~bzr-pqm/bzr/bzr.dev

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
# (C) 2005 Canonical Ltd

# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2 of the License, or
# (at your option) any later version.

# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU General Public License for more details.

# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA

# TODO: Up-front, stat all files in order and remove those which are deleted or 
# out-of-date.  Don't actually re-read them until they're needed.  That ought 
# to bring all the inodes into core so that future stats to them are fast, and 
# it preserves the nice property that any caller will always get up-to-date
# data except in unavoidable cases.

# TODO: Perhaps return more details on the file to avoid statting it
# again: nonexistent, file type, size, etc

# TODO: Perhaps use a Python pickle instead of a text file; might be faster.



CACHE_HEADER = "### bzr hashcache v5\n"

import os, stat, time
import sha

from bzrlib.osutils import sha_file, pathjoin, safe_unicode
from bzrlib.trace import mutter, warning
from bzrlib.atomicfile import AtomicFile
from bzrlib.errors import BzrError


FP_MTIME_COLUMN = 1
FP_CTIME_COLUMN = 2
FP_MODE_COLUMN = 5

def _fingerprint(abspath):
    try:
        fs = os.lstat(abspath)
    except OSError:
        # might be missing, etc
        return None

    if stat.S_ISDIR(fs.st_mode):
        return None

    # we discard any high precision because it's not reliable; perhaps we
    # could do better on some systems?
    return (fs.st_size, long(fs.st_mtime),
            long(fs.st_ctime), fs.st_ino, fs.st_dev, fs.st_mode)


class HashCache(object):
    """Cache for looking up file SHA-1.

    Files are considered to match the cached value if the fingerprint
    of the file has not changed.  This includes its mtime, ctime,
    device number, inode number, and size.  This should catch
    modifications or replacement of the file by a new one.

    This may not catch modifications that do not change the file's
    size and that occur within the resolution window of the
    timestamps.  To handle this we specifically do not cache files
    which have changed since the start of the present second, since
    they could undetectably change again.

    This scheme may fail if the machine's clock steps backwards.
    Don't do that.

    This does not canonicalize the paths passed in; that should be
    done by the caller.

    _cache
        Indexed by path, points to a two-tuple of the SHA-1 of the file.
        and its fingerprint.

    stat_count
        number of times files have been statted

    hit_count
        number of times files have been retrieved from the cache, avoiding a
        re-read
        
    miss_count
        number of misses (times files have been completely re-read)
    """
    needs_write = False

    def __init__(self, root, cache_file_name, mode=None):
        """Create a hash cache in base dir, and set the file mode to mode."""
        self.root = safe_unicode(root)
        self.hit_count = 0
        self.miss_count = 0
        self.stat_count = 0
        self.danger_count = 0
        self.removed_count = 0
        self.update_count = 0
        self._cache = {}
        self._mode = mode
        self._cache_file_name = safe_unicode(cache_file_name)

    def cache_file_name(self):
        return self._cache_file_name

    def clear(self):
        """Discard all cached information.

        This does not reset the counters."""
        if self._cache:
            self.needs_write = True
            self._cache = {}

    def scan(self):
        """Scan all files and remove entries where the cache entry is obsolete.
        
        Obsolete entries are those where the file has been modified or deleted
        since the entry was inserted.        
        """
        # FIXME optimisation opportunity, on linux [and check other oses]:
        # rather than iteritems order, stat in inode order.
        prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
        prep.sort()
        
        for inum, path, cache_entry in prep:
            abspath = pathjoin(self.root, path)
            fp = _fingerprint(abspath)
            self.stat_count += 1
            
            cache_fp = cache_entry[1]
    
            if (not fp) or (cache_fp != fp):
                # not here or not a regular file anymore
                self.removed_count += 1
                self.needs_write = True
                del self._cache[path]


    def get_sha1(self, path):
        """Return the sha1 of a file.
        """
        abspath = pathjoin(self.root, path)
        self.stat_count += 1
        file_fp = _fingerprint(abspath)
        
        if not file_fp:
            # not a regular file or not existing
            if path in self._cache:
                self.removed_count += 1
                self.needs_write = True
                del self._cache[path]
            return None        

        if path in self._cache:
            cache_sha1, cache_fp = self._cache[path]
        else:
            cache_sha1, cache_fp = None, None

        if cache_fp == file_fp:
            self.hit_count += 1
            return cache_sha1
        
        self.miss_count += 1


        mode = file_fp[FP_MODE_COLUMN]
        if stat.S_ISREG(mode):
            digest = sha_file(file(abspath, 'rb', buffering=65000))
        elif stat.S_ISLNK(mode):
            digest = sha.new(os.readlink(abspath)).hexdigest()
        else:
            raise BzrError("file %r: unknown file stat mode: %o"%(abspath,mode))

        now = int(time.time())
        if file_fp[FP_MTIME_COLUMN] >= now or file_fp[FP_CTIME_COLUMN] >= now:
            # changed too recently; can't be cached.  we can
            # return the result and it could possibly be cached
            # next time.
            #
            # the point is that we only want to cache when we are sure that any
            # subsequent modifications of the file can be detected.  If a
            # modification neither changes the inode, the device, the size, nor
            # the mode, then we can only distinguish it by time; therefore we
            # need to let sufficient time elapse before we may cache this entry
            # again.  If we didn't do this, then, for example, a very quick 1
            # byte replacement in the file might go undetected.
            self.danger_count += 1 
            if cache_fp:
                self.removed_count += 1
                self.needs_write = True
                del self._cache[path]
        else:
            self.update_count += 1
            self.needs_write = True
            self._cache[path] = (digest, file_fp)
        return digest
        
    def write(self):
        """Write contents of cache to file."""
        outf = AtomicFile(self.cache_file_name(), 'wb', new_mode=self._mode)
        try:
            print >>outf, CACHE_HEADER,

            for path, c  in self._cache.iteritems():
                assert '//' not in path, path
                outf.write(path.encode('utf-8'))
                outf.write('// ')
                print >>outf, c[0],     # hex sha1
                for fld in c[1]:
                    print >>outf, "%d" % fld,
                print >>outf

            outf.commit()
            self.needs_write = False
        finally:
            if not outf.closed:
                outf.abort()

    def read(self):
        """Reinstate cache from file.

        Overwrites existing cache.

        If the cache file has the wrong version marker, this just clears 
        the cache."""
        self._cache = {}

        fn = self.cache_file_name()
        try:
            inf = file(fn, 'rb', buffering=65000)
        except IOError, e:
            mutter("failed to open %s: %s", fn, e)
            # better write it now so it is valid
            self.needs_write = True
            return


        hdr = inf.readline()
        if hdr != CACHE_HEADER:
            mutter('cache header marker not found at top of %s;'
                   ' discarding cache', fn)
            self.needs_write = True
            return

        for l in inf:
            pos = l.index('// ')
            path = l[:pos].decode('utf-8')
            if path in self._cache:
                warning('duplicated path %r in cache' % path)
                continue

            pos += 3
            fields = l[pos:].split(' ')
            if len(fields) != 7:
                warning("bad line in hashcache: %r" % l)
                continue

            sha1 = fields[0]
            if len(sha1) != 40:
                warning("bad sha1 in hashcache: %r" % sha1)
                continue

            fp = tuple(map(long, fields[1:]))

            self._cache[path] = (sha1, fp)

        self.needs_write = False