1
# Copyright (C) 2005, 2006 Canonical Ltd
1
# (C) 2005 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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
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# out-of-date.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
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# out-of-date.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"
39
filters as _mod_filters,
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
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)
51
59
class HashCache(object):
79
87
number of times files have been retrieved from the cache, avoiding a
83
91
number of misses (times files have been completely re-read)
85
93
needs_write = False
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 ?
95
def __init__(self, basedir):
96
self.basedir = basedir
99
98
self.miss_count = 0
100
99
self.stat_count = 0
117
115
self.needs_write = True
121
120
"""Scan all files and remove entries where the cache entry is obsolete.
123
122
Obsolete entries are those where the file has been modified or deleted
124
since the entry was inserted.
123
since the entry was inserted.
126
# FIXME optimisation opportunity, on linux [and check other oses]:
127
# rather than iteritems order, stat in inode order.
128
125
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
131
128
for inum, path, cache_entry in prep:
132
abspath = osutils.pathjoin(self.root, path)
133
fp = self._fingerprint(abspath)
129
abspath = pathjoin(self.basedir, path)
130
fp = _fingerprint(abspath)
134
131
self.stat_count += 1
136
133
cache_fp = cache_entry[1]
138
135
if (not fp) or (cache_fp != fp):
139
136
# not here or not a regular file anymore
140
137
self.removed_count += 1
141
138
self.needs_write = True
142
139
del self._cache[path]
144
def get_sha1(self, path, stat_value=None):
142
def get_sha1(self, path):
145
143
"""Return the sha1 of a file.
147
if path.__class__ is str:
148
abspath = osutils.pathjoin(self.root_utf8, path)
150
abspath = osutils.pathjoin(self.root, path)
145
abspath = pathjoin(self.basedir, path)
151
146
self.stat_count += 1
152
file_fp = self._fingerprint(abspath, stat_value)
147
file_fp = _fingerprint(abspath)
155
150
# not a regular file or not existing
156
151
if path in self._cache:
157
152
self.removed_count += 1
158
153
self.needs_write = True
159
154
del self._cache[path]
162
157
if path in self._cache:
163
158
cache_sha1, cache_fp = self._cache[path]
165
160
cache_sha1, cache_fp = None, None
167
162
if cache_fp == file_fp:
168
## mutter("hashcache hit for %s %r -> %s", path, file_fp, cache_sha1)
169
## mutter("now = %s", time.time())
170
163
self.hit_count += 1
171
164
return cache_sha1
173
166
self.miss_count += 1
175
169
mode = file_fp[FP_MODE_COLUMN]
176
170
if stat.S_ISREG(mode):
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)
171
digest = sha_file(file(abspath, 'rb', buffering=65000))
182
172
elif stat.S_ISLNK(mode):
183
target = osutils.readlink(osutils.safe_unicode(abspath))
184
digest = osutils.sha_string(target.encode('UTF-8'))
173
link_target = os.readlink(abspath)
174
digest = sha.new(os.readlink(abspath)).hexdigest()
186
raise errors.BzrError("file %r: unknown file stat mode: %o"
176
raise BzrError("file %r: unknown file stat mode: %o"%(abspath,mode))
189
# window of 3 seconds to allow for 2s resolution on windows,
190
# unsynchronized file servers, etc.
191
cutoff = self._cutoff_time()
192
if file_fp[FP_MTIME_COLUMN] >= cutoff \
193
or file_fp[FP_CTIME_COLUMN] >= cutoff:
178
now = int(time.time())
179
if file_fp[1] >= now or file_fp[2] >= now:
194
180
# changed too recently; can't be cached. we can
195
181
# return the result and it could possibly be cached
198
# the point is that we only want to cache when we are sure that any
199
# subsequent modifications of the file can be detected. If a
200
# modification neither changes the inode, the device, the size, nor
201
# the mode, then we can only distinguish it by time; therefore we
202
# need to let sufficient time elapse before we may cache this entry
203
# again. If we didn't do this, then, for example, a very quick 1
204
# byte replacement in the file might go undetected.
205
## mutter('%r modified too recently; not caching', path)
206
self.danger_count += 1
183
self.danger_count += 1
208
185
self.removed_count += 1
209
186
self.needs_write = True
210
187
del self._cache[path]
212
## mutter('%r added to cache: now=%f, mtime=%d, ctime=%d',
213
## path, time.time(), file_fp[FP_MTIME_COLUMN],
214
## file_fp[FP_CTIME_COLUMN])
215
189
self.update_count += 1
216
190
self.needs_write = True
217
191
self._cache[path] = (digest, file_fp)
220
def _really_sha1_file(self, abspath, filters):
221
"""Calculate the SHA1 of a file by reading the full text"""
222
return _mod_filters.internal_size_sha_file_byname(abspath, filters)[1]
225
195
"""Write contents of cache to file."""
226
outf = atomicfile.AtomicFile(self.cache_file_name(), 'wb',
196
outf = AtomicFile(self.cache_file_name(), 'wb')
229
outf.write(CACHE_HEADER)
198
print >>outf, CACHE_HEADER,
231
200
for path, c in self._cache.iteritems():
232
line_info = [path.encode('utf-8'), '// ', c[0], ' ']
233
line_info.append(' '.join([str(fld) for fld in c[1]]))
234
line_info.append('\n')
235
outf.write(''.join(line_info))
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,
237
210
self.needs_write = False
238
## mutter("write hash cache: %s hits=%d misses=%d stat=%d recent=%d updates=%d",
239
## self.cache_file_name(), self.hit_count, self.miss_count,
241
## self.danger_count, self.update_count)
246
216
"""Reinstate cache from file.
248
218
Overwrites existing cache.
250
If the cache file has the wrong version marker, this just clears
220
If the cache file has the wrong version marker, this just clears
290
261
self._cache[path] = (sha1, fp)
292
263
self.needs_write = False
294
def _cutoff_time(self):
295
"""Return cutoff time.
297
Files modified more recently than this time are at risk of being
298
undetectably modified and so can't be cached.
300
return int(time.time()) - 3
302
def _fingerprint(self, abspath, stat_value=None):
303
if stat_value is None:
305
stat_value = os.lstat(abspath)
307
# might be missing, etc
309
if stat.S_ISDIR(stat_value.st_mode):
311
# we discard any high precision because it's not reliable; perhaps we
312
# could do better on some systems?
313
return (stat_value.st_size, long(stat_value.st_mtime),
314
long(stat_value.st_ctime), stat_value.st_ino,
315
stat_value.st_dev, stat_value.st_mode)