1
# Copyright (C) 2005-2010 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
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
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# out-of-date.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
22
# it preserves the nice property that any caller will always get up-to-date
23
# data except in unavoidable cases.
25
# TODO: Perhaps return more details on the file to avoid statting it
26
# again: nonexistent, file type, size, etc
28
# TODO: Perhaps use a Python pickle instead of a text file; might be faster.
32
CACHE_HEADER = "### bzr hashcache v5\n"
41
filters as _mod_filters,
53
class HashCache(object):
54
"""Cache for looking up file SHA-1.
56
Files are considered to match the cached value if the fingerprint
57
of the file has not changed. This includes its mtime, ctime,
58
device number, inode number, and size. This should catch
59
modifications or replacement of the file by a new one.
61
This may not catch modifications that do not change the file's
62
size and that occur within the resolution window of the
63
timestamps. To handle this we specifically do not cache files
64
which have changed since the start of the present second, since
65
they could undetectably change again.
67
This scheme may fail if the machine's clock steps backwards.
70
This does not canonicalize the paths passed in; that should be
74
Indexed by path, points to a two-tuple of the SHA-1 of the file.
78
number of times files have been statted
81
number of times files have been retrieved from the cache, avoiding a
85
number of misses (times files have been completely re-read)
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 ?
103
self.danger_count = 0
104
self.removed_count = 0
105
self.update_count = 0
108
self._cache_file_name = osutils.safe_unicode(cache_file_name)
109
self._filter_provider = content_filter_stack_provider
111
def cache_file_name(self):
112
return self._cache_file_name
115
"""Discard all cached information.
117
This does not reset the counters."""
119
self.needs_write = True
123
"""Scan all files and remove entries where the cache entry is obsolete.
125
Obsolete entries are those where the file has been modified or deleted
126
since the entry was inserted.
128
# FIXME optimisation opportunity, on linux [and check other oses]:
129
# rather than iteritems order, stat in inode order.
130
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
133
for inum, path, cache_entry in prep:
134
abspath = osutils.pathjoin(self.root, path)
135
fp = self._fingerprint(abspath)
138
cache_fp = cache_entry[1]
140
if (not fp) or (cache_fp != fp):
141
# not here or not a regular file anymore
142
self.removed_count += 1
143
self.needs_write = True
144
del self._cache[path]
146
def get_sha1(self, path, stat_value=None):
147
"""Return the sha1 of a file.
149
if path.__class__ is str:
150
abspath = osutils.pathjoin(self.root_utf8, path)
152
abspath = osutils.pathjoin(self.root, path)
154
file_fp = self._fingerprint(abspath, stat_value)
157
# not a regular file or not existing
158
if path in self._cache:
159
self.removed_count += 1
160
self.needs_write = True
161
del self._cache[path]
164
if path in self._cache:
165
cache_sha1, cache_fp = self._cache[path]
167
cache_sha1, cache_fp = None, None
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())
177
mode = file_fp[FP_MODE_COLUMN]
178
if stat.S_ISREG(mode):
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)
184
elif stat.S_ISLNK(mode):
185
target = osutils.readlink(osutils.safe_unicode(abspath))
186
digest = osutils.sha_string(target.encode('UTF-8'))
188
raise errors.BzrError("file %r: unknown file stat mode: %o"
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:
196
# changed too recently; can't be cached. we can
197
# return the result and it could possibly be cached
200
# the point is that we only want to cache when we are sure that any
201
# subsequent modifications of the file can be detected. If a
202
# modification neither changes the inode, the device, the size, nor
203
# the mode, then we can only distinguish it by time; therefore we
204
# need to let sufficient time elapse before we may cache this entry
205
# again. If we didn't do this, then, for example, a very quick 1
206
# byte replacement in the file might go undetected.
207
## mutter('%r modified too recently; not caching', path)
208
self.danger_count += 1
210
self.removed_count += 1
211
self.needs_write = True
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])
217
self.update_count += 1
218
self.needs_write = True
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]
227
"""Write contents of cache to file."""
228
outf = atomicfile.AtomicFile(self.cache_file_name(), 'wb',
231
outf.write(CACHE_HEADER)
233
for path, c in self._cache.iteritems():
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))
239
self.needs_write = False
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)
248
"""Reinstate cache from file.
250
Overwrites existing cache.
252
If the cache file has the wrong version marker, this just clears
256
fn = self.cache_file_name()
258
inf = file(fn, 'rb', buffering=65000)
260
trace.mutter("failed to open %s: %s", fn, e)
261
# better write it now so it is valid
262
self.needs_write = True
266
if hdr != CACHE_HEADER:
267
trace.mutter('cache header marker not found at top of %s;'
268
' discarding cache', fn)
269
self.needs_write = True
274
path = l[:pos].decode('utf-8')
275
if path in self._cache:
276
trace.warning('duplicated path %r in cache' % path)
280
fields = l[pos:].split(' ')
282
trace.warning("bad line in hashcache: %r" % l)
287
trace.warning("bad sha1 in hashcache: %r" % sha1)
290
fp = tuple(map(long, fields[1:]))
292
self._cache[path] = (sha1, fp)
294
# GZ 2009-09-20: Should really use a try/finally block to ensure close
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
