89
70
O_BINARY = getattr(os, 'O_BINARY', 0)
92
def get_unicode_argv():
94
user_encoding = get_user_encoding()
95
return [a.decode(user_encoding) for a in sys.argv[1:]]
96
except UnicodeDecodeError:
97
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
101
73
def make_readonly(filename):
102
74
"""Make a filename read-only."""
103
mod = os.lstat(filename).st_mode
104
if not stat.S_ISLNK(mod):
106
os.chmod(filename, mod)
75
mod = os.stat(filename).st_mode
77
os.chmod(filename, mod)
109
80
def make_writable(filename):
110
mod = os.lstat(filename).st_mode
111
if not stat.S_ISLNK(mod):
113
os.chmod(filename, mod)
116
def minimum_path_selection(paths):
117
"""Return the smallset subset of paths which are outside paths.
119
:param paths: A container (and hence not None) of paths.
120
:return: A set of paths sufficient to include everything in paths via
121
is_inside, drawn from the paths parameter.
127
return path.split('/')
128
sorted_paths = sorted(list(paths), key=sort_key)
130
search_paths = [sorted_paths[0]]
131
for path in sorted_paths[1:]:
132
if not is_inside(search_paths[-1], path):
133
# This path is unique, add it
134
search_paths.append(path)
136
return set(search_paths)
81
mod = os.stat(filename).st_mode
83
os.chmod(filename, mod)
201
177
def fancy_rename(old, new, rename_func, unlink_func):
202
178
"""A fancy rename, when you don't have atomic rename.
204
180
:param old: The old path, to rename from
205
181
:param new: The new path, to rename to
206
182
:param rename_func: The potentially non-atomic rename function
207
:param unlink_func: A way to delete the target file if the full rename
183
:param unlink_func: A way to delete the target file if the full rename succeeds
210
186
# sftp rename doesn't allow overwriting, so play tricks:
211
188
base = os.path.basename(new)
212
189
dirname = os.path.dirname(new)
213
# callers use different encodings for the paths so the following MUST
214
# respect that. We rely on python upcasting to unicode if new is unicode
215
# and keeping a str if not.
216
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
217
os.getpid(), rand_chars(10))
190
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
218
191
tmp_name = pathjoin(dirname, tmp_name)
220
193
# Rename the file out of the way, but keep track if it didn't exist
556
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
557
report_activity=None, direction='read'):
558
"""Copy contents of one file to another.
560
The read_length can either be -1 to read to end-of-file (EOF) or
561
it can specify the maximum number of bytes to read.
563
The buff_size represents the maximum size for each read operation
564
performed on from_file.
566
:param report_activity: Call this as bytes are read, see
567
Transport._report_activity
568
:param direction: Will be passed to report_activity
570
:return: The number of bytes copied.
574
# read specified number of bytes
576
while read_length > 0:
577
num_bytes_to_read = min(read_length, buff_size)
579
block = from_file.read(num_bytes_to_read)
583
if report_activity is not None:
584
report_activity(len(block), direction)
587
actual_bytes_read = len(block)
588
read_length -= actual_bytes_read
589
length += actual_bytes_read
593
block = from_file.read(buff_size)
597
if report_activity is not None:
598
report_activity(len(block), direction)
604
def pump_string_file(bytes, file_handle, segment_size=None):
605
"""Write bytes to file_handle in many smaller writes.
607
:param bytes: The string to write.
608
:param file_handle: The file to write to.
610
# Write data in chunks rather than all at once, because very large
611
# writes fail on some platforms (e.g. Windows with SMB mounted
614
segment_size = 5242880 # 5MB
615
segments = range(len(bytes) / segment_size + 1)
616
write = file_handle.write
617
for segment_index in segments:
618
segment = buffer(bytes, segment_index * segment_size, segment_size)
535
def pumpfile(fromfile, tofile):
536
"""Copy contents of one file to another."""
539
b = fromfile.read(BUFSIZE)
622
545
def file_iterator(input_file, readsize=32768):
723
615
:param timezone: How to display the time: 'utc', 'original' for the
724
616
timezone specified by offset, or 'local' for the process's current
726
:param date_fmt: strftime format.
727
:param show_offset: Whether to append the timezone.
729
(date_fmt, tt, offset_str) = \
730
_format_date(t, offset, timezone, date_fmt, show_offset)
731
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
732
date_str = time.strftime(date_fmt, tt)
733
return date_str + offset_str
736
# Cache of formatted offset strings
740
def format_date_with_offset_in_original_timezone(t, offset=0,
741
_cache=_offset_cache):
742
"""Return a formatted date string in the original timezone.
744
This routine may be faster then format_date.
746
:param t: Seconds since the epoch.
747
:param offset: Timezone offset in seconds east of utc.
751
tt = time.gmtime(t + offset)
752
date_fmt = _default_format_by_weekday_num[tt[6]]
753
date_str = time.strftime(date_fmt, tt)
754
offset_str = _cache.get(offset, None)
755
if offset_str is None:
756
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
757
_cache[offset] = offset_str
758
return date_str + offset_str
761
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
763
"""Return an unicode date string formatted according to the current locale.
765
:param t: Seconds since the epoch.
766
:param offset: Timezone offset in seconds east of utc.
767
:param timezone: How to display the time: 'utc', 'original' for the
768
timezone specified by offset, or 'local' for the process's current
770
:param date_fmt: strftime format.
771
:param show_offset: Whether to append the timezone.
773
(date_fmt, tt, offset_str) = \
774
_format_date(t, offset, timezone, date_fmt, show_offset)
775
date_str = time.strftime(date_fmt, tt)
776
if not isinstance(date_str, unicode):
777
date_str = date_str.decode(get_user_encoding(), 'replace')
778
return date_str + offset_str
781
def _format_date(t, offset, timezone, date_fmt, show_offset):
618
:param show_offset: Whether to append the timezone.
619
:param date_fmt: strftime format.
782
621
if timezone == 'utc':
783
622
tt = time.gmtime(t)
759
assert isinstance(p, (list, tuple))
919
761
if (f == '..') or (f is None) or (f == ''):
920
762
raise errors.BzrError("sorry, %r not allowed in path" % f)
921
763
return pathjoin(*p)
924
def parent_directories(filename):
925
"""Return the list of parent directories, deepest first.
927
For example, parent_directories("a/b/c") -> ["a/b", "a"].
930
parts = splitpath(dirname(filename))
932
parents.append(joinpath(parts))
937
_extension_load_failures = []
940
def failed_to_load_extension(exception):
941
"""Handle failing to load a binary extension.
943
This should be called from the ImportError block guarding the attempt to
944
import the native extension. If this function returns, the pure-Python
945
implementation should be loaded instead::
948
>>> import bzrlib._fictional_extension_pyx
949
>>> except ImportError, e:
950
>>> bzrlib.osutils.failed_to_load_extension(e)
951
>>> import bzrlib._fictional_extension_py
953
# NB: This docstring is just an example, not a doctest, because doctest
954
# currently can't cope with the use of lazy imports in this namespace --
957
# This currently doesn't report the failure at the time it occurs, because
958
# they tend to happen very early in startup when we can't check config
959
# files etc, and also we want to report all failures but not spam the user
961
from bzrlib import trace
962
exception_str = str(exception)
963
if exception_str not in _extension_load_failures:
964
trace.mutter("failed to load compiled extension: %s" % exception_str)
965
_extension_load_failures.append(exception_str)
968
def report_extension_load_failures():
969
if not _extension_load_failures:
971
from bzrlib.config import GlobalConfig
972
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
974
# the warnings framework should by default show this only once
975
from bzrlib.trace import warning
977
"bzr: warning: some compiled extensions could not be loaded; "
978
"see <https://answers.launchpad.net/bzr/+faq/703>")
979
# we no longer show the specific missing extensions here, because it makes
980
# the message too long and scary - see
981
# https://bugs.launchpad.net/bzr/+bug/430529
985
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
986
except ImportError, e:
987
failed_to_load_extension(e)
988
from bzrlib._chunks_to_lines_py import chunks_to_lines
766
@deprecated_function(zero_nine)
767
def appendpath(p1, p2):
771
return pathjoin(p1, p2)
991
774
def split_lines(s):
992
775
"""Split s into lines, but without removing the newline characters."""
993
# Trivially convert a fulltext into a 'chunked' representation, and let
994
# chunks_to_lines do the heavy lifting.
995
if isinstance(s, str):
996
# chunks_to_lines only supports 8-bit strings
997
return chunks_to_lines([s])
999
return _split_lines(s)
1002
def _split_lines(s):
1003
"""Split s into lines, but without removing the newline characters.
1005
This supports Unicode or plain string objects.
1007
776
lines = s.split('\n')
1008
777
result = [line + '\n' for line in lines[:-1]]
1028
797
shutil.copyfile(src, dest)
1031
def delete_any(path):
1032
"""Delete a file, symlink or directory.
1034
Will delete even if readonly.
799
def delete_any(full_path):
800
"""Delete a file or directory."""
1037
_delete_file_or_dir(path)
1038
except (OSError, IOError), e:
1039
if e.errno in (errno.EPERM, errno.EACCES):
1040
# make writable and try again
1043
except (OSError, IOError):
1045
_delete_file_or_dir(path)
804
# We may be renaming a dangling inventory id
805
if e.errno not in (errno.EISDIR, errno.EACCES, errno.EPERM):
1050
def _delete_file_or_dir(path):
1051
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1052
# Forgiveness than Permission (EAFP) because:
1053
# - root can damage a solaris file system by using unlink,
1054
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1055
# EACCES, OSX: EPERM) when invoked on a directory.
1056
if isdir(path): # Takes care of symlinks
1062
810
def has_symlinks():
1132
854
avoids that problem.
1135
if len(base) < MIN_ABS_PATHLENGTH:
1136
# must have space for e.g. a drive letter
1137
raise ValueError('%r is too short to calculate a relative path'
857
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
858
' exceed the platform minimum length (which is %d)' %
1140
861
rp = abspath(path)
1145
if len(head) <= len(base) and head != base:
1146
raise errors.PathNotChild(rp, base)
865
while len(head) >= len(base):
1147
866
if head == base:
1149
head, tail = split(head)
868
head, tail = os.path.split(head)
872
raise errors.PathNotChild(rp, base)
1154
return pathjoin(*reversed(s))
1159
def _cicp_canonical_relpath(base, path):
1160
"""Return the canonical path relative to base.
1162
Like relpath, but on case-insensitive-case-preserving file-systems, this
1163
will return the relpath as stored on the file-system rather than in the
1164
case specified in the input string, for all existing portions of the path.
1166
This will cause O(N) behaviour if called for every path in a tree; if you
1167
have a number of paths to convert, you should use canonical_relpaths().
1169
# TODO: it should be possible to optimize this for Windows by using the
1170
# win32 API FindFiles function to look for the specified name - but using
1171
# os.listdir() still gives us the correct, platform agnostic semantics in
1174
rel = relpath(base, path)
1175
# '.' will have been turned into ''
1179
abs_base = abspath(base)
1181
_listdir = os.listdir
1183
# use an explicit iterator so we can easily consume the rest on early exit.
1184
bit_iter = iter(rel.split('/'))
1185
for bit in bit_iter:
1188
next_entries = _listdir(current)
1189
except OSError: # enoent, eperm, etc
1190
# We can't find this in the filesystem, so just append the
1192
current = pathjoin(current, bit, *list(bit_iter))
1194
for look in next_entries:
1195
if lbit == look.lower():
1196
current = pathjoin(current, look)
1199
# got to the end, nothing matched, so we just return the
1200
# non-existing bits as they were specified (the filename may be
1201
# the target of a move, for example).
1202
current = pathjoin(current, bit, *list(bit_iter))
1204
return current[len(abs_base):].lstrip('/')
1206
# XXX - TODO - we need better detection/integration of case-insensitive
1207
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1208
# filesystems), for example, so could probably benefit from the same basic
1209
# support there. For now though, only Windows and OSX get that support, and
1210
# they get it for *all* file-systems!
1211
if sys.platform in ('win32', 'darwin'):
1212
canonical_relpath = _cicp_canonical_relpath
1214
canonical_relpath = relpath
1216
def canonical_relpaths(base, paths):
1217
"""Create an iterable to canonicalize a sequence of relative paths.
1219
The intent is for this implementation to use a cache, vastly speeding
1220
up multiple transformations in the same directory.
1222
# but for now, we haven't optimized...
1223
return [canonical_relpath(base, p) for p in paths]
1225
880
def safe_unicode(unicode_or_utf8_string):
1226
881
"""Coerce unicode_or_utf8_string into unicode.
1228
883
If it is unicode, it is returned.
1229
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1230
wrapped in a BzrBadParameterNotUnicode exception.
884
Otherwise it is decoded from utf-8. If a decoding error
885
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
886
as a BzrBadParameter exception.
1232
888
if isinstance(unicode_or_utf8_string, unicode):
1233
889
return unicode_or_utf8_string
1321
977
On platforms where the system normalizes filenames (Mac OSX),
1322
978
you can access a file by any path which will normalize correctly.
1323
On platforms where the system does not normalize filenames
979
On platforms where the system does not normalize filenames
1324
980
(Windows, Linux), you have to access a file by its exact path.
1326
Internally, bzr only supports NFC normalization, since that is
982
Internally, bzr only supports NFC/NFKC normalization, since that is
1327
983
the standard for XML documents.
1329
985
So return the normalized path, and a flag indicating if the file
1330
986
can be accessed by that path.
1333
return unicodedata.normalize('NFC', unicode(path)), True
989
return unicodedata.normalize('NFKC', unicode(path)), True
1336
992
def _inaccessible_normalized_filename(path):
1337
993
__doc__ = _accessible_normalized_filename.__doc__
1339
normalized = unicodedata.normalize('NFC', unicode(path))
995
normalized = unicodedata.normalize('NFKC', unicode(path))
1340
996
return normalized, normalized == path
1346
1002
normalized_filename = _inaccessible_normalized_filename
1349
default_terminal_width = 80
1350
"""The default terminal width for ttys.
1352
This is defined so that higher levels can share a common fallback value when
1353
terminal_width() returns None.
1357
1005
def terminal_width():
1358
"""Return terminal width.
1360
None is returned if the width can't established precisely.
1363
- if BZR_COLUMNS is set, returns its value
1364
- if there is no controlling terminal, returns None
1365
- if COLUMNS is set, returns its value,
1367
From there, we need to query the OS to get the size of the controlling
1371
- get termios.TIOCGWINSZ
1372
- if an error occurs or a negative value is obtained, returns None
1376
- win32utils.get_console_size() decides,
1377
- returns None on error (provided default value)
1380
# If BZR_COLUMNS is set, take it, user is always right
1382
return int(os.environ['BZR_COLUMNS'])
1383
except (KeyError, ValueError):
1386
isatty = getattr(sys.stdout, 'isatty', None)
1387
if isatty is None or not isatty():
1388
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1391
# If COLUMNS is set, take it, the terminal knows better (even inside a
1392
# given terminal, the application can decide to set COLUMNS to a lower
1393
# value (splitted screen) or a bigger value (scroll bars))
1395
return int(os.environ['COLUMNS'])
1396
except (KeyError, ValueError):
1399
width, height = _terminal_size(None, None)
1401
# Consider invalid values as meaning no width
1407
def _win32_terminal_size(width, height):
1408
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1409
return width, height
1412
def _ioctl_terminal_size(width, height):
1006
"""Return estimated terminal width."""
1007
if sys.platform == 'win32':
1008
return win32utils.get_console_size()[0]
1414
1011
import struct, fcntl, termios
1415
1012
s = struct.pack('HHHH', 0, 0, 0, 0)
1416
1013
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1417
height, width = struct.unpack('HHHH', x)[0:2]
1418
except (IOError, AttributeError):
1014
width = struct.unpack('HHHH', x)[1]
1420
return width, height
1422
_terminal_size = None
1423
"""Returns the terminal size as (width, height).
1425
:param width: Default value for width.
1426
:param height: Default value for height.
1428
This is defined specifically for each OS and query the size of the controlling
1429
terminal. If any error occurs, the provided default values should be returned.
1431
if sys.platform == 'win32':
1432
_terminal_size = _win32_terminal_size
1434
_terminal_size = _ioctl_terminal_size
1437
def _terminal_size_changed(signum, frame):
1438
"""Set COLUMNS upon receiving a SIGnal for WINdow size CHange."""
1439
width, height = _terminal_size(None, None)
1440
if width is not None:
1441
os.environ['COLUMNS'] = str(width)
1444
_registered_sigwinch = False
1446
def watch_sigwinch():
1447
"""Register for SIGWINCH, once and only once."""
1448
global _registered_sigwinch
1449
if not _registered_sigwinch:
1450
if sys.platform == 'win32':
1451
# Martin (gz) mentioned WINDOW_BUFFER_SIZE_RECORD from
1452
# ReadConsoleInput but I've no idea how to plug that in
1453
# the current design -- vila 20091216
1019
width = int(os.environ['COLUMNS'])
1456
signal.signal(signal.SIGWINCH, _terminal_size_changed)
1457
_registered_sigwinch = True
1460
1028
def supports_executable():
1507
1075
raise errors.IllegalPath(path)
1510
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1512
def _is_error_enotdir(e):
1513
"""Check if this exception represents ENOTDIR.
1515
Unfortunately, python is very inconsistent about the exception
1516
here. The cases are:
1517
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1518
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1519
which is the windows error code.
1520
3) Windows, Python2.5 uses errno == EINVAL and
1521
winerror == ERROR_DIRECTORY
1523
:param e: An Exception object (expected to be OSError with an errno
1524
attribute, but we should be able to cope with anything)
1525
:return: True if this represents an ENOTDIR error. False otherwise.
1527
en = getattr(e, 'errno', None)
1528
if (en == errno.ENOTDIR
1529
or (sys.platform == 'win32'
1530
and (en == _WIN32_ERROR_DIRECTORY
1531
or (en == errno.EINVAL
1532
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1538
1078
def walkdirs(top, prefix=""):
1539
1079
"""Yield data about all the directories in a tree.
1541
1081
This yields all the data about the contents of a directory at a time.
1542
1082
After each directory has been yielded, if the caller has mutated the list
1543
1083
to exclude some directories, they are then not descended into.
1545
1085
The data yielded is of the form:
1546
1086
((directory-relpath, directory-path-from-top),
1547
[(relpath, basename, kind, lstat, path-from-top), ...]),
1087
[(directory-relpath, basename, kind, lstat, path-from-top), ...]),
1548
1088
- directory-relpath is the relative path of the directory being returned
1549
1089
with respect to top. prefix is prepended to this.
1550
- directory-path-from-root is the path including top for this directory.
1090
- directory-path-from-root is the path including top for this directory.
1551
1091
It is suitable for use with os functions.
1552
1092
- relpath is the relative path within the subtree being walked.
1553
1093
- basename is the basename of the path
1555
1095
present within the tree - but it may be recorded as versioned. See
1556
1096
versioned_kind.
1557
1097
- lstat is the stat data *if* the file was statted.
1558
- planned, not implemented:
1098
- planned, not implemented:
1559
1099
path_from_tree_root is the path from the root of the tree.
1561
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1101
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1562
1102
allows one to walk a subtree but get paths that are relative to a tree
1563
1103
rooted higher up.
1564
1104
:return: an iterator over the dirs.
1566
1106
#TODO there is a bit of a smell where the results of the directory-
1567
# summary in this, and the path from the root, may not agree
1107
# summary in this, and the path from the root, may not agree
1568
1108
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1569
1109
# potentially confusing output. We should make this more robust - but
1570
1110
# not at a speed cost. RBC 20060731
1571
1111
_lstat = os.lstat
1572
1112
_directory = _directory_kind
1573
1113
_listdir = os.listdir
1574
_kind_from_mode = file_kind_from_stat_mode
1114
_kind_from_mode = _formats.get
1575
1115
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1577
1117
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1586
1126
append = dirblock.append
1588
names = sorted(_listdir(top))
1590
if not _is_error_enotdir(e):
1594
abspath = top_slash + name
1595
statvalue = _lstat(abspath)
1596
kind = _kind_from_mode(statvalue.st_mode)
1597
append((relprefix + name, name, kind, statvalue, abspath))
1127
for name in sorted(_listdir(top)):
1128
abspath = top_slash + name
1129
statvalue = _lstat(abspath)
1130
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1131
append((relprefix + name, name, kind, statvalue, abspath))
1598
1132
yield (relroot, top), dirblock
1600
1134
# push the user specified dirs from dirblock
1601
1135
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1604
class DirReader(object):
1605
"""An interface for reading directories."""
1607
def top_prefix_to_starting_dir(self, top, prefix=""):
1608
"""Converts top and prefix to a starting dir entry
1610
:param top: A utf8 path
1611
:param prefix: An optional utf8 path to prefix output relative paths
1613
:return: A tuple starting with prefix, and ending with the native
1616
raise NotImplementedError(self.top_prefix_to_starting_dir)
1618
def read_dir(self, prefix, top):
1619
"""Read a specific dir.
1621
:param prefix: A utf8 prefix to be preprended to the path basenames.
1622
:param top: A natively encoded path to read.
1623
:return: A list of the directories contents. Each item contains:
1624
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1626
raise NotImplementedError(self.read_dir)
1629
_selected_dir_reader = None
1632
1138
def _walkdirs_utf8(top, prefix=""):
1633
1139
"""Yield data about all the directories in a tree.
1643
1149
path-from-top might be unicode or utf8, but it is the correct path to
1644
1150
pass to os functions to affect the file in question. (such as os.lstat)
1646
global _selected_dir_reader
1647
if _selected_dir_reader is None:
1648
fs_encoding = _fs_enc.upper()
1649
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1650
# Win98 doesn't have unicode apis like FindFirstFileW
1651
# TODO: We possibly could support Win98 by falling back to the
1652
# original FindFirstFile, and using TCHAR instead of WCHAR,
1653
# but that gets a bit tricky, and requires custom compiling
1656
from bzrlib._walkdirs_win32 import Win32ReadDir
1657
_selected_dir_reader = Win32ReadDir()
1660
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1661
# ANSI_X3.4-1968 is a form of ASCII
1663
from bzrlib._readdir_pyx import UTF8DirReader
1664
_selected_dir_reader = UTF8DirReader()
1665
except ImportError, e:
1666
failed_to_load_extension(e)
1669
if _selected_dir_reader is None:
1670
# Fallback to the python version
1671
_selected_dir_reader = UnicodeDirReader()
1152
fs_encoding = _fs_enc.upper()
1153
if (sys.platform == 'win32' or
1154
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1155
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1157
return _walkdirs_fs_utf8(top, prefix=prefix)
1160
def _walkdirs_fs_utf8(top, prefix=""):
1161
"""See _walkdirs_utf8.
1163
This sub-function is called when we know the filesystem is already in utf8
1164
encoding. So we don't need to transcode filenames.
1167
_directory = _directory_kind
1168
_listdir = os.listdir
1169
_kind_from_mode = _formats.get
1673
1171
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1674
1172
# But we don't actually uses 1-3 in pending, so set them to None
1675
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1676
read_dir = _selected_dir_reader.read_dir
1677
_directory = _directory_kind
1173
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1679
relroot, _, _, _, top = pending[-1].pop()
1682
dirblock = sorted(read_dir(relroot, top))
1175
relroot, _, _, _, top = pending.pop()
1177
relprefix = relroot + '/'
1180
top_slash = top + '/'
1183
append = dirblock.append
1184
for name in sorted(_listdir(top)):
1185
abspath = top_slash + name
1186
statvalue = _lstat(abspath)
1187
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1188
append((relprefix + name, name, kind, statvalue, abspath))
1683
1189
yield (relroot, top), dirblock
1684
1191
# push the user specified dirs from dirblock
1685
next = [d for d in reversed(dirblock) if d[2] == _directory]
1687
pending.append(next)
1690
class UnicodeDirReader(DirReader):
1691
"""A dir reader for non-utf8 file systems, which transcodes."""
1693
__slots__ = ['_utf8_encode']
1696
self._utf8_encode = codecs.getencoder('utf8')
1698
def top_prefix_to_starting_dir(self, top, prefix=""):
1699
"""See DirReader.top_prefix_to_starting_dir."""
1700
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1702
def read_dir(self, prefix, top):
1703
"""Read a single directory from a non-utf8 file system.
1705
top, and the abspath element in the output are unicode, all other paths
1706
are utf8. Local disk IO is done via unicode calls to listdir etc.
1708
This is currently the fallback code path when the filesystem encoding is
1709
not UTF-8. It may be better to implement an alternative so that we can
1710
safely handle paths that are not properly decodable in the current
1713
See DirReader.read_dir for details.
1715
_utf8_encode = self._utf8_encode
1717
_listdir = os.listdir
1718
_kind_from_mode = file_kind_from_stat_mode
1721
relprefix = prefix + '/'
1192
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1195
def _walkdirs_unicode_to_utf8(top, prefix=""):
1196
"""See _walkdirs_utf8
1198
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1200
This is currently the fallback code path when the filesystem encoding is
1201
not UTF-8. It may be better to implement an alternative so that we can
1202
safely handle paths that are not properly decodable in the current
1205
_utf8_encode = codecs.getencoder('utf8')
1207
_directory = _directory_kind
1208
_listdir = os.listdir
1209
_kind_from_mode = _formats.get
1211
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1213
relroot, _, _, _, top = pending.pop()
1215
relprefix = relroot + '/'
1724
1218
top_slash = top + u'/'
1727
1221
append = dirblock.append
1728
1222
for name in sorted(_listdir(top)):
1730
name_utf8 = _utf8_encode(name)[0]
1731
except UnicodeDecodeError:
1732
raise errors.BadFilenameEncoding(
1733
_utf8_encode(relprefix)[0] + name, _fs_enc)
1223
name_utf8 = _utf8_encode(name)[0]
1734
1224
abspath = top_slash + name
1735
1225
statvalue = _lstat(abspath)
1736
kind = _kind_from_mode(statvalue.st_mode)
1226
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1737
1227
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1228
yield (relroot, top), dirblock
1230
# push the user specified dirs from dirblock
1231
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1741
1234
def copy_tree(from_path, to_path, handlers={}):
1742
1235
"""Copy all of the entries in from_path into to_path.
1744
:param from_path: The base directory to copy.
1237
:param from_path: The base directory to copy.
1745
1238
:param to_path: The target directory. If it does not exist, it will
1747
1240
:param handlers: A dictionary of functions, which takes a source and
1933
1382
# The pathjoin for '.' is a workaround for Python bug #1213894.
1934
1383
# (initial path components aren't dereferenced)
1935
1384
return pathjoin(realpath(pathjoin('.', parent)), base)
1938
def supports_mapi():
1939
"""Return True if we can use MAPI to launch a mail client."""
1940
return sys.platform == "win32"
1943
def resource_string(package, resource_name):
1944
"""Load a resource from a package and return it as a string.
1946
Note: Only packages that start with bzrlib are currently supported.
1948
This is designed to be a lightweight implementation of resource
1949
loading in a way which is API compatible with the same API from
1951
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1952
If and when pkg_resources becomes a standard library, this routine
1955
# Check package name is within bzrlib
1956
if package == "bzrlib":
1957
resource_relpath = resource_name
1958
elif package.startswith("bzrlib."):
1959
package = package[len("bzrlib."):].replace('.', os.sep)
1960
resource_relpath = pathjoin(package, resource_name)
1962
raise errors.BzrError('resource package %s not in bzrlib' % package)
1964
# Map the resource to a file and read its contents
1965
base = dirname(bzrlib.__file__)
1966
if getattr(sys, 'frozen', None): # bzr.exe
1967
base = abspath(pathjoin(base, '..', '..'))
1968
filename = pathjoin(base, resource_relpath)
1969
return open(filename, 'rU').read()
1972
def file_kind_from_stat_mode_thunk(mode):
1973
global file_kind_from_stat_mode
1974
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
1976
from bzrlib._readdir_pyx import UTF8DirReader
1977
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
1978
except ImportError, e:
1979
# This is one time where we won't warn that an extension failed to
1980
# load. The extension is never available on Windows anyway.
1981
from bzrlib._readdir_py import (
1982
_kind_from_mode as file_kind_from_stat_mode
1984
return file_kind_from_stat_mode(mode)
1985
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
1988
def file_kind(f, _lstat=os.lstat):
1990
return file_kind_from_stat_mode(_lstat(f).st_mode)
1992
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
1993
raise errors.NoSuchFile(f)
1997
def until_no_eintr(f, *a, **kw):
1998
"""Run f(*a, **kw), retrying if an EINTR error occurs."""
1999
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2003
except (IOError, OSError), e:
2004
if e.errno == errno.EINTR:
2008
def re_compile_checked(re_string, flags=0, where=""):
2009
"""Return a compiled re, or raise a sensible error.
2011
This should only be used when compiling user-supplied REs.
2013
:param re_string: Text form of regular expression.
2014
:param flags: eg re.IGNORECASE
2015
:param where: Message explaining to the user the context where
2016
it occurred, eg 'log search filter'.
2018
# from https://bugs.launchpad.net/bzr/+bug/251352
2020
re_obj = re.compile(re_string, flags)
2025
where = ' in ' + where
2026
# despite the name 'error' is a type
2027
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2028
% (where, re_string, e))
2031
if sys.platform == "win32":
2034
return msvcrt.getch()
2039
fd = sys.stdin.fileno()
2040
settings = termios.tcgetattr(fd)
2043
ch = sys.stdin.read(1)
2045
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2049
if sys.platform == 'linux2':
2050
def _local_concurrency():
2052
prefix = 'processor'
2053
for line in file('/proc/cpuinfo', 'rb'):
2054
if line.startswith(prefix):
2055
concurrency = int(line[line.find(':')+1:]) + 1
2057
elif sys.platform == 'darwin':
2058
def _local_concurrency():
2059
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2060
stdout=subprocess.PIPE).communicate()[0]
2061
elif sys.platform[0:7] == 'freebsd':
2062
def _local_concurrency():
2063
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2064
stdout=subprocess.PIPE).communicate()[0]
2065
elif sys.platform == 'sunos5':
2066
def _local_concurrency():
2067
return subprocess.Popen(['psrinfo', '-p',],
2068
stdout=subprocess.PIPE).communicate()[0]
2069
elif sys.platform == "win32":
2070
def _local_concurrency():
2071
# This appears to return the number of cores.
2072
return os.environ.get('NUMBER_OF_PROCESSORS')
2074
def _local_concurrency():
2079
_cached_local_concurrency = None
2081
def local_concurrency(use_cache=True):
2082
"""Return how many processes can be run concurrently.
2084
Rely on platform specific implementations and default to 1 (one) if
2085
anything goes wrong.
2087
global _cached_local_concurrency
2089
if _cached_local_concurrency is not None and use_cache:
2090
return _cached_local_concurrency
2092
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2093
if concurrency is None:
2095
concurrency = _local_concurrency()
2096
except (OSError, IOError):
2099
concurrency = int(concurrency)
2100
except (TypeError, ValueError):
2103
_cached_concurrency = concurrency
2107
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2108
"""A stream writer that doesn't decode str arguments."""
2110
def __init__(self, encode, stream, errors='strict'):
2111
codecs.StreamWriter.__init__(self, stream, errors)
2112
self.encode = encode
2114
def write(self, object):
2115
if type(object) is str:
2116
self.stream.write(object)
2118
data, _ = self.encode(object, self.errors)
2119
self.stream.write(data)
added
removed
bzrlib/osutils.py
