1
# Copyright (C) 2005-2010 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20
from stat import (S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE,
21
S_ISCHR, S_ISBLK, S_ISFIFO, S_ISSOCK)
27
from bzrlib.lazy_import import lazy_import
28
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,
45
from tempfile import (
57
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
59
if sys.version_info < (2, 5):
60
import md5 as _mod_md5
62
import sha as _mod_sha
72
from bzrlib import symbol_versioning
75
# Cross platform wall-clock time functionality with decent resolution.
76
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
77
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
78
# synchronized with ``time.time()``, this is only meant to be used to find
79
# delta times by subtracting from another call to this function.
80
timer_func = time.time
81
if sys.platform == 'win32':
82
timer_func = time.clock
84
# On win32, O_BINARY is used to indicate the file should
85
# be opened in binary mode, rather than text mode.
86
# On other platforms, O_BINARY doesn't exist, because
87
# they always open in binary mode, so it is okay to
88
# OR with 0 on those platforms.
89
# O_NOINHERIT and O_TEXT exists only on win32 too.
90
O_BINARY = getattr(os, 'O_BINARY', 0)
91
O_TEXT = getattr(os, 'O_TEXT', 0)
92
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
95
def get_unicode_argv():
97
user_encoding = get_user_encoding()
98
return [a.decode(user_encoding) for a in sys.argv[1:]]
99
except UnicodeDecodeError:
100
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
104
def make_readonly(filename):
105
"""Make a filename read-only."""
106
mod = os.lstat(filename).st_mode
107
if not stat.S_ISLNK(mod):
109
os.chmod(filename, mod)
112
def make_writable(filename):
113
mod = os.lstat(filename).st_mode
114
if not stat.S_ISLNK(mod):
116
os.chmod(filename, mod)
119
def minimum_path_selection(paths):
120
"""Return the smallset subset of paths which are outside paths.
122
:param paths: A container (and hence not None) of paths.
123
:return: A set of paths sufficient to include everything in paths via
124
is_inside, drawn from the paths parameter.
130
return path.split('/')
131
sorted_paths = sorted(list(paths), key=sort_key)
133
search_paths = [sorted_paths[0]]
134
for path in sorted_paths[1:]:
135
if not is_inside(search_paths[-1], path):
136
# This path is unique, add it
137
search_paths.append(path)
139
return set(search_paths)
146
"""Return a quoted filename filename
148
This previously used backslash quoting, but that works poorly on
150
# TODO: I'm not really sure this is the best format either.x
152
if _QUOTE_RE is None:
153
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
155
if _QUOTE_RE.search(f):
161
_directory_kind = 'directory'
164
"""Return the current umask"""
165
# Assume that people aren't messing with the umask while running
166
# XXX: This is not thread safe, but there is no way to get the
167
# umask without setting it
175
_directory_kind: "/",
177
'tree-reference': '+',
181
def kind_marker(kind):
183
return _kind_marker_map[kind]
185
# Slightly faster than using .get(, '') when the common case is that
190
lexists = getattr(os.path, 'lexists', None)
194
stat = getattr(os, 'lstat', os.stat)
198
if e.errno == errno.ENOENT:
201
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
204
def fancy_rename(old, new, rename_func, unlink_func):
205
"""A fancy rename, when you don't have atomic rename.
207
:param old: The old path, to rename from
208
:param new: The new path, to rename to
209
:param rename_func: The potentially non-atomic rename function
210
:param unlink_func: A way to delete the target file if the full rename
213
# sftp rename doesn't allow overwriting, so play tricks:
214
base = os.path.basename(new)
215
dirname = os.path.dirname(new)
216
# callers use different encodings for the paths so the following MUST
217
# respect that. We rely on python upcasting to unicode if new is unicode
218
# and keeping a str if not.
219
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
220
os.getpid(), rand_chars(10))
221
tmp_name = pathjoin(dirname, tmp_name)
223
# Rename the file out of the way, but keep track if it didn't exist
224
# We don't want to grab just any exception
225
# something like EACCES should prevent us from continuing
226
# The downside is that the rename_func has to throw an exception
227
# with an errno = ENOENT, or NoSuchFile
230
rename_func(new, tmp_name)
231
except (errors.NoSuchFile,), e:
234
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
235
# function raises an IOError with errno is None when a rename fails.
236
# This then gets caught here.
237
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
240
if (getattr(e, 'errno', None) is None
241
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
250
# This may throw an exception, in which case success will
252
rename_func(old, new)
254
except (IOError, OSError), e:
255
# source and target may be aliases of each other (e.g. on a
256
# case-insensitive filesystem), so we may have accidentally renamed
257
# source by when we tried to rename target
258
failure_exc = sys.exc_info()
259
if (file_existed and e.errno in (None, errno.ENOENT)
260
and old.lower() == new.lower()):
261
# source and target are the same file on a case-insensitive
262
# filesystem, so we don't generate an exception
266
# If the file used to exist, rename it back into place
267
# otherwise just delete it from the tmp location
269
unlink_func(tmp_name)
271
rename_func(tmp_name, new)
272
if failure_exc is not None:
273
raise failure_exc[0], failure_exc[1], failure_exc[2]
276
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
277
# choke on a Unicode string containing a relative path if
278
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
280
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
281
def _posix_abspath(path):
282
# jam 20060426 rather than encoding to fsencoding
283
# copy posixpath.abspath, but use os.getcwdu instead
284
if not posixpath.isabs(path):
285
path = posixpath.join(getcwd(), path)
286
return posixpath.normpath(path)
289
def _posix_realpath(path):
290
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
293
def _win32_fixdrive(path):
294
"""Force drive letters to be consistent.
296
win32 is inconsistent whether it returns lower or upper case
297
and even if it was consistent the user might type the other
298
so we force it to uppercase
299
running python.exe under cmd.exe return capital C:\\
300
running win32 python inside a cygwin shell returns lowercase c:\\
302
drive, path = _nt_splitdrive(path)
303
return drive.upper() + path
306
def _win32_abspath(path):
307
# Real _nt_abspath doesn't have a problem with a unicode cwd
308
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
311
def _win98_abspath(path):
312
"""Return the absolute version of a path.
313
Windows 98 safe implementation (python reimplementation
314
of Win32 API function GetFullPathNameW)
319
# \\HOST\path => //HOST/path
320
# //HOST/path => //HOST/path
321
# path => C:/cwd/path
324
# check for absolute path
325
drive = _nt_splitdrive(path)[0]
326
if drive == '' and path[:2] not in('//','\\\\'):
328
# we cannot simply os.path.join cwd and path
329
# because os.path.join('C:','/path') produce '/path'
330
# and this is incorrect
331
if path[:1] in ('/','\\'):
332
cwd = _nt_splitdrive(cwd)[0]
334
path = cwd + '\\' + path
335
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
338
def _win32_realpath(path):
339
# Real _nt_realpath doesn't have a problem with a unicode cwd
340
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
343
def _win32_pathjoin(*args):
344
return _nt_join(*args).replace('\\', '/')
347
def _win32_normpath(path):
348
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
352
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
355
def _win32_mkdtemp(*args, **kwargs):
356
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
359
def _win32_rename(old, new):
360
"""We expect to be able to atomically replace 'new' with old.
362
On win32, if new exists, it must be moved out of the way first,
366
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
368
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
369
# If we try to rename a non-existant file onto cwd, we get
370
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
371
# if the old path doesn't exist, sometimes we get EACCES
372
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
378
return unicodedata.normalize('NFC', os.getcwdu())
381
# Default is to just use the python builtins, but these can be rebound on
382
# particular platforms.
383
abspath = _posix_abspath
384
realpath = _posix_realpath
385
pathjoin = os.path.join
386
normpath = os.path.normpath
389
dirname = os.path.dirname
390
basename = os.path.basename
391
split = os.path.split
392
splitext = os.path.splitext
393
# These were already imported into local scope
394
# mkdtemp = tempfile.mkdtemp
395
# rmtree = shutil.rmtree
397
MIN_ABS_PATHLENGTH = 1
400
if sys.platform == 'win32':
401
if win32utils.winver == 'Windows 98':
402
abspath = _win98_abspath
404
abspath = _win32_abspath
405
realpath = _win32_realpath
406
pathjoin = _win32_pathjoin
407
normpath = _win32_normpath
408
getcwd = _win32_getcwd
409
mkdtemp = _win32_mkdtemp
410
rename = _win32_rename
412
MIN_ABS_PATHLENGTH = 3
414
def _win32_delete_readonly(function, path, excinfo):
415
"""Error handler for shutil.rmtree function [for win32]
416
Helps to remove files and dirs marked as read-only.
418
exception = excinfo[1]
419
if function in (os.remove, os.rmdir) \
420
and isinstance(exception, OSError) \
421
and exception.errno == errno.EACCES:
427
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
428
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
429
return shutil.rmtree(path, ignore_errors, onerror)
431
f = win32utils.get_unicode_argv # special function or None
435
elif sys.platform == 'darwin':
439
def get_terminal_encoding():
440
"""Find the best encoding for printing to the screen.
442
This attempts to check both sys.stdout and sys.stdin to see
443
what encoding they are in, and if that fails it falls back to
444
osutils.get_user_encoding().
445
The problem is that on Windows, locale.getpreferredencoding()
446
is not the same encoding as that used by the console:
447
http://mail.python.org/pipermail/python-list/2003-May/162357.html
449
On my standard US Windows XP, the preferred encoding is
450
cp1252, but the console is cp437
452
from bzrlib.trace import mutter
453
output_encoding = getattr(sys.stdout, 'encoding', None)
454
if not output_encoding:
455
input_encoding = getattr(sys.stdin, 'encoding', None)
456
if not input_encoding:
457
output_encoding = get_user_encoding()
458
mutter('encoding stdout as osutils.get_user_encoding() %r',
461
output_encoding = input_encoding
462
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
464
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
465
if output_encoding == 'cp0':
466
# invalid encoding (cp0 means 'no codepage' on Windows)
467
output_encoding = get_user_encoding()
468
mutter('cp0 is invalid encoding.'
469
' encoding stdout as osutils.get_user_encoding() %r',
473
codecs.lookup(output_encoding)
475
sys.stderr.write('bzr: warning:'
476
' unknown terminal encoding %s.\n'
477
' Using encoding %s instead.\n'
478
% (output_encoding, get_user_encoding())
480
output_encoding = get_user_encoding()
482
return output_encoding
485
def normalizepath(f):
486
if getattr(os.path, 'realpath', None) is not None:
490
[p,e] = os.path.split(f)
491
if e == "" or e == "." or e == "..":
494
return pathjoin(F(p), e)
498
"""True if f is an accessible directory."""
500
return S_ISDIR(os.lstat(f)[ST_MODE])
506
"""True if f is a regular file."""
508
return S_ISREG(os.lstat(f)[ST_MODE])
513
"""True if f is a symlink."""
515
return S_ISLNK(os.lstat(f)[ST_MODE])
519
def is_inside(dir, fname):
520
"""True if fname is inside dir.
522
The parameters should typically be passed to osutils.normpath first, so
523
that . and .. and repeated slashes are eliminated, and the separators
524
are canonical for the platform.
526
The empty string as a dir name is taken as top-of-tree and matches
529
# XXX: Most callers of this can actually do something smarter by
530
# looking at the inventory
540
return fname.startswith(dir)
543
def is_inside_any(dir_list, fname):
544
"""True if fname is inside any of given dirs."""
545
for dirname in dir_list:
546
if is_inside(dirname, fname):
551
def is_inside_or_parent_of_any(dir_list, fname):
552
"""True if fname is a child or a parent of any of the given files."""
553
for dirname in dir_list:
554
if is_inside(dirname, fname) or is_inside(fname, dirname):
559
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
560
report_activity=None, direction='read'):
561
"""Copy contents of one file to another.
563
The read_length can either be -1 to read to end-of-file (EOF) or
564
it can specify the maximum number of bytes to read.
566
The buff_size represents the maximum size for each read operation
567
performed on from_file.
569
:param report_activity: Call this as bytes are read, see
570
Transport._report_activity
571
:param direction: Will be passed to report_activity
573
:return: The number of bytes copied.
577
# read specified number of bytes
579
while read_length > 0:
580
num_bytes_to_read = min(read_length, buff_size)
582
block = from_file.read(num_bytes_to_read)
586
if report_activity is not None:
587
report_activity(len(block), direction)
590
actual_bytes_read = len(block)
591
read_length -= actual_bytes_read
592
length += actual_bytes_read
596
block = from_file.read(buff_size)
600
if report_activity is not None:
601
report_activity(len(block), direction)
607
def pump_string_file(bytes, file_handle, segment_size=None):
608
"""Write bytes to file_handle in many smaller writes.
610
:param bytes: The string to write.
611
:param file_handle: The file to write to.
613
# Write data in chunks rather than all at once, because very large
614
# writes fail on some platforms (e.g. Windows with SMB mounted
617
segment_size = 5242880 # 5MB
618
segments = range(len(bytes) / segment_size + 1)
619
write = file_handle.write
620
for segment_index in segments:
621
segment = buffer(bytes, segment_index * segment_size, segment_size)
625
def file_iterator(input_file, readsize=32768):
627
b = input_file.read(readsize)
634
"""Calculate the hexdigest of an open file.
636
The file cursor should be already at the start.
648
def size_sha_file(f):
649
"""Calculate the size and hexdigest of an open file.
651
The file cursor should be already at the start and
652
the caller is responsible for closing the file afterwards.
663
return size, s.hexdigest()
666
def sha_file_by_name(fname):
667
"""Calculate the SHA1 of a file by reading the full text"""
669
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
672
b = os.read(f, 1<<16)
680
def sha_strings(strings, _factory=sha):
681
"""Return the sha-1 of concatenation of strings"""
683
map(s.update, strings)
687
def sha_string(f, _factory=sha):
688
return _factory(f).hexdigest()
691
def fingerprint_file(f):
693
return {'size': len(b),
694
'sha1': sha(b).hexdigest()}
697
def compare_files(a, b):
698
"""Returns true if equal in contents"""
709
def local_time_offset(t=None):
710
"""Return offset of local zone from GMT, either at present or at time t."""
713
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
714
return offset.days * 86400 + offset.seconds
716
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
717
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
720
def format_date(t, offset=0, timezone='original', date_fmt=None,
722
"""Return a formatted date string.
724
:param t: Seconds since the epoch.
725
:param offset: Timezone offset in seconds east of utc.
726
:param timezone: How to display the time: 'utc', 'original' for the
727
timezone specified by offset, or 'local' for the process's current
729
:param date_fmt: strftime format.
730
:param show_offset: Whether to append the timezone.
732
(date_fmt, tt, offset_str) = \
733
_format_date(t, offset, timezone, date_fmt, show_offset)
734
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
735
date_str = time.strftime(date_fmt, tt)
736
return date_str + offset_str
739
# Cache of formatted offset strings
743
def format_date_with_offset_in_original_timezone(t, offset=0,
744
_cache=_offset_cache):
745
"""Return a formatted date string in the original timezone.
747
This routine may be faster then format_date.
749
:param t: Seconds since the epoch.
750
:param offset: Timezone offset in seconds east of utc.
754
tt = time.gmtime(t + offset)
755
date_fmt = _default_format_by_weekday_num[tt[6]]
756
date_str = time.strftime(date_fmt, tt)
757
offset_str = _cache.get(offset, None)
758
if offset_str is None:
759
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
760
_cache[offset] = offset_str
761
return date_str + offset_str
764
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
766
"""Return an unicode date string formatted according to the current locale.
768
:param t: Seconds since the epoch.
769
:param offset: Timezone offset in seconds east of utc.
770
:param timezone: How to display the time: 'utc', 'original' for the
771
timezone specified by offset, or 'local' for the process's current
773
:param date_fmt: strftime format.
774
:param show_offset: Whether to append the timezone.
776
(date_fmt, tt, offset_str) = \
777
_format_date(t, offset, timezone, date_fmt, show_offset)
778
date_str = time.strftime(date_fmt, tt)
779
if not isinstance(date_str, unicode):
780
date_str = date_str.decode(get_user_encoding(), 'replace')
781
return date_str + offset_str
784
def _format_date(t, offset, timezone, date_fmt, show_offset):
785
if timezone == 'utc':
788
elif timezone == 'original':
791
tt = time.gmtime(t + offset)
792
elif timezone == 'local':
793
tt = time.localtime(t)
794
offset = local_time_offset(t)
796
raise errors.UnsupportedTimezoneFormat(timezone)
798
date_fmt = "%a %Y-%m-%d %H:%M:%S"
800
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
803
return (date_fmt, tt, offset_str)
806
def compact_date(when):
807
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
810
def format_delta(delta):
811
"""Get a nice looking string for a time delta.
813
:param delta: The time difference in seconds, can be positive or negative.
814
positive indicates time in the past, negative indicates time in the
815
future. (usually time.time() - stored_time)
816
:return: String formatted to show approximate resolution
822
direction = 'in the future'
826
if seconds < 90: # print seconds up to 90 seconds
828
return '%d second %s' % (seconds, direction,)
830
return '%d seconds %s' % (seconds, direction)
832
minutes = int(seconds / 60)
833
seconds -= 60 * minutes
838
if minutes < 90: # print minutes, seconds up to 90 minutes
840
return '%d minute, %d second%s %s' % (
841
minutes, seconds, plural_seconds, direction)
843
return '%d minutes, %d second%s %s' % (
844
minutes, seconds, plural_seconds, direction)
846
hours = int(minutes / 60)
847
minutes -= 60 * hours
854
return '%d hour, %d minute%s %s' % (hours, minutes,
855
plural_minutes, direction)
856
return '%d hours, %d minute%s %s' % (hours, minutes,
857
plural_minutes, direction)
860
"""Return size of given open file."""
861
return os.fstat(f.fileno())[ST_SIZE]
864
# Define rand_bytes based on platform.
866
# Python 2.4 and later have os.urandom,
867
# but it doesn't work on some arches
869
rand_bytes = os.urandom
870
except (NotImplementedError, AttributeError):
871
# If python doesn't have os.urandom, or it doesn't work,
872
# then try to first pull random data from /dev/urandom
874
rand_bytes = file('/dev/urandom', 'rb').read
875
# Otherwise, use this hack as a last resort
876
except (IOError, OSError):
877
# not well seeded, but better than nothing
882
s += chr(random.randint(0, 255))
887
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
889
"""Return a random string of num alphanumeric characters
891
The result only contains lowercase chars because it may be used on
892
case-insensitive filesystems.
895
for raw_byte in rand_bytes(num):
896
s += ALNUM[ord(raw_byte) % 36]
900
## TODO: We could later have path objects that remember their list
901
## decomposition (might be too tricksy though.)
904
"""Turn string into list of parts."""
905
# split on either delimiter because people might use either on
907
ps = re.split(r'[\\/]', p)
912
raise errors.BzrError("sorry, %r not allowed in path" % f)
913
elif (f == '.') or (f == ''):
922
if (f == '..') or (f is None) or (f == ''):
923
raise errors.BzrError("sorry, %r not allowed in path" % f)
927
def parent_directories(filename):
928
"""Return the list of parent directories, deepest first.
930
For example, parent_directories("a/b/c") -> ["a/b", "a"].
933
parts = splitpath(dirname(filename))
935
parents.append(joinpath(parts))
940
_extension_load_failures = []
943
def failed_to_load_extension(exception):
944
"""Handle failing to load a binary extension.
946
This should be called from the ImportError block guarding the attempt to
947
import the native extension. If this function returns, the pure-Python
948
implementation should be loaded instead::
951
>>> import bzrlib._fictional_extension_pyx
952
>>> except ImportError, e:
953
>>> bzrlib.osutils.failed_to_load_extension(e)
954
>>> import bzrlib._fictional_extension_py
956
# NB: This docstring is just an example, not a doctest, because doctest
957
# currently can't cope with the use of lazy imports in this namespace --
960
# This currently doesn't report the failure at the time it occurs, because
961
# they tend to happen very early in startup when we can't check config
962
# files etc, and also we want to report all failures but not spam the user
964
from bzrlib import trace
965
exception_str = str(exception)
966
if exception_str not in _extension_load_failures:
967
trace.mutter("failed to load compiled extension: %s" % exception_str)
968
_extension_load_failures.append(exception_str)
971
def report_extension_load_failures():
972
if not _extension_load_failures:
974
from bzrlib.config import GlobalConfig
975
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
977
# the warnings framework should by default show this only once
978
from bzrlib.trace import warning
980
"bzr: warning: some compiled extensions could not be loaded; "
981
"see <https://answers.launchpad.net/bzr/+faq/703>")
982
# we no longer show the specific missing extensions here, because it makes
983
# the message too long and scary - see
984
# https://bugs.launchpad.net/bzr/+bug/430529
988
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
989
except ImportError, e:
990
failed_to_load_extension(e)
991
from bzrlib._chunks_to_lines_py import chunks_to_lines
995
"""Split s into lines, but without removing the newline characters."""
996
# Trivially convert a fulltext into a 'chunked' representation, and let
997
# chunks_to_lines do the heavy lifting.
998
if isinstance(s, str):
999
# chunks_to_lines only supports 8-bit strings
1000
return chunks_to_lines([s])
1002
return _split_lines(s)
1005
def _split_lines(s):
1006
"""Split s into lines, but without removing the newline characters.
1008
This supports Unicode or plain string objects.
1010
lines = s.split('\n')
1011
result = [line + '\n' for line in lines[:-1]]
1013
result.append(lines[-1])
1017
def hardlinks_good():
1018
return sys.platform not in ('win32', 'cygwin', 'darwin')
1021
def link_or_copy(src, dest):
1022
"""Hardlink a file, or copy it if it can't be hardlinked."""
1023
if not hardlinks_good():
1024
shutil.copyfile(src, dest)
1028
except (OSError, IOError), e:
1029
if e.errno != errno.EXDEV:
1031
shutil.copyfile(src, dest)
1034
def delete_any(path):
1035
"""Delete a file, symlink or directory.
1037
Will delete even if readonly.
1040
_delete_file_or_dir(path)
1041
except (OSError, IOError), e:
1042
if e.errno in (errno.EPERM, errno.EACCES):
1043
# make writable and try again
1046
except (OSError, IOError):
1048
_delete_file_or_dir(path)
1053
def _delete_file_or_dir(path):
1054
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1055
# Forgiveness than Permission (EAFP) because:
1056
# - root can damage a solaris file system by using unlink,
1057
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1058
# EACCES, OSX: EPERM) when invoked on a directory.
1059
if isdir(path): # Takes care of symlinks
1066
if getattr(os, 'symlink', None) is not None:
1072
def has_hardlinks():
1073
if getattr(os, 'link', None) is not None:
1079
def host_os_dereferences_symlinks():
1080
return (has_symlinks()
1081
and sys.platform not in ('cygwin', 'win32'))
1084
def readlink(abspath):
1085
"""Return a string representing the path to which the symbolic link points.
1087
:param abspath: The link absolute unicode path.
1089
This his guaranteed to return the symbolic link in unicode in all python
1092
link = abspath.encode(_fs_enc)
1093
target = os.readlink(link)
1094
target = target.decode(_fs_enc)
1098
def contains_whitespace(s):
1099
"""True if there are any whitespace characters in s."""
1100
# string.whitespace can include '\xa0' in certain locales, because it is
1101
# considered "non-breaking-space" as part of ISO-8859-1. But it
1102
# 1) Isn't a breaking whitespace
1103
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1105
# 3) '\xa0' isn't unicode safe since it is >128.
1107
# This should *not* be a unicode set of characters in case the source
1108
# string is not a Unicode string. We can auto-up-cast the characters since
1109
# they are ascii, but we don't want to auto-up-cast the string in case it
1111
for ch in ' \t\n\r\v\f':
1118
def contains_linebreaks(s):
1119
"""True if there is any vertical whitespace in s."""
1127
def relpath(base, path):
1128
"""Return path relative to base, or raise exception.
1130
The path may be either an absolute path or a path relative to the
1131
current working directory.
1133
os.path.commonprefix (python2.4) has a bad bug that it works just
1134
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1135
avoids that problem.
1138
if len(base) < MIN_ABS_PATHLENGTH:
1139
# must have space for e.g. a drive letter
1140
raise ValueError('%r is too short to calculate a relative path'
1148
if len(head) <= len(base) and head != base:
1149
raise errors.PathNotChild(rp, base)
1152
head, tail = split(head)
1157
return pathjoin(*reversed(s))
1162
def _cicp_canonical_relpath(base, path):
1163
"""Return the canonical path relative to base.
1165
Like relpath, but on case-insensitive-case-preserving file-systems, this
1166
will return the relpath as stored on the file-system rather than in the
1167
case specified in the input string, for all existing portions of the path.
1169
This will cause O(N) behaviour if called for every path in a tree; if you
1170
have a number of paths to convert, you should use canonical_relpaths().
1172
# TODO: it should be possible to optimize this for Windows by using the
1173
# win32 API FindFiles function to look for the specified name - but using
1174
# os.listdir() still gives us the correct, platform agnostic semantics in
1177
rel = relpath(base, path)
1178
# '.' will have been turned into ''
1182
abs_base = abspath(base)
1184
_listdir = os.listdir
1186
# use an explicit iterator so we can easily consume the rest on early exit.
1187
bit_iter = iter(rel.split('/'))
1188
for bit in bit_iter:
1191
next_entries = _listdir(current)
1192
except OSError: # enoent, eperm, etc
1193
# We can't find this in the filesystem, so just append the
1195
current = pathjoin(current, bit, *list(bit_iter))
1197
for look in next_entries:
1198
if lbit == look.lower():
1199
current = pathjoin(current, look)
1202
# got to the end, nothing matched, so we just return the
1203
# non-existing bits as they were specified (the filename may be
1204
# the target of a move, for example).
1205
current = pathjoin(current, bit, *list(bit_iter))
1207
return current[len(abs_base):].lstrip('/')
1209
# XXX - TODO - we need better detection/integration of case-insensitive
1210
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1211
# filesystems), for example, so could probably benefit from the same basic
1212
# support there. For now though, only Windows and OSX get that support, and
1213
# they get it for *all* file-systems!
1214
if sys.platform in ('win32', 'darwin'):
1215
canonical_relpath = _cicp_canonical_relpath
1217
canonical_relpath = relpath
1219
def canonical_relpaths(base, paths):
1220
"""Create an iterable to canonicalize a sequence of relative paths.
1222
The intent is for this implementation to use a cache, vastly speeding
1223
up multiple transformations in the same directory.
1225
# but for now, we haven't optimized...
1226
return [canonical_relpath(base, p) for p in paths]
1228
def safe_unicode(unicode_or_utf8_string):
1229
"""Coerce unicode_or_utf8_string into unicode.
1231
If it is unicode, it is returned.
1232
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1233
wrapped in a BzrBadParameterNotUnicode exception.
1235
if isinstance(unicode_or_utf8_string, unicode):
1236
return unicode_or_utf8_string
1238
return unicode_or_utf8_string.decode('utf8')
1239
except UnicodeDecodeError:
1240
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1243
def safe_utf8(unicode_or_utf8_string):
1244
"""Coerce unicode_or_utf8_string to a utf8 string.
1246
If it is a str, it is returned.
1247
If it is Unicode, it is encoded into a utf-8 string.
1249
if isinstance(unicode_or_utf8_string, str):
1250
# TODO: jam 20070209 This is overkill, and probably has an impact on
1251
# performance if we are dealing with lots of apis that want a
1254
# Make sure it is a valid utf-8 string
1255
unicode_or_utf8_string.decode('utf-8')
1256
except UnicodeDecodeError:
1257
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1258
return unicode_or_utf8_string
1259
return unicode_or_utf8_string.encode('utf-8')
1262
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1263
' Revision id generators should be creating utf8'
1267
def safe_revision_id(unicode_or_utf8_string, warn=True):
1268
"""Revision ids should now be utf8, but at one point they were unicode.
1270
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1272
:param warn: Functions that are sanitizing user data can set warn=False
1273
:return: None or a utf8 revision id.
1275
if (unicode_or_utf8_string is None
1276
or unicode_or_utf8_string.__class__ == str):
1277
return unicode_or_utf8_string
1279
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1281
return cache_utf8.encode(unicode_or_utf8_string)
1284
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1285
' generators should be creating utf8 file ids.')
1288
def safe_file_id(unicode_or_utf8_string, warn=True):
1289
"""File ids should now be utf8, but at one point they were unicode.
1291
This is the same as safe_utf8, except it uses the cached encode functions
1292
to save a little bit of performance.
1294
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1296
:param warn: Functions that are sanitizing user data can set warn=False
1297
:return: None or a utf8 file id.
1299
if (unicode_or_utf8_string is None
1300
or unicode_or_utf8_string.__class__ == str):
1301
return unicode_or_utf8_string
1303
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1305
return cache_utf8.encode(unicode_or_utf8_string)
1308
_platform_normalizes_filenames = False
1309
if sys.platform == 'darwin':
1310
_platform_normalizes_filenames = True
1313
def normalizes_filenames():
1314
"""Return True if this platform normalizes unicode filenames.
1316
Mac OSX does, Windows/Linux do not.
1318
return _platform_normalizes_filenames
1321
def _accessible_normalized_filename(path):
1322
"""Get the unicode normalized path, and if you can access the file.
1324
On platforms where the system normalizes filenames (Mac OSX),
1325
you can access a file by any path which will normalize correctly.
1326
On platforms where the system does not normalize filenames
1327
(Windows, Linux), you have to access a file by its exact path.
1329
Internally, bzr only supports NFC normalization, since that is
1330
the standard for XML documents.
1332
So return the normalized path, and a flag indicating if the file
1333
can be accessed by that path.
1336
return unicodedata.normalize('NFC', unicode(path)), True
1339
def _inaccessible_normalized_filename(path):
1340
__doc__ = _accessible_normalized_filename.__doc__
1342
normalized = unicodedata.normalize('NFC', unicode(path))
1343
return normalized, normalized == path
1346
if _platform_normalizes_filenames:
1347
normalized_filename = _accessible_normalized_filename
1349
normalized_filename = _inaccessible_normalized_filename
1352
def set_signal_handler(signum, handler, restart_syscall=True):
1353
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1354
on platforms that support that.
1356
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1357
automatically restart (by calling `signal.siginterrupt(signum,
1358
False)`). May be ignored if the feature is not available on this
1359
platform or Python version.
1361
old_handler = signal.signal(signum, handler)
1364
siginterrupt = signal.siginterrupt
1365
except AttributeError: # siginterrupt doesn't exist on this platform, or for this version of
1369
siginterrupt(signum, False)
1373
default_terminal_width = 80
1374
"""The default terminal width for ttys.
1376
This is defined so that higher levels can share a common fallback value when
1377
terminal_width() returns None.
1381
def terminal_width():
1382
"""Return terminal width.
1384
None is returned if the width can't established precisely.
1387
- if BZR_COLUMNS is set, returns its value
1388
- if there is no controlling terminal, returns None
1389
- if COLUMNS is set, returns its value,
1391
From there, we need to query the OS to get the size of the controlling
1395
- get termios.TIOCGWINSZ
1396
- if an error occurs or a negative value is obtained, returns None
1400
- win32utils.get_console_size() decides,
1401
- returns None on error (provided default value)
1404
# If BZR_COLUMNS is set, take it, user is always right
1406
return int(os.environ['BZR_COLUMNS'])
1407
except (KeyError, ValueError):
1410
isatty = getattr(sys.stdout, 'isatty', None)
1411
if isatty is None or not isatty():
1412
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1415
# If COLUMNS is set, take it, the terminal knows better (even inside a
1416
# given terminal, the application can decide to set COLUMNS to a lower
1417
# value (splitted screen) or a bigger value (scroll bars))
1419
return int(os.environ['COLUMNS'])
1420
except (KeyError, ValueError):
1423
width, height = _terminal_size(None, None)
1425
# Consider invalid values as meaning no width
1431
def _win32_terminal_size(width, height):
1432
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1433
return width, height
1436
def _ioctl_terminal_size(width, height):
1438
import struct, fcntl, termios
1439
s = struct.pack('HHHH', 0, 0, 0, 0)
1440
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1441
height, width = struct.unpack('HHHH', x)[0:2]
1442
except (IOError, AttributeError):
1444
return width, height
1446
_terminal_size = None
1447
"""Returns the terminal size as (width, height).
1449
:param width: Default value for width.
1450
:param height: Default value for height.
1452
This is defined specifically for each OS and query the size of the controlling
1453
terminal. If any error occurs, the provided default values should be returned.
1455
if sys.platform == 'win32':
1456
_terminal_size = _win32_terminal_size
1458
_terminal_size = _ioctl_terminal_size
1461
def _terminal_size_changed(signum, frame):
1462
"""Set COLUMNS upon receiving a SIGnal for WINdow size CHange."""
1463
width, height = _terminal_size(None, None)
1464
if width is not None:
1465
os.environ['COLUMNS'] = str(width)
1468
_registered_sigwinch = False
1470
def watch_sigwinch():
1471
"""Register for SIGWINCH, once and only once."""
1472
global _registered_sigwinch
1473
if not _registered_sigwinch:
1474
if sys.platform == 'win32':
1475
# Martin (gz) mentioned WINDOW_BUFFER_SIZE_RECORD from
1476
# ReadConsoleInput but I've no idea how to plug that in
1477
# the current design -- vila 20091216
1480
set_signal_handler(signal.SIGWINCH, _terminal_size_changed)
1481
_registered_sigwinch = True
1484
def supports_executable():
1485
return sys.platform != "win32"
1488
def supports_posix_readonly():
1489
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1491
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1492
directory controls creation/deletion, etc.
1494
And under win32, readonly means that the directory itself cannot be
1495
deleted. The contents of a readonly directory can be changed, unlike POSIX
1496
where files in readonly directories cannot be added, deleted or renamed.
1498
return sys.platform != "win32"
1501
def set_or_unset_env(env_variable, value):
1502
"""Modify the environment, setting or removing the env_variable.
1504
:param env_variable: The environment variable in question
1505
:param value: The value to set the environment to. If None, then
1506
the variable will be removed.
1507
:return: The original value of the environment variable.
1509
orig_val = os.environ.get(env_variable)
1511
if orig_val is not None:
1512
del os.environ[env_variable]
1514
if isinstance(value, unicode):
1515
value = value.encode(get_user_encoding())
1516
os.environ[env_variable] = value
1520
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1523
def check_legal_path(path):
1524
"""Check whether the supplied path is legal.
1525
This is only required on Windows, so we don't test on other platforms
1528
if sys.platform != "win32":
1530
if _validWin32PathRE.match(path) is None:
1531
raise errors.IllegalPath(path)
1534
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1536
def _is_error_enotdir(e):
1537
"""Check if this exception represents ENOTDIR.
1539
Unfortunately, python is very inconsistent about the exception
1540
here. The cases are:
1541
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1542
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1543
which is the windows error code.
1544
3) Windows, Python2.5 uses errno == EINVAL and
1545
winerror == ERROR_DIRECTORY
1547
:param e: An Exception object (expected to be OSError with an errno
1548
attribute, but we should be able to cope with anything)
1549
:return: True if this represents an ENOTDIR error. False otherwise.
1551
en = getattr(e, 'errno', None)
1552
if (en == errno.ENOTDIR
1553
or (sys.platform == 'win32'
1554
and (en == _WIN32_ERROR_DIRECTORY
1555
or (en == errno.EINVAL
1556
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1562
def walkdirs(top, prefix=""):
1563
"""Yield data about all the directories in a tree.
1565
This yields all the data about the contents of a directory at a time.
1566
After each directory has been yielded, if the caller has mutated the list
1567
to exclude some directories, they are then not descended into.
1569
The data yielded is of the form:
1570
((directory-relpath, directory-path-from-top),
1571
[(relpath, basename, kind, lstat, path-from-top), ...]),
1572
- directory-relpath is the relative path of the directory being returned
1573
with respect to top. prefix is prepended to this.
1574
- directory-path-from-root is the path including top for this directory.
1575
It is suitable for use with os functions.
1576
- relpath is the relative path within the subtree being walked.
1577
- basename is the basename of the path
1578
- kind is the kind of the file now. If unknown then the file is not
1579
present within the tree - but it may be recorded as versioned. See
1581
- lstat is the stat data *if* the file was statted.
1582
- planned, not implemented:
1583
path_from_tree_root is the path from the root of the tree.
1585
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1586
allows one to walk a subtree but get paths that are relative to a tree
1588
:return: an iterator over the dirs.
1590
#TODO there is a bit of a smell where the results of the directory-
1591
# summary in this, and the path from the root, may not agree
1592
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1593
# potentially confusing output. We should make this more robust - but
1594
# not at a speed cost. RBC 20060731
1596
_directory = _directory_kind
1597
_listdir = os.listdir
1598
_kind_from_mode = file_kind_from_stat_mode
1599
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1601
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1602
relroot, _, _, _, top = pending.pop()
1604
relprefix = relroot + u'/'
1607
top_slash = top + u'/'
1610
append = dirblock.append
1612
names = sorted(_listdir(top))
1614
if not _is_error_enotdir(e):
1618
abspath = top_slash + name
1619
statvalue = _lstat(abspath)
1620
kind = _kind_from_mode(statvalue.st_mode)
1621
append((relprefix + name, name, kind, statvalue, abspath))
1622
yield (relroot, top), dirblock
1624
# push the user specified dirs from dirblock
1625
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1628
class DirReader(object):
1629
"""An interface for reading directories."""
1631
def top_prefix_to_starting_dir(self, top, prefix=""):
1632
"""Converts top and prefix to a starting dir entry
1634
:param top: A utf8 path
1635
:param prefix: An optional utf8 path to prefix output relative paths
1637
:return: A tuple starting with prefix, and ending with the native
1640
raise NotImplementedError(self.top_prefix_to_starting_dir)
1642
def read_dir(self, prefix, top):
1643
"""Read a specific dir.
1645
:param prefix: A utf8 prefix to be preprended to the path basenames.
1646
:param top: A natively encoded path to read.
1647
:return: A list of the directories contents. Each item contains:
1648
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1650
raise NotImplementedError(self.read_dir)
1653
_selected_dir_reader = None
1656
def _walkdirs_utf8(top, prefix=""):
1657
"""Yield data about all the directories in a tree.
1659
This yields the same information as walkdirs() only each entry is yielded
1660
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1661
are returned as exact byte-strings.
1663
:return: yields a tuple of (dir_info, [file_info])
1664
dir_info is (utf8_relpath, path-from-top)
1665
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1666
if top is an absolute path, path-from-top is also an absolute path.
1667
path-from-top might be unicode or utf8, but it is the correct path to
1668
pass to os functions to affect the file in question. (such as os.lstat)
1670
global _selected_dir_reader
1671
if _selected_dir_reader is None:
1672
fs_encoding = _fs_enc.upper()
1673
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1674
# Win98 doesn't have unicode apis like FindFirstFileW
1675
# TODO: We possibly could support Win98 by falling back to the
1676
# original FindFirstFile, and using TCHAR instead of WCHAR,
1677
# but that gets a bit tricky, and requires custom compiling
1680
from bzrlib._walkdirs_win32 import Win32ReadDir
1681
_selected_dir_reader = Win32ReadDir()
1684
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1685
# ANSI_X3.4-1968 is a form of ASCII
1687
from bzrlib._readdir_pyx import UTF8DirReader
1688
_selected_dir_reader = UTF8DirReader()
1689
except ImportError, e:
1690
failed_to_load_extension(e)
1693
if _selected_dir_reader is None:
1694
# Fallback to the python version
1695
_selected_dir_reader = UnicodeDirReader()
1697
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1698
# But we don't actually uses 1-3 in pending, so set them to None
1699
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1700
read_dir = _selected_dir_reader.read_dir
1701
_directory = _directory_kind
1703
relroot, _, _, _, top = pending[-1].pop()
1706
dirblock = sorted(read_dir(relroot, top))
1707
yield (relroot, top), dirblock
1708
# push the user specified dirs from dirblock
1709
next = [d for d in reversed(dirblock) if d[2] == _directory]
1711
pending.append(next)
1714
class UnicodeDirReader(DirReader):
1715
"""A dir reader for non-utf8 file systems, which transcodes."""
1717
__slots__ = ['_utf8_encode']
1720
self._utf8_encode = codecs.getencoder('utf8')
1722
def top_prefix_to_starting_dir(self, top, prefix=""):
1723
"""See DirReader.top_prefix_to_starting_dir."""
1724
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1726
def read_dir(self, prefix, top):
1727
"""Read a single directory from a non-utf8 file system.
1729
top, and the abspath element in the output are unicode, all other paths
1730
are utf8. Local disk IO is done via unicode calls to listdir etc.
1732
This is currently the fallback code path when the filesystem encoding is
1733
not UTF-8. It may be better to implement an alternative so that we can
1734
safely handle paths that are not properly decodable in the current
1737
See DirReader.read_dir for details.
1739
_utf8_encode = self._utf8_encode
1741
_listdir = os.listdir
1742
_kind_from_mode = file_kind_from_stat_mode
1745
relprefix = prefix + '/'
1748
top_slash = top + u'/'
1751
append = dirblock.append
1752
for name in sorted(_listdir(top)):
1754
name_utf8 = _utf8_encode(name)[0]
1755
except UnicodeDecodeError:
1756
raise errors.BadFilenameEncoding(
1757
_utf8_encode(relprefix)[0] + name, _fs_enc)
1758
abspath = top_slash + name
1759
statvalue = _lstat(abspath)
1760
kind = _kind_from_mode(statvalue.st_mode)
1761
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1765
def copy_tree(from_path, to_path, handlers={}):
1766
"""Copy all of the entries in from_path into to_path.
1768
:param from_path: The base directory to copy.
1769
:param to_path: The target directory. If it does not exist, it will
1771
:param handlers: A dictionary of functions, which takes a source and
1772
destinations for files, directories, etc.
1773
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1774
'file', 'directory', and 'symlink' should always exist.
1775
If they are missing, they will be replaced with 'os.mkdir()',
1776
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1778
# Now, just copy the existing cached tree to the new location
1779
# We use a cheap trick here.
1780
# Absolute paths are prefixed with the first parameter
1781
# relative paths are prefixed with the second.
1782
# So we can get both the source and target returned
1783
# without any extra work.
1785
def copy_dir(source, dest):
1788
def copy_link(source, dest):
1789
"""Copy the contents of a symlink"""
1790
link_to = os.readlink(source)
1791
os.symlink(link_to, dest)
1793
real_handlers = {'file':shutil.copy2,
1794
'symlink':copy_link,
1795
'directory':copy_dir,
1797
real_handlers.update(handlers)
1799
if not os.path.exists(to_path):
1800
real_handlers['directory'](from_path, to_path)
1802
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1803
for relpath, name, kind, st, abspath in entries:
1804
real_handlers[kind](abspath, relpath)
1807
def path_prefix_key(path):
1808
"""Generate a prefix-order path key for path.
1810
This can be used to sort paths in the same way that walkdirs does.
1812
return (dirname(path) , path)
1815
def compare_paths_prefix_order(path_a, path_b):
1816
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1817
key_a = path_prefix_key(path_a)
1818
key_b = path_prefix_key(path_b)
1819
return cmp(key_a, key_b)
1822
_cached_user_encoding = None
1825
def get_user_encoding(use_cache=True):
1826
"""Find out what the preferred user encoding is.
1828
This is generally the encoding that is used for command line parameters
1829
and file contents. This may be different from the terminal encoding
1830
or the filesystem encoding.
1832
:param use_cache: Enable cache for detected encoding.
1833
(This parameter is turned on by default,
1834
and required only for selftesting)
1836
:return: A string defining the preferred user encoding
1838
global _cached_user_encoding
1839
if _cached_user_encoding is not None and use_cache:
1840
return _cached_user_encoding
1842
if sys.platform == 'darwin':
1843
# python locale.getpreferredencoding() always return
1844
# 'mac-roman' on darwin. That's a lie.
1845
sys.platform = 'posix'
1847
if os.environ.get('LANG', None) is None:
1848
# If LANG is not set, we end up with 'ascii', which is bad
1849
# ('mac-roman' is more than ascii), so we set a default which
1850
# will give us UTF-8 (which appears to work in all cases on
1851
# OSX). Users are still free to override LANG of course, as
1852
# long as it give us something meaningful. This work-around
1853
# *may* not be needed with python 3k and/or OSX 10.5, but will
1854
# work with them too -- vila 20080908
1855
os.environ['LANG'] = 'en_US.UTF-8'
1858
sys.platform = 'darwin'
1863
user_encoding = locale.getpreferredencoding()
1864
except locale.Error, e:
1865
sys.stderr.write('bzr: warning: %s\n'
1866
' Could not determine what text encoding to use.\n'
1867
' This error usually means your Python interpreter\n'
1868
' doesn\'t support the locale set by $LANG (%s)\n'
1869
" Continuing with ascii encoding.\n"
1870
% (e, os.environ.get('LANG')))
1871
user_encoding = 'ascii'
1873
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1874
# treat that as ASCII, and not support printing unicode characters to the
1877
# For python scripts run under vim, we get '', so also treat that as ASCII
1878
if user_encoding in (None, 'cp0', ''):
1879
user_encoding = 'ascii'
1883
codecs.lookup(user_encoding)
1885
sys.stderr.write('bzr: warning:'
1886
' unknown encoding %s.'
1887
' Continuing with ascii encoding.\n'
1890
user_encoding = 'ascii'
1893
_cached_user_encoding = user_encoding
1895
return user_encoding
1898
def get_host_name():
1899
"""Return the current unicode host name.
1901
This is meant to be used in place of socket.gethostname() because that
1902
behaves inconsistently on different platforms.
1904
if sys.platform == "win32":
1906
return win32utils.get_host_name()
1909
return socket.gethostname().decode(get_user_encoding())
1912
def recv_all(socket, bytes):
1913
"""Receive an exact number of bytes.
1915
Regular Socket.recv() may return less than the requested number of bytes,
1916
dependning on what's in the OS buffer. MSG_WAITALL is not available
1917
on all platforms, but this should work everywhere. This will return
1918
less than the requested amount if the remote end closes.
1920
This isn't optimized and is intended mostly for use in testing.
1923
while len(b) < bytes:
1924
new = until_no_eintr(socket.recv, bytes - len(b))
1931
def send_all(socket, bytes, report_activity=None):
1932
"""Send all bytes on a socket.
1934
Regular socket.sendall() can give socket error 10053 on Windows. This
1935
implementation sends no more than 64k at a time, which avoids this problem.
1937
:param report_activity: Call this as bytes are read, see
1938
Transport._report_activity
1941
for pos in xrange(0, len(bytes), chunk_size):
1942
block = bytes[pos:pos+chunk_size]
1943
if report_activity is not None:
1944
report_activity(len(block), 'write')
1945
until_no_eintr(socket.sendall, block)
1948
def dereference_path(path):
1949
"""Determine the real path to a file.
1951
All parent elements are dereferenced. But the file itself is not
1953
:param path: The original path. May be absolute or relative.
1954
:return: the real path *to* the file
1956
parent, base = os.path.split(path)
1957
# The pathjoin for '.' is a workaround for Python bug #1213894.
1958
# (initial path components aren't dereferenced)
1959
return pathjoin(realpath(pathjoin('.', parent)), base)
1962
def supports_mapi():
1963
"""Return True if we can use MAPI to launch a mail client."""
1964
return sys.platform == "win32"
1967
def resource_string(package, resource_name):
1968
"""Load a resource from a package and return it as a string.
1970
Note: Only packages that start with bzrlib are currently supported.
1972
This is designed to be a lightweight implementation of resource
1973
loading in a way which is API compatible with the same API from
1975
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1976
If and when pkg_resources becomes a standard library, this routine
1979
# Check package name is within bzrlib
1980
if package == "bzrlib":
1981
resource_relpath = resource_name
1982
elif package.startswith("bzrlib."):
1983
package = package[len("bzrlib."):].replace('.', os.sep)
1984
resource_relpath = pathjoin(package, resource_name)
1986
raise errors.BzrError('resource package %s not in bzrlib' % package)
1988
# Map the resource to a file and read its contents
1989
base = dirname(bzrlib.__file__)
1990
if getattr(sys, 'frozen', None): # bzr.exe
1991
base = abspath(pathjoin(base, '..', '..'))
1992
filename = pathjoin(base, resource_relpath)
1993
return open(filename, 'rU').read()
1996
def file_kind_from_stat_mode_thunk(mode):
1997
global file_kind_from_stat_mode
1998
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2000
from bzrlib._readdir_pyx import UTF8DirReader
2001
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2002
except ImportError, e:
2003
# This is one time where we won't warn that an extension failed to
2004
# load. The extension is never available on Windows anyway.
2005
from bzrlib._readdir_py import (
2006
_kind_from_mode as file_kind_from_stat_mode
2008
return file_kind_from_stat_mode(mode)
2009
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2012
def file_kind(f, _lstat=os.lstat):
2014
return file_kind_from_stat_mode(_lstat(f).st_mode)
2016
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2017
raise errors.NoSuchFile(f)
2021
def until_no_eintr(f, *a, **kw):
2022
"""Run f(*a, **kw), retrying if an EINTR error occurs."""
2023
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2027
except (IOError, OSError), e:
2028
if e.errno == errno.EINTR:
2032
def re_compile_checked(re_string, flags=0, where=""):
2033
"""Return a compiled re, or raise a sensible error.
2035
This should only be used when compiling user-supplied REs.
2037
:param re_string: Text form of regular expression.
2038
:param flags: eg re.IGNORECASE
2039
:param where: Message explaining to the user the context where
2040
it occurred, eg 'log search filter'.
2042
# from https://bugs.launchpad.net/bzr/+bug/251352
2044
re_obj = re.compile(re_string, flags)
2049
where = ' in ' + where
2050
# despite the name 'error' is a type
2051
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2052
% (where, re_string, e))
2055
if sys.platform == "win32":
2058
return msvcrt.getch()
2063
fd = sys.stdin.fileno()
2064
settings = termios.tcgetattr(fd)
2067
ch = sys.stdin.read(1)
2069
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2073
if sys.platform == 'linux2':
2074
def _local_concurrency():
2076
prefix = 'processor'
2077
for line in file('/proc/cpuinfo', 'rb'):
2078
if line.startswith(prefix):
2079
concurrency = int(line[line.find(':')+1:]) + 1
2081
elif sys.platform == 'darwin':
2082
def _local_concurrency():
2083
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2084
stdout=subprocess.PIPE).communicate()[0]
2085
elif sys.platform[0:7] == 'freebsd':
2086
def _local_concurrency():
2087
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2088
stdout=subprocess.PIPE).communicate()[0]
2089
elif sys.platform == 'sunos5':
2090
def _local_concurrency():
2091
return subprocess.Popen(['psrinfo', '-p',],
2092
stdout=subprocess.PIPE).communicate()[0]
2093
elif sys.platform == "win32":
2094
def _local_concurrency():
2095
# This appears to return the number of cores.
2096
return os.environ.get('NUMBER_OF_PROCESSORS')
2098
def _local_concurrency():
2103
_cached_local_concurrency = None
2105
def local_concurrency(use_cache=True):
2106
"""Return how many processes can be run concurrently.
2108
Rely on platform specific implementations and default to 1 (one) if
2109
anything goes wrong.
2111
global _cached_local_concurrency
2113
if _cached_local_concurrency is not None and use_cache:
2114
return _cached_local_concurrency
2116
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2117
if concurrency is None:
2119
concurrency = _local_concurrency()
2120
except (OSError, IOError):
2123
concurrency = int(concurrency)
2124
except (TypeError, ValueError):
2127
_cached_concurrency = concurrency
2131
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2132
"""A stream writer that doesn't decode str arguments."""
2134
def __init__(self, encode, stream, errors='strict'):
2135
codecs.StreamWriter.__init__(self, stream, errors)
2136
self.encode = encode
2138
def write(self, object):
2139
if type(object) is str:
2140
self.stream.write(object)
2142
data, _ = self.encode(object, self.errors)
2143
self.stream.write(data)
2145
if sys.platform == 'win32':
2146
def open_file(filename, mode='r', bufsize=-1):
2147
"""This function is used to override the ``open`` builtin.
2149
But it uses O_NOINHERIT flag so the file handle is not inherited by
2150
child processes. Deleting or renaming a closed file opened with this
2151
function is not blocking child processes.
2153
writing = 'w' in mode
2154
appending = 'a' in mode
2155
updating = '+' in mode
2156
binary = 'b' in mode
2159
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2160
# for flags for each modes.
2170
flags |= os.O_WRONLY
2171
flags |= os.O_CREAT | os.O_TRUNC
2176
flags |= os.O_WRONLY
2177
flags |= os.O_CREAT | os.O_APPEND
2182
flags |= os.O_RDONLY
2184
return os.fdopen(os.open(filename, flags), mode, bufsize)
added
removed
bzrlib/osutils.py
