1
# Bazaar-NG -- distributed version control
3
# Copyright (C) 2005 by Canonical Ltd
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
5
3
# This program is free software; you can redistribute it and/or modify
6
4
# it under the terms of the GNU General Public License as published by
7
5
# the Free Software Foundation; either version 2 of the License, or
8
6
# (at your option) any later version.
10
8
# This program is distributed in the hope that it will be useful,
11
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
12
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13
11
# GNU General Public License for more details.
15
13
# You should have received a copy of the GNU General Public License
16
14
# along with this program; if not, write to the Free Software
17
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
19
import os, types, re, time, errno, sys
20
from stat import S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE
22
from bzrlib.errors import BzrError
17
from cStringIO import StringIO
21
from stat import (S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE,
22
S_ISCHR, S_ISBLK, S_ISFIFO, S_ISSOCK)
26
from bzrlib.lazy_import import lazy_import
27
lazy_import(globals(), """
29
from datetime import datetime
31
from ntpath import (abspath as _nt_abspath,
33
normpath as _nt_normpath,
34
realpath as _nt_realpath,
35
splitdrive as _nt_splitdrive,
44
from tempfile import (
57
from bzrlib import symbol_versioning
58
from bzrlib.symbol_versioning import (
23
61
from bzrlib.trace import mutter
64
# On win32, O_BINARY is used to indicate the file should
65
# be opened in binary mode, rather than text mode.
66
# On other platforms, O_BINARY doesn't exist, because
67
# they always open in binary mode, so it is okay to
68
# OR with 0 on those platforms
69
O_BINARY = getattr(os, 'O_BINARY', 0)
71
# On posix, use lstat instead of stat so that we can
72
# operate on broken symlinks. On Windows revert to stat.
73
lstat = getattr(os, 'lstat', os.stat)
26
75
def make_readonly(filename):
27
76
"""Make a filename read-only."""
28
# TODO: probably needs to be fixed for windows
29
mod = os.stat(filename).st_mode
31
os.chmod(filename, mod)
77
mod = lstat(filename).st_mode
78
if not stat.S_ISLNK(mod):
80
os.chmod(filename, mod)
34
83
def make_writable(filename):
35
mod = os.stat(filename).st_mode
37
os.chmod(filename, mod)
40
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/_~-])')
84
mod = lstat(filename).st_mode
85
if not stat.S_ISLNK(mod):
87
os.chmod(filename, mod)
90
def minimum_path_selection(paths):
91
"""Return the smallset subset of paths which are outside paths.
93
:param paths: A container (and hence not None) of paths.
94
:return: A set of paths sufficient to include everything in paths via
95
is_inside_any, drawn from the paths parameter.
100
other_paths = paths.difference([path])
101
if not is_inside_any(other_paths, path):
102
# this is a top level path, we must check it.
103
search_paths.add(path)
42
111
"""Return a quoted filename filename
44
113
This previously used backslash quoting, but that works poorly on
46
115
# TODO: I'm not really sure this is the best format either.x
117
if _QUOTE_RE is None:
118
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
47
120
if _QUOTE_RE.search(f):
48
121
return '"' + f + '"'
54
mode = os.lstat(f)[ST_MODE]
62
raise BzrError("can't handle file kind with mode %o of %r" % (mode, f))
126
_directory_kind = 'directory'
129
stat.S_IFDIR:_directory_kind,
130
stat.S_IFCHR:'chardev',
131
stat.S_IFBLK:'block',
134
stat.S_IFLNK:'symlink',
135
stat.S_IFSOCK:'socket',
139
def file_kind_from_stat_mode(stat_mode, _formats=_formats, _unknown='unknown'):
140
"""Generate a file kind from a stat mode. This is used in walkdirs.
142
Its performance is critical: Do not mutate without careful benchmarking.
145
return _formats[stat_mode & 0170000]
150
def file_kind(f, _lstat=os.lstat, _mapper=file_kind_from_stat_mode):
152
return _mapper(_lstat(f).st_mode)
154
if getattr(e, 'errno', None) == errno.ENOENT:
155
raise errors.NoSuchFile(f)
160
"""Return the current umask"""
161
# Assume that people aren't messing with the umask while running
162
# XXX: This is not thread safe, but there is no way to get the
163
# umask without setting it
171
_directory_kind: "/",
173
'tree-reference': '+',
65
177
def kind_marker(kind):
68
elif kind == 'directory':
70
elif kind == 'symlink':
73
raise BzrError('invalid file kind %r' % kind)
179
return _kind_marker_map[kind]
181
raise errors.BzrError('invalid file kind %r' % kind)
184
lexists = getattr(os.path, 'lexists', None)
188
stat = getattr(os, 'lstat', os.stat)
192
if e.errno == errno.ENOENT:
195
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
198
def fancy_rename(old, new, rename_func, unlink_func):
199
"""A fancy rename, when you don't have atomic rename.
201
:param old: The old path, to rename from
202
:param new: The new path, to rename to
203
:param rename_func: The potentially non-atomic rename function
204
:param unlink_func: A way to delete the target file if the full rename succeeds
207
# sftp rename doesn't allow overwriting, so play tricks:
209
base = os.path.basename(new)
210
dirname = os.path.dirname(new)
211
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
212
tmp_name = pathjoin(dirname, tmp_name)
214
# Rename the file out of the way, but keep track if it didn't exist
215
# We don't want to grab just any exception
216
# something like EACCES should prevent us from continuing
217
# The downside is that the rename_func has to throw an exception
218
# with an errno = ENOENT, or NoSuchFile
221
rename_func(new, tmp_name)
222
except (errors.NoSuchFile,), e:
225
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
226
# function raises an IOError with errno is None when a rename fails.
227
# This then gets caught here.
228
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
231
if (getattr(e, 'errno', None) is None
232
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
239
# This may throw an exception, in which case success will
241
rename_func(old, new)
245
# If the file used to exist, rename it back into place
246
# otherwise just delete it from the tmp location
248
unlink_func(tmp_name)
250
rename_func(tmp_name, new)
253
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
254
# choke on a Unicode string containing a relative path if
255
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
257
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
258
def _posix_abspath(path):
259
# jam 20060426 rather than encoding to fsencoding
260
# copy posixpath.abspath, but use os.getcwdu instead
261
if not posixpath.isabs(path):
262
path = posixpath.join(getcwd(), path)
263
return posixpath.normpath(path)
266
def _posix_realpath(path):
267
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
270
def _win32_fixdrive(path):
271
"""Force drive letters to be consistent.
273
win32 is inconsistent whether it returns lower or upper case
274
and even if it was consistent the user might type the other
275
so we force it to uppercase
276
running python.exe under cmd.exe return capital C:\\
277
running win32 python inside a cygwin shell returns lowercase c:\\
279
drive, path = _nt_splitdrive(path)
280
return drive.upper() + path
283
def _win32_abspath(path):
284
# Real _nt_abspath doesn't have a problem with a unicode cwd
285
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
288
def _win98_abspath(path):
289
"""Return the absolute version of a path.
290
Windows 98 safe implementation (python reimplementation
291
of Win32 API function GetFullPathNameW)
296
# \\HOST\path => //HOST/path
297
# //HOST/path => //HOST/path
298
# path => C:/cwd/path
301
# check for absolute path
302
drive = _nt_splitdrive(path)[0]
303
if drive == '' and path[:2] not in('//','\\\\'):
305
# we cannot simply os.path.join cwd and path
306
# because os.path.join('C:','/path') produce '/path'
307
# and this is incorrect
308
if path[:1] in ('/','\\'):
309
cwd = _nt_splitdrive(cwd)[0]
311
path = cwd + '\\' + path
312
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
314
if win32utils.winver == 'Windows 98':
315
_win32_abspath = _win98_abspath
318
def _win32_realpath(path):
319
# Real _nt_realpath doesn't have a problem with a unicode cwd
320
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
323
def _win32_pathjoin(*args):
324
return _nt_join(*args).replace('\\', '/')
327
def _win32_normpath(path):
328
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
332
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
335
def _win32_mkdtemp(*args, **kwargs):
336
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
339
def _win32_rename(old, new):
340
"""We expect to be able to atomically replace 'new' with old.
342
On win32, if new exists, it must be moved out of the way first,
346
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
348
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
349
# If we try to rename a non-existant file onto cwd, we get
350
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
351
# if the old path doesn't exist, sometimes we get EACCES
352
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
358
return unicodedata.normalize('NFKC', os.getcwdu())
361
# Default is to just use the python builtins, but these can be rebound on
362
# particular platforms.
363
abspath = _posix_abspath
364
realpath = _posix_realpath
365
pathjoin = os.path.join
366
normpath = os.path.normpath
369
dirname = os.path.dirname
370
basename = os.path.basename
371
split = os.path.split
372
splitext = os.path.splitext
373
# These were already imported into local scope
374
# mkdtemp = tempfile.mkdtemp
375
# rmtree = shutil.rmtree
377
MIN_ABS_PATHLENGTH = 1
380
if sys.platform == 'win32':
381
abspath = _win32_abspath
382
realpath = _win32_realpath
383
pathjoin = _win32_pathjoin
384
normpath = _win32_normpath
385
getcwd = _win32_getcwd
386
mkdtemp = _win32_mkdtemp
387
rename = _win32_rename
389
MIN_ABS_PATHLENGTH = 3
391
def _win32_delete_readonly(function, path, excinfo):
392
"""Error handler for shutil.rmtree function [for win32]
393
Helps to remove files and dirs marked as read-only.
395
exception = excinfo[1]
396
if function in (os.remove, os.rmdir) \
397
and isinstance(exception, OSError) \
398
and exception.errno == errno.EACCES:
404
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
405
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
406
return shutil.rmtree(path, ignore_errors, onerror)
407
elif sys.platform == 'darwin':
411
def get_terminal_encoding():
412
"""Find the best encoding for printing to the screen.
414
This attempts to check both sys.stdout and sys.stdin to see
415
what encoding they are in, and if that fails it falls back to
416
bzrlib.user_encoding.
417
The problem is that on Windows, locale.getpreferredencoding()
418
is not the same encoding as that used by the console:
419
http://mail.python.org/pipermail/python-list/2003-May/162357.html
421
On my standard US Windows XP, the preferred encoding is
422
cp1252, but the console is cp437
424
output_encoding = getattr(sys.stdout, 'encoding', None)
425
if not output_encoding:
426
input_encoding = getattr(sys.stdin, 'encoding', None)
427
if not input_encoding:
428
output_encoding = bzrlib.user_encoding
429
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
431
output_encoding = input_encoding
432
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
434
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
435
if output_encoding == 'cp0':
436
# invalid encoding (cp0 means 'no codepage' on Windows)
437
output_encoding = bzrlib.user_encoding
438
mutter('cp0 is invalid encoding.'
439
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
442
codecs.lookup(output_encoding)
444
sys.stderr.write('bzr: warning:'
445
' unknown terminal encoding %s.\n'
446
' Using encoding %s instead.\n'
447
% (output_encoding, bzrlib.user_encoding)
449
output_encoding = bzrlib.user_encoding
451
return output_encoding
454
def normalizepath(f):
455
if getattr(os.path, 'realpath', None) is not None:
459
[p,e] = os.path.split(f)
460
if e == "" or e == "." or e == "..":
463
return pathjoin(F(p), e)
77
466
def backup_file(fn):
312
622
def local_time_offset(t=None):
313
623
"""Return offset of local zone from GMT, either at present or at time t."""
314
# python2.3 localtime() can't take None
318
if time.localtime(t).tm_isdst and time.daylight:
321
return -time.timezone
324
def format_date(t, offset=0, timezone='original'):
325
## TODO: Perhaps a global option to use either universal or local time?
326
## Or perhaps just let people set $TZ?
327
assert isinstance(t, float)
626
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
627
return offset.days * 86400 + offset.seconds
630
def format_date(t, offset=0, timezone='original', date_fmt=None,
632
"""Return a formatted date string.
634
:param t: Seconds since the epoch.
635
:param offset: Timezone offset in seconds east of utc.
636
:param timezone: How to display the time: 'utc', 'original' for the
637
timezone specified by offset, or 'local' for the process's current
639
:param show_offset: Whether to append the timezone.
640
:param date_fmt: strftime format.
329
642
if timezone == 'utc':
330
643
tt = time.gmtime(t)
332
645
elif timezone == 'original':
335
648
tt = time.gmtime(t + offset)
336
649
elif timezone == 'local':
337
650
tt = time.localtime(t)
338
651
offset = local_time_offset(t)
340
raise BzrError("unsupported timezone format %r",
341
['options are "utc", "original", "local"'])
343
return (time.strftime("%a %Y-%m-%d %H:%M:%S", tt)
344
+ ' %+03d%02d' % (offset / 3600, (offset / 60) % 60))
653
raise errors.BzrError("unsupported timezone format %r" % timezone,
654
['options are "utc", "original", "local"'])
656
date_fmt = "%a %Y-%m-%d %H:%M:%S"
658
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
661
return (time.strftime(date_fmt, tt) + offset_str)
347
664
def compact_date(when):
348
665
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
668
def format_delta(delta):
669
"""Get a nice looking string for a time delta.
671
:param delta: The time difference in seconds, can be positive or negative.
672
positive indicates time in the past, negative indicates time in the
673
future. (usually time.time() - stored_time)
674
:return: String formatted to show approximate resolution
680
direction = 'in the future'
684
if seconds < 90: # print seconds up to 90 seconds
686
return '%d second %s' % (seconds, direction,)
688
return '%d seconds %s' % (seconds, direction)
690
minutes = int(seconds / 60)
691
seconds -= 60 * minutes
696
if minutes < 90: # print minutes, seconds up to 90 minutes
698
return '%d minute, %d second%s %s' % (
699
minutes, seconds, plural_seconds, direction)
701
return '%d minutes, %d second%s %s' % (
702
minutes, seconds, plural_seconds, direction)
704
hours = int(minutes / 60)
705
minutes -= 60 * hours
712
return '%d hour, %d minute%s %s' % (hours, minutes,
713
plural_minutes, direction)
714
return '%d hours, %d minute%s %s' % (hours, minutes,
715
plural_minutes, direction)
353
718
"""Return size of given open file."""
354
719
return os.fstat(f.fileno())[ST_SIZE]
357
if hasattr(os, 'urandom'): # python 2.4 and later
722
# Define rand_bytes based on platform.
724
# Python 2.4 and later have os.urandom,
725
# but it doesn't work on some arches
358
727
rand_bytes = os.urandom
359
elif sys.platform == 'linux2':
360
rand_bytes = file('/dev/urandom', 'rb').read
362
# not well seeded, but better than nothing
367
s += chr(random.randint(0, 255))
728
except (NotImplementedError, AttributeError):
729
# If python doesn't have os.urandom, or it doesn't work,
730
# then try to first pull random data from /dev/urandom
732
rand_bytes = file('/dev/urandom', 'rb').read
733
# Otherwise, use this hack as a last resort
734
except (IOError, OSError):
735
# not well seeded, but better than nothing
740
s += chr(random.randint(0, 255))
745
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
747
"""Return a random string of num alphanumeric characters
749
The result only contains lowercase chars because it may be used on
750
case-insensitive filesystems.
753
for raw_byte in rand_bytes(num):
754
s += ALNUM[ord(raw_byte) % 36]
372
758
## TODO: We could later have path objects that remember their list
373
759
## decomposition (might be too tricksy though.)
375
761
def splitpath(p):
376
"""Turn string into list of parts.
382
>>> splitpath('a/./b')
384
>>> splitpath('a/.b')
386
>>> splitpath('a/../b')
387
Traceback (most recent call last):
389
BzrError: sorry, '..' not allowed in path
391
assert isinstance(p, types.StringTypes)
762
"""Turn string into list of parts."""
763
assert isinstance(p, basestring)
393
765
# split on either delimiter because people might use either on
408
assert isinstance(p, list)
780
assert isinstance(p, (list, tuple))
410
if (f == '..') or (f == None) or (f == ''):
411
raise BzrError("sorry, %r not allowed in path" % f)
412
return os.path.join(*p)
415
def appendpath(p1, p2):
419
return os.path.join(p1, p2)
422
def extern_command(cmd, ignore_errors = False):
423
mutter('external command: %s' % `cmd`)
425
if not ignore_errors:
426
raise BzrError('command failed')
429
def _read_config_value(name):
430
"""Read a config value from the file ~/.bzr.conf/<name>
431
Return None if the file does not exist"""
782
if (f == '..') or (f is None) or (f == ''):
783
raise errors.BzrError("sorry, %r not allowed in path" % f)
788
"""Split s into lines, but without removing the newline characters."""
789
lines = s.split('\n')
790
result = [line + '\n' for line in lines[:-1]]
792
result.append(lines[-1])
796
def hardlinks_good():
797
return sys.platform not in ('win32', 'cygwin', 'darwin')
800
def link_or_copy(src, dest):
801
"""Hardlink a file, or copy it if it can't be hardlinked."""
802
if not hardlinks_good():
803
shutil.copyfile(src, dest)
433
f = file(os.path.join(config_dir(), name), "r")
434
return f.read().decode(bzrlib.user_encoding).rstrip("\r\n")
436
if e.errno == errno.ENOENT:
442
"""Return a sequence of possible editor binaries for the current platform"""
443
e = _read_config_value("editor")
447
if os.name == "windows":
449
elif os.name == "posix":
451
yield os.environ["EDITOR"]
456
def _run_editor(filename):
457
"""Try to execute an editor to edit the commit message. Returns True on success,
459
for e in _get_editor():
460
x = os.spawnvp(os.P_WAIT, e, (e, filename))
807
except (OSError, IOError), e:
808
if e.errno != errno.EXDEV:
810
shutil.copyfile(src, dest)
813
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
814
# Forgiveness than Permission (EAFP) because:
815
# - root can damage a solaris file system by using unlink,
816
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
817
# EACCES, OSX: EPERM) when invoked on a directory.
818
def delete_any(path):
819
"""Delete a file or directory."""
820
if isdir(path): # Takes care of symlinks
827
if getattr(os, 'symlink', None) is not None:
833
def contains_whitespace(s):
834
"""True if there are any whitespace characters in s."""
835
# string.whitespace can include '\xa0' in certain locales, because it is
836
# considered "non-breaking-space" as part of ISO-8859-1. But it
837
# 1) Isn't a breaking whitespace
838
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
840
# 3) '\xa0' isn't unicode safe since it is >128.
842
# This should *not* be a unicode set of characters in case the source
843
# string is not a Unicode string. We can auto-up-cast the characters since
844
# they are ascii, but we don't want to auto-up-cast the string in case it
846
for ch in ' \t\n\r\v\f':
853
def contains_linebreaks(s):
854
"""True if there is any vertical whitespace in s."""
862
def relpath(base, path):
863
"""Return path relative to base, or raise exception.
865
The path may be either an absolute path or a path relative to the
866
current working directory.
868
os.path.commonprefix (python2.4) has a bad bug that it works just
869
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
873
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
874
' exceed the platform minimum length (which is %d)' %
881
while len(head) >= len(base):
467
raise BzrError("Could not start any editor. Please specify $EDITOR or use ~/.bzr.conf/editor")
471
def get_text_message(infotext, ignoreline = "default"):
474
if ignoreline == "default":
475
ignoreline = "-- This line and the following will be ignored --"
478
tmp_fileno, msgfilename = tempfile.mkstemp()
479
msgfile = os.close(tmp_fileno)
480
if infotext is not None and infotext != "":
482
msgfile = file(msgfilename, "w")
483
msgfile.write("\n\n%s\n\n%s" % (ignoreline, infotext))
488
if not _run_editor(msgfilename):
493
lastline, nlines = 0, 0
494
for line in file(msgfilename, "r"):
495
stripped_line = line.strip()
496
# strip empty line before the log message starts
498
if stripped_line != "":
502
# check for the ignore line only if there
503
# is additional information at the end
504
if hasinfo and stripped_line == ignoreline:
507
# keep track of the last line that had some content
508
if stripped_line != "":
514
# delete empty lines at the end
516
# add a newline at the end, if needed
517
if not msg[-1].endswith("\n"):
518
return "%s%s" % ("".join(msg), "\n")
522
# delete the msg file in any case
523
try: os.unlink(msgfilename)
884
head, tail = os.path.split(head)
888
raise errors.PathNotChild(rp, base)
896
def safe_unicode(unicode_or_utf8_string):
897
"""Coerce unicode_or_utf8_string into unicode.
899
If it is unicode, it is returned.
900
Otherwise it is decoded from utf-8. If a decoding error
901
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
902
as a BzrBadParameter exception.
904
if isinstance(unicode_or_utf8_string, unicode):
905
return unicode_or_utf8_string
907
return unicode_or_utf8_string.decode('utf8')
908
except UnicodeDecodeError:
909
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
912
def safe_utf8(unicode_or_utf8_string):
913
"""Coerce unicode_or_utf8_string to a utf8 string.
915
If it is a str, it is returned.
916
If it is Unicode, it is encoded into a utf-8 string.
918
if isinstance(unicode_or_utf8_string, str):
919
# TODO: jam 20070209 This is overkill, and probably has an impact on
920
# performance if we are dealing with lots of apis that want a
923
# Make sure it is a valid utf-8 string
924
unicode_or_utf8_string.decode('utf-8')
925
except UnicodeDecodeError:
926
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
927
return unicode_or_utf8_string
928
return unicode_or_utf8_string.encode('utf-8')
931
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
932
' Revision id generators should be creating utf8'
936
def safe_revision_id(unicode_or_utf8_string, warn=True):
937
"""Revision ids should now be utf8, but at one point they were unicode.
939
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
941
:param warn: Functions that are sanitizing user data can set warn=False
942
:return: None or a utf8 revision id.
944
if (unicode_or_utf8_string is None
945
or unicode_or_utf8_string.__class__ == str):
946
return unicode_or_utf8_string
948
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
950
return cache_utf8.encode(unicode_or_utf8_string)
953
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
954
' generators should be creating utf8 file ids.')
957
def safe_file_id(unicode_or_utf8_string, warn=True):
958
"""File ids should now be utf8, but at one point they were unicode.
960
This is the same as safe_utf8, except it uses the cached encode functions
961
to save a little bit of performance.
963
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
965
:param warn: Functions that are sanitizing user data can set warn=False
966
:return: None or a utf8 file id.
968
if (unicode_or_utf8_string is None
969
or unicode_or_utf8_string.__class__ == str):
970
return unicode_or_utf8_string
972
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
974
return cache_utf8.encode(unicode_or_utf8_string)
977
_platform_normalizes_filenames = False
978
if sys.platform == 'darwin':
979
_platform_normalizes_filenames = True
982
def normalizes_filenames():
983
"""Return True if this platform normalizes unicode filenames.
985
Mac OSX does, Windows/Linux do not.
987
return _platform_normalizes_filenames
990
def _accessible_normalized_filename(path):
991
"""Get the unicode normalized path, and if you can access the file.
993
On platforms where the system normalizes filenames (Mac OSX),
994
you can access a file by any path which will normalize correctly.
995
On platforms where the system does not normalize filenames
996
(Windows, Linux), you have to access a file by its exact path.
998
Internally, bzr only supports NFC/NFKC normalization, since that is
999
the standard for XML documents.
1001
So return the normalized path, and a flag indicating if the file
1002
can be accessed by that path.
1005
return unicodedata.normalize('NFKC', unicode(path)), True
1008
def _inaccessible_normalized_filename(path):
1009
__doc__ = _accessible_normalized_filename.__doc__
1011
normalized = unicodedata.normalize('NFKC', unicode(path))
1012
return normalized, normalized == path
1015
if _platform_normalizes_filenames:
1016
normalized_filename = _accessible_normalized_filename
1018
normalized_filename = _inaccessible_normalized_filename
1021
def terminal_width():
1022
"""Return estimated terminal width."""
1023
if sys.platform == 'win32':
1024
return win32utils.get_console_size()[0]
1027
import struct, fcntl, termios
1028
s = struct.pack('HHHH', 0, 0, 0, 0)
1029
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1030
width = struct.unpack('HHHH', x)[1]
1035
width = int(os.environ['COLUMNS'])
1044
def supports_executable():
1045
return sys.platform != "win32"
1048
def supports_posix_readonly():
1049
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1051
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1052
directory controls creation/deletion, etc.
1054
And under win32, readonly means that the directory itself cannot be
1055
deleted. The contents of a readonly directory can be changed, unlike POSIX
1056
where files in readonly directories cannot be added, deleted or renamed.
1058
return sys.platform != "win32"
1061
def set_or_unset_env(env_variable, value):
1062
"""Modify the environment, setting or removing the env_variable.
1064
:param env_variable: The environment variable in question
1065
:param value: The value to set the environment to. If None, then
1066
the variable will be removed.
1067
:return: The original value of the environment variable.
1069
orig_val = os.environ.get(env_variable)
1071
if orig_val is not None:
1072
del os.environ[env_variable]
1074
if isinstance(value, unicode):
1075
value = value.encode(bzrlib.user_encoding)
1076
os.environ[env_variable] = value
1080
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1083
def check_legal_path(path):
1084
"""Check whether the supplied path is legal.
1085
This is only required on Windows, so we don't test on other platforms
1088
if sys.platform != "win32":
1090
if _validWin32PathRE.match(path) is None:
1091
raise errors.IllegalPath(path)
1094
def walkdirs(top, prefix=""):
1095
"""Yield data about all the directories in a tree.
1097
This yields all the data about the contents of a directory at a time.
1098
After each directory has been yielded, if the caller has mutated the list
1099
to exclude some directories, they are then not descended into.
1101
The data yielded is of the form:
1102
((directory-relpath, directory-path-from-top),
1103
[(relpath, basename, kind, lstat, path-from-top), ...]),
1104
- directory-relpath is the relative path of the directory being returned
1105
with respect to top. prefix is prepended to this.
1106
- directory-path-from-root is the path including top for this directory.
1107
It is suitable for use with os functions.
1108
- relpath is the relative path within the subtree being walked.
1109
- basename is the basename of the path
1110
- kind is the kind of the file now. If unknown then the file is not
1111
present within the tree - but it may be recorded as versioned. See
1113
- lstat is the stat data *if* the file was statted.
1114
- planned, not implemented:
1115
path_from_tree_root is the path from the root of the tree.
1117
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1118
allows one to walk a subtree but get paths that are relative to a tree
1120
:return: an iterator over the dirs.
1122
#TODO there is a bit of a smell where the results of the directory-
1123
# summary in this, and the path from the root, may not agree
1124
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1125
# potentially confusing output. We should make this more robust - but
1126
# not at a speed cost. RBC 20060731
1128
_directory = _directory_kind
1129
_listdir = os.listdir
1130
_kind_from_mode = _formats.get
1131
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1133
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1134
relroot, _, _, _, top = pending.pop()
1136
relprefix = relroot + u'/'
1139
top_slash = top + u'/'
1142
append = dirblock.append
1143
for name in sorted(_listdir(top)):
1144
abspath = top_slash + name
1145
statvalue = _lstat(abspath)
1146
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1147
append((relprefix + name, name, kind, statvalue, abspath))
1148
yield (relroot, top), dirblock
1150
# push the user specified dirs from dirblock
1151
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1154
def _walkdirs_utf8(top, prefix=""):
1155
"""Yield data about all the directories in a tree.
1157
This yields the same information as walkdirs() only each entry is yielded
1158
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1159
are returned as exact byte-strings.
1161
:return: yields a tuple of (dir_info, [file_info])
1162
dir_info is (utf8_relpath, path-from-top)
1163
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1164
if top is an absolute path, path-from-top is also an absolute path.
1165
path-from-top might be unicode or utf8, but it is the correct path to
1166
pass to os functions to affect the file in question. (such as os.lstat)
1168
fs_encoding = _fs_enc.upper()
1169
if (sys.platform == 'win32' or
1170
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1171
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1173
return _walkdirs_fs_utf8(top, prefix=prefix)
1176
def _walkdirs_fs_utf8(top, prefix=""):
1177
"""See _walkdirs_utf8.
1179
This sub-function is called when we know the filesystem is already in utf8
1180
encoding. So we don't need to transcode filenames.
1183
_directory = _directory_kind
1184
_listdir = os.listdir
1185
_kind_from_mode = _formats.get
1187
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1188
# But we don't actually uses 1-3 in pending, so set them to None
1189
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1191
relroot, _, _, _, top = pending.pop()
1193
relprefix = relroot + '/'
1196
top_slash = top + '/'
1199
append = dirblock.append
1200
for name in sorted(_listdir(top)):
1201
abspath = top_slash + name
1202
statvalue = _lstat(abspath)
1203
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1204
append((relprefix + name, name, kind, statvalue, abspath))
1205
yield (relroot, top), dirblock
1207
# push the user specified dirs from dirblock
1208
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1211
def _walkdirs_unicode_to_utf8(top, prefix=""):
1212
"""See _walkdirs_utf8
1214
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1216
This is currently the fallback code path when the filesystem encoding is
1217
not UTF-8. It may be better to implement an alternative so that we can
1218
safely handle paths that are not properly decodable in the current
1221
_utf8_encode = codecs.getencoder('utf8')
1223
_directory = _directory_kind
1224
_listdir = os.listdir
1225
_kind_from_mode = _formats.get
1227
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1229
relroot, _, _, _, top = pending.pop()
1231
relprefix = relroot + '/'
1234
top_slash = top + u'/'
1237
append = dirblock.append
1238
for name in sorted(_listdir(top)):
1239
name_utf8 = _utf8_encode(name)[0]
1240
abspath = top_slash + name
1241
statvalue = _lstat(abspath)
1242
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1243
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1244
yield (relroot, top), dirblock
1246
# push the user specified dirs from dirblock
1247
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1250
def copy_tree(from_path, to_path, handlers={}):
1251
"""Copy all of the entries in from_path into to_path.
1253
:param from_path: The base directory to copy.
1254
:param to_path: The target directory. If it does not exist, it will
1256
:param handlers: A dictionary of functions, which takes a source and
1257
destinations for files, directories, etc.
1258
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1259
'file', 'directory', and 'symlink' should always exist.
1260
If they are missing, they will be replaced with 'os.mkdir()',
1261
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1263
# Now, just copy the existing cached tree to the new location
1264
# We use a cheap trick here.
1265
# Absolute paths are prefixed with the first parameter
1266
# relative paths are prefixed with the second.
1267
# So we can get both the source and target returned
1268
# without any extra work.
1270
def copy_dir(source, dest):
1273
def copy_link(source, dest):
1274
"""Copy the contents of a symlink"""
1275
link_to = os.readlink(source)
1276
os.symlink(link_to, dest)
1278
real_handlers = {'file':shutil.copy2,
1279
'symlink':copy_link,
1280
'directory':copy_dir,
1282
real_handlers.update(handlers)
1284
if not os.path.exists(to_path):
1285
real_handlers['directory'](from_path, to_path)
1287
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1288
for relpath, name, kind, st, abspath in entries:
1289
real_handlers[kind](abspath, relpath)
1292
def path_prefix_key(path):
1293
"""Generate a prefix-order path key for path.
1295
This can be used to sort paths in the same way that walkdirs does.
1297
return (dirname(path) , path)
1300
def compare_paths_prefix_order(path_a, path_b):
1301
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1302
key_a = path_prefix_key(path_a)
1303
key_b = path_prefix_key(path_b)
1304
return cmp(key_a, key_b)
1307
_cached_user_encoding = None
1310
def get_user_encoding(use_cache=True):
1311
"""Find out what the preferred user encoding is.
1313
This is generally the encoding that is used for command line parameters
1314
and file contents. This may be different from the terminal encoding
1315
or the filesystem encoding.
1317
:param use_cache: Enable cache for detected encoding.
1318
(This parameter is turned on by default,
1319
and required only for selftesting)
1321
:return: A string defining the preferred user encoding
1323
global _cached_user_encoding
1324
if _cached_user_encoding is not None and use_cache:
1325
return _cached_user_encoding
1327
if sys.platform == 'darwin':
1328
# work around egregious python 2.4 bug
1329
sys.platform = 'posix'
1333
sys.platform = 'darwin'
1338
user_encoding = locale.getpreferredencoding()
1339
except locale.Error, e:
1340
sys.stderr.write('bzr: warning: %s\n'
1341
' Could not determine what text encoding to use.\n'
1342
' This error usually means your Python interpreter\n'
1343
' doesn\'t support the locale set by $LANG (%s)\n'
1344
" Continuing with ascii encoding.\n"
1345
% (e, os.environ.get('LANG')))
1346
user_encoding = 'ascii'
1348
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1349
# treat that as ASCII, and not support printing unicode characters to the
1351
if user_encoding in (None, 'cp0'):
1352
user_encoding = 'ascii'
1356
codecs.lookup(user_encoding)
1358
sys.stderr.write('bzr: warning:'
1359
' unknown encoding %s.'
1360
' Continuing with ascii encoding.\n'
1363
user_encoding = 'ascii'
1366
_cached_user_encoding = user_encoding
1368
return user_encoding
1371
def recv_all(socket, bytes):
1372
"""Receive an exact number of bytes.
1374
Regular Socket.recv() may return less than the requested number of bytes,
1375
dependning on what's in the OS buffer. MSG_WAITALL is not available
1376
on all platforms, but this should work everywhere. This will return
1377
less than the requested amount if the remote end closes.
1379
This isn't optimized and is intended mostly for use in testing.
1382
while len(b) < bytes:
1383
new = socket.recv(bytes - len(b))
1389
def dereference_path(path):
1390
"""Determine the real path to a file.
1392
All parent elements are dereferenced. But the file itself is not
1394
:param path: The original path. May be absolute or relative.
1395
:return: the real path *to* the file
1397
parent, base = os.path.split(path)
1398
# The pathjoin for '.' is a workaround for Python bug #1213894.
1399
# (initial path components aren't dereferenced)
1400
return pathjoin(realpath(pathjoin('.', parent)), base)
1403
def supports_mapi():
1404
"""Return True if we can use MAPI to launch a mail client."""
1405
return sys.platform == "win32"
added
removed
bzrlib/osutils.py
