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
17
from __future__ import absolute_import
19
# TODO: Up-front, stat all files in order and remove those which are deleted or
20
# out-of-date. Don't actually re-read them until they're needed. That ought
21
# to bring all the inodes into core so that future stats to them are fast, and
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# out-of-date.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
22
20
# it preserves the nice property that any caller will always get up-to-date
23
21
# data except in unavoidable cases.
25
23
# TODO: Perhaps return more details on the file to avoid statting it
26
24
# again: nonexistent, file type, size, etc
28
# TODO: Perhaps use a Python pickle instead of a text file; might be faster.
32
28
CACHE_HEADER = "### bzr hashcache v5\n"
41
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)
53
53
class HashCache(object):
81
81
number of times files have been retrieved from the cache, avoiding a
85
85
number of misses (times files have been completely re-read)
87
87
needs_write = False
89
def __init__(self, root, cache_file_name, mode=None,
90
content_filter_stack_provider=None):
91
"""Create a hash cache in base dir, and set the file mode to mode.
93
:param content_filter_stack_provider: a function that takes a
94
path (relative to the top of the tree) and a file-id as
95
parameters and returns a stack of ContentFilters.
96
If None, no content filtering is performed.
98
self.root = osutils.safe_unicode(root)
99
self.root_utf8 = self.root.encode('utf8') # where is the filesystem encoding ?
89
def __init__(self, basedir):
90
self.basedir = basedir
100
91
self.hit_count = 0
101
92
self.miss_count = 0
102
93
self.stat_count = 0
119
111
self.needs_write = True
123
116
"""Scan all files and remove entries where the cache entry is obsolete.
125
118
Obsolete entries are those where the file has been modified or deleted
126
since the entry was inserted.
119
since the entry was inserted.
128
# FIXME optimisation opportunity, on linux [and check other oses]:
129
# rather than iteritems order, stat in inode order.
130
121
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
133
124
for inum, path, cache_entry in prep:
134
abspath = osutils.pathjoin(self.root, path)
135
fp = self._fingerprint(abspath)
125
abspath = os.sep.join([self.basedir, path])
126
fp = _fingerprint(abspath)
136
127
self.stat_count += 1
138
129
cache_fp = cache_entry[1]
140
131
if (not fp) or (cache_fp != fp):
141
132
# not here or not a regular file anymore
142
133
self.removed_count += 1
143
134
self.needs_write = True
144
135
del self._cache[path]
146
def get_sha1(self, path, stat_value=None):
139
def get_sha1(self, path):
147
140
"""Return the sha1 of a file.
149
if path.__class__ is str:
150
abspath = osutils.pathjoin(self.root_utf8, path)
152
abspath = osutils.pathjoin(self.root, path)
142
abspath = os.sep.join([self.basedir, path])
153
143
self.stat_count += 1
154
file_fp = self._fingerprint(abspath, stat_value)
144
file_fp = _fingerprint(abspath)
157
147
# not a regular file or not existing
158
148
if path in self._cache:
159
149
self.removed_count += 1
160
150
self.needs_write = True
161
151
del self._cache[path]
164
154
if path in self._cache:
165
155
cache_sha1, cache_fp = self._cache[path]
167
157
cache_sha1, cache_fp = None, None
169
159
if cache_fp == file_fp:
170
## mutter("hashcache hit for %s %r -> %s", path, file_fp, cache_sha1)
171
## mutter("now = %s", time.time())
172
160
self.hit_count += 1
173
161
return cache_sha1
175
163
self.miss_count += 1
177
mode = file_fp[FP_MODE_COLUMN]
178
if stat.S_ISREG(mode):
179
if self._filter_provider is None:
182
filters = self._filter_provider(path=path, file_id=None)
183
digest = self._really_sha1_file(abspath, filters)
184
elif stat.S_ISLNK(mode):
185
target = osutils.readlink(osutils.safe_unicode(abspath))
186
digest = osutils.sha_string(target.encode('UTF-8'))
188
raise errors.BzrError("file %r: unknown file stat mode: %o"
191
# window of 3 seconds to allow for 2s resolution on windows,
192
# unsynchronized file servers, etc.
193
cutoff = self._cutoff_time()
194
if file_fp[FP_MTIME_COLUMN] >= cutoff \
195
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:
196
168
# changed too recently; can't be cached. we can
197
169
# return the result and it could possibly be cached
200
# the point is that we only want to cache when we are sure that any
201
# subsequent modifications of the file can be detected. If a
202
# modification neither changes the inode, the device, the size, nor
203
# the mode, then we can only distinguish it by time; therefore we
204
# need to let sufficient time elapse before we may cache this entry
205
# again. If we didn't do this, then, for example, a very quick 1
206
# byte replacement in the file might go undetected.
207
## mutter('%r modified too recently; not caching', path)
208
self.danger_count += 1
171
self.danger_count += 1
210
173
self.removed_count += 1
211
174
self.needs_write = True
212
175
del self._cache[path]
214
## mutter('%r added to cache: now=%f, mtime=%d, ctime=%d',
215
## path, time.time(), file_fp[FP_MTIME_COLUMN],
216
## file_fp[FP_CTIME_COLUMN])
217
177
self.update_count += 1
218
178
self.needs_write = True
219
179
self._cache[path] = (digest, file_fp)
222
def _really_sha1_file(self, abspath, filters):
223
"""Calculate the SHA1 of a file by reading the full text"""
224
return _mod_filters.internal_size_sha_file_byname(abspath, filters)[1]
227
187
"""Write contents of cache to file."""
228
outf = atomicfile.AtomicFile(self.cache_file_name(), 'wb',
188
from atomicfile import AtomicFile
190
outf = AtomicFile(self.cache_file_name(), 'wb')
231
outf.write(CACHE_HEADER)
192
print >>outf, CACHE_HEADER,
233
194
for path, c in self._cache.iteritems():
234
line_info = [path.encode('utf-8'), '// ', c[0], ' ']
235
line_info.append(' '.join([str(fld) for fld in c[1]]))
236
line_info.append('\n')
237
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,
239
204
self.needs_write = False
240
## mutter("write hash cache: %s hits=%d misses=%d stat=%d recent=%d updates=%d",
241
## self.cache_file_name(), self.hit_count, self.miss_count,
243
## self.danger_count, self.update_count)
248
212
"""Reinstate cache from file.
250
214
Overwrites existing cache.
252
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
273
238
pos = l.index('// ')
274
239
path = l[:pos].decode('utf-8')
275
240
if path in self._cache:
276
trace.warning('duplicated path %r in cache' % path)
241
warning('duplicated path %r in cache' % path)
280
245
fields = l[pos:].split(' ')
282
trace.warning("bad line in hashcache: %r" % l)
247
warning("bad line in hashcache: %r" % l)
286
251
if len(sha1) != 40:
287
trace.warning("bad sha1 in hashcache: %r" % sha1)
252
warning("bad sha1 in hashcache: %r" % sha1)
290
255
fp = tuple(map(long, fields[1:]))
292
257
self._cache[path] = (sha1, fp)
294
# GZ 2009-09-20: Should really use a try/finally block to ensure close
297
259
self.needs_write = False
299
def _cutoff_time(self):
300
"""Return cutoff time.
302
Files modified more recently than this time are at risk of being
303
undetectably modified and so can't be cached.
305
return int(time.time()) - 3
307
def _fingerprint(self, abspath, stat_value=None):
308
if stat_value is None:
310
stat_value = os.lstat(abspath)
312
# might be missing, etc
314
if stat.S_ISDIR(stat_value.st_mode):
316
# we discard any high precision because it's not reliable; perhaps we
317
# could do better on some systems?
318
return (stat_value.st_size, long(stat_value.st_mtime),
319
long(stat_value.st_ctime), stat_value.st_ino,
320
stat_value.st_dev, stat_value.st_mode)
added
removed
bzrlib/hashcache.py
