1
# Copyright (C) 2005, 2006 by 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# out-of-date.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)
90
self.removed_count = 0
94
self._cache_file_name = safe_unicode(cache_file_name)
96
def cache_file_name(self):
97
return self._cache_file_name
100
"""Discard all cached information.
102
This does not reset the counters."""
104
self.needs_write = True
108
"""Scan all files and remove entries where the cache entry is obsolete.
110
Obsolete entries are those where the file has been modified or deleted
111
since the entry was inserted.
113
# FIXME optimisation opportunity, on linux [and check other oses]:
114
# rather than iteritems order, stat in inode order.
115
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
118
for inum, path, cache_entry in prep:
119
abspath = pathjoin(self.root, path)
120
fp = self._fingerprint(abspath)
123
cache_fp = cache_entry[1]
125
if (not fp) or (cache_fp != fp):
126
# not here or not a regular file anymore
127
self.removed_count += 1
128
self.needs_write = True
129
del self._cache[path]
131
def get_sha1(self, path):
132
"""Return the sha1 of a file.
134
abspath = pathjoin(self.root, path)
136
file_fp = self._fingerprint(abspath)
139
# not a regular file or not existing
140
if path in self._cache:
141
self.removed_count += 1
142
self.needs_write = True
143
del self._cache[path]
146
if path in self._cache:
147
cache_sha1, cache_fp = self._cache[path]
149
cache_sha1, cache_fp = None, None
151
if cache_fp == file_fp:
152
## mutter("hashcache hit for %s %r -> %s", path, file_fp, cache_sha1)
153
## mutter("now = %s", time.time())
159
mode = file_fp[FP_MODE_COLUMN]
160
if stat.S_ISREG(mode):
161
digest = self._really_sha1_file(abspath)
162
elif stat.S_ISLNK(mode):
163
digest = sha.new(os.readlink(abspath)).hexdigest()
165
raise BzrError("file %r: unknown file stat mode: %o"%(abspath,mode))
167
# window of 3 seconds to allow for 2s resolution on windows,
168
# unsynchronized file servers, etc.
169
cutoff = self._cutoff_time()
170
if file_fp[FP_MTIME_COLUMN] >= cutoff \
171
or file_fp[FP_CTIME_COLUMN] >= cutoff:
172
# changed too recently; can't be cached. we can
173
# return the result and it could possibly be cached
176
# the point is that we only want to cache when we are sure that any
177
# subsequent modifications of the file can be detected. If a
178
# modification neither changes the inode, the device, the size, nor
179
# the mode, then we can only distinguish it by time; therefore we
180
# need to let sufficient time elapse before we may cache this entry
181
# again. If we didn't do this, then, for example, a very quick 1
182
# byte replacement in the file might go undetected.
183
## mutter('%r modified too recently; not caching', path)
184
self.danger_count += 1
186
self.removed_count += 1
187
self.needs_write = True
188
del self._cache[path]
190
## mutter('%r added to cache: now=%f, mtime=%d, ctime=%d',
191
## path, time.time(), file_fp[FP_MTIME_COLUMN],
192
## file_fp[FP_CTIME_COLUMN])
193
self.update_count += 1
194
self.needs_write = True
195
self._cache[path] = (digest, file_fp)
198
def _really_sha1_file(self, abspath):
199
"""Calculate the SHA1 of a file by reading the full text"""
200
return sha_file(file(abspath, 'rb', buffering=65000))
203
"""Write contents of cache to file."""
204
outf = AtomicFile(self.cache_file_name(), 'wb', new_mode=self._mode)
206
print >>outf, CACHE_HEADER,
208
for path, c in self._cache.iteritems():
209
assert '//' not in path, path
210
outf.write(path.encode('utf-8'))
212
print >>outf, c[0], # hex sha1
214
print >>outf, "%d" % fld,
217
self.needs_write = False
218
## mutter("write hash cache: %s hits=%d misses=%d stat=%d recent=%d updates=%d",
219
## self.cache_file_name(), self.hit_count, self.miss_count,
221
## self.danger_count, self.update_count)
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
245
if hdr != CACHE_HEADER:
246
mutter('cache header marker not found at top of %s;'
247
' discarding cache', fn)
248
self.needs_write = True
253
path = l[:pos].decode('utf-8')
254
if path in self._cache:
255
warning('duplicated path %r in cache' % path)
259
fields = l[pos:].split(' ')
261
warning("bad line in hashcache: %r" % l)
266
warning("bad sha1 in hashcache: %r" % sha1)
269
fp = tuple(map(long, fields[1:]))
271
self._cache[path] = (sha1, fp)
273
self.needs_write = False
275
def _cutoff_time(self):
276
"""Return cutoff time.
278
Files modified more recently than this time are at risk of being
279
undetectably modified and so can't be cached.
281
return int(time.time()) - 3
283
def _fingerprint(self, abspath):
285
fs = os.lstat(abspath)
287
# might be missing, etc
289
if stat.S_ISDIR(fs.st_mode):
291
# we discard any high precision because it's not reliable; perhaps we
292
# could do better on some systems?
293
return (fs.st_size, long(fs.st_mtime),
294
long(fs.st_ctime), fs.st_ino, fs.st_dev, fs.st_mode)
added
removed
bzrlib/hashcache.py
