14
14
# along with this program; if not, write to the Free Software
15
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# out-of-date.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
20
# it preserves the nice property that any caller will always get up-to-date
21
# data except in unavoidable cases.
23
# TODO: Perhaps return more details on the file to avoid statting it
24
# again: nonexistent, file type, size, etc
26
# TODO: Perhaps use a Python pickle instead of a text file; might be faster.
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
45
20
def _fingerprint(abspath):
47
24
fs = os.lstat(abspath)
93
70
number of misses (times files have been completely re-read)
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)
72
def __init__(self, basedir):
73
self.basedir = basedir
100
74
self.hit_count = 0
101
75
self.miss_count = 0
102
76
self.stat_count = 0
103
77
self.danger_count = 0
104
self.removed_count = 0
105
self.update_count = 0
108
self._cache_file_name = safe_unicode(cache_file_name)
110
def cache_file_name(self):
111
return self._cache_file_name
114
"""Discard all cached information.
116
This does not reset the counters."""
118
self.needs_write = True
122
"""Scan all files and remove entries where the cache entry is obsolete.
124
Obsolete entries are those where the file has been modified or deleted
125
since the entry was inserted.
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
cache_fp = cache_entry[1]
139
if (not fp) or (cache_fp != fp):
140
# not here or not a regular file anymore
141
self.removed_count += 1
142
self.needs_write = True
143
del self._cache[path]
83
"""Discard all cached information."""
146
88
def get_sha1(self, path):
147
"""Return the sha1 of a file.
89
"""Return the hex SHA-1 of the contents of the file at path.
91
XXX: If the file does not exist or is not a plain file???
149
abspath = pathjoin(self.root, path)
95
from bzrlib.osutils import sha_file
97
abspath = os.path.join(self.basedir, path)
98
fp = _fingerprint(abspath)
99
cache_fp = self.validator.get(path)
150
101
self.stat_count += 1
151
file_fp = _fingerprint(abspath)
154
# not a regular file or not existing
155
if path in self._cache:
156
self.removed_count += 1
157
self.needs_write = True
158
del self._cache[path]
161
if path in self._cache:
162
cache_sha1, cache_fp = self._cache[path]
164
cache_sha1, cache_fp = None, None
166
if cache_fp == file_fp:
106
elif cache_fp and (cache_fp == fp):
167
107
self.hit_count += 1
173
mode = file_fp[FP_MODE_COLUMN]
174
if stat.S_ISREG(mode):
175
digest = sha_file(file(abspath, 'rb', buffering=65000))
176
elif stat.S_ISLNK(mode):
177
digest = sha.new(os.readlink(abspath)).hexdigest()
179
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
# changed too recently; can't be cached. we can
184
# return the result and it could possibly be cached
187
# the point is that we only want to cache when we are sure that any
188
# subsequent modifications of the file can be detected. If a
189
# modification neither changes the inode, the device, the size, nor
190
# the mode, then we can only distinguish it by time; therefore we
191
# need to let sufficient time elapse before we may cache this entry
192
# again. If we didn't do this, then, for example, a very quick 1
193
# byte replacement in the file might go undetected.
194
self.danger_count += 1
196
self.removed_count += 1
197
self.needs_write = True
198
del self._cache[path]
200
self.update_count += 1
201
self.needs_write = True
202
self._cache[path] = (digest, file_fp)
206
"""Write contents of cache to file."""
207
outf = AtomicFile(self.cache_file_name(), 'wb', new_mode=self._mode)
209
print >>outf, CACHE_HEADER,
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,
220
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)
230
"""Reinstate cache from file.
232
Overwrites existing cache.
234
If the cache file has the wrong version marker, this just clears
238
fn = self.cache_file_name()
240
inf = file(fn, 'rb', buffering=65000)
242
mutter("failed to open %s: %s", fn, e)
243
# better write it now so it is valid
244
self.needs_write = True
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)
277
self.needs_write = False
108
return self.cache_sha1[path]
111
digest = sha_file(file(abspath, 'rb'))
113
now = int(time.time())
114
if fp[1] >= now or fp[2] >= now:
115
# changed too recently; can't be cached. we can
116
# return the result and it could possibly be cached
118
self.danger_count += 1
120
del self.validator[path]
121
del self.cache_sha1[path]
123
self.validator[path] = fp
124
self.cache_sha1[path] = digest