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)
824
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
826
from bzrlib._chunks_to_lines_py import chunks_to_lines
830
"""Split s into lines, but without removing the newline characters."""
831
# Trivially convert a fulltext into a 'chunked' representation, and let
832
# chunks_to_lines do the heavy lifting.
833
if isinstance(s, str):
834
# chunks_to_lines only supports 8-bit strings
835
return chunks_to_lines([s])
837
return _split_lines(s)
841
"""Split s into lines, but without removing the newline characters.
843
This supports Unicode or plain string objects.
845
lines = s.split('\n')
846
result = [line + '\n' for line in lines[:-1]]
848
result.append(lines[-1])
852
def hardlinks_good():
853
return sys.platform not in ('win32', 'cygwin', 'darwin')
856
def link_or_copy(src, dest):
857
"""Hardlink a file, or copy it if it can't be hardlinked."""
858
if not hardlinks_good():
859
shutil.copyfile(src, dest)
863
except (OSError, IOError), e:
864
if e.errno != errno.EXDEV:
866
shutil.copyfile(src, dest)
869
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
870
# Forgiveness than Permission (EAFP) because:
871
# - root can damage a solaris file system by using unlink,
872
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
873
# EACCES, OSX: EPERM) when invoked on a directory.
874
def delete_any(path):
875
"""Delete a file or directory."""
876
if isdir(path): # Takes care of symlinks
883
if getattr(os, 'symlink', None) is not None:
890
if getattr(os, 'link', None) is not None:
896
def host_os_dereferences_symlinks():
897
return (has_symlinks()
898
and sys.platform not in ('cygwin', 'win32'))
901
def contains_whitespace(s):
902
"""True if there are any whitespace characters in s."""
903
# string.whitespace can include '\xa0' in certain locales, because it is
904
# considered "non-breaking-space" as part of ISO-8859-1. But it
905
# 1) Isn't a breaking whitespace
906
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
908
# 3) '\xa0' isn't unicode safe since it is >128.
910
# This should *not* be a unicode set of characters in case the source
911
# string is not a Unicode string. We can auto-up-cast the characters since
912
# they are ascii, but we don't want to auto-up-cast the string in case it
914
for ch in ' \t\n\r\v\f':
921
def contains_linebreaks(s):
922
"""True if there is any vertical whitespace in s."""
930
def relpath(base, path):
931
"""Return path relative to base, or raise exception.
933
The path may be either an absolute path or a path relative to the
934
current working directory.
936
os.path.commonprefix (python2.4) has a bad bug that it works just
937
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
941
if len(base) < MIN_ABS_PATHLENGTH:
942
# must have space for e.g. a drive letter
943
raise ValueError('%r is too short to calculate a relative path'
950
while len(head) >= len(base):
953
head, tail = os.path.split(head)
957
raise errors.PathNotChild(rp, base)
965
def _cicp_canonical_relpath(base, path):
966
"""Return the canonical path relative to base.
968
Like relpath, but on case-insensitive-case-preserving file-systems, this
969
will return the relpath as stored on the file-system rather than in the
970
case specified in the input string, for all existing portions of the path.
972
This will cause O(N) behaviour if called for every path in a tree; if you
973
have a number of paths to convert, you should use canonical_relpaths().
975
# TODO: it should be possible to optimize this for Windows by using the
976
# win32 API FindFiles function to look for the specified name - but using
977
# os.listdir() still gives us the correct, platform agnostic semantics in
980
rel = relpath(base, path)
981
# '.' will have been turned into ''
985
abs_base = abspath(base)
987
_listdir = os.listdir
989
# use an explicit iterator so we can easily consume the rest on early exit.
990
bit_iter = iter(rel.split('/'))
993
for look in _listdir(current):
994
if lbit == look.lower():
995
current = pathjoin(current, look)
998
# got to the end, nothing matched, so we just return the
999
# non-existing bits as they were specified (the filename may be
1000
# the target of a move, for example).
1001
current = pathjoin(current, bit, *list(bit_iter))
1003
return current[len(abs_base)+1:]
1005
# XXX - TODO - we need better detection/integration of case-insensitive
1006
# file-systems; Linux often sees FAT32 devices, for example, so could
1007
# probably benefit from the same basic support there. For now though, only
1008
# Windows gets that support, and it gets it for *all* file-systems!
1009
if sys.platform == "win32":
1010
canonical_relpath = _cicp_canonical_relpath
1012
canonical_relpath = relpath
1014
def canonical_relpaths(base, paths):
1015
"""Create an iterable to canonicalize a sequence of relative paths.
1017
The intent is for this implementation to use a cache, vastly speeding
1018
up multiple transformations in the same directory.
1020
# but for now, we haven't optimized...
1021
return [canonical_relpath(base, p) for p in paths]
1023
def safe_unicode(unicode_or_utf8_string):
1024
"""Coerce unicode_or_utf8_string into unicode.
1026
If it is unicode, it is returned.
1027
Otherwise it is decoded from utf-8. If a decoding error
1028
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
1029
as a BzrBadParameter exception.
1031
if isinstance(unicode_or_utf8_string, unicode):
1032
return unicode_or_utf8_string
1034
return unicode_or_utf8_string.decode('utf8')
1035
except UnicodeDecodeError:
1036
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1039
def safe_utf8(unicode_or_utf8_string):
1040
"""Coerce unicode_or_utf8_string to a utf8 string.
1042
If it is a str, it is returned.
1043
If it is Unicode, it is encoded into a utf-8 string.
1045
if isinstance(unicode_or_utf8_string, str):
1046
# TODO: jam 20070209 This is overkill, and probably has an impact on
1047
# performance if we are dealing with lots of apis that want a
1050
# Make sure it is a valid utf-8 string
1051
unicode_or_utf8_string.decode('utf-8')
1052
except UnicodeDecodeError:
1053
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1054
return unicode_or_utf8_string
1055
return unicode_or_utf8_string.encode('utf-8')
1058
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1059
' Revision id generators should be creating utf8'
1063
def safe_revision_id(unicode_or_utf8_string, warn=True):
1064
"""Revision ids should now be utf8, but at one point they were unicode.
1066
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1068
:param warn: Functions that are sanitizing user data can set warn=False
1069
:return: None or a utf8 revision id.
1071
if (unicode_or_utf8_string is None
1072
or unicode_or_utf8_string.__class__ == str):
1073
return unicode_or_utf8_string
1075
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1077
return cache_utf8.encode(unicode_or_utf8_string)
1080
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1081
' generators should be creating utf8 file ids.')
1084
def safe_file_id(unicode_or_utf8_string, warn=True):
1085
"""File ids should now be utf8, but at one point they were unicode.
1087
This is the same as safe_utf8, except it uses the cached encode functions
1088
to save a little bit of performance.
1090
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1092
:param warn: Functions that are sanitizing user data can set warn=False
1093
:return: None or a utf8 file id.
1095
if (unicode_or_utf8_string is None
1096
or unicode_or_utf8_string.__class__ == str):
1097
return unicode_or_utf8_string
1099
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1101
return cache_utf8.encode(unicode_or_utf8_string)
1104
_platform_normalizes_filenames = False
1105
if sys.platform == 'darwin':
1106
_platform_normalizes_filenames = True
1109
def normalizes_filenames():
1110
"""Return True if this platform normalizes unicode filenames.
1112
Mac OSX does, Windows/Linux do not.
1114
return _platform_normalizes_filenames
1117
def _accessible_normalized_filename(path):
1118
"""Get the unicode normalized path, and if you can access the file.
1120
On platforms where the system normalizes filenames (Mac OSX),
1121
you can access a file by any path which will normalize correctly.
1122
On platforms where the system does not normalize filenames
1123
(Windows, Linux), you have to access a file by its exact path.
1125
Internally, bzr only supports NFC normalization, since that is
1126
the standard for XML documents.
1128
So return the normalized path, and a flag indicating if the file
1129
can be accessed by that path.
1132
return unicodedata.normalize('NFC', unicode(path)), True
1135
def _inaccessible_normalized_filename(path):
1136
__doc__ = _accessible_normalized_filename.__doc__
1138
normalized = unicodedata.normalize('NFC', unicode(path))
1139
return normalized, normalized == path
1142
if _platform_normalizes_filenames:
1143
normalized_filename = _accessible_normalized_filename
1145
normalized_filename = _inaccessible_normalized_filename
1148
def terminal_width():
1149
"""Return estimated terminal width."""
1150
if sys.platform == 'win32':
1151
return win32utils.get_console_size()[0]
1154
import struct, fcntl, termios
1155
s = struct.pack('HHHH', 0, 0, 0, 0)
1156
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1157
width = struct.unpack('HHHH', x)[1]
1162
width = int(os.environ['COLUMNS'])
1171
def supports_executable():
1172
return sys.platform != "win32"
1175
def supports_posix_readonly():
1176
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1178
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1179
directory controls creation/deletion, etc.
1181
And under win32, readonly means that the directory itself cannot be
1182
deleted. The contents of a readonly directory can be changed, unlike POSIX
1183
where files in readonly directories cannot be added, deleted or renamed.
1185
return sys.platform != "win32"
1188
def set_or_unset_env(env_variable, value):
1189
"""Modify the environment, setting or removing the env_variable.
1191
:param env_variable: The environment variable in question
1192
:param value: The value to set the environment to. If None, then
1193
the variable will be removed.
1194
:return: The original value of the environment variable.
1196
orig_val = os.environ.get(env_variable)
1198
if orig_val is not None:
1199
del os.environ[env_variable]
1201
if isinstance(value, unicode):
1202
value = value.encode(get_user_encoding())
1203
os.environ[env_variable] = value
1207
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1210
def check_legal_path(path):
1211
"""Check whether the supplied path is legal.
1212
This is only required on Windows, so we don't test on other platforms
1215
if sys.platform != "win32":
1217
if _validWin32PathRE.match(path) is None:
1218
raise errors.IllegalPath(path)
1221
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1223
def _is_error_enotdir(e):
1224
"""Check if this exception represents ENOTDIR.
1226
Unfortunately, python is very inconsistent about the exception
1227
here. The cases are:
1228
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1229
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1230
which is the windows error code.
1231
3) Windows, Python2.5 uses errno == EINVAL and
1232
winerror == ERROR_DIRECTORY
1234
:param e: An Exception object (expected to be OSError with an errno
1235
attribute, but we should be able to cope with anything)
1236
:return: True if this represents an ENOTDIR error. False otherwise.
1238
en = getattr(e, 'errno', None)
1239
if (en == errno.ENOTDIR
1240
or (sys.platform == 'win32'
1241
and (en == _WIN32_ERROR_DIRECTORY
1242
or (en == errno.EINVAL
1243
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1249
def walkdirs(top, prefix=""):
1250
"""Yield data about all the directories in a tree.
1252
This yields all the data about the contents of a directory at a time.
1253
After each directory has been yielded, if the caller has mutated the list
1254
to exclude some directories, they are then not descended into.
1256
The data yielded is of the form:
1257
((directory-relpath, directory-path-from-top),
1258
[(relpath, basename, kind, lstat, path-from-top), ...]),
1259
- directory-relpath is the relative path of the directory being returned
1260
with respect to top. prefix is prepended to this.
1261
- directory-path-from-root is the path including top for this directory.
1262
It is suitable for use with os functions.
1263
- relpath is the relative path within the subtree being walked.
1264
- basename is the basename of the path
1265
- kind is the kind of the file now. If unknown then the file is not
1266
present within the tree - but it may be recorded as versioned. See
1268
- lstat is the stat data *if* the file was statted.
1269
- planned, not implemented:
1270
path_from_tree_root is the path from the root of the tree.
1272
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1273
allows one to walk a subtree but get paths that are relative to a tree
1275
:return: an iterator over the dirs.
1277
#TODO there is a bit of a smell where the results of the directory-
1278
# summary in this, and the path from the root, may not agree
1279
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1280
# potentially confusing output. We should make this more robust - but
1281
# not at a speed cost. RBC 20060731
1283
_directory = _directory_kind
1284
_listdir = os.listdir
1285
_kind_from_mode = file_kind_from_stat_mode
1286
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1288
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1289
relroot, _, _, _, top = pending.pop()
1291
relprefix = relroot + u'/'
1294
top_slash = top + u'/'
1297
append = dirblock.append
1299
names = sorted(_listdir(top))
1301
if not _is_error_enotdir(e):
1305
abspath = top_slash + name
1306
statvalue = _lstat(abspath)
1307
kind = _kind_from_mode(statvalue.st_mode)
1308
append((relprefix + name, name, kind, statvalue, abspath))
1309
yield (relroot, top), dirblock
1311
# push the user specified dirs from dirblock
1312
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1315
class DirReader(object):
1316
"""An interface for reading directories."""
1318
def top_prefix_to_starting_dir(self, top, prefix=""):
1319
"""Converts top and prefix to a starting dir entry
1321
:param top: A utf8 path
1322
:param prefix: An optional utf8 path to prefix output relative paths
1324
:return: A tuple starting with prefix, and ending with the native
1327
raise NotImplementedError(self.top_prefix_to_starting_dir)
1329
def read_dir(self, prefix, top):
1330
"""Read a specific dir.
1332
:param prefix: A utf8 prefix to be preprended to the path basenames.
1333
:param top: A natively encoded path to read.
1334
:return: A list of the directories contents. Each item contains:
1335
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1337
raise NotImplementedError(self.read_dir)
1340
_selected_dir_reader = None
1343
def _walkdirs_utf8(top, prefix=""):
1344
"""Yield data about all the directories in a tree.
1346
This yields the same information as walkdirs() only each entry is yielded
1347
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1348
are returned as exact byte-strings.
1350
:return: yields a tuple of (dir_info, [file_info])
1351
dir_info is (utf8_relpath, path-from-top)
1352
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1353
if top is an absolute path, path-from-top is also an absolute path.
1354
path-from-top might be unicode or utf8, but it is the correct path to
1355
pass to os functions to affect the file in question. (such as os.lstat)
1357
global _selected_dir_reader
1358
if _selected_dir_reader is None:
1359
fs_encoding = _fs_enc.upper()
1360
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1361
# Win98 doesn't have unicode apis like FindFirstFileW
1362
# TODO: We possibly could support Win98 by falling back to the
1363
# original FindFirstFile, and using TCHAR instead of WCHAR,
1364
# but that gets a bit tricky, and requires custom compiling
1367
from bzrlib._walkdirs_win32 import Win32ReadDir
1369
_selected_dir_reader = UnicodeDirReader()
1371
_selected_dir_reader = Win32ReadDir()
1372
elif fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1373
# ANSI_X3.4-1968 is a form of ASCII
1374
_selected_dir_reader = UnicodeDirReader()
1377
from bzrlib._readdir_pyx import UTF8DirReader
1379
# No optimised code path
1380
_selected_dir_reader = UnicodeDirReader()
1382
_selected_dir_reader = UTF8DirReader()
1383
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1384
# But we don't actually uses 1-3 in pending, so set them to None
1385
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1386
read_dir = _selected_dir_reader.read_dir
1387
_directory = _directory_kind
1389
relroot, _, _, _, top = pending[-1].pop()
1392
dirblock = sorted(read_dir(relroot, top))
1393
yield (relroot, top), dirblock
1394
# push the user specified dirs from dirblock
1395
next = [d for d in reversed(dirblock) if d[2] == _directory]
1397
pending.append(next)
1400
class UnicodeDirReader(DirReader):
1401
"""A dir reader for non-utf8 file systems, which transcodes."""
1403
__slots__ = ['_utf8_encode']
1406
self._utf8_encode = codecs.getencoder('utf8')
1408
def top_prefix_to_starting_dir(self, top, prefix=""):
1409
"""See DirReader.top_prefix_to_starting_dir."""
1410
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1412
def read_dir(self, prefix, top):
1413
"""Read a single directory from a non-utf8 file system.
1415
top, and the abspath element in the output are unicode, all other paths
1416
are utf8. Local disk IO is done via unicode calls to listdir etc.
1418
This is currently the fallback code path when the filesystem encoding is
1419
not UTF-8. It may be better to implement an alternative so that we can
1420
safely handle paths that are not properly decodable in the current
1423
See DirReader.read_dir for details.
1425
_utf8_encode = self._utf8_encode
1427
_listdir = os.listdir
1428
_kind_from_mode = file_kind_from_stat_mode
1431
relprefix = prefix + '/'
1434
top_slash = top + u'/'
1437
append = dirblock.append
1438
for name in sorted(_listdir(top)):
1440
name_utf8 = _utf8_encode(name)[0]
1441
except UnicodeDecodeError:
1442
raise errors.BadFilenameEncoding(
1443
_utf8_encode(relprefix)[0] + name, _fs_enc)
1444
abspath = top_slash + name
1445
statvalue = _lstat(abspath)
1446
kind = _kind_from_mode(statvalue.st_mode)
1447
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1451
def copy_tree(from_path, to_path, handlers={}):
1452
"""Copy all of the entries in from_path into to_path.
1454
:param from_path: The base directory to copy.
1455
:param to_path: The target directory. If it does not exist, it will
1457
:param handlers: A dictionary of functions, which takes a source and
1458
destinations for files, directories, etc.
1459
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1460
'file', 'directory', and 'symlink' should always exist.
1461
If they are missing, they will be replaced with 'os.mkdir()',
1462
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1464
# Now, just copy the existing cached tree to the new location
1465
# We use a cheap trick here.
1466
# Absolute paths are prefixed with the first parameter
1467
# relative paths are prefixed with the second.
1468
# So we can get both the source and target returned
1469
# without any extra work.
1471
def copy_dir(source, dest):
1474
def copy_link(source, dest):
1475
"""Copy the contents of a symlink"""
1476
link_to = os.readlink(source)
1477
os.symlink(link_to, dest)
1479
real_handlers = {'file':shutil.copy2,
1480
'symlink':copy_link,
1481
'directory':copy_dir,
1483
real_handlers.update(handlers)
1485
if not os.path.exists(to_path):
1486
real_handlers['directory'](from_path, to_path)
1488
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1489
for relpath, name, kind, st, abspath in entries:
1490
real_handlers[kind](abspath, relpath)
1493
def path_prefix_key(path):
1494
"""Generate a prefix-order path key for path.
1496
This can be used to sort paths in the same way that walkdirs does.
1498
return (dirname(path) , path)
1501
def compare_paths_prefix_order(path_a, path_b):
1502
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1503
key_a = path_prefix_key(path_a)
1504
key_b = path_prefix_key(path_b)
1505
return cmp(key_a, key_b)
1508
_cached_user_encoding = None
1511
def get_user_encoding(use_cache=True):
1512
"""Find out what the preferred user encoding is.
1514
This is generally the encoding that is used for command line parameters
1515
and file contents. This may be different from the terminal encoding
1516
or the filesystem encoding.
1518
:param use_cache: Enable cache for detected encoding.
1519
(This parameter is turned on by default,
1520
and required only for selftesting)
1522
:return: A string defining the preferred user encoding
1524
global _cached_user_encoding
1525
if _cached_user_encoding is not None and use_cache:
1526
return _cached_user_encoding
1528
if sys.platform == 'darwin':
1529
# python locale.getpreferredencoding() always return
1530
# 'mac-roman' on darwin. That's a lie.
1531
sys.platform = 'posix'
1533
if os.environ.get('LANG', None) is None:
1534
# If LANG is not set, we end up with 'ascii', which is bad
1535
# ('mac-roman' is more than ascii), so we set a default which
1536
# will give us UTF-8 (which appears to work in all cases on
1537
# OSX). Users are still free to override LANG of course, as
1538
# long as it give us something meaningful. This work-around
1539
# *may* not be needed with python 3k and/or OSX 10.5, but will
1540
# work with them too -- vila 20080908
1541
os.environ['LANG'] = 'en_US.UTF-8'
1544
sys.platform = 'darwin'
1549
user_encoding = locale.getpreferredencoding()
1550
except locale.Error, e:
1551
sys.stderr.write('bzr: warning: %s\n'
1552
' Could not determine what text encoding to use.\n'
1553
' This error usually means your Python interpreter\n'
1554
' doesn\'t support the locale set by $LANG (%s)\n'
1555
" Continuing with ascii encoding.\n"
1556
% (e, os.environ.get('LANG')))
1557
user_encoding = 'ascii'
1559
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1560
# treat that as ASCII, and not support printing unicode characters to the
1563
# For python scripts run under vim, we get '', so also treat that as ASCII
1564
if user_encoding in (None, 'cp0', ''):
1565
user_encoding = 'ascii'
1569
codecs.lookup(user_encoding)
1571
sys.stderr.write('bzr: warning:'
1572
' unknown encoding %s.'
1573
' Continuing with ascii encoding.\n'
1576
user_encoding = 'ascii'
1579
_cached_user_encoding = user_encoding
1581
return user_encoding
1584
def get_host_name():
1585
"""Return the current unicode host name.
1587
This is meant to be used in place of socket.gethostname() because that
1588
behaves inconsistently on different platforms.
1590
if sys.platform == "win32":
1592
return win32utils.get_host_name()
1595
return socket.gethostname().decode(get_user_encoding())
1598
def recv_all(socket, bytes):
1599
"""Receive an exact number of bytes.
1601
Regular Socket.recv() may return less than the requested number of bytes,
1602
dependning on what's in the OS buffer. MSG_WAITALL is not available
1603
on all platforms, but this should work everywhere. This will return
1604
less than the requested amount if the remote end closes.
1606
This isn't optimized and is intended mostly for use in testing.
1609
while len(b) < bytes:
1610
new = until_no_eintr(socket.recv, bytes - len(b))
1617
def send_all(socket, bytes):
1618
"""Send all bytes on a socket.
1620
Regular socket.sendall() can give socket error 10053 on Windows. This
1621
implementation sends no more than 64k at a time, which avoids this problem.
1624
for pos in xrange(0, len(bytes), chunk_size):
1625
until_no_eintr(socket.sendall, bytes[pos:pos+chunk_size])
1628
def dereference_path(path):
1629
"""Determine the real path to a file.
1631
All parent elements are dereferenced. But the file itself is not
1633
:param path: The original path. May be absolute or relative.
1634
:return: the real path *to* the file
1636
parent, base = os.path.split(path)
1637
# The pathjoin for '.' is a workaround for Python bug #1213894.
1638
# (initial path components aren't dereferenced)
1639
return pathjoin(realpath(pathjoin('.', parent)), base)
1642
def supports_mapi():
1643
"""Return True if we can use MAPI to launch a mail client."""
1644
return sys.platform == "win32"
1647
def resource_string(package, resource_name):
1648
"""Load a resource from a package and return it as a string.
1650
Note: Only packages that start with bzrlib are currently supported.
1652
This is designed to be a lightweight implementation of resource
1653
loading in a way which is API compatible with the same API from
1655
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1656
If and when pkg_resources becomes a standard library, this routine
1659
# Check package name is within bzrlib
1660
if package == "bzrlib":
1661
resource_relpath = resource_name
1662
elif package.startswith("bzrlib."):
1663
package = package[len("bzrlib."):].replace('.', os.sep)
1664
resource_relpath = pathjoin(package, resource_name)
1666
raise errors.BzrError('resource package %s not in bzrlib' % package)
1668
# Map the resource to a file and read its contents
1669
base = dirname(bzrlib.__file__)
1670
if getattr(sys, 'frozen', None): # bzr.exe
1671
base = abspath(pathjoin(base, '..', '..'))
1672
filename = pathjoin(base, resource_relpath)
1673
return open(filename, 'rU').read()
1676
def file_kind_from_stat_mode_thunk(mode):
1677
global file_kind_from_stat_mode
1678
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
1680
from bzrlib._readdir_pyx import UTF8DirReader
1681
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
1683
from bzrlib._readdir_py import (
1684
_kind_from_mode as file_kind_from_stat_mode
1686
return file_kind_from_stat_mode(mode)
1687
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
1690
def file_kind(f, _lstat=os.lstat):
1692
return file_kind_from_stat_mode(_lstat(f).st_mode)
1694
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
1695
raise errors.NoSuchFile(f)
1699
def until_no_eintr(f, *a, **kw):
1700
"""Run f(*a, **kw), retrying if an EINTR error occurs."""
1701
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
1705
except (IOError, OSError), e:
1706
if e.errno == errno.EINTR:
1711
if sys.platform == "win32":
1714
return msvcrt.getch()
1719
fd = sys.stdin.fileno()
1720
settings = termios.tcgetattr(fd)
1723
ch = sys.stdin.read(1)
1725
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
added
removed
bzrlib/osutils.py
