1
# Copyright (C) 2005, 2006 by 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# out-of-date.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# out-of-date.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, pathjoin, safe_unicode
36
from bzrlib.trace import mutter, warning
37
from bzrlib.atomicfile import AtomicFile
38
from bzrlib.errors import BzrError
41
filters as _mod_filters,
41
47
FP_MTIME_COLUMN = 1
42
48
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
53
class HashCache(object):
89
81
number of times files have been retrieved from the cache, avoiding a
93
85
number of misses (times files have been completely re-read)
95
87
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)
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 ?
100
100
self.hit_count = 0
101
101
self.miss_count = 0
102
102
self.stat_count = 0
122
123
"""Scan all files and remove entries where the cache entry is obsolete.
124
125
Obsolete entries are those where the file has been modified or deleted
125
since the entry was inserted.
126
since the entry was inserted.
127
128
# FIXME optimisation opportunity, on linux [and check other oses]:
128
129
# rather than iteritems order, stat in inode order.
129
130
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
132
133
for inum, path, cache_entry in prep:
133
abspath = pathjoin(self.root, path)
134
fp = _fingerprint(abspath)
134
abspath = osutils.pathjoin(self.root, path)
135
fp = self._fingerprint(abspath)
135
136
self.stat_count += 1
137
138
cache_fp = cache_entry[1]
139
140
if (not fp) or (cache_fp != fp):
140
141
# not here or not a regular file anymore
141
142
self.removed_count += 1
142
143
self.needs_write = True
143
144
del self._cache[path]
146
def get_sha1(self, path):
146
def get_sha1(self, path, stat_value=None):
147
147
"""Return the sha1 of a file.
149
abspath = pathjoin(self.root, path)
149
if path.__class__ is str:
150
abspath = osutils.pathjoin(self.root_utf8, path)
152
abspath = osutils.pathjoin(self.root, path)
150
153
self.stat_count += 1
151
file_fp = _fingerprint(abspath)
154
file_fp = self._fingerprint(abspath, stat_value)
154
157
# not a regular file or not existing
155
158
if path in self._cache:
156
159
self.removed_count += 1
157
160
self.needs_write = True
158
161
del self._cache[path]
161
164
if path in self._cache:
162
165
cache_sha1, cache_fp = self._cache[path]
164
167
cache_sha1, cache_fp = None, None
166
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())
167
172
self.hit_count += 1
168
173
return cache_sha1
170
175
self.miss_count += 1
173
177
mode = file_fp[FP_MODE_COLUMN]
174
178
if stat.S_ISREG(mode):
175
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)
176
184
elif stat.S_ISLNK(mode):
177
digest = sha.new(os.readlink(abspath)).hexdigest()
185
target = osutils.readlink(osutils.safe_unicode(abspath))
186
digest = osutils.sha_string(target.encode('UTF-8'))
179
raise BzrError("file %r: unknown file stat mode: %o"%(abspath,mode))
188
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:
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:
183
196
# changed too recently; can't be cached. we can
184
197
# return the result and it could possibly be cached
191
204
# need to let sufficient time elapse before we may cache this entry
192
205
# again. If we didn't do this, then, for example, a very quick 1
193
206
# byte replacement in the file might go undetected.
194
self.danger_count += 1
207
## mutter('%r modified too recently; not caching', path)
208
self.danger_count += 1
196
210
self.removed_count += 1
197
211
self.needs_write = True
198
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])
200
217
self.update_count += 1
201
218
self.needs_write = True
202
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]
206
227
"""Write contents of cache to file."""
207
outf = AtomicFile(self.cache_file_name(), 'wb', new_mode=self._mode)
228
outf = atomicfile.AtomicFile(self.cache_file_name(), 'wb',
209
print >>outf, CACHE_HEADER,
231
outf.write(CACHE_HEADER)
211
233
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,
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))
220
239
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)
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)
230
248
"""Reinstate cache from file.
232
250
Overwrites existing cache.
234
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
240
258
inf = file(fn, 'rb', buffering=65000)
241
259
except IOError, e:
242
mutter("failed to open %s: %s", fn, e)
260
trace.mutter("failed to open %s: %s", fn, e)
243
261
# better write it now so it is valid
244
262
self.needs_write = True
248
265
hdr = inf.readline()
249
266
if hdr != CACHE_HEADER:
250
mutter('cache header marker not found at top of %s;'
251
' discarding cache', fn)
267
trace.mutter('cache header marker not found at top of %s;'
268
' discarding cache', fn)
252
269
self.needs_write = True
256
273
pos = l.index('// ')
257
274
path = l[:pos].decode('utf-8')
258
275
if path in self._cache:
259
warning('duplicated path %r in cache' % path)
276
trace.warning('duplicated path %r in cache' % path)
263
280
fields = l[pos:].split(' ')
264
281
if len(fields) != 7:
265
warning("bad line in hashcache: %r" % l)
282
trace.warning("bad line in hashcache: %r" % l)
269
286
if len(sha1) != 40:
270
warning("bad sha1 in hashcache: %r" % sha1)
287
trace.warning("bad sha1 in hashcache: %r" % sha1)
273
290
fp = tuple(map(long, fields[1:]))
275
292
self._cache[path] = (sha1, fp)
294
# GZ 2009-09-20: Should really use a try/finally block to ensure close
277
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
