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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 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# out-of-date.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.filters import internal_size_sha_file_byname
35
from bzrlib.osutils import sha_file, sha_string, 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
content_filter_stack_provider=None):
85
"""Create a hash cache in base dir, and set the file mode to mode.
87
:param content_filter_stack_provider: a function that takes a
88
path (relative to the top of the tree) and a file-id as
89
parameters and returns a stack of ContentFilters.
90
If None, no content filtering is performed.
92
self.root = safe_unicode(root)
93
self.root_utf8 = self.root.encode('utf8') # where is the filesystem encoding ?
98
self.removed_count = 0
102
self._cache_file_name = safe_unicode(cache_file_name)
103
self._filter_provider = content_filter_stack_provider
105
def cache_file_name(self):
106
return self._cache_file_name
109
"""Discard all cached information.
111
This does not reset the counters."""
113
self.needs_write = True
117
"""Scan all files and remove entries where the cache entry is obsolete.
119
Obsolete entries are those where the file has been modified or deleted
120
since the entry was inserted.
122
# FIXME optimisation opportunity, on linux [and check other oses]:
123
# rather than iteritems order, stat in inode order.
124
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
127
for inum, path, cache_entry in prep:
128
abspath = pathjoin(self.root, path)
129
fp = self._fingerprint(abspath)
132
cache_fp = cache_entry[1]
134
if (not fp) or (cache_fp != fp):
135
# not here or not a regular file anymore
136
self.removed_count += 1
137
self.needs_write = True
138
del self._cache[path]
140
def get_sha1(self, path, stat_value=None):
141
"""Return the sha1 of a file.
143
if path.__class__ is str:
144
abspath = pathjoin(self.root_utf8, path)
146
abspath = pathjoin(self.root, path)
148
file_fp = self._fingerprint(abspath, stat_value)
151
# not a regular file or not existing
152
if path in self._cache:
153
self.removed_count += 1
154
self.needs_write = True
155
del self._cache[path]
158
if path in self._cache:
159
cache_sha1, cache_fp = self._cache[path]
161
cache_sha1, cache_fp = None, None
163
if cache_fp == file_fp:
164
## mutter("hashcache hit for %s %r -> %s", path, file_fp, cache_sha1)
165
## mutter("now = %s", time.time())
171
mode = file_fp[FP_MODE_COLUMN]
172
if stat.S_ISREG(mode):
173
if self._filter_provider is None:
176
filters = self._filter_provider(path=path, file_id=None)
177
digest = self._really_sha1_file(abspath, filters)
178
elif stat.S_ISLNK(mode):
179
digest = sha_string(os.readlink(abspath))
181
raise BzrError("file %r: unknown file stat mode: %o"%(abspath,mode))
183
# window of 3 seconds to allow for 2s resolution on windows,
184
# unsynchronized file servers, etc.
185
cutoff = self._cutoff_time()
186
if file_fp[FP_MTIME_COLUMN] >= cutoff \
187
or file_fp[FP_CTIME_COLUMN] >= cutoff:
188
# changed too recently; can't be cached. we can
189
# return the result and it could possibly be cached
192
# the point is that we only want to cache when we are sure that any
193
# subsequent modifications of the file can be detected. If a
194
# modification neither changes the inode, the device, the size, nor
195
# the mode, then we can only distinguish it by time; therefore we
196
# need to let sufficient time elapse before we may cache this entry
197
# again. If we didn't do this, then, for example, a very quick 1
198
# byte replacement in the file might go undetected.
199
## mutter('%r modified too recently; not caching', path)
200
self.danger_count += 1
202
self.removed_count += 1
203
self.needs_write = True
204
del self._cache[path]
206
## mutter('%r added to cache: now=%f, mtime=%d, ctime=%d',
207
## path, time.time(), file_fp[FP_MTIME_COLUMN],
208
## file_fp[FP_CTIME_COLUMN])
209
self.update_count += 1
210
self.needs_write = True
211
self._cache[path] = (digest, file_fp)
214
def _really_sha1_file(self, abspath, filters):
215
"""Calculate the SHA1 of a file by reading the full text"""
216
return internal_size_sha_file_byname(abspath, filters)[1]
219
"""Write contents of cache to file."""
220
outf = AtomicFile(self.cache_file_name(), 'wb', new_mode=self._mode)
222
outf.write(CACHE_HEADER)
224
for path, c in self._cache.iteritems():
225
line_info = [path.encode('utf-8'), '// ', c[0], ' ']
226
line_info.append(' '.join([str(fld) for fld in c[1]]))
227
line_info.append('\n')
228
outf.write(''.join(line_info))
230
self.needs_write = False
231
## mutter("write hash cache: %s hits=%d misses=%d stat=%d recent=%d updates=%d",
232
## self.cache_file_name(), self.hit_count, self.miss_count,
234
## self.danger_count, self.update_count)
239
"""Reinstate cache from file.
241
Overwrites existing cache.
243
If the cache file has the wrong version marker, this just clears
247
fn = self.cache_file_name()
249
inf = file(fn, 'rb', buffering=65000)
251
mutter("failed to open %s: %s", fn, e)
252
# better write it now so it is valid
253
self.needs_write = True
257
if hdr != CACHE_HEADER:
258
mutter('cache header marker not found at top of %s;'
259
' discarding cache', fn)
260
self.needs_write = True
265
path = l[:pos].decode('utf-8')
266
if path in self._cache:
267
warning('duplicated path %r in cache' % path)
271
fields = l[pos:].split(' ')
273
warning("bad line in hashcache: %r" % l)
278
warning("bad sha1 in hashcache: %r" % sha1)
281
fp = tuple(map(long, fields[1:]))
283
self._cache[path] = (sha1, fp)
285
self.needs_write = False
287
def _cutoff_time(self):
288
"""Return cutoff time.
290
Files modified more recently than this time are at risk of being
291
undetectably modified and so can't be cached.
293
return int(time.time()) - 3
295
def _fingerprint(self, abspath, stat_value=None):
296
if stat_value is None:
298
stat_value = os.lstat(abspath)
300
# might be missing, etc
302
if stat.S_ISDIR(stat_value.st_mode):
304
# we discard any high precision because it's not reliable; perhaps we
305
# could do better on some systems?
306
return (stat_value.st_size, long(stat_value.st_mtime),
307
long(stat_value.st_ctime), stat_value.st_ino,
308
stat_value.st_dev, stat_value.st_mode)
added
removed
bzrlib/hashcache.py
