1
# Copyright (C) 2005-2010 Canonical Ltd
1
# Copyright (C) 2005, 2006 by 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
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# out-of-date.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
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# out-of-date.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
22
20
# it preserves the nice property that any caller will always get up-to-date
23
21
# data except in unavoidable cases.
28
26
# TODO: Perhaps use a Python pickle instead of a text file; might be faster.
31
CACHE_HEADER = b"### bzr hashcache v5\n"
40
filters as _mod_filters,
30
CACHE_HEADER = "### bzr hashcache v5\n"
35
from bzrlib.osutils import sha_file, pathjoin, safe_unicode
36
from bzrlib.trace import mutter, warning
37
from bzrlib.atomicfile import AtomicFile
38
from bzrlib.errors import BzrError
50
41
FP_MTIME_COLUMN = 1
51
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)
55
61
class HashCache(object):
56
62
"""Cache for looking up file SHA-1.
83
89
number of times files have been retrieved from the cache, avoiding a
87
93
number of misses (times files have been completely re-read)
89
95
needs_write = False
91
def __init__(self, root, cache_file_name, mode=None,
92
content_filter_stack_provider=None):
93
"""Create a hash cache in base dir, and set the file mode to mode.
95
:param content_filter_stack_provider: a function that takes a
96
path (relative to the top of the tree) and a file-id as
97
parameters and returns a stack of ContentFilters.
98
If None, no content filtering is performed.
100
if not isinstance(root, text_type):
101
raise ValueError("Base dir for hashcache must be text")
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."""
99
self.root = safe_unicode(root)
103
100
self.hit_count = 0
104
101
self.miss_count = 0
105
102
self.stat_count = 0
126
122
"""Scan all files and remove entries where the cache entry is obsolete.
128
124
Obsolete entries are those where the file has been modified or deleted
129
since the entry was inserted.
125
since the entry was inserted.
131
# Stat in inode order as optimisation for at least linux.
132
def inode_order(path_and_cache):
133
return path_and_cache[1][1][3]
134
for path, cache_val in sorted(viewitems(self._cache), key=inode_order):
135
abspath = osutils.pathjoin(self.root, path)
136
fp = self._fingerprint(abspath)
127
# FIXME optimisation opportunity, on linux [and check other oses]:
128
# rather than iteritems order, stat in inode order.
129
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
132
for inum, path, cache_entry in prep:
133
abspath = pathjoin(self.root, path)
134
fp = _fingerprint(abspath)
137
135
self.stat_count += 1
139
if not fp or cache_val[1] != fp:
137
cache_fp = cache_entry[1]
139
if (not fp) or (cache_fp != fp):
140
140
# not here or not a regular file anymore
141
141
self.removed_count += 1
142
142
self.needs_write = True
143
143
del self._cache[path]
145
def get_sha1(self, path, stat_value=None):
146
def get_sha1(self, path):
146
147
"""Return the sha1 of a file.
148
abspath = osutils.pathjoin(self.root, path)
149
abspath = pathjoin(self.root, path)
149
150
self.stat_count += 1
150
file_fp = self._fingerprint(abspath, stat_value)
151
file_fp = _fingerprint(abspath)
153
154
# not a regular file or not existing
154
155
if path in self._cache:
155
156
self.removed_count += 1
156
157
self.needs_write = True
157
158
del self._cache[path]
160
161
if path in self._cache:
161
162
cache_sha1, cache_fp = self._cache[path]
165
166
if cache_fp == file_fp:
166
167
self.hit_count += 1
167
168
return cache_sha1
169
170
self.miss_count += 1
171
173
mode = file_fp[FP_MODE_COLUMN]
172
174
if stat.S_ISREG(mode):
173
if self._filter_provider is None:
176
filters = self._filter_provider(path=path)
177
digest = self._really_sha1_file(abspath, filters)
175
digest = sha_file(file(abspath, 'rb', buffering=65000))
178
176
elif stat.S_ISLNK(mode):
179
target = osutils.readlink(abspath)
180
digest = osutils.sha_string(target.encode('UTF-8'))
177
digest = sha.new(os.readlink(abspath)).hexdigest()
182
raise errors.BzrError("file %r: unknown file stat mode: %o"
179
raise BzrError("file %r: unknown file stat mode: %o"%(abspath,mode))
185
# window of 3 seconds to allow for 2s resolution on windows,
186
# unsynchronized file servers, etc.
187
cutoff = self._cutoff_time()
188
if file_fp[FP_MTIME_COLUMN] >= cutoff \
189
or file_fp[FP_CTIME_COLUMN] >= cutoff:
181
now = int(time.time())
182
if file_fp[FP_MTIME_COLUMN] >= now or file_fp[FP_CTIME_COLUMN] >= now:
190
183
# changed too recently; can't be cached. we can
191
184
# return the result and it could possibly be cached
198
191
# need to let sufficient time elapse before we may cache this entry
199
192
# again. If we didn't do this, then, for example, a very quick 1
200
193
# byte replacement in the file might go undetected.
201
## mutter('%r modified too recently; not caching', path)
202
self.danger_count += 1
194
self.danger_count += 1
204
196
self.removed_count += 1
205
197
self.needs_write = True
206
198
del self._cache[path]
208
# mutter('%r added to cache: now=%f, mtime=%d, ctime=%d',
209
## path, time.time(), file_fp[FP_MTIME_COLUMN],
210
# file_fp[FP_CTIME_COLUMN])
211
200
self.update_count += 1
212
201
self.needs_write = True
213
202
self._cache[path] = (digest, file_fp)
216
def _really_sha1_file(self, abspath, filters):
217
"""Calculate the SHA1 of a file by reading the full text"""
218
return _mod_filters.internal_size_sha_file_byname(abspath, filters)[1]
221
206
"""Write contents of cache to file."""
222
with atomicfile.AtomicFile(self.cache_file_name(), 'wb',
223
new_mode=self._mode) as outf:
224
outf.write(CACHE_HEADER)
207
outf = AtomicFile(self.cache_file_name(), 'wb', new_mode=self._mode)
209
print >>outf, CACHE_HEADER,
226
for path, c in viewitems(self._cache):
227
line_info = [path.encode('utf-8'), b'// ', c[0], b' ']
228
line_info.append(b'%d %d %d %d %d %d' % c[1])
229
line_info.append(b'\n')
230
outf.write(b''.join(line_info))
211
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,
231
220
self.needs_write = False
232
# mutter("write hash cache: %s hits=%d misses=%d stat=%d recent=%d updates=%d",
233
# self.cache_file_name(), self.hit_count, self.miss_count,
235
# self.danger_count, self.update_count)
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)
238
230
"""Reinstate cache from file.
240
232
Overwrites existing cache.
242
If the cache file has the wrong version marker, this just clears
234
If the cache file has the wrong version marker, this just clears
246
238
fn = self.cache_file_name()
248
inf = open(fn, 'rb', buffering=65000)
250
trace.mutter("failed to open %s: %s", fn, str(e))
240
inf = file(fn, 'rb', buffering=65000)
242
mutter("failed to open %s: %s", fn, e)
251
243
# better write it now so it is valid
252
244
self.needs_write = True
257
if hdr != CACHE_HEADER:
258
trace.mutter('cache header marker not found at top of %s;'
259
' discarding cache', fn)
260
self.needs_write = True
264
pos = l.index(b'// ')
265
path = l[:pos].decode('utf-8')
266
if path in self._cache:
267
trace.warning('duplicated path %r in cache' % path)
271
fields = l[pos:].split(b' ')
273
trace.warning("bad line in hashcache: %r" % l)
278
trace.warning("bad sha1 in hashcache: %r" % sha1)
281
fp = tuple(map(int, fields[1:]))
283
self._cache[path] = (sha1, fp)
249
if hdr != CACHE_HEADER:
250
mutter('cache header marker not found at top of %s;'
251
' discarding cache', fn)
252
self.needs_write = True
257
path = l[:pos].decode('utf-8')
258
if path in self._cache:
259
warning('duplicated path %r in cache' % path)
263
fields = l[pos:].split(' ')
265
warning("bad line in hashcache: %r" % l)
270
warning("bad sha1 in hashcache: %r" % sha1)
273
fp = tuple(map(long, fields[1:]))
275
self._cache[path] = (sha1, fp)
285
277
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, int(stat_value.st_mtime),
307
int(stat_value.st_ctime), stat_value.st_ino,
308
stat_value.st_dev, stat_value.st_mode)