Show More
@@ -1,443 +1,448 b'' | |||
|
1 | 1 | # store.py - repository store handling for Mercurial |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2008 Matt Mackall <mpm@selenic.com> |
|
4 | 4 | # |
|
5 | 5 | # This software may be used and distributed according to the terms of the |
|
6 | 6 | # GNU General Public License version 2 or any later version. |
|
7 | 7 | |
|
8 | 8 | from i18n import _ |
|
9 | 9 | import osutil, scmutil, util |
|
10 | 10 | import os, stat, errno |
|
11 | 11 | |
|
12 | 12 | _sha = util.sha1 |
|
13 | 13 | |
|
14 | 14 | # This avoids a collision between a file named foo and a dir named |
|
15 | 15 | # foo.i or foo.d |
|
16 | 16 | def encodedir(path): |
|
17 | 17 | ''' |
|
18 | 18 | >>> encodedir('data/foo.i') |
|
19 | 19 | 'data/foo.i' |
|
20 | 20 | >>> encodedir('data/foo.i/bla.i') |
|
21 | 21 | 'data/foo.i.hg/bla.i' |
|
22 | 22 | >>> encodedir('data/foo.i.hg/bla.i') |
|
23 | 23 | 'data/foo.i.hg.hg/bla.i' |
|
24 | 24 | ''' |
|
25 | 25 | return (path |
|
26 | 26 | .replace(".hg/", ".hg.hg/") |
|
27 | 27 | .replace(".i/", ".i.hg/") |
|
28 | 28 | .replace(".d/", ".d.hg/")) |
|
29 | 29 | |
|
30 | 30 | def decodedir(path): |
|
31 | 31 | ''' |
|
32 | 32 | >>> decodedir('data/foo.i') |
|
33 | 33 | 'data/foo.i' |
|
34 | 34 | >>> decodedir('data/foo.i.hg/bla.i') |
|
35 | 35 | 'data/foo.i/bla.i' |
|
36 | 36 | >>> decodedir('data/foo.i.hg.hg/bla.i') |
|
37 | 37 | 'data/foo.i.hg/bla.i' |
|
38 | 38 | ''' |
|
39 | 39 | if ".hg/" not in path: |
|
40 | 40 | return path |
|
41 | 41 | return (path |
|
42 | 42 | .replace(".d.hg/", ".d/") |
|
43 | 43 | .replace(".i.hg/", ".i/") |
|
44 | 44 | .replace(".hg.hg/", ".hg/")) |
|
45 | 45 | |
|
46 | 46 | def _buildencodefun(): |
|
47 | 47 | ''' |
|
48 | 48 | >>> enc, dec = _buildencodefun() |
|
49 | 49 | |
|
50 | 50 | >>> enc('nothing/special.txt') |
|
51 | 51 | 'nothing/special.txt' |
|
52 | 52 | >>> dec('nothing/special.txt') |
|
53 | 53 | 'nothing/special.txt' |
|
54 | 54 | |
|
55 | 55 | >>> enc('HELLO') |
|
56 | 56 | '_h_e_l_l_o' |
|
57 | 57 | >>> dec('_h_e_l_l_o') |
|
58 | 58 | 'HELLO' |
|
59 | 59 | |
|
60 | 60 | >>> enc('hello:world?') |
|
61 | 61 | 'hello~3aworld~3f' |
|
62 | 62 | >>> dec('hello~3aworld~3f') |
|
63 | 63 | 'hello:world?' |
|
64 | 64 | |
|
65 | 65 | >>> enc('the\x07quick\xADshot') |
|
66 | 66 | 'the~07quick~adshot' |
|
67 | 67 | >>> dec('the~07quick~adshot') |
|
68 | 68 | 'the\\x07quick\\xadshot' |
|
69 | 69 | ''' |
|
70 | 70 | e = '_' |
|
71 | 71 | winreserved = [ord(x) for x in '\\:*?"<>|'] |
|
72 | 72 | cmap = dict([(chr(x), chr(x)) for x in xrange(127)]) |
|
73 | 73 | for x in (range(32) + range(126, 256) + winreserved): |
|
74 | 74 | cmap[chr(x)] = "~%02x" % x |
|
75 | 75 | for x in range(ord("A"), ord("Z")+1) + [ord(e)]: |
|
76 | 76 | cmap[chr(x)] = e + chr(x).lower() |
|
77 | 77 | dmap = {} |
|
78 | 78 | for k, v in cmap.iteritems(): |
|
79 | 79 | dmap[v] = k |
|
80 | 80 | def decode(s): |
|
81 | 81 | i = 0 |
|
82 | 82 | while i < len(s): |
|
83 | 83 | for l in xrange(1, 4): |
|
84 | 84 | try: |
|
85 | 85 | yield dmap[s[i:i + l]] |
|
86 | 86 | i += l |
|
87 | 87 | break |
|
88 | 88 | except KeyError: |
|
89 | 89 | pass |
|
90 | 90 | else: |
|
91 | 91 | raise KeyError |
|
92 | 92 | return (lambda s: "".join([cmap[c] for c in encodedir(s)]), |
|
93 | 93 | lambda s: decodedir("".join(list(decode(s))))) |
|
94 | 94 | |
|
95 | 95 | encodefilename, decodefilename = _buildencodefun() |
|
96 | 96 | |
|
97 | 97 | def _buildlowerencodefun(): |
|
98 | 98 | ''' |
|
99 | 99 | >>> f = _buildlowerencodefun() |
|
100 | 100 | >>> f('nothing/special.txt') |
|
101 | 101 | 'nothing/special.txt' |
|
102 | 102 | >>> f('HELLO') |
|
103 | 103 | 'hello' |
|
104 | 104 | >>> f('hello:world?') |
|
105 | 105 | 'hello~3aworld~3f' |
|
106 | 106 | >>> f('the\x07quick\xADshot') |
|
107 | 107 | 'the~07quick~adshot' |
|
108 | 108 | ''' |
|
109 | 109 | winreserved = [ord(x) for x in '\\:*?"<>|'] |
|
110 | 110 | cmap = dict([(chr(x), chr(x)) for x in xrange(127)]) |
|
111 | 111 | for x in (range(32) + range(126, 256) + winreserved): |
|
112 | 112 | cmap[chr(x)] = "~%02x" % x |
|
113 | 113 | for x in range(ord("A"), ord("Z")+1): |
|
114 | 114 | cmap[chr(x)] = chr(x).lower() |
|
115 | 115 | return lambda s: "".join([cmap[c] for c in s]) |
|
116 | 116 | |
|
117 | 117 | lowerencode = _buildlowerencodefun() |
|
118 | 118 | |
|
119 | 119 | # Windows reserved names: con, prn, aux, nul, com1..com9, lpt1..lpt9 |
|
120 | 120 | _winres3 = ('aux', 'con', 'prn', 'nul') # length 3 |
|
121 | 121 | _winres4 = ('com', 'lpt') # length 4 (with trailing 1..9) |
|
122 | 122 | def _auxencode(path, dotencode): |
|
123 | 123 | ''' |
|
124 | 124 | Encodes filenames containing names reserved by Windows or which end in |
|
125 | 125 | period or space. Does not touch other single reserved characters c. |
|
126 | 126 | Specifically, c in '\\:*?"<>|' or ord(c) <= 31 are *not* encoded here. |
|
127 | 127 | Additionally encodes space or period at the beginning, if dotencode is |
|
128 | 128 | True. Parameter path is assumed to be all lowercase. |
|
129 | 129 | A segment only needs encoding if a reserved name appears as a |
|
130 | 130 | basename (e.g. "aux", "aux.foo"). A directory or file named "foo.aux" |
|
131 | 131 | doesn't need encoding. |
|
132 | 132 | |
|
133 | 133 | >>> _auxencode('.foo/aux.txt/txt.aux/con/prn/nul/foo.', True) |
|
134 | 134 | ['~2efoo', 'au~78.txt', 'txt.aux', 'co~6e', 'pr~6e', 'nu~6c', 'foo~2e'] |
|
135 | 135 | >>> _auxencode('.com1com2/lpt9.lpt4.lpt1/conprn/com0/lpt0/foo.', False) |
|
136 | 136 | ['.com1com2', 'lp~749.lpt4.lpt1', 'conprn', 'com0', 'lpt0', 'foo~2e'] |
|
137 | 137 | >>> _auxencode('foo. ', True) |
|
138 | 138 | ['foo.~20'] |
|
139 | 139 | >>> _auxencode(' .foo', True) |
|
140 | 140 | ['~20.foo'] |
|
141 | 141 | ''' |
|
142 | 142 | res = path.split('/') |
|
143 | 143 | for i, n in enumerate(res): |
|
144 | 144 | if not n: |
|
145 | 145 | continue |
|
146 | 146 | if dotencode and n[0] in '. ': |
|
147 | 147 | n = "~%02x" % ord(n[0]) + n[1:] |
|
148 | 148 | res[i] = n |
|
149 | 149 | else: |
|
150 | 150 | l = n.find('.') |
|
151 | 151 | if l == -1: |
|
152 | 152 | l = len(n) |
|
153 | 153 | if ((l == 3 and n[:3] in _winres3) or |
|
154 | 154 | (l == 4 and n[3] <= '9' and n[3] >= '1' |
|
155 | 155 | and n[:3] in _winres4)): |
|
156 | 156 | # encode third letter ('aux' -> 'au~78') |
|
157 | 157 | ec = "~%02x" % ord(n[2]) |
|
158 | 158 | n = n[0:2] + ec + n[3:] |
|
159 | 159 | res[i] = n |
|
160 | 160 | if n[-1] in '. ': |
|
161 | 161 | # encode last period or space ('foo...' -> 'foo..~2e') |
|
162 | 162 | res[i] = n[:-1] + "~%02x" % ord(n[-1]) |
|
163 | 163 | return res |
|
164 | 164 | |
|
165 | 165 | _maxstorepathlen = 120 |
|
166 | 166 | _dirprefixlen = 8 |
|
167 | 167 | _maxshortdirslen = 8 * (_dirprefixlen + 1) - 4 |
|
168 | 168 | def _hybridencode(path, auxencode): |
|
169 | 169 | '''encodes path with a length limit |
|
170 | 170 | |
|
171 | 171 | Encodes all paths that begin with 'data/', according to the following. |
|
172 | 172 | |
|
173 | 173 | Default encoding (reversible): |
|
174 | 174 | |
|
175 | 175 | Encodes all uppercase letters 'X' as '_x'. All reserved or illegal |
|
176 | 176 | characters are encoded as '~xx', where xx is the two digit hex code |
|
177 | 177 | of the character (see encodefilename). |
|
178 | 178 | Relevant path components consisting of Windows reserved filenames are |
|
179 | 179 | masked by encoding the third character ('aux' -> 'au~78', see auxencode). |
|
180 | 180 | |
|
181 | 181 | Hashed encoding (not reversible): |
|
182 | 182 | |
|
183 | 183 | If the default-encoded path is longer than _maxstorepathlen, a |
|
184 | 184 | non-reversible hybrid hashing of the path is done instead. |
|
185 | 185 | This encoding uses up to _dirprefixlen characters of all directory |
|
186 | 186 | levels of the lowerencoded path, but not more levels than can fit into |
|
187 | 187 | _maxshortdirslen. |
|
188 | 188 | Then follows the filler followed by the sha digest of the full path. |
|
189 | 189 | The filler is the beginning of the basename of the lowerencoded path |
|
190 | 190 | (the basename is everything after the last path separator). The filler |
|
191 | 191 | is as long as possible, filling in characters from the basename until |
|
192 | 192 | the encoded path has _maxstorepathlen characters (or all chars of the |
|
193 | 193 | basename have been taken). |
|
194 | 194 | The extension (e.g. '.i' or '.d') is preserved. |
|
195 | 195 | |
|
196 | 196 | The string 'data/' at the beginning is replaced with 'dh/', if the hashed |
|
197 | 197 | encoding was used. |
|
198 | 198 | ''' |
|
199 | 199 | res = '/'.join(auxencode(encodefilename(path))) |
|
200 | 200 | if len(res) > _maxstorepathlen: |
|
201 | 201 | path = encodedir(path) |
|
202 | 202 | digest = _sha(path).hexdigest() |
|
203 | 203 | parts = auxencode(lowerencode(path))[1:] |
|
204 | 204 | basename = parts[-1] |
|
205 | 205 | _root, ext = os.path.splitext(basename) |
|
206 | 206 | sdirs = [] |
|
207 | sdirslen = 0 | |
|
207 | 208 | for p in parts[:-1]: |
|
208 | 209 | d = p[:_dirprefixlen] |
|
209 | 210 | if d[-1] in '. ': |
|
210 | 211 | # Windows can't access dirs ending in period or space |
|
211 | 212 | d = d[:-1] + '_' |
|
212 | t = '/'.join(sdirs) + '/' + d | |
|
213 | if len(t) > _maxshortdirslen: | |
|
214 |
|
|
|
213 | if sdirslen == 0: | |
|
214 | t = len(d) | |
|
215 | else: | |
|
216 | t = sdirslen + 1 + len(d) | |
|
217 | if t > _maxshortdirslen: | |
|
218 | break | |
|
215 | 219 | sdirs.append(d) |
|
220 | sdirslen = t | |
|
216 | 221 | dirs = '/'.join(sdirs) |
|
217 | 222 | if len(dirs) > 0: |
|
218 | 223 | dirs += '/' |
|
219 | 224 | res = 'dh/' + dirs + digest + ext |
|
220 | 225 | spaceleft = _maxstorepathlen - len(res) |
|
221 | 226 | if spaceleft > 0: |
|
222 | 227 | filler = basename[:spaceleft] |
|
223 | 228 | res = 'dh/' + dirs + filler + digest + ext |
|
224 | 229 | return res |
|
225 | 230 | |
|
226 | 231 | def _calcmode(path): |
|
227 | 232 | try: |
|
228 | 233 | # files in .hg/ will be created using this mode |
|
229 | 234 | mode = os.stat(path).st_mode |
|
230 | 235 | # avoid some useless chmods |
|
231 | 236 | if (0777 & ~util.umask) == (0777 & mode): |
|
232 | 237 | mode = None |
|
233 | 238 | except OSError: |
|
234 | 239 | mode = None |
|
235 | 240 | return mode |
|
236 | 241 | |
|
237 | 242 | _data = ('data 00manifest.d 00manifest.i 00changelog.d 00changelog.i' |
|
238 | 243 | ' phaseroots obsstore') |
|
239 | 244 | |
|
240 | 245 | class basicstore(object): |
|
241 | 246 | '''base class for local repository stores''' |
|
242 | 247 | def __init__(self, path, openertype): |
|
243 | 248 | self.path = path |
|
244 | 249 | self.createmode = _calcmode(path) |
|
245 | 250 | op = openertype(self.path) |
|
246 | 251 | op.createmode = self.createmode |
|
247 | 252 | self.opener = scmutil.filteropener(op, encodedir) |
|
248 | 253 | |
|
249 | 254 | def join(self, f): |
|
250 | 255 | return self.path + '/' + encodedir(f) |
|
251 | 256 | |
|
252 | 257 | def _walk(self, relpath, recurse): |
|
253 | 258 | '''yields (unencoded, encoded, size)''' |
|
254 | 259 | path = self.path |
|
255 | 260 | if relpath: |
|
256 | 261 | path += '/' + relpath |
|
257 | 262 | striplen = len(self.path) + 1 |
|
258 | 263 | l = [] |
|
259 | 264 | if os.path.isdir(path): |
|
260 | 265 | visit = [path] |
|
261 | 266 | while visit: |
|
262 | 267 | p = visit.pop() |
|
263 | 268 | for f, kind, st in osutil.listdir(p, stat=True): |
|
264 | 269 | fp = p + '/' + f |
|
265 | 270 | if kind == stat.S_IFREG and f[-2:] in ('.d', '.i'): |
|
266 | 271 | n = util.pconvert(fp[striplen:]) |
|
267 | 272 | l.append((decodedir(n), n, st.st_size)) |
|
268 | 273 | elif kind == stat.S_IFDIR and recurse: |
|
269 | 274 | visit.append(fp) |
|
270 | 275 | l.sort() |
|
271 | 276 | return l |
|
272 | 277 | |
|
273 | 278 | def datafiles(self): |
|
274 | 279 | return self._walk('data', True) |
|
275 | 280 | |
|
276 | 281 | def walk(self): |
|
277 | 282 | '''yields (unencoded, encoded, size)''' |
|
278 | 283 | # yield data files first |
|
279 | 284 | for x in self.datafiles(): |
|
280 | 285 | yield x |
|
281 | 286 | # yield manifest before changelog |
|
282 | 287 | for x in reversed(self._walk('', False)): |
|
283 | 288 | yield x |
|
284 | 289 | |
|
285 | 290 | def copylist(self): |
|
286 | 291 | return ['requires'] + _data.split() |
|
287 | 292 | |
|
288 | 293 | def write(self): |
|
289 | 294 | pass |
|
290 | 295 | |
|
291 | 296 | class encodedstore(basicstore): |
|
292 | 297 | def __init__(self, path, openertype): |
|
293 | 298 | self.path = path + '/store' |
|
294 | 299 | self.createmode = _calcmode(self.path) |
|
295 | 300 | op = openertype(self.path) |
|
296 | 301 | op.createmode = self.createmode |
|
297 | 302 | self.opener = scmutil.filteropener(op, encodefilename) |
|
298 | 303 | |
|
299 | 304 | def datafiles(self): |
|
300 | 305 | for a, b, size in self._walk('data', True): |
|
301 | 306 | try: |
|
302 | 307 | a = decodefilename(a) |
|
303 | 308 | except KeyError: |
|
304 | 309 | a = None |
|
305 | 310 | yield a, b, size |
|
306 | 311 | |
|
307 | 312 | def join(self, f): |
|
308 | 313 | return self.path + '/' + encodefilename(f) |
|
309 | 314 | |
|
310 | 315 | def copylist(self): |
|
311 | 316 | return (['requires', '00changelog.i'] + |
|
312 | 317 | ['store/' + f for f in _data.split()]) |
|
313 | 318 | |
|
314 | 319 | class fncache(object): |
|
315 | 320 | # the filename used to be partially encoded |
|
316 | 321 | # hence the encodedir/decodedir dance |
|
317 | 322 | def __init__(self, opener): |
|
318 | 323 | self.opener = opener |
|
319 | 324 | self.entries = None |
|
320 | 325 | self._dirty = False |
|
321 | 326 | |
|
322 | 327 | def _load(self): |
|
323 | 328 | '''fill the entries from the fncache file''' |
|
324 | 329 | self._dirty = False |
|
325 | 330 | try: |
|
326 | 331 | fp = self.opener('fncache', mode='rb') |
|
327 | 332 | except IOError: |
|
328 | 333 | # skip nonexistent file |
|
329 | 334 | self.entries = set() |
|
330 | 335 | return |
|
331 | 336 | self.entries = set(map(decodedir, fp.read().splitlines())) |
|
332 | 337 | if '' in self.entries: |
|
333 | 338 | fp.seek(0) |
|
334 | 339 | for n, line in enumerate(fp): |
|
335 | 340 | if not line.rstrip('\n'): |
|
336 | 341 | t = _('invalid entry in fncache, line %s') % (n + 1) |
|
337 | 342 | raise util.Abort(t) |
|
338 | 343 | fp.close() |
|
339 | 344 | |
|
340 | 345 | def _write(self, files, atomictemp): |
|
341 | 346 | fp = self.opener('fncache', mode='wb', atomictemp=atomictemp) |
|
342 | 347 | if files: |
|
343 | 348 | fp.write('\n'.join(map(encodedir, files)) + '\n') |
|
344 | 349 | fp.close() |
|
345 | 350 | self._dirty = False |
|
346 | 351 | |
|
347 | 352 | def rewrite(self, files): |
|
348 | 353 | self._write(files, False) |
|
349 | 354 | self.entries = set(files) |
|
350 | 355 | |
|
351 | 356 | def write(self): |
|
352 | 357 | if self._dirty: |
|
353 | 358 | self._write(self.entries, True) |
|
354 | 359 | |
|
355 | 360 | def add(self, fn): |
|
356 | 361 | if self.entries is None: |
|
357 | 362 | self._load() |
|
358 | 363 | if fn not in self.entries: |
|
359 | 364 | self._dirty = True |
|
360 | 365 | self.entries.add(fn) |
|
361 | 366 | |
|
362 | 367 | def __contains__(self, fn): |
|
363 | 368 | if self.entries is None: |
|
364 | 369 | self._load() |
|
365 | 370 | return fn in self.entries |
|
366 | 371 | |
|
367 | 372 | def __iter__(self): |
|
368 | 373 | if self.entries is None: |
|
369 | 374 | self._load() |
|
370 | 375 | return iter(self.entries) |
|
371 | 376 | |
|
372 | 377 | class _fncacheopener(scmutil.abstractopener): |
|
373 | 378 | def __init__(self, op, fnc, encode): |
|
374 | 379 | self.opener = op |
|
375 | 380 | self.fncache = fnc |
|
376 | 381 | self.encode = encode |
|
377 | 382 | |
|
378 | 383 | def _getmustaudit(self): |
|
379 | 384 | return self.opener.mustaudit |
|
380 | 385 | |
|
381 | 386 | def _setmustaudit(self, onoff): |
|
382 | 387 | self.opener.mustaudit = onoff |
|
383 | 388 | |
|
384 | 389 | mustaudit = property(_getmustaudit, _setmustaudit) |
|
385 | 390 | |
|
386 | 391 | def __call__(self, path, mode='r', *args, **kw): |
|
387 | 392 | if mode not in ('r', 'rb') and path.startswith('data/'): |
|
388 | 393 | self.fncache.add(path) |
|
389 | 394 | return self.opener(self.encode(path), mode, *args, **kw) |
|
390 | 395 | |
|
391 | 396 | class fncachestore(basicstore): |
|
392 | 397 | def __init__(self, path, openertype, encode): |
|
393 | 398 | self.encode = encode |
|
394 | 399 | self.path = path + '/store' |
|
395 | 400 | self.pathsep = self.path + '/' |
|
396 | 401 | self.createmode = _calcmode(self.path) |
|
397 | 402 | op = openertype(self.path) |
|
398 | 403 | op.createmode = self.createmode |
|
399 | 404 | fnc = fncache(op) |
|
400 | 405 | self.fncache = fnc |
|
401 | 406 | self.opener = _fncacheopener(op, fnc, encode) |
|
402 | 407 | |
|
403 | 408 | def join(self, f): |
|
404 | 409 | return self.pathsep + self.encode(f) |
|
405 | 410 | |
|
406 | 411 | def getsize(self, path): |
|
407 | 412 | return os.stat(self.pathsep + path).st_size |
|
408 | 413 | |
|
409 | 414 | def datafiles(self): |
|
410 | 415 | rewrite = False |
|
411 | 416 | existing = [] |
|
412 | 417 | for f in sorted(self.fncache): |
|
413 | 418 | ef = self.encode(f) |
|
414 | 419 | try: |
|
415 | 420 | yield f, ef, self.getsize(ef) |
|
416 | 421 | existing.append(f) |
|
417 | 422 | except OSError, err: |
|
418 | 423 | if err.errno != errno.ENOENT: |
|
419 | 424 | raise |
|
420 | 425 | # nonexistent entry |
|
421 | 426 | rewrite = True |
|
422 | 427 | if rewrite: |
|
423 | 428 | # rewrite fncache to remove nonexistent entries |
|
424 | 429 | # (may be caused by rollback / strip) |
|
425 | 430 | self.fncache.rewrite(existing) |
|
426 | 431 | |
|
427 | 432 | def copylist(self): |
|
428 | 433 | d = ('data dh fncache phaseroots obsstore' |
|
429 | 434 | ' 00manifest.d 00manifest.i 00changelog.d 00changelog.i') |
|
430 | 435 | return (['requires', '00changelog.i'] + |
|
431 | 436 | ['store/' + f for f in d.split()]) |
|
432 | 437 | |
|
433 | 438 | def write(self): |
|
434 | 439 | self.fncache.write() |
|
435 | 440 | |
|
436 | 441 | def store(requirements, path, openertype): |
|
437 | 442 | if 'store' in requirements: |
|
438 | 443 | if 'fncache' in requirements: |
|
439 | 444 | auxencode = lambda f: _auxencode(f, 'dotencode' in requirements) |
|
440 | 445 | encode = lambda f: _hybridencode(f, auxencode) |
|
441 | 446 | return fncachestore(path, openertype, encode) |
|
442 | 447 | return encodedstore(path, openertype) |
|
443 | 448 | return basicstore(path, openertype) |
General Comments 0
You need to be logged in to leave comments.
Login now