1
# Copyright (C) 2005, 2006 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# out-of-date.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
47
class HashCache(object):
48
"""Cache for looking up file SHA-1.
50
Files are considered to match the cached value if the fingerprint
51
of the file has not changed. This includes its mtime, ctime,
52
device number, inode number, and size. This should catch
53
modifications or replacement of the file by a new one.
55
This may not catch modifications that do not change the file's
56
size and that occur within the resolution window of the
57
timestamps. To handle this we specifically do not cache files
58
which have changed since the start of the present second, since
59
they could undetectably change again.
61
This scheme may fail if the machine's clock steps backwards.
64
This does not canonicalize the paths passed in; that should be
68
Indexed by path, points to a two-tuple of the SHA-1 of the file.
72
number of times files have been statted
75
number of times files have been retrieved from the cache, avoiding a
79
number of misses (times files have been completely re-read)
83
def __init__(self, root, cache_file_name, mode=None):
84
"""Create a hash cache in base dir, and set the file mode to mode."""
85
self.root = safe_unicode(root)
86
self.root_utf8 = self.root.encode('utf8') # where is the filesystem encoding ?
91
self.removed_count = 0
95
self._cache_file_name = safe_unicode(cache_file_name)
97
def cache_file_name(self):
98
return self._cache_file_name
101
"""Discard all cached information.
103
This does not reset the counters."""
105
self.needs_write = True
109
"""Scan all files and remove entries where the cache entry is obsolete.
111
Obsolete entries are those where the file has been modified or deleted
112
since the entry was inserted.
114
# FIXME optimisation opportunity, on linux [and check other oses]:
115
# rather than iteritems order, stat in inode order.
116
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
119
for inum, path, cache_entry in prep:
120
abspath = pathjoin(self.root, path)
121
fp = self._fingerprint(abspath)
124
cache_fp = cache_entry[1]
126
if (not fp) or (cache_fp != fp):
127
# not here or not a regular file anymore
128
self.removed_count += 1
129
self.needs_write = True
130
del self._cache[path]
132
def get_sha1(self, path, stat_value=None):
133
"""Return the sha1 of a file.
135
if path.__class__ is str:
136
abspath = pathjoin(self.root_utf8, path)
138
abspath = pathjoin(self.root, path)
140
file_fp = self._fingerprint(abspath, stat_value)
143
# not a regular file or not existing
144
if path in self._cache:
145
self.removed_count += 1
146
self.needs_write = True
147
del self._cache[path]
150
if path in self._cache:
151
cache_sha1, cache_fp = self._cache[path]
153
cache_sha1, cache_fp = None, None
155
if cache_fp == file_fp:
156
## mutter("hashcache hit for %s %r -> %s", path, file_fp, cache_sha1)
157
## mutter("now = %s", time.time())
163
mode = file_fp[FP_MODE_COLUMN]
164
if stat.S_ISREG(mode):
165
digest = self._really_sha1_file(abspath)
166
elif stat.S_ISLNK(mode):
167
digest = sha.new(os.readlink(abspath)).hexdigest()
169
raise BzrError("file %r: unknown file stat mode: %o"%(abspath,mode))
171
# window of 3 seconds to allow for 2s resolution on windows,
172
# unsynchronized file servers, etc.
173
cutoff = self._cutoff_time()
174
if file_fp[FP_MTIME_COLUMN] >= cutoff \
175
or file_fp[FP_CTIME_COLUMN] >= cutoff:
176
# changed too recently; can't be cached. we can
177
# return the result and it could possibly be cached
180
# the point is that we only want to cache when we are sure that any
181
# subsequent modifications of the file can be detected. If a
182
# modification neither changes the inode, the device, the size, nor
183
# the mode, then we can only distinguish it by time; therefore we
184
# need to let sufficient time elapse before we may cache this entry
185
# again. If we didn't do this, then, for example, a very quick 1
186
# byte replacement in the file might go undetected.
187
## mutter('%r modified too recently; not caching', path)
188
self.danger_count += 1
190
self.removed_count += 1
191
self.needs_write = True
192
del self._cache[path]
194
## mutter('%r added to cache: now=%f, mtime=%d, ctime=%d',
195
## path, time.time(), file_fp[FP_MTIME_COLUMN],
196
## file_fp[FP_CTIME_COLUMN])
197
self.update_count += 1
198
self.needs_write = True
199
self._cache[path] = (digest, file_fp)
202
def _really_sha1_file(self, abspath):
203
"""Calculate the SHA1 of a file by reading the full text"""
204
return sha_file(file(abspath, 'rb', buffering=65000))
207
"""Write contents of cache to file."""
208
outf = AtomicFile(self.cache_file_name(), 'wb', new_mode=self._mode)
210
outf.write(CACHE_HEADER)
212
for path, c in self._cache.iteritems():
213
assert '//' not in path, path
214
line_info = [path.encode('utf-8'), '// ', c[0], ' ']
215
line_info.append(' '.join([str(fld) for fld in c[1]]))
216
line_info.append('\n')
217
outf.write(''.join(line_info))
219
self.needs_write = False
220
## mutter("write hash cache: %s hits=%d misses=%d stat=%d recent=%d updates=%d",
221
## self.cache_file_name(), self.hit_count, self.miss_count,
223
## self.danger_count, self.update_count)
228
"""Reinstate cache from file.
230
Overwrites existing cache.
232
If the cache file has the wrong version marker, this just clears
236
fn = self.cache_file_name()
238
inf = file(fn, 'rb', buffering=65000)
240
mutter("failed to open %s: %s", fn, e)
241
# better write it now so it is valid
242
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
276
def _cutoff_time(self):
277
"""Return cutoff time.
279
Files modified more recently than this time are at risk of being
280
undetectably modified and so can't be cached.
282
return int(time.time()) - 3
284
def _fingerprint(self, abspath, stat_value=None):
285
if stat_value is None:
287
stat_value = os.lstat(abspath)
289
# might be missing, etc
291
if stat.S_ISDIR(stat_value.st_mode):
293
# we discard any high precision because it's not reliable; perhaps we
294
# could do better on some systems?
295
return (stat_value.st_size, long(stat_value.st_mtime),
296
long(stat_value.st_ctime), stat_value.st_ino,
297
stat_value.st_dev, stat_value.st_mode)
added
removed
bzrlib/hashcache.py
