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# out-of-date.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 = b"### bzr hashcache v5\n"
41
filters as _mod_filters,
57
class HashCache(object):
58
"""Cache for looking up file SHA-1.
60
Files are considered to match the cached value if the fingerprint
61
of the file has not changed. This includes its mtime, ctime,
62
device number, inode number, and size. This should catch
63
modifications or replacement of the file by a new one.
65
This may not catch modifications that do not change the file's
66
size and that occur within the resolution window of the
67
timestamps. To handle this we specifically do not cache files
68
which have changed since the start of the present second, since
69
they could undetectably change again.
71
This scheme may fail if the machine's clock steps backwards.
74
This does not canonicalize the paths passed in; that should be
78
Indexed by path, points to a two-tuple of the SHA-1 of the file.
82
number of times files have been statted
85
number of times files have been retrieved from the cache, avoiding a
89
number of misses (times files have been completely re-read)
93
def __init__(self, root, cache_file_name, mode=None,
94
content_filter_stack_provider=None):
95
"""Create a hash cache in base dir, and set the file mode to mode.
97
:param content_filter_stack_provider: a function that takes a
98
path (relative to the top of the tree) and a file-id as
99
parameters and returns a stack of ContentFilters.
100
If None, no content filtering is performed.
102
if not isinstance(root, text_type):
103
raise ValueError("Base dir for hashcache must be text")
108
self.danger_count = 0
109
self.removed_count = 0
110
self.update_count = 0
113
self._cache_file_name = cache_file_name
114
self._filter_provider = content_filter_stack_provider
116
def cache_file_name(self):
117
return self._cache_file_name
120
"""Discard all cached information.
122
This does not reset the counters."""
124
self.needs_write = True
128
"""Scan all files and remove entries where the cache entry is obsolete.
130
Obsolete entries are those where the file has been modified or deleted
131
since the entry was inserted.
133
# Stat in inode order as optimisation for at least linux.
134
def inode_order(path_and_cache):
135
return path_and_cache[1][1][3]
136
for path, cache_val in sorted(viewitems(self._cache), key=inode_order):
137
abspath = osutils.pathjoin(self.root, path)
138
fp = self._fingerprint(abspath)
141
if not fp or cache_val[1] != fp:
142
# not here or not a regular file anymore
143
self.removed_count += 1
144
self.needs_write = True
145
del self._cache[path]
147
def get_sha1(self, path, stat_value=None):
148
"""Return the sha1 of a file.
150
abspath = osutils.pathjoin(self.root, path)
152
file_fp = self._fingerprint(abspath, stat_value)
155
# not a regular file or not existing
156
if path in self._cache:
157
self.removed_count += 1
158
self.needs_write = True
159
del self._cache[path]
162
if path in self._cache:
163
cache_sha1, cache_fp = self._cache[path]
165
cache_sha1, cache_fp = None, None
167
if cache_fp == file_fp:
168
## mutter("hashcache hit for %s %r -> %s", path, file_fp, cache_sha1)
169
## mutter("now = %s", time.time())
175
mode = file_fp[FP_MODE_COLUMN]
176
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)
182
elif stat.S_ISLNK(mode):
183
target = osutils.readlink(abspath)
184
digest = osutils.sha_string(target.encode('UTF-8'))
186
raise errors.BzrError("file %r: unknown file stat mode: %o"
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:
194
# changed too recently; can't be cached. we can
195
# 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
208
self.removed_count += 1
209
self.needs_write = True
210
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
self.update_count += 1
216
self.needs_write = True
217
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
"""Write contents of cache to file."""
226
outf = atomicfile.AtomicFile(self.cache_file_name(), 'wb',
229
outf.write(CACHE_HEADER)
231
for path, c in viewitems(self._cache):
232
line_info = [path.encode('utf-8'), b'// ', c[0], b' ']
233
line_info.append(b'%d %d %d %d %d %d' % c[1])
234
line_info.append(b'\n')
235
outf.write(b''.join(line_info))
237
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
"""Reinstate cache from file.
248
Overwrites existing cache.
250
If the cache file has the wrong version marker, this just clears
254
fn = self.cache_file_name()
256
inf = open(fn, 'rb', buffering=65000)
258
trace.mutter("failed to open %s: %s", fn, str(e))
259
# better write it now so it is valid
260
self.needs_write = True
265
if hdr != CACHE_HEADER:
266
trace.mutter('cache header marker not found at top of %s;'
267
' discarding cache', fn)
268
self.needs_write = True
272
pos = l.index(b'// ')
273
path = l[:pos].decode('utf-8')
274
if path in self._cache:
275
trace.warning('duplicated path %r in cache' % path)
279
fields = l[pos:].split(b' ')
281
trace.warning("bad line in hashcache: %r" % l)
286
trace.warning("bad sha1 in hashcache: %r" % sha1)
289
fp = tuple(map(int, fields[1:]))
291
self._cache[path] = (sha1, fp)
293
self.needs_write = False
295
def _cutoff_time(self):
296
"""Return cutoff time.
298
Files modified more recently than this time are at risk of being
299
undetectably modified and so can't be cached.
301
return int(time.time()) - 3
303
def _fingerprint(self, abspath, stat_value=None):
304
if stat_value is None:
306
stat_value = os.lstat(abspath)
308
# might be missing, etc
310
if stat.S_ISDIR(stat_value.st_mode):
312
# we discard any high precision because it's not reliable; perhaps we
313
# could do better on some systems?
314
return (stat_value.st_size, int(stat_value.st_mtime),
315
int(stat_value.st_ctime), stat_value.st_ino,
316
stat_value.st_dev, stat_value.st_mode)