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
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# out-of-date.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
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# out-of-date.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
20
22
# it preserves the nice property that any caller will always get up-to-date
21
23
# data except in unavoidable cases.
30
32
CACHE_HEADER = "### bzr hashcache v5\n"
35
from bzrlib.osutils import sha_file
36
from bzrlib.trace import mutter, warning
37
from bzrlib.atomicfile import AtomicFile
38
from bzrlib.osutils import pathjoin
41
filters as _mod_filters,
43
def _fingerprint(abspath):
45
fs = os.lstat(abspath)
47
# might be missing, etc
50
if stat.S_ISDIR(fs.st_mode):
53
# we discard any high precision because it's not reliable; perhaps we
54
# could do better on some systems?
55
return (fs.st_size, long(fs.st_mtime),
56
long(fs.st_ctime), fs.st_ino, fs.st_dev, fs.st_mode)
59
53
class HashCache(object):
87
81
number of times files have been retrieved from the cache, avoiding a
91
85
number of misses (times files have been completely re-read)
93
87
needs_write = False
95
def __init__(self, basedir):
96
self.basedir = basedir
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 ?
97
100
self.hit_count = 0
98
101
self.miss_count = 0
99
102
self.stat_count = 0
115
119
self.needs_write = True
120
123
"""Scan all files and remove entries where the cache entry is obsolete.
122
125
Obsolete entries are those where the file has been modified or deleted
123
since the entry was inserted.
126
since the entry was inserted.
128
# FIXME optimisation opportunity, on linux [and check other oses]:
129
# rather than iteritems order, stat in inode order.
125
130
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
128
133
for inum, path, cache_entry in prep:
129
abspath = pathjoin(self.basedir, path)
130
fp = _fingerprint(abspath)
134
abspath = osutils.pathjoin(self.root, path)
135
fp = self._fingerprint(abspath)
131
136
self.stat_count += 1
133
138
cache_fp = cache_entry[1]
135
140
if (not fp) or (cache_fp != fp):
136
141
# not here or not a regular file anymore
137
142
self.removed_count += 1
138
143
self.needs_write = True
139
144
del self._cache[path]
142
def get_sha1(self, path):
146
def get_sha1(self, path, stat_value=None):
143
147
"""Return the sha1 of a file.
145
abspath = pathjoin(self.basedir, path)
149
if path.__class__ is str:
150
abspath = osutils.pathjoin(self.root_utf8, path)
152
abspath = osutils.pathjoin(self.root, path)
146
153
self.stat_count += 1
147
file_fp = _fingerprint(abspath)
154
file_fp = self._fingerprint(abspath, stat_value)
150
157
# not a regular file or not existing
151
158
if path in self._cache:
152
159
self.removed_count += 1
153
160
self.needs_write = True
154
161
del self._cache[path]
157
164
if path in self._cache:
158
165
cache_sha1, cache_fp = self._cache[path]
160
167
cache_sha1, cache_fp = None, None
162
169
if cache_fp == file_fp:
170
## mutter("hashcache hit for %s %r -> %s", path, file_fp, cache_sha1)
171
## mutter("now = %s", time.time())
163
172
self.hit_count += 1
164
173
return cache_sha1
166
175
self.miss_count += 1
169
177
mode = file_fp[FP_MODE_COLUMN]
170
178
if stat.S_ISREG(mode):
171
digest = sha_file(file(abspath, 'rb', buffering=65000))
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)
172
184
elif stat.S_ISLNK(mode):
173
link_target = os.readlink(abspath)
174
digest = sha.new(os.readlink(abspath)).hexdigest()
185
target = osutils.readlink(osutils.safe_unicode(abspath))
186
digest = osutils.sha_string(target.encode('UTF-8'))
176
raise BzrError("file %r: unknown file stat mode: %o"%(abspath,mode))
188
raise errors.BzrError("file %r: unknown file stat mode: %o"
178
now = int(time.time())
179
if file_fp[1] >= now or file_fp[2] >= now:
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:
180
196
# changed too recently; can't be cached. we can
181
197
# return the result and it could possibly be cached
183
self.danger_count += 1
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
185
210
self.removed_count += 1
186
211
self.needs_write = True
187
212
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])
189
217
self.update_count += 1
190
218
self.needs_write = True
191
219
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]
195
227
"""Write contents of cache to file."""
196
outf = AtomicFile(self.cache_file_name(), 'wb')
228
outf = atomicfile.AtomicFile(self.cache_file_name(), 'wb',
198
print >>outf, CACHE_HEADER,
231
outf.write(CACHE_HEADER)
200
233
for path, c in self._cache.iteritems():
201
assert '//' not in path, path
202
outf.write(path.encode('utf-8'))
204
print >>outf, c[0], # hex sha1
206
print >>outf, "%d" % fld,
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))
210
239
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)
216
248
"""Reinstate cache from file.
218
250
Overwrites existing cache.
220
If the cache file has the wrong version marker, this just clears
252
If the cache file has the wrong version marker, this just clears
242
273
pos = l.index('// ')
243
274
path = l[:pos].decode('utf-8')
244
275
if path in self._cache:
245
warning('duplicated path %r in cache' % path)
276
trace.warning('duplicated path %r in cache' % path)
249
280
fields = l[pos:].split(' ')
250
281
if len(fields) != 7:
251
warning("bad line in hashcache: %r" % l)
282
trace.warning("bad line in hashcache: %r" % l)
255
286
if len(sha1) != 40:
256
warning("bad sha1 in hashcache: %r" % sha1)
287
trace.warning("bad sha1 in hashcache: %r" % sha1)
259
290
fp = tuple(map(long, fields[1:]))
261
292
self._cache[path] = (sha1, fp)
294
# GZ 2009-09-20: Should really use a try/finally block to ensure close
263
297
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
