|
1 # store.py - repository store handling for Mercurial |
|
2 # |
|
3 # Copyright 2008 Matt Mackall <mpm@selenic.com> |
|
4 # |
|
5 # This software may be used and distributed according to the terms of the |
|
6 # GNU General Public License version 2 or any later version. |
|
7 |
|
8 from i18n import _ |
|
9 import osutil, util |
|
10 import os, stat |
|
11 |
|
12 _sha = util.sha1 |
|
13 |
|
14 # This avoids a collision between a file named foo and a dir named |
|
15 # foo.i or foo.d |
|
16 def encodedir(path): |
|
17 if not path.startswith('data/'): |
|
18 return path |
|
19 return (path |
|
20 .replace(".hg/", ".hg.hg/") |
|
21 .replace(".i/", ".i.hg/") |
|
22 .replace(".d/", ".d.hg/")) |
|
23 |
|
24 def decodedir(path): |
|
25 if not path.startswith('data/') or ".hg/" not in path: |
|
26 return path |
|
27 return (path |
|
28 .replace(".d.hg/", ".d/") |
|
29 .replace(".i.hg/", ".i/") |
|
30 .replace(".hg.hg/", ".hg/")) |
|
31 |
|
32 def _buildencodefun(): |
|
33 e = '_' |
|
34 win_reserved = [ord(x) for x in '\\:*?"<>|'] |
|
35 cmap = dict([(chr(x), chr(x)) for x in xrange(127)]) |
|
36 for x in (range(32) + range(126, 256) + win_reserved): |
|
37 cmap[chr(x)] = "~%02x" % x |
|
38 for x in range(ord("A"), ord("Z")+1) + [ord(e)]: |
|
39 cmap[chr(x)] = e + chr(x).lower() |
|
40 dmap = {} |
|
41 for k, v in cmap.iteritems(): |
|
42 dmap[v] = k |
|
43 def decode(s): |
|
44 i = 0 |
|
45 while i < len(s): |
|
46 for l in xrange(1, 4): |
|
47 try: |
|
48 yield dmap[s[i:i + l]] |
|
49 i += l |
|
50 break |
|
51 except KeyError: |
|
52 pass |
|
53 else: |
|
54 raise KeyError |
|
55 return (lambda s: "".join([cmap[c] for c in encodedir(s)]), |
|
56 lambda s: decodedir("".join(list(decode(s))))) |
|
57 |
|
58 encodefilename, decodefilename = _buildencodefun() |
|
59 |
|
60 def _build_lower_encodefun(): |
|
61 win_reserved = [ord(x) for x in '\\:*?"<>|'] |
|
62 cmap = dict([(chr(x), chr(x)) for x in xrange(127)]) |
|
63 for x in (range(32) + range(126, 256) + win_reserved): |
|
64 cmap[chr(x)] = "~%02x" % x |
|
65 for x in range(ord("A"), ord("Z")+1): |
|
66 cmap[chr(x)] = chr(x).lower() |
|
67 return lambda s: "".join([cmap[c] for c in s]) |
|
68 |
|
69 lowerencode = _build_lower_encodefun() |
|
70 |
|
71 _windows_reserved_filenames = '''con prn aux nul |
|
72 com1 com2 com3 com4 com5 com6 com7 com8 com9 |
|
73 lpt1 lpt2 lpt3 lpt4 lpt5 lpt6 lpt7 lpt8 lpt9'''.split() |
|
74 def _auxencode(path, dotencode): |
|
75 res = [] |
|
76 for n in path.split('/'): |
|
77 if n: |
|
78 base = n.split('.')[0] |
|
79 if base and (base in _windows_reserved_filenames): |
|
80 # encode third letter ('aux' -> 'au~78') |
|
81 ec = "~%02x" % ord(n[2]) |
|
82 n = n[0:2] + ec + n[3:] |
|
83 if n[-1] in '. ': |
|
84 # encode last period or space ('foo...' -> 'foo..~2e') |
|
85 n = n[:-1] + "~%02x" % ord(n[-1]) |
|
86 if dotencode and n[0] in '. ': |
|
87 n = "~%02x" % ord(n[0]) + n[1:] |
|
88 res.append(n) |
|
89 return '/'.join(res) |
|
90 |
|
91 MAX_PATH_LEN_IN_HGSTORE = 120 |
|
92 DIR_PREFIX_LEN = 8 |
|
93 _MAX_SHORTENED_DIRS_LEN = 8 * (DIR_PREFIX_LEN + 1) - 4 |
|
94 def _hybridencode(path, auxencode): |
|
95 '''encodes path with a length limit |
|
96 |
|
97 Encodes all paths that begin with 'data/', according to the following. |
|
98 |
|
99 Default encoding (reversible): |
|
100 |
|
101 Encodes all uppercase letters 'X' as '_x'. All reserved or illegal |
|
102 characters are encoded as '~xx', where xx is the two digit hex code |
|
103 of the character (see encodefilename). |
|
104 Relevant path components consisting of Windows reserved filenames are |
|
105 masked by encoding the third character ('aux' -> 'au~78', see auxencode). |
|
106 |
|
107 Hashed encoding (not reversible): |
|
108 |
|
109 If the default-encoded path is longer than MAX_PATH_LEN_IN_HGSTORE, a |
|
110 non-reversible hybrid hashing of the path is done instead. |
|
111 This encoding uses up to DIR_PREFIX_LEN characters of all directory |
|
112 levels of the lowerencoded path, but not more levels than can fit into |
|
113 _MAX_SHORTENED_DIRS_LEN. |
|
114 Then follows the filler followed by the sha digest of the full path. |
|
115 The filler is the beginning of the basename of the lowerencoded path |
|
116 (the basename is everything after the last path separator). The filler |
|
117 is as long as possible, filling in characters from the basename until |
|
118 the encoded path has MAX_PATH_LEN_IN_HGSTORE characters (or all chars |
|
119 of the basename have been taken). |
|
120 The extension (e.g. '.i' or '.d') is preserved. |
|
121 |
|
122 The string 'data/' at the beginning is replaced with 'dh/', if the hashed |
|
123 encoding was used. |
|
124 ''' |
|
125 if not path.startswith('data/'): |
|
126 return path |
|
127 # escape directories ending with .i and .d |
|
128 path = encodedir(path) |
|
129 ndpath = path[len('data/'):] |
|
130 res = 'data/' + auxencode(encodefilename(ndpath)) |
|
131 if len(res) > MAX_PATH_LEN_IN_HGSTORE: |
|
132 digest = _sha(path).hexdigest() |
|
133 aep = auxencode(lowerencode(ndpath)) |
|
134 _root, ext = os.path.splitext(aep) |
|
135 parts = aep.split('/') |
|
136 basename = parts[-1] |
|
137 sdirs = [] |
|
138 for p in parts[:-1]: |
|
139 d = p[:DIR_PREFIX_LEN] |
|
140 if d[-1] in '. ': |
|
141 # Windows can't access dirs ending in period or space |
|
142 d = d[:-1] + '_' |
|
143 t = '/'.join(sdirs) + '/' + d |
|
144 if len(t) > _MAX_SHORTENED_DIRS_LEN: |
|
145 break |
|
146 sdirs.append(d) |
|
147 dirs = '/'.join(sdirs) |
|
148 if len(dirs) > 0: |
|
149 dirs += '/' |
|
150 res = 'dh/' + dirs + digest + ext |
|
151 space_left = MAX_PATH_LEN_IN_HGSTORE - len(res) |
|
152 if space_left > 0: |
|
153 filler = basename[:space_left] |
|
154 res = 'dh/' + dirs + filler + digest + ext |
|
155 return res |
|
156 |
|
157 def _calcmode(path): |
|
158 try: |
|
159 # files in .hg/ will be created using this mode |
|
160 mode = os.stat(path).st_mode |
|
161 # avoid some useless chmods |
|
162 if (0777 & ~util.umask) == (0777 & mode): |
|
163 mode = None |
|
164 except OSError: |
|
165 mode = None |
|
166 return mode |
|
167 |
|
168 _data = 'data 00manifest.d 00manifest.i 00changelog.d 00changelog.i' |
|
169 |
|
170 class basicstore(object): |
|
171 '''base class for local repository stores''' |
|
172 def __init__(self, path, opener, pathjoiner): |
|
173 self.pathjoiner = pathjoiner |
|
174 self.path = path |
|
175 self.createmode = _calcmode(path) |
|
176 op = opener(self.path) |
|
177 op.createmode = self.createmode |
|
178 self.opener = lambda f, *args, **kw: op(encodedir(f), *args, **kw) |
|
179 |
|
180 def join(self, f): |
|
181 return self.pathjoiner(self.path, encodedir(f)) |
|
182 |
|
183 def _walk(self, relpath, recurse): |
|
184 '''yields (unencoded, encoded, size)''' |
|
185 path = self.pathjoiner(self.path, relpath) |
|
186 striplen = len(self.path) + len(os.sep) |
|
187 l = [] |
|
188 if os.path.isdir(path): |
|
189 visit = [path] |
|
190 while visit: |
|
191 p = visit.pop() |
|
192 for f, kind, st in osutil.listdir(p, stat=True): |
|
193 fp = self.pathjoiner(p, f) |
|
194 if kind == stat.S_IFREG and f[-2:] in ('.d', '.i'): |
|
195 n = util.pconvert(fp[striplen:]) |
|
196 l.append((decodedir(n), n, st.st_size)) |
|
197 elif kind == stat.S_IFDIR and recurse: |
|
198 visit.append(fp) |
|
199 return sorted(l) |
|
200 |
|
201 def datafiles(self): |
|
202 return self._walk('data', True) |
|
203 |
|
204 def walk(self): |
|
205 '''yields (unencoded, encoded, size)''' |
|
206 # yield data files first |
|
207 for x in self.datafiles(): |
|
208 yield x |
|
209 # yield manifest before changelog |
|
210 for x in reversed(self._walk('', False)): |
|
211 yield x |
|
212 |
|
213 def copylist(self): |
|
214 return ['requires'] + _data.split() |
|
215 |
|
216 class encodedstore(basicstore): |
|
217 def __init__(self, path, opener, pathjoiner): |
|
218 self.pathjoiner = pathjoiner |
|
219 self.path = self.pathjoiner(path, 'store') |
|
220 self.createmode = _calcmode(self.path) |
|
221 op = opener(self.path) |
|
222 op.createmode = self.createmode |
|
223 self.opener = lambda f, *args, **kw: op(encodefilename(f), *args, **kw) |
|
224 |
|
225 def datafiles(self): |
|
226 for a, b, size in self._walk('data', True): |
|
227 try: |
|
228 a = decodefilename(a) |
|
229 except KeyError: |
|
230 a = None |
|
231 yield a, b, size |
|
232 |
|
233 def join(self, f): |
|
234 return self.pathjoiner(self.path, encodefilename(f)) |
|
235 |
|
236 def copylist(self): |
|
237 return (['requires', '00changelog.i'] + |
|
238 [self.pathjoiner('store', f) for f in _data.split()]) |
|
239 |
|
240 class fncache(object): |
|
241 # the filename used to be partially encoded |
|
242 # hence the encodedir/decodedir dance |
|
243 def __init__(self, opener): |
|
244 self.opener = opener |
|
245 self.entries = None |
|
246 |
|
247 def _load(self): |
|
248 '''fill the entries from the fncache file''' |
|
249 self.entries = set() |
|
250 try: |
|
251 fp = self.opener('fncache', mode='rb') |
|
252 except IOError: |
|
253 # skip nonexistent file |
|
254 return |
|
255 for n, line in enumerate(fp): |
|
256 if (len(line) < 2) or (line[-1] != '\n'): |
|
257 t = _('invalid entry in fncache, line %s') % (n + 1) |
|
258 raise util.Abort(t) |
|
259 self.entries.add(decodedir(line[:-1])) |
|
260 fp.close() |
|
261 |
|
262 def rewrite(self, files): |
|
263 fp = self.opener('fncache', mode='wb') |
|
264 for p in files: |
|
265 fp.write(encodedir(p) + '\n') |
|
266 fp.close() |
|
267 self.entries = set(files) |
|
268 |
|
269 def add(self, fn): |
|
270 if self.entries is None: |
|
271 self._load() |
|
272 if fn not in self.entries: |
|
273 self.opener('fncache', 'ab').write(encodedir(fn) + '\n') |
|
274 self.entries.add(fn) |
|
275 |
|
276 def __contains__(self, fn): |
|
277 if self.entries is None: |
|
278 self._load() |
|
279 return fn in self.entries |
|
280 |
|
281 def __iter__(self): |
|
282 if self.entries is None: |
|
283 self._load() |
|
284 return iter(self.entries) |
|
285 |
|
286 class fncachestore(basicstore): |
|
287 def __init__(self, path, opener, pathjoiner, encode): |
|
288 self.encode = encode |
|
289 self.pathjoiner = pathjoiner |
|
290 self.path = self.pathjoiner(path, 'store') |
|
291 self.createmode = _calcmode(self.path) |
|
292 op = opener(self.path) |
|
293 op.createmode = self.createmode |
|
294 fnc = fncache(op) |
|
295 self.fncache = fnc |
|
296 |
|
297 def fncacheopener(path, mode='r', *args, **kw): |
|
298 if mode not in ('r', 'rb') and path.startswith('data/'): |
|
299 fnc.add(path) |
|
300 return op(self.encode(path), mode, *args, **kw) |
|
301 self.opener = fncacheopener |
|
302 |
|
303 def join(self, f): |
|
304 return self.pathjoiner(self.path, self.encode(f)) |
|
305 |
|
306 def datafiles(self): |
|
307 rewrite = False |
|
308 existing = [] |
|
309 pjoin = self.pathjoiner |
|
310 spath = self.path |
|
311 for f in self.fncache: |
|
312 ef = self.encode(f) |
|
313 try: |
|
314 st = os.stat(pjoin(spath, ef)) |
|
315 yield f, ef, st.st_size |
|
316 existing.append(f) |
|
317 except OSError: |
|
318 # nonexistent entry |
|
319 rewrite = True |
|
320 if rewrite: |
|
321 # rewrite fncache to remove nonexistent entries |
|
322 # (may be caused by rollback / strip) |
|
323 self.fncache.rewrite(existing) |
|
324 |
|
325 def copylist(self): |
|
326 d = ('data dh fncache' |
|
327 ' 00manifest.d 00manifest.i 00changelog.d 00changelog.i') |
|
328 return (['requires', '00changelog.i'] + |
|
329 [self.pathjoiner('store', f) for f in d.split()]) |
|
330 |
|
331 def store(requirements, path, opener, pathjoiner=None): |
|
332 pathjoiner = pathjoiner or os.path.join |
|
333 if 'store' in requirements: |
|
334 if 'fncache' in requirements: |
|
335 auxencode = lambda f: _auxencode(f, 'dotencode' in requirements) |
|
336 encode = lambda f: _hybridencode(f, auxencode) |
|
337 return fncachestore(path, opener, pathjoiner, encode) |
|
338 return encodedstore(path, opener, pathjoiner) |
|
339 return basicstore(path, opener, pathjoiner) |