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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 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
report_activity=None, direction='read'):
513
"""Copy contents of one file to another.
515
The read_length can either be -1 to read to end-of-file (EOF) or
516
it can specify the maximum number of bytes to read.
518
The buff_size represents the maximum size for each read operation
519
performed on from_file.
521
:param report_activity: Call this as bytes are read, see
522
Transport._report_activity
523
:param direction: Will be passed to report_activity
525
:return: The number of bytes copied.
529
# read specified number of bytes
531
while read_length > 0:
532
num_bytes_to_read = min(read_length, buff_size)
534
block = from_file.read(num_bytes_to_read)
538
if report_activity is not None:
539
report_activity(len(block), direction)
542
actual_bytes_read = len(block)
543
read_length -= actual_bytes_read
544
length += actual_bytes_read
548
block = from_file.read(buff_size)
552
if report_activity is not None:
553
report_activity(len(block), direction)
559
def pump_string_file(bytes, file_handle, segment_size=None):
560
"""Write bytes to file_handle in many smaller writes.
562
:param bytes: The string to write.
563
:param file_handle: The file to write to.
565
# Write data in chunks rather than all at once, because very large
566
# writes fail on some platforms (e.g. Windows with SMB mounted
569
segment_size = 5242880 # 5MB
570
segments = range(len(bytes) / segment_size + 1)
571
write = file_handle.write
572
for segment_index in segments:
573
segment = buffer(bytes, segment_index * segment_size, segment_size)
577
def file_iterator(input_file, readsize=32768):
579
b = input_file.read(readsize)
586
"""Calculate the hexdigest of an open file.
588
The file cursor should be already at the start.
600
def size_sha_file(f):
601
"""Calculate the size and hexdigest of an open file.
603
The file cursor should be already at the start and
604
the caller is responsible for closing the file afterwards.
615
return size, s.hexdigest()
618
def sha_file_by_name(fname):
619
"""Calculate the SHA1 of a file by reading the full text"""
621
f = os.open(fname, os.O_RDONLY | O_BINARY)
624
b = os.read(f, 1<<16)
632
def sha_strings(strings, _factory=sha):
633
"""Return the sha-1 of concatenation of strings"""
635
map(s.update, strings)
639
def sha_string(f, _factory=sha):
640
return _factory(f).hexdigest()
643
def fingerprint_file(f):
645
return {'size': len(b),
646
'sha1': sha(b).hexdigest()}
649
def compare_files(a, b):
650
"""Returns true if equal in contents"""
661
def local_time_offset(t=None):
662
"""Return offset of local zone from GMT, either at present or at time t."""
665
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
666
return offset.days * 86400 + offset.seconds
668
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
670
def format_date(t, offset=0, timezone='original', date_fmt=None,
672
"""Return a formatted date string.
674
:param t: Seconds since the epoch.
675
:param offset: Timezone offset in seconds east of utc.
676
:param timezone: How to display the time: 'utc', 'original' for the
677
timezone specified by offset, or 'local' for the process's current
679
:param date_fmt: strftime format.
680
:param show_offset: Whether to append the timezone.
682
(date_fmt, tt, offset_str) = \
683
_format_date(t, offset, timezone, date_fmt, show_offset)
684
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
685
date_str = time.strftime(date_fmt, tt)
686
return date_str + offset_str
688
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
690
"""Return an unicode date string formatted according to the current locale.
692
:param t: Seconds since the epoch.
693
:param offset: Timezone offset in seconds east of utc.
694
:param timezone: How to display the time: 'utc', 'original' for the
695
timezone specified by offset, or 'local' for the process's current
697
:param date_fmt: strftime format.
698
:param show_offset: Whether to append the timezone.
700
(date_fmt, tt, offset_str) = \
701
_format_date(t, offset, timezone, date_fmt, show_offset)
702
date_str = time.strftime(date_fmt, tt)
703
if not isinstance(date_str, unicode):
704
date_str = date_str.decode(bzrlib.user_encoding, 'replace')
705
return date_str + offset_str
707
def _format_date(t, offset, timezone, date_fmt, show_offset):
708
if timezone == 'utc':
711
elif timezone == 'original':
714
tt = time.gmtime(t + offset)
715
elif timezone == 'local':
716
tt = time.localtime(t)
717
offset = local_time_offset(t)
719
raise errors.UnsupportedTimezoneFormat(timezone)
721
date_fmt = "%a %Y-%m-%d %H:%M:%S"
723
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
726
return (date_fmt, tt, offset_str)
729
def compact_date(when):
730
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
733
def format_delta(delta):
734
"""Get a nice looking string for a time delta.
736
:param delta: The time difference in seconds, can be positive or negative.
737
positive indicates time in the past, negative indicates time in the
738
future. (usually time.time() - stored_time)
739
:return: String formatted to show approximate resolution
745
direction = 'in the future'
749
if seconds < 90: # print seconds up to 90 seconds
751
return '%d second %s' % (seconds, direction,)
753
return '%d seconds %s' % (seconds, direction)
755
minutes = int(seconds / 60)
756
seconds -= 60 * minutes
761
if minutes < 90: # print minutes, seconds up to 90 minutes
763
return '%d minute, %d second%s %s' % (
764
minutes, seconds, plural_seconds, direction)
766
return '%d minutes, %d second%s %s' % (
767
minutes, seconds, plural_seconds, direction)
769
hours = int(minutes / 60)
770
minutes -= 60 * hours
777
return '%d hour, %d minute%s %s' % (hours, minutes,
778
plural_minutes, direction)
779
return '%d hours, %d minute%s %s' % (hours, minutes,
780
plural_minutes, direction)
783
"""Return size of given open file."""
784
return os.fstat(f.fileno())[ST_SIZE]
787
# Define rand_bytes based on platform.
789
# Python 2.4 and later have os.urandom,
790
# but it doesn't work on some arches
792
rand_bytes = os.urandom
793
except (NotImplementedError, AttributeError):
794
# If python doesn't have os.urandom, or it doesn't work,
795
# then try to first pull random data from /dev/urandom
797
rand_bytes = file('/dev/urandom', 'rb').read
798
# Otherwise, use this hack as a last resort
799
except (IOError, OSError):
800
# not well seeded, but better than nothing
805
s += chr(random.randint(0, 255))
810
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
812
"""Return a random string of num alphanumeric characters
814
The result only contains lowercase chars because it may be used on
815
case-insensitive filesystems.
818
for raw_byte in rand_bytes(num):
819
s += ALNUM[ord(raw_byte) % 36]
823
## TODO: We could later have path objects that remember their list
824
## decomposition (might be too tricksy though.)
827
"""Turn string into list of parts."""
828
# split on either delimiter because people might use either on
830
ps = re.split(r'[\\/]', p)
835
raise errors.BzrError("sorry, %r not allowed in path" % f)
836
elif (f == '.') or (f == ''):
845
if (f == '..') or (f is None) or (f == ''):
846
raise errors.BzrError("sorry, %r not allowed in path" % f)
851
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
853
from bzrlib._chunks_to_lines_py import chunks_to_lines
857
"""Split s into lines, but without removing the newline characters."""
858
# Trivially convert a fulltext into a 'chunked' representation, and let
859
# chunks_to_lines do the heavy lifting.
860
if isinstance(s, str):
861
# chunks_to_lines only supports 8-bit strings
862
return chunks_to_lines([s])
864
return _split_lines(s)
868
"""Split s into lines, but without removing the newline characters.
870
This supports Unicode or plain string objects.
872
lines = s.split('\n')
873
result = [line + '\n' for line in lines[:-1]]
875
result.append(lines[-1])
879
def hardlinks_good():
880
return sys.platform not in ('win32', 'cygwin', 'darwin')
883
def link_or_copy(src, dest):
884
"""Hardlink a file, or copy it if it can't be hardlinked."""
885
if not hardlinks_good():
886
shutil.copyfile(src, dest)
890
except (OSError, IOError), e:
891
if e.errno != errno.EXDEV:
893
shutil.copyfile(src, dest)
896
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
897
# Forgiveness than Permission (EAFP) because:
898
# - root can damage a solaris file system by using unlink,
899
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
900
# EACCES, OSX: EPERM) when invoked on a directory.
901
def delete_any(path):
902
"""Delete a file or directory."""
903
if isdir(path): # Takes care of symlinks
910
if getattr(os, 'symlink', None) is not None:
917
if getattr(os, 'link', None) is not None:
923
def host_os_dereferences_symlinks():
924
return (has_symlinks()
925
and sys.platform not in ('cygwin', 'win32'))
928
def contains_whitespace(s):
929
"""True if there are any whitespace characters in s."""
930
# string.whitespace can include '\xa0' in certain locales, because it is
931
# considered "non-breaking-space" as part of ISO-8859-1. But it
932
# 1) Isn't a breaking whitespace
933
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
935
# 3) '\xa0' isn't unicode safe since it is >128.
937
# This should *not* be a unicode set of characters in case the source
938
# string is not a Unicode string. We can auto-up-cast the characters since
939
# they are ascii, but we don't want to auto-up-cast the string in case it
941
for ch in ' \t\n\r\v\f':
948
def contains_linebreaks(s):
949
"""True if there is any vertical whitespace in s."""
957
def relpath(base, path):
958
"""Return path relative to base, or raise exception.
960
The path may be either an absolute path or a path relative to the
961
current working directory.
963
os.path.commonprefix (python2.4) has a bad bug that it works just
964
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
968
if len(base) < MIN_ABS_PATHLENGTH:
969
# must have space for e.g. a drive letter
970
raise ValueError('%r is too short to calculate a relative path'
977
while len(head) >= len(base):
980
head, tail = os.path.split(head)
984
raise errors.PathNotChild(rp, base)
992
def _cicp_canonical_relpath(base, path):
993
"""Return the canonical path relative to base.
995
Like relpath, but on case-insensitive-case-preserving file-systems, this
996
will return the relpath as stored on the file-system rather than in the
997
case specified in the input string, for all existing portions of the path.
999
This will cause O(N) behaviour if called for every path in a tree; if you
1000
have a number of paths to convert, you should use canonical_relpaths().
1002
# TODO: it should be possible to optimize this for Windows by using the
1003
# win32 API FindFiles function to look for the specified name - but using
1004
# os.listdir() still gives us the correct, platform agnostic semantics in
1007
rel = relpath(base, path)
1008
# '.' will have been turned into ''
1012
abs_base = abspath(base)
1014
_listdir = os.listdir
1016
# use an explicit iterator so we can easily consume the rest on early exit.
1017
bit_iter = iter(rel.split('/'))
1018
for bit in bit_iter:
1020
for look in _listdir(current):
1021
if lbit == look.lower():
1022
current = pathjoin(current, look)
1025
# got to the end, nothing matched, so we just return the
1026
# non-existing bits as they were specified (the filename may be
1027
# the target of a move, for example).
1028
current = pathjoin(current, bit, *list(bit_iter))
1030
return current[len(abs_base)+1:]
1032
# XXX - TODO - we need better detection/integration of case-insensitive
1033
# file-systems; Linux often sees FAT32 devices, for example, so could
1034
# probably benefit from the same basic support there. For now though, only
1035
# Windows gets that support, and it gets it for *all* file-systems!
1036
if sys.platform == "win32":
1037
canonical_relpath = _cicp_canonical_relpath
1039
canonical_relpath = relpath
1041
def canonical_relpaths(base, paths):
1042
"""Create an iterable to canonicalize a sequence of relative paths.
1044
The intent is for this implementation to use a cache, vastly speeding
1045
up multiple transformations in the same directory.
1047
# but for now, we haven't optimized...
1048
return [canonical_relpath(base, p) for p in paths]
1050
def safe_unicode(unicode_or_utf8_string):
1051
"""Coerce unicode_or_utf8_string into unicode.
1053
If it is unicode, it is returned.
1054
Otherwise it is decoded from utf-8. If a decoding error
1055
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
1056
as a BzrBadParameter exception.
1058
if isinstance(unicode_or_utf8_string, unicode):
1059
return unicode_or_utf8_string
1061
return unicode_or_utf8_string.decode('utf8')
1062
except UnicodeDecodeError:
1063
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1066
def safe_utf8(unicode_or_utf8_string):
1067
"""Coerce unicode_or_utf8_string to a utf8 string.
1069
If it is a str, it is returned.
1070
If it is Unicode, it is encoded into a utf-8 string.
1072
if isinstance(unicode_or_utf8_string, str):
1073
# TODO: jam 20070209 This is overkill, and probably has an impact on
1074
# performance if we are dealing with lots of apis that want a
1077
# Make sure it is a valid utf-8 string
1078
unicode_or_utf8_string.decode('utf-8')
1079
except UnicodeDecodeError:
1080
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1081
return unicode_or_utf8_string
1082
return unicode_or_utf8_string.encode('utf-8')
1085
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1086
' Revision id generators should be creating utf8'
1090
def safe_revision_id(unicode_or_utf8_string, warn=True):
1091
"""Revision ids should now be utf8, but at one point they were unicode.
1093
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1095
:param warn: Functions that are sanitizing user data can set warn=False
1096
:return: None or a utf8 revision id.
1098
if (unicode_or_utf8_string is None
1099
or unicode_or_utf8_string.__class__ == str):
1100
return unicode_or_utf8_string
1102
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1104
return cache_utf8.encode(unicode_or_utf8_string)
1107
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1108
' generators should be creating utf8 file ids.')
1111
def safe_file_id(unicode_or_utf8_string, warn=True):
1112
"""File ids should now be utf8, but at one point they were unicode.
1114
This is the same as safe_utf8, except it uses the cached encode functions
1115
to save a little bit of performance.
1117
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1119
:param warn: Functions that are sanitizing user data can set warn=False
1120
:return: None or a utf8 file id.
1122
if (unicode_or_utf8_string is None
1123
or unicode_or_utf8_string.__class__ == str):
1124
return unicode_or_utf8_string
1126
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1128
return cache_utf8.encode(unicode_or_utf8_string)
1131
_platform_normalizes_filenames = False
1132
if sys.platform == 'darwin':
1133
_platform_normalizes_filenames = True
1136
def normalizes_filenames():
1137
"""Return True if this platform normalizes unicode filenames.
1139
Mac OSX does, Windows/Linux do not.
1141
return _platform_normalizes_filenames
1144
def _accessible_normalized_filename(path):
1145
"""Get the unicode normalized path, and if you can access the file.
1147
On platforms where the system normalizes filenames (Mac OSX),
1148
you can access a file by any path which will normalize correctly.
1149
On platforms where the system does not normalize filenames
1150
(Windows, Linux), you have to access a file by its exact path.
1152
Internally, bzr only supports NFC normalization, since that is
1153
the standard for XML documents.
1155
So return the normalized path, and a flag indicating if the file
1156
can be accessed by that path.
1159
return unicodedata.normalize('NFC', unicode(path)), True
1162
def _inaccessible_normalized_filename(path):
1163
__doc__ = _accessible_normalized_filename.__doc__
1165
normalized = unicodedata.normalize('NFC', unicode(path))
1166
return normalized, normalized == path
1169
if _platform_normalizes_filenames:
1170
normalized_filename = _accessible_normalized_filename
1172
normalized_filename = _inaccessible_normalized_filename
1175
def terminal_width():
1176
"""Return estimated terminal width."""
1177
if sys.platform == 'win32':
1178
return win32utils.get_console_size()[0]
1181
import struct, fcntl, termios
1182
s = struct.pack('HHHH', 0, 0, 0, 0)
1183
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1184
width = struct.unpack('HHHH', x)[1]
1189
width = int(os.environ['COLUMNS'])
1198
def supports_executable():
1199
return sys.platform != "win32"
1202
def supports_posix_readonly():
1203
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1205
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1206
directory controls creation/deletion, etc.
1208
And under win32, readonly means that the directory itself cannot be
1209
deleted. The contents of a readonly directory can be changed, unlike POSIX
1210
where files in readonly directories cannot be added, deleted or renamed.
1212
return sys.platform != "win32"
1215
def set_or_unset_env(env_variable, value):
1216
"""Modify the environment, setting or removing the env_variable.
1218
:param env_variable: The environment variable in question
1219
:param value: The value to set the environment to. If None, then
1220
the variable will be removed.
1221
:return: The original value of the environment variable.
1223
orig_val = os.environ.get(env_variable)
1225
if orig_val is not None:
1226
del os.environ[env_variable]
1228
if isinstance(value, unicode):
1229
value = value.encode(get_user_encoding())
1230
os.environ[env_variable] = value
1234
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1237
def check_legal_path(path):
1238
"""Check whether the supplied path is legal.
1239
This is only required on Windows, so we don't test on other platforms
1242
if sys.platform != "win32":
1244
if _validWin32PathRE.match(path) is None:
1245
raise errors.IllegalPath(path)
1248
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1250
def _is_error_enotdir(e):
1251
"""Check if this exception represents ENOTDIR.
1253
Unfortunately, python is very inconsistent about the exception
1254
here. The cases are:
1255
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1256
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1257
which is the windows error code.
1258
3) Windows, Python2.5 uses errno == EINVAL and
1259
winerror == ERROR_DIRECTORY
1261
:param e: An Exception object (expected to be OSError with an errno
1262
attribute, but we should be able to cope with anything)
1263
:return: True if this represents an ENOTDIR error. False otherwise.
1265
en = getattr(e, 'errno', None)
1266
if (en == errno.ENOTDIR
1267
or (sys.platform == 'win32'
1268
and (en == _WIN32_ERROR_DIRECTORY
1269
or (en == errno.EINVAL
1270
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1276
def walkdirs(top, prefix=""):
1277
"""Yield data about all the directories in a tree.
1279
This yields all the data about the contents of a directory at a time.
1280
After each directory has been yielded, if the caller has mutated the list
1281
to exclude some directories, they are then not descended into.
1283
The data yielded is of the form:
1284
((directory-relpath, directory-path-from-top),
1285
[(relpath, basename, kind, lstat, path-from-top), ...]),
1286
- directory-relpath is the relative path of the directory being returned
1287
with respect to top. prefix is prepended to this.
1288
- directory-path-from-root is the path including top for this directory.
1289
It is suitable for use with os functions.
1290
- relpath is the relative path within the subtree being walked.
1291
- basename is the basename of the path
1292
- kind is the kind of the file now. If unknown then the file is not
1293
present within the tree - but it may be recorded as versioned. See
1295
- lstat is the stat data *if* the file was statted.
1296
- planned, not implemented:
1297
path_from_tree_root is the path from the root of the tree.
1299
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1300
allows one to walk a subtree but get paths that are relative to a tree
1302
:return: an iterator over the dirs.
1304
#TODO there is a bit of a smell where the results of the directory-
1305
# summary in this, and the path from the root, may not agree
1306
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1307
# potentially confusing output. We should make this more robust - but
1308
# not at a speed cost. RBC 20060731
1310
_directory = _directory_kind
1311
_listdir = os.listdir
1312
_kind_from_mode = file_kind_from_stat_mode
1313
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1315
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1316
relroot, _, _, _, top = pending.pop()
1318
relprefix = relroot + u'/'
1321
top_slash = top + u'/'
1324
append = dirblock.append
1326
names = sorted(_listdir(top))
1328
if not _is_error_enotdir(e):
1332
abspath = top_slash + name
1333
statvalue = _lstat(abspath)
1334
kind = _kind_from_mode(statvalue.st_mode)
1335
append((relprefix + name, name, kind, statvalue, abspath))
1336
yield (relroot, top), dirblock
1338
# push the user specified dirs from dirblock
1339
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1342
class DirReader(object):
1343
"""An interface for reading directories."""
1345
def top_prefix_to_starting_dir(self, top, prefix=""):
1346
"""Converts top and prefix to a starting dir entry
1348
:param top: A utf8 path
1349
:param prefix: An optional utf8 path to prefix output relative paths
1351
:return: A tuple starting with prefix, and ending with the native
1354
raise NotImplementedError(self.top_prefix_to_starting_dir)
1356
def read_dir(self, prefix, top):
1357
"""Read a specific dir.
1359
:param prefix: A utf8 prefix to be preprended to the path basenames.
1360
:param top: A natively encoded path to read.
1361
:return: A list of the directories contents. Each item contains:
1362
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1364
raise NotImplementedError(self.read_dir)
1367
_selected_dir_reader = None
1370
def _walkdirs_utf8(top, prefix=""):
1371
"""Yield data about all the directories in a tree.
1373
This yields the same information as walkdirs() only each entry is yielded
1374
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1375
are returned as exact byte-strings.
1377
:return: yields a tuple of (dir_info, [file_info])
1378
dir_info is (utf8_relpath, path-from-top)
1379
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1380
if top is an absolute path, path-from-top is also an absolute path.
1381
path-from-top might be unicode or utf8, but it is the correct path to
1382
pass to os functions to affect the file in question. (such as os.lstat)
1384
global _selected_dir_reader
1385
if _selected_dir_reader is None:
1386
fs_encoding = _fs_enc.upper()
1387
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1388
# Win98 doesn't have unicode apis like FindFirstFileW
1389
# TODO: We possibly could support Win98 by falling back to the
1390
# original FindFirstFile, and using TCHAR instead of WCHAR,
1391
# but that gets a bit tricky, and requires custom compiling
1394
from bzrlib._walkdirs_win32 import Win32ReadDir
1396
_selected_dir_reader = UnicodeDirReader()
1398
_selected_dir_reader = Win32ReadDir()
1399
elif fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1400
# ANSI_X3.4-1968 is a form of ASCII
1401
_selected_dir_reader = UnicodeDirReader()
1404
from bzrlib._readdir_pyx import UTF8DirReader
1406
# No optimised code path
1407
_selected_dir_reader = UnicodeDirReader()
1409
_selected_dir_reader = UTF8DirReader()
1410
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1411
# But we don't actually uses 1-3 in pending, so set them to None
1412
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1413
read_dir = _selected_dir_reader.read_dir
1414
_directory = _directory_kind
1416
relroot, _, _, _, top = pending[-1].pop()
1419
dirblock = sorted(read_dir(relroot, top))
1420
yield (relroot, top), dirblock
1421
# push the user specified dirs from dirblock
1422
next = [d for d in reversed(dirblock) if d[2] == _directory]
1424
pending.append(next)
1427
class UnicodeDirReader(DirReader):
1428
"""A dir reader for non-utf8 file systems, which transcodes."""
1430
__slots__ = ['_utf8_encode']
1433
self._utf8_encode = codecs.getencoder('utf8')
1435
def top_prefix_to_starting_dir(self, top, prefix=""):
1436
"""See DirReader.top_prefix_to_starting_dir."""
1437
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1439
def read_dir(self, prefix, top):
1440
"""Read a single directory from a non-utf8 file system.
1442
top, and the abspath element in the output are unicode, all other paths
1443
are utf8. Local disk IO is done via unicode calls to listdir etc.
1445
This is currently the fallback code path when the filesystem encoding is
1446
not UTF-8. It may be better to implement an alternative so that we can
1447
safely handle paths that are not properly decodable in the current
1450
See DirReader.read_dir for details.
1452
_utf8_encode = self._utf8_encode
1454
_listdir = os.listdir
1455
_kind_from_mode = file_kind_from_stat_mode
1458
relprefix = prefix + '/'
1461
top_slash = top + u'/'
1464
append = dirblock.append
1465
for name in sorted(_listdir(top)):
1467
name_utf8 = _utf8_encode(name)[0]
1468
except UnicodeDecodeError:
1469
raise errors.BadFilenameEncoding(
1470
_utf8_encode(relprefix)[0] + name, _fs_enc)
1471
abspath = top_slash + name
1472
statvalue = _lstat(abspath)
1473
kind = _kind_from_mode(statvalue.st_mode)
1474
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1478
def copy_tree(from_path, to_path, handlers={}):
1479
"""Copy all of the entries in from_path into to_path.
1481
:param from_path: The base directory to copy.
1482
:param to_path: The target directory. If it does not exist, it will
1484
:param handlers: A dictionary of functions, which takes a source and
1485
destinations for files, directories, etc.
1486
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1487
'file', 'directory', and 'symlink' should always exist.
1488
If they are missing, they will be replaced with 'os.mkdir()',
1489
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1491
# Now, just copy the existing cached tree to the new location
1492
# We use a cheap trick here.
1493
# Absolute paths are prefixed with the first parameter
1494
# relative paths are prefixed with the second.
1495
# So we can get both the source and target returned
1496
# without any extra work.
1498
def copy_dir(source, dest):
1501
def copy_link(source, dest):
1502
"""Copy the contents of a symlink"""
1503
link_to = os.readlink(source)
1504
os.symlink(link_to, dest)
1506
real_handlers = {'file':shutil.copy2,
1507
'symlink':copy_link,
1508
'directory':copy_dir,
1510
real_handlers.update(handlers)
1512
if not os.path.exists(to_path):
1513
real_handlers['directory'](from_path, to_path)
1515
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1516
for relpath, name, kind, st, abspath in entries:
1517
real_handlers[kind](abspath, relpath)
1520
def path_prefix_key(path):
1521
"""Generate a prefix-order path key for path.
1523
This can be used to sort paths in the same way that walkdirs does.
1525
return (dirname(path) , path)
1528
def compare_paths_prefix_order(path_a, path_b):
1529
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1530
key_a = path_prefix_key(path_a)
1531
key_b = path_prefix_key(path_b)
1532
return cmp(key_a, key_b)
1535
_cached_user_encoding = None
1538
def get_user_encoding(use_cache=True):
1539
"""Find out what the preferred user encoding is.
1541
This is generally the encoding that is used for command line parameters
1542
and file contents. This may be different from the terminal encoding
1543
or the filesystem encoding.
1545
:param use_cache: Enable cache for detected encoding.
1546
(This parameter is turned on by default,
1547
and required only for selftesting)
1549
:return: A string defining the preferred user encoding
1551
global _cached_user_encoding
1552
if _cached_user_encoding is not None and use_cache:
1553
return _cached_user_encoding
1555
if sys.platform == 'darwin':
1556
# python locale.getpreferredencoding() always return
1557
# 'mac-roman' on darwin. That's a lie.
1558
sys.platform = 'posix'
1560
if os.environ.get('LANG', None) is None:
1561
# If LANG is not set, we end up with 'ascii', which is bad
1562
# ('mac-roman' is more than ascii), so we set a default which
1563
# will give us UTF-8 (which appears to work in all cases on
1564
# OSX). Users are still free to override LANG of course, as
1565
# long as it give us something meaningful. This work-around
1566
# *may* not be needed with python 3k and/or OSX 10.5, but will
1567
# work with them too -- vila 20080908
1568
os.environ['LANG'] = 'en_US.UTF-8'
1571
sys.platform = 'darwin'
1576
user_encoding = locale.getpreferredencoding()
1577
except locale.Error, e:
1578
sys.stderr.write('bzr: warning: %s\n'
1579
' Could not determine what text encoding to use.\n'
1580
' This error usually means your Python interpreter\n'
1581
' doesn\'t support the locale set by $LANG (%s)\n'
1582
" Continuing with ascii encoding.\n"
1583
% (e, os.environ.get('LANG')))
1584
user_encoding = 'ascii'
1586
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1587
# treat that as ASCII, and not support printing unicode characters to the
1590
# For python scripts run under vim, we get '', so also treat that as ASCII
1591
if user_encoding in (None, 'cp0', ''):
1592
user_encoding = 'ascii'
1596
codecs.lookup(user_encoding)
1598
sys.stderr.write('bzr: warning:'
1599
' unknown encoding %s.'
1600
' Continuing with ascii encoding.\n'
1603
user_encoding = 'ascii'
1606
_cached_user_encoding = user_encoding
1608
return user_encoding
1611
def get_host_name():
1612
"""Return the current unicode host name.
1614
This is meant to be used in place of socket.gethostname() because that
1615
behaves inconsistently on different platforms.
1617
if sys.platform == "win32":
1619
return win32utils.get_host_name()
1622
return socket.gethostname().decode(get_user_encoding())
1625
def recv_all(socket, bytes):
1626
"""Receive an exact number of bytes.
1628
Regular Socket.recv() may return less than the requested number of bytes,
1629
dependning on what's in the OS buffer. MSG_WAITALL is not available
1630
on all platforms, but this should work everywhere. This will return
1631
less than the requested amount if the remote end closes.
1633
This isn't optimized and is intended mostly for use in testing.
1636
while len(b) < bytes:
1637
new = until_no_eintr(socket.recv, bytes - len(b))
1644
def send_all(socket, bytes, report_activity=None):
1645
"""Send all bytes on a socket.
1647
Regular socket.sendall() can give socket error 10053 on Windows. This
1648
implementation sends no more than 64k at a time, which avoids this problem.
1650
:param report_activity: Call this as bytes are read, see
1651
Transport._report_activity
1654
for pos in xrange(0, len(bytes), chunk_size):
1655
block = bytes[pos:pos+chunk_size]
1656
if report_activity is not None:
1657
report_activity(len(block), 'write')
1658
until_no_eintr(socket.sendall, block)
1661
def dereference_path(path):
1662
"""Determine the real path to a file.
1664
All parent elements are dereferenced. But the file itself is not
1666
:param path: The original path. May be absolute or relative.
1667
:return: the real path *to* the file
1669
parent, base = os.path.split(path)
1670
# The pathjoin for '.' is a workaround for Python bug #1213894.
1671
# (initial path components aren't dereferenced)
1672
return pathjoin(realpath(pathjoin('.', parent)), base)
1675
def supports_mapi():
1676
"""Return True if we can use MAPI to launch a mail client."""
1677
return sys.platform == "win32"
1680
def resource_string(package, resource_name):
1681
"""Load a resource from a package and return it as a string.
1683
Note: Only packages that start with bzrlib are currently supported.
1685
This is designed to be a lightweight implementation of resource
1686
loading in a way which is API compatible with the same API from
1688
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1689
If and when pkg_resources becomes a standard library, this routine
1692
# Check package name is within bzrlib
1693
if package == "bzrlib":
1694
resource_relpath = resource_name
1695
elif package.startswith("bzrlib."):
1696
package = package[len("bzrlib."):].replace('.', os.sep)
1697
resource_relpath = pathjoin(package, resource_name)
1699
raise errors.BzrError('resource package %s not in bzrlib' % package)
1701
# Map the resource to a file and read its contents
1702
base = dirname(bzrlib.__file__)
1703
if getattr(sys, 'frozen', None): # bzr.exe
1704
base = abspath(pathjoin(base, '..', '..'))
1705
filename = pathjoin(base, resource_relpath)
1706
return open(filename, 'rU').read()
1709
def file_kind_from_stat_mode_thunk(mode):
1710
global file_kind_from_stat_mode
1711
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
1713
from bzrlib._readdir_pyx import UTF8DirReader
1714
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
1716
from bzrlib._readdir_py import (
1717
_kind_from_mode as file_kind_from_stat_mode
1719
return file_kind_from_stat_mode(mode)
1720
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
1723
def file_kind(f, _lstat=os.lstat):
1725
return file_kind_from_stat_mode(_lstat(f).st_mode)
1727
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
1728
raise errors.NoSuchFile(f)
1732
def until_no_eintr(f, *a, **kw):
1733
"""Run f(*a, **kw), retrying if an EINTR error occurs."""
1734
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
1738
except (IOError, OSError), e:
1739
if e.errno == errno.EINTR:
1744
if sys.platform == "win32":
1747
return msvcrt.getch()
1752
fd = sys.stdin.fileno()
1753
settings = termios.tcgetattr(fd)
1756
ch = sys.stdin.read(1)
1758
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
added
removed
bzrlib/osutils.py
