revset: move order constants from revsetlang

Thanks to the recent refactor, the ordering rule is fully processed at
runtime.

mercurial/revset.py

@@ -1,2154 +1,2191 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.
 
 from __future__ import absolute_import
 
 import re
 
 from .i18n import _
 from . import (
     dagop,
     destutil,
     encoding,
     error,
     hbisect,
     match as matchmod,
     node,
     obsolete as obsmod,
     obsutil,
     pathutil,
     phases,
     registrar,
     repoview,
     revsetlang,
     scmutil,
     smartset,
     util,
 )
 
 # helpers for processing parsed tree
 getsymbol = revsetlang.getsymbol
 getstring = revsetlang.getstring
 getinteger = revsetlang.getinteger
 getboolean = revsetlang.getboolean
 getlist = revsetlang.getlist
 getrange = revsetlang.getrange
 getargs = revsetlang.getargs
 getargsdict = revsetlang.getargsdict
 
-# constants used as an argument of match() and matchany()
-anyorder = revsetlang.anyorder
-defineorder = revsetlang.defineorder
-followorder = revsetlang.followorder
-
 baseset = smartset.baseset
 generatorset = smartset.generatorset
 spanset = smartset.spanset
 fullreposet = smartset.fullreposet
 
+# Constants for ordering requirement, used in getset():
+#
+# If 'define', any nested functions and operations MAY change the ordering of
+# the entries in the set (but if changes the ordering, it MUST ALWAYS change
+# it). If 'follow', any nested functions and operations MUST take the ordering
+# specified by the first operand to the '&' operator.
+#
+# For instance,
+#
+#   X & (Y | Z)
+#   ^   ^^^^^^^
+#   |   follow
+#   define
+#
+# will be evaluated as 'or(y(x()), z(x()))', where 'x()' can change the order
+# of the entries in the set, but 'y()', 'z()' and 'or()' shouldn't.
+#
+# 'any' means the order doesn't matter. For instance,
+#
+#   (X & Y) | ancestors(Z)
+#        ^              ^
+#        any            any
+#
+# For 'X & Y', 'X' decides order so the order of 'Y' does not matter. For
+# 'ancestors(Z)', Z's order does not matter since 'ancestors' does not care
+# about the order of its argument.
+#
+# Currently, most revsets do not care about the order, so 'define' is
+# equivalent to 'follow' for them, and the resulting order is based on the
+# 'subset' parameter passed down to them:
+#
+#   m = revset.match(..., order=defineorder)
+#   m(repo, subset)
+#           ^^^^^^
+#      For most revsets, 'define' means using the order this subset provides
+#
+# There are a few revsets that always redefine the order if 'define' is
+# specified: 'sort(X)', 'reverse(X)', 'x:y'.
+anyorder = 'any'        # don't care the order, could be even random-shuffled
+defineorder = 'define'  # ALWAYS redefine, or ALWAYS follow the current order
+followorder = 'follow'  # MUST follow the current order
+
 # helpers
 
 def getset(repo, subset, x, order=defineorder):
     if not x:
         raise error.ParseError(_("missing argument"))
     return methods[x[0]](repo, subset, *x[1:], order=order)
 
 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, order):
     x = scmutil.intrev(repo[x])
     if (x in subset
         or x == node.nullrev and isinstance(subset, fullreposet)):
         return baseset([x])
     return baseset()
 
 def rangeset(repo, subset, x, y, order):
     m = getset(repo, fullreposet(repo), x)
     n = getset(repo, fullreposet(repo), y)
 
     if not m or not n:
         return baseset()
     return _makerangeset(repo, subset, m.first(), n.last(), order)
 
 def rangeall(repo, subset, x, order):
     assert x is None
     return _makerangeset(repo, subset, 0, len(repo) - 1, order)
 
 def rangepre(repo, subset, y, order):
     # ':y' can't be rewritten to '0:y' since '0' may be hidden
     n = getset(repo, fullreposet(repo), y)
     if not n:
         return baseset()
     return _makerangeset(repo, subset, 0, n.last(), order)
 
 def rangepost(repo, subset, x, order):
     m = getset(repo, fullreposet(repo), x)
     if not m:
         return baseset()
     return _makerangeset(repo, subset, m.first(), len(repo) - 1, order)
 
 def _makerangeset(repo, subset, m, n, order):
     if m == n:
         r = baseset([m])
     elif n == node.wdirrev:
         r = spanset(repo, m, len(repo)) + baseset([n])
     elif m == node.wdirrev:
         r = baseset([m]) + spanset(repo, len(repo) - 1, n - 1)
     elif m < n:
         r = spanset(repo, m, n + 1)
     else:
         r = spanset(repo, m, n - 1)
 
     if order == defineorder:
         return r & subset
     else:
         # carrying the sorting over when possible would be more efficient
         return subset & r
 
 def dagrange(repo, subset, x, y, order):
     r = fullreposet(repo)
     xs = dagop.reachableroots(repo, getset(repo, r, x), getset(repo, r, y),
                               includepath=True)
     return subset & xs
 
 def andset(repo, subset, x, y, order):
     if order == anyorder:
         yorder = anyorder
     else:
         yorder = followorder
     return getset(repo, getset(repo, subset, x, order), y, yorder)
 
 def flipandset(repo, subset, y, x, order):
     # 'flipand(y, x)' is equivalent to 'and(x, y)', but faster when y is small
     if order == anyorder:
         yorder = anyorder
     else:
         yorder = followorder
     return getset(repo, getset(repo, subset, y, yorder), x, order)
 
 def differenceset(repo, subset, x, y, order):
     return getset(repo, subset, x, order) - getset(repo, subset, y, anyorder)
 
 def _orsetlist(repo, subset, xs, order):
     assert xs
     if len(xs) == 1:
         return getset(repo, subset, xs[0], order)
     p = len(xs) // 2
     a = _orsetlist(repo, subset, xs[:p], order)
     b = _orsetlist(repo, subset, xs[p:], order)
     return a + b
 
 def orset(repo, subset, x, order):
     xs = getlist(x)
     if order == followorder:
         # slow path to take the subset order
         return subset & _orsetlist(repo, fullreposet(repo), xs, anyorder)
     else:
         return _orsetlist(repo, subset, xs, order)
 
 def notset(repo, subset, x, order):
     return subset - getset(repo, subset, x, anyorder)
 
 def relationset(repo, subset, x, y, order):
     raise error.ParseError(_("can't use a relation in this context"))
 
 def relsubscriptset(repo, subset, x, y, z, order):
     # this is pretty basic implementation of 'x#y[z]' operator, still
     # experimental so undocumented. see the wiki for further ideas.
     # https://www.mercurial-scm.org/wiki/RevsetOperatorPlan
     rel = getsymbol(y)
     n = getinteger(z, _("relation subscript must be an integer"))
 
     # TODO: perhaps this should be a table of relation functions
     if rel in ('g', 'generations'):
         # TODO: support range, rewrite tests, and drop startdepth argument
         # from ancestors() and descendants() predicates
         if n <= 0:
             n = -n
             return _ancestors(repo, subset, x, startdepth=n, stopdepth=n + 1)
         else:
             return _descendants(repo, subset, x, startdepth=n, stopdepth=n + 1)
 
     raise error.UnknownIdentifier(rel, ['generations'])
 
 def subscriptset(repo, subset, x, y, order):
     raise error.ParseError(_("can't use a subscript in this context"))
 
 def listset(repo, subset, *xs, **opts):
     raise error.ParseError(_("can't use a list in this context"),
                            hint=_('see hg help "revsets.x or y"'))
 
 def keyvaluepair(repo, subset, k, v, order):
     raise error.ParseError(_("can't use a key-value pair in this context"))
 
 def func(repo, subset, a, b, order):
     f = getsymbol(a)
     if f in symbols:
         func = symbols[f]
         if getattr(func, '_takeorder', False):
             return func(repo, subset, b, order)
         return func(repo, subset, b)
 
     keep = lambda fn: getattr(fn, '__doc__', None) is not None
 
     syms = [s for (s, fn) in symbols.items() if keep(fn)]
     raise error.UnknownIdentifier(f, syms)
 
 # functions
 
 # symbols are callables like:
 #   fn(repo, subset, x)
 # with:
 #   repo - current repository instance
 #   subset - of revisions to be examined
 #   x - argument in tree form
 symbols = {}
 
 # 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()
 
 predicate = registrar.revsetpredicate()
 
 @predicate('_destupdate')
 def _destupdate(repo, subset, x):
     # experimental revset for update destination
     args = getargsdict(x, 'limit', 'clean')
     return subset & baseset([destutil.destupdate(repo, **args)[0]])
 
 @predicate('_destmerge')
 def _destmerge(repo, subset, x):
     # experimental revset for merge destination
     sourceset = None
     if x is not None:
         sourceset = getset(repo, fullreposet(repo), x)
     return subset & baseset([destutil.destmerge(repo, sourceset=sourceset)])
 
 @predicate('adds(pattern)', safe=True)
 def adds(repo, subset, x):
     """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)
 
 @predicate('ancestor(*changeset)', safe=True)
 def ancestor(repo, subset, x):
     """A 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 = fullreposet(repo)
     anc = None
 
     # (getset(repo, rl, i) for i in l) generates a list of lists
     for revs in (getset(repo, rl, i) for i in l):
         for r in revs:
             if anc is None:
                 anc = repo[r]
             else:
                 anc = anc.ancestor(repo[r])
 
     if anc is not None and anc.rev() in subset:
         return baseset([anc.rev()])
     return baseset()
 
 def _ancestors(repo, subset, x, followfirst=False, startdepth=None,
                stopdepth=None):
     heads = getset(repo, fullreposet(repo), x)
     if not heads:
         return baseset()
     s = dagop.revancestors(repo, heads, followfirst, startdepth, stopdepth)
     return subset & s
 
 @predicate('ancestors(set[, depth])', safe=True)
 def ancestors(repo, subset, x):
     """Changesets that are ancestors of changesets in set, including the
     given changesets themselves.
 
     If depth is specified, the result only includes changesets up to
     the specified generation.
     """
     # startdepth is for internal use only until we can decide the UI
     args = getargsdict(x, 'ancestors', 'set depth startdepth')
     if 'set' not in args:
         # i18n: "ancestors" is a keyword
         raise error.ParseError(_('ancestors takes at least 1 argument'))
     startdepth = stopdepth = None
     if 'startdepth' in args:
         n = getinteger(args['startdepth'],
                        "ancestors expects an integer startdepth")
         if n < 0:
             raise error.ParseError("negative startdepth")
         startdepth = n
     if 'depth' in args:
         # i18n: "ancestors" is a keyword
         n = getinteger(args['depth'], _("ancestors expects an integer depth"))
         if n < 0:
             raise error.ParseError(_("negative depth"))
         stopdepth = n + 1
     return _ancestors(repo, subset, args['set'],
                       startdepth=startdepth, stopdepth=stopdepth)
 
 @predicate('_firstancestors', safe=True)
 def _firstancestors(repo, subset, x):
     # ``_firstancestors(set)``
     # Like ``ancestors(set)`` but follows only the first parents.
     return _ancestors(repo, subset, x, followfirst=True)
 
 def _childrenspec(repo, subset, x, n, order):
     """Changesets that are the Nth child of a changeset
     in set.
     """
     cs = set()
     for r in getset(repo, fullreposet(repo), x):
         for i in range(n):
             c = repo[r].children()
             if len(c) == 0:
                 break
             if len(c) > 1:
                 raise error.RepoLookupError(
                     _("revision in set has more than one child"))
             r = c[0].rev()
         else:
             cs.add(r)
     return subset & cs
 
 def ancestorspec(repo, subset, x, n, order):
     """``set~n``
     Changesets that are the Nth ancestor (first parents only) of a changeset
     in set.
     """
     n = getinteger(n, _("~ expects a number"))
     if n < 0:
         # children lookup
         return _childrenspec(repo, subset, x, -n, order)
     ps = set()
     cl = repo.changelog
     for r in getset(repo, fullreposet(repo), x):
         for i in range(n):
             try:
                 r = cl.parentrevs(r)[0]
             except error.WdirUnsupported:
                 r = repo[r].parents()[0].rev()
         ps.add(r)
     return subset & ps
 
 @predicate('author(string)', safe=True)
 def author(repo, subset, x):
     """Alias for ``user(string)``.
     """
     # i18n: "author" is a keyword
     n = getstring(x, _("author requires a string"))
     kind, pattern, matcher = _substringmatcher(n, casesensitive=False)
     return subset.filter(lambda x: matcher(repo[x].user()),
                          condrepr=('<user %r>', n))
 
 @predicate('bisect(string)', safe=True)
 def bisect(repo, subset, x):
     """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 subset & state
 
 # Backward-compatibility
 # - no help entry so that we do not advertise it any more
 @predicate('bisected', safe=True)
 def bisected(repo, subset, x):
     return bisect(repo, subset, x)
 
 @predicate('bookmark([name])', safe=True)
 def bookmark(repo, subset, x):
     """The named bookmark or all bookmarks.
 
     Pattern matching is supported for `name`. See :hg:`help revisions.patterns`.
     """
     # 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 = util.stringmatcher(bm)
         bms = set()
         if kind == 'literal':
             bmrev = repo._bookmarks.get(pattern, None)
             if not bmrev:
                 raise error.RepoLookupError(_("bookmark '%s' does not exist")
                                             % pattern)
             bms.add(repo[bmrev].rev())
         else:
             matchrevs = set()
             for name, bmrev in repo._bookmarks.iteritems():
                 if matcher(name):
                     matchrevs.add(bmrev)
             if not matchrevs:
                 raise error.RepoLookupError(_("no bookmarks exist"
                                               " that match '%s'") % pattern)
             for bmrev in matchrevs:
                 bms.add(repo[bmrev].rev())
     else:
         bms = {repo[r].rev() for r in repo._bookmarks.values()}
     bms -= {node.nullrev}
     return subset & bms
 
 @predicate('branch(string or set)', safe=True)
 def branch(repo, subset, x):
     """
     All changesets belonging to the given branch or the branches of the given
     changesets.
 
     Pattern matching is supported for `string`. See
     :hg:`help revisions.patterns`.
     """
     getbi = repo.revbranchcache().branchinfo
     def getbranch(r):
         try:
             return getbi(r)[0]
         except error.WdirUnsupported:
             return repo[r].branch()
 
     try:
         b = getstring(x, '')
     except error.ParseError:
         # not a string, but another revspec, e.g. tip()
         pass
     else:
         kind, pattern, matcher = util.stringmatcher(b)
         if kind == 'literal':
             # note: falls through to the revspec case if no branch with
             # this name exists and pattern kind is not specified explicitly
             if pattern in repo.branchmap():
                 return subset.filter(lambda r: matcher(getbranch(r)),
                                      condrepr=('<branch %r>', b))
             if b.startswith('literal:'):
                 raise error.RepoLookupError(_("branch '%s' does not exist")
                                             % pattern)
         else:
             return subset.filter(lambda r: matcher(getbranch(r)),
                                  condrepr=('<branch %r>', b))
 
     s = getset(repo, fullreposet(repo), x)
     b = set()
     for r in s:
         b.add(getbranch(r))
     c = s.__contains__
     return subset.filter(lambda r: c(r) or getbranch(r) in b,
                          condrepr=lambda: '<branch %r>' % sorted(b))
 
 @predicate('bumped()', safe=True)
 def bumped(repo, subset, x):
     msg = ("'bumped()' is deprecated, "
            "use 'phasedivergent()'")
     repo.ui.deprecwarn(msg, '4.4')
 
     return phasedivergent(repo, subset, x)
 
 @predicate('phasedivergent()', safe=True)
 def phasedivergent(repo, subset, x):
     """Mutable changesets marked as successors of public changesets.
 
     Only non-public and non-obsolete changesets can be `phasedivergent`.
     (EXPERIMENTAL)
     """
     # i18n: "phasedivergent" is a keyword
     getargs(x, 0, 0, _("phasedivergent takes no arguments"))
     phasedivergent = obsmod.getrevs(repo, 'phasedivergent')
     return subset & phasedivergent
 
 @predicate('bundle()', safe=True)
 def bundle(repo, subset, x):
     """Changesets in the bundle.
 
     Bundle must be specified by the -R option."""
 
     try:
         bundlerevs = repo.changelog.bundlerevs
     except AttributeError:
         raise error.Abort(_("no bundle provided - specify with -R"))
     return subset & bundlerevs
 
 def checkstatus(repo, subset, pat, field):
     hasset = matchmod.patkind(pat) == 'set'
 
     mcache = [None]
     def matches(x):
         c = repo[x]
         if not mcache[0] or hasset:
             mcache[0] = matchmod.match(repo.root, repo.getcwd(), [pat], ctx=c)
         m = mcache[0]
         fname = None
         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 subset.filter(matches, condrepr=('<status[%r] %r>', field, pat))
 
 def _children(repo, subset, parentset):
     if not parentset:
         return baseset()
     cs = set()
     pr = repo.changelog.parentrevs
     minrev = parentset.min()
     nullrev = node.nullrev
     for r in subset:
         if r <= minrev:
             continue
         p1, p2 = pr(r)
         if p1 in parentset:
             cs.add(r)
         if p2 != nullrev and p2 in parentset:
             cs.add(r)
     return baseset(cs)
 
 @predicate('children(set)', safe=True)
 def children(repo, subset, x):
     """Child changesets of changesets in set.
     """
     s = getset(repo, fullreposet(repo), x)
     cs = _children(repo, subset, s)
     return subset & cs
 
 @predicate('closed()', safe=True)
 def closed(repo, subset, x):
     """Changeset is closed.
     """
     # i18n: "closed" is a keyword
     getargs(x, 0, 0, _("closed takes no arguments"))
     return subset.filter(lambda r: repo[r].closesbranch(),
                          condrepr='<branch closed>')
 
 @predicate('contains(pattern)')
 def contains(repo, subset, x):
     """The revision's manifest contains a file matching pattern (but might not
     modify it). 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 subset.filter(matches, condrepr=('<contains %r>', pat))
 
 @predicate('converted([id])', safe=True)
 def converted(repo, subset, x):
     """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 subset.filter(lambda r: _matchvalue(r),
                          condrepr=('<converted %r>', rev))
 
 @predicate('date(interval)', safe=True)
 def date(repo, subset, x):
     """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 subset.filter(lambda x: dm(repo[x].date()[0]),
                          condrepr=('<date %r>', ds))
 
 @predicate('desc(string)', safe=True)
 def desc(repo, subset, x):
     """Search commit message for string. The match is case-insensitive.
 
     Pattern matching is supported for `string`. See
     :hg:`help revisions.patterns`.
     """
     # i18n: "desc" is a keyword
     ds = getstring(x, _("desc requires a string"))
 
     kind, pattern, matcher = _substringmatcher(ds, casesensitive=False)
 
     return subset.filter(lambda r: matcher(repo[r].description()),
                          condrepr=('<desc %r>', ds))
 
 def _descendants(repo, subset, x, followfirst=False, startdepth=None,
                  stopdepth=None):
     roots = getset(repo, fullreposet(repo), x)
     if not roots:
         return baseset()
     s = dagop.revdescendants(repo, roots, followfirst, startdepth, stopdepth)
     return subset & s
 
 @predicate('descendants(set[, depth])', safe=True)
 def descendants(repo, subset, x):
     """Changesets which are descendants of changesets in set, including the
     given changesets themselves.
 
     If depth is specified, the result only includes changesets up to
     the specified generation.
     """
     # startdepth is for internal use only until we can decide the UI
     args = getargsdict(x, 'descendants', 'set depth startdepth')
     if 'set' not in args:
         # i18n: "descendants" is a keyword
         raise error.ParseError(_('descendants takes at least 1 argument'))
     startdepth = stopdepth = None
     if 'startdepth' in args:
         n = getinteger(args['startdepth'],
                        "descendants expects an integer startdepth")
         if n < 0:
             raise error.ParseError("negative startdepth")
         startdepth = n
     if 'depth' in args:
         # i18n: "descendants" is a keyword
         n = getinteger(args['depth'], _("descendants expects an integer depth"))
         if n < 0:
             raise error.ParseError(_("negative depth"))
         stopdepth = n + 1
     return _descendants(repo, subset, args['set'],
                         startdepth=startdepth, stopdepth=stopdepth)
 
 @predicate('_firstdescendants', safe=True)
 def _firstdescendants(repo, subset, x):
     # ``_firstdescendants(set)``
     # Like ``descendants(set)`` but follows only the first parents.
     return _descendants(repo, subset, x, followfirst=True)
 
 @predicate('destination([set])', safe=True)
 def destination(repo, subset, x):
     """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:
         sources = getset(repo, fullreposet(repo), x)
     else:
         sources = fullreposet(repo)
 
     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 arg set.
     # Even if the immediate src of r is not in the arg set, 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 sources or src in dests:
                 dests.update(lineage)
                 break
 
             r = src
             src = _getrevsource(repo, r)
 
     return subset.filter(dests.__contains__,
                          condrepr=lambda: '<destination %r>' % sorted(dests))
 
 @predicate('divergent()', safe=True)
 def divergent(repo, subset, x):
     msg = ("'divergent()' is deprecated, "
            "use 'contentdivergent()'")
     repo.ui.deprecwarn(msg, '4.4')
 
     return contentdivergent(repo, subset, x)
 
 @predicate('contentdivergent()', safe=True)
 def contentdivergent(repo, subset, x):
     """
     Final successors of changesets with an alternative set of final
     successors. (EXPERIMENTAL)
     """
     # i18n: "contentdivergent" is a keyword
     getargs(x, 0, 0, _("contentdivergent takes no arguments"))
     contentdivergent = obsmod.getrevs(repo, 'contentdivergent')
     return subset & contentdivergent
 
 @predicate('extinct()', safe=True)
 def extinct(repo, subset, x):
     """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
 
 @predicate('extra(label, [value])', safe=True)
 def extra(repo, subset, x):
     """Changesets with the given label in the extra metadata, with the given
     optional value.
 
     Pattern matching is supported for `value`. See
     :hg:`help revisions.patterns`.
     """
     args = getargsdict(x, 'extra', 'label value')
     if 'label' not in args:
         # i18n: "extra" is a keyword
         raise error.ParseError(_('extra takes at least 1 argument'))
     # i18n: "extra" is a keyword
     label = getstring(args['label'], _('first argument to extra must be '
                                        'a string'))
     value = None
 
     if 'value' in args:
         # i18n: "extra" is a keyword
         value = getstring(args['value'], _('second argument to extra must be '
                                            'a string'))
         kind, value, matcher = util.stringmatcher(value)
 
     def _matchvalue(r):
         extra = repo[r].extra()
         return label in extra and (value is None or matcher(extra[label]))
 
     return subset.filter(lambda r: _matchvalue(r),
                          condrepr=('<extra[%r] %r>', label, value))
 
 @predicate('filelog(pattern)', safe=True)
 def filelog(repo, subset, x):
     """Changesets connected to the specified filelog.
 
     For performance reasons, visits only revisions mentioned in the file-level
     filelog, rather than filtering through all changesets (much faster, but
     doesn't include deletes or duplicate changes). 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.
 
     If some linkrev points to revisions filtered by the current repoview, we'll
     work around it to return a non-filtered value.
     """
 
     # i18n: "filelog" is a keyword
     pat = getstring(x, _("filelog requires a pattern"))
     s = set()
     cl = repo.changelog
 
     if not matchmod.patkind(pat):
         f = pathutil.canonpath(repo.root, repo.getcwd(), pat)
         files = [f]
     else:
         m = matchmod.match(repo.root, repo.getcwd(), [pat], ctx=repo[None])
         files = (f for f in repo[None] if m(f))
 
     for f in files:
         fl = repo.file(f)
         known = {}
         scanpos = 0
         for fr in list(fl):
             fn = fl.node(fr)
             if fn in known:
                 s.add(known[fn])
                 continue
 
             lr = fl.linkrev(fr)
             if lr in cl:
                 s.add(lr)
             elif scanpos is not None:
                 # lowest matching changeset is filtered, scan further
                 # ahead in changelog
                 start = max(lr, scanpos) + 1
                 scanpos = None
                 for r in cl.revs(start):
                     # minimize parsing of non-matching entries
                     if f in cl.revision(r) and f in cl.readfiles(r):
                         try:
                             # try to use manifest delta fastpath
                             n = repo[r].filenode(f)
                             if n not in known:
                                 if n == fn:
                                     s.add(r)
                                     scanpos = r
                                     break
                                 else:
                                     known[n] = r
                         except error.ManifestLookupError:
                             # deletion in changelog
                             continue
 
     return subset & s
 
 @predicate('first(set, [n])', safe=True, takeorder=True)
 def first(repo, subset, x, order):
     """An alias for limit().
     """
     return limit(repo, subset, x, order)
 
 def _follow(repo, subset, x, name, followfirst=False):
     l = getargs(x, 0, 2, _("%s takes no arguments or a pattern "
                            "and an optional revset") % name)
     c = repo['.']
     if l:
         x = getstring(l[0], _("%s expected a pattern") % name)
         rev = None
         if len(l) >= 2:
             revs = getset(repo, fullreposet(repo), l[1])
             if len(revs) != 1:
                 raise error.RepoLookupError(
                         _("%s expected one starting revision") % name)
             rev = revs.last()
             c = repo[rev]
         matcher = matchmod.match(repo.root, repo.getcwd(), [x],
                                  ctx=repo[rev], default='path')
 
         files = c.manifest().walk(matcher)
 
         s = set()
         for fname in files:
             fctx = c[fname]
             s = s.union(set(c.rev() for c in fctx.ancestors(followfirst)))
             # include the revision responsible for the most recent version
             s.add(fctx.introrev())
     else:
         s = dagop.revancestors(repo, baseset([c.rev()]), followfirst)
 
     return subset & s
 
 @predicate('follow([pattern[, startrev]])', safe=True)
 def follow(repo, subset, x):
     """
     An alias for ``::.`` (ancestors of the working directory's first parent).
     If pattern is specified, the histories of files matching given
     pattern in the revision given by startrev are followed, including copies.
     """
     return _follow(repo, subset, x, 'follow')
 
 @predicate('_followfirst', safe=True)
 def _followfirst(repo, subset, x):
     # ``followfirst([pattern[, startrev]])``
     # Like ``follow([pattern[, startrev]])`` but follows only the first parent
     # of every revisions or files revisions.
     return _follow(repo, subset, x, '_followfirst', followfirst=True)
 
 @predicate('followlines(file, fromline:toline[, startrev=., descend=False])',
            safe=True)
 def followlines(repo, subset, x):
     """Changesets modifying `file` in line range ('fromline', 'toline').
 
     Line range corresponds to 'file' content at 'startrev' and should hence be
     consistent with file size. If startrev is not specified, working directory's
     parent is used.
 
     By default, ancestors of 'startrev' are returned. If 'descend' is True,
     descendants of 'startrev' are returned though renames are (currently) not
     followed in this direction.
     """
     args = getargsdict(x, 'followlines', 'file *lines startrev descend')
     if len(args['lines']) != 1:
         raise error.ParseError(_("followlines requires a line range"))
 
     rev = '.'
     if 'startrev' in args:
         revs = getset(repo, fullreposet(repo), args['startrev'])
         if len(revs) != 1:
             raise error.ParseError(
                 # i18n: "followlines" is a keyword
                 _("followlines expects exactly one revision"))
         rev = revs.last()
 
     pat = getstring(args['file'], _("followlines requires a pattern"))
     if not matchmod.patkind(pat):
         fname = pathutil.canonpath(repo.root, repo.getcwd(), pat)
     else:
         m = matchmod.match(repo.root, repo.getcwd(), [pat], ctx=repo[rev])
         files = [f for f in repo[rev] if m(f)]
         if len(files) != 1:
             # i18n: "followlines" is a keyword
             raise error.ParseError(_("followlines expects exactly one file"))
         fname = files[0]
 
     # i18n: "followlines" is a keyword
     lr = getrange(args['lines'][0], _("followlines expects a line range"))
     fromline, toline = [getinteger(a, _("line range bounds must be integers"))
                         for a in lr]
     fromline, toline = util.processlinerange(fromline, toline)
 
     fctx = repo[rev].filectx(fname)
     descend = False
     if 'descend' in args:
         descend = getboolean(args['descend'],
                              # i18n: "descend" is a keyword
                              _("descend argument must be a boolean"))
     if descend:
         rs = generatorset(
             (c.rev() for c, _linerange
              in dagop.blockdescendants(fctx, fromline, toline)),
             iterasc=True)
     else:
         rs = generatorset(
             (c.rev() for c, _linerange
              in dagop.blockancestors(fctx, fromline, toline)),
             iterasc=False)
     return subset & rs
 
 @predicate('all()', safe=True)
 def getall(repo, subset, x):
     """All changesets, the same as ``0:tip``.
     """
     # i18n: "all" is a keyword
     getargs(x, 0, 0, _("all takes no arguments"))
     return subset & spanset(repo)  # drop "null" if any
 
 @predicate('grep(regex)')
 def grep(repo, subset, x):
     """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 as 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 subset.filter(matches, condrepr=('<grep %r>', gr.pattern))
 
 @predicate('_matchfiles', safe=True)
 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.
 
     l = getargs(x, 1, -1, "_matchfiles requires at least one argument")
     pats, inc, exc = [], [], []
     rev, default = None, None
     for arg in l:
         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:
                 raise error.ParseError('_matchfiles expected at most one '
                                        'revision')
             if value != '': # empty means working directory; leave rev as None
                 rev = value
         elif prefix == 'd:':
             if default is not None:
                 raise error.ParseError('_matchfiles expected at most one '
                                        'default mode')
             default = value
         else:
             raise error.ParseError('invalid _matchfiles prefix: %s' % prefix)
     if not default:
         default = 'glob'
 
     m = matchmod.match(repo.root, repo.getcwd(), pats, include=inc,
                        exclude=exc, ctx=repo[rev], default=default)
 
     # This directly read the changelog data as creating changectx for all
     # revisions is quite expensive.
     getfiles = repo.changelog.readfiles
     wdirrev = node.wdirrev
     def matches(x):
         if x == wdirrev:
             files = repo[x].files()
         else:
             files = getfiles(x)
         for f in files:
             if m(f):
                 return True
         return False
 
     return subset.filter(matches,
                          condrepr=('<matchfiles patterns=%r, include=%r '
                                    'exclude=%r, default=%r, rev=%r>',
                                    pats, inc, exc, default, rev))
 
 @predicate('file(pattern)', safe=True)
 def hasfile(repo, subset, x):
     """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))
 
 @predicate('head()', safe=True)
 def head(repo, subset, x):
     """Changeset is a named branch head.
     """
     # i18n: "head" is a keyword
     getargs(x, 0, 0, _("head takes no arguments"))
     hs = set()
     cl = repo.changelog
     for ls in repo.branchmap().itervalues():
         hs.update(cl.rev(h) for h in ls)
     return subset & baseset(hs)
 
 @predicate('heads(set)', safe=True)
 def heads(repo, subset, x):
     """Members of set with no children in set.
     """
     s = getset(repo, subset, x)
     ps = parents(repo, subset, x)
     return s - ps
 
 @predicate('hidden()', safe=True)
 def hidden(repo, subset, x):
     """Hidden changesets.
     """
     # i18n: "hidden" is a keyword
     getargs(x, 0, 0, _("hidden takes no arguments"))
     hiddenrevs = repoview.filterrevs(repo, 'visible')
     return subset & hiddenrevs
 
 @predicate('keyword(string)', safe=True)
 def keyword(repo, subset, x):
     """Search commit message, user name, and names of changed files for
     string. The match is case-insensitive.
 
     For a regular expression or case sensitive search of these fields, use
     ``grep(regex)``.
     """
     # i18n: "keyword" is a keyword
     kw = encoding.lower(getstring(x, _("keyword requires a string")))
 
     def matches(r):
         c = repo[r]
         return any(kw in encoding.lower(t)
                    for t in c.files() + [c.user(), c.description()])
 
     return subset.filter(matches, condrepr=('<keyword %r>', kw))
 
 @predicate('limit(set[, n[, offset]])', safe=True, takeorder=True)
 def limit(repo, subset, x, order):
     """First n members of set, defaulting to 1, starting from offset.
     """
     args = getargsdict(x, 'limit', 'set n offset')
     if 'set' not in args:
         # i18n: "limit" is a keyword
         raise error.ParseError(_("limit requires one to three arguments"))
     # i18n: "limit" is a keyword
     lim = getinteger(args.get('n'), _("limit expects a number"), default=1)
     if lim < 0:
         raise error.ParseError(_("negative number to select"))
     # i18n: "limit" is a keyword
     ofs = getinteger(args.get('offset'), _("limit expects a number"), default=0)
     if ofs < 0:
         raise error.ParseError(_("negative offset"))
     os = getset(repo, fullreposet(repo), args['set'])
     ls = os.slice(ofs, ofs + lim)
     if order == followorder and lim > 1:
         return subset & ls
     return ls & subset
 
 @predicate('last(set, [n])', safe=True, takeorder=True)
 def last(repo, subset, x, order):
     """Last n members of set, defaulting to 1.
     """
     # i18n: "last" is a keyword
     l = getargs(x, 1, 2, _("last requires one or two arguments"))
     lim = 1
     if len(l) == 2:
         # i18n: "last" is a keyword
         lim = getinteger(l[1], _("last expects a number"))
     if lim < 0:
         raise error.ParseError(_("negative number to select"))
     os = getset(repo, fullreposet(repo), l[0])
     os.reverse()
     ls = os.slice(0, lim)
     if order == followorder and lim > 1:
         return subset & ls
     ls.reverse()
     return ls & subset
 
 @predicate('max(set)', safe=True)
 def maxrev(repo, subset, x):
     """Changeset with highest revision number in set.
     """
     os = getset(repo, fullreposet(repo), x)
     try:
         m = os.max()
         if m in subset:
             return baseset([m], datarepr=('<max %r, %r>', subset, os))
     except ValueError:
         # os.max() throws a ValueError when the collection is empty.
         # Same as python's max().
         pass
     return baseset(datarepr=('<max %r, %r>', subset, os))
 
 @predicate('merge()', safe=True)
 def merge(repo, subset, x):
     """Changeset is a merge changeset.
     """
     # i18n: "merge" is a keyword
     getargs(x, 0, 0, _("merge takes no arguments"))
     cl = repo.changelog
     return subset.filter(lambda r: cl.parentrevs(r)[1] != -1,
                          condrepr='<merge>')
 
 @predicate('branchpoint()', safe=True)
 def branchpoint(repo, subset, x):
     """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()
     # XXX this should be 'parentset.min()' assuming 'parentset' is a smartset
     # (and if it is not, it should.)
     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 subset.filter(lambda r: parentscount[r - baserev] > 1,
                          condrepr='<branchpoint>')
 
 @predicate('min(set)', safe=True)
 def minrev(repo, subset, x):
     """Changeset with lowest revision number in set.
     """
     os = getset(repo, fullreposet(repo), x)
     try:
         m = os.min()
         if m in subset:
             return baseset([m], datarepr=('<min %r, %r>', subset, os))
     except ValueError:
         # os.min() throws a ValueError when the collection is empty.
         # Same as python's min().
         pass
     return baseset(datarepr=('<min %r, %r>', subset, os))
 
 @predicate('modifies(pattern)', safe=True)
 def modifies(repo, subset, x):
     """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)
 
 @predicate('named(namespace)')
 def named(repo, subset, x):
     """The changesets in a given namespace.
 
     Pattern matching is supported for `namespace`. See
     :hg:`help revisions.patterns`.
     """
     # i18n: "named" is a keyword
     args = getargs(x, 1, 1, _('named requires a namespace argument'))
 
     ns = getstring(args[0],
                    # i18n: "named" is a keyword
                    _('the argument to named must be a string'))
     kind, pattern, matcher = util.stringmatcher(ns)
     namespaces = set()
     if kind == 'literal':
         if pattern not in repo.names:
             raise error.RepoLookupError(_("namespace '%s' does not exist")
                                         % ns)
         namespaces.add(repo.names[pattern])
     else:
         for name, ns in repo.names.iteritems():
             if matcher(name):
                 namespaces.add(ns)
         if not namespaces:
             raise error.RepoLookupError(_("no namespace exists"
                                           " that match '%s'") % pattern)
 
     names = set()
     for ns in namespaces:
         for name in ns.listnames(repo):
             if name not in ns.deprecated:
                 names.update(repo[n].rev() for n in ns.nodes(repo, name))
 
     names -= {node.nullrev}
     return subset & names
 
 @predicate('id(string)', safe=True)
 def node_(repo, subset, x):
     """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:
         try:
             rn = repo.changelog.rev(node.bin(n))
         except error.WdirUnsupported:
             rn = node.wdirrev
         except (LookupError, TypeError):
             rn = None
     else:
         rn = None
         try:
             pm = repo.changelog._partialmatch(n)
             if pm is not None:
                 rn = repo.changelog.rev(pm)
         except error.WdirUnsupported:
             rn = node.wdirrev
 
     if rn is None:
         return baseset()
     result = baseset([rn])
     return result & subset
 
 @predicate('obsolete()', safe=True)
 def obsolete(repo, subset, x):
     """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
 
 @predicate('only(set, [set])', safe=True)
 def only(repo, subset, x):
     """Changesets that are ancestors of the first set that are not ancestors
     of any other head in the repo. If a second set is specified, the result
     is ancestors of the first set that are not ancestors of the second set
     (i.e. ::<set1> - ::<set2>).
     """
     cl = repo.changelog
     # i18n: "only" is a keyword
     args = getargs(x, 1, 2, _('only takes one or two arguments'))
     include = getset(repo, fullreposet(repo), args[0])
     if len(args) == 1:
         if not include:
             return baseset()
 
         descendants = set(dagop.revdescendants(repo, include, False))
         exclude = [rev for rev in cl.headrevs()
             if not rev in descendants and not rev in include]
     else:
         exclude = getset(repo, fullreposet(repo), args[1])
 
     results = set(cl.findmissingrevs(common=exclude, heads=include))
     # XXX we should turn this into a baseset instead of a set, smartset may do
     # some optimizations from the fact this is a baseset.
     return subset & results
 
 @predicate('origin([set])', safe=True)
 def origin(repo, subset, x):
     """
     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:
         dests = getset(repo, fullreposet(repo), x)
     else:
         dests = fullreposet(repo)
 
     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 = {_firstsrc(r) for r in dests}
     o -= {None}
     # XXX we should turn this into a baseset instead of a set, smartset may do
     # some optimizations from the fact this is a baseset.
     return subset & o
 
 @predicate('outgoing([path])', safe=False)
 def outgoing(repo, subset, x):
     """Changesets not found in the specified destination repository, or the
     default push location.
     """
     # Avoid cycles.
     from . import (
         discovery,
         hg,
     )
     # 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 = {cl.rev(r) for r in outgoing.missing}
     return subset & o
 
 @predicate('p1([set])', safe=True)
 def p1(repo, subset, x):
     """First parent of changesets in set, or the working directory.
     """
     if x is None:
         p = repo[x].p1().rev()
         if p >= 0:
             return subset & baseset([p])
         return baseset()
 
     ps = set()
     cl = repo.changelog
     for r in getset(repo, fullreposet(repo), x):
         try:
             ps.add(cl.parentrevs(r)[0])
         except error.WdirUnsupported:
             ps.add(repo[r].parents()[0].rev())
     ps -= {node.nullrev}
     # XXX we should turn this into a baseset instead of a set, smartset may do
     # some optimizations from the fact this is a baseset.
     return subset & ps
 
 @predicate('p2([set])', safe=True)
 def p2(repo, subset, x):
     """Second parent of changesets in set, or the working directory.
     """
     if x is None:
         ps = repo[x].parents()
         try:
             p = ps[1].rev()
             if p >= 0:
                 return subset & baseset([p])
             return baseset()
         except IndexError:
             return baseset()
 
     ps = set()
     cl = repo.changelog
     for r in getset(repo, fullreposet(repo), x):
         try:
             ps.add(cl.parentrevs(r)[1])
         except error.WdirUnsupported:
             parents = repo[r].parents()
             if len(parents) == 2:
                 ps.add(parents[1])
     ps -= {node.nullrev}
     # XXX we should turn this into a baseset instead of a set, smartset may do
     # some optimizations from the fact this is a baseset.
     return subset & ps
 
 def parentpost(repo, subset, x, order):
     return p1(repo, subset, x)
 
 @predicate('parents([set])', safe=True)
 def parents(repo, subset, x):
     """
     The set of all parents for all changesets in set, or the working directory.
     """
     if x is None:
         ps = set(p.rev() for p in repo[x].parents())
     else:
         ps = set()
         cl = repo.changelog
         up = ps.update
         parentrevs = cl.parentrevs
         for r in getset(repo, fullreposet(repo), x):
             try:
                 up(parentrevs(r))
             except error.WdirUnsupported:
                 up(p.rev() for p in repo[r].parents())
     ps -= {node.nullrev}
     return subset & ps
 
 def _phase(repo, subset, *targets):
     """helper to select all rev in <targets> phases"""
     s = repo._phasecache.getrevset(repo, targets)
     return subset & s
 
 @predicate('draft()', safe=True)
 def draft(repo, subset, x):
     """Changeset in draft phase."""
     # i18n: "draft" is a keyword
     getargs(x, 0, 0, _("draft takes no arguments"))
     target = phases.draft
     return _phase(repo, subset, target)
 
 @predicate('secret()', safe=True)
 def secret(repo, subset, x):
     """Changeset in secret phase."""
     # i18n: "secret" is a keyword
     getargs(x, 0, 0, _("secret takes no arguments"))
     target = phases.secret
     return _phase(repo, subset, target)
 
 def parentspec(repo, subset, x, n, order):
     """``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, fullreposet(repo), x):
         if n == 0:
             ps.add(r)
         elif n == 1:
             try:
                 ps.add(cl.parentrevs(r)[0])
             except error.WdirUnsupported:
                 ps.add(repo[r].parents()[0].rev())
         else:
             try:
                 parents = cl.parentrevs(r)
                 if parents[1] != node.nullrev:
                     ps.add(parents[1])
             except error.WdirUnsupported:
                 parents = repo[r].parents()
                 if len(parents) == 2:
                     ps.add(parents[1].rev())
     return subset & ps
 
 @predicate('present(set)', safe=True, takeorder=True)
 def present(repo, subset, x, order):
     """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, order)
     except error.RepoLookupError:
         return baseset()
 
 # for internal use
 @predicate('_notpublic', safe=True)
 def _notpublic(repo, subset, x):
     getargs(x, 0, 0, "_notpublic takes no arguments")
     return _phase(repo, subset, phases.draft, phases.secret)
 
 @predicate('public()', safe=True)
 def public(repo, subset, x):
     """Changeset in public phase."""
     # i18n: "public" is a keyword
     getargs(x, 0, 0, _("public takes no arguments"))
     phase = repo._phasecache.phase
     target = phases.public
     condition = lambda r: phase(repo, r) == target
     return subset.filter(condition, condrepr=('<phase %r>', target),
                          cache=False)
 
 @predicate('remote([id [,path]])', safe=False)
 def remote(repo, subset, x):
     """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.
     """
 
     from . import hg # avoid start-up nasties
     # i18n: "remote" is a keyword
     l = getargs(x, 0, 2, _("remote takes zero, one, or two 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()
 
 @predicate('removes(pattern)', safe=True)
 def removes(repo, subset, x):
     """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)
 
 @predicate('rev(number)', safe=True)
 def rev(repo, subset, x):
     """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"))
     if l not in repo.changelog and l not in (node.nullrev, node.wdirrev):
         return baseset()
     return subset & baseset([l])
 
 @predicate('matching(revision [, field])', safe=True)
 def matching(repo, subset, x):
     """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, fullreposet(repo), 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 subset.filter(matches, condrepr=('<matching%r %r>', fields, revs))
 
 @predicate('reverse(set)', safe=True, takeorder=True)
 def reverse(repo, subset, x, order):
     """Reverse order of set.
     """
     l = getset(repo, subset, x, order)
     if order == defineorder:
         l.reverse()
     return l
 
 @predicate('roots(set)', safe=True)
 def roots(repo, subset, x):
     """Changesets in set with no parent changeset in set.
     """
     s = getset(repo, fullreposet(repo), x)
     parents = repo.changelog.parentrevs
     def filter(r):
         for p in parents(r):
             if 0 <= p and p in s:
                 return False
         return True
     return subset & s.filter(filter, condrepr='<roots>')
 
 _sortkeyfuncs = {
     'rev': lambda c: c.rev(),
     'branch': lambda c: c.branch(),
     'desc': lambda c: c.description(),
     'user': lambda c: c.user(),
     'author': lambda c: c.user(),
     'date': lambda c: c.date()[0],
 }
 
 def _getsortargs(x):
     """Parse sort options into (set, [(key, reverse)], opts)"""
     args = getargsdict(x, 'sort', 'set keys topo.firstbranch')
     if 'set' not in args:
         # i18n: "sort" is a keyword
         raise error.ParseError(_('sort requires one or two arguments'))
     keys = "rev"
     if 'keys' in args:
         # i18n: "sort" is a keyword
         keys = getstring(args['keys'], _("sort spec must be a string"))
 
     keyflags = []
     for k in keys.split():
         fk = k
         reverse = (k[0] == '-')
         if reverse:
             k = k[1:]
         if k not in _sortkeyfuncs and k != 'topo':
             raise error.ParseError(_("unknown sort key %r") % fk)
         keyflags.append((k, reverse))
 
     if len(keyflags) > 1 and any(k == 'topo' for k, reverse in keyflags):
         # i18n: "topo" is a keyword
         raise error.ParseError(_('topo sort order cannot be combined '
                                  'with other sort keys'))
 
     opts = {}
     if 'topo.firstbranch' in args:
         if any(k == 'topo' for k, reverse in keyflags):
             opts['topo.firstbranch'] = args['topo.firstbranch']
         else:
             # i18n: "topo" and "topo.firstbranch" are keywords
             raise error.ParseError(_('topo.firstbranch can only be used '
                                      'when using the topo sort key'))
 
     return args['set'], keyflags, opts
 
 @predicate('sort(set[, [-]key... [, ...]])', safe=True, takeorder=True)
 def sort(repo, subset, x, order):
     """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
     - ``topo`` for a reverse topographical sort
 
     The ``topo`` sort order cannot be combined with other sort keys. This sort
     takes one optional argument, ``topo.firstbranch``, which takes a revset that
     specifies what topographical branches to prioritize in the sort.
 
     """
     s, keyflags, opts = _getsortargs(x)
     revs = getset(repo, subset, s, order)
 
     if not keyflags or order != defineorder:
         return revs
     if len(keyflags) == 1 and keyflags[0][0] == "rev":
         revs.sort(reverse=keyflags[0][1])
         return revs
     elif keyflags[0][0] == "topo":
         firstbranch = ()
         if 'topo.firstbranch' in opts:
             firstbranch = getset(repo, subset, opts['topo.firstbranch'])
         revs = baseset(dagop.toposort(revs, repo.changelog.parentrevs,
                                       firstbranch),
                        istopo=True)
         if keyflags[0][1]:
             revs.reverse()
         return revs
 
     # sort() is guaranteed to be stable
     ctxs = [repo[r] for r in revs]
     for k, reverse in reversed(keyflags):
         ctxs.sort(key=_sortkeyfuncs[k], reverse=reverse)
     return baseset([c.rev() for c in ctxs])
 
 @predicate('subrepo([pattern])')
 def subrepo(repo, subset, x):
     """Changesets that add, modify or remove the given subrepo.  If no subrepo
     pattern is named, any subrepo changes are returned.
     """
     # i18n: "subrepo" is a keyword
     args = getargs(x, 0, 1, _('subrepo takes at most one argument'))
     pat = None
     if len(args) != 0:
         pat = getstring(args[0], _("subrepo requires a pattern"))
 
     m = matchmod.exact(repo.root, repo.root, ['.hgsubstate'])
 
     def submatches(names):
         k, p, m = util.stringmatcher(pat)
         for name in names:
             if m(name):
                 yield name
 
     def matches(x):
         c = repo[x]
         s = repo.status(c.p1().node(), c.node(), match=m)
 
         if pat is None:
             return s.added or s.modified or s.removed
 
         if s.added:
             return any(submatches(c.substate.keys()))
 
         if s.modified:
             subs = set(c.p1().substate.keys())
             subs.update(c.substate.keys())
 
             for path in submatches(subs):
                 if c.p1().substate.get(path) != c.substate.get(path):
                     return True
 
         if s.removed:
             return any(submatches(c.p1().substate.keys()))
 
         return False
 
     return subset.filter(matches, condrepr=('<subrepo %r>', pat))
 
 def _mapbynodefunc(repo, s, f):
     """(repo, smartset, [node] -> [node]) -> smartset
 
     Helper method to map a smartset to another smartset given a function only
     talking about nodes. Handles converting between rev numbers and nodes, and
     filtering.
     """
     cl = repo.unfiltered().changelog
     torev = cl.rev
     tonode = cl.node
     nodemap = cl.nodemap
     result = set(torev(n) for n in f(tonode(r) for r in s) if n in nodemap)
     return smartset.baseset(result - repo.changelog.filteredrevs)
 
 @predicate('successors(set)', safe=True)
 def successors(repo, subset, x):
     """All successors for set, including the given set themselves"""
     s = getset(repo, fullreposet(repo), x)
     f = lambda nodes: obsutil.allsuccessors(repo.obsstore, nodes)
     d = _mapbynodefunc(repo, s, f)
     return subset & d
 
 def _substringmatcher(pattern, casesensitive=True):
     kind, pattern, matcher = util.stringmatcher(pattern,
                                                 casesensitive=casesensitive)
     if kind == 'literal':
         if not casesensitive:
             pattern = encoding.lower(pattern)
             matcher = lambda s: pattern in encoding.lower(s)
         else:
             matcher = lambda s: pattern in s
     return kind, pattern, matcher
 
 @predicate('tag([name])', safe=True)
 def tag(repo, subset, x):
     """The specified tag by name, or all tagged revisions if no name is given.
 
     Pattern matching is supported for `name`. See
     :hg:`help revisions.patterns`.
     """
     # 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 = util.stringmatcher(pattern)
         if kind == 'literal':
             # avoid resolving all tags
             tn = repo._tagscache.tags.get(pattern, None)
             if tn is None:
                 raise error.RepoLookupError(_("tag '%s' does not exist")
                                             % pattern)
             s = {repo[tn].rev()}
         else:
             s = {cl.rev(n) for t, n in repo.tagslist() if matcher(t)}
     else:
         s = {cl.rev(n) for t, n in repo.tagslist() if t != 'tip'}
     return subset & s
 
 @predicate('tagged', safe=True)
 def tagged(repo, subset, x):
     return tag(repo, subset, x)
 
 @predicate('unstable()', safe=True)
 def unstable(repo, subset, x):
     msg = ("'unstable()' is deprecated, "
            "use 'orphan()'")
     repo.ui.deprecwarn(msg, '4.4')
 
     return orphan(repo, subset, x)
 
 @predicate('orphan()', safe=True)
 def orphan(repo, subset, x):
     """Non-obsolete changesets with obsolete ancestors. (EXPERIMENTAL)
     """
     # i18n: "orphan" is a keyword
     getargs(x, 0, 0, _("orphan takes no arguments"))
     orphan = obsmod.getrevs(repo, 'orphan')
     return subset & orphan
 
 
 @predicate('user(string)', safe=True)
 def user(repo, subset, x):
     """User name contains string. The match is case-insensitive.
 
     Pattern matching is supported for `string`. See
     :hg:`help revisions.patterns`.
     """
     return author(repo, subset, x)
 
 @predicate('wdir()', safe=True)
 def wdir(repo, subset, x):
     """Working directory. (EXPERIMENTAL)"""
     # i18n: "wdir" is a keyword
     getargs(x, 0, 0, _("wdir takes no arguments"))
     if node.wdirrev in subset or isinstance(subset, fullreposet):
         return baseset([node.wdirrev])
     return baseset()
 
 def _orderedlist(repo, subset, x):
     s = getstring(x, "internal error")
     if not s:
         return baseset()
     # remove duplicates here. it's difficult for caller to deduplicate sets
     # because different symbols can point to the same rev.
     cl = repo.changelog
     ls = []
     seen = set()
     for t in s.split('\0'):
         try:
             # fast path for integer revision
             r = int(t)
             if str(r) != t or r not in cl:
                 raise ValueError
             revs = [r]
         except ValueError:
             revs = stringset(repo, subset, t, defineorder)
 
         for r in revs:
             if r in seen:
                 continue
             if (r in subset
                 or r == node.nullrev and isinstance(subset, fullreposet)):
                 ls.append(r)
             seen.add(r)
     return baseset(ls)
 
 # for internal use
 @predicate('_list', safe=True, takeorder=True)
 def _list(repo, subset, x, order):
     if order == followorder:
         # slow path to take the subset order
         return subset & _orderedlist(repo, fullreposet(repo), x)
     else:
         return _orderedlist(repo, subset, x)
 
 def _orderedintlist(repo, subset, x):
     s = getstring(x, "internal error")
     if not s:
         return baseset()
     ls = [int(r) for r in s.split('\0')]
     s = subset
     return baseset([r for r in ls if r in s])
 
 # for internal use
 @predicate('_intlist', safe=True, takeorder=True)
 def _intlist(repo, subset, x, order):
     if order == followorder:
         # slow path to take the subset order
         return subset & _orderedintlist(repo, fullreposet(repo), x)
     else:
         return _orderedintlist(repo, subset, x)
 
 def _orderedhexlist(repo, subset, x):
     s = getstring(x, "internal error")
     if not s:
         return baseset()
     cl = repo.changelog
     ls = [cl.rev(node.bin(r)) for r in s.split('\0')]
     s = subset
     return baseset([r for r in ls if r in s])
 
 # for internal use
 @predicate('_hexlist', safe=True, takeorder=True)
 def _hexlist(repo, subset, x, order):
     if order == followorder:
         # slow path to take the subset order
         return subset & _orderedhexlist(repo, fullreposet(repo), x)
     else:
         return _orderedhexlist(repo, subset, x)
 
 methods = {
     "range": rangeset,
     "rangeall": rangeall,
     "rangepre": rangepre,
     "rangepost": rangepost,
     "dagrange": dagrange,
     "string": stringset,
     "symbol": stringset,
     "and": andset,
     "flipand": flipandset,
     "or": orset,
     "not": notset,
     "difference": differenceset,
     "relation": relationset,
     "relsubscript": relsubscriptset,
     "subscript": subscriptset,
     "list": listset,
     "keyvalue": keyvaluepair,
     "func": func,
     "ancestor": ancestorspec,
     "parent": parentspec,
     "parentpost": parentpost,
 }
 
 def posttreebuilthook(tree, repo):
     # hook for extensions to execute code on the optimized tree
     pass
 
 def match(ui, spec, repo=None, order=defineorder):
     """Create a matcher for a single revision spec
 
     If order=followorder, a matcher takes the ordering specified by the input
     set.
     """
     return matchany(ui, [spec], repo=repo, order=order)
 
 def matchany(ui, specs, repo=None, order=defineorder, localalias=None):
     """Create a matcher that will include any revisions matching one of the
     given specs
 
     If order=followorder, a matcher takes the ordering specified by the input
     set.
 
     If localalias is not None, it is a dict {name: definitionstring}. It takes
     precedence over [revsetalias] config section.
     """
     if not specs:
         def mfunc(repo, subset=None):
             return baseset()
         return mfunc
     if not all(specs):
         raise error.ParseError(_("empty query"))
     lookup = None
     if repo:
         lookup = repo.__contains__
     if len(specs) == 1:
         tree = revsetlang.parse(specs[0], lookup)
     else:
         tree = ('or',
                 ('list',) + tuple(revsetlang.parse(s, lookup) for s in specs))
 
     aliases = []
     warn = None
     if ui:
         aliases.extend(ui.configitems('revsetalias'))
         warn = ui.warn
     if localalias:
         aliases.extend(localalias.items())
     if aliases:
         tree = revsetlang.expandaliases(tree, aliases, warn=warn)
     tree = revsetlang.foldconcat(tree)
     tree = revsetlang.analyze(tree)
     tree = revsetlang.optimize(tree)
     posttreebuilthook(tree, repo)
     return makematcher(tree, order)
 
 def makematcher(tree, order=defineorder):
     """Create a matcher from an evaluatable tree"""
     def mfunc(repo, subset=None):
         if subset is None:
             subset = fullreposet(repo)
         return getset(repo, subset, tree, order)
     return mfunc
 
 def loadpredicate(ui, extname, registrarobj):
     """Load revset predicates from specified registrarobj
     """
     for name, func in registrarobj._table.iteritems():
         symbols[name] = func
         if func._safe:
             safesymbols.add(name)
 
 # load built-in predicates explicitly to setup safesymbols
 loadpredicate(None, None, predicate)
 
 # tell hggettext to extract docstrings from these functions:
 i18nfunctions = symbols.values()

mercurial/revsetlang.py

@@ -1,713 +1,671 b''
 # revsetlang.py - parser, tokenizer and utility for revision set language
 #
 # 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.
 
 from __future__ import absolute_import
 
 import string
 
 from .i18n import _
 from . import (
     error,
     node,
     parser,
     pycompat,
     util,
 )
 
 elements = {
     # token-type: binding-strength, primary, prefix, infix, suffix
     "(": (21, None, ("group", 1, ")"), ("func", 1, ")"), None),
     "[": (21, None, None, ("subscript", 1, "]"), None),
     "#": (21, None, None, ("relation", 21), None),
     "##": (20, None, None, ("_concat", 20), None),
     "~": (18, None, None, ("ancestor", 18), None),
     "^": (18, None, None, ("parent", 18), "parentpost"),
     "-": (5, None, ("negate", 19), ("minus", 5), None),
     "::": (17, None, ("dagrangepre", 17), ("dagrange", 17), "dagrangepost"),
     "..": (17, None, ("dagrangepre", 17), ("dagrange", 17), "dagrangepost"),
     ":": (15, "rangeall", ("rangepre", 15), ("range", 15), "rangepost"),
     "not": (10, None, ("not", 10), None, None),
     "!": (10, None, ("not", 10), None, None),
     "and": (5, None, None, ("and", 5), None),
     "&": (5, None, None, ("and", 5), None),
     "%": (5, None, None, ("only", 5), "onlypost"),
     "or": (4, None, None, ("or", 4), None),
     "|": (4, None, None, ("or", 4), None),
     "+": (4, None, None, ("or", 4), None),
     "=": (3, None, None, ("keyvalue", 3), None),
     ",": (2, None, None, ("list", 2), None),
     ")": (0, None, None, None, None),
     "]": (0, None, None, None, None),
     "symbol": (0, "symbol", None, None, None),
     "string": (0, "string", None, None, None),
     "end": (0, None, None, None, None),
 }
 
 keywords = {'and', 'or', 'not'}
 
 _quoteletters = {'"', "'"}
 _simpleopletters = set(pycompat.iterbytestr("()[]#:=,-|&+!~^%"))
 
 # default set of valid characters for the initial letter of symbols
 _syminitletters = set(pycompat.iterbytestr(
     string.ascii_letters.encode('ascii') +
     string.digits.encode('ascii') +
     '._@')) | set(map(pycompat.bytechr, xrange(128, 256)))
 
 # default set of valid characters for non-initial letters of symbols
 _symletters = _syminitletters | set(pycompat.iterbytestr('-/'))
 
 def tokenize(program, lookup=None, syminitletters=None, symletters=None):
     '''
     Parse a revset statement into a stream of tokens
 
     ``syminitletters`` is the set of valid characters for the initial
     letter of symbols.
 
     By default, character ``c`` is recognized as valid for initial
     letter of symbols, if ``c.isalnum() or c in '._@' or ord(c) > 127``.
 
     ``symletters`` is the set of valid characters for non-initial
     letters of symbols.
 
     By default, character ``c`` is recognized as valid for non-initial
     letters of symbols, if ``c.isalnum() or c in '-._/@' or ord(c) > 127``.
 
     Check that @ is a valid unquoted token character (issue3686):
     >>> list(tokenize("@::"))
     [('symbol', '@', 0), ('::', None, 1), ('end', None, 3)]
 
     '''
     program = pycompat.bytestr(program)
     if syminitletters is None:
         syminitletters = _syminitletters
     if symletters is None:
         symletters = _symletters
 
     if program and lookup:
         # attempt to parse old-style ranges first to deal with
         # things like old-tag which contain query metacharacters
         parts = program.split(':', 1)
         if all(lookup(sym) for sym in parts if sym):
             if parts[0]:
                 yield ('symbol', parts[0], 0)
             if len(parts) > 1:
                 s = len(parts[0])
                 yield (':', None, s)
                 if parts[1]:
                     yield ('symbol', parts[1], s + 1)
             yield ('end', None, len(program))
             return
 
     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 == '#' and program[pos:pos + 2] == '##': # look ahead carefully
             yield ('##', None, pos)
             pos += 1 # skip ahead
         elif c in _simpleopletters: # handle simple operators
             yield (c, None, pos)
         elif (c in _quoteletters 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 = parser.unescapestr
             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 in syminitletters:
             s = pos
             pos += 1
             while pos < l: # find end of symbol
                 d = program[pos]
                 if d not in symletters:
                     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)
             elif '-' in sym:
                 # some jerk gave us foo-bar-baz, try to check if it's a symbol
                 if lookup and lookup(sym):
                     # looks like a real symbol
                     yield ('symbol', sym, s)
                 else:
                     # looks like an expression
                     parts = sym.split('-')
                     for p in parts[:-1]:
                         if p: # possible consecutive -
                             yield ('symbol', p, s)
                         s += len(p)
                         yield ('-', None, pos)
                         s += 1
                     if parts[-1]: # possible trailing -
                         yield ('symbol', parts[-1], s)
             else:
                 yield ('symbol', sym, s)
             pos -= 1
         else:
             raise error.ParseError(_("syntax error in revset '%s'") %
                                    program, pos)
         pos += 1
     yield ('end', None, pos)
 
 # helpers
 
 _notset = object()
 
 def getsymbol(x):
     if x and x[0] == 'symbol':
         return x[1]
     raise error.ParseError(_('not a symbol'))
 
 def getstring(x, err):
     if x and (x[0] == 'string' or x[0] == 'symbol'):
         return x[1]
     raise error.ParseError(err)
 
 def getinteger(x, err, default=_notset):
     if not x and default is not _notset:
         return default
     try:
         return int(getstring(x, err))
     except ValueError:
         raise error.ParseError(err)
 
 def getboolean(x, err):
     value = util.parsebool(getsymbol(x))
     if value is not None:
         return value
     raise error.ParseError(err)
 
 def getlist(x):
     if not x:
         return []
     if x[0] == 'list':
         return list(x[1:])
     return [x]
 
 def getrange(x, err):
     if not x:
         raise error.ParseError(err)
     op = x[0]
     if op == 'range':
         return x[1], x[2]
     elif op == 'rangepre':
         return None, x[1]
     elif op == 'rangepost':
         return x[1], None
     elif op == 'rangeall':
         return None, None
     raise error.ParseError(err)
 
 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 getargsdict(x, funcname, keys):
     return parser.buildargsdict(getlist(x), funcname, parser.splitargspec(keys),
                                 keyvaluenode='keyvalue', keynode='symbol')
 
 def _isnamedfunc(x, funcname):
     """Check if given tree matches named function"""
     return x and x[0] == 'func' and getsymbol(x[1]) == funcname
 
 def _isposargs(x, n):
     """Check if given tree is n-length list of positional arguments"""
     l = getlist(x)
     return len(l) == n and all(y and y[0] != 'keyvalue' for y in l)
 
 def _matchnamedfunc(x, funcname):
     """Return args tree if given tree matches named function; otherwise None
 
     This can't be used for testing a nullary function since its args tree
     is also None. Use _isnamedfunc() instead.
     """
     if not _isnamedfunc(x, funcname):
         return
     return x[2]
 
-# Constants for ordering requirement, used in getset():
-#
-# If 'define', any nested functions and operations MAY change the ordering of
-# the entries in the set (but if changes the ordering, it MUST ALWAYS change
-# it). If 'follow', any nested functions and operations MUST take the ordering
-# specified by the first operand to the '&' operator.
-#
-# For instance,
-#
-#   X & (Y | Z)
-#   ^   ^^^^^^^
-#   |   follow
-#   define
-#
-# will be evaluated as 'or(y(x()), z(x()))', where 'x()' can change the order
-# of the entries in the set, but 'y()', 'z()' and 'or()' shouldn't.
-#
-# 'any' means the order doesn't matter. For instance,
-#
-#   (X & Y) | ancestors(Z)
-#        ^              ^
-#        any            any
-#
-# For 'X & Y', 'X' decides order so the order of 'Y' does not matter. For
-# 'ancestors(Z)', Z's order does not matter since 'ancestors' does not care
-# about the order of its argument.
-#
-# Currently, most revsets do not care about the order, so 'define' is
-# equivalent to 'follow' for them, and the resulting order is based on the
-# 'subset' parameter passed down to them:
-#
-#   m = revset.match(..., order=defineorder)
-#   m(repo, subset)
-#           ^^^^^^
-#      For most revsets, 'define' means using the order this subset provides
-#
-# There are a few revsets that always redefine the order if 'define' is
-# specified: 'sort(X)', 'reverse(X)', 'x:y'.
-anyorder = 'any'        # don't care the order, could be even random-shuffled
-defineorder = 'define'  # ALWAYS redefine, or ALWAYS follow the current order
-followorder = 'follow'  # MUST follow the current order
-
 def _matchonly(revs, bases):
     """
     >>> f = lambda *args: _matchonly(*map(parse, args))
     >>> f('ancestors(A)', 'not ancestors(B)')
     ('list', ('symbol', 'A'), ('symbol', 'B'))
     """
     ta = _matchnamedfunc(revs, 'ancestors')
     tb = bases and bases[0] == 'not' and _matchnamedfunc(bases[1], 'ancestors')
     if _isposargs(ta, 1) and _isposargs(tb, 1):
         return ('list', ta, tb)
 
 def _fixops(x):
     """Rewrite raw parsed tree to resolve ambiguous syntax which cannot be
     handled well by our simple top-down parser"""
     if not isinstance(x, tuple):
         return x
 
     op = x[0]
     if op == 'parent':
         # x^:y means (x^) : y, not x ^ (:y)
         # x^:  means (x^) :,   not x ^ (:)
         post = ('parentpost', x[1])
         if x[2][0] == 'dagrangepre':
             return _fixops(('dagrange', post, x[2][1]))
         elif x[2][0] == 'rangepre':
             return _fixops(('range', post, x[2][1]))
         elif x[2][0] == 'rangeall':
             return _fixops(('rangepost', post))
     elif op == 'or':
         # make number of arguments deterministic:
         # x + y + z -> (or x y z) -> (or (list x y z))
         return (op, _fixops(('list',) + x[1:]))
     elif op == 'subscript' and x[1][0] == 'relation':
         # x#y[z] ternary
         return _fixops(('relsubscript', x[1][1], x[1][2], x[2]))
 
     return (op,) + tuple(_fixops(y) for y in x[1:])
 
 def _analyze(x):
     if x is None:
         return x
 
     op = x[0]
     if op == 'minus':
         return _analyze(('and', x[1], ('not', x[2])))
     elif op == 'only':
         t = ('func', ('symbol', 'only'), ('list', x[1], x[2]))
         return _analyze(t)
     elif op == 'onlypost':
         return _analyze(('func', ('symbol', 'only'), x[1]))
     elif op == 'dagrangepre':
         return _analyze(('func', ('symbol', 'ancestors'), x[1]))
     elif op == 'dagrangepost':
         return _analyze(('func', ('symbol', 'descendants'), x[1]))
     elif op == 'negate':
         s = getstring(x[1], _("can't negate that"))
         return _analyze(('string', '-' + s))
     elif op in ('string', 'symbol'):
         return x
     elif op == 'rangeall':
         return (op, None)
     elif op in {'or', 'not', 'rangepre', 'rangepost', 'parentpost'}:
         return (op, _analyze(x[1]))
     elif op == 'group':
         return _analyze(x[1])
     elif op in {'and', 'dagrange', 'range', 'parent', 'ancestor', 'relation',
                 'subscript'}:
         ta = _analyze(x[1])
         tb = _analyze(x[2])
         return (op, ta, tb)
     elif op == 'relsubscript':
         ta = _analyze(x[1])
         tb = _analyze(x[2])
         tc = _analyze(x[3])
         return (op, ta, tb, tc)
     elif op == 'list':
         return (op,) + tuple(_analyze(y) for y in x[1:])
     elif op == 'keyvalue':
         return (op, x[1], _analyze(x[2]))
     elif op == 'func':
         return (op, x[1], _analyze(x[2]))
     raise ValueError('invalid operator %r' % op)
 
 def analyze(x):
     """Transform raw parsed tree to evaluatable tree which can be fed to
     optimize() or getset()
 
     All pseudo operations should be mapped to real operations or functions
     defined in methods or symbols table respectively.
     """
     return _analyze(x)
 
 def _optimize(x, small):
     if x is None:
         return 0, x
 
     smallbonus = 1
     if small:
         smallbonus = .5
 
     op = x[0]
     if op in ('string', 'symbol'):
         return smallbonus, x # single revisions are small
     elif op == 'and':
         wa, ta = _optimize(x[1], True)
         wb, tb = _optimize(x[2], True)
         w = min(wa, wb)
 
         # (::x and not ::y)/(not ::y and ::x) have a fast path
         tm = _matchonly(ta, tb) or _matchonly(tb, ta)
         if tm:
             return w, ('func', ('symbol', 'only'), tm)
 
         if tb is not None and tb[0] == 'not':
             return wa, ('difference', ta, tb[1])
         if wa > wb:
             return w, ('flipand', tb, ta)
         return w, (op, ta, tb)
     elif op == 'or':
         # fast path for machine-generated expression, that is likely to have
         # lots of trivial revisions: 'a + b + c()' to '_list(a b) + c()'
         ws, ts, ss = [], [], []
         def flushss():
             if not ss:
                 return
             if len(ss) == 1:
                 w, t = ss[0]
             else:
                 s = '\0'.join(t[1] for w, t in ss)
                 y = ('func', ('symbol', '_list'), ('string', s))
                 w, t = _optimize(y, False)
             ws.append(w)
             ts.append(t)
             del ss[:]
         for y in getlist(x[1]):
             w, t = _optimize(y, False)
             if t is not None and (t[0] == 'string' or t[0] == 'symbol'):
                 ss.append((w, t))
                 continue
             flushss()
             ws.append(w)
             ts.append(t)
         flushss()
         if len(ts) == 1:
             return ws[0], ts[0] # 'or' operation is fully optimized out
         return max(ws), (op, ('list',) + tuple(ts))
     elif op == 'not':
         # Optimize not public() to _notpublic() because we have a fast version
         if x[1][:3] == ('func', ('symbol', 'public'), None):
             newsym = ('func', ('symbol', '_notpublic'), None)
             o = _optimize(newsym, not small)
             return o[0], o[1]
         else:
             o = _optimize(x[1], not small)
             return o[0], (op, o[1])
     elif op == 'rangeall':
         return smallbonus, x
     elif op in ('rangepre', 'rangepost', 'parentpost'):
         o = _optimize(x[1], small)
         return o[0], (op, o[1])
     elif op in ('dagrange', 'range'):
         wa, ta = _optimize(x[1], small)
         wb, tb = _optimize(x[2], small)
         return wa + wb, (op, ta, tb)
     elif op in ('parent', 'ancestor', 'relation', 'subscript'):
         w, t = _optimize(x[1], small)
         return w, (op, t, x[2])
     elif op == 'relsubscript':
         w, t = _optimize(x[1], small)
         return w, (op, t, x[2], x[3])
     elif op == 'list':
         ws, ts = zip(*(_optimize(y, small) for y in x[1:]))
         return sum(ws), (op,) + ts
     elif op == 'keyvalue':
         w, t = _optimize(x[2], small)
         return w, (op, x[1], t)
     elif op == 'func':
         f = getsymbol(x[1])
         wa, ta = _optimize(x[2], small)
         if f in ('author', 'branch', 'closed', 'date', 'desc', 'file', 'grep',
                  'keyword', 'outgoing', 'user', 'destination'):
             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', 'wdir', '_intlist'):
             w = 0
         elif f == "sort":
             w = 10 # assume most sorts look at changelog
         else:
             w = 1
         return w + wa, (op, x[1], ta)
     raise ValueError('invalid operator %r' % op)
 
 def optimize(tree):
     """Optimize evaluatable tree
 
     All pseudo operations should be transformed beforehand.
     """
     _weight, newtree = _optimize(tree, small=True)
     return newtree
 
 # the set of valid characters for the initial letter of symbols in
 # alias declarations and definitions
 _aliassyminitletters = _syminitletters | set(pycompat.sysstr('$'))
 
 def _parsewith(spec, lookup=None, syminitletters=None):
     """Generate a parse tree of given spec with given tokenizing options
 
     >>> _parsewith('foo($1)', syminitletters=_aliassyminitletters)
     ('func', ('symbol', 'foo'), ('symbol', '$1'))
     >>> _parsewith('$1')
     Traceback (most recent call last):
       ...
     ParseError: ("syntax error in revset '$1'", 0)
     >>> _parsewith('foo bar')
     Traceback (most recent call last):
       ...
     ParseError: ('invalid token', 4)
     """
     p = parser.parser(elements)
     tree, pos = p.parse(tokenize(spec, lookup=lookup,
                                  syminitletters=syminitletters))
     if pos != len(spec):
         raise error.ParseError(_('invalid token'), pos)
     return _fixops(parser.simplifyinfixops(tree, ('list', 'or')))
 
 class _aliasrules(parser.basealiasrules):
     """Parsing and expansion rule set of revset aliases"""
     _section = _('revset alias')
 
     @staticmethod
     def _parse(spec):
         """Parse alias declaration/definition ``spec``
 
         This allows symbol names to use also ``$`` as an initial letter
         (for backward compatibility), and callers of this function should
         examine whether ``$`` is used also for unexpected symbols or not.
         """
         return _parsewith(spec, syminitletters=_aliassyminitletters)
 
     @staticmethod
     def _trygetfunc(tree):
         if tree[0] == 'func' and tree[1][0] == 'symbol':
             return tree[1][1], getlist(tree[2])
 
 def expandaliases(tree, aliases, warn=None):
     """Expand aliases in a tree, aliases is a list of (name, value) tuples"""
     aliases = _aliasrules.buildmap(aliases)
     tree = _aliasrules.expand(aliases, tree)
     # warn about problematic (but not referred) aliases
     if warn is not None:
         for name, alias in sorted(aliases.iteritems()):
             if alias.error and not alias.warned:
                 warn(_('warning: %s\n') % (alias.error))
                 alias.warned = True
     return tree
 
 def foldconcat(tree):
     """Fold elements to be concatenated by `##`
     """
     if not isinstance(tree, tuple) or tree[0] in ('string', 'symbol'):
         return tree
     if tree[0] == '_concat':
         pending = [tree]
         l = []
         while pending:
             e = pending.pop()
             if e[0] == '_concat':
                 pending.extend(reversed(e[1:]))
             elif e[0] in ('string', 'symbol'):
                 l.append(e[1])
             else:
                 msg = _("\"##\" can't concatenate \"%s\" element") % (e[0])
                 raise error.ParseError(msg)
         return ('string', ''.join(l))
     else:
         return tuple(foldconcat(t) for t in tree)
 
 def parse(spec, lookup=None):
     return _parsewith(spec, lookup=lookup)
 
 def _quote(s):
     r"""Quote a value in order to make it safe for the revset engine.
 
     >>> _quote('asdf')
     "'asdf'"
     >>> _quote("asdf'\"")
     '\'asdf\\\'"\''
     >>> _quote('asdf\'')
     "'asdf\\''"
     >>> _quote(1)
     "'1'"
     """
     return "'%s'" % util.escapestr(pycompat.bytestr(s))
 
 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 argtype(c, arg):
         if c == 'd':
             return '%d' % 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 "_intlist('%s')" % "\0".join('%d' % int(a) for a in s)
         elif t == 's':
             return "_list('%s')" % "\0".join(s)
         elif t == 'n':
             return "_hexlist('%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))
 
     expr = pycompat.bytestr(expr)
     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 error.Abort(_('unexpected revspec format character %s')
                                   % d)
         else:
             ret += c
         pos += 1
 
     return ret
 
 def prettyformat(tree):
     return parser.prettyformat(tree, ('string', 'symbol'))
 
 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