1
# Copyright (C) 2005-2010 Canonical Ltd
1
# (C) 2005 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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
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# out-of-date.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
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# out-of-date.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.
30
28
CACHE_HEADER = "### bzr hashcache v5\n"
39
filters as _mod_filters,
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)
51
53
class HashCache(object):
79
81
number of times files have been retrieved from the cache, avoiding a
83
85
number of misses (times files have been completely re-read)
85
87
needs_write = False
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 ?
89
def __init__(self, basedir):
90
self.basedir = basedir
99
92
self.miss_count = 0
100
93
self.stat_count = 0
117
111
self.needs_write = True
121
116
"""Scan all files and remove entries where the cache entry is obsolete.
123
118
Obsolete entries are those where the file has been modified or deleted
124
since the entry was inserted.
119
since the entry was inserted.
126
# FIXME optimisation opportunity, on linux [and check other oses]:
127
# rather than iteritems order, stat in inode order.
128
121
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
131
124
for inum, path, cache_entry in prep:
132
abspath = osutils.pathjoin(self.root, path)
133
fp = self._fingerprint(abspath)
125
abspath = os.sep.join([self.basedir, path])
126
fp = _fingerprint(abspath)
134
127
self.stat_count += 1
136
129
cache_fp = cache_entry[1]
138
131
if (not fp) or (cache_fp != fp):
139
132
# not here or not a regular file anymore
140
133
self.removed_count += 1
141
134
self.needs_write = True
142
135
del self._cache[path]
144
def get_sha1(self, path, stat_value=None):
139
def get_sha1(self, path):
145
140
"""Return the sha1 of a file.
147
if path.__class__ is str:
148
abspath = osutils.pathjoin(self.root_utf8, path)
150
abspath = osutils.pathjoin(self.root, path)
142
abspath = os.sep.join([self.basedir, path])
151
143
self.stat_count += 1
152
file_fp = self._fingerprint(abspath, stat_value)
144
file_fp = _fingerprint(abspath)
155
147
# not a regular file or not existing
156
148
if path in self._cache:
157
149
self.removed_count += 1
158
150
self.needs_write = True
159
151
del self._cache[path]
162
154
if path in self._cache:
163
155
cache_sha1, cache_fp = self._cache[path]
165
157
cache_sha1, cache_fp = None, None
167
159
if cache_fp == file_fp:
168
## mutter("hashcache hit for %s %r -> %s", path, file_fp, cache_sha1)
169
## mutter("now = %s", time.time())
170
160
self.hit_count += 1
171
161
return cache_sha1
173
163
self.miss_count += 1
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:
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:
194
168
# changed too recently; can't be cached. we can
195
169
# return the result and it could possibly be cached
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
171
self.danger_count += 1
208
173
self.removed_count += 1
209
174
self.needs_write = True
210
175
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])
215
177
self.update_count += 1
216
178
self.needs_write = True
217
179
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]
225
187
"""Write contents of cache to file."""
226
outf = atomicfile.AtomicFile(self.cache_file_name(), 'wb',
188
from atomicfile import AtomicFile
190
outf = AtomicFile(self.cache_file_name(), 'wb')
229
outf.write(CACHE_HEADER)
192
print >>outf, CACHE_HEADER,
231
194
for path, c in self._cache.iteritems():
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))
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,
237
204
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)
246
212
"""Reinstate cache from file.
248
214
Overwrites existing cache.
250
If the cache file has the wrong version marker, this just clears
216
If the cache file has the wrong version marker, this just clears
256
222
inf = file(fn, 'rb', buffering=65000)
257
223
except IOError, 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
224
mutter("failed to open %s: %s" % (fn, e))
263
228
hdr = inf.readline()
264
229
if hdr != CACHE_HEADER:
265
trace.mutter('cache header marker not found at top of %s;'
266
' discarding cache', fn)
267
self.needs_write = True
230
mutter('cache header marker not found at top of %s; discarding cache'
271
235
pos = l.index('// ')
272
236
path = l[:pos].decode('utf-8')
273
237
if path in self._cache:
274
trace.warning('duplicated path %r in cache' % path)
238
warning('duplicated path %r in cache' % path)
278
242
fields = l[pos:].split(' ')
280
trace.warning("bad line in hashcache: %r" % l)
244
warning("bad line in hashcache: %r" % l)
284
248
if len(sha1) != 40:
285
trace.warning("bad sha1 in hashcache: %r" % sha1)
249
warning("bad sha1 in hashcache: %r" % sha1)
288
252
fp = tuple(map(long, fields[1:]))
290
254
self._cache[path] = (sha1, fp)
292
# GZ 2009-09-20: Should really use a try/finally block to ensure close
295
256
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
