revset: optimize missing ancestor expressions

A missing ancestor expression is any expression of the form (::x - ::y) or
equivalent. Such expressions are remarkably common, and so far have involved
multiple walks down the DAG, followed by a set difference operation.

With this patch, such expressions will be transformed into uses of the fast
algorithm at ancestor.missingancestor.

For a repository with over 600,000 revisions, perfrevset for '::tip - ::-10000'
returns:

Before: ! wall 3.999575 comb 4.000000 user 3.910000 sys 0.090000 (best of 3)
After:  ! wall 0.132423 comb 0.130000 user 0.130000 sys 0.000000 (best of 75)

mercurial/revset.py

@@ -1,2197 +1,2214 b''
 # revset.py - revision set queries for mercurial
 #
 # Copyright 2010 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.
 
 import re
 import parser, util, error, discovery, hbisect, phases
 import node
 import match as matchmod
 from i18n import _
 import encoding
 import obsolete as obsmod
 import pathutil
 import repoview
 
 def _revancestors(repo, revs, followfirst):
     """Like revlog.ancestors(), but supports followfirst."""
     cut = followfirst and 1 or None
     cl = repo.changelog
     visit = util.deque(revs)
     seen = set([node.nullrev])
     while visit:
         for parent in cl.parentrevs(visit.popleft())[:cut]:
             if parent not in seen:
                 visit.append(parent)
                 seen.add(parent)
                 yield parent
 
 def _revdescendants(repo, revs, followfirst):
     """Like revlog.descendants() but supports followfirst."""
     cut = followfirst and 1 or None
     cl = repo.changelog
     first = min(revs)
     nullrev = node.nullrev
     if first == nullrev:
         # Are there nodes with a null first parent and a non-null
         # second one? Maybe. Do we care? Probably not.
         for i in cl:
             yield i
         return
 
     seen = set(revs)
     for i in cl.revs(first + 1):
         for x in cl.parentrevs(i)[:cut]:
             if x != nullrev and x in seen:
                 seen.add(i)
                 yield i
                 break
 
 def _revsbetween(repo, roots, heads):
     """Return all paths between roots and heads, inclusive of both endpoint
     sets."""
     if not roots:
         return baseset([])
     parentrevs = repo.changelog.parentrevs
     visit = baseset(heads)
     reachable = set()
     seen = {}
     minroot = min(roots)
     roots = set(roots)
     # open-code the post-order traversal due to the tiny size of
     # sys.getrecursionlimit()
     while visit:
         rev = visit.pop()
         if rev in roots:
             reachable.add(rev)
         parents = parentrevs(rev)
         seen[rev] = parents
         for parent in parents:
             if parent >= minroot and parent not in seen:
                 visit.append(parent)
     if not reachable:
         return baseset([])
     for rev in sorted(seen):
         for parent in seen[rev]:
             if parent in reachable:
                 reachable.add(rev)
     return baseset(sorted(reachable))
 
 elements = {
     "(": (20, ("group", 1, ")"), ("func", 1, ")")),
     "~": (18, None, ("ancestor", 18)),
     "^": (18, None, ("parent", 18), ("parentpost", 18)),
     "-": (5, ("negate", 19), ("minus", 5)),
     "::": (17, ("dagrangepre", 17), ("dagrange", 17),
            ("dagrangepost", 17)),
     "..": (17, ("dagrangepre", 17), ("dagrange", 17),
            ("dagrangepost", 17)),
     ":": (15, ("rangepre", 15), ("range", 15), ("rangepost", 15)),
     "not": (10, ("not", 10)),
     "!": (10, ("not", 10)),
     "and": (5, None, ("and", 5)),
     "&": (5, None, ("and", 5)),
     "or": (4, None, ("or", 4)),
     "|": (4, None, ("or", 4)),
     "+": (4, None, ("or", 4)),
     ",": (2, None, ("list", 2)),
     ")": (0, None, None),
     "symbol": (0, ("symbol",), None),
     "string": (0, ("string",), None),
     "end": (0, None, None),
 }
 
 keywords = set(['and', 'or', 'not'])
 
 def tokenize(program):
     '''
     Parse a revset statement into a stream of tokens
 
     Check that @ is a valid unquoted token character (issue3686):
     >>> list(tokenize("@::"))
     [('symbol', '@', 0), ('::', None, 1), ('end', None, 3)]
 
     '''
 
     pos, l = 0, len(program)
     while pos < l:
         c = program[pos]
         if c.isspace(): # skip inter-token whitespace
             pass
         elif c == ':' and program[pos:pos + 2] == '::': # look ahead carefully
             yield ('::', None, pos)
             pos += 1 # skip ahead
         elif c == '.' and program[pos:pos + 2] == '..': # look ahead carefully
             yield ('..', None, pos)
             pos += 1 # skip ahead
         elif c in "():,-|&+!~^": # handle simple operators
             yield (c, None, pos)
         elif (c in '"\'' or c == 'r' and
               program[pos:pos + 2] in ("r'", 'r"')): # handle quoted strings
             if c == 'r':
                 pos += 1
                 c = program[pos]
                 decode = lambda x: x
             else:
                 decode = lambda x: x.decode('string-escape')
             pos += 1
             s = pos
             while pos < l: # find closing quote
                 d = program[pos]
                 if d == '\\': # skip over escaped characters
                     pos += 2
                     continue
                 if d == c:
                     yield ('string', decode(program[s:pos]), s)
                     break
                 pos += 1
             else:
                 raise error.ParseError(_("unterminated string"), s)
         # gather up a symbol/keyword
         elif c.isalnum() or c in '._@' or ord(c) > 127:
             s = pos
             pos += 1
             while pos < l: # find end of symbol
                 d = program[pos]
                 if not (d.isalnum() or d in "._/@" or ord(d) > 127):
                     break
                 if d == '.' and program[pos - 1] == '.': # special case for ..
                     pos -= 1
                     break
                 pos += 1
             sym = program[s:pos]
             if sym in keywords: # operator keywords
                 yield (sym, None, s)
             else:
                 yield ('symbol', sym, s)
             pos -= 1
         else:
             raise error.ParseError(_("syntax error"), pos)
         pos += 1
     yield ('end', None, pos)
 
 # helpers
 
 def getstring(x, err):
     if x and (x[0] == 'string' or x[0] == 'symbol'):
         return x[1]
     raise error.ParseError(err)
 
 def getlist(x):
     if not x:
         return []
     if x[0] == 'list':
         return getlist(x[1]) + [x[2]]
     return [x]
 
 def getargs(x, min, max, err):
     l = getlist(x)
     if len(l) < min or (max >= 0 and len(l) > max):
         raise error.ParseError(err)
     return l
 
 def getset(repo, subset, x):
     if not x:
         raise error.ParseError(_("missing argument"))
     return methods[x[0]](repo, subset, *x[1:])
 
 def _getrevsource(repo, r):
     extra = repo[r].extra()
     for label in ('source', 'transplant_source', 'rebase_source'):
         if label in extra:
             try:
                 return repo[extra[label]].rev()
             except error.RepoLookupError:
                 pass
     return None
 
 # operator methods
 
 def stringset(repo, subset, x):
     x = repo[x].rev()
     if x == -1 and len(subset) == len(repo):
         return baseset([-1])
     if len(subset) == len(repo) or x in subset:
         return baseset([x])
     return baseset([])
 
 def symbolset(repo, subset, x):
     if x in symbols:
         raise error.ParseError(_("can't use %s here") % x)
     return stringset(repo, subset, x)
 
 def rangeset(repo, subset, x, y):
     cl = baseset(repo.changelog)
     m = getset(repo, cl, x)
     n = getset(repo, cl, y)
 
     if not m or not n:
         return baseset([])
     m, n = m[0], n[-1]
 
     if m < n:
         r = range(m, n + 1)
     else:
         r = range(m, n - 1, -1)
     s = subset.set()
     return baseset([x for x in r if x in s])
 
 def dagrange(repo, subset, x, y):
     r = baseset(repo)
     xs = _revsbetween(repo, getset(repo, r, x), getset(repo, r, y))
     s = subset.set()
     return baseset([r for r in xs if r in s])
 
 def andset(repo, subset, x, y):
     return getset(repo, getset(repo, subset, x), y)
 
 def orset(repo, subset, x, y):
     xl = getset(repo, subset, x)
     yl = getset(repo, subset - xl, y)
     return xl + yl
 
 def notset(repo, subset, x):
     return subset - getset(repo, subset, x)
 
 def listset(repo, subset, a, b):
     raise error.ParseError(_("can't use a list in this context"))
 
 def func(repo, subset, a, b):
     if a[0] == 'symbol' and a[1] in symbols:
         return symbols[a[1]](repo, subset, b)
     raise error.ParseError(_("not a function: %s") % a[1])
 
 # functions
 
 def adds(repo, subset, x):
     """``adds(pattern)``
     Changesets that add a file matching pattern.
 
     The pattern without explicit kind like ``glob:`` is expected to be
     relative to the current directory and match against a file or a
     directory.
     """
     # i18n: "adds" is a keyword
     pat = getstring(x, _("adds requires a pattern"))
     return checkstatus(repo, subset, pat, 1)
 
 def ancestor(repo, subset, x):
     """``ancestor(*changeset)``
     Greatest common ancestor of the changesets.
 
     Accepts 0 or more changesets.
     Will return empty list when passed no args.
     Greatest common ancestor of a single changeset is that changeset.
     """
     # i18n: "ancestor" is a keyword
     l = getlist(x)
     rl = baseset(repo)
     anc = None
 
     # (getset(repo, rl, i) for i in l) generates a list of lists
     rev = repo.changelog.rev
     ancestor = repo.changelog.ancestor
     node = repo.changelog.node
     for revs in (getset(repo, rl, i) for i in l):
         for r in revs:
             if anc is None:
                 anc = r
             else:
                 anc = rev(ancestor(node(anc), node(r)))
 
     if anc is not None and anc in subset:
         return baseset([anc])
     return baseset([])
 
 def _ancestors(repo, subset, x, followfirst=False):
     args = getset(repo, baseset(repo), x)
     if not args:
         return baseset([])
     s = set(_revancestors(repo, args, followfirst)) | set(args)
     return baseset([r for r in subset if r in s])
 
 def ancestors(repo, subset, x):
     """``ancestors(set)``
     Changesets that are ancestors of a changeset in set.
     """
     return _ancestors(repo, subset, x)
 
 def _firstancestors(repo, subset, x):
     # ``_firstancestors(set)``
     # Like ``ancestors(set)`` but follows only the first parents.
     return _ancestors(repo, subset, x, followfirst=True)
 
 def ancestorspec(repo, subset, x, n):
     """``set~n``
     Changesets that are the Nth ancestor (first parents only) of a changeset
     in set.
     """
     try:
         n = int(n[1])
     except (TypeError, ValueError):
         raise error.ParseError(_("~ expects a number"))
     ps = set()
     cl = repo.changelog
     for r in getset(repo, baseset(cl), x):
         for i in range(n):
             r = cl.parentrevs(r)[0]
         ps.add(r)
     return baseset([r for r in subset if r in ps])
 
 def author(repo, subset, x):
     """``author(string)``
     Alias for ``user(string)``.
     """
     # i18n: "author" is a keyword
     n = encoding.lower(getstring(x, _("author requires a string")))
     kind, pattern, matcher = _substringmatcher(n)
     return lazyset(subset, lambda x: matcher(encoding.lower(repo[x].user())))
 
 def bisect(repo, subset, x):
     """``bisect(string)``
     Changesets marked in the specified bisect status:
 
     - ``good``, ``bad``, ``skip``: csets explicitly marked as good/bad/skip
     - ``goods``, ``bads``      : csets topologically good/bad
     - ``range``              : csets taking part in the bisection
     - ``pruned``             : csets that are goods, bads or skipped
     - ``untested``           : csets whose fate is yet unknown
     - ``ignored``            : csets ignored due to DAG topology
     - ``current``            : the cset currently being bisected
     """
     # i18n: "bisect" is a keyword
     status = getstring(x, _("bisect requires a string")).lower()
     state = set(hbisect.get(repo, status))
     return baseset([r for r in subset if r in state])
 
 # Backward-compatibility
 # - no help entry so that we do not advertise it any more
 def bisected(repo, subset, x):
     return bisect(repo, subset, x)
 
 def bookmark(repo, subset, x):
     """``bookmark([name])``
     The named bookmark or all bookmarks.
 
     If `name` starts with `re:`, the remainder of the name is treated as
     a regular expression. To match a bookmark that actually starts with `re:`,
     use the prefix `literal:`.
     """
     # i18n: "bookmark" is a keyword
     args = getargs(x, 0, 1, _('bookmark takes one or no arguments'))
     if args:
         bm = getstring(args[0],
                        # i18n: "bookmark" is a keyword
                        _('the argument to bookmark must be a string'))
         kind, pattern, matcher = _stringmatcher(bm)
         if kind == 'literal':
             bmrev = repo._bookmarks.get(bm, None)
             if not bmrev:
                 raise util.Abort(_("bookmark '%s' does not exist") % bm)
             bmrev = repo[bmrev].rev()
             return lazyset(subset, lambda r: r == bmrev)
         else:
             matchrevs = set()
             for name, bmrev in repo._bookmarks.iteritems():
                 if matcher(name):
                     matchrevs.add(bmrev)
             if not matchrevs:
                 raise util.Abort(_("no bookmarks exist that match '%s'")
                                  % pattern)
             bmrevs = set()
             for bmrev in matchrevs:
                 bmrevs.add(repo[bmrev].rev())
             return subset & bmrevs
 
     bms = set([repo[r].rev()
                for r in repo._bookmarks.values()])
     return lazyset(subset, lambda r: r in bms)
 
 def branch(repo, subset, x):
     """``branch(string or set)``
     All changesets belonging to the given branch or the branches of the given
     changesets.
 
     If `string` starts with `re:`, the remainder of the name is treated as
     a regular expression. To match a branch that actually starts with `re:`,
     use the prefix `literal:`.
     """
     try:
         b = getstring(x, '')
     except error.ParseError:
         # not a string, but another revspec, e.g. tip()
         pass
     else:
         kind, pattern, matcher = _stringmatcher(b)
         if kind == 'literal':
             # note: falls through to the revspec case if no branch with
             # this name exists
             if pattern in repo.branchmap():
                 return lazyset(subset, lambda r: matcher(repo[r].branch()))
         else:
             return lazyset(subset, lambda r: matcher(repo[r].branch()))
 
     s = getset(repo, baseset(repo), x)
     b = set()
     for r in s:
         b.add(repo[r].branch())
     s = s.set()
     return lazyset(subset, lambda r: r in s or repo[r].branch() in b)
 
 def bumped(repo, subset, x):
     """``bumped()``
     Mutable changesets marked as successors of public changesets.
 
     Only non-public and non-obsolete changesets can be `bumped`.
     """
     # i18n: "bumped" is a keyword
     getargs(x, 0, 0, _("bumped takes no arguments"))
     bumped = obsmod.getrevs(repo, 'bumped')
     return subset & bumped
 
 def bundle(repo, subset, x):
     """``bundle()``
     Changesets in the bundle.
 
     Bundle must be specified by the -R option."""
 
     try:
         bundlerevs = repo.changelog.bundlerevs
     except AttributeError:
         raise util.Abort(_("no bundle provided - specify with -R"))
     return subset & bundlerevs
 
 def checkstatus(repo, subset, pat, field):
     hasset = matchmod.patkind(pat) == 'set'
 
     def matches(x):
         m = None
         fname = None
         c = repo[x]
         if not m or hasset:
             m = matchmod.match(repo.root, repo.getcwd(), [pat], ctx=c)
             if not m.anypats() and len(m.files()) == 1:
                 fname = m.files()[0]
         if fname is not None:
             if fname not in c.files():
                 return False
         else:
             for f in c.files():
                 if m(f):
                     break
             else:
                 return False
         files = repo.status(c.p1().node(), c.node())[field]
         if fname is not None:
             if fname in files:
                 return True
         else:
             for f in files:
                 if m(f):
                     return True
 
     return lazyset(subset, matches)
 
 def _children(repo, narrow, parentset):
     cs = set()
     if not parentset:
         return baseset(cs)
     pr = repo.changelog.parentrevs
     minrev = min(parentset)
     for r in narrow:
         if r <= minrev:
             continue
         for p in pr(r):
             if p in parentset:
                 cs.add(r)
     return baseset(cs)
 
 def children(repo, subset, x):
     """``children(set)``
     Child changesets of changesets in set.
     """
     s = getset(repo, baseset(repo), x).set()
     cs = _children(repo, subset, s)
     return subset & cs
 
 def closed(repo, subset, x):
     """``closed()``
     Changeset is closed.
     """
     # i18n: "closed" is a keyword
     getargs(x, 0, 0, _("closed takes no arguments"))
     return lazyset(subset, lambda r: repo[r].closesbranch())
 
 def contains(repo, subset, x):
     """``contains(pattern)``
     Revision contains a file matching pattern. See :hg:`help patterns`
     for information about file patterns.
 
     The pattern without explicit kind like ``glob:`` is expected to be
     relative to the current directory and match against a file exactly
     for efficiency.
     """
     # i18n: "contains" is a keyword
     pat = getstring(x, _("contains requires a pattern"))
 
     def matches(x):
         if not matchmod.patkind(pat):
             pats = pathutil.canonpath(repo.root, repo.getcwd(), pat)
             if pats in repo[x]:
                 return True
         else:
             c = repo[x]
             m = matchmod.match(repo.root, repo.getcwd(), [pat], ctx=c)
             for f in c.manifest():
                 if m(f):
                     return True
         return False
 
     return lazyset(subset, matches)
 
 def converted(repo, subset, x):
     """``converted([id])``
     Changesets converted from the given identifier in the old repository if
     present, or all converted changesets if no identifier is specified.
     """
 
     # There is exactly no chance of resolving the revision, so do a simple
     # string compare and hope for the best
 
     rev = None
     # i18n: "converted" is a keyword
     l = getargs(x, 0, 1, _('converted takes one or no arguments'))
     if l:
         # i18n: "converted" is a keyword
         rev = getstring(l[0], _('converted requires a revision'))
 
     def _matchvalue(r):
         source = repo[r].extra().get('convert_revision', None)
         return source is not None and (rev is None or source.startswith(rev))
 
     return lazyset(subset, lambda r: _matchvalue(r))
 
 def date(repo, subset, x):
     """``date(interval)``
     Changesets within the interval, see :hg:`help dates`.
     """
     # i18n: "date" is a keyword
     ds = getstring(x, _("date requires a string"))
     dm = util.matchdate(ds)
     return lazyset(subset, lambda x: dm(repo[x].date()[0]))
 
 def desc(repo, subset, x):
     """``desc(string)``
     Search commit message for string. The match is case-insensitive.
     """
     # i18n: "desc" is a keyword
     ds = encoding.lower(getstring(x, _("desc requires a string")))
 
     def matches(x):
         c = repo[x]
         return ds in encoding.lower(c.description())
 
     return lazyset(subset, matches)
 
 def _descendants(repo, subset, x, followfirst=False):
     args = getset(repo, baseset(repo), x)
     if not args:
         return baseset([])
     s = set(_revdescendants(repo, args, followfirst)) | set(args)
     return baseset([r for r in subset if r in s])
 
 def descendants(repo, subset, x):
     """``descendants(set)``
     Changesets which are descendants of changesets in set.
     """
     return _descendants(repo, subset, x)
 
 def _firstdescendants(repo, subset, x):
     # ``_firstdescendants(set)``
     # Like ``descendants(set)`` but follows only the first parents.
     return _descendants(repo, subset, x, followfirst=True)
 
 def destination(repo, subset, x):
     """``destination([set])``
     Changesets that were created by a graft, transplant or rebase operation,
     with the given revisions specified as the source.  Omitting the optional set
     is the same as passing all().
     """
     if x is not None:
         args = getset(repo, baseset(repo), x).set()
     else:
         args = getall(repo, baseset(repo), x).set()
 
     dests = set()
 
     # subset contains all of the possible destinations that can be returned, so
     # iterate over them and see if their source(s) were provided in the args.
     # Even if the immediate src of r is not in the args, src's source (or
     # further back) may be.  Scanning back further than the immediate src allows
     # transitive transplants and rebases to yield the same results as transitive
     # grafts.
     for r in subset:
         src = _getrevsource(repo, r)
         lineage = None
 
         while src is not None:
             if lineage is None:
                 lineage = list()
 
             lineage.append(r)
 
             # The visited lineage is a match if the current source is in the arg
             # set.  Since every candidate dest is visited by way of iterating
             # subset, any dests further back in the lineage will be tested by a
             # different iteration over subset.  Likewise, if the src was already
             # selected, the current lineage can be selected without going back
             # further.
             if src in args or src in dests:
                 dests.update(lineage)
                 break
 
             r = src
             src = _getrevsource(repo, r)
 
     return baseset([r for r in subset if r in dests])
 
 def divergent(repo, subset, x):
     """``divergent()``
     Final successors of changesets with an alternative set of final successors.
     """
     # i18n: "divergent" is a keyword
     getargs(x, 0, 0, _("divergent takes no arguments"))
     divergent = obsmod.getrevs(repo, 'divergent')
     return baseset([r for r in subset if r in divergent])
 
 def draft(repo, subset, x):
     """``draft()``
     Changeset in draft phase."""
     # i18n: "draft" is a keyword
     getargs(x, 0, 0, _("draft takes no arguments"))
     pc = repo._phasecache
     return lazyset(subset, lambda r: pc.phase(repo, r) == phases.draft)
 
 def extinct(repo, subset, x):
     """``extinct()``
     Obsolete changesets with obsolete descendants only.
     """
     # i18n: "extinct" is a keyword
     getargs(x, 0, 0, _("extinct takes no arguments"))
     extincts = obsmod.getrevs(repo, 'extinct')
     return subset & extincts
 
 def extra(repo, subset, x):
     """``extra(label, [value])``
     Changesets with the given label in the extra metadata, with the given
     optional value.
 
     If `value` starts with `re:`, the remainder of the value is treated as
     a regular expression. To match a value that actually starts with `re:`,
     use the prefix `literal:`.
     """
 
     # i18n: "extra" is a keyword
     l = getargs(x, 1, 2, _('extra takes at least 1 and at most 2 arguments'))
     # i18n: "extra" is a keyword
     label = getstring(l[0], _('first argument to extra must be a string'))
     value = None
 
     if len(l) > 1:
         # i18n: "extra" is a keyword
         value = getstring(l[1], _('second argument to extra must be a string'))
         kind, value, matcher = _stringmatcher(value)
 
     def _matchvalue(r):
         extra = repo[r].extra()
         return label in extra and (value is None or matcher(extra[label]))
 
     return lazyset(subset, lambda r: _matchvalue(r))
 
 def filelog(repo, subset, x):
     """``filelog(pattern)``
     Changesets connected to the specified filelog.
 
     For performance reasons, ``filelog()`` does not show every changeset
     that affects the requested file(s). See :hg:`help log` for details. For
     a slower, more accurate result, use ``file()``.
 
     The pattern without explicit kind like ``glob:`` is expected to be
     relative to the current directory and match against a file exactly
     for efficiency.
     """
 
     # i18n: "filelog" is a keyword
     pat = getstring(x, _("filelog requires a pattern"))
     s = set()
 
     if not matchmod.patkind(pat):
         f = pathutil.canonpath(repo.root, repo.getcwd(), pat)
         fl = repo.file(f)
         for fr in fl:
             s.add(fl.linkrev(fr))
     else:
         m = matchmod.match(repo.root, repo.getcwd(), [pat], ctx=repo[None])
         for f in repo[None]:
             if m(f):
                 fl = repo.file(f)
                 for fr in fl:
                     s.add(fl.linkrev(fr))
 
     return baseset([r for r in subset if r in s])
 
 def first(repo, subset, x):
     """``first(set, [n])``
     An alias for limit().
     """
     return limit(repo, subset, x)
 
 def _follow(repo, subset, x, name, followfirst=False):
     l = getargs(x, 0, 1, _("%s takes no arguments or a filename") % name)
     c = repo['.']
     if l:
         x = getstring(l[0], _("%s expected a filename") % name)
         if x in c:
             cx = c[x]
             s = set(ctx.rev() for ctx in cx.ancestors(followfirst=followfirst))
             # include the revision responsible for the most recent version
             s.add(cx.linkrev())
         else:
             return baseset([])
     else:
         s = set(_revancestors(repo, [c.rev()], followfirst)) | set([c.rev()])
 
     return baseset([r for r in subset if r in s])
 
 def follow(repo, subset, x):
     """``follow([file])``
     An alias for ``::.`` (ancestors of the working copy's first parent).
     If a filename is specified, the history of the given file is followed,
     including copies.
     """
     return _follow(repo, subset, x, 'follow')
 
 def _followfirst(repo, subset, x):
     # ``followfirst([file])``
     # Like ``follow([file])`` but follows only the first parent of
     # every revision or file revision.
     return _follow(repo, subset, x, '_followfirst', followfirst=True)
 
 def getall(repo, subset, x):
     """``all()``
     All changesets, the same as ``0:tip``.
     """
     # i18n: "all" is a keyword
     getargs(x, 0, 0, _("all takes no arguments"))
     return subset
 
 def grep(repo, subset, x):
     """``grep(regex)``
     Like ``keyword(string)`` but accepts a regex. Use ``grep(r'...')``
     to ensure special escape characters are handled correctly. Unlike
     ``keyword(string)``, the match is case-sensitive.
     """
     try:
         # i18n: "grep" is a keyword
         gr = re.compile(getstring(x, _("grep requires a string")))
     except re.error, e:
         raise error.ParseError(_('invalid match pattern: %s') % e)
 
     def matches(x):
         c = repo[x]
         for e in c.files() + [c.user(), c.description()]:
             if gr.search(e):
                 return True
         return False
 
     return lazyset(subset, matches)
 
 def _matchfiles(repo, subset, x):
     # _matchfiles takes a revset list of prefixed arguments:
     #
     #   [p:foo, i:bar, x:baz]
     #
     # builds a match object from them and filters subset. Allowed
     # prefixes are 'p:' for regular patterns, 'i:' for include
     # patterns and 'x:' for exclude patterns. Use 'r:' prefix to pass
     # a revision identifier, or the empty string to reference the
     # working directory, from which the match object is
     # initialized. Use 'd:' to set the default matching mode, default
     # to 'glob'. At most one 'r:' and 'd:' argument can be passed.
 
     # i18n: "_matchfiles" is a keyword
     l = getargs(x, 1, -1, _("_matchfiles requires at least one argument"))
     pats, inc, exc = [], [], []
     hasset = False
     rev, default = None, None
     for arg in l:
         # i18n: "_matchfiles" is a keyword
         s = getstring(arg, _("_matchfiles requires string arguments"))
         prefix, value = s[:2], s[2:]
         if prefix == 'p:':
             pats.append(value)
         elif prefix == 'i:':
             inc.append(value)
         elif prefix == 'x:':
             exc.append(value)
         elif prefix == 'r:':
             if rev is not None:
                 # i18n: "_matchfiles" is a keyword
                 raise error.ParseError(_('_matchfiles expected at most one '
                                          'revision'))
             rev = value
         elif prefix == 'd:':
             if default is not None:
                 # i18n: "_matchfiles" is a keyword
                 raise error.ParseError(_('_matchfiles expected at most one '
                                          'default mode'))
             default = value
         else:
             # i18n: "_matchfiles" is a keyword
             raise error.ParseError(_('invalid _matchfiles prefix: %s') % prefix)
         if not hasset and matchmod.patkind(value) == 'set':
             hasset = True
     if not default:
         default = 'glob'
 
     def matches(x):
         m = None
         c = repo[x]
         if not m or (hasset and rev is None):
             ctx = c
             if rev is not None:
                 ctx = repo[rev or None]
             m = matchmod.match(repo.root, repo.getcwd(), pats, include=inc,
                                exclude=exc, ctx=ctx, default=default)
         for f in c.files():
             if m(f):
                 return True
         return False
 
     return lazyset(subset, matches)
 
 def hasfile(repo, subset, x):
     """``file(pattern)``
     Changesets affecting files matched by pattern.
 
     For a faster but less accurate result, consider using ``filelog()``
     instead.
 
     This predicate uses ``glob:`` as the default kind of pattern.
     """
     # i18n: "file" is a keyword
     pat = getstring(x, _("file requires a pattern"))
     return _matchfiles(repo, subset, ('string', 'p:' + pat))
 
 def head(repo, subset, x):
     """``head()``
     Changeset is a named branch head.
     """
     # i18n: "head" is a keyword
     getargs(x, 0, 0, _("head takes no arguments"))
     hs = set()
     for b, ls in repo.branchmap().iteritems():
         hs.update(repo[h].rev() for h in ls)
     return baseset([r for r in subset if r in hs])
 
 def heads(repo, subset, x):
     """``heads(set)``
     Members of set with no children in set.
     """
     s = getset(repo, subset, x)
     ps = parents(repo, subset, x)
     return s - ps
 
 def hidden(repo, subset, x):
     """``hidden()``
     Hidden changesets.
     """
     # i18n: "hidden" is a keyword
     getargs(x, 0, 0, _("hidden takes no arguments"))
     hiddenrevs = repoview.filterrevs(repo, 'visible')
     return subset & hiddenrevs
 
 def keyword(repo, subset, x):
     """``keyword(string)``
     Search commit message, user name, and names of changed files for
     string. The match is case-insensitive.
     """
     # i18n: "keyword" is a keyword
     kw = encoding.lower(getstring(x, _("keyword requires a string")))
 
     def matches(r):
         c = repo[r]
         return util.any(kw in encoding.lower(t) for t in c.files() + [c.user(),
             c.description()])
 
     return lazyset(subset, matches)
 
 def limit(repo, subset, x):
     """``limit(set, [n])``
     First n members of set, defaulting to 1.
     """
     # i18n: "limit" is a keyword
     l = getargs(x, 1, 2, _("limit requires one or two arguments"))
     try:
         lim = 1
         if len(l) == 2:
             # i18n: "limit" is a keyword
             lim = int(getstring(l[1], _("limit requires a number")))
     except (TypeError, ValueError):
         # i18n: "limit" is a keyword
         raise error.ParseError(_("limit expects a number"))
     ss = subset.set()
     os = getset(repo, baseset(repo), l[0])
     bs = baseset([])
     it = iter(os)
     for x in xrange(lim):
         try:
             y = it.next()
             if y in ss:
                 bs.append(y)
         except (StopIteration):
             break
     return bs
 
 def last(repo, subset, x):
     """``last(set, [n])``
     Last n members of set, defaulting to 1.
     """
     # i18n: "last" is a keyword
     l = getargs(x, 1, 2, _("last requires one or two arguments"))
     try:
         lim = 1
         if len(l) == 2:
             # i18n: "last" is a keyword
             lim = int(getstring(l[1], _("last requires a number")))
     except (TypeError, ValueError):
         # i18n: "last" is a keyword
         raise error.ParseError(_("last expects a number"))
     ss = subset.set()
     os = getset(repo, baseset(repo), l[0])[-lim:]
     return baseset([r for r in os if r in ss])
 
 def maxrev(repo, subset, x):
     """``max(set)``
     Changeset with highest revision number in set.
     """
     os = getset(repo, baseset(repo), x)
     if os:
         m = max(os)
         if m in subset:
             return baseset([m])
     return baseset([])
 
 def merge(repo, subset, x):
     """``merge()``
     Changeset is a merge changeset.
     """
     # i18n: "merge" is a keyword
     getargs(x, 0, 0, _("merge takes no arguments"))
     cl = repo.changelog
     return lazyset(subset, lambda r: cl.parentrevs(r)[1] != -1)
 
 def branchpoint(repo, subset, x):
     """``branchpoint()``
     Changesets with more than one child.
     """
     # i18n: "branchpoint" is a keyword
     getargs(x, 0, 0, _("branchpoint takes no arguments"))
     cl = repo.changelog
     if not subset:
         return baseset([])
     baserev = min(subset)
     parentscount = [0]*(len(repo) - baserev)
     for r in cl.revs(start=baserev + 1):
         for p in cl.parentrevs(r):
             if p >= baserev:
                 parentscount[p - baserev] += 1
     return baseset([r for r in subset if (parentscount[r - baserev] > 1)])
 
 def minrev(repo, subset, x):
     """``min(set)``
     Changeset with lowest revision number in set.
     """
     os = getset(repo, baseset(repo), x)
     if os:
         m = min(os)
         if m in subset:
             return baseset([m])
     return baseset([])
 
 def _missingancestors(repo, subset, x):
     # i18n: "_missingancestors" is a keyword
     revs, bases = getargs(x, 2, 2,
                           _("_missingancestors requires two arguments"))
     rs = baseset(repo)
     revs = getset(repo, rs, revs)
     bases = getset(repo, rs, bases)
     missing = set(repo.changelog.findmissingrevs(bases, revs))
     return baseset([r for r in subset if r in missing])
 
 def modifies(repo, subset, x):
     """``modifies(pattern)``
     Changesets modifying files matched by pattern.
 
     The pattern without explicit kind like ``glob:`` is expected to be
     relative to the current directory and match against a file or a
     directory.
     """
     # i18n: "modifies" is a keyword
     pat = getstring(x, _("modifies requires a pattern"))
     return checkstatus(repo, subset, pat, 0)
 
 def node_(repo, subset, x):
     """``id(string)``
     Revision non-ambiguously specified by the given hex string prefix.
     """
     # i18n: "id" is a keyword
     l = getargs(x, 1, 1, _("id requires one argument"))
     # i18n: "id" is a keyword
     n = getstring(l[0], _("id requires a string"))
     if len(n) == 40:
         rn = repo[n].rev()
     else:
         rn = None
         pm = repo.changelog._partialmatch(n)
         if pm is not None:
             rn = repo.changelog.rev(pm)
 
     return baseset([r for r in subset if r == rn])
 
 def obsolete(repo, subset, x):
     """``obsolete()``
     Mutable changeset with a newer version."""
     # i18n: "obsolete" is a keyword
     getargs(x, 0, 0, _("obsolete takes no arguments"))
     obsoletes = obsmod.getrevs(repo, 'obsolete')
     return subset & obsoletes
 
 def origin(repo, subset, x):
     """``origin([set])``
     Changesets that were specified as a source for the grafts, transplants or
     rebases that created the given revisions.  Omitting the optional set is the
     same as passing all().  If a changeset created by these operations is itself
     specified as a source for one of these operations, only the source changeset
     for the first operation is selected.
     """
     if x is not None:
         args = getset(repo, baseset(repo), x).set()
     else:
         args = getall(repo, baseset(repo), x).set()
 
     def _firstsrc(rev):
         src = _getrevsource(repo, rev)
         if src is None:
             return None
 
         while True:
             prev = _getrevsource(repo, src)
 
             if prev is None:
                 return src
             src = prev
 
     o = set([_firstsrc(r) for r in args])
     return baseset([r for r in subset if r in o])
 
 def outgoing(repo, subset, x):
     """``outgoing([path])``
     Changesets not found in the specified destination repository, or the
     default push location.
     """
     import hg # avoid start-up nasties
     # i18n: "outgoing" is a keyword
     l = getargs(x, 0, 1, _("outgoing takes one or no arguments"))
     # i18n: "outgoing" is a keyword
     dest = l and getstring(l[0], _("outgoing requires a repository path")) or ''
     dest = repo.ui.expandpath(dest or 'default-push', dest or 'default')
     dest, branches = hg.parseurl(dest)
     revs, checkout = hg.addbranchrevs(repo, repo, branches, [])
     if revs:
         revs = [repo.lookup(rev) for rev in revs]
     other = hg.peer(repo, {}, dest)
     repo.ui.pushbuffer()
     outgoing = discovery.findcommonoutgoing(repo, other, onlyheads=revs)
     repo.ui.popbuffer()
     cl = repo.changelog
     o = set([cl.rev(r) for r in outgoing.missing])
     return baseset([r for r in subset if r in o])
 
 def p1(repo, subset, x):
     """``p1([set])``
     First parent of changesets in set, or the working directory.
     """
     if x is None:
         p = repo[x].p1().rev()
         return baseset([r for r in subset if r == p])
 
     ps = set()
     cl = repo.changelog
     for r in getset(repo, baseset(repo), x):
         ps.add(cl.parentrevs(r)[0])
     return subset & ps
 
 def p2(repo, subset, x):
     """``p2([set])``
     Second parent of changesets in set, or the working directory.
     """
     if x is None:
         ps = repo[x].parents()
         try:
             p = ps[1].rev()
             return baseset([r for r in subset if r == p])
         except IndexError:
             return baseset([])
 
     ps = set()
     cl = repo.changelog
     for r in getset(repo, baseset(repo), x):
         ps.add(cl.parentrevs(r)[1])
     return subset & ps
 
 def parents(repo, subset, x):
     """``parents([set])``
     The set of all parents for all changesets in set, or the working directory.
     """
     if x is None:
         ps = tuple(p.rev() for p in repo[x].parents())
         return subset & ps
 
     ps = set()
     cl = repo.changelog
     for r in getset(repo, baseset(repo), x):
         ps.update(cl.parentrevs(r))
     return subset & ps
 
 def parentspec(repo, subset, x, n):
     """``set^0``
     The set.
     ``set^1`` (or ``set^``), ``set^2``
     First or second parent, respectively, of all changesets in set.
     """
     try:
         n = int(n[1])
         if n not in (0, 1, 2):
             raise ValueError
     except (TypeError, ValueError):
         raise error.ParseError(_("^ expects a number 0, 1, or 2"))
     ps = set()
     cl = repo.changelog
     for r in getset(repo, baseset(cl), x):
         if n == 0:
             ps.add(r)
         elif n == 1:
             ps.add(cl.parentrevs(r)[0])
         elif n == 2:
             parents = cl.parentrevs(r)
             if len(parents) > 1:
                 ps.add(parents[1])
     return subset & ps
 
 def present(repo, subset, x):
     """``present(set)``
     An empty set, if any revision in set isn't found; otherwise,
     all revisions in set.
 
     If any of specified revisions is not present in the local repository,
     the query is normally aborted. But this predicate allows the query
     to continue even in such cases.
     """
     try:
         return getset(repo, subset, x)
     except error.RepoLookupError:
         return baseset([])
 
 def public(repo, subset, x):
     """``public()``
     Changeset in public phase."""
     # i18n: "public" is a keyword
     getargs(x, 0, 0, _("public takes no arguments"))
     pc = repo._phasecache
     return lazyset(subset, lambda r: pc.phase(repo, r) == phases.public)
 
 def remote(repo, subset, x):
     """``remote([id [,path]])``
     Local revision that corresponds to the given identifier in a
     remote repository, if present. Here, the '.' identifier is a
     synonym for the current local branch.
     """
 
     import hg # avoid start-up nasties
     # i18n: "remote" is a keyword
     l = getargs(x, 0, 2, _("remote takes one, two or no arguments"))
 
     q = '.'
     if len(l) > 0:
     # i18n: "remote" is a keyword
         q = getstring(l[0], _("remote requires a string id"))
     if q == '.':
         q = repo['.'].branch()
 
     dest = ''
     if len(l) > 1:
         # i18n: "remote" is a keyword
         dest = getstring(l[1], _("remote requires a repository path"))
     dest = repo.ui.expandpath(dest or 'default')
     dest, branches = hg.parseurl(dest)
     revs, checkout = hg.addbranchrevs(repo, repo, branches, [])
     if revs:
         revs = [repo.lookup(rev) for rev in revs]
     other = hg.peer(repo, {}, dest)
     n = other.lookup(q)
     if n in repo:
         r = repo[n].rev()
         if r in subset:
             return baseset([r])
     return baseset([])
 
 def removes(repo, subset, x):
     """``removes(pattern)``
     Changesets which remove files matching pattern.
 
     The pattern without explicit kind like ``glob:`` is expected to be
     relative to the current directory and match against a file or a
     directory.
     """
     # i18n: "removes" is a keyword
     pat = getstring(x, _("removes requires a pattern"))
     return checkstatus(repo, subset, pat, 2)
 
 def rev(repo, subset, x):
     """``rev(number)``
     Revision with the given numeric identifier.
     """
     # i18n: "rev" is a keyword
     l = getargs(x, 1, 1, _("rev requires one argument"))
     try:
         # i18n: "rev" is a keyword
         l = int(getstring(l[0], _("rev requires a number")))
     except (TypeError, ValueError):
         # i18n: "rev" is a keyword
         raise error.ParseError(_("rev expects a number"))
     return baseset([r for r in subset if r == l])
 
 def matching(repo, subset, x):
     """``matching(revision [, field])``
     Changesets in which a given set of fields match the set of fields in the
     selected revision or set.
 
     To match more than one field pass the list of fields to match separated
     by spaces (e.g. ``author description``).
 
     Valid fields are most regular revision fields and some special fields.
 
     Regular revision fields are ``description``, ``author``, ``branch``,
     ``date``, ``files``, ``phase``, ``parents``, ``substate``, ``user``
     and ``diff``.
     Note that ``author`` and ``user`` are synonyms. ``diff`` refers to the
     contents of the revision. Two revisions matching their ``diff`` will
     also match their ``files``.
 
     Special fields are ``summary`` and ``metadata``:
     ``summary`` matches the first line of the description.
     ``metadata`` is equivalent to matching ``description user date``
     (i.e. it matches the main metadata fields).
 
     ``metadata`` is the default field which is used when no fields are
     specified. You can match more than one field at a time.
     """
     # i18n: "matching" is a keyword
     l = getargs(x, 1, 2, _("matching takes 1 or 2 arguments"))
 
     revs = getset(repo, baseset(repo.changelog), l[0])
 
     fieldlist = ['metadata']
     if len(l) > 1:
             fieldlist = getstring(l[1],
                 # i18n: "matching" is a keyword
                 _("matching requires a string "
                 "as its second argument")).split()
 
     # Make sure that there are no repeated fields,
     # expand the 'special' 'metadata' field type
     # and check the 'files' whenever we check the 'diff'
     fields = []
     for field in fieldlist:
         if field == 'metadata':
             fields += ['user', 'description', 'date']
         elif field == 'diff':
             # a revision matching the diff must also match the files
             # since matching the diff is very costly, make sure to
             # also match the files first
             fields += ['files', 'diff']
         else:
             if field == 'author':
                 field = 'user'
             fields.append(field)
     fields = set(fields)
     if 'summary' in fields and 'description' in fields:
         # If a revision matches its description it also matches its summary
         fields.discard('summary')
 
     # We may want to match more than one field
     # Not all fields take the same amount of time to be matched
     # Sort the selected fields in order of increasing matching cost
     fieldorder = ['phase', 'parents', 'user', 'date', 'branch', 'summary',
         'files', 'description', 'substate', 'diff']
     def fieldkeyfunc(f):
         try:
             return fieldorder.index(f)
         except ValueError:
             # assume an unknown field is very costly
             return len(fieldorder)
     fields = list(fields)
     fields.sort(key=fieldkeyfunc)
 
     # Each field will be matched with its own "getfield" function
     # which will be added to the getfieldfuncs array of functions
     getfieldfuncs = []
     _funcs = {
         'user': lambda r: repo[r].user(),
         'branch': lambda r: repo[r].branch(),
         'date': lambda r: repo[r].date(),
         'description': lambda r: repo[r].description(),
         'files': lambda r: repo[r].files(),
         'parents': lambda r: repo[r].parents(),
         'phase': lambda r: repo[r].phase(),
         'substate': lambda r: repo[r].substate,
         'summary': lambda r: repo[r].description().splitlines()[0],
         'diff': lambda r: list(repo[r].diff(git=True),)
     }
     for info in fields:
         getfield = _funcs.get(info, None)
         if getfield is None:
             raise error.ParseError(
                 # i18n: "matching" is a keyword
                 _("unexpected field name passed to matching: %s") % info)
         getfieldfuncs.append(getfield)
     # convert the getfield array of functions into a "getinfo" function
     # which returns an array of field values (or a single value if there
     # is only one field to match)
     getinfo = lambda r: [f(r) for f in getfieldfuncs]
 
     def matches(x):
         for rev in revs:
             target = getinfo(rev)
             match = True
             for n, f in enumerate(getfieldfuncs):
                 if target[n] != f(x):
                     match = False
             if match:
                 return True
         return False
 
     return lazyset(subset, matches)
 
 def reverse(repo, subset, x):
     """``reverse(set)``
     Reverse order of set.
     """
     l = getset(repo, subset, x)
     l.reverse()
     return l
 
 def roots(repo, subset, x):
     """``roots(set)``
     Changesets in set with no parent changeset in set.
     """
     s = getset(repo, baseset(repo.changelog), x).set()
     subset = baseset([r for r in subset if r in s])
     cs = _children(repo, subset, s)
     return subset - cs
 
 def secret(repo, subset, x):
     """``secret()``
     Changeset in secret phase."""
     # i18n: "secret" is a keyword
     getargs(x, 0, 0, _("secret takes no arguments"))
     pc = repo._phasecache
     return lazyset(subset, lambda x: pc.phase(repo, x) == phases.secret)
 
 def sort(repo, subset, x):
     """``sort(set[, [-]key...])``
     Sort set by keys. The default sort order is ascending, specify a key
     as ``-key`` to sort in descending order.
 
     The keys can be:
 
     - ``rev`` for the revision number,
     - ``branch`` for the branch name,
     - ``desc`` for the commit message (description),
     - ``user`` for user name (``author`` can be used as an alias),
     - ``date`` for the commit date
     """
     # i18n: "sort" is a keyword
     l = getargs(x, 1, 2, _("sort requires one or two arguments"))
     keys = "rev"
     if len(l) == 2:
         # i18n: "sort" is a keyword
         keys = getstring(l[1], _("sort spec must be a string"))
 
     s = l[0]
     keys = keys.split()
     l = []
     def invert(s):
         return "".join(chr(255 - ord(c)) for c in s)
     for r in getset(repo, subset, s):
         c = repo[r]
         e = []
         for k in keys:
             if k == 'rev':
                 e.append(r)
             elif k == '-rev':
                 e.append(-r)
             elif k == 'branch':
                 e.append(c.branch())
             elif k == '-branch':
                 e.append(invert(c.branch()))
             elif k == 'desc':
                 e.append(c.description())
             elif k == '-desc':
                 e.append(invert(c.description()))
             elif k in 'user author':
                 e.append(c.user())
             elif k in '-user -author':
                 e.append(invert(c.user()))
             elif k == 'date':
                 e.append(c.date()[0])
             elif k == '-date':
                 e.append(-c.date()[0])
             else:
                 raise error.ParseError(_("unknown sort key %r") % k)
         e.append(r)
         l.append(e)
     l.sort()
     return baseset([e[-1] for e in l])
 
 def _stringmatcher(pattern):
     """
     accepts a string, possibly starting with 're:' or 'literal:' prefix.
     returns the matcher name, pattern, and matcher function.
     missing or unknown prefixes are treated as literal matches.
 
     helper for tests:
     >>> def test(pattern, *tests):
     ...     kind, pattern, matcher = _stringmatcher(pattern)
     ...     return (kind, pattern, [bool(matcher(t)) for t in tests])
 
     exact matching (no prefix):
     >>> test('abcdefg', 'abc', 'def', 'abcdefg')
     ('literal', 'abcdefg', [False, False, True])
 
     regex matching ('re:' prefix)
     >>> test('re:a.+b', 'nomatch', 'fooadef', 'fooadefbar')
     ('re', 'a.+b', [False, False, True])
 
     force exact matches ('literal:' prefix)
     >>> test('literal:re:foobar', 'foobar', 're:foobar')
     ('literal', 're:foobar', [False, True])
 
     unknown prefixes are ignored and treated as literals
     >>> test('foo:bar', 'foo', 'bar', 'foo:bar')
     ('literal', 'foo:bar', [False, False, True])
     """
     if pattern.startswith('re:'):
         pattern = pattern[3:]
         try:
             regex = re.compile(pattern)
         except re.error, e:
             raise error.ParseError(_('invalid regular expression: %s')
                                    % e)
         return 're', pattern, regex.search
     elif pattern.startswith('literal:'):
         pattern = pattern[8:]
     return 'literal', pattern, pattern.__eq__
 
 def _substringmatcher(pattern):
     kind, pattern, matcher = _stringmatcher(pattern)
     if kind == 'literal':
         matcher = lambda s: pattern in s
     return kind, pattern, matcher
 
 def tag(repo, subset, x):
     """``tag([name])``
     The specified tag by name, or all tagged revisions if no name is given.
     """
     # i18n: "tag" is a keyword
     args = getargs(x, 0, 1, _("tag takes one or no arguments"))
     cl = repo.changelog
     if args:
         pattern = getstring(args[0],
                             # i18n: "tag" is a keyword
                             _('the argument to tag must be a string'))
         kind, pattern, matcher = _stringmatcher(pattern)
         if kind == 'literal':
             # avoid resolving all tags
             tn = repo._tagscache.tags.get(pattern, None)
             if tn is None:
                 raise util.Abort(_("tag '%s' does not exist") % pattern)
             s = set([repo[tn].rev()])
         else:
             s = set([cl.rev(n) for t, n in repo.tagslist() if matcher(t)])
     else:
         s = set([cl.rev(n) for t, n in repo.tagslist() if t != 'tip'])
     return subset & s
 
 def tagged(repo, subset, x):
     return tag(repo, subset, x)
 
 def unstable(repo, subset, x):
     """``unstable()``
     Non-obsolete changesets with obsolete ancestors.
     """
     # i18n: "unstable" is a keyword
     getargs(x, 0, 0, _("unstable takes no arguments"))
     unstables = obsmod.getrevs(repo, 'unstable')
     return subset & unstables
 
 
 def user(repo, subset, x):
     """``user(string)``
     User name contains string. The match is case-insensitive.
 
     If `string` starts with `re:`, the remainder of the string is treated as
     a regular expression. To match a user that actually contains `re:`, use
     the prefix `literal:`.
     """
     return author(repo, subset, x)
 
 # for internal use
 def _list(repo, subset, x):
     s = getstring(x, "internal error")
     if not s:
         return baseset([])
     ls = [repo[r].rev() for r in s.split('\0')]
     s = subset.set()
     return baseset([r for r in ls if r in s])
 
 symbols = {
     "adds": adds,
     "all": getall,
     "ancestor": ancestor,
     "ancestors": ancestors,
     "_firstancestors": _firstancestors,
     "author": author,
     "bisect": bisect,
     "bisected": bisected,
     "bookmark": bookmark,
     "branch": branch,
     "branchpoint": branchpoint,
     "bumped": bumped,
     "bundle": bundle,
     "children": children,
     "closed": closed,
     "contains": contains,
     "converted": converted,
     "date": date,
     "desc": desc,
     "descendants": descendants,
     "_firstdescendants": _firstdescendants,
     "destination": destination,
     "divergent": divergent,
     "draft": draft,
     "extinct": extinct,
     "extra": extra,
     "file": hasfile,
     "filelog": filelog,
     "first": first,
     "follow": follow,
     "_followfirst": _followfirst,
     "grep": grep,
     "head": head,
     "heads": heads,
     "hidden": hidden,
     "id": node_,
     "keyword": keyword,
     "last": last,
     "limit": limit,
     "_matchfiles": _matchfiles,
     "max": maxrev,
     "merge": merge,
     "min": minrev,
     "_missingancestors": _missingancestors,
     "modifies": modifies,
     "obsolete": obsolete,
     "origin": origin,
     "outgoing": outgoing,
     "p1": p1,
     "p2": p2,
     "parents": parents,
     "present": present,
     "public": public,
     "remote": remote,
     "removes": removes,
     "rev": rev,
     "reverse": reverse,
     "roots": roots,
     "sort": sort,
     "secret": secret,
     "matching": matching,
     "tag": tag,
     "tagged": tagged,
     "user": user,
     "unstable": unstable,
     "_list": _list,
 }
 
 # symbols which can't be used for a DoS attack for any given input
 # (e.g. those which accept regexes as plain strings shouldn't be included)
 # functions that just return a lot of changesets (like all) don't count here
 safesymbols = set([
     "adds",
     "all",
     "ancestor",
     "ancestors",
     "_firstancestors",
     "author",
     "bisect",
     "bisected",
     "bookmark",
     "branch",
     "branchpoint",
     "bumped",
     "bundle",
     "children",
     "closed",
     "converted",
     "date",
     "desc",
     "descendants",
     "_firstdescendants",
     "destination",
     "divergent",
     "draft",
     "extinct",
     "extra",
     "file",
     "filelog",
     "first",
     "follow",
     "_followfirst",
     "head",
     "heads",
     "hidden",
     "id",
     "keyword",
     "last",
     "limit",
     "_matchfiles",
     "max",
     "merge",
     "min",
     "_missingancestors",
     "modifies",
     "obsolete",
     "origin",
     "outgoing",
     "p1",
     "p2",
     "parents",
     "present",
     "public",
     "remote",
     "removes",
     "rev",
     "reverse",
     "roots",
     "sort",
     "secret",
     "matching",
     "tag",
     "tagged",
     "user",
     "unstable",
     "_list",
 ])
 
 methods = {
     "range": rangeset,
     "dagrange": dagrange,
     "string": stringset,
     "symbol": symbolset,
     "and": andset,
     "or": orset,
     "not": notset,
     "list": listset,
     "func": func,
     "ancestor": ancestorspec,
     "parent": parentspec,
     "parentpost": p1,
 }
 
 def optimize(x, small):
     if x is None:
         return 0, x
 
     smallbonus = 1
     if small:
         smallbonus = .5
 
     op = x[0]
     if op == 'minus':
         return optimize(('and', x[1], ('not', x[2])), small)
     elif op == 'dagrangepre':
         return optimize(('func', ('symbol', 'ancestors'), x[1]), small)
     elif op == 'dagrangepost':
         return optimize(('func', ('symbol', 'descendants'), x[1]), small)
     elif op == 'rangepre':
         return optimize(('range', ('string', '0'), x[1]), small)
     elif op == 'rangepost':
         return optimize(('range', x[1], ('string', 'tip')), small)
     elif op == 'negate':
         return optimize(('string',
                          '-' + getstring(x[1], _("can't negate that"))), small)
     elif op in 'string symbol negate':
         return smallbonus, x # single revisions are small
     elif op == 'and':
         wa, ta = optimize(x[1], True)
         wb, tb = optimize(x[2], True)
+
+        # (::x and not ::y)/(not ::y and ::x) have a fast path
+        def ismissingancestors(revs, bases):
+            return (
+                revs[0] == 'func'
+                and getstring(revs[1], _('not a symbol')) == 'ancestors'
+                and bases[0] == 'not'
+                and bases[1][0] == 'func'
+                and getstring(bases[1][1], _('not a symbol')) == 'ancestors')
+
         w = min(wa, wb)
+        if ismissingancestors(ta, tb):
+            return w, ('func', ('symbol', '_missingancestors'),
+                       ('list', ta[2], tb[1][2]))
+        if ismissingancestors(tb, ta):
+            return w, ('func', ('symbol', '_missingancestors'),
+                       ('list', tb[2], ta[1][2]))
+
         if wa > wb:
             return w, (op, tb, ta)
         return w, (op, ta, tb)
     elif op == 'or':
         wa, ta = optimize(x[1], False)
         wb, tb = optimize(x[2], False)
         if wb < wa:
             wb, wa = wa, wb
         return max(wa, wb), (op, ta, tb)
     elif op == 'not':
         o = optimize(x[1], not small)
         return o[0], (op, o[1])
     elif op == 'parentpost':
         o = optimize(x[1], small)
         return o[0], (op, o[1])
     elif op == 'group':
         return optimize(x[1], small)
     elif op in 'dagrange range list parent ancestorspec':
         if op == 'parent':
             # x^:y means (x^) : y, not x ^ (:y)
             post = ('parentpost', x[1])
             if x[2][0] == 'dagrangepre':
                 return optimize(('dagrange', post, x[2][1]), small)
             elif x[2][0] == 'rangepre':
                 return optimize(('range', post, x[2][1]), small)
 
         wa, ta = optimize(x[1], small)
         wb, tb = optimize(x[2], small)
         return wa + wb, (op, ta, tb)
     elif op == 'func':
         f = getstring(x[1], _("not a symbol"))
         wa, ta = optimize(x[2], small)
         if f in ("author branch closed date desc file grep keyword "
                  "outgoing user"):
             w = 10 # slow
         elif f in "modifies adds removes":
             w = 30 # slower
         elif f == "contains":
             w = 100 # very slow
         elif f == "ancestor":
             w = 1 * smallbonus
         elif f in "reverse limit first":
             w = 0
         elif f in "sort":
             w = 10 # assume most sorts look at changelog
         else:
             w = 1
         return w + wa, (op, x[1], ta)
     return 1, x
 
 _aliasarg = ('func', ('symbol', '_aliasarg'))
 def _getaliasarg(tree):
     """If tree matches ('func', ('symbol', '_aliasarg'), ('string', X))
     return X, None otherwise.
     """
     if (len(tree) == 3 and tree[:2] == _aliasarg
         and tree[2][0] == 'string'):
         return tree[2][1]
     return None
 
 def _checkaliasarg(tree, known=None):
     """Check tree contains no _aliasarg construct or only ones which
     value is in known. Used to avoid alias placeholders injection.
     """
     if isinstance(tree, tuple):
         arg = _getaliasarg(tree)
         if arg is not None and (not known or arg not in known):
             raise error.ParseError(_("not a function: %s") % '_aliasarg')
         for t in tree:
             _checkaliasarg(t, known)
 
 class revsetalias(object):
     funcre = re.compile('^([^(]+)\(([^)]+)\)$')
     args = None
 
     def __init__(self, name, value):
         '''Aliases like:
 
         h = heads(default)
         b($1) = ancestors($1) - ancestors(default)
         '''
         m = self.funcre.search(name)
         if m:
             self.name = m.group(1)
             self.tree = ('func', ('symbol', m.group(1)))
             self.args = [x.strip() for x in m.group(2).split(',')]
             for arg in self.args:
                 # _aliasarg() is an unknown symbol only used separate
                 # alias argument placeholders from regular strings.
                 value = value.replace(arg, '_aliasarg(%r)' % (arg,))
         else:
             self.name = name
             self.tree = ('symbol', name)
 
         self.replacement, pos = parse(value)
         if pos != len(value):
             raise error.ParseError(_('invalid token'), pos)
         # Check for placeholder injection
         _checkaliasarg(self.replacement, self.args)
 
 def _getalias(aliases, tree):
     """If tree looks like an unexpanded alias, return it. Return None
     otherwise.
     """
     if isinstance(tree, tuple) and tree:
         if tree[0] == 'symbol' and len(tree) == 2:
             name = tree[1]
             alias = aliases.get(name)
             if alias and alias.args is None and alias.tree == tree:
                 return alias
         if tree[0] == 'func' and len(tree) > 1:
             if tree[1][0] == 'symbol' and len(tree[1]) == 2:
                 name = tree[1][1]
                 alias = aliases.get(name)
                 if alias and alias.args is not None and alias.tree == tree[:2]:
                     return alias
     return None
 
 def _expandargs(tree, args):
     """Replace _aliasarg instances with the substitution value of the
     same name in args, recursively.
     """
     if not tree or not isinstance(tree, tuple):
         return tree
     arg = _getaliasarg(tree)
     if arg is not None:
         return args[arg]
     return tuple(_expandargs(t, args) for t in tree)
 
 def _expandaliases(aliases, tree, expanding, cache):
     """Expand aliases in tree, recursively.
 
     'aliases' is a dictionary mapping user defined aliases to
     revsetalias objects.
     """
     if not isinstance(tree, tuple):
         # Do not expand raw strings
         return tree
     alias = _getalias(aliases, tree)
     if alias is not None:
         if alias in expanding:
             raise error.ParseError(_('infinite expansion of revset alias "%s" '
                                      'detected') % alias.name)
         expanding.append(alias)
         if alias.name not in cache:
             cache[alias.name] = _expandaliases(aliases, alias.replacement,
                                                expanding, cache)
         result = cache[alias.name]
         expanding.pop()
         if alias.args is not None:
             l = getlist(tree[2])
             if len(l) != len(alias.args):
                 raise error.ParseError(
                     _('invalid number of arguments: %s') % len(l))
             l = [_expandaliases(aliases, a, [], cache) for a in l]
             result = _expandargs(result, dict(zip(alias.args, l)))
     else:
         result = tuple(_expandaliases(aliases, t, expanding, cache)
                        for t in tree)
     return result
 
 def findaliases(ui, tree):
     _checkaliasarg(tree)
     aliases = {}
     for k, v in ui.configitems('revsetalias'):
         alias = revsetalias(k, v)
         aliases[alias.name] = alias
     return _expandaliases(aliases, tree, [], {})
 
 def parse(spec):
     p = parser.parser(tokenize, elements)
     return p.parse(spec)
 
 def match(ui, spec):
     if not spec:
         raise error.ParseError(_("empty query"))
     tree, pos = parse(spec)
     if (pos != len(spec)):
         raise error.ParseError(_("invalid token"), pos)
     if ui:
         tree = findaliases(ui, tree)
     weight, tree = optimize(tree, True)
     def mfunc(repo, subset):
         return getset(repo, subset, tree)
     return mfunc
 
 def formatspec(expr, *args):
     '''
     This is a convenience function for using revsets internally, and
     escapes arguments appropriately. Aliases are intentionally ignored
     so that intended expression behavior isn't accidentally subverted.
 
     Supported arguments:
 
     %r = revset expression, parenthesized
     %d = int(arg), no quoting
     %s = string(arg), escaped and single-quoted
     %b = arg.branch(), escaped and single-quoted
     %n = hex(arg), single-quoted
     %% = a literal '%'
 
     Prefixing the type with 'l' specifies a parenthesized list of that type.
 
     >>> formatspec('%r:: and %lr', '10 or 11', ("this()", "that()"))
     '(10 or 11):: and ((this()) or (that()))'
     >>> formatspec('%d:: and not %d::', 10, 20)
     '10:: and not 20::'
     >>> formatspec('%ld or %ld', [], [1])
     "_list('') or 1"
     >>> formatspec('keyword(%s)', 'foo\\xe9')
     "keyword('foo\\\\xe9')"
     >>> b = lambda: 'default'
     >>> b.branch = b
     >>> formatspec('branch(%b)', b)
     "branch('default')"
     >>> formatspec('root(%ls)', ['a', 'b', 'c', 'd'])
     "root(_list('a\\x00b\\x00c\\x00d'))"
     '''
 
     def quote(s):
         return repr(str(s))
 
     def argtype(c, arg):
         if c == 'd':
             return str(int(arg))
         elif c == 's':
             return quote(arg)
         elif c == 'r':
             parse(arg) # make sure syntax errors are confined
             return '(%s)' % arg
         elif c == 'n':
             return quote(node.hex(arg))
         elif c == 'b':
             return quote(arg.branch())
 
     def listexp(s, t):
         l = len(s)
         if l == 0:
             return "_list('')"
         elif l == 1:
             return argtype(t, s[0])
         elif t == 'd':
             return "_list('%s')" % "\0".join(str(int(a)) for a in s)
         elif t == 's':
             return "_list('%s')" % "\0".join(s)
         elif t == 'n':
             return "_list('%s')" % "\0".join(node.hex(a) for a in s)
         elif t == 'b':
             return "_list('%s')" % "\0".join(a.branch() for a in s)
 
         m = l // 2
         return '(%s or %s)' % (listexp(s[:m], t), listexp(s[m:], t))
 
     ret = ''
     pos = 0
     arg = 0
     while pos < len(expr):
         c = expr[pos]
         if c == '%':
             pos += 1
             d = expr[pos]
             if d == '%':
                 ret += d
             elif d in 'dsnbr':
                 ret += argtype(d, args[arg])
                 arg += 1
             elif d == 'l':
                 # a list of some type
                 pos += 1
                 d = expr[pos]
                 ret += listexp(list(args[arg]), d)
                 arg += 1
             else:
                 raise util.Abort('unexpected revspec format character %s' % d)
         else:
             ret += c
         pos += 1
 
     return ret
 
 def prettyformat(tree):
     def _prettyformat(tree, level, lines):
         if not isinstance(tree, tuple) or tree[0] in ('string', 'symbol'):
             lines.append((level, str(tree)))
         else:
             lines.append((level, '(%s' % tree[0]))
             for s in tree[1:]:
                 _prettyformat(s, level + 1, lines)
             lines[-1:] = [(lines[-1][0], lines[-1][1] + ')')]
 
     lines = []
     _prettyformat(tree, 0, lines)
     output = '\n'.join(('  '*l + s) for l, s in lines)
     return output
 
 def depth(tree):
     if isinstance(tree, tuple):
         return max(map(depth, tree)) + 1
     else:
         return 0
 
 def funcsused(tree):
     if not isinstance(tree, tuple) or tree[0] in ('string', 'symbol'):
         return set()
     else:
         funcs = set()
         for s in tree[1:]:
             funcs |= funcsused(s)
         if tree[0] == 'func':
             funcs.add(tree[1][1])
         return funcs
 
 class baseset(list):
     """Basic data structure that represents a revset and contains the basic
     operation that it should be able to perform.
     """
     def __init__(self, data):
         super(baseset, self).__init__(data)
         self._set = None
 
     def set(self):
         if not self._set:
             self._set = set(self)
         return self._set
 
     def __sub__(self, x):
         if isinstance(x, baseset):
             s = x.set()
         else:
             s = set(x)
         return baseset(self.set() - s)
 
     def __and__(self, x):
         if isinstance(x, baseset):
             x = x.set()
         return baseset([y for y in self if y in x])
 
     def __add__(self, x):
         s = self.set()
         l = [r for r in x if r not in s]
         return baseset(list(self) + l)
 
 class lazyset(object):
     """Duck type for baseset class which iterates lazily over the revisions in
     the subset and contains a function which tests for membership in the
     revset
     """
     def __init__(self, subset, condition):
         self._subset = subset
         self._condition = condition
 
     def __contains__(self, x):
         return x in self._subset and self._condition(x)
 
     def __iter__(self):
         cond = self._condition
         for x in self._subset:
             if cond(x):
                 yield x
 
     def __and__(self, x):
         return lazyset(self, lambda r: r in x)
 
     def __sub__(self, x):
         return lazyset(self, lambda r: r not in x)
 
     def __add__(self, x):
         l = baseset([r for r in self])
         return l + baseset(x)
 
     def __len__(self):
         # Basic implementation to be changed in future patches.
         l = baseset([r for r in self])
         return len(l)
 
     def __getitem__(self, x):
         # Basic implementation to be changed in future patches.
         l = baseset([r for r in self])
         return l[x]
 
     def sort(self, reverse=False):
         # Basic implementation to be changed in future patches.
         self._subset = baseset(self._subset)
         self._subset.sort(reverse=reverse)
 
     def reverse(self):
         self._subset.reverse()
 
     def set(self):
         return set([r for r in self])
 
 class spanset(object):
     """Duck type for baseset class which represents a range of revisions and
     can work lazily and without having all the range in memory
     """
     def __init__(self, start, end):
         self._start = start
         self._end = end
 
     def __iter__(self):
         if self._start <= self._end:
             for r in xrange(self._start, self._end):
                 yield r
         else:
             for r in xrange(self._start, self._end, -1):
                 yield r
 
     def __contains__(self, x):
         return (x <= self._start and x > self._end) or (x >= self._start and x<
                 self._end)
 
     def __and__(self, x):
         return lazyset(self, lambda r: r in x)
 
     def __sub__(self, x):
         return lazyset(self, lambda r: r not in x)
 
     def __add__(self, x):
         l = baseset(self)
         return l + baseset(x)
 
     def __len__(self):
         return abs(self._end - self._start)
 
     def __getitem__(self, x):
         # Basic implementation to be changed in future patches.
         l = baseset([r for r in self])
         return l[x]
 
     def sort(self, reverse=False):
         # Basic implementation to be changed in future patches.
         if reverse:
             self.reverse()
 
     def reverse(self):
         if self._start <= self._end:
             self._start, self._end = self._end - 1, self._start - 1
         else:
             self._start, self._end = self._end + 1, self._start + 1
 
     def set(self):
         return self
 
 # tell hggettext to extract docstrings from these functions:
 i18nfunctions = symbols.values()

tests/test-revset.t

@@ -1,895 +1,959 b''
   $ HGENCODING=utf-8
   $ export HGENCODING
 
   $ try() {
   >   hg debugrevspec --debug "$@"
   > }
 
   $ log() {
   >   hg log --template '{rev}\n' -r "$1"
   > }
 
   $ hg init repo
   $ cd repo
 
   $ echo a > a
   $ hg branch a
   marked working directory as branch a
   (branches are permanent and global, did you want a bookmark?)
   $ hg ci -Aqm0
 
   $ echo b > b
   $ hg branch b
   marked working directory as branch b
   (branches are permanent and global, did you want a bookmark?)
   $ hg ci -Aqm1
 
   $ rm a
   $ hg branch a-b-c-
   marked working directory as branch a-b-c-
   (branches are permanent and global, did you want a bookmark?)
   $ hg ci -Aqm2 -u Bob
 
   $ hg log -r "extra('branch', 'a-b-c-')" --template '{rev}\n'
   2
   $ hg log -r "extra('branch')" --template '{rev}\n'
   0
   1
   2
   $ hg log -r "extra('branch', 're:a')" --template '{rev} {branch}\n'
   0 a
   2 a-b-c-
 
   $ hg co 1
   1 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg branch +a+b+c+
   marked working directory as branch +a+b+c+
   (branches are permanent and global, did you want a bookmark?)
   $ hg ci -Aqm3
 
   $ hg co 2  # interleave
   0 files updated, 0 files merged, 1 files removed, 0 files unresolved
   $ echo bb > b
   $ hg branch -- -a-b-c-
   marked working directory as branch -a-b-c-
   (branches are permanent and global, did you want a bookmark?)
   $ hg ci -Aqm4 -d "May 12 2005"
 
   $ hg co 3
   2 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg branch !a/b/c/
   marked working directory as branch !a/b/c/
   (branches are permanent and global, did you want a bookmark?)
   $ hg ci -Aqm"5 bug"
 
   $ hg merge 4
   1 files updated, 0 files merged, 1 files removed, 0 files unresolved
   (branch merge, don't forget to commit)
   $ hg branch _a_b_c_
   marked working directory as branch _a_b_c_
   (branches are permanent and global, did you want a bookmark?)
   $ hg ci -Aqm"6 issue619"
 
   $ hg branch .a.b.c.
   marked working directory as branch .a.b.c.
   (branches are permanent and global, did you want a bookmark?)
   $ hg ci -Aqm7
 
   $ hg branch all
   marked working directory as branch all
   (branches are permanent and global, did you want a bookmark?)
 
   $ hg co 4
   0 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg branch é
   marked working directory as branch \xc3\xa9 (esc)
   (branches are permanent and global, did you want a bookmark?)
   $ hg ci -Aqm9
 
   $ hg tag -r6 1.0
 
   $ hg clone --quiet -U -r 7 . ../remote1
   $ hg clone --quiet -U -r 8 . ../remote2
   $ echo "[paths]" >> .hg/hgrc
   $ echo "default = ../remote1" >> .hg/hgrc
 
 names that should work without quoting
 
   $ try a
   ('symbol', 'a')
   0
   $ try b-a
   (minus
     ('symbol', 'b')
     ('symbol', 'a'))
   1
   $ try _a_b_c_
   ('symbol', '_a_b_c_')
   6
   $ try _a_b_c_-a
   (minus
     ('symbol', '_a_b_c_')
     ('symbol', 'a'))
   6
   $ try .a.b.c.
   ('symbol', '.a.b.c.')
   7
   $ try .a.b.c.-a
   (minus
     ('symbol', '.a.b.c.')
     ('symbol', 'a'))
   7
   $ try -- '-a-b-c-' # complains
   hg: parse error at 7: not a prefix: end
   [255]
   $ log -a-b-c- # succeeds with fallback
   4
   $ try -- -a-b-c--a # complains
   (minus
     (minus
       (minus
         (negate
           ('symbol', 'a'))
         ('symbol', 'b'))
       ('symbol', 'c'))
     (negate
       ('symbol', 'a')))
   abort: unknown revision '-a'!
   [255]
   $ try é
   ('symbol', '\xc3\xa9')
   9
 
 quoting needed
 
   $ try '"-a-b-c-"-a'
   (minus
     ('string', '-a-b-c-')
     ('symbol', 'a'))
   4
 
   $ log '1 or 2'
   1
   2
   $ log '1|2'
   1
   2
   $ log '1 and 2'
   $ log '1&2'
   $ try '1&2|3' # precedence - and is higher
   (or
     (and
       ('symbol', '1')
       ('symbol', '2'))
     ('symbol', '3'))
   3
   $ try '1|2&3'
   (or
     ('symbol', '1')
     (and
       ('symbol', '2')
       ('symbol', '3')))
   1
   $ try '1&2&3' # associativity
   (and
     (and
       ('symbol', '1')
       ('symbol', '2'))
     ('symbol', '3'))
   $ try '1|(2|3)'
   (or
     ('symbol', '1')
     (group
       (or
         ('symbol', '2')
         ('symbol', '3'))))
   1
   2
   3
   $ log '1.0' # tag
   6
   $ log 'a' # branch
   0
   $ log '2785f51ee'
   0
   $ log 'date(2005)'
   4
   $ log 'date(this is a test)'
   hg: parse error at 10: unexpected token: symbol
   [255]
   $ log 'date()'
   hg: parse error: date requires a string
   [255]
   $ log 'date'
   hg: parse error: can't use date here
   [255]
   $ log 'date('
   hg: parse error at 5: not a prefix: end
   [255]
   $ log 'date(tip)'
   abort: invalid date: 'tip'
   [255]
   $ log '"date"'
   abort: unknown revision 'date'!
   [255]
   $ log 'date(2005) and 1::'
   4
 
 ancestor can accept 0 or more arguments
 
   $ log 'ancestor()'
   $ log 'ancestor(1)'
   1
   $ log 'ancestor(4,5)'
   1
   $ log 'ancestor(4,5) and 4'
   $ log 'ancestor(0,0,1,3)'
   0
   $ log 'ancestor(3,1,5,3,5,1)'
   1
   $ log 'ancestor(0,1,3,5)'
   0
   $ log 'ancestor(1,2,3,4,5)'
   1
   $ log 'ancestors(5)'
   0
   1
   3
   5
   $ log 'ancestor(ancestors(5))'
   0
   $ log 'author(bob)'
   2
   $ log 'author("re:bob|test")'
   0
   1
   2
   3
   4
   5
   6
   7
   8
   9
   $ log 'branch(é)'
   8
   9
   $ log 'branch(a)'
   0
   $ hg log -r 'branch("re:a")' --template '{rev} {branch}\n'
   0 a
   2 a-b-c-
   3 +a+b+c+
   4 -a-b-c-
   5 !a/b/c/
   6 _a_b_c_
   7 .a.b.c.
   $ log 'children(ancestor(4,5))'
   2
   3
   $ log 'closed()'
   $ log 'contains(a)'
   0
   1
   3
   5
   $ log 'contains("../repo/a")'
   0
   1
   3
   5
   $ log 'desc(B)'
   5
   $ log 'descendants(2 or 3)'
   2
   3
   4
   5
   6
   7
   8
   9
   $ log 'file("b*")'
   1
   4
   $ log 'filelog("b")'
   1
   4
   $ log 'filelog("../repo/b")'
   1
   4
   $ log 'follow()'
   0
   1
   2
   4
   8
   9
   $ log 'grep("issue\d+")'
   6
   $ try 'grep("(")' # invalid regular expression
   (func
     ('symbol', 'grep')
     ('string', '('))
   hg: parse error: invalid match pattern: unbalanced parenthesis
   [255]
   $ try 'grep("\bissue\d+")'
   (func
     ('symbol', 'grep')
     ('string', '\x08issue\\d+'))
   $ try 'grep(r"\bissue\d+")'
   (func
     ('symbol', 'grep')
     ('string', '\\bissue\\d+'))
   6
   $ try 'grep(r"\")'
   hg: parse error at 7: unterminated string
   [255]
   $ log 'head()'
   0
   1
   2
   3
   4
   5
   6
   7
   9
   $ log 'heads(6::)'
   7
   $ log 'keyword(issue)'
   6
   $ log 'keyword("test a")'
   $ log 'limit(head(), 1)'
   0
   $ log 'matching(6)'
   6
   $ log 'matching(6:7, "phase parents user date branch summary files description substate")'
   6
   7
   $ log 'max(contains(a))'
   5
   $ log 'min(contains(a))'
   0
   $ log 'merge()'
   6
   $ log 'branchpoint()'
   1
   4
   $ log 'modifies(b)'
   4
   $ log 'modifies("path:b")'
   4
   $ log 'modifies("*")'
   4
   6
   $ log 'modifies("set:modified()")'
   4
   $ log 'id(5)'
   2
   $ log 'outgoing()'
   8
   9
   $ log 'outgoing("../remote1")'
   8
   9
   $ log 'outgoing("../remote2")'
   3
   5
   6
   7
   9
   $ log 'p1(merge())'
   5
   $ log 'p2(merge())'
   4
   $ log 'parents(merge())'
   4
   5
   $ log 'p1(branchpoint())'
   0
   2
   $ log 'p2(branchpoint())'
   $ log 'parents(branchpoint())'
   0
   2
   $ log 'removes(a)'
   2
   6
   $ log 'roots(all())'
   0
   $ log 'reverse(2 or 3 or 4 or 5)'
   5
   4
   3
   2
   $ log 'reverse(all())'
   9
   8
   7
   6
   5
   4
   3
   2
   1
   0
   $ log '1:: and reverse(all())'
   9
   8
   7
   6
   5
   4
   3
   2
   1
   $ log 'rev(5)'
   5
   $ log 'sort(limit(reverse(all()), 3))'
   7
   8
   9
   $ log 'sort(2 or 3 or 4 or 5, date)'
   2
   3
   5
   4
   $ log 'tagged()'
   6
   $ log 'tag()'
   6
   $ log 'tag(1.0)'
   6
   $ log 'tag(tip)'
   9
 
+check that conversion to _missingancestors works
+  $ try --optimize '::3 - ::1'
+  (minus
+    (dagrangepre
+      ('symbol', '3'))
+    (dagrangepre
+      ('symbol', '1')))
+  * optimized:
+  (func
+    ('symbol', '_missingancestors')
+    (list
+      ('symbol', '3')
+      ('symbol', '1')))
+  3
+  $ try --optimize 'ancestors(1) - ancestors(3)'
+  (minus
+    (func
+      ('symbol', 'ancestors')
+      ('symbol', '1'))
+    (func
+      ('symbol', 'ancestors')
+      ('symbol', '3')))
+  * optimized:
+  (func
+    ('symbol', '_missingancestors')
+    (list
+      ('symbol', '1')
+      ('symbol', '3')))
+  $ try --optimize 'not ::2 and ::6'
+  (and
+    (not
+      (dagrangepre
+        ('symbol', '2')))
+    (dagrangepre
+      ('symbol', '6')))
+  * optimized:
+  (func
+    ('symbol', '_missingancestors')
+    (list
+      ('symbol', '6')
+      ('symbol', '2')))
+  3
+  4
+  5
+  6
+  $ try --optimize 'ancestors(6) and not ancestors(4)'
+  (and
+    (func
+      ('symbol', 'ancestors')
+      ('symbol', '6'))
+    (not
+      (func
+        ('symbol', 'ancestors')
+        ('symbol', '4'))))
+  * optimized:
+  (func
+    ('symbol', '_missingancestors')
+    (list
+      ('symbol', '6')
+      ('symbol', '4')))
+  3
+  5
+  6
+
 we can use patterns when searching for tags
 
   $ log 'tag("1..*")'
   abort: tag '1..*' does not exist
   [255]
   $ log 'tag("re:1..*")'
   6
   $ log 'tag("re:[0-9].[0-9]")'
   6
   $ log 'tag("literal:1.0")'
   6
   $ log 'tag("re:0..*")'
 
   $ log 'tag(unknown)'
   abort: tag 'unknown' does not exist
   [255]
   $ log 'branch(unknown)'
   abort: unknown revision 'unknown'!
   [255]
   $ log 'user(bob)'
   2
 
   $ log '4::8'
   4
   8
   $ log '4:8'
   4
   5
   6
   7
   8
 
   $ log 'sort(!merge() & (modifies(b) | user(bob) | keyword(bug) | keyword(issue) & 1::9), "-date")'
   4
   2
   5
 
   $ log 'not 0 and 0:2'
   1
   2
   $ log 'not 1 and 0:2'
   0
   2
   $ log 'not 2 and 0:2'
   0
   1
   $ log '(1 and 2)::'
   $ log '(1 and 2):'
   $ log '(1 and 2):3'
   $ log 'sort(head(), -rev)'
   9
   7
   6
   5
   4
   3
   2
   1
   0
   $ log '4::8 - 8'
   4
   $ log 'matching(1 or 2 or 3) and (2 or 3 or 1)'
   2
   3
   1
 
 issue2437
 
   $ log '3 and p1(5)'
   3
   $ log '4 and p2(6)'
   4
   $ log '1 and parents(:2)'
   1
   $ log '2 and children(1:)'
   2
   $ log 'roots(all()) or roots(all())'
   0
   $ hg debugrevspec 'roots(all()) or roots(all())'
   0
   $ log 'heads(branch(é)) or heads(branch(é))'
   9
   $ log 'ancestors(8) and (heads(branch("-a-b-c-")) or heads(branch(é)))'
   4
 
 issue2654: report a parse error if the revset was not completely parsed
 
   $ log '1 OR 2'
   hg: parse error at 2: invalid token
   [255]
 
 or operator should preserve ordering:
   $ log 'reverse(2::4) or tip'
   4
   2
   9
 
 parentrevspec
 
   $ log 'merge()^0'
   6
   $ log 'merge()^'
   5
   $ log 'merge()^1'
   5
   $ log 'merge()^2'
   4
   $ log 'merge()^^'
   3
   $ log 'merge()^1^'
   3
   $ log 'merge()^^^'
   1
 
   $ log 'merge()~0'
   6
   $ log 'merge()~1'
   5
   $ log 'merge()~2'
   3
   $ log 'merge()~2^1'
   1
   $ log 'merge()~3'
   1
 
   $ log '(-3:tip)^'
   4
   6
   8
 
   $ log 'tip^foo'
   hg: parse error: ^ expects a number 0, 1, or 2
   [255]
 
 aliases:
 
   $ echo '[revsetalias]' >> .hg/hgrc
   $ echo 'm = merge()' >> .hg/hgrc
   $ echo 'sincem = descendants(m)' >> .hg/hgrc
   $ echo 'd($1) = reverse(sort($1, date))' >> .hg/hgrc
   $ echo 'rs(ARG1, ARG2) = reverse(sort(ARG1, ARG2))' >> .hg/hgrc
   $ echo 'rs4(ARG1, ARGA, ARGB, ARG2) = reverse(sort(ARG1, ARG2))' >> .hg/hgrc
 
   $ try m
   ('symbol', 'm')
   (func
     ('symbol', 'merge')
     None)
   6
 
 test alias recursion
 
   $ try sincem
   ('symbol', 'sincem')
   (func
     ('symbol', 'descendants')
     (func
       ('symbol', 'merge')
       None))
   6
   7
 
 test infinite recursion
 
   $ echo 'recurse1 = recurse2' >> .hg/hgrc
   $ echo 'recurse2 = recurse1' >> .hg/hgrc
   $ try recurse1
   ('symbol', 'recurse1')
   hg: parse error: infinite expansion of revset alias "recurse1" detected
   [255]
 
   $ echo 'level1($1, $2) = $1 or $2' >> .hg/hgrc
   $ echo 'level2($1, $2) = level1($2, $1)' >> .hg/hgrc
   $ try "level2(level1(1, 2), 3)"
   (func
     ('symbol', 'level2')
     (list
       (func
         ('symbol', 'level1')
         (list
           ('symbol', '1')
           ('symbol', '2')))
       ('symbol', '3')))
   (or
     ('symbol', '3')
     (or
       ('symbol', '1')
       ('symbol', '2')))
   3
   1
   2
 
 test nesting and variable passing
 
   $ echo 'nested($1) = nested2($1)' >> .hg/hgrc
   $ echo 'nested2($1) = nested3($1)' >> .hg/hgrc
   $ echo 'nested3($1) = max($1)' >> .hg/hgrc
   $ try 'nested(2:5)'
   (func
     ('symbol', 'nested')
     (range
       ('symbol', '2')
       ('symbol', '5')))
   (func
     ('symbol', 'max')
     (range
       ('symbol', '2')
       ('symbol', '5')))
   5
 
 test variable isolation, variable placeholders are rewritten as string
 then parsed and matched again as string. Check they do not leak too
 far away.
 
   $ echo 'injectparamasstring = max("$1")' >> .hg/hgrc
   $ echo 'callinjection($1) = descendants(injectparamasstring)' >> .hg/hgrc
   $ try 'callinjection(2:5)'
   (func
     ('symbol', 'callinjection')
     (range
       ('symbol', '2')
       ('symbol', '5')))
   (func
     ('symbol', 'descendants')
     (func
       ('symbol', 'max')
       ('string', '$1')))
   abort: unknown revision '$1'!
   [255]
 
   $ echo 'injectparamasstring2 = max(_aliasarg("$1"))' >> .hg/hgrc
   $ echo 'callinjection2($1) = descendants(injectparamasstring2)' >> .hg/hgrc
   $ try 'callinjection2(2:5)'
   (func
     ('symbol', 'callinjection2')
     (range
       ('symbol', '2')
       ('symbol', '5')))
   hg: parse error: not a function: _aliasarg
   [255]
   >>> data = file('.hg/hgrc', 'rb').read()
   >>> file('.hg/hgrc', 'wb').write(data.replace('_aliasarg', ''))
 
   $ try 'd(2:5)'
   (func
     ('symbol', 'd')
     (range
       ('symbol', '2')
       ('symbol', '5')))
   (func
     ('symbol', 'reverse')
     (func
       ('symbol', 'sort')
       (list
         (range
           ('symbol', '2')
           ('symbol', '5'))
         ('symbol', 'date'))))
   4
   5
   3
   2
   $ try 'rs(2 or 3, date)'
   (func
     ('symbol', 'rs')
     (list
       (or
         ('symbol', '2')
         ('symbol', '3'))
       ('symbol', 'date')))
   (func
     ('symbol', 'reverse')
     (func
       ('symbol', 'sort')
       (list
         (or
           ('symbol', '2')
           ('symbol', '3'))
         ('symbol', 'date'))))
   3
   2
   $ try 'rs()'
   (func
     ('symbol', 'rs')
     None)
   hg: parse error: invalid number of arguments: 0
   [255]
   $ try 'rs(2)'
   (func
     ('symbol', 'rs')
     ('symbol', '2'))
   hg: parse error: invalid number of arguments: 1
   [255]
   $ try 'rs(2, data, 7)'
   (func
     ('symbol', 'rs')
     (list
       (list
         ('symbol', '2')
         ('symbol', 'data'))
       ('symbol', '7')))
   hg: parse error: invalid number of arguments: 3
   [255]
   $ try 'rs4(2 or 3, x, x, date)'
   (func
     ('symbol', 'rs4')
     (list
       (list
         (list
           (or
             ('symbol', '2')
             ('symbol', '3'))
           ('symbol', 'x'))
         ('symbol', 'x'))
       ('symbol', 'date')))
   (func
     ('symbol', 'reverse')
     (func
       ('symbol', 'sort')
       (list
         (or
           ('symbol', '2')
           ('symbol', '3'))
         ('symbol', 'date'))))
   3
   2
 
 issue2549 - correct optimizations
 
   $ log 'limit(1 or 2 or 3, 2) and not 2'
   1
   $ log 'max(1 or 2) and not 2'
   $ log 'min(1 or 2) and not 1'
   $ log 'last(1 or 2, 1) and not 2'
 
 test revsets started with 40-chars hash (issue3669)
 
   $ ISSUE3669_TIP=`hg tip --template '{node}'`
   $ hg log -r "${ISSUE3669_TIP}" --template '{rev}\n'
   9
   $ hg log -r "${ISSUE3669_TIP}^" --template '{rev}\n'
   8
 
 test or-ed indirect predicates (issue3775)
 
   $ log '6 or 6^1' | sort
   5
   6
   $ log '6^1 or 6' | sort
   5
   6
   $ log '4 or 4~1' | sort
   2
   4
   $ log '4~1 or 4' | sort
   2
   4
   $ log '(0 or 2):(4 or 6) or 0 or 6' | sort
   0
   1
   2
   3
   4
   5
   6
   $ log '0 or 6 or (0 or 2):(4 or 6)' | sort
   0
   1
   2
   3
   4
   5
   6
 
 tests for 'remote()' predicate:
 #.  (csets in remote) (id)            (remote)
 1.  less than local   current branch  "default"
 2.  same with local   specified       "default"
 3.  more than local   specified       specified
 
   $ hg clone --quiet -U . ../remote3
   $ cd ../remote3
   $ hg update -q 7
   $ echo r > r
   $ hg ci -Aqm 10
   $ log 'remote()'
   7
   $ log 'remote("a-b-c-")'
   2
   $ cd ../repo
   $ log 'remote(".a.b.c.", "../remote3")'
 
   $ cd ..
 
 test author/desc/keyword in problematic encoding
 # unicode: cp932:
 # u30A2    0x83 0x41(= 'A')
 # u30C2    0x83 0x61(= 'a')
 
   $ hg init problematicencoding
   $ cd problematicencoding
 
   $ python > setup.sh <<EOF
   > print u'''
   > echo a > text
   > hg add text
   > hg --encoding utf-8 commit -u '\u30A2' -m none
   > echo b > text
   > hg --encoding utf-8 commit -u '\u30C2' -m none
   > echo c > text
   > hg --encoding utf-8 commit -u none -m '\u30A2'
   > echo d > text
   > hg --encoding utf-8 commit -u none -m '\u30C2'
   > '''.encode('utf-8')
   > EOF
   $ sh < setup.sh
 
 test in problematic encoding
   $ python > test.sh <<EOF
   > print u'''
   > hg --encoding cp932 log --template '{rev}\\n' -r 'author(\u30A2)'
   > echo ====
   > hg --encoding cp932 log --template '{rev}\\n' -r 'author(\u30C2)'
   > echo ====
   > hg --encoding cp932 log --template '{rev}\\n' -r 'desc(\u30A2)'
   > echo ====
   > hg --encoding cp932 log --template '{rev}\\n' -r 'desc(\u30C2)'
   > echo ====
   > hg --encoding cp932 log --template '{rev}\\n' -r 'keyword(\u30A2)'
   > echo ====
   > hg --encoding cp932 log --template '{rev}\\n' -r 'keyword(\u30C2)'
   > '''.encode('cp932')
   > EOF
   $ sh < test.sh
   0
   ====
   1
   ====
   2
   ====
   3
   ====
   0
   2
   ====
   1
   3
 
   $ cd ..