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)
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,
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
raise errors.BzrError('invalid file kind %r' % kind)
188
lexists = getattr(os.path, 'lexists', None)
192
stat = getattr(os, 'lstat', os.stat)
196
if e.errno == errno.ENOENT:
199
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
202
def fancy_rename(old, new, rename_func, unlink_func):
203
"""A fancy rename, when you don't have atomic rename.
205
:param old: The old path, to rename from
206
:param new: The new path, to rename to
207
:param rename_func: The potentially non-atomic rename function
208
:param unlink_func: A way to delete the target file if the full rename
211
# sftp rename doesn't allow overwriting, so play tricks:
212
base = os.path.basename(new)
213
dirname = os.path.dirname(new)
214
# callers use different encodings for the paths so the following MUST
215
# respect that. We rely on python upcasting to unicode if new is unicode
216
# and keeping a str if not.
217
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
218
os.getpid(), rand_chars(10))
219
tmp_name = pathjoin(dirname, tmp_name)
221
# Rename the file out of the way, but keep track if it didn't exist
222
# We don't want to grab just any exception
223
# something like EACCES should prevent us from continuing
224
# The downside is that the rename_func has to throw an exception
225
# with an errno = ENOENT, or NoSuchFile
228
rename_func(new, tmp_name)
229
except (errors.NoSuchFile,), e:
232
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
233
# function raises an IOError with errno is None when a rename fails.
234
# This then gets caught here.
235
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
238
if (getattr(e, 'errno', None) is None
239
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
248
# This may throw an exception, in which case success will
250
rename_func(old, new)
252
except (IOError, OSError), e:
253
# source and target may be aliases of each other (e.g. on a
254
# case-insensitive filesystem), so we may have accidentally renamed
255
# source by when we tried to rename target
256
failure_exc = sys.exc_info()
257
if (file_existed and e.errno in (None, errno.ENOENT)
258
and old.lower() == new.lower()):
259
# source and target are the same file on a case-insensitive
260
# filesystem, so we don't generate an exception
264
# If the file used to exist, rename it back into place
265
# otherwise just delete it from the tmp location
267
unlink_func(tmp_name)
269
rename_func(tmp_name, new)
270
if failure_exc is not None:
271
raise failure_exc[0], failure_exc[1], failure_exc[2]
274
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
275
# choke on a Unicode string containing a relative path if
276
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
278
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
279
def _posix_abspath(path):
280
# jam 20060426 rather than encoding to fsencoding
281
# copy posixpath.abspath, but use os.getcwdu instead
282
if not posixpath.isabs(path):
283
path = posixpath.join(getcwd(), path)
284
return posixpath.normpath(path)
287
def _posix_realpath(path):
288
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
291
def _win32_fixdrive(path):
292
"""Force drive letters to be consistent.
294
win32 is inconsistent whether it returns lower or upper case
295
and even if it was consistent the user might type the other
296
so we force it to uppercase
297
running python.exe under cmd.exe return capital C:\\
298
running win32 python inside a cygwin shell returns lowercase c:\\
300
drive, path = _nt_splitdrive(path)
301
return drive.upper() + path
304
def _win32_abspath(path):
305
# Real _nt_abspath doesn't have a problem with a unicode cwd
306
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
309
def _win98_abspath(path):
310
"""Return the absolute version of a path.
311
Windows 98 safe implementation (python reimplementation
312
of Win32 API function GetFullPathNameW)
317
# \\HOST\path => //HOST/path
318
# //HOST/path => //HOST/path
319
# path => C:/cwd/path
322
# check for absolute path
323
drive = _nt_splitdrive(path)[0]
324
if drive == '' and path[:2] not in('//','\\\\'):
326
# we cannot simply os.path.join cwd and path
327
# because os.path.join('C:','/path') produce '/path'
328
# and this is incorrect
329
if path[:1] in ('/','\\'):
330
cwd = _nt_splitdrive(cwd)[0]
332
path = cwd + '\\' + path
333
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
336
def _win32_realpath(path):
337
# Real _nt_realpath doesn't have a problem with a unicode cwd
338
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
341
def _win32_pathjoin(*args):
342
return _nt_join(*args).replace('\\', '/')
345
def _win32_normpath(path):
346
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
350
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
353
def _win32_mkdtemp(*args, **kwargs):
354
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
357
def _win32_rename(old, new):
358
"""We expect to be able to atomically replace 'new' with old.
360
On win32, if new exists, it must be moved out of the way first,
364
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
366
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
367
# If we try to rename a non-existant file onto cwd, we get
368
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
369
# if the old path doesn't exist, sometimes we get EACCES
370
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
376
return unicodedata.normalize('NFC', os.getcwdu())
379
# Default is to just use the python builtins, but these can be rebound on
380
# particular platforms.
381
abspath = _posix_abspath
382
realpath = _posix_realpath
383
pathjoin = os.path.join
384
normpath = os.path.normpath
387
dirname = os.path.dirname
388
basename = os.path.basename
389
split = os.path.split
390
splitext = os.path.splitext
391
# These were already imported into local scope
392
# mkdtemp = tempfile.mkdtemp
393
# rmtree = shutil.rmtree
395
MIN_ABS_PATHLENGTH = 1
398
if sys.platform == 'win32':
399
if win32utils.winver == 'Windows 98':
400
abspath = _win98_abspath
402
abspath = _win32_abspath
403
realpath = _win32_realpath
404
pathjoin = _win32_pathjoin
405
normpath = _win32_normpath
406
getcwd = _win32_getcwd
407
mkdtemp = _win32_mkdtemp
408
rename = _win32_rename
410
MIN_ABS_PATHLENGTH = 3
412
def _win32_delete_readonly(function, path, excinfo):
413
"""Error handler for shutil.rmtree function [for win32]
414
Helps to remove files and dirs marked as read-only.
416
exception = excinfo[1]
417
if function in (os.remove, os.rmdir) \
418
and isinstance(exception, OSError) \
419
and exception.errno == errno.EACCES:
425
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
426
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
427
return shutil.rmtree(path, ignore_errors, onerror)
429
f = win32utils.get_unicode_argv # special function or None
433
elif sys.platform == 'darwin':
437
def get_terminal_encoding():
438
"""Find the best encoding for printing to the screen.
440
This attempts to check both sys.stdout and sys.stdin to see
441
what encoding they are in, and if that fails it falls back to
442
osutils.get_user_encoding().
443
The problem is that on Windows, locale.getpreferredencoding()
444
is not the same encoding as that used by the console:
445
http://mail.python.org/pipermail/python-list/2003-May/162357.html
447
On my standard US Windows XP, the preferred encoding is
448
cp1252, but the console is cp437
450
from bzrlib.trace import mutter
451
output_encoding = getattr(sys.stdout, 'encoding', None)
452
if not output_encoding:
453
input_encoding = getattr(sys.stdin, 'encoding', None)
454
if not input_encoding:
455
output_encoding = get_user_encoding()
456
mutter('encoding stdout as osutils.get_user_encoding() %r',
459
output_encoding = input_encoding
460
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
462
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
463
if output_encoding == 'cp0':
464
# invalid encoding (cp0 means 'no codepage' on Windows)
465
output_encoding = get_user_encoding()
466
mutter('cp0 is invalid encoding.'
467
' encoding stdout as osutils.get_user_encoding() %r',
471
codecs.lookup(output_encoding)
473
sys.stderr.write('bzr: warning:'
474
' unknown terminal encoding %s.\n'
475
' Using encoding %s instead.\n'
476
% (output_encoding, get_user_encoding())
478
output_encoding = get_user_encoding()
480
return output_encoding
483
def normalizepath(f):
484
if getattr(os.path, 'realpath', None) is not None:
488
[p,e] = os.path.split(f)
489
if e == "" or e == "." or e == "..":
492
return pathjoin(F(p), e)
496
"""True if f is an accessible directory."""
498
return S_ISDIR(os.lstat(f)[ST_MODE])
504
"""True if f is a regular file."""
506
return S_ISREG(os.lstat(f)[ST_MODE])
511
"""True if f is a symlink."""
513
return S_ISLNK(os.lstat(f)[ST_MODE])
517
def is_inside(dir, fname):
518
"""True if fname is inside dir.
520
The parameters should typically be passed to osutils.normpath first, so
521
that . and .. and repeated slashes are eliminated, and the separators
522
are canonical for the platform.
524
The empty string as a dir name is taken as top-of-tree and matches
527
# XXX: Most callers of this can actually do something smarter by
528
# looking at the inventory
538
return fname.startswith(dir)
541
def is_inside_any(dir_list, fname):
542
"""True if fname is inside any of given dirs."""
543
for dirname in dir_list:
544
if is_inside(dirname, fname):
549
def is_inside_or_parent_of_any(dir_list, fname):
550
"""True if fname is a child or a parent of any of the given files."""
551
for dirname in dir_list:
552
if is_inside(dirname, fname) or is_inside(fname, dirname):
557
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
558
report_activity=None, direction='read'):
559
"""Copy contents of one file to another.
561
The read_length can either be -1 to read to end-of-file (EOF) or
562
it can specify the maximum number of bytes to read.
564
The buff_size represents the maximum size for each read operation
565
performed on from_file.
567
:param report_activity: Call this as bytes are read, see
568
Transport._report_activity
569
:param direction: Will be passed to report_activity
571
:return: The number of bytes copied.
575
# read specified number of bytes
577
while read_length > 0:
578
num_bytes_to_read = min(read_length, buff_size)
580
block = from_file.read(num_bytes_to_read)
584
if report_activity is not None:
585
report_activity(len(block), direction)
588
actual_bytes_read = len(block)
589
read_length -= actual_bytes_read
590
length += actual_bytes_read
594
block = from_file.read(buff_size)
598
if report_activity is not None:
599
report_activity(len(block), direction)
605
def pump_string_file(bytes, file_handle, segment_size=None):
606
"""Write bytes to file_handle in many smaller writes.
608
:param bytes: The string to write.
609
:param file_handle: The file to write to.
611
# Write data in chunks rather than all at once, because very large
612
# writes fail on some platforms (e.g. Windows with SMB mounted
615
segment_size = 5242880 # 5MB
616
segments = range(len(bytes) / segment_size + 1)
617
write = file_handle.write
618
for segment_index in segments:
619
segment = buffer(bytes, segment_index * segment_size, segment_size)
623
def file_iterator(input_file, readsize=32768):
625
b = input_file.read(readsize)
632
"""Calculate the hexdigest of an open file.
634
The file cursor should be already at the start.
646
def size_sha_file(f):
647
"""Calculate the size and hexdigest of an open file.
649
The file cursor should be already at the start and
650
the caller is responsible for closing the file afterwards.
661
return size, s.hexdigest()
664
def sha_file_by_name(fname):
665
"""Calculate the SHA1 of a file by reading the full text"""
667
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
670
b = os.read(f, 1<<16)
678
def sha_strings(strings, _factory=sha):
679
"""Return the sha-1 of concatenation of strings"""
681
map(s.update, strings)
685
def sha_string(f, _factory=sha):
686
return _factory(f).hexdigest()
689
def fingerprint_file(f):
691
return {'size': len(b),
692
'sha1': sha(b).hexdigest()}
695
def compare_files(a, b):
696
"""Returns true if equal in contents"""
707
def local_time_offset(t=None):
708
"""Return offset of local zone from GMT, either at present or at time t."""
711
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
712
return offset.days * 86400 + offset.seconds
714
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
715
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
718
def format_date(t, offset=0, timezone='original', date_fmt=None,
720
"""Return a formatted date string.
722
:param t: Seconds since the epoch.
723
:param offset: Timezone offset in seconds east of utc.
724
:param timezone: How to display the time: 'utc', 'original' for the
725
timezone specified by offset, or 'local' for the process's current
727
:param date_fmt: strftime format.
728
:param show_offset: Whether to append the timezone.
730
(date_fmt, tt, offset_str) = \
731
_format_date(t, offset, timezone, date_fmt, show_offset)
732
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
733
date_str = time.strftime(date_fmt, tt)
734
return date_str + offset_str
737
# Cache of formatted offset strings
741
def format_date_with_offset_in_original_timezone(t, offset=0,
742
_cache=_offset_cache):
743
"""Return a formatted date string in the original timezone.
745
This routine may be faster then format_date.
747
:param t: Seconds since the epoch.
748
:param offset: Timezone offset in seconds east of utc.
752
tt = time.gmtime(t + offset)
753
date_fmt = _default_format_by_weekday_num[tt[6]]
754
date_str = time.strftime(date_fmt, tt)
755
offset_str = _cache.get(offset, None)
756
if offset_str is None:
757
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
758
_cache[offset] = offset_str
759
return date_str + offset_str
762
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
764
"""Return an unicode date string formatted according to the current locale.
766
:param t: Seconds since the epoch.
767
:param offset: Timezone offset in seconds east of utc.
768
:param timezone: How to display the time: 'utc', 'original' for the
769
timezone specified by offset, or 'local' for the process's current
771
:param date_fmt: strftime format.
772
:param show_offset: Whether to append the timezone.
774
(date_fmt, tt, offset_str) = \
775
_format_date(t, offset, timezone, date_fmt, show_offset)
776
date_str = time.strftime(date_fmt, tt)
777
if not isinstance(date_str, unicode):
778
date_str = date_str.decode(get_user_encoding(), 'replace')
779
return date_str + offset_str
782
def _format_date(t, offset, timezone, date_fmt, show_offset):
783
if timezone == 'utc':
786
elif timezone == 'original':
789
tt = time.gmtime(t + offset)
790
elif timezone == 'local':
791
tt = time.localtime(t)
792
offset = local_time_offset(t)
794
raise errors.UnsupportedTimezoneFormat(timezone)
796
date_fmt = "%a %Y-%m-%d %H:%M:%S"
798
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
801
return (date_fmt, tt, offset_str)
804
def compact_date(when):
805
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
808
def format_delta(delta):
809
"""Get a nice looking string for a time delta.
811
:param delta: The time difference in seconds, can be positive or negative.
812
positive indicates time in the past, negative indicates time in the
813
future. (usually time.time() - stored_time)
814
:return: String formatted to show approximate resolution
820
direction = 'in the future'
824
if seconds < 90: # print seconds up to 90 seconds
826
return '%d second %s' % (seconds, direction,)
828
return '%d seconds %s' % (seconds, direction)
830
minutes = int(seconds / 60)
831
seconds -= 60 * minutes
836
if minutes < 90: # print minutes, seconds up to 90 minutes
838
return '%d minute, %d second%s %s' % (
839
minutes, seconds, plural_seconds, direction)
841
return '%d minutes, %d second%s %s' % (
842
minutes, seconds, plural_seconds, direction)
844
hours = int(minutes / 60)
845
minutes -= 60 * hours
852
return '%d hour, %d minute%s %s' % (hours, minutes,
853
plural_minutes, direction)
854
return '%d hours, %d minute%s %s' % (hours, minutes,
855
plural_minutes, direction)
858
"""Return size of given open file."""
859
return os.fstat(f.fileno())[ST_SIZE]
862
# Define rand_bytes based on platform.
864
# Python 2.4 and later have os.urandom,
865
# but it doesn't work on some arches
867
rand_bytes = os.urandom
868
except (NotImplementedError, AttributeError):
869
# If python doesn't have os.urandom, or it doesn't work,
870
# then try to first pull random data from /dev/urandom
872
rand_bytes = file('/dev/urandom', 'rb').read
873
# Otherwise, use this hack as a last resort
874
except (IOError, OSError):
875
# not well seeded, but better than nothing
880
s += chr(random.randint(0, 255))
885
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
887
"""Return a random string of num alphanumeric characters
889
The result only contains lowercase chars because it may be used on
890
case-insensitive filesystems.
893
for raw_byte in rand_bytes(num):
894
s += ALNUM[ord(raw_byte) % 36]
898
## TODO: We could later have path objects that remember their list
899
## decomposition (might be too tricksy though.)
902
"""Turn string into list of parts."""
903
# split on either delimiter because people might use either on
905
ps = re.split(r'[\\/]', p)
910
raise errors.BzrError("sorry, %r not allowed in path" % f)
911
elif (f == '.') or (f == ''):
920
if (f == '..') or (f is None) or (f == ''):
921
raise errors.BzrError("sorry, %r not allowed in path" % f)
925
def parent_directories(filename):
926
"""Return the list of parent directories, deepest first.
928
For example, parent_directories("a/b/c") -> ["a/b", "a"].
931
parts = splitpath(dirname(filename))
933
parents.append(joinpath(parts))
938
_extension_load_failures = []
941
def failed_to_load_extension(exception):
942
"""Handle failing to load a binary extension.
944
This should be called from the ImportError block guarding the attempt to
945
import the native extension. If this function returns, the pure-Python
946
implementation should be loaded instead::
949
>>> import bzrlib._fictional_extension_pyx
950
>>> except ImportError, e:
951
>>> bzrlib.osutils.failed_to_load_extension(e)
952
>>> import bzrlib._fictional_extension_py
954
# NB: This docstring is just an example, not a doctest, because doctest
955
# currently can't cope with the use of lazy imports in this namespace --
958
# This currently doesn't report the failure at the time it occurs, because
959
# they tend to happen very early in startup when we can't check config
960
# files etc, and also we want to report all failures but not spam the user
962
from bzrlib import trace
963
exception_str = str(exception)
964
if exception_str not in _extension_load_failures:
965
trace.mutter("failed to load compiled extension: %s" % exception_str)
966
_extension_load_failures.append(exception_str)
969
def report_extension_load_failures():
970
if not _extension_load_failures:
972
from bzrlib.config import GlobalConfig
973
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
975
# the warnings framework should by default show this only once
976
from bzrlib.trace import warning
978
"bzr: warning: some compiled extensions could not be loaded; "
979
"see <https://answers.launchpad.net/bzr/+faq/703>")
980
# we no longer show the specific missing extensions here, because it makes
981
# the message too long and scary - see
982
# https://bugs.launchpad.net/bzr/+bug/430529
986
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
987
except ImportError, e:
988
failed_to_load_extension(e)
989
from bzrlib._chunks_to_lines_py import chunks_to_lines
993
"""Split s into lines, but without removing the newline characters."""
994
# Trivially convert a fulltext into a 'chunked' representation, and let
995
# chunks_to_lines do the heavy lifting.
996
if isinstance(s, str):
997
# chunks_to_lines only supports 8-bit strings
998
return chunks_to_lines([s])
1000
return _split_lines(s)
1003
def _split_lines(s):
1004
"""Split s into lines, but without removing the newline characters.
1006
This supports Unicode or plain string objects.
1008
lines = s.split('\n')
1009
result = [line + '\n' for line in lines[:-1]]
1011
result.append(lines[-1])
1015
def hardlinks_good():
1016
return sys.platform not in ('win32', 'cygwin', 'darwin')
1019
def link_or_copy(src, dest):
1020
"""Hardlink a file, or copy it if it can't be hardlinked."""
1021
if not hardlinks_good():
1022
shutil.copyfile(src, dest)
1026
except (OSError, IOError), e:
1027
if e.errno != errno.EXDEV:
1029
shutil.copyfile(src, dest)
1032
def delete_any(path):
1033
"""Delete a file, symlink or directory.
1035
Will delete even if readonly.
1038
_delete_file_or_dir(path)
1039
except (OSError, IOError), e:
1040
if e.errno in (errno.EPERM, errno.EACCES):
1041
# make writable and try again
1044
except (OSError, IOError):
1046
_delete_file_or_dir(path)
1051
def _delete_file_or_dir(path):
1052
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1053
# Forgiveness than Permission (EAFP) because:
1054
# - root can damage a solaris file system by using unlink,
1055
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1056
# EACCES, OSX: EPERM) when invoked on a directory.
1057
if isdir(path): # Takes care of symlinks
1064
if getattr(os, 'symlink', None) is not None:
1070
def has_hardlinks():
1071
if getattr(os, 'link', None) is not None:
1077
def host_os_dereferences_symlinks():
1078
return (has_symlinks()
1079
and sys.platform not in ('cygwin', 'win32'))
1082
def readlink(abspath):
1083
"""Return a string representing the path to which the symbolic link points.
1085
:param abspath: The link absolute unicode path.
1087
This his guaranteed to return the symbolic link in unicode in all python
1090
link = abspath.encode(_fs_enc)
1091
target = os.readlink(link)
1092
target = target.decode(_fs_enc)
1096
def contains_whitespace(s):
1097
"""True if there are any whitespace characters in s."""
1098
# string.whitespace can include '\xa0' in certain locales, because it is
1099
# considered "non-breaking-space" as part of ISO-8859-1. But it
1100
# 1) Isn't a breaking whitespace
1101
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1103
# 3) '\xa0' isn't unicode safe since it is >128.
1105
# This should *not* be a unicode set of characters in case the source
1106
# string is not a Unicode string. We can auto-up-cast the characters since
1107
# they are ascii, but we don't want to auto-up-cast the string in case it
1109
for ch in ' \t\n\r\v\f':
1116
def contains_linebreaks(s):
1117
"""True if there is any vertical whitespace in s."""
1125
def relpath(base, path):
1126
"""Return path relative to base, or raise exception.
1128
The path may be either an absolute path or a path relative to the
1129
current working directory.
1131
os.path.commonprefix (python2.4) has a bad bug that it works just
1132
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1133
avoids that problem.
1136
if len(base) < MIN_ABS_PATHLENGTH:
1137
# must have space for e.g. a drive letter
1138
raise ValueError('%r is too short to calculate a relative path'
1146
if len(head) <= len(base) and head != base:
1147
raise errors.PathNotChild(rp, base)
1150
head, tail = split(head)
1155
return pathjoin(*reversed(s))
1160
def _cicp_canonical_relpath(base, path):
1161
"""Return the canonical path relative to base.
1163
Like relpath, but on case-insensitive-case-preserving file-systems, this
1164
will return the relpath as stored on the file-system rather than in the
1165
case specified in the input string, for all existing portions of the path.
1167
This will cause O(N) behaviour if called for every path in a tree; if you
1168
have a number of paths to convert, you should use canonical_relpaths().
1170
# TODO: it should be possible to optimize this for Windows by using the
1171
# win32 API FindFiles function to look for the specified name - but using
1172
# os.listdir() still gives us the correct, platform agnostic semantics in
1175
rel = relpath(base, path)
1176
# '.' will have been turned into ''
1180
abs_base = abspath(base)
1182
_listdir = os.listdir
1184
# use an explicit iterator so we can easily consume the rest on early exit.
1185
bit_iter = iter(rel.split('/'))
1186
for bit in bit_iter:
1189
next_entries = _listdir(current)
1190
except OSError: # enoent, eperm, etc
1191
# We can't find this in the filesystem, so just append the
1193
current = pathjoin(current, bit, *list(bit_iter))
1195
for look in next_entries:
1196
if lbit == look.lower():
1197
current = pathjoin(current, look)
1200
# got to the end, nothing matched, so we just return the
1201
# non-existing bits as they were specified (the filename may be
1202
# the target of a move, for example).
1203
current = pathjoin(current, bit, *list(bit_iter))
1205
return current[len(abs_base):].lstrip('/')
1207
# XXX - TODO - we need better detection/integration of case-insensitive
1208
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1209
# filesystems), for example, so could probably benefit from the same basic
1210
# support there. For now though, only Windows and OSX get that support, and
1211
# they get it for *all* file-systems!
1212
if sys.platform in ('win32', 'darwin'):
1213
canonical_relpath = _cicp_canonical_relpath
1215
canonical_relpath = relpath
1217
def canonical_relpaths(base, paths):
1218
"""Create an iterable to canonicalize a sequence of relative paths.
1220
The intent is for this implementation to use a cache, vastly speeding
1221
up multiple transformations in the same directory.
1223
# but for now, we haven't optimized...
1224
return [canonical_relpath(base, p) for p in paths]
1226
def safe_unicode(unicode_or_utf8_string):
1227
"""Coerce unicode_or_utf8_string into unicode.
1229
If it is unicode, it is returned.
1230
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1231
wrapped in a BzrBadParameterNotUnicode exception.
1233
if isinstance(unicode_or_utf8_string, unicode):
1234
return unicode_or_utf8_string
1236
return unicode_or_utf8_string.decode('utf8')
1237
except UnicodeDecodeError:
1238
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1241
def safe_utf8(unicode_or_utf8_string):
1242
"""Coerce unicode_or_utf8_string to a utf8 string.
1244
If it is a str, it is returned.
1245
If it is Unicode, it is encoded into a utf-8 string.
1247
if isinstance(unicode_or_utf8_string, str):
1248
# TODO: jam 20070209 This is overkill, and probably has an impact on
1249
# performance if we are dealing with lots of apis that want a
1252
# Make sure it is a valid utf-8 string
1253
unicode_or_utf8_string.decode('utf-8')
1254
except UnicodeDecodeError:
1255
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1256
return unicode_or_utf8_string
1257
return unicode_or_utf8_string.encode('utf-8')
1260
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1261
' Revision id generators should be creating utf8'
1265
def safe_revision_id(unicode_or_utf8_string, warn=True):
1266
"""Revision ids should now be utf8, but at one point they were unicode.
1268
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1270
:param warn: Functions that are sanitizing user data can set warn=False
1271
:return: None or a utf8 revision id.
1273
if (unicode_or_utf8_string is None
1274
or unicode_or_utf8_string.__class__ == str):
1275
return unicode_or_utf8_string
1277
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1279
return cache_utf8.encode(unicode_or_utf8_string)
1282
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1283
' generators should be creating utf8 file ids.')
1286
def safe_file_id(unicode_or_utf8_string, warn=True):
1287
"""File ids should now be utf8, but at one point they were unicode.
1289
This is the same as safe_utf8, except it uses the cached encode functions
1290
to save a little bit of performance.
1292
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1294
:param warn: Functions that are sanitizing user data can set warn=False
1295
:return: None or a utf8 file id.
1297
if (unicode_or_utf8_string is None
1298
or unicode_or_utf8_string.__class__ == str):
1299
return unicode_or_utf8_string
1301
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1303
return cache_utf8.encode(unicode_or_utf8_string)
1306
_platform_normalizes_filenames = False
1307
if sys.platform == 'darwin':
1308
_platform_normalizes_filenames = True
1311
def normalizes_filenames():
1312
"""Return True if this platform normalizes unicode filenames.
1314
Mac OSX does, Windows/Linux do not.
1316
return _platform_normalizes_filenames
1319
def _accessible_normalized_filename(path):
1320
"""Get the unicode normalized path, and if you can access the file.
1322
On platforms where the system normalizes filenames (Mac OSX),
1323
you can access a file by any path which will normalize correctly.
1324
On platforms where the system does not normalize filenames
1325
(Windows, Linux), you have to access a file by its exact path.
1327
Internally, bzr only supports NFC normalization, since that is
1328
the standard for XML documents.
1330
So return the normalized path, and a flag indicating if the file
1331
can be accessed by that path.
1334
return unicodedata.normalize('NFC', unicode(path)), True
1337
def _inaccessible_normalized_filename(path):
1338
__doc__ = _accessible_normalized_filename.__doc__
1340
normalized = unicodedata.normalize('NFC', unicode(path))
1341
return normalized, normalized == path
1344
if _platform_normalizes_filenames:
1345
normalized_filename = _accessible_normalized_filename
1347
normalized_filename = _inaccessible_normalized_filename
1350
def set_signal_handler(signum, handler, restart_syscall=True):
1351
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1352
on platforms that support that.
1354
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1355
automatically restart (by calling `signal.siginterrupt(signum,
1356
False)`). May be ignored if the feature is not available on this
1357
platform or Python version.
1359
old_handler = signal.signal(signum, handler)
1362
siginterrupt = signal.siginterrupt
1363
except AttributeError: # siginterrupt doesn't exist on this platform, or for this version of
1367
siginterrupt(signum, False)
1371
default_terminal_width = 80
1372
"""The default terminal width for ttys.
1374
This is defined so that higher levels can share a common fallback value when
1375
terminal_width() returns None.
1379
def terminal_width():
1380
"""Return terminal width.
1382
None is returned if the width can't established precisely.
1385
- if BZR_COLUMNS is set, returns its value
1386
- if there is no controlling terminal, returns None
1387
- if COLUMNS is set, returns its value,
1389
From there, we need to query the OS to get the size of the controlling
1393
- get termios.TIOCGWINSZ
1394
- if an error occurs or a negative value is obtained, returns None
1398
- win32utils.get_console_size() decides,
1399
- returns None on error (provided default value)
1402
# If BZR_COLUMNS is set, take it, user is always right
1404
return int(os.environ['BZR_COLUMNS'])
1405
except (KeyError, ValueError):
1408
isatty = getattr(sys.stdout, 'isatty', None)
1409
if isatty is None or not isatty():
1410
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1413
# If COLUMNS is set, take it, the terminal knows better (even inside a
1414
# given terminal, the application can decide to set COLUMNS to a lower
1415
# value (splitted screen) or a bigger value (scroll bars))
1417
return int(os.environ['COLUMNS'])
1418
except (KeyError, ValueError):
1421
width, height = _terminal_size(None, None)
1423
# Consider invalid values as meaning no width
1429
def _win32_terminal_size(width, height):
1430
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1431
return width, height
1434
def _ioctl_terminal_size(width, height):
1436
import struct, fcntl, termios
1437
s = struct.pack('HHHH', 0, 0, 0, 0)
1438
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1439
height, width = struct.unpack('HHHH', x)[0:2]
1440
except (IOError, AttributeError):
1442
return width, height
1444
_terminal_size = None
1445
"""Returns the terminal size as (width, height).
1447
:param width: Default value for width.
1448
:param height: Default value for height.
1450
This is defined specifically for each OS and query the size of the controlling
1451
terminal. If any error occurs, the provided default values should be returned.
1453
if sys.platform == 'win32':
1454
_terminal_size = _win32_terminal_size
1456
_terminal_size = _ioctl_terminal_size
1459
def _terminal_size_changed(signum, frame):
1460
"""Set COLUMNS upon receiving a SIGnal for WINdow size CHange."""
1461
width, height = _terminal_size(None, None)
1462
if width is not None:
1463
os.environ['COLUMNS'] = str(width)
1466
_registered_sigwinch = False
1468
def watch_sigwinch():
1469
"""Register for SIGWINCH, once and only once."""
1470
global _registered_sigwinch
1471
if not _registered_sigwinch:
1472
if sys.platform == 'win32':
1473
# Martin (gz) mentioned WINDOW_BUFFER_SIZE_RECORD from
1474
# ReadConsoleInput but I've no idea how to plug that in
1475
# the current design -- vila 20091216
1478
set_signal_handler(signal.SIGWINCH, _terminal_size_changed)
1479
_registered_sigwinch = True
1482
def supports_executable():
1483
return sys.platform != "win32"
1486
def supports_posix_readonly():
1487
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1489
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1490
directory controls creation/deletion, etc.
1492
And under win32, readonly means that the directory itself cannot be
1493
deleted. The contents of a readonly directory can be changed, unlike POSIX
1494
where files in readonly directories cannot be added, deleted or renamed.
1496
return sys.platform != "win32"
1499
def set_or_unset_env(env_variable, value):
1500
"""Modify the environment, setting or removing the env_variable.
1502
:param env_variable: The environment variable in question
1503
:param value: The value to set the environment to. If None, then
1504
the variable will be removed.
1505
:return: The original value of the environment variable.
1507
orig_val = os.environ.get(env_variable)
1509
if orig_val is not None:
1510
del os.environ[env_variable]
1512
if isinstance(value, unicode):
1513
value = value.encode(get_user_encoding())
1514
os.environ[env_variable] = value
1518
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1521
def check_legal_path(path):
1522
"""Check whether the supplied path is legal.
1523
This is only required on Windows, so we don't test on other platforms
1526
if sys.platform != "win32":
1528
if _validWin32PathRE.match(path) is None:
1529
raise errors.IllegalPath(path)
1532
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1534
def _is_error_enotdir(e):
1535
"""Check if this exception represents ENOTDIR.
1537
Unfortunately, python is very inconsistent about the exception
1538
here. The cases are:
1539
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1540
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1541
which is the windows error code.
1542
3) Windows, Python2.5 uses errno == EINVAL and
1543
winerror == ERROR_DIRECTORY
1545
:param e: An Exception object (expected to be OSError with an errno
1546
attribute, but we should be able to cope with anything)
1547
:return: True if this represents an ENOTDIR error. False otherwise.
1549
en = getattr(e, 'errno', None)
1550
if (en == errno.ENOTDIR
1551
or (sys.platform == 'win32'
1552
and (en == _WIN32_ERROR_DIRECTORY
1553
or (en == errno.EINVAL
1554
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1560
def walkdirs(top, prefix=""):
1561
"""Yield data about all the directories in a tree.
1563
This yields all the data about the contents of a directory at a time.
1564
After each directory has been yielded, if the caller has mutated the list
1565
to exclude some directories, they are then not descended into.
1567
The data yielded is of the form:
1568
((directory-relpath, directory-path-from-top),
1569
[(relpath, basename, kind, lstat, path-from-top), ...]),
1570
- directory-relpath is the relative path of the directory being returned
1571
with respect to top. prefix is prepended to this.
1572
- directory-path-from-root is the path including top for this directory.
1573
It is suitable for use with os functions.
1574
- relpath is the relative path within the subtree being walked.
1575
- basename is the basename of the path
1576
- kind is the kind of the file now. If unknown then the file is not
1577
present within the tree - but it may be recorded as versioned. See
1579
- lstat is the stat data *if* the file was statted.
1580
- planned, not implemented:
1581
path_from_tree_root is the path from the root of the tree.
1583
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1584
allows one to walk a subtree but get paths that are relative to a tree
1586
:return: an iterator over the dirs.
1588
#TODO there is a bit of a smell where the results of the directory-
1589
# summary in this, and the path from the root, may not agree
1590
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1591
# potentially confusing output. We should make this more robust - but
1592
# not at a speed cost. RBC 20060731
1594
_directory = _directory_kind
1595
_listdir = os.listdir
1596
_kind_from_mode = file_kind_from_stat_mode
1597
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1599
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1600
relroot, _, _, _, top = pending.pop()
1602
relprefix = relroot + u'/'
1605
top_slash = top + u'/'
1608
append = dirblock.append
1610
names = sorted(_listdir(top))
1612
if not _is_error_enotdir(e):
1616
abspath = top_slash + name
1617
statvalue = _lstat(abspath)
1618
kind = _kind_from_mode(statvalue.st_mode)
1619
append((relprefix + name, name, kind, statvalue, abspath))
1620
yield (relroot, top), dirblock
1622
# push the user specified dirs from dirblock
1623
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1626
class DirReader(object):
1627
"""An interface for reading directories."""
1629
def top_prefix_to_starting_dir(self, top, prefix=""):
1630
"""Converts top and prefix to a starting dir entry
1632
:param top: A utf8 path
1633
:param prefix: An optional utf8 path to prefix output relative paths
1635
:return: A tuple starting with prefix, and ending with the native
1638
raise NotImplementedError(self.top_prefix_to_starting_dir)
1640
def read_dir(self, prefix, top):
1641
"""Read a specific dir.
1643
:param prefix: A utf8 prefix to be preprended to the path basenames.
1644
:param top: A natively encoded path to read.
1645
:return: A list of the directories contents. Each item contains:
1646
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1648
raise NotImplementedError(self.read_dir)
1651
_selected_dir_reader = None
1654
def _walkdirs_utf8(top, prefix=""):
1655
"""Yield data about all the directories in a tree.
1657
This yields the same information as walkdirs() only each entry is yielded
1658
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1659
are returned as exact byte-strings.
1661
:return: yields a tuple of (dir_info, [file_info])
1662
dir_info is (utf8_relpath, path-from-top)
1663
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1664
if top is an absolute path, path-from-top is also an absolute path.
1665
path-from-top might be unicode or utf8, but it is the correct path to
1666
pass to os functions to affect the file in question. (such as os.lstat)
1668
global _selected_dir_reader
1669
if _selected_dir_reader is None:
1670
fs_encoding = _fs_enc.upper()
1671
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1672
# Win98 doesn't have unicode apis like FindFirstFileW
1673
# TODO: We possibly could support Win98 by falling back to the
1674
# original FindFirstFile, and using TCHAR instead of WCHAR,
1675
# but that gets a bit tricky, and requires custom compiling
1678
from bzrlib._walkdirs_win32 import Win32ReadDir
1679
_selected_dir_reader = Win32ReadDir()
1682
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1683
# ANSI_X3.4-1968 is a form of ASCII
1685
from bzrlib._readdir_pyx import UTF8DirReader
1686
_selected_dir_reader = UTF8DirReader()
1687
except ImportError, e:
1688
failed_to_load_extension(e)
1691
if _selected_dir_reader is None:
1692
# Fallback to the python version
1693
_selected_dir_reader = UnicodeDirReader()
1695
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1696
# But we don't actually uses 1-3 in pending, so set them to None
1697
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1698
read_dir = _selected_dir_reader.read_dir
1699
_directory = _directory_kind
1701
relroot, _, _, _, top = pending[-1].pop()
1704
dirblock = sorted(read_dir(relroot, top))
1705
yield (relroot, top), dirblock
1706
# push the user specified dirs from dirblock
1707
next = [d for d in reversed(dirblock) if d[2] == _directory]
1709
pending.append(next)
1712
class UnicodeDirReader(DirReader):
1713
"""A dir reader for non-utf8 file systems, which transcodes."""
1715
__slots__ = ['_utf8_encode']
1718
self._utf8_encode = codecs.getencoder('utf8')
1720
def top_prefix_to_starting_dir(self, top, prefix=""):
1721
"""See DirReader.top_prefix_to_starting_dir."""
1722
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1724
def read_dir(self, prefix, top):
1725
"""Read a single directory from a non-utf8 file system.
1727
top, and the abspath element in the output are unicode, all other paths
1728
are utf8. Local disk IO is done via unicode calls to listdir etc.
1730
This is currently the fallback code path when the filesystem encoding is
1731
not UTF-8. It may be better to implement an alternative so that we can
1732
safely handle paths that are not properly decodable in the current
1735
See DirReader.read_dir for details.
1737
_utf8_encode = self._utf8_encode
1739
_listdir = os.listdir
1740
_kind_from_mode = file_kind_from_stat_mode
1743
relprefix = prefix + '/'
1746
top_slash = top + u'/'
1749
append = dirblock.append
1750
for name in sorted(_listdir(top)):
1752
name_utf8 = _utf8_encode(name)[0]
1753
except UnicodeDecodeError:
1754
raise errors.BadFilenameEncoding(
1755
_utf8_encode(relprefix)[0] + name, _fs_enc)
1756
abspath = top_slash + name
1757
statvalue = _lstat(abspath)
1758
kind = _kind_from_mode(statvalue.st_mode)
1759
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1763
def copy_tree(from_path, to_path, handlers={}):
1764
"""Copy all of the entries in from_path into to_path.
1766
:param from_path: The base directory to copy.
1767
:param to_path: The target directory. If it does not exist, it will
1769
:param handlers: A dictionary of functions, which takes a source and
1770
destinations for files, directories, etc.
1771
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1772
'file', 'directory', and 'symlink' should always exist.
1773
If they are missing, they will be replaced with 'os.mkdir()',
1774
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1776
# Now, just copy the existing cached tree to the new location
1777
# We use a cheap trick here.
1778
# Absolute paths are prefixed with the first parameter
1779
# relative paths are prefixed with the second.
1780
# So we can get both the source and target returned
1781
# without any extra work.
1783
def copy_dir(source, dest):
1786
def copy_link(source, dest):
1787
"""Copy the contents of a symlink"""
1788
link_to = os.readlink(source)
1789
os.symlink(link_to, dest)
1791
real_handlers = {'file':shutil.copy2,
1792
'symlink':copy_link,
1793
'directory':copy_dir,
1795
real_handlers.update(handlers)
1797
if not os.path.exists(to_path):
1798
real_handlers['directory'](from_path, to_path)
1800
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1801
for relpath, name, kind, st, abspath in entries:
1802
real_handlers[kind](abspath, relpath)
1805
def path_prefix_key(path):
1806
"""Generate a prefix-order path key for path.
1808
This can be used to sort paths in the same way that walkdirs does.
1810
return (dirname(path) , path)
1813
def compare_paths_prefix_order(path_a, path_b):
1814
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1815
key_a = path_prefix_key(path_a)
1816
key_b = path_prefix_key(path_b)
1817
return cmp(key_a, key_b)
1820
_cached_user_encoding = None
1823
def get_user_encoding(use_cache=True):
1824
"""Find out what the preferred user encoding is.
1826
This is generally the encoding that is used for command line parameters
1827
and file contents. This may be different from the terminal encoding
1828
or the filesystem encoding.
1830
:param use_cache: Enable cache for detected encoding.
1831
(This parameter is turned on by default,
1832
and required only for selftesting)
1834
:return: A string defining the preferred user encoding
1836
global _cached_user_encoding
1837
if _cached_user_encoding is not None and use_cache:
1838
return _cached_user_encoding
1840
if sys.platform == 'darwin':
1841
# python locale.getpreferredencoding() always return
1842
# 'mac-roman' on darwin. That's a lie.
1843
sys.platform = 'posix'
1845
if os.environ.get('LANG', None) is None:
1846
# If LANG is not set, we end up with 'ascii', which is bad
1847
# ('mac-roman' is more than ascii), so we set a default which
1848
# will give us UTF-8 (which appears to work in all cases on
1849
# OSX). Users are still free to override LANG of course, as
1850
# long as it give us something meaningful. This work-around
1851
# *may* not be needed with python 3k and/or OSX 10.5, but will
1852
# work with them too -- vila 20080908
1853
os.environ['LANG'] = 'en_US.UTF-8'
1856
sys.platform = 'darwin'
1861
user_encoding = locale.getpreferredencoding()
1862
except locale.Error, e:
1863
sys.stderr.write('bzr: warning: %s\n'
1864
' Could not determine what text encoding to use.\n'
1865
' This error usually means your Python interpreter\n'
1866
' doesn\'t support the locale set by $LANG (%s)\n'
1867
" Continuing with ascii encoding.\n"
1868
% (e, os.environ.get('LANG')))
1869
user_encoding = 'ascii'
1871
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1872
# treat that as ASCII, and not support printing unicode characters to the
1875
# For python scripts run under vim, we get '', so also treat that as ASCII
1876
if user_encoding in (None, 'cp0', ''):
1877
user_encoding = 'ascii'
1881
codecs.lookup(user_encoding)
1883
sys.stderr.write('bzr: warning:'
1884
' unknown encoding %s.'
1885
' Continuing with ascii encoding.\n'
1888
user_encoding = 'ascii'
1891
_cached_user_encoding = user_encoding
1893
return user_encoding
1896
def get_host_name():
1897
"""Return the current unicode host name.
1899
This is meant to be used in place of socket.gethostname() because that
1900
behaves inconsistently on different platforms.
1902
if sys.platform == "win32":
1904
return win32utils.get_host_name()
1907
return socket.gethostname().decode(get_user_encoding())
1910
def recv_all(socket, bytes):
1911
"""Receive an exact number of bytes.
1913
Regular Socket.recv() may return less than the requested number of bytes,
1914
dependning on what's in the OS buffer. MSG_WAITALL is not available
1915
on all platforms, but this should work everywhere. This will return
1916
less than the requested amount if the remote end closes.
1918
This isn't optimized and is intended mostly for use in testing.
1921
while len(b) < bytes:
1922
new = until_no_eintr(socket.recv, bytes - len(b))
1929
def send_all(socket, bytes, report_activity=None):
1930
"""Send all bytes on a socket.
1932
Regular socket.sendall() can give socket error 10053 on Windows. This
1933
implementation sends no more than 64k at a time, which avoids this problem.
1935
:param report_activity: Call this as bytes are read, see
1936
Transport._report_activity
1939
for pos in xrange(0, len(bytes), chunk_size):
1940
block = bytes[pos:pos+chunk_size]
1941
if report_activity is not None:
1942
report_activity(len(block), 'write')
1943
until_no_eintr(socket.sendall, block)
1946
def dereference_path(path):
1947
"""Determine the real path to a file.
1949
All parent elements are dereferenced. But the file itself is not
1951
:param path: The original path. May be absolute or relative.
1952
:return: the real path *to* the file
1954
parent, base = os.path.split(path)
1955
# The pathjoin for '.' is a workaround for Python bug #1213894.
1956
# (initial path components aren't dereferenced)
1957
return pathjoin(realpath(pathjoin('.', parent)), base)
1960
def supports_mapi():
1961
"""Return True if we can use MAPI to launch a mail client."""
1962
return sys.platform == "win32"
1965
def resource_string(package, resource_name):
1966
"""Load a resource from a package and return it as a string.
1968
Note: Only packages that start with bzrlib are currently supported.
1970
This is designed to be a lightweight implementation of resource
1971
loading in a way which is API compatible with the same API from
1973
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1974
If and when pkg_resources becomes a standard library, this routine
1977
# Check package name is within bzrlib
1978
if package == "bzrlib":
1979
resource_relpath = resource_name
1980
elif package.startswith("bzrlib."):
1981
package = package[len("bzrlib."):].replace('.', os.sep)
1982
resource_relpath = pathjoin(package, resource_name)
1984
raise errors.BzrError('resource package %s not in bzrlib' % package)
1986
# Map the resource to a file and read its contents
1987
base = dirname(bzrlib.__file__)
1988
if getattr(sys, 'frozen', None): # bzr.exe
1989
base = abspath(pathjoin(base, '..', '..'))
1990
filename = pathjoin(base, resource_relpath)
1991
return open(filename, 'rU').read()
1994
def file_kind_from_stat_mode_thunk(mode):
1995
global file_kind_from_stat_mode
1996
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
1998
from bzrlib._readdir_pyx import UTF8DirReader
1999
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2000
except ImportError, e:
2001
# This is one time where we won't warn that an extension failed to
2002
# load. The extension is never available on Windows anyway.
2003
from bzrlib._readdir_py import (
2004
_kind_from_mode as file_kind_from_stat_mode
2006
return file_kind_from_stat_mode(mode)
2007
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2010
def file_kind(f, _lstat=os.lstat):
2012
return file_kind_from_stat_mode(_lstat(f).st_mode)
2014
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2015
raise errors.NoSuchFile(f)
2019
def until_no_eintr(f, *a, **kw):
2020
"""Run f(*a, **kw), retrying if an EINTR error occurs."""
2021
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2025
except (IOError, OSError), e:
2026
if e.errno == errno.EINTR:
2030
def re_compile_checked(re_string, flags=0, where=""):
2031
"""Return a compiled re, or raise a sensible error.
2033
This should only be used when compiling user-supplied REs.
2035
:param re_string: Text form of regular expression.
2036
:param flags: eg re.IGNORECASE
2037
:param where: Message explaining to the user the context where
2038
it occurred, eg 'log search filter'.
2040
# from https://bugs.launchpad.net/bzr/+bug/251352
2042
re_obj = re.compile(re_string, flags)
2047
where = ' in ' + where
2048
# despite the name 'error' is a type
2049
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2050
% (where, re_string, e))
2053
if sys.platform == "win32":
2056
return msvcrt.getch()
2061
fd = sys.stdin.fileno()
2062
settings = termios.tcgetattr(fd)
2065
ch = sys.stdin.read(1)
2067
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2071
if sys.platform == 'linux2':
2072
def _local_concurrency():
2074
prefix = 'processor'
2075
for line in file('/proc/cpuinfo', 'rb'):
2076
if line.startswith(prefix):
2077
concurrency = int(line[line.find(':')+1:]) + 1
2079
elif sys.platform == 'darwin':
2080
def _local_concurrency():
2081
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2082
stdout=subprocess.PIPE).communicate()[0]
2083
elif sys.platform[0:7] == 'freebsd':
2084
def _local_concurrency():
2085
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2086
stdout=subprocess.PIPE).communicate()[0]
2087
elif sys.platform == 'sunos5':
2088
def _local_concurrency():
2089
return subprocess.Popen(['psrinfo', '-p',],
2090
stdout=subprocess.PIPE).communicate()[0]
2091
elif sys.platform == "win32":
2092
def _local_concurrency():
2093
# This appears to return the number of cores.
2094
return os.environ.get('NUMBER_OF_PROCESSORS')
2096
def _local_concurrency():
2101
_cached_local_concurrency = None
2103
def local_concurrency(use_cache=True):
2104
"""Return how many processes can be run concurrently.
2106
Rely on platform specific implementations and default to 1 (one) if
2107
anything goes wrong.
2109
global _cached_local_concurrency
2111
if _cached_local_concurrency is not None and use_cache:
2112
return _cached_local_concurrency
2114
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2115
if concurrency is None:
2117
concurrency = _local_concurrency()
2118
except (OSError, IOError):
2121
concurrency = int(concurrency)
2122
except (TypeError, ValueError):
2125
_cached_concurrency = concurrency
2129
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2130
"""A stream writer that doesn't decode str arguments."""
2132
def __init__(self, encode, stream, errors='strict'):
2133
codecs.StreamWriter.__init__(self, stream, errors)
2134
self.encode = encode
2136
def write(self, object):
2137
if type(object) is str:
2138
self.stream.write(object)
2140
data, _ = self.encode(object, self.errors)
2141
self.stream.write(data)
2143
if sys.platform == 'win32':
2144
def open_file(filename, mode='r', bufsize=-1):
2145
"""This function is used to override the ``open`` builtin.
2147
But it uses O_NOINHERIT flag so the file handle is not inherited by
2148
child processes. Deleting or renaming a closed file opened with this
2149
function is not blocking child processes.
2151
writing = 'w' in mode
2152
appending = 'a' in mode
2153
updating = '+' in mode
2154
binary = 'b' in mode
2157
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2158
# for flags for each modes.
2168
flags |= os.O_WRONLY
2169
flags |= os.O_CREAT | os.O_TRUNC
2174
flags |= os.O_WRONLY
2175
flags |= os.O_CREAT | os.O_APPEND
2180
flags |= os.O_RDONLY
2182
return os.fdopen(os.open(filename, flags), mode, bufsize)
added
removed
bzrlib/osutils.py
