ancestor: optimize _lazyancestorsiter() for contiguous chains

If there's no revision between p1 and current, p1 must be the next revision
to visit. In this case, we can get around the overhead of heappop/push
operations. Note that this is faster than using heapreplace().

'current - p1 == 1' could be generalized as 'all(r not in seen for r in
xrange(p1, current)', but Python is too slow to do such thing.

mercurial/ancestor.py

@@ -1,374 +1,382 b''
 # ancestor.py - generic DAG ancestor algorithm for mercurial
 #
 # Copyright 2006 Matt Mackall <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 from __future__ import absolute_import
 
 import heapq
 
 from .node import nullrev
 from . import (
     pycompat,
 )
 
 def commonancestorsheads(pfunc, *nodes):
     """Returns a set with the heads of all common ancestors of all nodes,
     heads(::nodes[0] and ::nodes[1] and ...) .
 
     pfunc must return a list of parent vertices for a given vertex.
     """
     if not isinstance(nodes, set):
         nodes = set(nodes)
     if nullrev in nodes:
         return set()
     if len(nodes) <= 1:
         return nodes
 
     allseen = (1 << len(nodes)) - 1
     seen = [0] * (max(nodes) + 1)
     for i, n in enumerate(nodes):
         seen[n] = 1 << i
     poison = 1 << (i + 1)
 
     gca = set()
     interesting = len(nodes)
     nv = len(seen) - 1
     while nv >= 0 and interesting:
         v = nv
         nv -= 1
         if not seen[v]:
             continue
         sv = seen[v]
         if sv < poison:
             interesting -= 1
             if sv == allseen:
                 gca.add(v)
                 sv |= poison
                 if v in nodes:
                     # history is linear
                     return {v}
         if sv < poison:
             for p in pfunc(v):
                 sp = seen[p]
                 if p == nullrev:
                     continue
                 if sp == 0:
                     seen[p] = sv
                     interesting += 1
                 elif sp != sv:
                     seen[p] |= sv
         else:
             for p in pfunc(v):
                 if p == nullrev:
                     continue
                 sp = seen[p]
                 if sp and sp < poison:
                     interesting -= 1
                 seen[p] = sv
     return gca
 
 def ancestors(pfunc, *orignodes):
     """
     Returns the common ancestors of a and b that are furthest from a
     root (as measured by longest path).
 
     pfunc must return a list of parent vertices for a given vertex.
     """
     def deepest(nodes):
         interesting = {}
         count = max(nodes) + 1
         depth = [0] * count
         seen = [0] * count
         mapping = []
         for (i, n) in enumerate(sorted(nodes)):
             depth[n] = 1
             b = 1 << i
             seen[n] = b
             interesting[b] = 1
             mapping.append((b, n))
         nv = count - 1
         while nv >= 0 and len(interesting) > 1:
             v = nv
             nv -= 1
             dv = depth[v]
             if dv == 0:
                 continue
             sv = seen[v]
             for p in pfunc(v):
                 if p == nullrev:
                     continue
                 dp = depth[p]
                 nsp = sp = seen[p]
                 if dp <= dv:
                     depth[p] = dv + 1
                     if sp != sv:
                         interesting[sv] += 1
                         nsp = seen[p] = sv
                         if sp:
                             interesting[sp] -= 1
                             if interesting[sp] == 0:
                                 del interesting[sp]
                 elif dv == dp - 1:
                     nsp = sp | sv
                     if nsp == sp:
                         continue
                     seen[p] = nsp
                     interesting.setdefault(nsp, 0)
                     interesting[nsp] += 1
                     interesting[sp] -= 1
                     if interesting[sp] == 0:
                         del interesting[sp]
             interesting[sv] -= 1
             if interesting[sv] == 0:
                 del interesting[sv]
 
         if len(interesting) != 1:
             return []
 
         k = 0
         for i in interesting:
             k |= i
         return set(n for (i, n) in mapping if k & i)
 
     gca = commonancestorsheads(pfunc, *orignodes)
 
     if len(gca) <= 1:
         return gca
     return deepest(gca)
 
 class incrementalmissingancestors(object):
     '''persistent state used to calculate missing ancestors incrementally
 
     Although similar in spirit to lazyancestors below, this is a separate class
     because trying to support contains and missingancestors operations with the
     same internal data structures adds needless complexity.'''
     def __init__(self, pfunc, bases):
         self.bases = set(bases)
         if not self.bases:
             self.bases.add(nullrev)
         self.pfunc = pfunc
 
     def hasbases(self):
         '''whether the common set has any non-trivial bases'''
         return self.bases and self.bases != {nullrev}
 
     def addbases(self, newbases):
         '''grow the ancestor set by adding new bases'''
         self.bases.update(newbases)
 
     def removeancestorsfrom(self, revs):
         '''remove all ancestors of bases from the set revs (in place)'''
         bases = self.bases
         pfunc = self.pfunc
         revs.difference_update(bases)
         # nullrev is always an ancestor
         revs.discard(nullrev)
         if not revs:
             return
         # anything in revs > start is definitely not an ancestor of bases
         # revs <= start needs to be investigated
         start = max(bases)
         keepcount = sum(1 for r in revs if r > start)
         if len(revs) == keepcount:
             # no revs to consider
             return
 
         for curr in pycompat.xrange(start, min(revs) - 1, -1):
             if curr not in bases:
                 continue
             revs.discard(curr)
             bases.update(pfunc(curr))
             if len(revs) == keepcount:
                 # no more potential revs to discard
                 break
 
     def missingancestors(self, revs):
         '''return all the ancestors of revs that are not ancestors of self.bases
 
         This may include elements from revs.
 
         Equivalent to the revset (::revs - ::self.bases). Revs are returned in
         revision number order, which is a topological order.'''
         revsvisit = set(revs)
         basesvisit = self.bases
         pfunc = self.pfunc
         bothvisit = revsvisit.intersection(basesvisit)
         revsvisit.difference_update(bothvisit)
         if not revsvisit:
             return []
 
         start = max(max(revsvisit), max(basesvisit))
         # At this point, we hold the invariants that:
         # - revsvisit is the set of nodes we know are an ancestor of at least
         #   one of the nodes in revs
         # - basesvisit is the same for bases
         # - bothvisit is the set of nodes we know are ancestors of at least one
         #   of the nodes in revs and one of the nodes in bases. bothvisit and
         #   revsvisit are mutually exclusive, but bothvisit is a subset of
         #   basesvisit.
         # Now we walk down in reverse topo order, adding parents of nodes
         # already visited to the sets while maintaining the invariants. When a
         # node is found in both revsvisit and basesvisit, it is removed from
         # revsvisit and added to bothvisit. When revsvisit becomes empty, there
         # are no more ancestors of revs that aren't also ancestors of bases, so
         # exit.
 
         missing = []
         for curr in pycompat.xrange(start, nullrev, -1):
             if not revsvisit:
                 break
 
             if curr in bothvisit:
                 bothvisit.remove(curr)
                 # curr's parents might have made it into revsvisit through
                 # another path
                 for p in pfunc(curr):
                     revsvisit.discard(p)
                     basesvisit.add(p)
                     bothvisit.add(p)
                 continue
 
             if curr in revsvisit:
                 missing.append(curr)
                 revsvisit.remove(curr)
                 thisvisit = revsvisit
                 othervisit = basesvisit
             elif curr in basesvisit:
                 thisvisit = basesvisit
                 othervisit = revsvisit
             else:
                 # not an ancestor of revs or bases: ignore
                 continue
 
             for p in pfunc(curr):
                 if p == nullrev:
                     pass
                 elif p in othervisit or p in bothvisit:
                     # p is implicitly in thisvisit. This means p is or should be
                     # in bothvisit
                     revsvisit.discard(p)
                     basesvisit.add(p)
                     bothvisit.add(p)
                 else:
                     # visit later
                     thisvisit.add(p)
 
         missing.reverse()
         return missing
 
 # Extracted from lazyancestors.__iter__ to avoid a reference cycle
 def _lazyancestorsiter(parentrevs, initrevs, stoprev, inclusive):
     seen = {nullrev}
     schedule = heapq.heappush
     nextitem = heapq.heappop
     see = seen.add
 
     if inclusive:
         visit = [-r for r in initrevs]
         seen.update(initrevs)
         heapq.heapify(visit)
     else:
         visit = []
         heapq.heapify(visit)
         for r in initrevs:
             p1, p2 = parentrevs(r)
             if p1 not in seen:
                 schedule(visit, -p1)
                 see(p1)
             if p2 not in seen:
                 schedule(visit, -p2)
                 see(p2)
 
     while visit:
-        current = -nextitem(visit)
+        current = -visit[0]
         if current < stoprev:
             break
         yield current
+        # optimize out heapq operation if p1 is known to be the next highest
+        # revision, which is quite common in linear history.
         p1, p2 = parentrevs(current)
         if p1 not in seen:
-            schedule(visit, -p1)
+            if current - p1 == 1:
+                visit[0] = -p1
+            else:
+                nextitem(visit)
+                schedule(visit, -p1)
             see(p1)
+        else:
+            nextitem(visit)
         if p2 not in seen:
             schedule(visit, -p2)
             see(p2)
 
 class lazyancestors(object):
     def __init__(self, pfunc, revs, stoprev=0, inclusive=False):
         """Create a new object generating ancestors for the given revs. Does
         not generate revs lower than stoprev.
 
         This is computed lazily starting from revs. The object supports
         iteration and membership.
 
         cl should be a changelog and revs should be an iterable. inclusive is
         a boolean that indicates whether revs should be included. Revs lower
         than stoprev will not be generated.
 
         Result does not include the null revision."""
         self._parentrevs = pfunc
         self._initrevs = revs = [r for r in revs if r >= stoprev]
         self._stoprev = stoprev
         self._inclusive = inclusive
 
         self._containsseen = set()
         self._containsiter = _lazyancestorsiter(self._parentrevs,
                                                 self._initrevs,
                                                 self._stoprev,
                                                 self._inclusive)
 
     def __nonzero__(self):
         """False if the set is empty, True otherwise."""
         try:
             next(iter(self))
             return True
         except StopIteration:
             return False
 
     __bool__ = __nonzero__
 
     def __iter__(self):
         """Generate the ancestors of _initrevs in reverse topological order.
 
         If inclusive is False, yield a sequence of revision numbers starting
         with the parents of each revision in revs, i.e., each revision is
         *not* considered an ancestor of itself. Results are emitted in reverse
         revision number order. That order is also topological: a child is
         always emitted before its parent.
 
         If inclusive is True, the source revisions are also yielded. The
         reverse revision number order is still enforced."""
         for rev in _lazyancestorsiter(self._parentrevs, self._initrevs,
                                       self._stoprev, self._inclusive):
             yield rev
 
     def __contains__(self, target):
         """Test whether target is an ancestor of self._initrevs."""
         seen = self._containsseen
         if target in seen:
             return True
         iter = self._containsiter
         if iter is None:
             # Iterator exhausted
             return False
         # Only integer target is valid, but some callers expect 'None in self'
         # to be False. So we explicitly allow it.
         if target is None:
             return False
 
         see = seen.add
         try:
             while True:
                 rev = next(iter)
                 see(rev)
                 if rev == target:
                     return True
                 if rev < target:
                     return False
         except StopIteration:
             # Set to None to indicate fast-path can be used next time, and to
             # free up memory.
             self._containsiter = None
             return False

tests/test-ancestor.py

@@ -1,280 +1,284 b''
 from __future__ import absolute_import, print_function
 
 import binascii
 import getopt
 import math
 import os
 import random
 import sys
 import time
 
 from mercurial.node import nullrev
 from mercurial import (
     ancestor,
     debugcommands,
     hg,
     pycompat,
     ui as uimod,
     util,
 )
 
 if pycompat.ispy3:
     long = int
     xrange = range
 
 def buildgraph(rng, nodes=100, rootprob=0.05, mergeprob=0.2, prevprob=0.7):
     '''nodes: total number of nodes in the graph
     rootprob: probability that a new node (not 0) will be a root
     mergeprob: probability that, excluding a root a node will be a merge
     prevprob: probability that p1 will be the previous node
 
     return value is a graph represented as an adjacency list.
     '''
     graph = [None] * nodes
     for i in xrange(nodes):
         if i == 0 or rng.random() < rootprob:
             graph[i] = [nullrev]
         elif i == 1:
             graph[i] = [0]
         elif rng.random() < mergeprob:
             if i == 2 or rng.random() < prevprob:
                 # p1 is prev
                 p1 = i - 1
             else:
                 p1 = rng.randrange(i - 1)
             p2 = rng.choice(list(range(0, p1)) + list(range(p1 + 1, i)))
             graph[i] = [p1, p2]
         elif rng.random() < prevprob:
             graph[i] = [i - 1]
         else:
             graph[i] = [rng.randrange(i - 1)]
 
     return graph
 
 def buildancestorsets(graph):
     ancs = [None] * len(graph)
     for i in xrange(len(graph)):
         ancs[i] = {i}
         if graph[i] == [nullrev]:
             continue
         for p in graph[i]:
             ancs[i].update(ancs[p])
     return ancs
 
 class naiveincrementalmissingancestors(object):
     def __init__(self, ancs, bases):
         self.ancs = ancs
         self.bases = set(bases)
     def addbases(self, newbases):
         self.bases.update(newbases)
     def removeancestorsfrom(self, revs):
         for base in self.bases:
             if base != nullrev:
                 revs.difference_update(self.ancs[base])
         revs.discard(nullrev)
     def missingancestors(self, revs):
         res = set()
         for rev in revs:
             if rev != nullrev:
                 res.update(self.ancs[rev])
         for base in self.bases:
             if base != nullrev:
                 res.difference_update(self.ancs[base])
         return sorted(res)
 
 def test_missingancestors(seed, rng):
     # empirically observed to take around 1 second
     graphcount = 100
     testcount = 10
     inccount = 10
     nerrs = [0]
     # the default mu and sigma give us a nice distribution of mostly
     # single-digit counts (including 0) with some higher ones
     def lognormrandom(mu, sigma):
         return int(math.floor(rng.lognormvariate(mu, sigma)))
 
     def samplerevs(nodes, mu=1.1, sigma=0.8):
         count = min(lognormrandom(mu, sigma), len(nodes))
         return rng.sample(nodes, count)
 
     def err(seed, graph, bases, seq, output, expected):
         if nerrs[0] == 0:
             print('seed:', hex(seed)[:-1], file=sys.stderr)
         if gerrs[0] == 0:
             print('graph:', graph, file=sys.stderr)
         print('* bases:', bases, file=sys.stderr)
         print('* seq: ', seq, file=sys.stderr)
         print('*  output:  ', output, file=sys.stderr)
         print('*  expected:', expected, file=sys.stderr)
         nerrs[0] += 1
         gerrs[0] += 1
 
     for g in xrange(graphcount):
         graph = buildgraph(rng)
         ancs = buildancestorsets(graph)
         gerrs = [0]
         for _ in xrange(testcount):
             # start from nullrev to include it as a possibility
             graphnodes = range(nullrev, len(graph))
             bases = samplerevs(graphnodes)
 
             # fast algorithm
             inc = ancestor.incrementalmissingancestors(graph.__getitem__, bases)
             # reference slow algorithm
             naiveinc = naiveincrementalmissingancestors(ancs, bases)
             seq = []
             revs = []
             for _ in xrange(inccount):
                 if rng.random() < 0.2:
                     newbases = samplerevs(graphnodes)
                     seq.append(('addbases', newbases))
                     inc.addbases(newbases)
                     naiveinc.addbases(newbases)
                 if rng.random() < 0.4:
                     # larger set so that there are more revs to remove from
                     revs = samplerevs(graphnodes, mu=1.5)
                     seq.append(('removeancestorsfrom', revs))
                     hrevs = set(revs)
                     rrevs = set(revs)
                     inc.removeancestorsfrom(hrevs)
                     naiveinc.removeancestorsfrom(rrevs)
                     if hrevs != rrevs:
                         err(seed, graph, bases, seq, sorted(hrevs),
                             sorted(rrevs))
                 else:
                     revs = samplerevs(graphnodes)
                     seq.append(('missingancestors', revs))
                     h = inc.missingancestors(revs)
                     r = naiveinc.missingancestors(revs)
                     if h != r:
                         err(seed, graph, bases, seq, h, r)
 
 # graph is a dict of child->parent adjacency lists for this graph:
 # o  13
 # |
 # | o  12
 # | |
 # | | o    11
 # | | |\
 # | | | | o  10
 # | | | | |
 # | o---+ |  9
 # | | | | |
 # o | | | |  8
 #  / / / /
 # | | o |  7
 # | | | |
 # o---+ |  6
 #  / / /
 # | | o  5
 # | |/
 # | o  4
 # | |
 # o |  3
 # | |
 # | o  2
 # |/
 # o  1
 # |
 # o  0
 
 graph = {0: [-1, -1], 1: [0, -1], 2: [1, -1], 3: [1, -1], 4: [2, -1],
          5: [4, -1], 6: [4, -1], 7: [4, -1], 8: [-1, -1], 9: [6, 7],
          10: [5, -1], 11: [3, 7], 12: [9, -1], 13: [8, -1]}
 
 def genlazyancestors(revs, stoprev=0, inclusive=False):
     print(("%% lazy ancestor set for %s, stoprev = %s, inclusive = %s" %
            (revs, stoprev, inclusive)))
     return ancestor.lazyancestors(graph.get, revs, stoprev=stoprev,
                                   inclusive=inclusive)
 
 def printlazyancestors(s, l):
     print('membership: %r' % [n for n in l if n in s])
     print('iteration:  %r' % list(s))
 
 def test_lazyancestors():
     # Empty revs
     s = genlazyancestors([])
     printlazyancestors(s, [3, 0, -1])
 
     # Standard example
     s = genlazyancestors([11, 13])
     printlazyancestors(s, [11, 13, 7, 9, 8, 3, 6, 4, 1, -1, 0])
 
     # Standard with ancestry in the initial set (1 is ancestor of 3)
     s = genlazyancestors([1, 3])
     printlazyancestors(s, [1, -1, 0])
 
     # Including revs
     s = genlazyancestors([11, 13], inclusive=True)
     printlazyancestors(s, [11, 13, 7, 9, 8, 3, 6, 4, 1, -1, 0])
 
     # Test with stoprev
     s = genlazyancestors([11, 13], stoprev=6)
     printlazyancestors(s, [11, 13, 7, 9, 8, 3, 6, 4, 1, -1, 0])
     s = genlazyancestors([11, 13], stoprev=6, inclusive=True)
     printlazyancestors(s, [11, 13, 7, 9, 8, 3, 6, 4, 1, -1, 0])
 
     # Test with stoprev >= min(initrevs)
     s = genlazyancestors([11, 13], stoprev=11, inclusive=True)
     printlazyancestors(s, [11, 13, 7, 9, 8, 3, 6, 4, 1, -1, 0])
     s = genlazyancestors([11, 13], stoprev=12, inclusive=True)
     printlazyancestors(s, [11, 13, 7, 9, 8, 3, 6, 4, 1, -1, 0])
 
+    # Contiguous chains: 5->4, 2->1 (where 1 is in seen set), 1->0
+    s = genlazyancestors([10, 1], inclusive=True)
+    printlazyancestors(s, [2, 10, 4, 5, -1, 0, 1])
+
 # The C gca algorithm requires a real repo. These are textual descriptions of
 # DAGs that have been known to be problematic, and, optionally, known pairs
 # of revisions and their expected ancestor list.
 dagtests = [
     (b'+2*2*2/*3/2', {}),
     (b'+3*3/*2*2/*4*4/*4/2*4/2*2', {}),
     (b'+2*2*/2*4*/4*/3*2/4', {(6, 7): [3, 5]}),
 ]
 def test_gca():
     u = uimod.ui.load()
     for i, (dag, tests) in enumerate(dagtests):
         repo = hg.repository(u, b'gca%d' % i, create=1)
         cl = repo.changelog
         if not util.safehasattr(cl.index, 'ancestors'):
             # C version not available
             return
 
         debugcommands.debugbuilddag(u, repo, dag)
         # Compare the results of the Python and C versions. This does not
         # include choosing a winner when more than one gca exists -- we make
         # sure both return exactly the same set of gcas.
         # Also compare against expected results, if available.
         for a in cl:
             for b in cl:
                 cgcas = sorted(cl.index.ancestors(a, b))
                 pygcas = sorted(ancestor.ancestors(cl.parentrevs, a, b))
                 expected = None
                 if (a, b) in tests:
                     expected = tests[(a, b)]
                 if cgcas != pygcas or (expected and cgcas != expected):
                     print("test_gca: for dag %s, gcas for %d, %d:"
                           % (dag, a, b))
                     print("  C returned:      %s" % cgcas)
                     print("  Python returned: %s" % pygcas)
                     if expected:
                         print("  expected:        %s" % expected)
 
 def main():
     seed = None
     opts, args = getopt.getopt(sys.argv[1:], 's:', ['seed='])
     for o, a in opts:
         if o in ('-s', '--seed'):
             seed = long(a, base=0) # accepts base 10 or 16 strings
 
     if seed is None:
         try:
             seed = long(binascii.hexlify(os.urandom(16)), 16)
         except AttributeError:
             seed = long(time.time() * 1000)
 
     rng = random.Random(seed)
     test_missingancestors(seed, rng)
     test_lazyancestors()
     test_gca()
 
 if __name__ == '__main__':
     main()

tests/test-ancestor.py.out

@@ -1,24 +1,27 b''
 % lazy ancestor set for [], stoprev = 0, inclusive = False
 membership: []
 iteration:  []
 % lazy ancestor set for [11, 13], stoprev = 0, inclusive = False
 membership: [7, 8, 3, 4, 1, 0]
 iteration:  [8, 7, 4, 3, 2, 1, 0]
 % lazy ancestor set for [1, 3], stoprev = 0, inclusive = False
 membership: [1, 0]
 iteration:  [1, 0]
 % lazy ancestor set for [11, 13], stoprev = 0, inclusive = True
 membership: [11, 13, 7, 8, 3, 4, 1, 0]
 iteration:  [13, 11, 8, 7, 4, 3, 2, 1, 0]
 % lazy ancestor set for [11, 13], stoprev = 6, inclusive = False
 membership: [7, 8]
 iteration:  [8, 7]
 % lazy ancestor set for [11, 13], stoprev = 6, inclusive = True
 membership: [11, 13, 7, 8]
 iteration:  [13, 11, 8, 7]
 % lazy ancestor set for [11, 13], stoprev = 11, inclusive = True
 membership: [11, 13]
 iteration:  [13, 11]
 % lazy ancestor set for [11, 13], stoprev = 12, inclusive = True
 membership: [13]
 iteration:  [13]
+% lazy ancestor set for [10, 1], stoprev = 0, inclusive = True
+membership: [2, 10, 4, 5, 0, 1]
+iteration:  [10, 5, 4, 2, 1, 0]