1
# (C) 2005 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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
# TODO: Up-front, stat all files in order and remove those which are deleted or
18
# out-of-date. Don't actually re-read them until they're needed. That ought
19
# to bring all the inodes into core so that future stats to them are fast, and
20
# it preserves the nice property that any caller will always get up-to-date
21
# data except in unavoidable cases.
23
# TODO: Perhaps return more details on the file to avoid statting it
24
# again: nonexistent, file type, size, etc
26
# TODO: Perhaps use a Python pickle instead of a text file; might be faster.
30
CACHE_HEADER = "### bzr hashcache v5\n"
35
from bzrlib.osutils import sha_file, pathjoin, safe_unicode
36
from bzrlib.trace import mutter, warning
37
from bzrlib.atomicfile import AtomicFile
38
from bzrlib.errors import BzrError
45
def _fingerprint(abspath):
47
fs = os.lstat(abspath)
49
# might be missing, etc
52
if stat.S_ISDIR(fs.st_mode):
55
# we discard any high precision because it's not reliable; perhaps we
56
# could do better on some systems?
57
return (fs.st_size, long(fs.st_mtime),
58
long(fs.st_ctime), fs.st_ino, fs.st_dev, fs.st_mode)
61
class HashCache(object):
62
"""Cache for looking up file SHA-1.
64
Files are considered to match the cached value if the fingerprint
65
of the file has not changed. This includes its mtime, ctime,
66
device number, inode number, and size. This should catch
67
modifications or replacement of the file by a new one.
69
This may not catch modifications that do not change the file's
70
size and that occur within the resolution window of the
71
timestamps. To handle this we specifically do not cache files
72
which have changed since the start of the present second, since
73
they could undetectably change again.
75
This scheme may fail if the machine's clock steps backwards.
78
This does not canonicalize the paths passed in; that should be
82
Indexed by path, points to a two-tuple of the SHA-1 of the file.
86
number of times files have been statted
89
number of times files have been retrieved from the cache, avoiding a
93
number of misses (times files have been completely re-read)
97
def __init__(self, root, cache_file_name, mode=None):
98
"""Create a hash cache in base dir, and set the file mode to mode."""
99
self.root = safe_unicode(root)
103
self.danger_count = 0
104
self.removed_count = 0
105
self.update_count = 0
108
self._cache_file_name = safe_unicode(cache_file_name)
110
def cache_file_name(self):
111
return self._cache_file_name
114
"""Discard all cached information.
116
This does not reset the counters."""
118
self.needs_write = True
122
"""Scan all files and remove entries where the cache entry is obsolete.
124
Obsolete entries are those where the file has been modified or deleted
125
since the entry was inserted.
127
# FIXME optimisation opportunity, on linux [and check other oses]:
128
# rather than iteritems order, stat in inode order.
129
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
132
for inum, path, cache_entry in prep:
133
abspath = pathjoin(self.root, path)
134
fp = _fingerprint(abspath)
137
cache_fp = cache_entry[1]
139
if (not fp) or (cache_fp != fp):
140
# not here or not a regular file anymore
141
self.removed_count += 1
142
self.needs_write = True
143
del self._cache[path]
146
def get_sha1(self, path):
147
"""Return the sha1 of a file.
149
abspath = pathjoin(self.root, path)
151
file_fp = _fingerprint(abspath)
154
# not a regular file or not existing
155
if path in self._cache:
156
self.removed_count += 1
157
self.needs_write = True
158
del self._cache[path]
161
if path in self._cache:
162
cache_sha1, cache_fp = self._cache[path]
164
cache_sha1, cache_fp = None, None
166
if cache_fp == file_fp:
173
mode = file_fp[FP_MODE_COLUMN]
174
if stat.S_ISREG(mode):
175
digest = sha_file(file(abspath, 'rb', buffering=65000))
176
elif stat.S_ISLNK(mode):
177
digest = sha.new(os.readlink(abspath)).hexdigest()
179
raise BzrError("file %r: unknown file stat mode: %o"%(abspath,mode))
181
now = int(time.time())
182
if file_fp[FP_MTIME_COLUMN] >= now or file_fp[FP_CTIME_COLUMN] >= now:
183
# changed too recently; can't be cached. we can
184
# return the result and it could possibly be cached
187
# the point is that we only want to cache when we are sure that any
188
# subsequent modifications of the file can be detected. If a
189
# modification neither changes the inode, the device, the size, nor
190
# the mode, then we can only distinguish it by time; therefore we
191
# need to let sufficient time elapse before we may cache this entry
192
# again. If we didn't do this, then, for example, a very quick 1
193
# byte replacement in the file might go undetected.
194
self.danger_count += 1
196
self.removed_count += 1
197
self.needs_write = True
198
del self._cache[path]
200
self.update_count += 1
201
self.needs_write = True
202
self._cache[path] = (digest, file_fp)
206
"""Write contents of cache to file."""
207
outf = AtomicFile(self.cache_file_name(), 'wb', new_mode=self._mode)
209
print >>outf, CACHE_HEADER,
211
for path, c in self._cache.iteritems():
212
assert '//' not in path, path
213
outf.write(path.encode('utf-8'))
215
print >>outf, c[0], # hex sha1
217
print >>outf, "%d" % fld,
221
self.needs_write = False
227
"""Reinstate cache from file.
229
Overwrites existing cache.
231
If the cache file has the wrong version marker, this just clears
235
fn = self.cache_file_name()
237
inf = file(fn, 'rb', buffering=65000)
239
mutter("failed to open %s: %s", fn, e)
240
# better write it now so it is valid
241
self.needs_write = True
246
if hdr != CACHE_HEADER:
247
mutter('cache header marker not found at top of %s;'
248
' discarding cache', fn)
249
self.needs_write = True
254
path = l[:pos].decode('utf-8')
255
if path in self._cache:
256
warning('duplicated path %r in cache' % path)
260
fields = l[pos:].split(' ')
262
warning("bad line in hashcache: %r" % l)
267
warning("bad sha1 in hashcache: %r" % sha1)
270
fp = tuple(map(long, fields[1:]))
272
self._cache[path] = (sha1, fp)
274
self.needs_write = False
added
removed
bzrlib/hashcache.py
