1
# (C) 2005 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# out-of-date.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# out-of-date.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.
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.
28
30
CACHE_HEADER = "### bzr hashcache v5\n"
30
32
import os, stat, time
32
from bzrlib.osutils import sha_file
34
from bzrlib.filters import internal_size_sha_file_byname
35
from bzrlib.osutils import sha_file, sha_string, pathjoin, safe_unicode
33
36
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)
37
from bzrlib.atomicfile import AtomicFile
38
from bzrlib.errors import BzrError
53
47
class HashCache(object):
81
75
number of times files have been retrieved from the cache, avoiding a
85
79
number of misses (times files have been completely re-read)
87
81
needs_write = False
89
def __init__(self, basedir):
90
self.basedir = basedir
83
def __init__(self, root, cache_file_name, mode=None,
84
content_filter_stack_provider=None):
85
"""Create a hash cache in base dir, and set the file mode to mode.
87
:param content_filter_stack_provider: a function that takes a
88
path (relative to the top of the tree) and a file-id as
89
parameters and returns a stack of ContentFilters.
90
If None, no content filtering is performed.
92
self.root = safe_unicode(root)
93
self.root_utf8 = self.root.encode('utf8') # where is the filesystem encoding ?
92
95
self.miss_count = 0
93
96
self.stat_count = 0
111
113
self.needs_write = True
116
117
"""Scan all files and remove entries where the cache entry is obsolete.
118
119
Obsolete entries are those where the file has been modified or deleted
119
since the entry was inserted.
120
since the entry was inserted.
122
# FIXME optimisation opportunity, on linux [and check other oses]:
123
# rather than iteritems order, stat in inode order.
121
124
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
124
127
for inum, path, cache_entry in prep:
125
abspath = os.sep.join([self.basedir, path])
126
fp = _fingerprint(abspath)
128
abspath = pathjoin(self.root, path)
129
fp = self._fingerprint(abspath)
127
130
self.stat_count += 1
129
132
cache_fp = cache_entry[1]
131
134
if (not fp) or (cache_fp != fp):
132
135
# not here or not a regular file anymore
133
136
self.removed_count += 1
134
137
self.needs_write = True
135
138
del self._cache[path]
139
def get_sha1(self, path):
140
def get_sha1(self, path, stat_value=None):
140
141
"""Return the sha1 of a file.
142
abspath = os.sep.join([self.basedir, path])
143
if path.__class__ is str:
144
abspath = pathjoin(self.root_utf8, path)
146
abspath = pathjoin(self.root, path)
143
147
self.stat_count += 1
144
file_fp = _fingerprint(abspath)
148
file_fp = self._fingerprint(abspath, stat_value)
147
151
# not a regular file or not existing
148
152
if path in self._cache:
149
153
self.removed_count += 1
150
154
self.needs_write = True
151
155
del self._cache[path]
154
158
if path in self._cache:
155
159
cache_sha1, cache_fp = self._cache[path]
157
161
cache_sha1, cache_fp = None, None
159
163
if cache_fp == file_fp:
164
## mutter("hashcache hit for %s %r -> %s", path, file_fp, cache_sha1)
165
## mutter("now = %s", time.time())
160
166
self.hit_count += 1
161
167
return cache_sha1
163
169
self.miss_count += 1
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:
171
mode = file_fp[FP_MODE_COLUMN]
172
if stat.S_ISREG(mode):
173
if self._filter_provider is None:
176
filters = self._filter_provider(path=path, file_id=None)
177
digest = self._really_sha1_file(abspath, filters)
178
elif stat.S_ISLNK(mode):
179
digest = sha_string(os.readlink(abspath))
181
raise BzrError("file %r: unknown file stat mode: %o"%(abspath,mode))
183
# window of 3 seconds to allow for 2s resolution on windows,
184
# unsynchronized file servers, etc.
185
cutoff = self._cutoff_time()
186
if file_fp[FP_MTIME_COLUMN] >= cutoff \
187
or file_fp[FP_CTIME_COLUMN] >= cutoff:
168
188
# changed too recently; can't be cached. we can
169
189
# return the result and it could possibly be cached
171
self.danger_count += 1
192
# the point is that we only want to cache when we are sure that any
193
# subsequent modifications of the file can be detected. If a
194
# modification neither changes the inode, the device, the size, nor
195
# the mode, then we can only distinguish it by time; therefore we
196
# need to let sufficient time elapse before we may cache this entry
197
# again. If we didn't do this, then, for example, a very quick 1
198
# byte replacement in the file might go undetected.
199
## mutter('%r modified too recently; not caching', path)
200
self.danger_count += 1
173
202
self.removed_count += 1
174
203
self.needs_write = True
175
204
del self._cache[path]
206
## mutter('%r added to cache: now=%f, mtime=%d, ctime=%d',
207
## path, time.time(), file_fp[FP_MTIME_COLUMN],
208
## file_fp[FP_CTIME_COLUMN])
177
209
self.update_count += 1
178
210
self.needs_write = True
179
211
self._cache[path] = (digest, file_fp)
214
def _really_sha1_file(self, abspath, filters):
215
"""Calculate the SHA1 of a file by reading the full text"""
216
return internal_size_sha_file_byname(abspath, filters)[1]
187
219
"""Write contents of cache to file."""
188
from atomicfile import AtomicFile
190
outf = AtomicFile(self.cache_file_name(), 'wb')
220
outf = AtomicFile(self.cache_file_name(), 'wb', new_mode=self._mode)
192
print >>outf, CACHE_HEADER,
222
outf.write(CACHE_HEADER)
194
224
for path, c in self._cache.iteritems():
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,
225
line_info = [path.encode('utf-8'), '// ', c[0], ' ']
226
line_info.append(' '.join([str(fld) for fld in c[1]]))
227
line_info.append('\n')
228
outf.write(''.join(line_info))
204
230
self.needs_write = False
231
## mutter("write hash cache: %s hits=%d misses=%d stat=%d recent=%d updates=%d",
232
## self.cache_file_name(), self.hit_count, self.miss_count,
234
## self.danger_count, self.update_count)
212
239
"""Reinstate cache from file.
214
241
Overwrites existing cache.
216
If the cache file has the wrong version marker, this just clears
243
If the cache file has the wrong version marker, this just clears
254
283
self._cache[path] = (sha1, fp)
256
285
self.needs_write = False
287
def _cutoff_time(self):
288
"""Return cutoff time.
290
Files modified more recently than this time are at risk of being
291
undetectably modified and so can't be cached.
293
return int(time.time()) - 3
295
def _fingerprint(self, abspath, stat_value=None):
296
if stat_value is None:
298
stat_value = os.lstat(abspath)
300
# might be missing, etc
302
if stat.S_ISDIR(stat_value.st_mode):
304
# we discard any high precision because it's not reliable; perhaps we
305
# could do better on some systems?
306
return (stat_value.st_size, long(stat_value.st_mtime),
307
long(stat_value.st_ctime), stat_value.st_ino,
308
stat_value.st_dev, stat_value.st_mode)
added
removed
bzrlib/hashcache.py
