Show More
@@ -1,368 +1,374 b'' | |||
|
1 | 1 | # ancestor.py - generic DAG ancestor algorithm for mercurial |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2006 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 .node import nullrev |
|
13 | 13 | from . import ( |
|
14 | 14 | pycompat, |
|
15 | 15 | ) |
|
16 | 16 | |
|
17 | 17 | def commonancestorsheads(pfunc, *nodes): |
|
18 | 18 | """Returns a set with the heads of all common ancestors of all nodes, |
|
19 | 19 | heads(::nodes[0] and ::nodes[1] and ...) . |
|
20 | 20 | |
|
21 | 21 | pfunc must return a list of parent vertices for a given vertex. |
|
22 | 22 | """ |
|
23 | 23 | if not isinstance(nodes, set): |
|
24 | 24 | nodes = set(nodes) |
|
25 | 25 | if nullrev in nodes: |
|
26 | 26 | return set() |
|
27 | 27 | if len(nodes) <= 1: |
|
28 | 28 | return nodes |
|
29 | 29 | |
|
30 | 30 | allseen = (1 << len(nodes)) - 1 |
|
31 | 31 | seen = [0] * (max(nodes) + 1) |
|
32 | 32 | for i, n in enumerate(nodes): |
|
33 | 33 | seen[n] = 1 << i |
|
34 | 34 | poison = 1 << (i + 1) |
|
35 | 35 | |
|
36 | 36 | gca = set() |
|
37 | 37 | interesting = len(nodes) |
|
38 | 38 | nv = len(seen) - 1 |
|
39 | 39 | while nv >= 0 and interesting: |
|
40 | 40 | v = nv |
|
41 | 41 | nv -= 1 |
|
42 | 42 | if not seen[v]: |
|
43 | 43 | continue |
|
44 | 44 | sv = seen[v] |
|
45 | 45 | if sv < poison: |
|
46 | 46 | interesting -= 1 |
|
47 | 47 | if sv == allseen: |
|
48 | 48 | gca.add(v) |
|
49 | 49 | sv |= poison |
|
50 | 50 | if v in nodes: |
|
51 | 51 | # history is linear |
|
52 | 52 | return {v} |
|
53 | 53 | if sv < poison: |
|
54 | 54 | for p in pfunc(v): |
|
55 | 55 | sp = seen[p] |
|
56 | 56 | if p == nullrev: |
|
57 | 57 | continue |
|
58 | 58 | if sp == 0: |
|
59 | 59 | seen[p] = sv |
|
60 | 60 | interesting += 1 |
|
61 | 61 | elif sp != sv: |
|
62 | 62 | seen[p] |= sv |
|
63 | 63 | else: |
|
64 | 64 | for p in pfunc(v): |
|
65 | 65 | if p == nullrev: |
|
66 | 66 | continue |
|
67 | 67 | sp = seen[p] |
|
68 | 68 | if sp and sp < poison: |
|
69 | 69 | interesting -= 1 |
|
70 | 70 | seen[p] = sv |
|
71 | 71 | return gca |
|
72 | 72 | |
|
73 | 73 | def ancestors(pfunc, *orignodes): |
|
74 | 74 | """ |
|
75 | 75 | Returns the common ancestors of a and b that are furthest from a |
|
76 | 76 | root (as measured by longest path). |
|
77 | 77 | |
|
78 | 78 | pfunc must return a list of parent vertices for a given vertex. |
|
79 | 79 | """ |
|
80 | 80 | def deepest(nodes): |
|
81 | 81 | interesting = {} |
|
82 | 82 | count = max(nodes) + 1 |
|
83 | 83 | depth = [0] * count |
|
84 | 84 | seen = [0] * count |
|
85 | 85 | mapping = [] |
|
86 | 86 | for (i, n) in enumerate(sorted(nodes)): |
|
87 | 87 | depth[n] = 1 |
|
88 | 88 | b = 1 << i |
|
89 | 89 | seen[n] = b |
|
90 | 90 | interesting[b] = 1 |
|
91 | 91 | mapping.append((b, n)) |
|
92 | 92 | nv = count - 1 |
|
93 | 93 | while nv >= 0 and len(interesting) > 1: |
|
94 | 94 | v = nv |
|
95 | 95 | nv -= 1 |
|
96 | 96 | dv = depth[v] |
|
97 | 97 | if dv == 0: |
|
98 | 98 | continue |
|
99 | 99 | sv = seen[v] |
|
100 | 100 | for p in pfunc(v): |
|
101 | 101 | if p == nullrev: |
|
102 | 102 | continue |
|
103 | 103 | dp = depth[p] |
|
104 | 104 | nsp = sp = seen[p] |
|
105 | 105 | if dp <= dv: |
|
106 | 106 | depth[p] = dv + 1 |
|
107 | 107 | if sp != sv: |
|
108 | 108 | interesting[sv] += 1 |
|
109 | 109 | nsp = seen[p] = sv |
|
110 | 110 | if sp: |
|
111 | 111 | interesting[sp] -= 1 |
|
112 | 112 | if interesting[sp] == 0: |
|
113 | 113 | del interesting[sp] |
|
114 | 114 | elif dv == dp - 1: |
|
115 | 115 | nsp = sp | sv |
|
116 | 116 | if nsp == sp: |
|
117 | 117 | continue |
|
118 | 118 | seen[p] = nsp |
|
119 | 119 | interesting.setdefault(nsp, 0) |
|
120 | 120 | interesting[nsp] += 1 |
|
121 | 121 | interesting[sp] -= 1 |
|
122 | 122 | if interesting[sp] == 0: |
|
123 | 123 | del interesting[sp] |
|
124 | 124 | interesting[sv] -= 1 |
|
125 | 125 | if interesting[sv] == 0: |
|
126 | 126 | del interesting[sv] |
|
127 | 127 | |
|
128 | 128 | if len(interesting) != 1: |
|
129 | 129 | return [] |
|
130 | 130 | |
|
131 | 131 | k = 0 |
|
132 | 132 | for i in interesting: |
|
133 | 133 | k |= i |
|
134 | 134 | return set(n for (i, n) in mapping if k & i) |
|
135 | 135 | |
|
136 | 136 | gca = commonancestorsheads(pfunc, *orignodes) |
|
137 | 137 | |
|
138 | 138 | if len(gca) <= 1: |
|
139 | 139 | return gca |
|
140 | 140 | return deepest(gca) |
|
141 | 141 | |
|
142 | 142 | class incrementalmissingancestors(object): |
|
143 | 143 | '''persistent state used to calculate missing ancestors incrementally |
|
144 | 144 | |
|
145 | 145 | Although similar in spirit to lazyancestors below, this is a separate class |
|
146 | 146 | because trying to support contains and missingancestors operations with the |
|
147 | 147 | same internal data structures adds needless complexity.''' |
|
148 | 148 | def __init__(self, pfunc, bases): |
|
149 | 149 | self.bases = set(bases) |
|
150 | 150 | if not self.bases: |
|
151 | 151 | self.bases.add(nullrev) |
|
152 | 152 | self.pfunc = pfunc |
|
153 | 153 | |
|
154 | 154 | def hasbases(self): |
|
155 | 155 | '''whether the common set has any non-trivial bases''' |
|
156 | 156 | return self.bases and self.bases != {nullrev} |
|
157 | 157 | |
|
158 | 158 | def addbases(self, newbases): |
|
159 | 159 | '''grow the ancestor set by adding new bases''' |
|
160 | 160 | self.bases.update(newbases) |
|
161 | 161 | |
|
162 | 162 | def removeancestorsfrom(self, revs): |
|
163 | 163 | '''remove all ancestors of bases from the set revs (in place)''' |
|
164 | 164 | bases = self.bases |
|
165 | 165 | pfunc = self.pfunc |
|
166 | 166 | revs.difference_update(bases) |
|
167 | 167 | # nullrev is always an ancestor |
|
168 | 168 | revs.discard(nullrev) |
|
169 | 169 | if not revs: |
|
170 | 170 | return |
|
171 | 171 | # anything in revs > start is definitely not an ancestor of bases |
|
172 | 172 | # revs <= start needs to be investigated |
|
173 | 173 | start = max(bases) |
|
174 | 174 | keepcount = sum(1 for r in revs if r > start) |
|
175 | 175 | if len(revs) == keepcount: |
|
176 | 176 | # no revs to consider |
|
177 | 177 | return |
|
178 | 178 | |
|
179 | 179 | for curr in pycompat.xrange(start, min(revs) - 1, -1): |
|
180 | 180 | if curr not in bases: |
|
181 | 181 | continue |
|
182 | 182 | revs.discard(curr) |
|
183 | 183 | bases.update(pfunc(curr)) |
|
184 | 184 | if len(revs) == keepcount: |
|
185 | 185 | # no more potential revs to discard |
|
186 | 186 | break |
|
187 | 187 | |
|
188 | 188 | def missingancestors(self, revs): |
|
189 | 189 | '''return all the ancestors of revs that are not ancestors of self.bases |
|
190 | 190 | |
|
191 | 191 | This may include elements from revs. |
|
192 | 192 | |
|
193 | 193 | Equivalent to the revset (::revs - ::self.bases). Revs are returned in |
|
194 | 194 | revision number order, which is a topological order.''' |
|
195 | 195 | revsvisit = set(revs) |
|
196 | 196 | basesvisit = self.bases |
|
197 | 197 | pfunc = self.pfunc |
|
198 | 198 | bothvisit = revsvisit.intersection(basesvisit) |
|
199 | 199 | revsvisit.difference_update(bothvisit) |
|
200 | 200 | if not revsvisit: |
|
201 | 201 | return [] |
|
202 | 202 | |
|
203 | 203 | start = max(max(revsvisit), max(basesvisit)) |
|
204 | 204 | # At this point, we hold the invariants that: |
|
205 | 205 | # - revsvisit is the set of nodes we know are an ancestor of at least |
|
206 | 206 | # one of the nodes in revs |
|
207 | 207 | # - basesvisit is the same for bases |
|
208 | 208 | # - bothvisit is the set of nodes we know are ancestors of at least one |
|
209 | 209 | # of the nodes in revs and one of the nodes in bases. bothvisit and |
|
210 | 210 | # revsvisit are mutually exclusive, but bothvisit is a subset of |
|
211 | 211 | # basesvisit. |
|
212 | 212 | # Now we walk down in reverse topo order, adding parents of nodes |
|
213 | 213 | # already visited to the sets while maintaining the invariants. When a |
|
214 | 214 | # node is found in both revsvisit and basesvisit, it is removed from |
|
215 | 215 | # revsvisit and added to bothvisit. When revsvisit becomes empty, there |
|
216 | 216 | # are no more ancestors of revs that aren't also ancestors of bases, so |
|
217 | 217 | # exit. |
|
218 | 218 | |
|
219 | 219 | missing = [] |
|
220 | 220 | for curr in pycompat.xrange(start, nullrev, -1): |
|
221 | 221 | if not revsvisit: |
|
222 | 222 | break |
|
223 | 223 | |
|
224 | 224 | if curr in bothvisit: |
|
225 | 225 | bothvisit.remove(curr) |
|
226 | 226 | # curr's parents might have made it into revsvisit through |
|
227 | 227 | # another path |
|
228 | 228 | for p in pfunc(curr): |
|
229 | 229 | revsvisit.discard(p) |
|
230 | 230 | basesvisit.add(p) |
|
231 | 231 | bothvisit.add(p) |
|
232 | 232 | continue |
|
233 | 233 | |
|
234 | 234 | if curr in revsvisit: |
|
235 | 235 | missing.append(curr) |
|
236 | 236 | revsvisit.remove(curr) |
|
237 | 237 | thisvisit = revsvisit |
|
238 | 238 | othervisit = basesvisit |
|
239 | 239 | elif curr in basesvisit: |
|
240 | 240 | thisvisit = basesvisit |
|
241 | 241 | othervisit = revsvisit |
|
242 | 242 | else: |
|
243 | 243 | # not an ancestor of revs or bases: ignore |
|
244 | 244 | continue |
|
245 | 245 | |
|
246 | 246 | for p in pfunc(curr): |
|
247 | 247 | if p == nullrev: |
|
248 | 248 | pass |
|
249 | 249 | elif p in othervisit or p in bothvisit: |
|
250 | 250 | # p is implicitly in thisvisit. This means p is or should be |
|
251 | 251 | # in bothvisit |
|
252 | 252 | revsvisit.discard(p) |
|
253 | 253 | basesvisit.add(p) |
|
254 | 254 | bothvisit.add(p) |
|
255 | 255 | else: |
|
256 | 256 | # visit later |
|
257 | 257 | thisvisit.add(p) |
|
258 | 258 | |
|
259 | 259 | missing.reverse() |
|
260 | 260 | return missing |
|
261 | 261 | |
|
262 | 262 | # Extracted from lazyancestors.__iter__ to avoid a reference cycle |
|
263 | 263 | def _lazyancestorsiter(parentrevs, initrevs, stoprev, inclusive): |
|
264 | 264 | seen = {nullrev} |
|
265 | 265 | schedule = heapq.heappush |
|
266 | 266 | nextitem = heapq.heappop |
|
267 | 267 | see = seen.add |
|
268 | 268 | |
|
269 | 269 | if inclusive: |
|
270 | 270 | visit = [-r for r in initrevs] |
|
271 | 271 | seen.update(initrevs) |
|
272 | 272 | heapq.heapify(visit) |
|
273 | 273 | else: |
|
274 | 274 | visit = [] |
|
275 | 275 | heapq.heapify(visit) |
|
276 | 276 | for r in initrevs: |
|
277 |
|
|
|
278 |
|
|
|
279 |
|
|
|
280 |
|
|
|
277 | p1, p2 = parentrevs(r) | |
|
278 | if p1 not in seen: | |
|
279 | schedule(visit, -p1) | |
|
280 | see(p1) | |
|
281 | if p2 not in seen: | |
|
282 | schedule(visit, -p2) | |
|
283 | see(p2) | |
|
281 | 284 | |
|
282 | 285 | while visit: |
|
283 | 286 | current = -nextitem(visit) |
|
284 | 287 | if current < stoprev: |
|
285 | 288 | break |
|
286 | 289 | yield current |
|
287 |
|
|
|
288 |
|
|
|
289 |
|
|
|
290 |
|
|
|
290 | p1, p2 = parentrevs(current) | |
|
291 | if p1 not in seen: | |
|
292 | schedule(visit, -p1) | |
|
293 | see(p1) | |
|
294 | if p2 not in seen: | |
|
295 | schedule(visit, -p2) | |
|
296 | see(p2) | |
|
291 | 297 | |
|
292 | 298 | class lazyancestors(object): |
|
293 | 299 | def __init__(self, pfunc, revs, stoprev=0, inclusive=False): |
|
294 | 300 | """Create a new object generating ancestors for the given revs. Does |
|
295 | 301 | not generate revs lower than stoprev. |
|
296 | 302 | |
|
297 | 303 | This is computed lazily starting from revs. The object supports |
|
298 | 304 | iteration and membership. |
|
299 | 305 | |
|
300 | 306 | cl should be a changelog and revs should be an iterable. inclusive is |
|
301 | 307 | a boolean that indicates whether revs should be included. Revs lower |
|
302 | 308 | than stoprev will not be generated. |
|
303 | 309 | |
|
304 | 310 | Result does not include the null revision.""" |
|
305 | 311 | self._parentrevs = pfunc |
|
306 | 312 | self._initrevs = revs = [r for r in revs if r >= stoprev] |
|
307 | 313 | self._stoprev = stoprev |
|
308 | 314 | self._inclusive = inclusive |
|
309 | 315 | |
|
310 | 316 | self._containsseen = set() |
|
311 | 317 | self._containsiter = _lazyancestorsiter(self._parentrevs, |
|
312 | 318 | self._initrevs, |
|
313 | 319 | self._stoprev, |
|
314 | 320 | self._inclusive) |
|
315 | 321 | |
|
316 | 322 | def __nonzero__(self): |
|
317 | 323 | """False if the set is empty, True otherwise.""" |
|
318 | 324 | try: |
|
319 | 325 | next(iter(self)) |
|
320 | 326 | return True |
|
321 | 327 | except StopIteration: |
|
322 | 328 | return False |
|
323 | 329 | |
|
324 | 330 | __bool__ = __nonzero__ |
|
325 | 331 | |
|
326 | 332 | def __iter__(self): |
|
327 | 333 | """Generate the ancestors of _initrevs in reverse topological order. |
|
328 | 334 | |
|
329 | 335 | If inclusive is False, yield a sequence of revision numbers starting |
|
330 | 336 | with the parents of each revision in revs, i.e., each revision is |
|
331 | 337 | *not* considered an ancestor of itself. Results are emitted in reverse |
|
332 | 338 | revision number order. That order is also topological: a child is |
|
333 | 339 | always emitted before its parent. |
|
334 | 340 | |
|
335 | 341 | If inclusive is True, the source revisions are also yielded. The |
|
336 | 342 | reverse revision number order is still enforced.""" |
|
337 | 343 | for rev in _lazyancestorsiter(self._parentrevs, self._initrevs, |
|
338 | 344 | self._stoprev, self._inclusive): |
|
339 | 345 | yield rev |
|
340 | 346 | |
|
341 | 347 | def __contains__(self, target): |
|
342 | 348 | """Test whether target is an ancestor of self._initrevs.""" |
|
343 | 349 | seen = self._containsseen |
|
344 | 350 | if target in seen: |
|
345 | 351 | return True |
|
346 | 352 | iter = self._containsiter |
|
347 | 353 | if iter is None: |
|
348 | 354 | # Iterator exhausted |
|
349 | 355 | return False |
|
350 | 356 | # Only integer target is valid, but some callers expect 'None in self' |
|
351 | 357 | # to be False. So we explicitly allow it. |
|
352 | 358 | if target is None: |
|
353 | 359 | return False |
|
354 | 360 | |
|
355 | 361 | see = seen.add |
|
356 | 362 | try: |
|
357 | 363 | while True: |
|
358 | 364 | rev = next(iter) |
|
359 | 365 | see(rev) |
|
360 | 366 | if rev == target: |
|
361 | 367 | return True |
|
362 | 368 | if rev < target: |
|
363 | 369 | return False |
|
364 | 370 | except StopIteration: |
|
365 | 371 | # Set to None to indicate fast-path can be used next time, and to |
|
366 | 372 | # free up memory. |
|
367 | 373 | self._containsiter = None |
|
368 | 374 | return False |
@@ -1,280 +1,280 b'' | |||
|
1 | 1 | from __future__ import absolute_import, print_function |
|
2 | 2 | |
|
3 | 3 | import binascii |
|
4 | 4 | import getopt |
|
5 | 5 | import math |
|
6 | 6 | import os |
|
7 | 7 | import random |
|
8 | 8 | import sys |
|
9 | 9 | import time |
|
10 | 10 | |
|
11 | 11 | from mercurial.node import nullrev |
|
12 | 12 | from mercurial import ( |
|
13 | 13 | ancestor, |
|
14 | 14 | debugcommands, |
|
15 | 15 | hg, |
|
16 | 16 | pycompat, |
|
17 | 17 | ui as uimod, |
|
18 | 18 | util, |
|
19 | 19 | ) |
|
20 | 20 | |
|
21 | 21 | if pycompat.ispy3: |
|
22 | 22 | long = int |
|
23 | 23 | xrange = range |
|
24 | 24 | |
|
25 | 25 | def buildgraph(rng, nodes=100, rootprob=0.05, mergeprob=0.2, prevprob=0.7): |
|
26 | 26 | '''nodes: total number of nodes in the graph |
|
27 | 27 | rootprob: probability that a new node (not 0) will be a root |
|
28 | 28 | mergeprob: probability that, excluding a root a node will be a merge |
|
29 | 29 | prevprob: probability that p1 will be the previous node |
|
30 | 30 | |
|
31 | 31 | return value is a graph represented as an adjacency list. |
|
32 | 32 | ''' |
|
33 | 33 | graph = [None] * nodes |
|
34 | 34 | for i in xrange(nodes): |
|
35 | 35 | if i == 0 or rng.random() < rootprob: |
|
36 | 36 | graph[i] = [nullrev] |
|
37 | 37 | elif i == 1: |
|
38 | 38 | graph[i] = [0] |
|
39 | 39 | elif rng.random() < mergeprob: |
|
40 | 40 | if i == 2 or rng.random() < prevprob: |
|
41 | 41 | # p1 is prev |
|
42 | 42 | p1 = i - 1 |
|
43 | 43 | else: |
|
44 | 44 | p1 = rng.randrange(i - 1) |
|
45 | 45 | p2 = rng.choice(list(range(0, p1)) + list(range(p1 + 1, i))) |
|
46 | 46 | graph[i] = [p1, p2] |
|
47 | 47 | elif rng.random() < prevprob: |
|
48 | 48 | graph[i] = [i - 1] |
|
49 | 49 | else: |
|
50 | 50 | graph[i] = [rng.randrange(i - 1)] |
|
51 | 51 | |
|
52 | 52 | return graph |
|
53 | 53 | |
|
54 | 54 | def buildancestorsets(graph): |
|
55 | 55 | ancs = [None] * len(graph) |
|
56 | 56 | for i in xrange(len(graph)): |
|
57 | 57 | ancs[i] = {i} |
|
58 | 58 | if graph[i] == [nullrev]: |
|
59 | 59 | continue |
|
60 | 60 | for p in graph[i]: |
|
61 | 61 | ancs[i].update(ancs[p]) |
|
62 | 62 | return ancs |
|
63 | 63 | |
|
64 | 64 | class naiveincrementalmissingancestors(object): |
|
65 | 65 | def __init__(self, ancs, bases): |
|
66 | 66 | self.ancs = ancs |
|
67 | 67 | self.bases = set(bases) |
|
68 | 68 | def addbases(self, newbases): |
|
69 | 69 | self.bases.update(newbases) |
|
70 | 70 | def removeancestorsfrom(self, revs): |
|
71 | 71 | for base in self.bases: |
|
72 | 72 | if base != nullrev: |
|
73 | 73 | revs.difference_update(self.ancs[base]) |
|
74 | 74 | revs.discard(nullrev) |
|
75 | 75 | def missingancestors(self, revs): |
|
76 | 76 | res = set() |
|
77 | 77 | for rev in revs: |
|
78 | 78 | if rev != nullrev: |
|
79 | 79 | res.update(self.ancs[rev]) |
|
80 | 80 | for base in self.bases: |
|
81 | 81 | if base != nullrev: |
|
82 | 82 | res.difference_update(self.ancs[base]) |
|
83 | 83 | return sorted(res) |
|
84 | 84 | |
|
85 | 85 | def test_missingancestors(seed, rng): |
|
86 | 86 | # empirically observed to take around 1 second |
|
87 | 87 | graphcount = 100 |
|
88 | 88 | testcount = 10 |
|
89 | 89 | inccount = 10 |
|
90 | 90 | nerrs = [0] |
|
91 | 91 | # the default mu and sigma give us a nice distribution of mostly |
|
92 | 92 | # single-digit counts (including 0) with some higher ones |
|
93 | 93 | def lognormrandom(mu, sigma): |
|
94 | 94 | return int(math.floor(rng.lognormvariate(mu, sigma))) |
|
95 | 95 | |
|
96 | 96 | def samplerevs(nodes, mu=1.1, sigma=0.8): |
|
97 | 97 | count = min(lognormrandom(mu, sigma), len(nodes)) |
|
98 | 98 | return rng.sample(nodes, count) |
|
99 | 99 | |
|
100 | 100 | def err(seed, graph, bases, seq, output, expected): |
|
101 | 101 | if nerrs[0] == 0: |
|
102 | 102 | print('seed:', hex(seed)[:-1], file=sys.stderr) |
|
103 | 103 | if gerrs[0] == 0: |
|
104 | 104 | print('graph:', graph, file=sys.stderr) |
|
105 | 105 | print('* bases:', bases, file=sys.stderr) |
|
106 | 106 | print('* seq: ', seq, file=sys.stderr) |
|
107 | 107 | print('* output: ', output, file=sys.stderr) |
|
108 | 108 | print('* expected:', expected, file=sys.stderr) |
|
109 | 109 | nerrs[0] += 1 |
|
110 | 110 | gerrs[0] += 1 |
|
111 | 111 | |
|
112 | 112 | for g in xrange(graphcount): |
|
113 | 113 | graph = buildgraph(rng) |
|
114 | 114 | ancs = buildancestorsets(graph) |
|
115 | 115 | gerrs = [0] |
|
116 | 116 | for _ in xrange(testcount): |
|
117 | 117 | # start from nullrev to include it as a possibility |
|
118 | 118 | graphnodes = range(nullrev, len(graph)) |
|
119 | 119 | bases = samplerevs(graphnodes) |
|
120 | 120 | |
|
121 | 121 | # fast algorithm |
|
122 | 122 | inc = ancestor.incrementalmissingancestors(graph.__getitem__, bases) |
|
123 | 123 | # reference slow algorithm |
|
124 | 124 | naiveinc = naiveincrementalmissingancestors(ancs, bases) |
|
125 | 125 | seq = [] |
|
126 | 126 | revs = [] |
|
127 | 127 | for _ in xrange(inccount): |
|
128 | 128 | if rng.random() < 0.2: |
|
129 | 129 | newbases = samplerevs(graphnodes) |
|
130 | 130 | seq.append(('addbases', newbases)) |
|
131 | 131 | inc.addbases(newbases) |
|
132 | 132 | naiveinc.addbases(newbases) |
|
133 | 133 | if rng.random() < 0.4: |
|
134 | 134 | # larger set so that there are more revs to remove from |
|
135 | 135 | revs = samplerevs(graphnodes, mu=1.5) |
|
136 | 136 | seq.append(('removeancestorsfrom', revs)) |
|
137 | 137 | hrevs = set(revs) |
|
138 | 138 | rrevs = set(revs) |
|
139 | 139 | inc.removeancestorsfrom(hrevs) |
|
140 | 140 | naiveinc.removeancestorsfrom(rrevs) |
|
141 | 141 | if hrevs != rrevs: |
|
142 | 142 | err(seed, graph, bases, seq, sorted(hrevs), |
|
143 | 143 | sorted(rrevs)) |
|
144 | 144 | else: |
|
145 | 145 | revs = samplerevs(graphnodes) |
|
146 | 146 | seq.append(('missingancestors', revs)) |
|
147 | 147 | h = inc.missingancestors(revs) |
|
148 | 148 | r = naiveinc.missingancestors(revs) |
|
149 | 149 | if h != r: |
|
150 | 150 | err(seed, graph, bases, seq, h, r) |
|
151 | 151 | |
|
152 | 152 | # graph is a dict of child->parent adjacency lists for this graph: |
|
153 | 153 | # o 13 |
|
154 | 154 | # | |
|
155 | 155 | # | o 12 |
|
156 | 156 | # | | |
|
157 | 157 | # | | o 11 |
|
158 | 158 | # | | |\ |
|
159 | 159 | # | | | | o 10 |
|
160 | 160 | # | | | | | |
|
161 | 161 | # | o---+ | 9 |
|
162 | 162 | # | | | | | |
|
163 | 163 | # o | | | | 8 |
|
164 | 164 | # / / / / |
|
165 | 165 | # | | o | 7 |
|
166 | 166 | # | | | | |
|
167 | 167 | # o---+ | 6 |
|
168 | 168 | # / / / |
|
169 | 169 | # | | o 5 |
|
170 | 170 | # | |/ |
|
171 | 171 | # | o 4 |
|
172 | 172 | # | | |
|
173 | 173 | # o | 3 |
|
174 | 174 | # | | |
|
175 | 175 | # | o 2 |
|
176 | 176 | # |/ |
|
177 | 177 | # o 1 |
|
178 | 178 | # | |
|
179 | 179 | # o 0 |
|
180 | 180 | |
|
181 |
graph = {0: [-1], 1: [0], 2: [1], 3: [1], 4: [2 |
|
|
182 | 7: [4], 8: [-1], 9: [6, 7], 10: [5], 11: [3, 7], 12: [9], | |
|
183 | 13: [8]} | |
|
181 | graph = {0: [-1, -1], 1: [0, -1], 2: [1, -1], 3: [1, -1], 4: [2, -1], | |
|
182 | 5: [4, -1], 6: [4, -1], 7: [4, -1], 8: [-1, -1], 9: [6, 7], | |
|
183 | 10: [5, -1], 11: [3, 7], 12: [9, -1], 13: [8, -1]} | |
|
184 | 184 | |
|
185 | 185 | def genlazyancestors(revs, stoprev=0, inclusive=False): |
|
186 | 186 | print(("%% lazy ancestor set for %s, stoprev = %s, inclusive = %s" % |
|
187 | 187 | (revs, stoprev, inclusive))) |
|
188 | 188 | return ancestor.lazyancestors(graph.get, revs, stoprev=stoprev, |
|
189 | 189 | inclusive=inclusive) |
|
190 | 190 | |
|
191 | 191 | def printlazyancestors(s, l): |
|
192 | 192 | print('membership: %r' % [n for n in l if n in s]) |
|
193 | 193 | print('iteration: %r' % list(s)) |
|
194 | 194 | |
|
195 | 195 | def test_lazyancestors(): |
|
196 | 196 | # Empty revs |
|
197 | 197 | s = genlazyancestors([]) |
|
198 | 198 | printlazyancestors(s, [3, 0, -1]) |
|
199 | 199 | |
|
200 | 200 | # Standard example |
|
201 | 201 | s = genlazyancestors([11, 13]) |
|
202 | 202 | printlazyancestors(s, [11, 13, 7, 9, 8, 3, 6, 4, 1, -1, 0]) |
|
203 | 203 | |
|
204 | 204 | # Standard with ancestry in the initial set (1 is ancestor of 3) |
|
205 | 205 | s = genlazyancestors([1, 3]) |
|
206 | 206 | printlazyancestors(s, [1, -1, 0]) |
|
207 | 207 | |
|
208 | 208 | # Including revs |
|
209 | 209 | s = genlazyancestors([11, 13], inclusive=True) |
|
210 | 210 | printlazyancestors(s, [11, 13, 7, 9, 8, 3, 6, 4, 1, -1, 0]) |
|
211 | 211 | |
|
212 | 212 | # Test with stoprev |
|
213 | 213 | s = genlazyancestors([11, 13], stoprev=6) |
|
214 | 214 | printlazyancestors(s, [11, 13, 7, 9, 8, 3, 6, 4, 1, -1, 0]) |
|
215 | 215 | s = genlazyancestors([11, 13], stoprev=6, inclusive=True) |
|
216 | 216 | printlazyancestors(s, [11, 13, 7, 9, 8, 3, 6, 4, 1, -1, 0]) |
|
217 | 217 | |
|
218 | 218 | # Test with stoprev >= min(initrevs) |
|
219 | 219 | s = genlazyancestors([11, 13], stoprev=11, inclusive=True) |
|
220 | 220 | printlazyancestors(s, [11, 13, 7, 9, 8, 3, 6, 4, 1, -1, 0]) |
|
221 | 221 | s = genlazyancestors([11, 13], stoprev=12, inclusive=True) |
|
222 | 222 | printlazyancestors(s, [11, 13, 7, 9, 8, 3, 6, 4, 1, -1, 0]) |
|
223 | 223 | |
|
224 | 224 | # The C gca algorithm requires a real repo. These are textual descriptions of |
|
225 | 225 | # DAGs that have been known to be problematic, and, optionally, known pairs |
|
226 | 226 | # of revisions and their expected ancestor list. |
|
227 | 227 | dagtests = [ |
|
228 | 228 | (b'+2*2*2/*3/2', {}), |
|
229 | 229 | (b'+3*3/*2*2/*4*4/*4/2*4/2*2', {}), |
|
230 | 230 | (b'+2*2*/2*4*/4*/3*2/4', {(6, 7): [3, 5]}), |
|
231 | 231 | ] |
|
232 | 232 | def test_gca(): |
|
233 | 233 | u = uimod.ui.load() |
|
234 | 234 | for i, (dag, tests) in enumerate(dagtests): |
|
235 | 235 | repo = hg.repository(u, b'gca%d' % i, create=1) |
|
236 | 236 | cl = repo.changelog |
|
237 | 237 | if not util.safehasattr(cl.index, 'ancestors'): |
|
238 | 238 | # C version not available |
|
239 | 239 | return |
|
240 | 240 | |
|
241 | 241 | debugcommands.debugbuilddag(u, repo, dag) |
|
242 | 242 | # Compare the results of the Python and C versions. This does not |
|
243 | 243 | # include choosing a winner when more than one gca exists -- we make |
|
244 | 244 | # sure both return exactly the same set of gcas. |
|
245 | 245 | # Also compare against expected results, if available. |
|
246 | 246 | for a in cl: |
|
247 | 247 | for b in cl: |
|
248 | 248 | cgcas = sorted(cl.index.ancestors(a, b)) |
|
249 | 249 | pygcas = sorted(ancestor.ancestors(cl.parentrevs, a, b)) |
|
250 | 250 | expected = None |
|
251 | 251 | if (a, b) in tests: |
|
252 | 252 | expected = tests[(a, b)] |
|
253 | 253 | if cgcas != pygcas or (expected and cgcas != expected): |
|
254 | 254 | print("test_gca: for dag %s, gcas for %d, %d:" |
|
255 | 255 | % (dag, a, b)) |
|
256 | 256 | print(" C returned: %s" % cgcas) |
|
257 | 257 | print(" Python returned: %s" % pygcas) |
|
258 | 258 | if expected: |
|
259 | 259 | print(" expected: %s" % expected) |
|
260 | 260 | |
|
261 | 261 | def main(): |
|
262 | 262 | seed = None |
|
263 | 263 | opts, args = getopt.getopt(sys.argv[1:], 's:', ['seed=']) |
|
264 | 264 | for o, a in opts: |
|
265 | 265 | if o in ('-s', '--seed'): |
|
266 | 266 | seed = long(a, base=0) # accepts base 10 or 16 strings |
|
267 | 267 | |
|
268 | 268 | if seed is None: |
|
269 | 269 | try: |
|
270 | 270 | seed = long(binascii.hexlify(os.urandom(16)), 16) |
|
271 | 271 | except AttributeError: |
|
272 | 272 | seed = long(time.time() * 1000) |
|
273 | 273 | |
|
274 | 274 | rng = random.Random(seed) |
|
275 | 275 | test_missingancestors(seed, rng) |
|
276 | 276 | test_lazyancestors() |
|
277 | 277 | test_gca() |
|
278 | 278 | |
|
279 | 279 | if __name__ == '__main__': |
|
280 | 280 | main() |
General Comments 0
You need to be logged in to leave comments.
Login now