"""Return a formatted date string.
647
:param t: Seconds since the epoch.
648
:param offset: Timezone offset in seconds east of utc.
649
:param timezone: How to display the time: 'utc', 'original' for the
650
timezone specified by offset, or 'local' for the process's current
652
:param date_fmt: strftime format.
653
:param show_offset: Whether to append the timezone.
655
(date_fmt, tt, offset_str) = \
656
_format_date(t, offset, timezone, date_fmt, show_offset)
657
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
658
date_str = time.strftime(date_fmt, tt)
659
return date_str + offset_str
661
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
663
"""Return an unicode date string formatted according to the current locale.
665
:param t: Seconds since the epoch.
666
:param offset: Timezone offset in seconds east of utc.
667
:param timezone: How to display the time: 'utc', 'original' for the
668
timezone specified by offset, or 'local' for the process's current
670
:param date_fmt: strftime format.
671
:param show_offset: Whether to append the timezone.
673
(date_fmt, tt, offset_str) = \
674
_format_date(t, offset, timezone, date_fmt, show_offset)
675
date_str = time.strftime(date_fmt, tt)
676
if not isinstance(date_str, unicode):
677
date_str = date_str.decode(bzrlib.user_encoding, 'replace')
678
return date_str + offset_str
680
def _format_date(t, offset, timezone, date_fmt, show_offset):
681
if timezone == 'utc':
684
elif timezone == 'original':
687
tt = time.gmtime(t + offset)
688
elif timezone == 'local':
689
tt = time.localtime(t)
690
offset = local_time_offset(t)
692
raise errors.UnsupportedTimezoneFormat(timezone)
694
date_fmt = "%a %Y-%m-%d %H:%M:%S"
696
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
699
return (date_fmt, tt, offset_str)
702
def compact_date(when):
703
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
706
def format_delta(delta):
707
"""Get a nice looking string for a time delta.
709
:param delta: The time difference in seconds, can be positive or negative.
710
positive indicates time in the past, negative indicates time in the
711
future. (usually time.time() - stored_time)
712
:return: String formatted to show approximate resolution
718
direction = 'in the future'
722
if seconds < 90: # print seconds up to 90 seconds
724
return '%d second %s' % (seconds, direction,)
726
return '%d seconds %s' % (seconds, direction)
728
minutes = int(seconds / 60)
729
seconds -= 60 * minutes
734
if minutes < 90: # print minutes, seconds up to 90 minutes
736
return '%d minute, %d second%s %s' % (
737
minutes, seconds, plural_seconds, direction)
739
return '%d minutes, %d second%s %s' % (
740
minutes, seconds, plural_seconds, direction)
742
hours = int(minutes / 60)
743
minutes -= 60 * hours
750
return '%d hour, %d minute%s %s' % (hours, minutes,
751
plural_minutes, direction)
752
return '%d hours, %d minute%s %s' % (hours, minutes,
753
plural_minutes, direction)
756
"""Return size of given open file."""
757
return os.fstat(f.fileno())[ST_SIZE]
760
# Define rand_bytes based on platform.
762
# Python 2.4 and later have os.urandom,
763
# but it doesn't work on some arches
765
rand_bytes = os.urandom
766
except (NotImplementedError, AttributeError):
767
# If python doesn't have os.urandom, or it doesn't work,
768
# then try to first pull random data from /dev/urandom
770
rand_bytes = file('/dev/urandom', 'rb').read
771
# Otherwise, use this hack as a last resort
772
except (IOError, OSError):
773
# not well seeded, but better than nothing
778
s += chr(random.randint(0, 255))
783
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
785
"""Return a random string of num alphanumeric characters
787
The result only contains lowercase chars because it may be used on
788
case-insensitive filesystems.
791
for raw_byte in rand_bytes(num):
792
s += ALNUM[ord(raw_byte) % 36]
796
## TODO: We could later have path objects that remember their list
797
## decomposition (might be too tricksy though.)
800
"""Turn string into list of parts."""
801
# split on either delimiter because people might use either on
803
ps = re.split(r'[\\/]', p)
808
raise errors.BzrError("sorry, %r not allowed in path" % f)
809
elif (f == '.') or (f == ''):
818
if (f == '..') or (f is None) or (f == ''):
819
raise errors.BzrError("sorry, %r not allowed in path" % f)
823
def chunks_to_lines(chunks):
824
"""Ensure that chunks is split cleanly into lines.
826
Each entry in the result should contain a single newline at the end. Except
827
for the last entry which may not have a final newline.
829
:param chunks: An iterable of strings
830
:return: A list of strings.
832
# Optimize for a very common case when chunks are already lines
836
# This is a bit ugly, but is the fastest way to check if all of the
837
# chunks are individual lines.
838
# You can't use function calls like .count(), .index(), or endswith()
839
# because they incur too much python overhead.
841
# if chunk is an empty string, it will raise IndexError, which will
843
# if chunk doesn't end with '\n' then we hit fail()
844
# if there is more than one '\n' then we hit fail()
845
# timing shows this loop to take 2.58ms rather than 3.18ms for
846
# split_lines(''.join(chunks))
847
# Further, it means we get to preserve the original lines, rather than
849
[(chunk[-1] == '\n' and '\n' not in chunk[:-1]) or fail()
854
return split_lines(''.join(chunks))
858
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
864
"""Split s into lines, but without removing the newline characters."""
865
lines = s.split('\n')
866
result = [line + '\n' for line in lines[:-1]]
868
result.append(lines[-1])
872
def hardlinks_good():
873
return sys.platform not in ('win32', 'cygwin', 'darwin')
876
def link_or_copy(src, dest):
877
"""Hardlink a file, or copy it if it can't be hardlinked."""
878
if not hardlinks_good():
879
shutil.copyfile(src, dest)
883
except (OSError, IOError), e:
884
if e.errno != errno.EXDEV:
886
shutil.copyfile(src, dest)
889
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
890
# Forgiveness than Permission (EAFP) because:
891
# - root can damage a solaris file system by using unlink,
892
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
893
# EACCES, OSX: EPERM) when invoked on a directory.
894
def delete_any(path):
895
"""Delete a file or directory."""
896
if isdir(path): # Takes care of symlinks
903
if getattr(os, 'symlink', None) is not None:
910
if getattr(os, 'link', None) is not None:
916
def host_os_dereferences_symlinks():
917
return (has_symlinks()
918
and sys.platform not in ('cygwin', 'win32'))
921
def contains_whitespace(s):
922
"""True if there are any whitespace characters in s."""
923
# string.whitespace can include '\xa0' in certain locales, because it is
924
# considered "non-breaking-space" as part of ISO-8859-1. But it
925
# 1) Isn't a breaking whitespace
926
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
928
# 3) '\xa0' isn't unicode safe since it is >128.
930
# This should *not* be a unicode set of characters in case the source
931
# string is not a Unicode string. We can auto-up-cast the characters since
932
# they are ascii, but we don't want to auto-up-cast the string in case it
934
for ch in ' \t\n\r\v\f':
941
def contains_linebreaks(s):
942
"""True if there is any vertical whitespace in s."""
950
def relpath(base, path):
951
"""Return path relative to base, or raise exception.
953
The path may be either an absolute path or a path relative to the
954
current working directory.
956
os.path.commonprefix (python2.4) has a bad bug that it works just
957
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
961
if len(base) < MIN_ABS_PATHLENGTH:
962
# must have space for e.g. a drive letter
963
raise ValueError('%r is too short to calculate a relative path'
970
while len(head) >= len(base):
973
head, tail = os.path.split(head)
977
raise errors.PathNotChild(rp, base)
985
def safe_unicode(unicode_or_utf8_string):
986
"""Coerce unicode_or_utf8_string into unicode.
988
If it is unicode, it is returned.
989
Otherwise it is decoded from utf-8. If a decoding error
990
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
991
as a BzrBadParameter exception.
993
if isinstance(unicode_or_utf8_string, unicode):
994
return unicode_or_utf8_string
996
return unicode_or_utf8_string.decode('utf8')
997
except UnicodeDecodeError:
998
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1001
def safe_utf8(unicode_or_utf8_string):
1002
"""Coerce unicode_or_utf8_string to a utf8 string.
1004
If it is a str, it is returned.
1005
If it is Unicode, it is encoded into a utf-8 string.
1007
if isinstance(unicode_or_utf8_string, str):
1008
# TODO: jam 20070209 This is overkill, and probably has an impact on
1009
# performance if we are dealing with lots of apis that want a
1012
# Make sure it is a valid utf-8 string
1013
unicode_or_utf8_string.decode('utf-8')
1014
except UnicodeDecodeError:
1015
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1016
return unicode_or_utf8_string
1017
return unicode_or_utf8_string.encode('utf-8')
1020
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1021
' Revision id generators should be creating utf8'
1025
def safe_revision_id(unicode_or_utf8_string, warn=True):
1026
"""Revision ids should now be utf8, but at one point they were unicode.
1028
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1030
:param warn: Functions that are sanitizing user data can set warn=False
1031
:return: None or a utf8 revision id.
1033
if (unicode_or_utf8_string is None
1034
or unicode_or_utf8_string.__class__ == str):
1035
return unicode_or_utf8_string
1037
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1039
return cache_utf8.encode(unicode_or_utf8_string)
1042
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1043
' generators should be creating utf8 file ids.')
1046
def safe_file_id(unicode_or_utf8_string, warn=True):
1047
"""File ids should now be utf8, but at one point they were unicode.
1049
This is the same as safe_utf8, except it uses the cached encode functions
1050
to save a little bit of performance.
1052
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1054
:param warn: Functions that are sanitizing user data can set warn=False
1055
:return: None or a utf8 file id.
1057
if (unicode_or_utf8_string is None
1058
or unicode_or_utf8_string.__class__ == str):
1059
return unicode_or_utf8_string
1061
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1063
return cache_utf8.encode(unicode_or_utf8_string)
1066
_platform_normalizes_filenames = False
1067
if sys.platform == 'darwin':
1068
_platform_normalizes_filenames = True
1071
def normalizes_filenames():
1072
"""Return True if this platform normalizes unicode filenames.
1074
Mac OSX does, Windows/Linux do not.
1076
return _platform_normalizes_filenames
1079
def _accessible_normalized_filename(path):
1080
"""Get the unicode normalized path, and if you can access the file.
1082
On platforms where the system normalizes filenames (Mac OSX),
1083
you can access a file by any path which will normalize correctly.
1084
On platforms where the system does not normalize filenames
1085
(Windows, Linux), you have to access a file by its exact path.
1087
Internally, bzr only supports NFC normalization, since that is
1088
the standard for XML documents.
1090
So return the normalized path, and a flag indicating if the file
1091
can be accessed by that path.
1094
return unicodedata.normalize('NFC', unicode(path)), True
1097
def _inaccessible_normalized_filename(path):
1098
__doc__ = _accessible_normalized_filename.__doc__
1100
normalized = unicodedata.normalize('NFC', unicode(path))
1101
return normalized, normalized == path
1104
if _platform_normalizes_filenames:
1105
normalized_filename = _accessible_normalized_filename
1107
normalized_filename = _inaccessible_normalized_filename
1110
def terminal_width():
1111
"""Return estimated terminal width."""
1112
if sys.platform == 'win32':
1113
return win32utils.get_console_size()[0]
1116
import struct, fcntl, termios
1117
s = struct.pack('HHHH', 0, 0, 0, 0)
1118
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1119
width = struct.unpack('HHHH', x)[1]
1124
width = int(os.environ['COLUMNS'])
1133
def supports_executable():
1134
return sys.platform != "win32"
1137
def supports_posix_readonly():
1138
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1140
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1141
directory controls creation/deletion, etc.
1143
And under win32, readonly means that the directory itself cannot be
1144
deleted. The contents of a readonly directory can be changed, unlike POSIX
1145
where files in readonly directories cannot be added, deleted or renamed.
1147
return sys.platform != "win32"
1150
def set_or_unset_env(env_variable, value):
1151
"""Modify the environment, setting or removing the env_variable.
1153
:param env_variable: The environment variable in question
1154
:param value: The value to set the environment to. If None, then
1155
the variable will be removed.
1156
:return: The original value of the environment variable.
1158
orig_val = os.environ.get(env_variable)
1160
if orig_val is not None:
1161
del os.environ[env_variable]
1163
if isinstance(value, unicode):
1164
value = value.encode(get_user_encoding())
1165
os.environ[env_variable] = value
1169
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1172
def check_legal_path(path):
1173
"""Check whether the supplied path is legal.
1174
This is only required on Windows, so we don't test on other platforms
1177
if sys.platform != "win32":
1179
if _validWin32PathRE.match(path) is None:
1180
raise errors.IllegalPath(path)
1183
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1185
def _is_error_enotdir(e):
1186
"""Check if this exception represents ENOTDIR.
1188
Unfortunately, python is very inconsistent about the exception
1189
here. The cases are:
1190
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1191
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1192
which is the windows error code.
1193
3) Windows, Python2.5 uses errno == EINVAL and
1194
winerror == ERROR_DIRECTORY
1196
:param e: An Exception object (expected to be OSError with an errno
1197
attribute, but we should be able to cope with anything)
1198
:return: True if this represents an ENOTDIR error. False otherwise.
1200
en = getattr(e, 'errno', None)
1201
if (en == errno.ENOTDIR
1202
or (sys.platform == 'win32'
1203
and (en == _WIN32_ERROR_DIRECTORY
1204
or (en == errno.EINVAL
1205
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1211
def walkdirs(top, prefix=""):
1212
"""Yield data about all the directories in a tree.
1214
This yields all the data about the contents of a directory at a time.
1215
After each directory has been yielded, if the caller has mutated the list
1216
to exclude some directories, they are then not descended into.
1218
The data yielded is of the form:
1219
((directory-relpath, directory-path-from-top),
1220
[(relpath, basename, kind, lstat, path-from-top), ...]),
1221
- directory-relpath is the relative path of the directory being returned
1222
with respect to top. prefix is prepended to this.
1223
- directory-path-from-root is the path including top for this directory.
1224
It is suitable for use with os functions.
1225
- relpath is the relative path within the subtree being walked.
1226
- basename is the basename of the path
1227
- kind is the kind of the file now. If unknown then the file is not
1228
present within the tree - but it may be recorded as versioned. See
1230
- lstat is the stat data *if* the file was statted.
1231
- planned, not implemented:
1232
path_from_tree_root is the path from the root of the tree.
1234
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1235
allows one to walk a subtree but get paths that are relative to a tree
1237
:return: an iterator over the dirs.
1239
#TODO there is a bit of a smell where the results of the directory-
1240
# summary in this, and the path from the root, may not agree
1241
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1242
# potentially confusing output. We should make this more robust - but
1243
# not at a speed cost. RBC 20060731
1245
_directory = _directory_kind
1246
_listdir = os.listdir
1247
_kind_from_mode = file_kind_from_stat_mode
1248
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1250
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1251
relroot, _, _, _, top = pending.pop()
1253
relprefix = relroot + u'/'
1256
top_slash = top + u'/'
1259
append = dirblock.append
1261
names = sorted(_listdir(top))
1263
if not _is_error_enotdir(e):
1267
abspath = top_slash + name
1268
statvalue = _lstat(abspath)
1269
kind = _kind_from_mode(statvalue.st_mode)
1270
append((relprefix + name, name, kind, statvalue, abspath))
1271
yield (relroot, top), dirblock
1273
# push the user specified dirs from dirblock
1274
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1277
class DirReader(object):
1278
"""An interface for reading directories."""
1280
def top_prefix_to_starting_dir(self, top, prefix=""):
1281
"""Converts top and prefix to a starting dir entry
1283
:param top: A utf8 path
1284
:param prefix: An optional utf8 path to prefix output relative paths
1286
:return: A tuple starting with prefix, and ending with the native
1289
raise NotImplementedError(self.top_prefix_to_starting_dir)
1291
def read_dir(self, prefix, top):
1292
"""Read a specific dir.
1294
:param prefix: A utf8 prefix to be preprended to the path basenames.
1295
:param top: A natively encoded path to read.
1296
:return: A list of the directories contents. Each item contains:
1297
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1299
raise NotImplementedError(self.read_dir)
1302
_selected_dir_reader = None
1305
def _walkdirs_utf8(top, prefix=""):
1306
"""Yield data about all the directories in a tree.
1308
This yields the same information as walkdirs() only each entry is yielded
1309
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1310
are returned as exact byte-strings.
1312
:return: yields a tuple of (dir_info, [file_info])
1313
dir_info is (utf8_relpath, path-from-top)
1314
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1315
if top is an absolute path, path-from-top is also an absolute path.
1316
path-from-top might be unicode or utf8, but it is the correct path to
1317
pass to os functions to affect the file in question. (such as os.lstat)
1319
global _selected_dir_reader
1320
if _selected_dir_reader is None:
1321
fs_encoding = _fs_enc.upper()
1322
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1323
# Win98 doesn't have unicode apis like FindFirstFileW
1324
# TODO: We possibly could support Win98 by falling back to the
1325
# original FindFirstFile, and using TCHAR instead of WCHAR,
1326
# but that gets a bit tricky, and requires custom compiling
1329
from bzrlib._walkdirs_win32 import Win32ReadDir
1331
_selected_dir_reader = UnicodeDirReader()
1333
_selected_dir_reader = Win32ReadDir()
1334
elif fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1335
# ANSI_X3.4-1968 is a form of ASCII
1336
_selected_dir_reader = UnicodeDirReader()
1339
from bzrlib._readdir_pyx import UTF8DirReader
1341
# No optimised code path
1342
_selected_dir_reader = UnicodeDirReader()
1344
_selected_dir_reader = UTF8DirReader()
1345
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1346
# But we don't actually uses 1-3 in pending, so set them to None
1347
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1348
read_dir = _selected_dir_reader.read_dir
1349
_directory = _directory_kind
1351
relroot, _, _, _, top = pending[-1].pop()
1354
dirblock = sorted(read_dir(relroot, top))
1355
yield (relroot, top), dirblock
1356
# push the user specified dirs from dirblock
1357
next = [d for d in reversed(dirblock) if d[2] == _directory]
1359
pending.append(next)
1362
class UnicodeDirReader(DirReader):
1363
"""A dir reader for non-utf8 file systems, which transcodes."""
1365
__slots__ = ['_utf8_encode']
1368
self._utf8_encode = codecs.getencoder('utf8')
1370
def top_prefix_to_starting_dir(self, top, prefix=""):
1371
"""See DirReader.top_prefix_to_starting_dir."""
1372
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1374
def read_dir(self, prefix, top):
1375
"""Read a single directory from a non-utf8 file system.
1377
top, and the abspath element in the output are unicode, all other paths
1378
are utf8. Local disk IO is done via unicode calls to listdir etc.
1380
This is currently the fallback code path when the filesystem encoding is
1381
not UTF-8. It may be better to implement an alternative so that we can
1382
safely handle paths that are not properly decodable in the current
1385
See DirReader.read_dir for details.
1387
_utf8_encode = self._utf8_encode
1389
_listdir = os.listdir
1390
_kind_from_mode = file_kind_from_stat_mode
1393
relprefix = prefix + '/'
1396
top_slash = top + u'/'
1399
append = dirblock.append
1400
for name in sorted(_listdir(top)):
1402
name_utf8 = _utf8_encode(name)[0]
1403
except UnicodeDecodeError:
1404
raise errors.BadFilenameEncoding(
1405
_utf8_encode(relprefix)[0] + name, _fs_enc)
1406
abspath = top_slash + name
1407
statvalue = _lstat(abspath)
1408
kind = _kind_from_mode(statvalue.st_mode)
1409
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1413
def copy_tree(from_path, to_path, handlers={}):
1414
"""Copy all of the entries in from_path into to_path.
1416
:param from_path: The base directory to copy.
1417
:param to_path: The target directory. If it does not exist, it will
1419
:param handlers: A dictionary of functions, which takes a source and
1420
destinations for files, directories, etc.
1421
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1422
'file', 'directory', and 'symlink' should always exist.
1423
If they are missing, they will be replaced with 'os.mkdir()',
1424
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1426
# Now, just copy the existing cached tree to the new location
1427
# We use a cheap trick here.
1428
# Absolute paths are prefixed with the first parameter
1429
# relative paths are prefixed with the second.
1430
# So we can get both the source and target returned
1431
# without any extra work.
1433
def copy_dir(source, dest):
1436
def copy_link(source, dest):
1437
"""Copy the contents of a symlink"""
1438
link_to = os.readlink(source)
1439
os.symlink(link_to, dest)
1441
real_handlers = {'file':shutil.copy2,
1442
'symlink':copy_link,
1443
'directory':copy_dir,
1445
real_handlers.update(handlers)
1447
if not os.path.exists(to_path):
1448
real_handlers['directory'](from_path, to_path)
1450
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1451
for relpath, name, kind, st, abspath in entries:
1452
real_handlers[kind](abspath, relpath)
1455
def path_prefix_key(path):
1456
"""Generate a prefix-order path key for path.
1458
This can be used to sort paths in the same way that walkdirs does.
1460
return (dirname(path) , path)
1463
def compare_paths_prefix_order(path_a, path_b):
1464
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1465
key_a = path_prefix_key(path_a)
1466
key_b = path_prefix_key(path_b)
1467
return cmp(key_a, key_b)
1470
_cached_user_encoding = None
1473
def get_user_encoding(use_cache=True):
1474
"""Find out what the preferred user encoding is.
1476
This is generally the encoding that is used for command line parameters
1477
and file contents. This may be different from the terminal encoding
1478
or the filesystem encoding.
1480
:param use_cache: Enable cache for detected encoding.
1481
(This parameter is turned on by default,
1482
and required only for selftesting)
1484
:return: A string defining the preferred user encoding
1486
global _cached_user_encoding
1487
if _cached_user_encoding is not None and use_cache:
1488
return _cached_user_encoding
1490
if sys.platform == 'darwin':
1491
# python locale.getpreferredencoding() always return
1492
# 'mac-roman' on darwin. That's a lie.
1493
sys.platform = 'posix'
1495
if os.environ.get('LANG', None) is None:
1496
# If LANG is not set, we end up with 'ascii', which is bad
1497
# ('mac-roman' is more than ascii), so we set a default which
1498
# will give us UTF-8 (which appears to work in all cases on
1499
# OSX). Users are still free to override LANG of course, as
1500
# long as it give us something meaningful. This work-around
1501
# *may* not be needed with python 3k and/or OSX 10.5, but will
1502
# work with them too -- vila 20080908
1503
os.environ['LANG'] = 'en_US.UTF-8'
1506
sys.platform = 'darwin'
1511
user_encoding = locale.getpreferredencoding()
1512
except locale.Error, e:
1513
sys.stderr.write('bzr: warning: %s\n'
1514
' Could not determine what text encoding to use.\n'
1515
' This error usually means your Python interpreter\n'
1516
' doesn\'t support the locale set by $LANG (%s)\n'
1517
" Continuing with ascii encoding.\n"
1518
% (e, os.environ.get('LANG')))
1519
user_encoding = 'ascii'
1521
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1522
# treat that as ASCII, and not support printing unicode characters to the
1525
# For python scripts run under vim, we get '', so also treat that as ASCII
1526
if user_encoding in (None, 'cp0', ''):
1527
user_encoding = 'ascii'
1531
codecs.lookup(user_encoding)
1533
sys.stderr.write('bzr: warning:'
1534
' unknown encoding %s.'
1535
' Continuing with ascii encoding.\n'
1538
user_encoding = 'ascii'
1541
_cached_user_encoding = user_encoding
1543
return user_encoding
1546
def get_host_name():
1547
"""Return the current unicode host name.
1549
This is meant to be used in place of socket.gethostname() because that
1550
behaves inconsistently on different platforms.
1552
if sys.platform == "win32":
1554
return win32utils.get_host_name()
1557
return socket.gethostname().decode(get_user_encoding())
1560
def recv_all(socket, bytes):
1561
"""Receive an exact number of bytes.
1563
Regular Socket.recv() may return less than the requested number of bytes,
1564
dependning on what's in the OS buffer. MSG_WAITALL is not available
1565
on all platforms, but this should work everywhere. This will return
1566
less than the requested amount if the remote end closes.
1568
This isn't optimized and is intended mostly for use in testing.
1571
while len(b) < bytes:
1572
new = socket.recv(bytes - len(b))
1579
def send_all(socket, bytes):
1580
"""Send all bytes on a socket.
1582
Regular socket.sendall() can give socket error 10053 on Windows. This
1583
implementation sends no more than 64k at a time, which avoids this problem.
1586
for pos in xrange(0, len(bytes), chunk_size):
1587
socket.sendall(bytes[pos:pos+chunk_size])
1590
def dereference_path(path):
1591
"""Determine the real path to a file.
1593
All parent elements are dereferenced. But the file itself is not
1595
:param path: The original path. May be absolute or relative.
1596
:return: the real path *to* the file
1598
parent, base = os.path.split(path)
1599
# The pathjoin for '.' is a workaround for Python bug #1213894.
1600
# (initial path components aren't dereferenced)
1601
return pathjoin(realpath(pathjoin('.', parent)), base)
1604
def supports_mapi():
1605
"""Return True if we can use MAPI to launch a mail client."""
1606
return sys.platform == "win32"
1609
def resource_string(package, resource_name):
1610
"""Load a resource from a package and return it as a string.
1612
Note: Only packages that start with bzrlib are currently supported.
1614
This is designed to be a lightweight implementation of resource
1615
loading in a way which is API compatible with the same API from
1617
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1618
If and when pkg_resources becomes a standard library, this routine
1621
# Check package name is within bzrlib
1622
if package == "bzrlib":
1623
resource_relpath = resource_name
1624
elif package.startswith("bzrlib."):
1625
package = package[len("bzrlib."):].replace('.', os.sep)
1626
resource_relpath = pathjoin(package, resource_name)
1628
raise errors.BzrError('resource package %s not in bzrlib' % package)
1630
# Map the resource to a file and read its contents
1631
base = dirname(bzrlib.__file__)
1632
if getattr(sys, 'frozen', None): # bzr.exe
1633
base = abspath(pathjoin(base, '..', '..'))
1634
filename = pathjoin(base, resource_relpath)
1635
return open(filename, 'rU').read()
1638
def file_kind_from_stat_mode_thunk(mode):
1639
global file_kind_from_stat_mode
1640
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
1642
from bzrlib._readdir_pyx import UTF8DirReader
1643
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
1645
from bzrlib._readdir_py import (
1646
_kind_from_mode as file_kind_from_stat_mode
1648
return file_kind_from_stat_mode(mode)
1649
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
1652
def file_kind(f, _lstat=os.lstat):
1654
return file_kind_from_stat_mode(_lstat(f).st_mode)
1656
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
1657
raise errors.NoSuchFile(f)
1660
if sys.platform == "win32":
1663
return msvcrt.getch()
1668
fd = sys.stdin.fileno()
1669
settings = termios.tcgetattr(fd)
1672
ch = sys.stdin.read(1)
1674
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
added
removed
bzrlib/osutils.py
