|
1 # Copyright 2007 Bryan O'Sullivan <bos@serpentine.com> |
|
2 # Copyright 2007 Alexis S. L. Carvalho <alexis@cecm.usp.br> |
|
3 # |
|
4 # This software may be used and distributed according to the terms of the |
|
5 # GNU General Public License version 2 or any later version. |
|
6 |
|
7 import shlex |
|
8 from mercurial.i18n import _ |
|
9 from mercurial import util |
|
10 from common import SKIPREV, converter_source |
|
11 |
|
12 def rpairs(name): |
|
13 e = len(name) |
|
14 while e != -1: |
|
15 yield name[:e], name[e + 1:] |
|
16 e = name.rfind('/', 0, e) |
|
17 yield '.', name |
|
18 |
|
19 class filemapper(object): |
|
20 '''Map and filter filenames when importing. |
|
21 A name can be mapped to itself, a new name, or None (omit from new |
|
22 repository).''' |
|
23 |
|
24 def __init__(self, ui, path=None): |
|
25 self.ui = ui |
|
26 self.include = {} |
|
27 self.exclude = {} |
|
28 self.rename = {} |
|
29 if path: |
|
30 if self.parse(path): |
|
31 raise util.Abort(_('errors in filemap')) |
|
32 |
|
33 def parse(self, path): |
|
34 errs = 0 |
|
35 def check(name, mapping, listname): |
|
36 if not name: |
|
37 self.ui.warn(_('%s:%d: path to %s is missing\n') % |
|
38 (lex.infile, lex.lineno, listname)) |
|
39 return 1 |
|
40 if name in mapping: |
|
41 self.ui.warn(_('%s:%d: %r already in %s list\n') % |
|
42 (lex.infile, lex.lineno, name, listname)) |
|
43 return 1 |
|
44 if (name.startswith('/') or |
|
45 name.endswith('/') or |
|
46 '//' in name): |
|
47 self.ui.warn(_('%s:%d: superfluous / in %s %r\n') % |
|
48 (lex.infile, lex.lineno, listname, name)) |
|
49 return 1 |
|
50 return 0 |
|
51 lex = shlex.shlex(open(path), path, True) |
|
52 lex.wordchars += '!@#$%^&*()-=+[]{}|;:,./<>?' |
|
53 cmd = lex.get_token() |
|
54 while cmd: |
|
55 if cmd == 'include': |
|
56 name = lex.get_token() |
|
57 errs += check(name, self.exclude, 'exclude') |
|
58 self.include[name] = name |
|
59 elif cmd == 'exclude': |
|
60 name = lex.get_token() |
|
61 errs += check(name, self.include, 'include') |
|
62 errs += check(name, self.rename, 'rename') |
|
63 self.exclude[name] = name |
|
64 elif cmd == 'rename': |
|
65 src = lex.get_token() |
|
66 dest = lex.get_token() |
|
67 errs += check(src, self.exclude, 'exclude') |
|
68 self.rename[src] = dest |
|
69 elif cmd == 'source': |
|
70 errs += self.parse(lex.get_token()) |
|
71 else: |
|
72 self.ui.warn(_('%s:%d: unknown directive %r\n') % |
|
73 (lex.infile, lex.lineno, cmd)) |
|
74 errs += 1 |
|
75 cmd = lex.get_token() |
|
76 return errs |
|
77 |
|
78 def lookup(self, name, mapping): |
|
79 for pre, suf in rpairs(name): |
|
80 try: |
|
81 return mapping[pre], pre, suf |
|
82 except KeyError: |
|
83 pass |
|
84 return '', name, '' |
|
85 |
|
86 def __call__(self, name): |
|
87 if self.include: |
|
88 inc = self.lookup(name, self.include)[0] |
|
89 else: |
|
90 inc = name |
|
91 if self.exclude: |
|
92 exc = self.lookup(name, self.exclude)[0] |
|
93 else: |
|
94 exc = '' |
|
95 if (not self.include and exc) or (len(inc) <= len(exc)): |
|
96 return None |
|
97 newpre, pre, suf = self.lookup(name, self.rename) |
|
98 if newpre: |
|
99 if newpre == '.': |
|
100 return suf |
|
101 if suf: |
|
102 return newpre + '/' + suf |
|
103 return newpre |
|
104 return name |
|
105 |
|
106 def active(self): |
|
107 return bool(self.include or self.exclude or self.rename) |
|
108 |
|
109 # This class does two additional things compared to a regular source: |
|
110 # |
|
111 # - Filter and rename files. This is mostly wrapped by the filemapper |
|
112 # class above. We hide the original filename in the revision that is |
|
113 # returned by getchanges to be able to find things later in getfile. |
|
114 # |
|
115 # - Return only revisions that matter for the files we're interested in. |
|
116 # This involves rewriting the parents of the original revision to |
|
117 # create a graph that is restricted to those revisions. |
|
118 # |
|
119 # This set of revisions includes not only revisions that directly |
|
120 # touch files we're interested in, but also merges that merge two |
|
121 # or more interesting revisions. |
|
122 |
|
123 class filemap_source(converter_source): |
|
124 def __init__(self, ui, baseconverter, filemap): |
|
125 super(filemap_source, self).__init__(ui) |
|
126 self.base = baseconverter |
|
127 self.filemapper = filemapper(ui, filemap) |
|
128 self.commits = {} |
|
129 # if a revision rev has parent p in the original revision graph, then |
|
130 # rev will have parent self.parentmap[p] in the restricted graph. |
|
131 self.parentmap = {} |
|
132 # self.wantedancestors[rev] is the set of all ancestors of rev that |
|
133 # are in the restricted graph. |
|
134 self.wantedancestors = {} |
|
135 self.convertedorder = None |
|
136 self._rebuilt = False |
|
137 self.origparents = {} |
|
138 self.children = {} |
|
139 self.seenchildren = {} |
|
140 |
|
141 def before(self): |
|
142 self.base.before() |
|
143 |
|
144 def after(self): |
|
145 self.base.after() |
|
146 |
|
147 def setrevmap(self, revmap): |
|
148 # rebuild our state to make things restartable |
|
149 # |
|
150 # To avoid calling getcommit for every revision that has already |
|
151 # been converted, we rebuild only the parentmap, delaying the |
|
152 # rebuild of wantedancestors until we need it (i.e. until a |
|
153 # merge). |
|
154 # |
|
155 # We assume the order argument lists the revisions in |
|
156 # topological order, so that we can infer which revisions were |
|
157 # wanted by previous runs. |
|
158 self._rebuilt = not revmap |
|
159 seen = {SKIPREV: SKIPREV} |
|
160 dummyset = set() |
|
161 converted = [] |
|
162 for rev in revmap.order: |
|
163 mapped = revmap[rev] |
|
164 wanted = mapped not in seen |
|
165 if wanted: |
|
166 seen[mapped] = rev |
|
167 self.parentmap[rev] = rev |
|
168 else: |
|
169 self.parentmap[rev] = seen[mapped] |
|
170 self.wantedancestors[rev] = dummyset |
|
171 arg = seen[mapped] |
|
172 if arg == SKIPREV: |
|
173 arg = None |
|
174 converted.append((rev, wanted, arg)) |
|
175 self.convertedorder = converted |
|
176 return self.base.setrevmap(revmap) |
|
177 |
|
178 def rebuild(self): |
|
179 if self._rebuilt: |
|
180 return True |
|
181 self._rebuilt = True |
|
182 self.parentmap.clear() |
|
183 self.wantedancestors.clear() |
|
184 self.seenchildren.clear() |
|
185 for rev, wanted, arg in self.convertedorder: |
|
186 if rev not in self.origparents: |
|
187 self.origparents[rev] = self.getcommit(rev).parents |
|
188 if arg is not None: |
|
189 self.children[arg] = self.children.get(arg, 0) + 1 |
|
190 |
|
191 for rev, wanted, arg in self.convertedorder: |
|
192 parents = self.origparents[rev] |
|
193 if wanted: |
|
194 self.mark_wanted(rev, parents) |
|
195 else: |
|
196 self.mark_not_wanted(rev, arg) |
|
197 self._discard(arg, *parents) |
|
198 |
|
199 return True |
|
200 |
|
201 def getheads(self): |
|
202 return self.base.getheads() |
|
203 |
|
204 def getcommit(self, rev): |
|
205 # We want to save a reference to the commit objects to be able |
|
206 # to rewrite their parents later on. |
|
207 c = self.commits[rev] = self.base.getcommit(rev) |
|
208 for p in c.parents: |
|
209 self.children[p] = self.children.get(p, 0) + 1 |
|
210 return c |
|
211 |
|
212 def _discard(self, *revs): |
|
213 for r in revs: |
|
214 if r is None: |
|
215 continue |
|
216 self.seenchildren[r] = self.seenchildren.get(r, 0) + 1 |
|
217 if self.seenchildren[r] == self.children[r]: |
|
218 del self.wantedancestors[r] |
|
219 del self.parentmap[r] |
|
220 del self.seenchildren[r] |
|
221 if self._rebuilt: |
|
222 del self.children[r] |
|
223 |
|
224 def wanted(self, rev, i): |
|
225 # Return True if we're directly interested in rev. |
|
226 # |
|
227 # i is an index selecting one of the parents of rev (if rev |
|
228 # has no parents, i is None). getchangedfiles will give us |
|
229 # the list of files that are different in rev and in the parent |
|
230 # indicated by i. If we're interested in any of these files, |
|
231 # we're interested in rev. |
|
232 try: |
|
233 files = self.base.getchangedfiles(rev, i) |
|
234 except NotImplementedError: |
|
235 raise util.Abort(_("source repository doesn't support --filemap")) |
|
236 for f in files: |
|
237 if self.filemapper(f): |
|
238 return True |
|
239 return False |
|
240 |
|
241 def mark_not_wanted(self, rev, p): |
|
242 # Mark rev as not interesting and update data structures. |
|
243 |
|
244 if p is None: |
|
245 # A root revision. Use SKIPREV to indicate that it doesn't |
|
246 # map to any revision in the restricted graph. Put SKIPREV |
|
247 # in the set of wanted ancestors to simplify code elsewhere |
|
248 self.parentmap[rev] = SKIPREV |
|
249 self.wantedancestors[rev] = set((SKIPREV,)) |
|
250 return |
|
251 |
|
252 # Reuse the data from our parent. |
|
253 self.parentmap[rev] = self.parentmap[p] |
|
254 self.wantedancestors[rev] = self.wantedancestors[p] |
|
255 |
|
256 def mark_wanted(self, rev, parents): |
|
257 # Mark rev ss wanted and update data structures. |
|
258 |
|
259 # rev will be in the restricted graph, so children of rev in |
|
260 # the original graph should still have rev as a parent in the |
|
261 # restricted graph. |
|
262 self.parentmap[rev] = rev |
|
263 |
|
264 # The set of wanted ancestors of rev is the union of the sets |
|
265 # of wanted ancestors of its parents. Plus rev itself. |
|
266 wrev = set() |
|
267 for p in parents: |
|
268 wrev.update(self.wantedancestors[p]) |
|
269 wrev.add(rev) |
|
270 self.wantedancestors[rev] = wrev |
|
271 |
|
272 def getchanges(self, rev): |
|
273 parents = self.commits[rev].parents |
|
274 if len(parents) > 1: |
|
275 self.rebuild() |
|
276 |
|
277 # To decide whether we're interested in rev we: |
|
278 # |
|
279 # - calculate what parents rev will have if it turns out we're |
|
280 # interested in it. If it's going to have more than 1 parent, |
|
281 # we're interested in it. |
|
282 # |
|
283 # - otherwise, we'll compare it with the single parent we found. |
|
284 # If any of the files we're interested in is different in the |
|
285 # the two revisions, we're interested in rev. |
|
286 |
|
287 # A parent p is interesting if its mapped version (self.parentmap[p]): |
|
288 # - is not SKIPREV |
|
289 # - is still not in the list of parents (we don't want duplicates) |
|
290 # - is not an ancestor of the mapped versions of the other parents |
|
291 mparents = [] |
|
292 wp = None |
|
293 for i, p1 in enumerate(parents): |
|
294 mp1 = self.parentmap[p1] |
|
295 if mp1 == SKIPREV or mp1 in mparents: |
|
296 continue |
|
297 for p2 in parents: |
|
298 if p1 == p2 or mp1 == self.parentmap[p2]: |
|
299 continue |
|
300 if mp1 in self.wantedancestors[p2]: |
|
301 break |
|
302 else: |
|
303 mparents.append(mp1) |
|
304 wp = i |
|
305 |
|
306 if wp is None and parents: |
|
307 wp = 0 |
|
308 |
|
309 self.origparents[rev] = parents |
|
310 |
|
311 closed = 'close' in self.commits[rev].extra |
|
312 |
|
313 if len(mparents) < 2 and not closed and not self.wanted(rev, wp): |
|
314 # We don't want this revision. |
|
315 # Update our state and tell the convert process to map this |
|
316 # revision to the same revision its parent as mapped to. |
|
317 p = None |
|
318 if parents: |
|
319 p = parents[wp] |
|
320 self.mark_not_wanted(rev, p) |
|
321 self.convertedorder.append((rev, False, p)) |
|
322 self._discard(*parents) |
|
323 return self.parentmap[rev] |
|
324 |
|
325 # We want this revision. |
|
326 # Rewrite the parents of the commit object |
|
327 self.commits[rev].parents = mparents |
|
328 self.mark_wanted(rev, parents) |
|
329 self.convertedorder.append((rev, True, None)) |
|
330 self._discard(*parents) |
|
331 |
|
332 # Get the real changes and do the filtering/mapping. To be |
|
333 # able to get the files later on in getfile, we hide the |
|
334 # original filename in the rev part of the return value. |
|
335 changes, copies = self.base.getchanges(rev) |
|
336 newnames = {} |
|
337 files = [] |
|
338 for f, r in changes: |
|
339 newf = self.filemapper(f) |
|
340 if newf: |
|
341 files.append((newf, (f, r))) |
|
342 newnames[f] = newf |
|
343 |
|
344 ncopies = {} |
|
345 for c in copies: |
|
346 newc = self.filemapper(c) |
|
347 if newc: |
|
348 newsource = self.filemapper(copies[c]) |
|
349 if newsource: |
|
350 ncopies[newc] = newsource |
|
351 |
|
352 return files, ncopies |
|
353 |
|
354 def getfile(self, name, rev): |
|
355 realname, realrev = rev |
|
356 return self.base.getfile(realname, realrev) |
|
357 |
|
358 def gettags(self): |
|
359 return self.base.gettags() |
|
360 |
|
361 def hasnativeorder(self): |
|
362 return self.base.hasnativeorder() |
|
363 |
|
364 def lookuprev(self, rev): |
|
365 return self.base.lookuprev(rev) |