Show More
@@ -1,805 +1,806 b'' | |||
|
1 | 1 | # copies.py - copy detection 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 __future__ import absolute_import |
|
9 | 9 | |
|
10 | 10 | import collections |
|
11 | 11 | import heapq |
|
12 | 12 | import os |
|
13 | 13 | |
|
14 | 14 | from .i18n import _ |
|
15 | 15 | |
|
16 | 16 | from . import ( |
|
17 | 17 | match as matchmod, |
|
18 | 18 | node, |
|
19 | 19 | pathutil, |
|
20 | 20 | util, |
|
21 | 21 | ) |
|
22 | 22 | from .utils import ( |
|
23 | 23 | stringutil, |
|
24 | 24 | ) |
|
25 | 25 | |
|
26 | 26 | def _findlimit(repo, ctxa, ctxb): |
|
27 | 27 | """ |
|
28 | 28 | Find the last revision that needs to be checked to ensure that a full |
|
29 | 29 | transitive closure for file copies can be properly calculated. |
|
30 | 30 | Generally, this means finding the earliest revision number that's an |
|
31 | 31 | ancestor of a or b but not both, except when a or b is a direct descendent |
|
32 | 32 | of the other, in which case we can return the minimum revnum of a and b. |
|
33 | 33 | """ |
|
34 | 34 | |
|
35 | 35 | # basic idea: |
|
36 | 36 | # - mark a and b with different sides |
|
37 | 37 | # - if a parent's children are all on the same side, the parent is |
|
38 | 38 | # on that side, otherwise it is on no side |
|
39 | 39 | # - walk the graph in topological order with the help of a heap; |
|
40 | 40 | # - add unseen parents to side map |
|
41 | 41 | # - clear side of any parent that has children on different sides |
|
42 | 42 | # - track number of interesting revs that might still be on a side |
|
43 | 43 | # - track the lowest interesting rev seen |
|
44 | 44 | # - quit when interesting revs is zero |
|
45 | 45 | |
|
46 | 46 | cl = repo.changelog |
|
47 | 47 | wdirparents = None |
|
48 | 48 | a = ctxa.rev() |
|
49 | 49 | b = ctxb.rev() |
|
50 | 50 | if a is None: |
|
51 | 51 | wdirparents = (ctxa.p1(), ctxa.p2()) |
|
52 | 52 | a = node.wdirrev |
|
53 | 53 | if b is None: |
|
54 | 54 | assert not wdirparents |
|
55 | 55 | wdirparents = (ctxb.p1(), ctxb.p2()) |
|
56 | 56 | b = node.wdirrev |
|
57 | 57 | |
|
58 | 58 | side = {a: -1, b: 1} |
|
59 | 59 | visit = [-a, -b] |
|
60 | 60 | heapq.heapify(visit) |
|
61 | 61 | interesting = len(visit) |
|
62 | 62 | limit = node.wdirrev |
|
63 | 63 | |
|
64 | 64 | while interesting: |
|
65 | 65 | r = -heapq.heappop(visit) |
|
66 | 66 | if r == node.wdirrev: |
|
67 | 67 | parents = [pctx.rev() for pctx in wdirparents] |
|
68 | 68 | else: |
|
69 | 69 | parents = cl.parentrevs(r) |
|
70 | 70 | if parents[1] == node.nullrev: |
|
71 | 71 | parents = parents[:1] |
|
72 | 72 | for p in parents: |
|
73 | 73 | if p not in side: |
|
74 | 74 | # first time we see p; add it to visit |
|
75 | 75 | side[p] = side[r] |
|
76 | 76 | if side[p]: |
|
77 | 77 | interesting += 1 |
|
78 | 78 | heapq.heappush(visit, -p) |
|
79 | 79 | elif side[p] and side[p] != side[r]: |
|
80 | 80 | # p was interesting but now we know better |
|
81 | 81 | side[p] = 0 |
|
82 | 82 | interesting -= 1 |
|
83 | 83 | if side[r]: |
|
84 | 84 | limit = r # lowest rev visited |
|
85 | 85 | interesting -= 1 |
|
86 | 86 | |
|
87 | 87 | # Consider the following flow (see test-commit-amend.t under issue4405): |
|
88 | 88 | # 1/ File 'a0' committed |
|
89 | 89 | # 2/ File renamed from 'a0' to 'a1' in a new commit (call it 'a1') |
|
90 | 90 | # 3/ Move back to first commit |
|
91 | 91 | # 4/ Create a new commit via revert to contents of 'a1' (call it 'a1-amend') |
|
92 | 92 | # 5/ Rename file from 'a1' to 'a2' and commit --amend 'a1-msg' |
|
93 | 93 | # |
|
94 | 94 | # During the amend in step five, we will be in this state: |
|
95 | 95 | # |
|
96 | 96 | # @ 3 temporary amend commit for a1-amend |
|
97 | 97 | # | |
|
98 | 98 | # o 2 a1-amend |
|
99 | 99 | # | |
|
100 | 100 | # | o 1 a1 |
|
101 | 101 | # |/ |
|
102 | 102 | # o 0 a0 |
|
103 | 103 | # |
|
104 | 104 | # When _findlimit is called, a and b are revs 3 and 0, so limit will be 2, |
|
105 | 105 | # yet the filelog has the copy information in rev 1 and we will not look |
|
106 | 106 | # back far enough unless we also look at the a and b as candidates. |
|
107 | 107 | # This only occurs when a is a descendent of b or visa-versa. |
|
108 | 108 | return min(limit, a, b) |
|
109 | 109 | |
|
110 | 110 | def _chainandfilter(src, dst, a, b): |
|
111 | 111 | """chain two sets of copies 'a' and 'b' and filter result""" |
|
112 | 112 | |
|
113 | 113 | # When chaining copies in 'a' (from 'src' via some other commit 'mid') with |
|
114 | 114 | # copies in 'b' (from 'mid' to 'dst'), we can get the different cases in the |
|
115 | 115 | # following table (not including trivial cases). For example, case 2 is |
|
116 | 116 | # where a file existed in 'src' and remained under that name in 'mid' and |
|
117 | 117 | # then was renamed between 'mid' and 'dst'. |
|
118 | 118 | # |
|
119 | 119 | # case src mid dst result |
|
120 | 120 | # 1 x y - - |
|
121 | 121 | # 2 x y y x->y |
|
122 | 122 | # 3 x y x - |
|
123 | 123 | # 4 x y z x->z |
|
124 | 124 | # 5 - x y - |
|
125 | 125 | # 6 x x y x->y |
|
126 | 126 | # |
|
127 | 127 | # _chain() takes care of chaining the copies in 'a' and 'b', but it |
|
128 | 128 | # cannot tell the difference between cases 1 and 2, between 3 and 4, or |
|
129 | 129 | # between 5 and 6, so it includes all cases in its result. |
|
130 | 130 | # Cases 1, 3, and 5 are then removed by _filter(). |
|
131 | 131 | |
|
132 | 132 | t = _chain(a, b) |
|
133 | 133 | _filter(src, dst, t) |
|
134 | 134 | return t |
|
135 | 135 | |
|
136 | 136 | def _filter(src, dst, t): |
|
137 | 137 | """filters out invalid copies after chaining""" |
|
138 | 138 | for k, v in list(t.items()): |
|
139 | 139 | # remove copies from files that didn't exist |
|
140 | 140 | if v not in src: |
|
141 | 141 | del t[k] |
|
142 | 142 | # remove criss-crossed copies |
|
143 | 143 | elif k in src and v in dst: |
|
144 | 144 | del t[k] |
|
145 | 145 | # remove copies to files that were then removed |
|
146 | 146 | elif k not in dst: |
|
147 | 147 | del t[k] |
|
148 | 148 | |
|
149 | 149 | def _chain(a, b): |
|
150 | 150 | """chain two sets of copies 'a' and 'b'""" |
|
151 | 151 | t = a.copy() |
|
152 | 152 | for k, v in b.iteritems(): |
|
153 | 153 | if v in t: |
|
154 | 154 | t[k] = t[v] |
|
155 | 155 | else: |
|
156 | 156 | t[k] = v |
|
157 | 157 | return t |
|
158 | 158 | |
|
159 | 159 | def _tracefile(fctx, am, limit): |
|
160 | 160 | """return file context that is the ancestor of fctx present in ancestor |
|
161 | 161 | manifest am, stopping after the first ancestor lower than limit""" |
|
162 | 162 | |
|
163 | 163 | for f in fctx.ancestors(): |
|
164 | if am.get(f.path(), None) == f.filenode(): | |
|
165 | return f | |
|
164 | path = f.path() | |
|
165 | if am.get(path, None) == f.filenode(): | |
|
166 | return path | |
|
166 | 167 | if not f.isintroducedafter(limit): |
|
167 | 168 | return None |
|
168 | 169 | |
|
169 | 170 | def _dirstatecopies(repo, match=None): |
|
170 | 171 | ds = repo.dirstate |
|
171 | 172 | c = ds.copies().copy() |
|
172 | 173 | for k in list(c): |
|
173 | 174 | if ds[k] not in 'anm' or (match and not match(k)): |
|
174 | 175 | del c[k] |
|
175 | 176 | return c |
|
176 | 177 | |
|
177 | 178 | def _computeforwardmissing(a, b, match=None): |
|
178 | 179 | """Computes which files are in b but not a. |
|
179 | 180 | This is its own function so extensions can easily wrap this call to see what |
|
180 | 181 | files _forwardcopies is about to process. |
|
181 | 182 | """ |
|
182 | 183 | ma = a.manifest() |
|
183 | 184 | mb = b.manifest() |
|
184 | 185 | return mb.filesnotin(ma, match=match) |
|
185 | 186 | |
|
186 | 187 | def usechangesetcentricalgo(repo): |
|
187 | 188 | """Checks if we should use changeset-centric copy algorithms""" |
|
188 | 189 | return (repo.ui.config('experimental', 'copies.read-from') in |
|
189 | 190 | ('changeset-only', 'compatibility')) |
|
190 | 191 | |
|
191 | 192 | def _committedforwardcopies(a, b, match): |
|
192 | 193 | """Like _forwardcopies(), but b.rev() cannot be None (working copy)""" |
|
193 | 194 | # files might have to be traced back to the fctx parent of the last |
|
194 | 195 | # one-side-only changeset, but not further back than that |
|
195 | 196 | repo = a._repo |
|
196 | 197 | |
|
197 | 198 | if usechangesetcentricalgo(repo): |
|
198 | 199 | return _changesetforwardcopies(a, b, match) |
|
199 | 200 | |
|
200 | 201 | debug = repo.ui.debugflag and repo.ui.configbool('devel', 'debug.copies') |
|
201 | 202 | dbg = repo.ui.debug |
|
202 | 203 | if debug: |
|
203 | 204 | dbg('debug.copies: looking into rename from %s to %s\n' |
|
204 | 205 | % (a, b)) |
|
205 | 206 | limit = _findlimit(repo, a, b) |
|
206 | 207 | if debug: |
|
207 | 208 | dbg('debug.copies: search limit: %d\n' % limit) |
|
208 | 209 | am = a.manifest() |
|
209 | 210 | |
|
210 | 211 | # find where new files came from |
|
211 | 212 | # we currently don't try to find where old files went, too expensive |
|
212 | 213 | # this means we can miss a case like 'hg rm b; hg cp a b' |
|
213 | 214 | cm = {} |
|
214 | 215 | |
|
215 | 216 | # Computing the forward missing is quite expensive on large manifests, since |
|
216 | 217 | # it compares the entire manifests. We can optimize it in the common use |
|
217 | 218 | # case of computing what copies are in a commit versus its parent (like |
|
218 | 219 | # during a rebase or histedit). Note, we exclude merge commits from this |
|
219 | 220 | # optimization, since the ctx.files() for a merge commit is not correct for |
|
220 | 221 | # this comparison. |
|
221 | 222 | forwardmissingmatch = match |
|
222 | 223 | if b.p1() == a and b.p2().node() == node.nullid: |
|
223 | 224 | filesmatcher = matchmod.exact(b.files()) |
|
224 | 225 | forwardmissingmatch = matchmod.intersectmatchers(match, filesmatcher) |
|
225 | 226 | missing = _computeforwardmissing(a, b, match=forwardmissingmatch) |
|
226 | 227 | |
|
227 | 228 | ancestrycontext = a._repo.changelog.ancestors([b.rev()], inclusive=True) |
|
228 | 229 | |
|
229 | 230 | if debug: |
|
230 | 231 | dbg('debug.copies: missing files to search: %d\n' % len(missing)) |
|
231 | 232 | |
|
232 | 233 | for f in sorted(missing): |
|
233 | 234 | if debug: |
|
234 | 235 | dbg('debug.copies: tracing file: %s\n' % f) |
|
235 | 236 | fctx = b[f] |
|
236 | 237 | fctx._ancestrycontext = ancestrycontext |
|
237 | 238 | |
|
238 | 239 | if debug: |
|
239 | 240 | start = util.timer() |
|
240 |
o |
|
|
241 |
if o |
|
|
241 | opath = _tracefile(fctx, am, limit) | |
|
242 | if opath: | |
|
242 | 243 | if debug: |
|
243 |
dbg('debug.copies: rename of: %s\n' % o |
|
|
244 |
cm[f] = o |
|
|
244 | dbg('debug.copies: rename of: %s\n' % opath) | |
|
245 | cm[f] = opath | |
|
245 | 246 | if debug: |
|
246 | 247 | dbg('debug.copies: time: %f seconds\n' |
|
247 | 248 | % (util.timer() - start)) |
|
248 | 249 | return cm |
|
249 | 250 | |
|
250 | 251 | def _changesetforwardcopies(a, b, match): |
|
251 | 252 | if a.rev() == node.nullrev: |
|
252 | 253 | return {} |
|
253 | 254 | |
|
254 | 255 | repo = a.repo() |
|
255 | 256 | children = {} |
|
256 | 257 | cl = repo.changelog |
|
257 | 258 | missingrevs = cl.findmissingrevs(common=[a.rev()], heads=[b.rev()]) |
|
258 | 259 | for r in missingrevs: |
|
259 | 260 | for p in cl.parentrevs(r): |
|
260 | 261 | if p == node.nullrev: |
|
261 | 262 | continue |
|
262 | 263 | if p not in children: |
|
263 | 264 | children[p] = [r] |
|
264 | 265 | else: |
|
265 | 266 | children[p].append(r) |
|
266 | 267 | |
|
267 | 268 | roots = set(children) - set(missingrevs) |
|
268 | 269 | # 'work' contains 3-tuples of a (revision number, parent number, copies). |
|
269 | 270 | # The parent number is only used for knowing which parent the copies dict |
|
270 | 271 | # came from. |
|
271 | 272 | # NOTE: To reduce costly copying the 'copies' dicts, we reuse the same |
|
272 | 273 | # instance for *one* of the child nodes (the last one). Once an instance |
|
273 | 274 | # has been put on the queue, it is thus no longer safe to modify it. |
|
274 | 275 | # Conversely, it *is* safe to modify an instance popped off the queue. |
|
275 | 276 | work = [(r, 1, {}) for r in roots] |
|
276 | 277 | heapq.heapify(work) |
|
277 | 278 | alwaysmatch = match.always() |
|
278 | 279 | while work: |
|
279 | 280 | r, i1, copies = heapq.heappop(work) |
|
280 | 281 | if work and work[0][0] == r: |
|
281 | 282 | # We are tracing copies from both parents |
|
282 | 283 | r, i2, copies2 = heapq.heappop(work) |
|
283 | 284 | for dst, src in copies2.items(): |
|
284 | 285 | # Unlike when copies are stored in the filelog, we consider |
|
285 | 286 | # it a copy even if the destination already existed on the |
|
286 | 287 | # other branch. It's simply too expensive to check if the |
|
287 | 288 | # file existed in the manifest. |
|
288 | 289 | if dst not in copies: |
|
289 | 290 | # If it was copied on the p1 side, leave it as copied from |
|
290 | 291 | # that side, even if it was also copied on the p2 side. |
|
291 | 292 | copies[dst] = copies2[dst] |
|
292 | 293 | if r == b.rev(): |
|
293 | 294 | _filter(a, b, copies) |
|
294 | 295 | return copies |
|
295 | 296 | for i, c in enumerate(children[r]): |
|
296 | 297 | childctx = repo[c] |
|
297 | 298 | if r == childctx.p1().rev(): |
|
298 | 299 | parent = 1 |
|
299 | 300 | childcopies = childctx.p1copies() |
|
300 | 301 | else: |
|
301 | 302 | assert r == childctx.p2().rev() |
|
302 | 303 | parent = 2 |
|
303 | 304 | childcopies = childctx.p2copies() |
|
304 | 305 | if not alwaysmatch: |
|
305 | 306 | childcopies = {dst: src for dst, src in childcopies.items() |
|
306 | 307 | if match(dst)} |
|
307 | 308 | # Copy the dict only if later iterations will also need it |
|
308 | 309 | if i != len(children[r]) - 1: |
|
309 | 310 | newcopies = copies.copy() |
|
310 | 311 | else: |
|
311 | 312 | newcopies = copies |
|
312 | 313 | if childcopies: |
|
313 | 314 | newcopies = _chain(newcopies, childcopies) |
|
314 | 315 | for f in childctx.filesremoved(): |
|
315 | 316 | if f in newcopies: |
|
316 | 317 | del newcopies[f] |
|
317 | 318 | heapq.heappush(work, (c, parent, newcopies)) |
|
318 | 319 | assert False |
|
319 | 320 | |
|
320 | 321 | def _forwardcopies(a, b, match=None): |
|
321 | 322 | """find {dst@b: src@a} copy mapping where a is an ancestor of b""" |
|
322 | 323 | |
|
323 | 324 | match = a.repo().narrowmatch(match) |
|
324 | 325 | # check for working copy |
|
325 | 326 | if b.rev() is None: |
|
326 | 327 | if a == b.p1(): |
|
327 | 328 | # short-circuit to avoid issues with merge states |
|
328 | 329 | return _dirstatecopies(b._repo, match) |
|
329 | 330 | |
|
330 | 331 | cm = _committedforwardcopies(a, b.p1(), match) |
|
331 | 332 | # combine copies from dirstate if necessary |
|
332 | 333 | return _chainandfilter(a, b, cm, _dirstatecopies(b._repo, match)) |
|
333 | 334 | return _committedforwardcopies(a, b, match) |
|
334 | 335 | |
|
335 | 336 | def _backwardrenames(a, b, match): |
|
336 | 337 | if a._repo.ui.config('experimental', 'copytrace') == 'off': |
|
337 | 338 | return {} |
|
338 | 339 | |
|
339 | 340 | # Even though we're not taking copies into account, 1:n rename situations |
|
340 | 341 | # can still exist (e.g. hg cp a b; hg mv a c). In those cases we |
|
341 | 342 | # arbitrarily pick one of the renames. |
|
342 | 343 | # We don't want to pass in "match" here, since that would filter |
|
343 | 344 | # the destination by it. Since we're reversing the copies, we want |
|
344 | 345 | # to filter the source instead. |
|
345 | 346 | f = _forwardcopies(b, a) |
|
346 | 347 | r = {} |
|
347 | 348 | for k, v in sorted(f.iteritems()): |
|
348 | 349 | if match and not match(v): |
|
349 | 350 | continue |
|
350 | 351 | # remove copies |
|
351 | 352 | if v in a: |
|
352 | 353 | continue |
|
353 | 354 | r[v] = k |
|
354 | 355 | return r |
|
355 | 356 | |
|
356 | 357 | def pathcopies(x, y, match=None): |
|
357 | 358 | """find {dst@y: src@x} copy mapping for directed compare""" |
|
358 | 359 | repo = x._repo |
|
359 | 360 | debug = repo.ui.debugflag and repo.ui.configbool('devel', 'debug.copies') |
|
360 | 361 | if debug: |
|
361 | 362 | repo.ui.debug('debug.copies: searching copies from %s to %s\n' |
|
362 | 363 | % (x, y)) |
|
363 | 364 | if x == y or not x or not y: |
|
364 | 365 | return {} |
|
365 | 366 | a = y.ancestor(x) |
|
366 | 367 | if a == x: |
|
367 | 368 | if debug: |
|
368 | 369 | repo.ui.debug('debug.copies: search mode: forward\n') |
|
369 | 370 | return _forwardcopies(x, y, match=match) |
|
370 | 371 | if a == y: |
|
371 | 372 | if debug: |
|
372 | 373 | repo.ui.debug('debug.copies: search mode: backward\n') |
|
373 | 374 | return _backwardrenames(x, y, match=match) |
|
374 | 375 | if debug: |
|
375 | 376 | repo.ui.debug('debug.copies: search mode: combined\n') |
|
376 | 377 | return _chainandfilter(x, y, _backwardrenames(x, a, match=match), |
|
377 | 378 | _forwardcopies(a, y, match=match)) |
|
378 | 379 | |
|
379 | 380 | def mergecopies(repo, c1, c2, base): |
|
380 | 381 | """ |
|
381 | 382 | Finds moves and copies between context c1 and c2 that are relevant for |
|
382 | 383 | merging. 'base' will be used as the merge base. |
|
383 | 384 | |
|
384 | 385 | Copytracing is used in commands like rebase, merge, unshelve, etc to merge |
|
385 | 386 | files that were moved/ copied in one merge parent and modified in another. |
|
386 | 387 | For example: |
|
387 | 388 | |
|
388 | 389 | o ---> 4 another commit |
|
389 | 390 | | |
|
390 | 391 | | o ---> 3 commit that modifies a.txt |
|
391 | 392 | | / |
|
392 | 393 | o / ---> 2 commit that moves a.txt to b.txt |
|
393 | 394 | |/ |
|
394 | 395 | o ---> 1 merge base |
|
395 | 396 | |
|
396 | 397 | If we try to rebase revision 3 on revision 4, since there is no a.txt in |
|
397 | 398 | revision 4, and if user have copytrace disabled, we prints the following |
|
398 | 399 | message: |
|
399 | 400 | |
|
400 | 401 | ```other changed <file> which local deleted``` |
|
401 | 402 | |
|
402 | 403 | Returns five dicts: "copy", "movewithdir", "diverge", "renamedelete" and |
|
403 | 404 | "dirmove". |
|
404 | 405 | |
|
405 | 406 | "copy" is a mapping from destination name -> source name, |
|
406 | 407 | where source is in c1 and destination is in c2 or vice-versa. |
|
407 | 408 | |
|
408 | 409 | "movewithdir" is a mapping from source name -> destination name, |
|
409 | 410 | where the file at source present in one context but not the other |
|
410 | 411 | needs to be moved to destination by the merge process, because the |
|
411 | 412 | other context moved the directory it is in. |
|
412 | 413 | |
|
413 | 414 | "diverge" is a mapping of source name -> list of destination names |
|
414 | 415 | for divergent renames. |
|
415 | 416 | |
|
416 | 417 | "renamedelete" is a mapping of source name -> list of destination |
|
417 | 418 | names for files deleted in c1 that were renamed in c2 or vice-versa. |
|
418 | 419 | |
|
419 | 420 | "dirmove" is a mapping of detected source dir -> destination dir renames. |
|
420 | 421 | This is needed for handling changes to new files previously grafted into |
|
421 | 422 | renamed directories. |
|
422 | 423 | |
|
423 | 424 | This function calls different copytracing algorithms based on config. |
|
424 | 425 | """ |
|
425 | 426 | # avoid silly behavior for update from empty dir |
|
426 | 427 | if not c1 or not c2 or c1 == c2: |
|
427 | 428 | return {}, {}, {}, {}, {} |
|
428 | 429 | |
|
429 | 430 | narrowmatch = c1.repo().narrowmatch() |
|
430 | 431 | |
|
431 | 432 | # avoid silly behavior for parent -> working dir |
|
432 | 433 | if c2.node() is None and c1.node() == repo.dirstate.p1(): |
|
433 | 434 | return _dirstatecopies(repo, narrowmatch), {}, {}, {}, {} |
|
434 | 435 | |
|
435 | 436 | copytracing = repo.ui.config('experimental', 'copytrace') |
|
436 | 437 | if stringutil.parsebool(copytracing) is False: |
|
437 | 438 | # stringutil.parsebool() returns None when it is unable to parse the |
|
438 | 439 | # value, so we should rely on making sure copytracing is on such cases |
|
439 | 440 | return {}, {}, {}, {}, {} |
|
440 | 441 | |
|
441 | 442 | if usechangesetcentricalgo(repo): |
|
442 | 443 | # The heuristics don't make sense when we need changeset-centric algos |
|
443 | 444 | return _fullcopytracing(repo, c1, c2, base) |
|
444 | 445 | |
|
445 | 446 | # Copy trace disabling is explicitly below the node == p1 logic above |
|
446 | 447 | # because the logic above is required for a simple copy to be kept across a |
|
447 | 448 | # rebase. |
|
448 | 449 | if copytracing == 'heuristics': |
|
449 | 450 | # Do full copytracing if only non-public revisions are involved as |
|
450 | 451 | # that will be fast enough and will also cover the copies which could |
|
451 | 452 | # be missed by heuristics |
|
452 | 453 | if _isfullcopytraceable(repo, c1, base): |
|
453 | 454 | return _fullcopytracing(repo, c1, c2, base) |
|
454 | 455 | return _heuristicscopytracing(repo, c1, c2, base) |
|
455 | 456 | else: |
|
456 | 457 | return _fullcopytracing(repo, c1, c2, base) |
|
457 | 458 | |
|
458 | 459 | def _isfullcopytraceable(repo, c1, base): |
|
459 | 460 | """ Checks that if base, source and destination are all no-public branches, |
|
460 | 461 | if yes let's use the full copytrace algorithm for increased capabilities |
|
461 | 462 | since it will be fast enough. |
|
462 | 463 | |
|
463 | 464 | `experimental.copytrace.sourcecommitlimit` can be used to set a limit for |
|
464 | 465 | number of changesets from c1 to base such that if number of changesets are |
|
465 | 466 | more than the limit, full copytracing algorithm won't be used. |
|
466 | 467 | """ |
|
467 | 468 | if c1.rev() is None: |
|
468 | 469 | c1 = c1.p1() |
|
469 | 470 | if c1.mutable() and base.mutable(): |
|
470 | 471 | sourcecommitlimit = repo.ui.configint('experimental', |
|
471 | 472 | 'copytrace.sourcecommitlimit') |
|
472 | 473 | commits = len(repo.revs('%d::%d', base.rev(), c1.rev())) |
|
473 | 474 | return commits < sourcecommitlimit |
|
474 | 475 | return False |
|
475 | 476 | |
|
476 | 477 | def _checksinglesidecopies(src, dsts1, m1, m2, mb, c2, base, |
|
477 | 478 | copy, renamedelete): |
|
478 | 479 | if src not in m2: |
|
479 | 480 | # deleted on side 2 |
|
480 | 481 | if src not in m1: |
|
481 | 482 | # renamed on side 1, deleted on side 2 |
|
482 | 483 | renamedelete[src] = dsts1 |
|
483 | 484 | elif m2[src] != mb[src]: |
|
484 | 485 | if not _related(c2[src], base[src]): |
|
485 | 486 | return |
|
486 | 487 | # modified on side 2 |
|
487 | 488 | for dst in dsts1: |
|
488 | 489 | if dst not in m2: |
|
489 | 490 | # dst not added on side 2 (handle as regular |
|
490 | 491 | # "both created" case in manifestmerge otherwise) |
|
491 | 492 | copy[dst] = src |
|
492 | 493 | |
|
493 | 494 | def _fullcopytracing(repo, c1, c2, base): |
|
494 | 495 | """ The full copytracing algorithm which finds all the new files that were |
|
495 | 496 | added from merge base up to the top commit and for each file it checks if |
|
496 | 497 | this file was copied from another file. |
|
497 | 498 | |
|
498 | 499 | This is pretty slow when a lot of changesets are involved but will track all |
|
499 | 500 | the copies. |
|
500 | 501 | """ |
|
501 | 502 | m1 = c1.manifest() |
|
502 | 503 | m2 = c2.manifest() |
|
503 | 504 | mb = base.manifest() |
|
504 | 505 | |
|
505 | 506 | copies1 = pathcopies(base, c1) |
|
506 | 507 | copies2 = pathcopies(base, c2) |
|
507 | 508 | |
|
508 | 509 | inversecopies1 = {} |
|
509 | 510 | inversecopies2 = {} |
|
510 | 511 | for dst, src in copies1.items(): |
|
511 | 512 | inversecopies1.setdefault(src, []).append(dst) |
|
512 | 513 | for dst, src in copies2.items(): |
|
513 | 514 | inversecopies2.setdefault(src, []).append(dst) |
|
514 | 515 | |
|
515 | 516 | copy = {} |
|
516 | 517 | diverge = {} |
|
517 | 518 | renamedelete = {} |
|
518 | 519 | allsources = set(inversecopies1) | set(inversecopies2) |
|
519 | 520 | for src in allsources: |
|
520 | 521 | dsts1 = inversecopies1.get(src) |
|
521 | 522 | dsts2 = inversecopies2.get(src) |
|
522 | 523 | if dsts1 and dsts2: |
|
523 | 524 | # copied/renamed on both sides |
|
524 | 525 | if src not in m1 and src not in m2: |
|
525 | 526 | # renamed on both sides |
|
526 | 527 | dsts1 = set(dsts1) |
|
527 | 528 | dsts2 = set(dsts2) |
|
528 | 529 | # If there's some overlap in the rename destinations, we |
|
529 | 530 | # consider it not divergent. For example, if side 1 copies 'a' |
|
530 | 531 | # to 'b' and 'c' and deletes 'a', and side 2 copies 'a' to 'c' |
|
531 | 532 | # and 'd' and deletes 'a'. |
|
532 | 533 | if dsts1 & dsts2: |
|
533 | 534 | for dst in (dsts1 & dsts2): |
|
534 | 535 | copy[dst] = src |
|
535 | 536 | else: |
|
536 | 537 | diverge[src] = sorted(dsts1 | dsts2) |
|
537 | 538 | elif src in m1 and src in m2: |
|
538 | 539 | # copied on both sides |
|
539 | 540 | dsts1 = set(dsts1) |
|
540 | 541 | dsts2 = set(dsts2) |
|
541 | 542 | for dst in (dsts1 & dsts2): |
|
542 | 543 | copy[dst] = src |
|
543 | 544 | # TODO: Handle cases where it was renamed on one side and copied |
|
544 | 545 | # on the other side |
|
545 | 546 | elif dsts1: |
|
546 | 547 | # copied/renamed only on side 1 |
|
547 | 548 | _checksinglesidecopies(src, dsts1, m1, m2, mb, c2, base, |
|
548 | 549 | copy, renamedelete) |
|
549 | 550 | elif dsts2: |
|
550 | 551 | # copied/renamed only on side 2 |
|
551 | 552 | _checksinglesidecopies(src, dsts2, m2, m1, mb, c1, base, |
|
552 | 553 | copy, renamedelete) |
|
553 | 554 | |
|
554 | 555 | renamedeleteset = set() |
|
555 | 556 | divergeset = set() |
|
556 | 557 | for dsts in diverge.values(): |
|
557 | 558 | divergeset.update(dsts) |
|
558 | 559 | for dsts in renamedelete.values(): |
|
559 | 560 | renamedeleteset.update(dsts) |
|
560 | 561 | |
|
561 | 562 | # find interesting file sets from manifests |
|
562 | 563 | addedinm1 = m1.filesnotin(mb, repo.narrowmatch()) |
|
563 | 564 | addedinm2 = m2.filesnotin(mb, repo.narrowmatch()) |
|
564 | 565 | u1 = sorted(addedinm1 - addedinm2) |
|
565 | 566 | u2 = sorted(addedinm2 - addedinm1) |
|
566 | 567 | |
|
567 | 568 | header = " unmatched files in %s" |
|
568 | 569 | if u1: |
|
569 | 570 | repo.ui.debug("%s:\n %s\n" % (header % 'local', "\n ".join(u1))) |
|
570 | 571 | if u2: |
|
571 | 572 | repo.ui.debug("%s:\n %s\n" % (header % 'other', "\n ".join(u2))) |
|
572 | 573 | |
|
573 | 574 | fullcopy = copies1.copy() |
|
574 | 575 | fullcopy.update(copies2) |
|
575 | 576 | if not fullcopy: |
|
576 | 577 | return copy, {}, diverge, renamedelete, {} |
|
577 | 578 | |
|
578 | 579 | if repo.ui.debugflag: |
|
579 | 580 | repo.ui.debug(" all copies found (* = to merge, ! = divergent, " |
|
580 | 581 | "% = renamed and deleted):\n") |
|
581 | 582 | for f in sorted(fullcopy): |
|
582 | 583 | note = "" |
|
583 | 584 | if f in copy: |
|
584 | 585 | note += "*" |
|
585 | 586 | if f in divergeset: |
|
586 | 587 | note += "!" |
|
587 | 588 | if f in renamedeleteset: |
|
588 | 589 | note += "%" |
|
589 | 590 | repo.ui.debug(" src: '%s' -> dst: '%s' %s\n" % (fullcopy[f], f, |
|
590 | 591 | note)) |
|
591 | 592 | del divergeset |
|
592 | 593 | |
|
593 | 594 | repo.ui.debug(" checking for directory renames\n") |
|
594 | 595 | |
|
595 | 596 | # generate a directory move map |
|
596 | 597 | d1, d2 = c1.dirs(), c2.dirs() |
|
597 | 598 | invalid = set() |
|
598 | 599 | dirmove = {} |
|
599 | 600 | |
|
600 | 601 | # examine each file copy for a potential directory move, which is |
|
601 | 602 | # when all the files in a directory are moved to a new directory |
|
602 | 603 | for dst, src in fullcopy.iteritems(): |
|
603 | 604 | dsrc, ddst = pathutil.dirname(src), pathutil.dirname(dst) |
|
604 | 605 | if dsrc in invalid: |
|
605 | 606 | # already seen to be uninteresting |
|
606 | 607 | continue |
|
607 | 608 | elif dsrc in d1 and ddst in d1: |
|
608 | 609 | # directory wasn't entirely moved locally |
|
609 | 610 | invalid.add(dsrc) |
|
610 | 611 | elif dsrc in d2 and ddst in d2: |
|
611 | 612 | # directory wasn't entirely moved remotely |
|
612 | 613 | invalid.add(dsrc) |
|
613 | 614 | elif dsrc in dirmove and dirmove[dsrc] != ddst: |
|
614 | 615 | # files from the same directory moved to two different places |
|
615 | 616 | invalid.add(dsrc) |
|
616 | 617 | else: |
|
617 | 618 | # looks good so far |
|
618 | 619 | dirmove[dsrc] = ddst |
|
619 | 620 | |
|
620 | 621 | for i in invalid: |
|
621 | 622 | if i in dirmove: |
|
622 | 623 | del dirmove[i] |
|
623 | 624 | del d1, d2, invalid |
|
624 | 625 | |
|
625 | 626 | if not dirmove: |
|
626 | 627 | return copy, {}, diverge, renamedelete, {} |
|
627 | 628 | |
|
628 | 629 | dirmove = {k + "/": v + "/" for k, v in dirmove.iteritems()} |
|
629 | 630 | |
|
630 | 631 | for d in dirmove: |
|
631 | 632 | repo.ui.debug(" discovered dir src: '%s' -> dst: '%s'\n" % |
|
632 | 633 | (d, dirmove[d])) |
|
633 | 634 | |
|
634 | 635 | movewithdir = {} |
|
635 | 636 | # check unaccounted nonoverlapping files against directory moves |
|
636 | 637 | for f in u1 + u2: |
|
637 | 638 | if f not in fullcopy: |
|
638 | 639 | for d in dirmove: |
|
639 | 640 | if f.startswith(d): |
|
640 | 641 | # new file added in a directory that was moved, move it |
|
641 | 642 | df = dirmove[d] + f[len(d):] |
|
642 | 643 | if df not in copy: |
|
643 | 644 | movewithdir[f] = df |
|
644 | 645 | repo.ui.debug((" pending file src: '%s' -> " |
|
645 | 646 | "dst: '%s'\n") % (f, df)) |
|
646 | 647 | break |
|
647 | 648 | |
|
648 | 649 | return copy, movewithdir, diverge, renamedelete, dirmove |
|
649 | 650 | |
|
650 | 651 | def _heuristicscopytracing(repo, c1, c2, base): |
|
651 | 652 | """ Fast copytracing using filename heuristics |
|
652 | 653 | |
|
653 | 654 | Assumes that moves or renames are of following two types: |
|
654 | 655 | |
|
655 | 656 | 1) Inside a directory only (same directory name but different filenames) |
|
656 | 657 | 2) Move from one directory to another |
|
657 | 658 | (same filenames but different directory names) |
|
658 | 659 | |
|
659 | 660 | Works only when there are no merge commits in the "source branch". |
|
660 | 661 | Source branch is commits from base up to c2 not including base. |
|
661 | 662 | |
|
662 | 663 | If merge is involved it fallbacks to _fullcopytracing(). |
|
663 | 664 | |
|
664 | 665 | Can be used by setting the following config: |
|
665 | 666 | |
|
666 | 667 | [experimental] |
|
667 | 668 | copytrace = heuristics |
|
668 | 669 | |
|
669 | 670 | In some cases the copy/move candidates found by heuristics can be very large |
|
670 | 671 | in number and that will make the algorithm slow. The number of possible |
|
671 | 672 | candidates to check can be limited by using the config |
|
672 | 673 | `experimental.copytrace.movecandidateslimit` which defaults to 100. |
|
673 | 674 | """ |
|
674 | 675 | |
|
675 | 676 | if c1.rev() is None: |
|
676 | 677 | c1 = c1.p1() |
|
677 | 678 | if c2.rev() is None: |
|
678 | 679 | c2 = c2.p1() |
|
679 | 680 | |
|
680 | 681 | copies = {} |
|
681 | 682 | |
|
682 | 683 | changedfiles = set() |
|
683 | 684 | m1 = c1.manifest() |
|
684 | 685 | if not repo.revs('%d::%d', base.rev(), c2.rev()): |
|
685 | 686 | # If base is not in c2 branch, we switch to fullcopytracing |
|
686 | 687 | repo.ui.debug("switching to full copytracing as base is not " |
|
687 | 688 | "an ancestor of c2\n") |
|
688 | 689 | return _fullcopytracing(repo, c1, c2, base) |
|
689 | 690 | |
|
690 | 691 | ctx = c2 |
|
691 | 692 | while ctx != base: |
|
692 | 693 | if len(ctx.parents()) == 2: |
|
693 | 694 | # To keep things simple let's not handle merges |
|
694 | 695 | repo.ui.debug("switching to full copytracing because of merges\n") |
|
695 | 696 | return _fullcopytracing(repo, c1, c2, base) |
|
696 | 697 | changedfiles.update(ctx.files()) |
|
697 | 698 | ctx = ctx.p1() |
|
698 | 699 | |
|
699 | 700 | cp = _forwardcopies(base, c2) |
|
700 | 701 | for dst, src in cp.iteritems(): |
|
701 | 702 | if src in m1: |
|
702 | 703 | copies[dst] = src |
|
703 | 704 | |
|
704 | 705 | # file is missing if it isn't present in the destination, but is present in |
|
705 | 706 | # the base and present in the source. |
|
706 | 707 | # Presence in the base is important to exclude added files, presence in the |
|
707 | 708 | # source is important to exclude removed files. |
|
708 | 709 | filt = lambda f: f not in m1 and f in base and f in c2 |
|
709 | 710 | missingfiles = [f for f in changedfiles if filt(f)] |
|
710 | 711 | |
|
711 | 712 | if missingfiles: |
|
712 | 713 | basenametofilename = collections.defaultdict(list) |
|
713 | 714 | dirnametofilename = collections.defaultdict(list) |
|
714 | 715 | |
|
715 | 716 | for f in m1.filesnotin(base.manifest()): |
|
716 | 717 | basename = os.path.basename(f) |
|
717 | 718 | dirname = os.path.dirname(f) |
|
718 | 719 | basenametofilename[basename].append(f) |
|
719 | 720 | dirnametofilename[dirname].append(f) |
|
720 | 721 | |
|
721 | 722 | for f in missingfiles: |
|
722 | 723 | basename = os.path.basename(f) |
|
723 | 724 | dirname = os.path.dirname(f) |
|
724 | 725 | samebasename = basenametofilename[basename] |
|
725 | 726 | samedirname = dirnametofilename[dirname] |
|
726 | 727 | movecandidates = samebasename + samedirname |
|
727 | 728 | # f is guaranteed to be present in c2, that's why |
|
728 | 729 | # c2.filectx(f) won't fail |
|
729 | 730 | f2 = c2.filectx(f) |
|
730 | 731 | # we can have a lot of candidates which can slow down the heuristics |
|
731 | 732 | # config value to limit the number of candidates moves to check |
|
732 | 733 | maxcandidates = repo.ui.configint('experimental', |
|
733 | 734 | 'copytrace.movecandidateslimit') |
|
734 | 735 | |
|
735 | 736 | if len(movecandidates) > maxcandidates: |
|
736 | 737 | repo.ui.status(_("skipping copytracing for '%s', more " |
|
737 | 738 | "candidates than the limit: %d\n") |
|
738 | 739 | % (f, len(movecandidates))) |
|
739 | 740 | continue |
|
740 | 741 | |
|
741 | 742 | for candidate in movecandidates: |
|
742 | 743 | f1 = c1.filectx(candidate) |
|
743 | 744 | if _related(f1, f2): |
|
744 | 745 | # if there are a few related copies then we'll merge |
|
745 | 746 | # changes into all of them. This matches the behaviour |
|
746 | 747 | # of upstream copytracing |
|
747 | 748 | copies[candidate] = f |
|
748 | 749 | |
|
749 | 750 | return copies, {}, {}, {}, {} |
|
750 | 751 | |
|
751 | 752 | def _related(f1, f2): |
|
752 | 753 | """return True if f1 and f2 filectx have a common ancestor |
|
753 | 754 | |
|
754 | 755 | Walk back to common ancestor to see if the two files originate |
|
755 | 756 | from the same file. Since workingfilectx's rev() is None it messes |
|
756 | 757 | up the integer comparison logic, hence the pre-step check for |
|
757 | 758 | None (f1 and f2 can only be workingfilectx's initially). |
|
758 | 759 | """ |
|
759 | 760 | |
|
760 | 761 | if f1 == f2: |
|
761 | 762 | return True # a match |
|
762 | 763 | |
|
763 | 764 | g1, g2 = f1.ancestors(), f2.ancestors() |
|
764 | 765 | try: |
|
765 | 766 | f1r, f2r = f1.linkrev(), f2.linkrev() |
|
766 | 767 | |
|
767 | 768 | if f1r is None: |
|
768 | 769 | f1 = next(g1) |
|
769 | 770 | if f2r is None: |
|
770 | 771 | f2 = next(g2) |
|
771 | 772 | |
|
772 | 773 | while True: |
|
773 | 774 | f1r, f2r = f1.linkrev(), f2.linkrev() |
|
774 | 775 | if f1r > f2r: |
|
775 | 776 | f1 = next(g1) |
|
776 | 777 | elif f2r > f1r: |
|
777 | 778 | f2 = next(g2) |
|
778 | 779 | else: # f1 and f2 point to files in the same linkrev |
|
779 | 780 | return f1 == f2 # true if they point to the same file |
|
780 | 781 | except StopIteration: |
|
781 | 782 | return False |
|
782 | 783 | |
|
783 | 784 | def duplicatecopies(repo, wctx, rev, fromrev, skiprev=None): |
|
784 | 785 | """reproduce copies from fromrev to rev in the dirstate |
|
785 | 786 | |
|
786 | 787 | If skiprev is specified, it's a revision that should be used to |
|
787 | 788 | filter copy records. Any copies that occur between fromrev and |
|
788 | 789 | skiprev will not be duplicated, even if they appear in the set of |
|
789 | 790 | copies between fromrev and rev. |
|
790 | 791 | """ |
|
791 | 792 | exclude = {} |
|
792 | 793 | ctraceconfig = repo.ui.config('experimental', 'copytrace') |
|
793 | 794 | bctrace = stringutil.parsebool(ctraceconfig) |
|
794 | 795 | if (skiprev is not None and |
|
795 | 796 | (ctraceconfig == 'heuristics' or bctrace or bctrace is None)): |
|
796 | 797 | # copytrace='off' skips this line, but not the entire function because |
|
797 | 798 | # the line below is O(size of the repo) during a rebase, while the rest |
|
798 | 799 | # of the function is much faster (and is required for carrying copy |
|
799 | 800 | # metadata across the rebase anyway). |
|
800 | 801 | exclude = pathcopies(repo[fromrev], repo[skiprev]) |
|
801 | 802 | for dst, src in pathcopies(repo[fromrev], repo[rev]).iteritems(): |
|
802 | 803 | if dst in exclude: |
|
803 | 804 | continue |
|
804 | 805 | if dst in wctx: |
|
805 | 806 | wctx[dst].markcopied(src) |
General Comments 0
You need to be logged in to leave comments.
Login now