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# out-of-date.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# out-of-date.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"
32
32
import os, stat, time
35
from bzrlib.osutils import sha_file, pathjoin, safe_unicode
34
from bzrlib.filters import internal_size_sha_file_byname
35
from bzrlib.osutils import sha_file, sha_string, pathjoin, safe_unicode
36
36
from bzrlib.trace import mutter, warning
37
37
from bzrlib.atomicfile import AtomicFile
38
38
from bzrlib.errors import BzrError
42
42
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
47
class HashCache(object):
89
75
number of times files have been retrieved from the cache, avoiding a
93
79
number of misses (times files have been completely re-read)
95
81
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."""
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.
99
92
self.root = safe_unicode(root)
93
self.root_utf8 = self.root.encode('utf8') # where is the filesystem encoding ?
100
94
self.hit_count = 0
101
95
self.miss_count = 0
102
96
self.stat_count = 0
122
117
"""Scan all files and remove entries where the cache entry is obsolete.
124
119
Obsolete entries are those where the file has been modified or deleted
125
since the entry was inserted.
120
since the entry was inserted.
127
122
# FIXME optimisation opportunity, on linux [and check other oses]:
128
123
# rather than iteritems order, stat in inode order.
129
124
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
132
127
for inum, path, cache_entry in prep:
133
128
abspath = pathjoin(self.root, path)
134
fp = _fingerprint(abspath)
129
fp = self._fingerprint(abspath)
135
130
self.stat_count += 1
137
132
cache_fp = cache_entry[1]
139
134
if (not fp) or (cache_fp != fp):
140
135
# not here or not a regular file anymore
141
136
self.removed_count += 1
142
137
self.needs_write = True
143
138
del self._cache[path]
146
def get_sha1(self, path):
140
def get_sha1(self, path, stat_value=None):
147
141
"""Return the sha1 of a file.
149
abspath = pathjoin(self.root, path)
143
if path.__class__ is str:
144
abspath = pathjoin(self.root_utf8, path)
146
abspath = pathjoin(self.root, path)
150
147
self.stat_count += 1
151
file_fp = _fingerprint(abspath)
148
file_fp = self._fingerprint(abspath, stat_value)
154
151
# not a regular file or not existing
155
152
if path in self._cache:
156
153
self.removed_count += 1
157
154
self.needs_write = True
158
155
del self._cache[path]
161
158
if path in self._cache:
162
159
cache_sha1, cache_fp = self._cache[path]
164
161
cache_sha1, cache_fp = None, None
166
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())
167
166
self.hit_count += 1
168
167
return cache_sha1
170
169
self.miss_count += 1
173
171
mode = file_fp[FP_MODE_COLUMN]
174
172
if stat.S_ISREG(mode):
175
digest = sha_file(file(abspath, 'rb', buffering=65000))
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)
176
178
elif stat.S_ISLNK(mode):
177
digest = sha.new(os.readlink(abspath)).hexdigest()
179
digest = sha_string(os.readlink(abspath))
179
181
raise BzrError("file %r: unknown file stat mode: %o"%(abspath,mode))
181
now = int(time.time())
182
if file_fp[FP_MTIME_COLUMN] >= now or file_fp[FP_CTIME_COLUMN] >= now:
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:
183
188
# changed too recently; can't be cached. we can
184
189
# return the result and it could possibly be cached
191
196
# need to let sufficient time elapse before we may cache this entry
192
197
# again. If we didn't do this, then, for example, a very quick 1
193
198
# byte replacement in the file might go undetected.
194
self.danger_count += 1
199
## mutter('%r modified too recently; not caching', path)
200
self.danger_count += 1
196
202
self.removed_count += 1
197
203
self.needs_write = True
198
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])
200
209
self.update_count += 1
201
210
self.needs_write = True
202
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]
206
219
"""Write contents of cache to file."""
207
220
outf = AtomicFile(self.cache_file_name(), 'wb', new_mode=self._mode)
209
print >>outf, CACHE_HEADER,
222
outf.write(CACHE_HEADER)
211
224
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,
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))
220
230
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)
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)
230
239
"""Reinstate cache from file.
232
241
Overwrites existing cache.
234
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
275
283
self._cache[path] = (sha1, fp)
277
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)