1
# (C) 2005 Canonical Ltd
1
# Copyright (C) 2005-2010 Canonical Ltd
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
5
5
# the Free Software Foundation; either version 2 of the License, or
6
6
# (at your option) any later version.
8
8
# This program is distributed in the hope that it will be useful,
9
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
11
# GNU General Public License for more details.
13
13
# You should have received a copy of the GNU General Public License
14
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
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# out-of-date.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
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# out-of-date.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
20
# it preserves the nice property that any caller will always get up-to-date
21
21
# data except in unavoidable cases.
23
23
# TODO: Perhaps return more details on the file to avoid statting it
24
24
# again: nonexistent, file type, size, etc
26
# TODO: Perhaps use a Python pickle instead of a text file; might be faster.
28
30
CACHE_HEADER = "### bzr hashcache v5\n"
32
from bzrlib.osutils import sha_file
33
from bzrlib.trace import mutter, warning
37
def _fingerprint(abspath):
39
fs = os.lstat(abspath)
41
# might be missing, etc
44
if stat.S_ISDIR(fs.st_mode):
47
# we discard any high precision because it's not reliable; perhaps we
48
# could do better on some systems?
49
return (fs.st_size, long(fs.st_mtime),
50
long(fs.st_ctime), fs.st_ino, fs.st_dev)
39
filters as _mod_filters,
53
51
class HashCache(object):
81
79
number of times files have been retrieved from the cache, avoiding a
85
83
number of misses (times files have been completely re-read)
87
85
needs_write = False
89
def __init__(self, basedir):
90
self.basedir = basedir
87
def __init__(self, root, cache_file_name, mode=None,
88
content_filter_stack_provider=None):
89
"""Create a hash cache in base dir, and set the file mode to mode.
91
:param content_filter_stack_provider: a function that takes a
92
path (relative to the top of the tree) and a file-id as
93
parameters and returns a stack of ContentFilters.
94
If None, no content filtering is performed.
96
self.root = osutils.safe_unicode(root)
97
self.root_utf8 = self.root.encode('utf8') # where is the filesystem encoding ?
92
99
self.miss_count = 0
93
100
self.stat_count = 0
111
117
self.needs_write = True
116
121
"""Scan all files and remove entries where the cache entry is obsolete.
118
123
Obsolete entries are those where the file has been modified or deleted
119
since the entry was inserted.
124
since the entry was inserted.
126
# FIXME optimisation opportunity, on linux [and check other oses]:
127
# rather than iteritems order, stat in inode order.
121
128
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
124
131
for inum, path, cache_entry in prep:
125
abspath = os.sep.join([self.basedir, path])
126
fp = _fingerprint(abspath)
132
abspath = osutils.pathjoin(self.root, path)
133
fp = self._fingerprint(abspath)
127
134
self.stat_count += 1
129
136
cache_fp = cache_entry[1]
131
138
if (not fp) or (cache_fp != fp):
132
139
# not here or not a regular file anymore
133
140
self.removed_count += 1
134
141
self.needs_write = True
135
142
del self._cache[path]
139
def get_sha1(self, path):
144
def get_sha1(self, path, stat_value=None):
140
145
"""Return the sha1 of a file.
142
abspath = os.sep.join([self.basedir, path])
147
if path.__class__ is str:
148
abspath = osutils.pathjoin(self.root_utf8, path)
150
abspath = osutils.pathjoin(self.root, path)
143
151
self.stat_count += 1
144
file_fp = _fingerprint(abspath)
152
file_fp = self._fingerprint(abspath, stat_value)
147
155
# not a regular file or not existing
148
156
if path in self._cache:
149
157
self.removed_count += 1
150
158
self.needs_write = True
151
159
del self._cache[path]
154
162
if path in self._cache:
155
163
cache_sha1, cache_fp = self._cache[path]
157
165
cache_sha1, cache_fp = None, None
159
167
if cache_fp == file_fp:
168
## mutter("hashcache hit for %s %r -> %s", path, file_fp, cache_sha1)
169
## mutter("now = %s", time.time())
160
170
self.hit_count += 1
161
171
return cache_sha1
163
173
self.miss_count += 1
164
digest = sha_file(file(abspath, 'rb', buffering=65000))
166
now = int(time.time())
167
if file_fp[1] >= now or file_fp[2] >= now:
175
mode = file_fp[FP_MODE_COLUMN]
176
if stat.S_ISREG(mode):
177
if self._filter_provider is None:
180
filters = self._filter_provider(path=path, file_id=None)
181
digest = self._really_sha1_file(abspath, filters)
182
elif stat.S_ISLNK(mode):
183
target = osutils.readlink(osutils.safe_unicode(abspath))
184
digest = osutils.sha_string(target.encode('UTF-8'))
186
raise errors.BzrError("file %r: unknown file stat mode: %o"
189
# window of 3 seconds to allow for 2s resolution on windows,
190
# unsynchronized file servers, etc.
191
cutoff = self._cutoff_time()
192
if file_fp[FP_MTIME_COLUMN] >= cutoff \
193
or file_fp[FP_CTIME_COLUMN] >= cutoff:
168
194
# changed too recently; can't be cached. we can
169
195
# return the result and it could possibly be cached
171
self.danger_count += 1
198
# the point is that we only want to cache when we are sure that any
199
# subsequent modifications of the file can be detected. If a
200
# modification neither changes the inode, the device, the size, nor
201
# the mode, then we can only distinguish it by time; therefore we
202
# need to let sufficient time elapse before we may cache this entry
203
# again. If we didn't do this, then, for example, a very quick 1
204
# byte replacement in the file might go undetected.
205
## mutter('%r modified too recently; not caching', path)
206
self.danger_count += 1
173
208
self.removed_count += 1
174
209
self.needs_write = True
175
210
del self._cache[path]
212
## mutter('%r added to cache: now=%f, mtime=%d, ctime=%d',
213
## path, time.time(), file_fp[FP_MTIME_COLUMN],
214
## file_fp[FP_CTIME_COLUMN])
177
215
self.update_count += 1
178
216
self.needs_write = True
179
217
self._cache[path] = (digest, file_fp)
220
def _really_sha1_file(self, abspath, filters):
221
"""Calculate the SHA1 of a file by reading the full text"""
222
return _mod_filters.internal_size_sha_file_byname(abspath, filters)[1]
187
225
"""Write contents of cache to file."""
188
from atomicfile import AtomicFile
190
outf = AtomicFile(self.cache_file_name(), 'wb')
226
outf = atomicfile.AtomicFile(self.cache_file_name(), 'wb',
192
print >>outf, CACHE_HEADER,
229
outf.write(CACHE_HEADER)
194
231
for path, c in self._cache.iteritems():
195
assert '//' not in path, path
196
outf.write(path.encode('utf-8'))
198
print >>outf, c[0], # hex sha1
200
print >>outf, "%d" % fld,
232
line_info = [path.encode('utf-8'), '// ', c[0], ' ']
233
line_info.append(' '.join([str(fld) for fld in c[1]]))
234
line_info.append('\n')
235
outf.write(''.join(line_info))
204
237
self.needs_write = False
238
## mutter("write hash cache: %s hits=%d misses=%d stat=%d recent=%d updates=%d",
239
## self.cache_file_name(), self.hit_count, self.miss_count,
241
## self.danger_count, self.update_count)
212
246
"""Reinstate cache from file.
214
248
Overwrites existing cache.
216
If the cache file has the wrong version marker, this just clears
250
If the cache file has the wrong version marker, this just clears
222
256
inf = file(fn, 'rb', buffering=65000)
223
257
except IOError, e:
224
mutter("failed to open %s: %s" % (fn, e))
258
trace.mutter("failed to open %s: %s", fn, e)
259
# better write it now so it is valid
260
self.needs_write = True
228
263
hdr = inf.readline()
229
264
if hdr != CACHE_HEADER:
230
mutter('cache header marker not found at top of %s; discarding cache'
265
trace.mutter('cache header marker not found at top of %s;'
266
' discarding cache', fn)
267
self.needs_write = True
235
271
pos = l.index('// ')
236
272
path = l[:pos].decode('utf-8')
237
273
if path in self._cache:
238
warning('duplicated path %r in cache' % path)
274
trace.warning('duplicated path %r in cache' % path)
242
278
fields = l[pos:].split(' ')
244
warning("bad line in hashcache: %r" % l)
280
trace.warning("bad line in hashcache: %r" % l)
248
284
if len(sha1) != 40:
249
warning("bad sha1 in hashcache: %r" % sha1)
285
trace.warning("bad sha1 in hashcache: %r" % sha1)
252
288
fp = tuple(map(long, fields[1:]))
254
290
self._cache[path] = (sha1, fp)
292
# GZ 2009-09-20: Should really use a try/finally block to ensure close
256
295
self.needs_write = False
297
def _cutoff_time(self):
298
"""Return cutoff time.
300
Files modified more recently than this time are at risk of being
301
undetectably modified and so can't be cached.
303
return int(time.time()) - 3
305
def _fingerprint(self, abspath, stat_value=None):
306
if stat_value is None:
308
stat_value = os.lstat(abspath)
310
# might be missing, etc
312
if stat.S_ISDIR(stat_value.st_mode):
314
# we discard any high precision because it's not reliable; perhaps we
315
# could do better on some systems?
316
return (stat_value.st_size, long(stat_value.st_mtime),
317
long(stat_value.st_ctime), stat_value.st_ino,
318
stat_value.st_dev, stat_value.st_mode)
added
removed
bzrlib/hashcache.py
