1
# Copyright (C) 2005, 2006 by Canonical Ltd
1
# Copyright (C) 2005, 2006 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# out-of-date.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# out-of-date.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.
30
30
CACHE_HEADER = "### bzr hashcache v5\n"
35
from bzrlib.osutils import sha_file, pathjoin, safe_unicode
36
from bzrlib.trace import mutter, warning
37
from bzrlib.atomicfile import AtomicFile
38
from bzrlib.errors import BzrError
39
filters as _mod_filters,
41
45
FP_MTIME_COLUMN = 1
42
46
FP_CTIME_COLUMN = 2
45
def _fingerprint(abspath):
47
fs = os.lstat(abspath)
49
# might be missing, etc
52
if stat.S_ISDIR(fs.st_mode):
55
# we discard any high precision because it's not reliable; perhaps we
56
# could do better on some systems?
57
return (fs.st_size, long(fs.st_mtime),
58
long(fs.st_ctime), fs.st_ino, fs.st_dev, fs.st_mode)
61
51
class HashCache(object):
89
79
number of times files have been retrieved from the cache, avoiding a
93
83
number of misses (times files have been completely re-read)
95
85
needs_write = False
97
def __init__(self, root, cache_file_name, mode=None):
98
"""Create a hash cache in base dir, and set the file mode to mode."""
99
self.root = safe_unicode(root)
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 ?
100
98
self.hit_count = 0
101
99
self.miss_count = 0
102
100
self.stat_count = 0
122
121
"""Scan all files and remove entries where the cache entry is obsolete.
124
123
Obsolete entries are those where the file has been modified or deleted
125
since the entry was inserted.
124
since the entry was inserted.
127
126
# FIXME optimisation opportunity, on linux [and check other oses]:
128
127
# rather than iteritems order, stat in inode order.
129
128
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
132
131
for inum, path, cache_entry in prep:
133
abspath = pathjoin(self.root, path)
134
fp = _fingerprint(abspath)
132
abspath = osutils.pathjoin(self.root, path)
133
fp = self._fingerprint(abspath)
135
134
self.stat_count += 1
137
136
cache_fp = cache_entry[1]
139
138
if (not fp) or (cache_fp != fp):
140
139
# not here or not a regular file anymore
141
140
self.removed_count += 1
142
141
self.needs_write = True
143
142
del self._cache[path]
146
def get_sha1(self, path):
144
def get_sha1(self, path, stat_value=None):
147
145
"""Return the sha1 of a file.
149
abspath = pathjoin(self.root, path)
147
if path.__class__ is str:
148
abspath = osutils.pathjoin(self.root_utf8, path)
150
abspath = osutils.pathjoin(self.root, path)
150
151
self.stat_count += 1
151
file_fp = _fingerprint(abspath)
152
file_fp = self._fingerprint(abspath, stat_value)
154
155
# not a regular file or not existing
155
156
if path in self._cache:
156
157
self.removed_count += 1
157
158
self.needs_write = True
158
159
del self._cache[path]
161
162
if path in self._cache:
162
163
cache_sha1, cache_fp = self._cache[path]
164
165
cache_sha1, cache_fp = None, None
166
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())
167
170
self.hit_count += 1
168
171
return cache_sha1
170
173
self.miss_count += 1
173
175
mode = file_fp[FP_MODE_COLUMN]
174
176
if stat.S_ISREG(mode):
175
digest = sha_file(file(abspath, 'rb', buffering=65000))
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)
176
182
elif stat.S_ISLNK(mode):
177
digest = sha.new(os.readlink(abspath)).hexdigest()
183
digest = osutils.sha_string(os.readlink(abspath))
179
raise BzrError("file %r: unknown file stat mode: %o"%(abspath,mode))
185
raise errors.BzrError("file %r: unknown file stat mode: %o"
181
now = int(time.time())
182
if file_fp[FP_MTIME_COLUMN] >= now or file_fp[FP_CTIME_COLUMN] >= now:
188
# window of 3 seconds to allow for 2s resolution on windows,
189
# unsynchronized file servers, etc.
190
cutoff = self._cutoff_time()
191
if file_fp[FP_MTIME_COLUMN] >= cutoff \
192
or file_fp[FP_CTIME_COLUMN] >= cutoff:
183
193
# changed too recently; can't be cached. we can
184
194
# return the result and it could possibly be cached
191
201
# need to let sufficient time elapse before we may cache this entry
192
202
# again. If we didn't do this, then, for example, a very quick 1
193
203
# byte replacement in the file might go undetected.
194
self.danger_count += 1
204
## mutter('%r modified too recently; not caching', path)
205
self.danger_count += 1
196
207
self.removed_count += 1
197
208
self.needs_write = True
198
209
del self._cache[path]
211
## mutter('%r added to cache: now=%f, mtime=%d, ctime=%d',
212
## path, time.time(), file_fp[FP_MTIME_COLUMN],
213
## file_fp[FP_CTIME_COLUMN])
200
214
self.update_count += 1
201
215
self.needs_write = True
202
216
self._cache[path] = (digest, file_fp)
219
def _really_sha1_file(self, abspath, filters):
220
"""Calculate the SHA1 of a file by reading the full text"""
221
return _mod_filters.internal_size_sha_file_byname(abspath, filters)[1]
206
224
"""Write contents of cache to file."""
207
outf = AtomicFile(self.cache_file_name(), 'wb', new_mode=self._mode)
225
outf = atomicfile.AtomicFile(self.cache_file_name(), 'wb',
209
print >>outf, CACHE_HEADER,
228
outf.write(CACHE_HEADER)
211
230
for path, c in self._cache.iteritems():
212
assert '//' not in path, path
213
outf.write(path.encode('utf-8'))
215
print >>outf, c[0], # hex sha1
217
print >>outf, "%d" % fld,
231
line_info = [path.encode('utf-8'), '// ', c[0], ' ']
232
line_info.append(' '.join([str(fld) for fld in c[1]]))
233
line_info.append('\n')
234
outf.write(''.join(line_info))
220
236
self.needs_write = False
221
mutter("write hash cache: %s hits=%d misses=%d stat=%d recent=%d updates=%d",
222
self.cache_file_name(), self.hit_count, self.miss_count,
224
self.danger_count, self.update_count)
237
## mutter("write hash cache: %s hits=%d misses=%d stat=%d recent=%d updates=%d",
238
## self.cache_file_name(), self.hit_count, self.miss_count,
240
## self.danger_count, self.update_count)
230
245
"""Reinstate cache from file.
232
247
Overwrites existing cache.
234
If the cache file has the wrong version marker, this just clears
249
If the cache file has the wrong version marker, this just clears
240
255
inf = file(fn, 'rb', buffering=65000)
241
256
except IOError, e:
242
mutter("failed to open %s: %s", fn, e)
257
trace.mutter("failed to open %s: %s", fn, e)
243
258
# better write it now so it is valid
244
259
self.needs_write = True
248
262
hdr = inf.readline()
249
263
if hdr != CACHE_HEADER:
250
mutter('cache header marker not found at top of %s;'
251
' discarding cache', fn)
264
trace.mutter('cache header marker not found at top of %s;'
265
' discarding cache', fn)
252
266
self.needs_write = True
256
270
pos = l.index('// ')
257
271
path = l[:pos].decode('utf-8')
258
272
if path in self._cache:
259
warning('duplicated path %r in cache' % path)
273
trace.warning('duplicated path %r in cache' % path)
263
277
fields = l[pos:].split(' ')
264
278
if len(fields) != 7:
265
warning("bad line in hashcache: %r" % l)
279
trace.warning("bad line in hashcache: %r" % l)
269
283
if len(sha1) != 40:
270
warning("bad sha1 in hashcache: %r" % sha1)
284
trace.warning("bad sha1 in hashcache: %r" % sha1)
273
287
fp = tuple(map(long, fields[1:]))
275
289
self._cache[path] = (sha1, fp)
277
291
self.needs_write = False
293
def _cutoff_time(self):
294
"""Return cutoff time.
296
Files modified more recently than this time are at risk of being
297
undetectably modified and so can't be cached.
299
return int(time.time()) - 3
301
def _fingerprint(self, abspath, stat_value=None):
302
if stat_value is None:
304
stat_value = os.lstat(abspath)
306
# might be missing, etc
308
if stat.S_ISDIR(stat_value.st_mode):
310
# we discard any high precision because it's not reliable; perhaps we
311
# could do better on some systems?
312
return (stat_value.st_size, long(stat_value.st_mtime),
313
long(stat_value.st_ctime), stat_value.st_ino,
314
stat_value.st_dev, stat_value.st_mode)
added
removed
bzrlib/hashcache.py
