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# out-of-date.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"
34
from bzrlib.osutils import sha_file, sha_string, pathjoin, safe_unicode
35
from bzrlib.trace import mutter, warning
36
from bzrlib.atomicfile import AtomicFile
37
from bzrlib.errors import BzrError
46
class HashCache(object):
47
"""Cache for looking up file SHA-1.
49
Files are considered to match the cached value if the fingerprint
50
of the file has not changed. This includes its mtime, ctime,
51
device number, inode number, and size. This should catch
52
modifications or replacement of the file by a new one.
54
This may not catch modifications that do not change the file's
55
size and that occur within the resolution window of the
56
timestamps. To handle this we specifically do not cache files
57
which have changed since the start of the present second, since
58
they could undetectably change again.
60
This scheme may fail if the machine's clock steps backwards.
63
This does not canonicalize the paths passed in; that should be
67
Indexed by path, points to a two-tuple of the SHA-1 of the file.
71
number of times files have been statted
74
number of times files have been retrieved from the cache, avoiding a
78
number of misses (times files have been completely re-read)
82
def __init__(self, root, cache_file_name, mode=None):
83
"""Create a hash cache in base dir, and set the file mode to mode."""
84
self.root = safe_unicode(root)
85
self.root_utf8 = self.root.encode('utf8') # where is the filesystem encoding ?
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, stat_value=None):
132
"""Return the sha1 of a file.
134
if path.__class__ is str:
135
abspath = pathjoin(self.root_utf8, path)
137
abspath = pathjoin(self.root, path)
139
file_fp = self._fingerprint(abspath, stat_value)
142
# not a regular file or not existing
143
if path in self._cache:
144
self.removed_count += 1
145
self.needs_write = True
146
del self._cache[path]
149
if path in self._cache:
150
cache_sha1, cache_fp = self._cache[path]
152
cache_sha1, cache_fp = None, None
154
if cache_fp == file_fp:
155
## mutter("hashcache hit for %s %r -> %s", path, file_fp, cache_sha1)
156
## mutter("now = %s", time.time())
162
mode = file_fp[FP_MODE_COLUMN]
163
if stat.S_ISREG(mode):
164
digest = self._really_sha1_file(abspath)
165
elif stat.S_ISLNK(mode):
166
digest = sha_string(os.readlink(abspath))
168
raise BzrError("file %r: unknown file stat mode: %o"%(abspath,mode))
170
# window of 3 seconds to allow for 2s resolution on windows,
171
# unsynchronized file servers, etc.
172
cutoff = self._cutoff_time()
173
if file_fp[FP_MTIME_COLUMN] >= cutoff \
174
or file_fp[FP_CTIME_COLUMN] >= cutoff:
175
# changed too recently; can't be cached. we can
176
# return the result and it could possibly be cached
179
# the point is that we only want to cache when we are sure that any
180
# subsequent modifications of the file can be detected. If a
181
# modification neither changes the inode, the device, the size, nor
182
# the mode, then we can only distinguish it by time; therefore we
183
# need to let sufficient time elapse before we may cache this entry
184
# again. If we didn't do this, then, for example, a very quick 1
185
# byte replacement in the file might go undetected.
186
## mutter('%r modified too recently; not caching', path)
187
self.danger_count += 1
189
self.removed_count += 1
190
self.needs_write = True
191
del self._cache[path]
193
## mutter('%r added to cache: now=%f, mtime=%d, ctime=%d',
194
## path, time.time(), file_fp[FP_MTIME_COLUMN],
195
## file_fp[FP_CTIME_COLUMN])
196
self.update_count += 1
197
self.needs_write = True
198
self._cache[path] = (digest, file_fp)
201
def _really_sha1_file(self, abspath):
202
"""Calculate the SHA1 of a file by reading the full text"""
203
return sha_file(file(abspath, 'rb', buffering=65000))
206
"""Write contents of cache to file."""
207
outf = AtomicFile(self.cache_file_name(), 'wb', new_mode=self._mode)
209
outf.write(CACHE_HEADER)
211
for path, c in self._cache.iteritems():
212
assert '//' not in path, path
213
line_info = [path.encode('utf-8'), '// ', c[0], ' ']
214
line_info.append(' '.join([str(fld) for fld in c[1]]))
215
line_info.append('\n')
216
outf.write(''.join(line_info))
218
self.needs_write = False
219
## mutter("write hash cache: %s hits=%d misses=%d stat=%d recent=%d updates=%d",
220
## self.cache_file_name(), self.hit_count, self.miss_count,
222
## 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, stat_value=None):
284
if stat_value is None:
286
stat_value = os.lstat(abspath)
288
# might be missing, etc
290
if stat.S_ISDIR(stat_value.st_mode):
292
# we discard any high precision because it's not reliable; perhaps we
293
# could do better on some systems?
294
return (stat_value.st_size, long(stat_value.st_mtime),
295
long(stat_value.st_ctime), stat_value.st_ino,
296
stat_value.st_dev, stat_value.st_mode)
added
removed
bzrlib/hashcache.py
