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, types
20
from stat import S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE
22
from errors import bailout
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 (
58
from bzrlib import symbol_versioning
59
from bzrlib.symbol_versioning import (
62
from bzrlib.trace import mutter
65
# On win32, O_BINARY is used to indicate the file should
66
# be opened in binary mode, rather than text mode.
67
# On other platforms, O_BINARY doesn't exist, because
68
# they always open in binary mode, so it is okay to
69
# OR with 0 on those platforms
70
O_BINARY = getattr(os, 'O_BINARY', 0)
24
73
def make_readonly(filename):
25
74
"""Make a filename read-only."""
26
# TODO: probably needs to be fixed for windows
27
mod = os.stat(filename).st_mode
29
os.chmod(filename, mod)
75
mod = os.lstat(filename).st_mode
76
if not stat.S_ISLNK(mod):
78
os.chmod(filename, mod)
32
81
def make_writable(filename):
33
mod = os.stat(filename).st_mode
35
os.chmod(filename, mod)
38
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/_~-])')
82
mod = os.lstat(filename).st_mode
83
if not stat.S_ISLNK(mod):
85
os.chmod(filename, mod)
88
def minimum_path_selection(paths):
89
"""Return the smallset subset of paths which are outside paths.
91
:param paths: A container (and hence not None) of paths.
92
:return: A set of paths sufficient to include everything in paths via
93
is_inside_any, drawn from the paths parameter.
98
other_paths = paths.difference([path])
99
if not is_inside_any(other_paths, path):
100
# this is a top level path, we must check it.
101
search_paths.add(path)
40
"""Return shell-quoted filename"""
41
## We could be a bit more terse by using double-quotes etc
42
f = _QUOTE_RE.sub(r'\\\1', f)
49
mode = os.lstat(f)[ST_MODE]
57
bailout("can't handle file kind with mode %o of %r" % (mode, f))
109
"""Return a quoted filename filename
111
This previously used backslash quoting, but that works poorly on
113
# TODO: I'm not really sure this is the best format either.x
115
if _QUOTE_RE is None:
116
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
118
if _QUOTE_RE.search(f):
124
_directory_kind = 'directory'
127
stat.S_IFDIR:_directory_kind,
128
stat.S_IFCHR:'chardev',
129
stat.S_IFBLK:'block',
132
stat.S_IFLNK:'symlink',
133
stat.S_IFSOCK:'socket',
137
def file_kind_from_stat_mode(stat_mode, _formats=_formats, _unknown='unknown'):
138
"""Generate a file kind from a stat mode. This is used in walkdirs.
140
Its performance is critical: Do not mutate without careful benchmarking.
143
return _formats[stat_mode & 0170000]
148
def file_kind(f, _lstat=os.lstat, _mapper=file_kind_from_stat_mode):
150
return _mapper(_lstat(f).st_mode)
152
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
153
raise errors.NoSuchFile(f)
158
"""Return the current umask"""
159
# Assume that people aren't messing with the umask while running
160
# XXX: This is not thread safe, but there is no way to get the
161
# umask without setting it
169
_directory_kind: "/",
171
'tree-reference': '+',
175
def kind_marker(kind):
177
return _kind_marker_map[kind]
179
raise errors.BzrError('invalid file kind %r' % kind)
182
lexists = getattr(os.path, 'lexists', None)
186
stat = getattr(os, 'lstat', os.stat)
190
if e.errno == errno.ENOENT:
193
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
196
def fancy_rename(old, new, rename_func, unlink_func):
197
"""A fancy rename, when you don't have atomic rename.
199
:param old: The old path, to rename from
200
:param new: The new path, to rename to
201
:param rename_func: The potentially non-atomic rename function
202
:param unlink_func: A way to delete the target file if the full rename succeeds
205
# sftp rename doesn't allow overwriting, so play tricks:
207
base = os.path.basename(new)
208
dirname = os.path.dirname(new)
209
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
210
tmp_name = pathjoin(dirname, tmp_name)
212
# Rename the file out of the way, but keep track if it didn't exist
213
# We don't want to grab just any exception
214
# something like EACCES should prevent us from continuing
215
# The downside is that the rename_func has to throw an exception
216
# with an errno = ENOENT, or NoSuchFile
219
rename_func(new, tmp_name)
220
except (errors.NoSuchFile,), e:
223
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
224
# function raises an IOError with errno is None when a rename fails.
225
# This then gets caught here.
226
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
229
if (getattr(e, 'errno', None) is None
230
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
238
# This may throw an exception, in which case success will
240
rename_func(old, new)
242
except (IOError, OSError), e:
243
# source and target may be aliases of each other (e.g. on a
244
# case-insensitive filesystem), so we may have accidentally renamed
245
# source by when we tried to rename target
246
if not (file_existed and e.errno in (None, errno.ENOENT)):
250
# If the file used to exist, rename it back into place
251
# otherwise just delete it from the tmp location
253
unlink_func(tmp_name)
255
rename_func(tmp_name, new)
258
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
259
# choke on a Unicode string containing a relative path if
260
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
262
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
263
def _posix_abspath(path):
264
# jam 20060426 rather than encoding to fsencoding
265
# copy posixpath.abspath, but use os.getcwdu instead
266
if not posixpath.isabs(path):
267
path = posixpath.join(getcwd(), path)
268
return posixpath.normpath(path)
271
def _posix_realpath(path):
272
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
275
def _win32_fixdrive(path):
276
"""Force drive letters to be consistent.
278
win32 is inconsistent whether it returns lower or upper case
279
and even if it was consistent the user might type the other
280
so we force it to uppercase
281
running python.exe under cmd.exe return capital C:\\
282
running win32 python inside a cygwin shell returns lowercase c:\\
284
drive, path = _nt_splitdrive(path)
285
return drive.upper() + path
288
def _win32_abspath(path):
289
# Real _nt_abspath doesn't have a problem with a unicode cwd
290
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
293
def _win98_abspath(path):
294
"""Return the absolute version of a path.
295
Windows 98 safe implementation (python reimplementation
296
of Win32 API function GetFullPathNameW)
301
# \\HOST\path => //HOST/path
302
# //HOST/path => //HOST/path
303
# path => C:/cwd/path
306
# check for absolute path
307
drive = _nt_splitdrive(path)[0]
308
if drive == '' and path[:2] not in('//','\\\\'):
310
# we cannot simply os.path.join cwd and path
311
# because os.path.join('C:','/path') produce '/path'
312
# and this is incorrect
313
if path[:1] in ('/','\\'):
314
cwd = _nt_splitdrive(cwd)[0]
316
path = cwd + '\\' + path
317
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
319
if win32utils.winver == 'Windows 98':
320
_win32_abspath = _win98_abspath
323
def _win32_realpath(path):
324
# Real _nt_realpath doesn't have a problem with a unicode cwd
325
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
328
def _win32_pathjoin(*args):
329
return _nt_join(*args).replace('\\', '/')
332
def _win32_normpath(path):
333
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
337
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
340
def _win32_mkdtemp(*args, **kwargs):
341
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
344
def _win32_rename(old, new):
345
"""We expect to be able to atomically replace 'new' with old.
347
On win32, if new exists, it must be moved out of the way first,
351
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
353
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
354
# If we try to rename a non-existant file onto cwd, we get
355
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
356
# if the old path doesn't exist, sometimes we get EACCES
357
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
363
return unicodedata.normalize('NFC', os.getcwdu())
366
# Default is to just use the python builtins, but these can be rebound on
367
# particular platforms.
368
abspath = _posix_abspath
369
realpath = _posix_realpath
370
pathjoin = os.path.join
371
normpath = os.path.normpath
374
dirname = os.path.dirname
375
basename = os.path.basename
376
split = os.path.split
377
splitext = os.path.splitext
378
# These were already imported into local scope
379
# mkdtemp = tempfile.mkdtemp
380
# rmtree = shutil.rmtree
382
MIN_ABS_PATHLENGTH = 1
385
if sys.platform == 'win32':
386
abspath = _win32_abspath
387
realpath = _win32_realpath
388
pathjoin = _win32_pathjoin
389
normpath = _win32_normpath
390
getcwd = _win32_getcwd
391
mkdtemp = _win32_mkdtemp
392
rename = _win32_rename
394
MIN_ABS_PATHLENGTH = 3
396
def _win32_delete_readonly(function, path, excinfo):
397
"""Error handler for shutil.rmtree function [for win32]
398
Helps to remove files and dirs marked as read-only.
400
exception = excinfo[1]
401
if function in (os.remove, os.rmdir) \
402
and isinstance(exception, OSError) \
403
and exception.errno == errno.EACCES:
409
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
410
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
411
return shutil.rmtree(path, ignore_errors, onerror)
412
elif sys.platform == 'darwin':
416
def get_terminal_encoding():
417
"""Find the best encoding for printing to the screen.
419
This attempts to check both sys.stdout and sys.stdin to see
420
what encoding they are in, and if that fails it falls back to
421
bzrlib.user_encoding.
422
The problem is that on Windows, locale.getpreferredencoding()
423
is not the same encoding as that used by the console:
424
http://mail.python.org/pipermail/python-list/2003-May/162357.html
426
On my standard US Windows XP, the preferred encoding is
427
cp1252, but the console is cp437
429
output_encoding = getattr(sys.stdout, 'encoding', None)
430
if not output_encoding:
431
input_encoding = getattr(sys.stdin, 'encoding', None)
432
if not input_encoding:
433
output_encoding = bzrlib.user_encoding
434
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
436
output_encoding = input_encoding
437
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
439
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
440
if output_encoding == 'cp0':
441
# invalid encoding (cp0 means 'no codepage' on Windows)
442
output_encoding = bzrlib.user_encoding
443
mutter('cp0 is invalid encoding.'
444
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
447
codecs.lookup(output_encoding)
449
sys.stderr.write('bzr: warning:'
450
' unknown terminal encoding %s.\n'
451
' Using encoding %s instead.\n'
452
% (output_encoding, bzrlib.user_encoding)
454
output_encoding = bzrlib.user_encoding
456
return output_encoding
459
def normalizepath(f):
460
if getattr(os.path, 'realpath', None) is not None:
464
[p,e] = os.path.split(f)
465
if e == "" or e == "." or e == "..":
468
return pathjoin(F(p), e)
78
def pumpfile(fromfile, tofile):
79
"""Copy contents of one file to another."""
80
tofile.write(fromfile.read())
84
"""Return a new UUID"""
86
## XXX: Could alternatively read /proc/sys/kernel/random/uuid on
87
## Linux, but we need something portable for other systems;
88
## preferably an implementation in Python.
89
bailout('uuids not allowed!')
90
return chomp(os.popen('uuidgen').readline())
93
if s and (s[-1] == '\n'):
487
"""True if f is a symlink."""
489
return S_ISLNK(os.lstat(f)[ST_MODE])
493
def is_inside(dir, fname):
494
"""True if fname is inside dir.
496
The parameters should typically be passed to osutils.normpath first, so
497
that . and .. and repeated slashes are eliminated, and the separators
498
are canonical for the platform.
500
The empty string as a dir name is taken as top-of-tree and matches
503
# XXX: Most callers of this can actually do something smarter by
504
# looking at the inventory
514
return fname.startswith(dir)
517
def is_inside_any(dir_list, fname):
518
"""True if fname is inside any of given dirs."""
519
for dirname in dir_list:
520
if is_inside(dirname, fname):
525
def is_inside_or_parent_of_any(dir_list, fname):
526
"""True if fname is a child or a parent of any of the given files."""
527
for dirname in dir_list:
528
if is_inside(dirname, fname) or is_inside(fname, dirname):
533
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768):
534
"""Copy contents of one file to another.
536
The read_length can either be -1 to read to end-of-file (EOF) or
537
it can specify the maximum number of bytes to read.
539
The buff_size represents the maximum size for each read operation
540
performed on from_file.
542
:return: The number of bytes copied.
546
# read specified number of bytes
548
while read_length > 0:
549
num_bytes_to_read = min(read_length, buff_size)
551
block = from_file.read(num_bytes_to_read)
557
actual_bytes_read = len(block)
558
read_length -= actual_bytes_read
559
length += actual_bytes_read
563
block = from_file.read(buff_size)
572
def file_iterator(input_file, readsize=32768):
574
b = input_file.read(readsize)
101
## TODO: Maybe read in chunks to handle big files
102
if hasattr(f, 'tell'):
118
"""Return email-style username.
120
Something similar to 'Martin Pool <mbp@sourcefrog.net>'
122
:todo: Check it's reasonably well-formed.
124
:todo: Allow taking it from a dotfile to help people on windows
125
who can't easily set variables.
127
:todo: Cope without pwd module, which is only on unix.
581
"""Calculate the hexdigest of an open file.
583
The file cursor should be already at the start.
129
e = os.environ.get('BZREMAIL') or os.environ.get('EMAIL')
595
def sha_file_by_name(fname):
596
"""Calculate the SHA1 of a file by reading the full text"""
598
f = os.open(fname, os.O_RDONLY | O_BINARY)
137
w = pwd.getpwuid(uid)
139
comma = gecos.find(',')
143
realname = gecos[:comma]
144
return '%s <%s@%s>' % (realname, w.pw_name, socket.getfqdn())
148
import getpass, socket
149
return '<%s@%s>' % (getpass.getuser(), socket.getfqdn())
153
"""Return just the email component of a username."""
154
e = os.environ.get('BZREMAIL') or os.environ.get('EMAIL')
157
m = re.search(r'[\w+.-]+@[\w+.-]+', e)
159
bailout('%r is not a reasonable email address' % e)
163
import getpass, socket
164
return '%s@%s' % (getpass.getuser(), socket.getfqdn())
601
b = os.read(f, 1<<16)
609
def sha_strings(strings, _factory=sha.new):
610
"""Return the sha-1 of concatenation of strings"""
612
map(s.update, strings)
616
def sha_string(f, _factory=sha.new):
617
return _factory(f).hexdigest()
620
def fingerprint_file(f):
622
return {'size': len(b),
623
'sha1': sha.new(b).hexdigest()}
169
626
def compare_files(a, b):
170
627
"""Returns true if equal in contents"""
171
# TODO: don't read the whole thing in one go.
172
result = a.read() == b.read()
177
def local_time_offset():
181
return -time.timezone
184
def format_date(t, offset=0, timezone='original'):
185
## TODO: Perhaps a global option to use either universal or local time?
186
## Or perhaps just let people set $TZ?
189
assert isinstance(t, float)
638
def local_time_offset(t=None):
639
"""Return offset of local zone from GMT, either at present or at time t."""
642
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
643
return offset.days * 86400 + offset.seconds
645
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
647
def format_date(t, offset=0, timezone='original', date_fmt=None,
649
"""Return a formatted date string.
651
:param t: Seconds since the epoch.
652
:param offset: Timezone offset in seconds east of utc.
653
:param timezone: How to display the time: 'utc', 'original' for the
654
timezone specified by offset, or 'local' for the process's current
656
:param show_offset: Whether to append the timezone.
657
:param date_fmt: strftime format.
191
659
if timezone == 'utc':
192
660
tt = time.gmtime(t)
194
662
elif timezone == 'original':
197
665
tt = time.gmtime(t + offset)
198
666
elif timezone == 'local':
199
667
tt = time.localtime(t)
200
offset = local_time_offset()
202
bailout("unsupported timezone format %r",
203
['options are "utc", "original", "local"'])
205
return (time.strftime("%a %Y-%m-%d %H:%M:%S", tt)
206
+ ' %+03d%02d' % (offset / 3600, (offset / 60) % 60))
668
offset = local_time_offset(t)
670
raise errors.UnsupportedTimezoneFormat(timezone)
672
date_fmt = "%a %Y-%m-%d %H:%M:%S"
674
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
677
# day of week depends on locale, so we do this ourself
678
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
679
return (time.strftime(date_fmt, tt) + offset_str)
209
682
def compact_date(when):
210
683
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
686
def format_delta(delta):
687
"""Get a nice looking string for a time delta.
689
:param delta: The time difference in seconds, can be positive or negative.
690
positive indicates time in the past, negative indicates time in the
691
future. (usually time.time() - stored_time)
692
:return: String formatted to show approximate resolution
698
direction = 'in the future'
702
if seconds < 90: # print seconds up to 90 seconds
704
return '%d second %s' % (seconds, direction,)
706
return '%d seconds %s' % (seconds, direction)
708
minutes = int(seconds / 60)
709
seconds -= 60 * minutes
714
if minutes < 90: # print minutes, seconds up to 90 minutes
716
return '%d minute, %d second%s %s' % (
717
minutes, seconds, plural_seconds, direction)
719
return '%d minutes, %d second%s %s' % (
720
minutes, seconds, plural_seconds, direction)
722
hours = int(minutes / 60)
723
minutes -= 60 * hours
730
return '%d hour, %d minute%s %s' % (hours, minutes,
731
plural_minutes, direction)
732
return '%d hours, %d minute%s %s' % (hours, minutes,
733
plural_minutes, direction)
215
736
"""Return size of given open file."""
216
737
return os.fstat(f.fileno())[ST_SIZE]
219
if hasattr(os, 'urandom'): # python 2.4 and later
740
# Define rand_bytes based on platform.
742
# Python 2.4 and later have os.urandom,
743
# but it doesn't work on some arches
220
745
rand_bytes = os.urandom
222
# FIXME: No good on non-Linux
223
_rand_file = file('/dev/urandom', 'rb')
224
rand_bytes = _rand_file.read
746
except (NotImplementedError, AttributeError):
747
# If python doesn't have os.urandom, or it doesn't work,
748
# then try to first pull random data from /dev/urandom
750
rand_bytes = file('/dev/urandom', 'rb').read
751
# Otherwise, use this hack as a last resort
752
except (IOError, OSError):
753
# not well seeded, but better than nothing
758
s += chr(random.randint(0, 255))
763
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
765
"""Return a random string of num alphanumeric characters
767
The result only contains lowercase chars because it may be used on
768
case-insensitive filesystems.
771
for raw_byte in rand_bytes(num):
772
s += ALNUM[ord(raw_byte) % 36]
227
776
## TODO: We could later have path objects that remember their list
228
777
## decomposition (might be too tricksy though.)
230
779
def splitpath(p):
231
"""Turn string into list of parts.
780
"""Turn string into list of parts."""
781
# split on either delimiter because people might use either on
783
ps = re.split(r'[\\/]', p)
237
>>> splitpath('a/./b')
239
>>> splitpath('a/.b')
241
>>> splitpath('a/../b')
242
Traceback (most recent call last):
244
BzrError: ("sorry, '..' not allowed in path", [])
246
assert isinstance(p, types.StringTypes)
247
ps = [f for f in p.split('/') if f != '.']
250
bailout("sorry, %r not allowed in path" % f)
788
raise errors.BzrError("sorry, %r not allowed in path" % f)
789
elif (f == '.') or (f == ''):
254
assert isinstance(p, list)
256
797
if (f == '..') or (f is None) or (f == ''):
257
bailout("sorry, %r not allowed in path" % f)
261
def appendpath(p1, p2):
268
def extern_command(cmd, ignore_errors = False):
269
mutter('external command: %s' % `cmd`)
271
if not ignore_errors:
272
bailout('command failed')
798
raise errors.BzrError("sorry, %r not allowed in path" % f)
803
"""Split s into lines, but without removing the newline characters."""
804
lines = s.split('\n')
805
result = [line + '\n' for line in lines[:-1]]
807
result.append(lines[-1])
811
def hardlinks_good():
812
return sys.platform not in ('win32', 'cygwin', 'darwin')
815
def link_or_copy(src, dest):
816
"""Hardlink a file, or copy it if it can't be hardlinked."""
817
if not hardlinks_good():
818
shutil.copyfile(src, dest)
822
except (OSError, IOError), e:
823
if e.errno != errno.EXDEV:
825
shutil.copyfile(src, dest)
828
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
829
# Forgiveness than Permission (EAFP) because:
830
# - root can damage a solaris file system by using unlink,
831
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
832
# EACCES, OSX: EPERM) when invoked on a directory.
833
def delete_any(path):
834
"""Delete a file or directory."""
835
if isdir(path): # Takes care of symlinks
842
if getattr(os, 'symlink', None) is not None:
849
if getattr(os, 'link', None) is not None:
855
def host_os_dereferences_symlinks():
856
return (has_symlinks()
857
and sys.platform not in ('cygwin', 'win32'))
860
def contains_whitespace(s):
861
"""True if there are any whitespace characters in s."""
862
# string.whitespace can include '\xa0' in certain locales, because it is
863
# considered "non-breaking-space" as part of ISO-8859-1. But it
864
# 1) Isn't a breaking whitespace
865
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
867
# 3) '\xa0' isn't unicode safe since it is >128.
869
# This should *not* be a unicode set of characters in case the source
870
# string is not a Unicode string. We can auto-up-cast the characters since
871
# they are ascii, but we don't want to auto-up-cast the string in case it
873
for ch in ' \t\n\r\v\f':
880
def contains_linebreaks(s):
881
"""True if there is any vertical whitespace in s."""
889
def relpath(base, path):
890
"""Return path relative to base, or raise exception.
892
The path may be either an absolute path or a path relative to the
893
current working directory.
895
os.path.commonprefix (python2.4) has a bad bug that it works just
896
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
900
if len(base) < MIN_ABS_PATHLENGTH:
901
# must have space for e.g. a drive letter
902
raise ValueError('%r is too short to calculate a relative path'
909
while len(head) >= len(base):
912
head, tail = os.path.split(head)
916
raise errors.PathNotChild(rp, base)
924
def safe_unicode(unicode_or_utf8_string):
925
"""Coerce unicode_or_utf8_string into unicode.
927
If it is unicode, it is returned.
928
Otherwise it is decoded from utf-8. If a decoding error
929
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
930
as a BzrBadParameter exception.
932
if isinstance(unicode_or_utf8_string, unicode):
933
return unicode_or_utf8_string
935
return unicode_or_utf8_string.decode('utf8')
936
except UnicodeDecodeError:
937
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
940
def safe_utf8(unicode_or_utf8_string):
941
"""Coerce unicode_or_utf8_string to a utf8 string.
943
If it is a str, it is returned.
944
If it is Unicode, it is encoded into a utf-8 string.
946
if isinstance(unicode_or_utf8_string, str):
947
# TODO: jam 20070209 This is overkill, and probably has an impact on
948
# performance if we are dealing with lots of apis that want a
951
# Make sure it is a valid utf-8 string
952
unicode_or_utf8_string.decode('utf-8')
953
except UnicodeDecodeError:
954
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
955
return unicode_or_utf8_string
956
return unicode_or_utf8_string.encode('utf-8')
959
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
960
' Revision id generators should be creating utf8'
964
def safe_revision_id(unicode_or_utf8_string, warn=True):
965
"""Revision ids should now be utf8, but at one point they were unicode.
967
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
969
:param warn: Functions that are sanitizing user data can set warn=False
970
:return: None or a utf8 revision id.
972
if (unicode_or_utf8_string is None
973
or unicode_or_utf8_string.__class__ == str):
974
return unicode_or_utf8_string
976
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
978
return cache_utf8.encode(unicode_or_utf8_string)
981
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
982
' generators should be creating utf8 file ids.')
985
def safe_file_id(unicode_or_utf8_string, warn=True):
986
"""File ids should now be utf8, but at one point they were unicode.
988
This is the same as safe_utf8, except it uses the cached encode functions
989
to save a little bit of performance.
991
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
993
:param warn: Functions that are sanitizing user data can set warn=False
994
:return: None or a utf8 file id.
996
if (unicode_or_utf8_string is None
997
or unicode_or_utf8_string.__class__ == str):
998
return unicode_or_utf8_string
1000
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1002
return cache_utf8.encode(unicode_or_utf8_string)
1005
_platform_normalizes_filenames = False
1006
if sys.platform == 'darwin':
1007
_platform_normalizes_filenames = True
1010
def normalizes_filenames():
1011
"""Return True if this platform normalizes unicode filenames.
1013
Mac OSX does, Windows/Linux do not.
1015
return _platform_normalizes_filenames
1018
def _accessible_normalized_filename(path):
1019
"""Get the unicode normalized path, and if you can access the file.
1021
On platforms where the system normalizes filenames (Mac OSX),
1022
you can access a file by any path which will normalize correctly.
1023
On platforms where the system does not normalize filenames
1024
(Windows, Linux), you have to access a file by its exact path.
1026
Internally, bzr only supports NFC normalization, since that is
1027
the standard for XML documents.
1029
So return the normalized path, and a flag indicating if the file
1030
can be accessed by that path.
1033
return unicodedata.normalize('NFC', unicode(path)), True
1036
def _inaccessible_normalized_filename(path):
1037
__doc__ = _accessible_normalized_filename.__doc__
1039
normalized = unicodedata.normalize('NFC', unicode(path))
1040
return normalized, normalized == path
1043
if _platform_normalizes_filenames:
1044
normalized_filename = _accessible_normalized_filename
1046
normalized_filename = _inaccessible_normalized_filename
1049
def terminal_width():
1050
"""Return estimated terminal width."""
1051
if sys.platform == 'win32':
1052
return win32utils.get_console_size()[0]
1055
import struct, fcntl, termios
1056
s = struct.pack('HHHH', 0, 0, 0, 0)
1057
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1058
width = struct.unpack('HHHH', x)[1]
1063
width = int(os.environ['COLUMNS'])
1072
def supports_executable():
1073
return sys.platform != "win32"
1076
def supports_posix_readonly():
1077
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1079
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1080
directory controls creation/deletion, etc.
1082
And under win32, readonly means that the directory itself cannot be
1083
deleted. The contents of a readonly directory can be changed, unlike POSIX
1084
where files in readonly directories cannot be added, deleted or renamed.
1086
return sys.platform != "win32"
1089
def set_or_unset_env(env_variable, value):
1090
"""Modify the environment, setting or removing the env_variable.
1092
:param env_variable: The environment variable in question
1093
:param value: The value to set the environment to. If None, then
1094
the variable will be removed.
1095
:return: The original value of the environment variable.
1097
orig_val = os.environ.get(env_variable)
1099
if orig_val is not None:
1100
del os.environ[env_variable]
1102
if isinstance(value, unicode):
1103
value = value.encode(bzrlib.user_encoding)
1104
os.environ[env_variable] = value
1108
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1111
def check_legal_path(path):
1112
"""Check whether the supplied path is legal.
1113
This is only required on Windows, so we don't test on other platforms
1116
if sys.platform != "win32":
1118
if _validWin32PathRE.match(path) is None:
1119
raise errors.IllegalPath(path)
1122
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1124
def _is_error_enotdir(e):
1125
"""Check if this exception represents ENOTDIR.
1127
Unfortunately, python is very inconsistent about the exception
1128
here. The cases are:
1129
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1130
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1131
which is the windows error code.
1132
3) Windows, Python2.5 uses errno == EINVAL and
1133
winerror == ERROR_DIRECTORY
1135
:param e: An Exception object (expected to be OSError with an errno
1136
attribute, but we should be able to cope with anything)
1137
:return: True if this represents an ENOTDIR error. False otherwise.
1139
en = getattr(e, 'errno', None)
1140
if (en == errno.ENOTDIR
1141
or (sys.platform == 'win32'
1142
and (en == _WIN32_ERROR_DIRECTORY
1143
or (en == errno.EINVAL
1144
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1150
def walkdirs(top, prefix=""):
1151
"""Yield data about all the directories in a tree.
1153
This yields all the data about the contents of a directory at a time.
1154
After each directory has been yielded, if the caller has mutated the list
1155
to exclude some directories, they are then not descended into.
1157
The data yielded is of the form:
1158
((directory-relpath, directory-path-from-top),
1159
[(relpath, basename, kind, lstat, path-from-top), ...]),
1160
- directory-relpath is the relative path of the directory being returned
1161
with respect to top. prefix is prepended to this.
1162
- directory-path-from-root is the path including top for this directory.
1163
It is suitable for use with os functions.
1164
- relpath is the relative path within the subtree being walked.
1165
- basename is the basename of the path
1166
- kind is the kind of the file now. If unknown then the file is not
1167
present within the tree - but it may be recorded as versioned. See
1169
- lstat is the stat data *if* the file was statted.
1170
- planned, not implemented:
1171
path_from_tree_root is the path from the root of the tree.
1173
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1174
allows one to walk a subtree but get paths that are relative to a tree
1176
:return: an iterator over the dirs.
1178
#TODO there is a bit of a smell where the results of the directory-
1179
# summary in this, and the path from the root, may not agree
1180
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1181
# potentially confusing output. We should make this more robust - but
1182
# not at a speed cost. RBC 20060731
1184
_directory = _directory_kind
1185
_listdir = os.listdir
1186
_kind_from_mode = _formats.get
1187
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1189
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1190
relroot, _, _, _, top = pending.pop()
1192
relprefix = relroot + u'/'
1195
top_slash = top + u'/'
1198
append = dirblock.append
1200
names = sorted(_listdir(top))
1202
if not _is_error_enotdir(e):
1206
abspath = top_slash + name
1207
statvalue = _lstat(abspath)
1208
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1209
append((relprefix + name, name, kind, statvalue, abspath))
1210
yield (relroot, top), dirblock
1212
# push the user specified dirs from dirblock
1213
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1216
_real_walkdirs_utf8 = None
1218
def _walkdirs_utf8(top, prefix=""):
1219
"""Yield data about all the directories in a tree.
1221
This yields the same information as walkdirs() only each entry is yielded
1222
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1223
are returned as exact byte-strings.
1225
:return: yields a tuple of (dir_info, [file_info])
1226
dir_info is (utf8_relpath, path-from-top)
1227
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1228
if top is an absolute path, path-from-top is also an absolute path.
1229
path-from-top might be unicode or utf8, but it is the correct path to
1230
pass to os functions to affect the file in question. (such as os.lstat)
1232
global _real_walkdirs_utf8
1233
if _real_walkdirs_utf8 is None:
1234
fs_encoding = _fs_enc.upper()
1235
if win32utils.winver == 'Windows NT':
1236
# Win98 doesn't have unicode apis like FindFirstFileW
1237
# TODO: We possibly could support Win98 by falling back to the
1238
# original FindFirstFile, and using TCHAR instead of WCHAR,
1239
# but that gets a bit tricky, and requires custom compiling
1242
from bzrlib._walkdirs_win32 import _walkdirs_utf8_win32_find_file
1244
_real_walkdirs_utf8 = _walkdirs_unicode_to_utf8
1246
_real_walkdirs_utf8 = _walkdirs_utf8_win32_find_file
1247
elif fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1248
# ANSI_X3.4-1968 is a form of ASCII
1249
_real_walkdirs_utf8 = _walkdirs_unicode_to_utf8
1251
_real_walkdirs_utf8 = _walkdirs_fs_utf8
1252
return _real_walkdirs_utf8(top, prefix=prefix)
1255
def _walkdirs_fs_utf8(top, prefix=""):
1256
"""See _walkdirs_utf8.
1258
This sub-function is called when we know the filesystem is already in utf8
1259
encoding. So we don't need to transcode filenames.
1262
_directory = _directory_kind
1263
_listdir = os.listdir
1264
_kind_from_mode = _formats.get
1266
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1267
# But we don't actually uses 1-3 in pending, so set them to None
1268
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1270
relroot, _, _, _, top = pending.pop()
1272
relprefix = relroot + '/'
1275
top_slash = top + '/'
1278
append = dirblock.append
1279
for name in sorted(_listdir(top)):
1280
abspath = top_slash + name
1281
statvalue = _lstat(abspath)
1282
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1283
append((relprefix + name, name, kind, statvalue, abspath))
1284
yield (relroot, top), dirblock
1286
# push the user specified dirs from dirblock
1287
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1290
def _walkdirs_unicode_to_utf8(top, prefix=""):
1291
"""See _walkdirs_utf8
1293
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1295
This is currently the fallback code path when the filesystem encoding is
1296
not UTF-8. It may be better to implement an alternative so that we can
1297
safely handle paths that are not properly decodable in the current
1300
_utf8_encode = codecs.getencoder('utf8')
1302
_directory = _directory_kind
1303
_listdir = os.listdir
1304
_kind_from_mode = _formats.get
1306
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1308
relroot, _, _, _, top = pending.pop()
1310
relprefix = relroot + '/'
1313
top_slash = top + u'/'
1316
append = dirblock.append
1317
for name in sorted(_listdir(top)):
1318
name_utf8 = _utf8_encode(name)[0]
1319
abspath = top_slash + name
1320
statvalue = _lstat(abspath)
1321
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1322
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1323
yield (relroot, top), dirblock
1325
# push the user specified dirs from dirblock
1326
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1329
def copy_tree(from_path, to_path, handlers={}):
1330
"""Copy all of the entries in from_path into to_path.
1332
:param from_path: The base directory to copy.
1333
:param to_path: The target directory. If it does not exist, it will
1335
:param handlers: A dictionary of functions, which takes a source and
1336
destinations for files, directories, etc.
1337
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1338
'file', 'directory', and 'symlink' should always exist.
1339
If they are missing, they will be replaced with 'os.mkdir()',
1340
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1342
# Now, just copy the existing cached tree to the new location
1343
# We use a cheap trick here.
1344
# Absolute paths are prefixed with the first parameter
1345
# relative paths are prefixed with the second.
1346
# So we can get both the source and target returned
1347
# without any extra work.
1349
def copy_dir(source, dest):
1352
def copy_link(source, dest):
1353
"""Copy the contents of a symlink"""
1354
link_to = os.readlink(source)
1355
os.symlink(link_to, dest)
1357
real_handlers = {'file':shutil.copy2,
1358
'symlink':copy_link,
1359
'directory':copy_dir,
1361
real_handlers.update(handlers)
1363
if not os.path.exists(to_path):
1364
real_handlers['directory'](from_path, to_path)
1366
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1367
for relpath, name, kind, st, abspath in entries:
1368
real_handlers[kind](abspath, relpath)
1371
def path_prefix_key(path):
1372
"""Generate a prefix-order path key for path.
1374
This can be used to sort paths in the same way that walkdirs does.
1376
return (dirname(path) , path)
1379
def compare_paths_prefix_order(path_a, path_b):
1380
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1381
key_a = path_prefix_key(path_a)
1382
key_b = path_prefix_key(path_b)
1383
return cmp(key_a, key_b)
1386
_cached_user_encoding = None
1389
def get_user_encoding(use_cache=True):
1390
"""Find out what the preferred user encoding is.
1392
This is generally the encoding that is used for command line parameters
1393
and file contents. This may be different from the terminal encoding
1394
or the filesystem encoding.
1396
:param use_cache: Enable cache for detected encoding.
1397
(This parameter is turned on by default,
1398
and required only for selftesting)
1400
:return: A string defining the preferred user encoding
1402
global _cached_user_encoding
1403
if _cached_user_encoding is not None and use_cache:
1404
return _cached_user_encoding
1406
if sys.platform == 'darwin':
1407
# work around egregious python 2.4 bug
1408
sys.platform = 'posix'
1412
sys.platform = 'darwin'
1417
user_encoding = locale.getpreferredencoding()
1418
except locale.Error, e:
1419
sys.stderr.write('bzr: warning: %s\n'
1420
' Could not determine what text encoding to use.\n'
1421
' This error usually means your Python interpreter\n'
1422
' doesn\'t support the locale set by $LANG (%s)\n'
1423
" Continuing with ascii encoding.\n"
1424
% (e, os.environ.get('LANG')))
1425
user_encoding = 'ascii'
1427
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1428
# treat that as ASCII, and not support printing unicode characters to the
1431
# For python scripts run under vim, we get '', so also treat that as ASCII
1432
if user_encoding in (None, 'cp0', ''):
1433
user_encoding = 'ascii'
1437
codecs.lookup(user_encoding)
1439
sys.stderr.write('bzr: warning:'
1440
' unknown encoding %s.'
1441
' Continuing with ascii encoding.\n'
1444
user_encoding = 'ascii'
1447
_cached_user_encoding = user_encoding
1449
return user_encoding
1452
def recv_all(socket, bytes):
1453
"""Receive an exact number of bytes.
1455
Regular Socket.recv() may return less than the requested number of bytes,
1456
dependning on what's in the OS buffer. MSG_WAITALL is not available
1457
on all platforms, but this should work everywhere. This will return
1458
less than the requested amount if the remote end closes.
1460
This isn't optimized and is intended mostly for use in testing.
1463
while len(b) < bytes:
1464
new = socket.recv(bytes - len(b))
1471
def send_all(socket, bytes):
1472
"""Send all bytes on a socket.
1474
Regular socket.sendall() can give socket error 10053 on Windows. This
1475
implementation sends no more than 64k at a time, which avoids this problem.
1478
for pos in xrange(0, len(bytes), chunk_size):
1479
socket.sendall(bytes[pos:pos+chunk_size])
1482
def dereference_path(path):
1483
"""Determine the real path to a file.
1485
All parent elements are dereferenced. But the file itself is not
1487
:param path: The original path. May be absolute or relative.
1488
:return: the real path *to* the file
1490
parent, base = os.path.split(path)
1491
# The pathjoin for '.' is a workaround for Python bug #1213894.
1492
# (initial path components aren't dereferenced)
1493
return pathjoin(realpath(pathjoin('.', parent)), base)
1496
def supports_mapi():
1497
"""Return True if we can use MAPI to launch a mail client."""
1498
return sys.platform == "win32"
1501
def resource_string(package, resource_name):
1502
"""Load a resource from a package and return it as a string.
1504
Note: Only packages that start with bzrlib are currently supported.
1506
This is designed to be a lightweight implementation of resource
1507
loading in a way which is API compatible with the same API from
1509
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1510
If and when pkg_resources becomes a standard library, this routine
1513
# Check package name is within bzrlib
1514
if package == "bzrlib":
1515
resource_relpath = resource_name
1516
elif package.startswith("bzrlib."):
1517
package = package[len("bzrlib."):].replace('.', os.sep)
1518
resource_relpath = pathjoin(package, resource_name)
1520
raise errors.BzrError('resource package %s not in bzrlib' % package)
1522
# Map the resource to a file and read its contents
1523
base = dirname(bzrlib.__file__)
1524
if getattr(sys, 'frozen', None): # bzr.exe
1525
base = abspath(pathjoin(base, '..', '..'))
1526
filename = pathjoin(base, resource_relpath)
1527
return open(filename, 'rU').read()
added
removed
bzrlib/osutils.py
