Show More
@@ -1,427 +1,427 b'' | |||
|
1 | 1 | # dagop.py - graph ancestry and topology algorithm for revset |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2010 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 heapq |
|
11 | 11 | |
|
12 | 12 | from . import ( |
|
13 | 13 | error, |
|
14 | 14 | mdiff, |
|
15 | 15 | node, |
|
16 | 16 | patch, |
|
17 | 17 | smartset, |
|
18 | 18 | ) |
|
19 | 19 | |
|
20 | 20 | baseset = smartset.baseset |
|
21 | 21 | generatorset = smartset.generatorset |
|
22 | 22 | |
|
23 | 23 | def _genrevancestors(repo, revs, followfirst): |
|
24 | 24 | if followfirst: |
|
25 | 25 | cut = 1 |
|
26 | 26 | else: |
|
27 | 27 | cut = None |
|
28 | 28 | cl = repo.changelog |
|
29 | 29 | |
|
30 | 30 | # load input revs lazily to heap so earlier revisions can be yielded |
|
31 | 31 | # without fully computing the input revs |
|
32 | 32 | revs.sort(reverse=True) |
|
33 | 33 | irevs = iter(revs) |
|
34 | h = [] | |
|
34 | pendingheap = [] | |
|
35 | 35 | |
|
36 | 36 | inputrev = next(irevs, None) |
|
37 | 37 | if inputrev is not None: |
|
38 | heapq.heappush(h, -inputrev) | |
|
38 | heapq.heappush(pendingheap, -inputrev) | |
|
39 | 39 | |
|
40 | 40 | seen = set() |
|
41 | while h: | |
|
42 |
curre |
|
|
43 |
if curre |
|
|
41 | while pendingheap: | |
|
42 | currev = -heapq.heappop(pendingheap) | |
|
43 | if currev == inputrev: | |
|
44 | 44 | inputrev = next(irevs, None) |
|
45 | 45 | if inputrev is not None: |
|
46 | heapq.heappush(h, -inputrev) | |
|
47 |
if curre |
|
|
48 |
seen.add(curre |
|
|
49 |
yield curre |
|
|
46 | heapq.heappush(pendingheap, -inputrev) | |
|
47 | if currev not in seen: | |
|
48 | seen.add(currev) | |
|
49 | yield currev | |
|
50 | 50 | try: |
|
51 |
for p |
|
|
52 |
if p |
|
|
53 |
heapq.heappush(h, -p |
|
|
51 | for prev in cl.parentrevs(currev)[:cut]: | |
|
52 | if prev != node.nullrev: | |
|
53 | heapq.heappush(pendingheap, -prev) | |
|
54 | 54 | except error.WdirUnsupported: |
|
55 |
for p |
|
|
56 |
if p |
|
|
57 |
heapq.heappush(h, -p |
|
|
55 | for pctx in repo[currev].parents()[:cut]: | |
|
56 | if pctx.rev() != node.nullrev: | |
|
57 | heapq.heappush(pendingheap, -pctx.rev()) | |
|
58 | 58 | |
|
59 | 59 | def revancestors(repo, revs, followfirst): |
|
60 | 60 | """Like revlog.ancestors(), but supports followfirst.""" |
|
61 | 61 | gen = _genrevancestors(repo, revs, followfirst) |
|
62 | 62 | return generatorset(gen, iterasc=False) |
|
63 | 63 | |
|
64 | 64 | def revdescendants(repo, revs, followfirst): |
|
65 | 65 | """Like revlog.descendants() but supports followfirst.""" |
|
66 | 66 | if followfirst: |
|
67 | 67 | cut = 1 |
|
68 | 68 | else: |
|
69 | 69 | cut = None |
|
70 | 70 | |
|
71 | 71 | def iterate(): |
|
72 | 72 | cl = repo.changelog |
|
73 | 73 | # XXX this should be 'parentset.min()' assuming 'parentset' is a |
|
74 | 74 | # smartset (and if it is not, it should.) |
|
75 | 75 | first = min(revs) |
|
76 | 76 | nullrev = node.nullrev |
|
77 | 77 | if first == nullrev: |
|
78 | 78 | # Are there nodes with a null first parent and a non-null |
|
79 | 79 | # second one? Maybe. Do we care? Probably not. |
|
80 | 80 | for i in cl: |
|
81 | 81 | yield i |
|
82 | 82 | else: |
|
83 | 83 | seen = set(revs) |
|
84 | 84 | for i in cl.revs(first + 1): |
|
85 | 85 | for x in cl.parentrevs(i)[:cut]: |
|
86 | 86 | if x != nullrev and x in seen: |
|
87 | 87 | seen.add(i) |
|
88 | 88 | yield i |
|
89 | 89 | break |
|
90 | 90 | |
|
91 | 91 | return generatorset(iterate(), iterasc=True) |
|
92 | 92 | |
|
93 | 93 | def _reachablerootspure(repo, minroot, roots, heads, includepath): |
|
94 | 94 | """return (heads(::<roots> and ::<heads>)) |
|
95 | 95 | |
|
96 | 96 | If includepath is True, return (<roots>::<heads>).""" |
|
97 | 97 | if not roots: |
|
98 | 98 | return [] |
|
99 | 99 | parentrevs = repo.changelog.parentrevs |
|
100 | 100 | roots = set(roots) |
|
101 | 101 | visit = list(heads) |
|
102 | 102 | reachable = set() |
|
103 | 103 | seen = {} |
|
104 | 104 | # prefetch all the things! (because python is slow) |
|
105 | 105 | reached = reachable.add |
|
106 | 106 | dovisit = visit.append |
|
107 | 107 | nextvisit = visit.pop |
|
108 | 108 | # open-code the post-order traversal due to the tiny size of |
|
109 | 109 | # sys.getrecursionlimit() |
|
110 | 110 | while visit: |
|
111 | 111 | rev = nextvisit() |
|
112 | 112 | if rev in roots: |
|
113 | 113 | reached(rev) |
|
114 | 114 | if not includepath: |
|
115 | 115 | continue |
|
116 | 116 | parents = parentrevs(rev) |
|
117 | 117 | seen[rev] = parents |
|
118 | 118 | for parent in parents: |
|
119 | 119 | if parent >= minroot and parent not in seen: |
|
120 | 120 | dovisit(parent) |
|
121 | 121 | if not reachable: |
|
122 | 122 | return baseset() |
|
123 | 123 | if not includepath: |
|
124 | 124 | return reachable |
|
125 | 125 | for rev in sorted(seen): |
|
126 | 126 | for parent in seen[rev]: |
|
127 | 127 | if parent in reachable: |
|
128 | 128 | reached(rev) |
|
129 | 129 | return reachable |
|
130 | 130 | |
|
131 | 131 | def reachableroots(repo, roots, heads, includepath=False): |
|
132 | 132 | """return (heads(::<roots> and ::<heads>)) |
|
133 | 133 | |
|
134 | 134 | If includepath is True, return (<roots>::<heads>).""" |
|
135 | 135 | if not roots: |
|
136 | 136 | return baseset() |
|
137 | 137 | minroot = roots.min() |
|
138 | 138 | roots = list(roots) |
|
139 | 139 | heads = list(heads) |
|
140 | 140 | try: |
|
141 | 141 | revs = repo.changelog.reachableroots(minroot, heads, roots, includepath) |
|
142 | 142 | except AttributeError: |
|
143 | 143 | revs = _reachablerootspure(repo, minroot, roots, heads, includepath) |
|
144 | 144 | revs = baseset(revs) |
|
145 | 145 | revs.sort() |
|
146 | 146 | return revs |
|
147 | 147 | |
|
148 | 148 | def _changesrange(fctx1, fctx2, linerange2, diffopts): |
|
149 | 149 | """Return `(diffinrange, linerange1)` where `diffinrange` is True |
|
150 | 150 | if diff from fctx2 to fctx1 has changes in linerange2 and |
|
151 | 151 | `linerange1` is the new line range for fctx1. |
|
152 | 152 | """ |
|
153 | 153 | blocks = mdiff.allblocks(fctx1.data(), fctx2.data(), diffopts) |
|
154 | 154 | filteredblocks, linerange1 = mdiff.blocksinrange(blocks, linerange2) |
|
155 | 155 | diffinrange = any(stype == '!' for _, stype in filteredblocks) |
|
156 | 156 | return diffinrange, linerange1 |
|
157 | 157 | |
|
158 | 158 | def blockancestors(fctx, fromline, toline, followfirst=False): |
|
159 | 159 | """Yield ancestors of `fctx` with respect to the block of lines within |
|
160 | 160 | `fromline`-`toline` range. |
|
161 | 161 | """ |
|
162 | 162 | diffopts = patch.diffopts(fctx._repo.ui) |
|
163 | 163 | introrev = fctx.introrev() |
|
164 | 164 | if fctx.rev() != introrev: |
|
165 | 165 | fctx = fctx.filectx(fctx.filenode(), changeid=introrev) |
|
166 | 166 | visit = {(fctx.linkrev(), fctx.filenode()): (fctx, (fromline, toline))} |
|
167 | 167 | while visit: |
|
168 | 168 | c, linerange2 = visit.pop(max(visit)) |
|
169 | 169 | pl = c.parents() |
|
170 | 170 | if followfirst: |
|
171 | 171 | pl = pl[:1] |
|
172 | 172 | if not pl: |
|
173 | 173 | # The block originates from the initial revision. |
|
174 | 174 | yield c, linerange2 |
|
175 | 175 | continue |
|
176 | 176 | inrange = False |
|
177 | 177 | for p in pl: |
|
178 | 178 | inrangep, linerange1 = _changesrange(p, c, linerange2, diffopts) |
|
179 | 179 | inrange = inrange or inrangep |
|
180 | 180 | if linerange1[0] == linerange1[1]: |
|
181 | 181 | # Parent's linerange is empty, meaning that the block got |
|
182 | 182 | # introduced in this revision; no need to go futher in this |
|
183 | 183 | # branch. |
|
184 | 184 | continue |
|
185 | 185 | # Set _descendantrev with 'c' (a known descendant) so that, when |
|
186 | 186 | # _adjustlinkrev is called for 'p', it receives this descendant |
|
187 | 187 | # (as srcrev) instead possibly topmost introrev. |
|
188 | 188 | p._descendantrev = c.rev() |
|
189 | 189 | visit[p.linkrev(), p.filenode()] = p, linerange1 |
|
190 | 190 | if inrange: |
|
191 | 191 | yield c, linerange2 |
|
192 | 192 | |
|
193 | 193 | def blockdescendants(fctx, fromline, toline): |
|
194 | 194 | """Yield descendants of `fctx` with respect to the block of lines within |
|
195 | 195 | `fromline`-`toline` range. |
|
196 | 196 | """ |
|
197 | 197 | # First possibly yield 'fctx' if it has changes in range with respect to |
|
198 | 198 | # its parents. |
|
199 | 199 | try: |
|
200 | 200 | c, linerange1 = next(blockancestors(fctx, fromline, toline)) |
|
201 | 201 | except StopIteration: |
|
202 | 202 | pass |
|
203 | 203 | else: |
|
204 | 204 | if c == fctx: |
|
205 | 205 | yield c, linerange1 |
|
206 | 206 | |
|
207 | 207 | diffopts = patch.diffopts(fctx._repo.ui) |
|
208 | 208 | fl = fctx.filelog() |
|
209 | 209 | seen = {fctx.filerev(): (fctx, (fromline, toline))} |
|
210 | 210 | for i in fl.descendants([fctx.filerev()]): |
|
211 | 211 | c = fctx.filectx(i) |
|
212 | 212 | inrange = False |
|
213 | 213 | for x in fl.parentrevs(i): |
|
214 | 214 | try: |
|
215 | 215 | p, linerange2 = seen[x] |
|
216 | 216 | except KeyError: |
|
217 | 217 | # nullrev or other branch |
|
218 | 218 | continue |
|
219 | 219 | inrangep, linerange1 = _changesrange(c, p, linerange2, diffopts) |
|
220 | 220 | inrange = inrange or inrangep |
|
221 | 221 | # If revision 'i' has been seen (it's a merge), we assume that its |
|
222 | 222 | # line range is the same independently of which parents was used |
|
223 | 223 | # to compute it. |
|
224 | 224 | assert i not in seen or seen[i][1] == linerange1, ( |
|
225 | 225 | 'computed line range for %s is not consistent between ' |
|
226 | 226 | 'ancestor branches' % c) |
|
227 | 227 | seen[i] = c, linerange1 |
|
228 | 228 | if inrange: |
|
229 | 229 | yield c, linerange1 |
|
230 | 230 | |
|
231 | 231 | def toposort(revs, parentsfunc, firstbranch=()): |
|
232 | 232 | """Yield revisions from heads to roots one (topo) branch at a time. |
|
233 | 233 | |
|
234 | 234 | This function aims to be used by a graph generator that wishes to minimize |
|
235 | 235 | the number of parallel branches and their interleaving. |
|
236 | 236 | |
|
237 | 237 | Example iteration order (numbers show the "true" order in a changelog): |
|
238 | 238 | |
|
239 | 239 | o 4 |
|
240 | 240 | | |
|
241 | 241 | o 1 |
|
242 | 242 | | |
|
243 | 243 | | o 3 |
|
244 | 244 | | | |
|
245 | 245 | | o 2 |
|
246 | 246 | |/ |
|
247 | 247 | o 0 |
|
248 | 248 | |
|
249 | 249 | Note that the ancestors of merges are understood by the current |
|
250 | 250 | algorithm to be on the same branch. This means no reordering will |
|
251 | 251 | occur behind a merge. |
|
252 | 252 | """ |
|
253 | 253 | |
|
254 | 254 | ### Quick summary of the algorithm |
|
255 | 255 | # |
|
256 | 256 | # This function is based around a "retention" principle. We keep revisions |
|
257 | 257 | # in memory until we are ready to emit a whole branch that immediately |
|
258 | 258 | # "merges" into an existing one. This reduces the number of parallel |
|
259 | 259 | # branches with interleaved revisions. |
|
260 | 260 | # |
|
261 | 261 | # During iteration revs are split into two groups: |
|
262 | 262 | # A) revision already emitted |
|
263 | 263 | # B) revision in "retention". They are stored as different subgroups. |
|
264 | 264 | # |
|
265 | 265 | # for each REV, we do the following logic: |
|
266 | 266 | # |
|
267 | 267 | # 1) if REV is a parent of (A), we will emit it. If there is a |
|
268 | 268 | # retention group ((B) above) that is blocked on REV being |
|
269 | 269 | # available, we emit all the revisions out of that retention |
|
270 | 270 | # group first. |
|
271 | 271 | # |
|
272 | 272 | # 2) else, we'll search for a subgroup in (B) awaiting for REV to be |
|
273 | 273 | # available, if such subgroup exist, we add REV to it and the subgroup is |
|
274 | 274 | # now awaiting for REV.parents() to be available. |
|
275 | 275 | # |
|
276 | 276 | # 3) finally if no such group existed in (B), we create a new subgroup. |
|
277 | 277 | # |
|
278 | 278 | # |
|
279 | 279 | # To bootstrap the algorithm, we emit the tipmost revision (which |
|
280 | 280 | # puts it in group (A) from above). |
|
281 | 281 | |
|
282 | 282 | revs.sort(reverse=True) |
|
283 | 283 | |
|
284 | 284 | # Set of parents of revision that have been emitted. They can be considered |
|
285 | 285 | # unblocked as the graph generator is already aware of them so there is no |
|
286 | 286 | # need to delay the revisions that reference them. |
|
287 | 287 | # |
|
288 | 288 | # If someone wants to prioritize a branch over the others, pre-filling this |
|
289 | 289 | # set will force all other branches to wait until this branch is ready to be |
|
290 | 290 | # emitted. |
|
291 | 291 | unblocked = set(firstbranch) |
|
292 | 292 | |
|
293 | 293 | # list of groups waiting to be displayed, each group is defined by: |
|
294 | 294 | # |
|
295 | 295 | # (revs: lists of revs waiting to be displayed, |
|
296 | 296 | # blocked: set of that cannot be displayed before those in 'revs') |
|
297 | 297 | # |
|
298 | 298 | # The second value ('blocked') correspond to parents of any revision in the |
|
299 | 299 | # group ('revs') that is not itself contained in the group. The main idea |
|
300 | 300 | # of this algorithm is to delay as much as possible the emission of any |
|
301 | 301 | # revision. This means waiting for the moment we are about to display |
|
302 | 302 | # these parents to display the revs in a group. |
|
303 | 303 | # |
|
304 | 304 | # This first implementation is smart until it encounters a merge: it will |
|
305 | 305 | # emit revs as soon as any parent is about to be emitted and can grow an |
|
306 | 306 | # arbitrary number of revs in 'blocked'. In practice this mean we properly |
|
307 | 307 | # retains new branches but gives up on any special ordering for ancestors |
|
308 | 308 | # of merges. The implementation can be improved to handle this better. |
|
309 | 309 | # |
|
310 | 310 | # The first subgroup is special. It corresponds to all the revision that |
|
311 | 311 | # were already emitted. The 'revs' lists is expected to be empty and the |
|
312 | 312 | # 'blocked' set contains the parents revisions of already emitted revision. |
|
313 | 313 | # |
|
314 | 314 | # You could pre-seed the <parents> set of groups[0] to a specific |
|
315 | 315 | # changesets to select what the first emitted branch should be. |
|
316 | 316 | groups = [([], unblocked)] |
|
317 | 317 | pendingheap = [] |
|
318 | 318 | pendingset = set() |
|
319 | 319 | |
|
320 | 320 | heapq.heapify(pendingheap) |
|
321 | 321 | heappop = heapq.heappop |
|
322 | 322 | heappush = heapq.heappush |
|
323 | 323 | for currentrev in revs: |
|
324 | 324 | # Heap works with smallest element, we want highest so we invert |
|
325 | 325 | if currentrev not in pendingset: |
|
326 | 326 | heappush(pendingheap, -currentrev) |
|
327 | 327 | pendingset.add(currentrev) |
|
328 | 328 | # iterates on pending rev until after the current rev have been |
|
329 | 329 | # processed. |
|
330 | 330 | rev = None |
|
331 | 331 | while rev != currentrev: |
|
332 | 332 | rev = -heappop(pendingheap) |
|
333 | 333 | pendingset.remove(rev) |
|
334 | 334 | |
|
335 | 335 | # Seek for a subgroup blocked, waiting for the current revision. |
|
336 | 336 | matching = [i for i, g in enumerate(groups) if rev in g[1]] |
|
337 | 337 | |
|
338 | 338 | if matching: |
|
339 | 339 | # The main idea is to gather together all sets that are blocked |
|
340 | 340 | # on the same revision. |
|
341 | 341 | # |
|
342 | 342 | # Groups are merged when a common blocking ancestor is |
|
343 | 343 | # observed. For example, given two groups: |
|
344 | 344 | # |
|
345 | 345 | # revs [5, 4] waiting for 1 |
|
346 | 346 | # revs [3, 2] waiting for 1 |
|
347 | 347 | # |
|
348 | 348 | # These two groups will be merged when we process |
|
349 | 349 | # 1. In theory, we could have merged the groups when |
|
350 | 350 | # we added 2 to the group it is now in (we could have |
|
351 | 351 | # noticed the groups were both blocked on 1 then), but |
|
352 | 352 | # the way it works now makes the algorithm simpler. |
|
353 | 353 | # |
|
354 | 354 | # We also always keep the oldest subgroup first. We can |
|
355 | 355 | # probably improve the behavior by having the longest set |
|
356 | 356 | # first. That way, graph algorithms could minimise the length |
|
357 | 357 | # of parallel lines their drawing. This is currently not done. |
|
358 | 358 | targetidx = matching.pop(0) |
|
359 | 359 | trevs, tparents = groups[targetidx] |
|
360 | 360 | for i in matching: |
|
361 | 361 | gr = groups[i] |
|
362 | 362 | trevs.extend(gr[0]) |
|
363 | 363 | tparents |= gr[1] |
|
364 | 364 | # delete all merged subgroups (except the one we kept) |
|
365 | 365 | # (starting from the last subgroup for performance and |
|
366 | 366 | # sanity reasons) |
|
367 | 367 | for i in reversed(matching): |
|
368 | 368 | del groups[i] |
|
369 | 369 | else: |
|
370 | 370 | # This is a new head. We create a new subgroup for it. |
|
371 | 371 | targetidx = len(groups) |
|
372 | 372 | groups.append(([], {rev})) |
|
373 | 373 | |
|
374 | 374 | gr = groups[targetidx] |
|
375 | 375 | |
|
376 | 376 | # We now add the current nodes to this subgroups. This is done |
|
377 | 377 | # after the subgroup merging because all elements from a subgroup |
|
378 | 378 | # that relied on this rev must precede it. |
|
379 | 379 | # |
|
380 | 380 | # we also update the <parents> set to include the parents of the |
|
381 | 381 | # new nodes. |
|
382 | 382 | if rev == currentrev: # only display stuff in rev |
|
383 | 383 | gr[0].append(rev) |
|
384 | 384 | gr[1].remove(rev) |
|
385 | 385 | parents = [p for p in parentsfunc(rev) if p > node.nullrev] |
|
386 | 386 | gr[1].update(parents) |
|
387 | 387 | for p in parents: |
|
388 | 388 | if p not in pendingset: |
|
389 | 389 | pendingset.add(p) |
|
390 | 390 | heappush(pendingheap, -p) |
|
391 | 391 | |
|
392 | 392 | # Look for a subgroup to display |
|
393 | 393 | # |
|
394 | 394 | # When unblocked is empty (if clause), we were not waiting for any |
|
395 | 395 | # revisions during the first iteration (if no priority was given) or |
|
396 | 396 | # if we emitted a whole disconnected set of the graph (reached a |
|
397 | 397 | # root). In that case we arbitrarily take the oldest known |
|
398 | 398 | # subgroup. The heuristic could probably be better. |
|
399 | 399 | # |
|
400 | 400 | # Otherwise (elif clause) if the subgroup is blocked on |
|
401 | 401 | # a revision we just emitted, we can safely emit it as |
|
402 | 402 | # well. |
|
403 | 403 | if not unblocked: |
|
404 | 404 | if len(groups) > 1: # display other subset |
|
405 | 405 | targetidx = 1 |
|
406 | 406 | gr = groups[1] |
|
407 | 407 | elif not gr[1] & unblocked: |
|
408 | 408 | gr = None |
|
409 | 409 | |
|
410 | 410 | if gr is not None: |
|
411 | 411 | # update the set of awaited revisions with the one from the |
|
412 | 412 | # subgroup |
|
413 | 413 | unblocked |= gr[1] |
|
414 | 414 | # output all revisions in the subgroup |
|
415 | 415 | for r in gr[0]: |
|
416 | 416 | yield r |
|
417 | 417 | # delete the subgroup that you just output |
|
418 | 418 | # unless it is groups[0] in which case you just empty it. |
|
419 | 419 | if targetidx: |
|
420 | 420 | del groups[targetidx] |
|
421 | 421 | else: |
|
422 | 422 | gr[0][:] = [] |
|
423 | 423 | # Check if we have some subgroup waiting for revisions we are not going to |
|
424 | 424 | # iterate over |
|
425 | 425 | for g in groups: |
|
426 | 426 | for r in g[0]: |
|
427 | 427 | yield r |
General Comments 0
You need to be logged in to leave comments.
Login now