Mercurial > hg-stable
changeset 20611:6490f8385391
revset: changed revset code to use filter method
Revset methods now use the filter code to apply a condition.
author | Lucas Moscovicz <lmoscovicz@fb.com> |
---|---|
date | Thu, 06 Feb 2014 09:28:41 -0800 |
parents | 34bb07e70c68 |
children | 60c308b932eb |
files | mercurial/revset.py |
diffstat | 1 files changed, 37 insertions(+), 37 deletions(-) [+] |
line wrap: on
line diff
--- a/mercurial/revset.py Thu Feb 06 17:18:11 2014 -0800 +++ b/mercurial/revset.py Thu Feb 06 09:28:41 2014 -0800 @@ -244,7 +244,7 @@ r = spanset(repo) xs = _revsbetween(repo, getset(repo, r, x), getset(repo, r, y)) s = subset.set() - return baseset([r for r in xs if r in s]) + return xs.filter(lambda r: r in s) def andset(repo, subset, x, y): return getset(repo, getset(repo, subset, x), y) @@ -312,7 +312,7 @@ if not args: return baseset([]) s = set(_revancestors(repo, args, followfirst)) | set(args) - return baseset([r for r in subset if r in s]) + return subset.filter(lambda r: r in s) def ancestors(repo, subset, x): """``ancestors(set)`` @@ -340,7 +340,7 @@ for i in range(n): r = cl.parentrevs(r)[0] ps.add(r) - return baseset([r for r in subset if r in ps]) + return subset.filter(lambda r: r in ps) def author(repo, subset, x): """``author(string)`` @@ -349,7 +349,7 @@ # i18n: "author" is a keyword n = encoding.lower(getstring(x, _("author requires a string"))) kind, pattern, matcher = _substringmatcher(n) - return lazyset(subset, lambda x: matcher(encoding.lower(repo[x].user()))) + return subset.filter(lambda x: matcher(encoding.lower(repo[x].user()))) def bisect(repo, subset, x): """``bisect(string)`` @@ -366,7 +366,7 @@ # i18n: "bisect" is a keyword status = getstring(x, _("bisect requires a string")).lower() state = set(hbisect.get(repo, status)) - return baseset([r for r in subset if r in state]) + return subset.filter(lambda r: r in state) # Backward-compatibility # - no help entry so that we do not advertise it any more @@ -393,7 +393,7 @@ if not bmrev: raise util.Abort(_("bookmark '%s' does not exist") % bm) bmrev = repo[bmrev].rev() - return lazyset(subset, lambda r: r == bmrev) + return subset.filter(lambda r: r == bmrev) else: matchrevs = set() for name, bmrev in repo._bookmarks.iteritems(): @@ -409,7 +409,7 @@ bms = set([repo[r].rev() for r in repo._bookmarks.values()]) - return lazyset(subset, lambda r: r in bms) + return subset.filter(lambda r: r in bms) def branch(repo, subset, x): """``branch(string or set)`` @@ -431,16 +431,16 @@ # note: falls through to the revspec case if no branch with # this name exists if pattern in repo.branchmap(): - return lazyset(subset, lambda r: matcher(repo[r].branch())) + return subset.filter(lambda r: matcher(repo[r].branch())) else: - return lazyset(subset, lambda r: matcher(repo[r].branch())) + return subset.filter(lambda r: matcher(repo[r].branch())) s = getset(repo, spanset(repo), x) b = set() for r in s: b.add(repo[r].branch()) s = s.set() - return lazyset(subset, lambda r: r in s or repo[r].branch() in b) + return subset.filter(lambda r: r in s or repo[r].branch() in b) def bumped(repo, subset, x): """``bumped()`` @@ -494,7 +494,7 @@ if m(f): return True - return lazyset(subset, matches) + return subset.filter(matches) def _children(repo, narrow, parentset): cs = set() @@ -524,7 +524,7 @@ """ # i18n: "closed" is a keyword getargs(x, 0, 0, _("closed takes no arguments")) - return lazyset(subset, lambda r: repo[r].closesbranch()) + return subset.filter(lambda r: repo[r].closesbranch()) def contains(repo, subset, x): """``contains(pattern)`` @@ -551,7 +551,7 @@ return True return False - return lazyset(subset, matches) + return subset.filter(matches) def converted(repo, subset, x): """``converted([id])`` @@ -573,7 +573,7 @@ source = repo[r].extra().get('convert_revision', None) return source is not None and (rev is None or source.startswith(rev)) - return lazyset(subset, lambda r: _matchvalue(r)) + return subset.filter(lambda r: _matchvalue(r)) def date(repo, subset, x): """``date(interval)`` @@ -582,7 +582,7 @@ # i18n: "date" is a keyword ds = getstring(x, _("date requires a string")) dm = util.matchdate(ds) - return lazyset(subset, lambda x: dm(repo[x].date()[0])) + return subset.filter(lambda x: dm(repo[x].date()[0])) def desc(repo, subset, x): """``desc(string)`` @@ -595,7 +595,7 @@ c = repo[x] return ds in encoding.lower(c.description()) - return lazyset(subset, matches) + return subset.filter(matches) def _descendants(repo, subset, x, followfirst=False): args = getset(repo, spanset(repo), x) @@ -657,7 +657,7 @@ r = src src = _getrevsource(repo, r) - return baseset([r for r in subset if r in dests]) + return subset.filter(lambda r: r in dests) def divergent(repo, subset, x): """``divergent()`` @@ -666,7 +666,7 @@ # i18n: "divergent" is a keyword getargs(x, 0, 0, _("divergent takes no arguments")) divergent = obsmod.getrevs(repo, 'divergent') - return baseset([r for r in subset if r in divergent]) + return subset.filter(lambda r: r in divergent) def draft(repo, subset, x): """``draft()`` @@ -674,7 +674,7 @@ # i18n: "draft" is a keyword getargs(x, 0, 0, _("draft takes no arguments")) pc = repo._phasecache - return lazyset(subset, lambda r: pc.phase(repo, r) == phases.draft) + return subset.filter(lambda r: pc.phase(repo, r) == phases.draft) def extinct(repo, subset, x): """``extinct()`` @@ -710,7 +710,7 @@ extra = repo[r].extra() return label in extra and (value is None or matcher(extra[label])) - return lazyset(subset, lambda r: _matchvalue(r)) + return subset.filter(lambda r: _matchvalue(r)) def filelog(repo, subset, x): """``filelog(pattern)`` @@ -742,7 +742,7 @@ for fr in fl: s.add(fl.linkrev(fr)) - return baseset([r for r in subset if r in s]) + return subset.filter(lambda r: r in s) def first(repo, subset, x): """``first(set, [n])`` @@ -765,7 +765,7 @@ else: s = set(_revancestors(repo, [c.rev()], followfirst)) | set([c.rev()]) - return baseset([r for r in subset if r in s]) + return subset.filter(lambda r: r in s) def follow(repo, subset, x): """``follow([file])`` @@ -808,7 +808,7 @@ return True return False - return lazyset(subset, matches) + return subset.filter(matches) def _matchfiles(repo, subset, x): # _matchfiles takes a revset list of prefixed arguments: @@ -872,7 +872,7 @@ return True return False - return lazyset(subset, matches) + return subset.filter(matches) def hasfile(repo, subset, x): """``file(pattern)`` @@ -896,7 +896,7 @@ hs = set() for b, ls in repo.branchmap().iteritems(): hs.update(repo[h].rev() for h in ls) - return baseset([r for r in subset if r in hs]) + return subset.filter(lambda r: r in hs) def heads(repo, subset, x): """``heads(set)`` @@ -928,7 +928,7 @@ return util.any(kw in encoding.lower(t) for t in c.files() + [c.user(), c.description()]) - return lazyset(subset, matches) + return subset.filter(matches) def limit(repo, subset, x): """``limit(set, [n])`` @@ -1003,7 +1003,7 @@ # i18n: "merge" is a keyword getargs(x, 0, 0, _("merge takes no arguments")) cl = repo.changelog - return lazyset(subset, lambda r: cl.parentrevs(r)[1] != -1) + return subset.filter(lambda r: cl.parentrevs(r)[1] != -1) def branchpoint(repo, subset, x): """``branchpoint()`` @@ -1020,7 +1020,7 @@ for p in cl.parentrevs(r): if p >= baserev: parentscount[p - baserev] += 1 - return baseset([r for r in subset if (parentscount[r - baserev] > 1)]) + return subset.filter(lambda r: parentscount[r - baserev] > 1) def minrev(repo, subset, x): """``min(set)`` @@ -1071,7 +1071,7 @@ if pm is not None: rn = repo.changelog.rev(pm) - return baseset([r for r in subset if r == rn]) + return subset.filter(lambda r: r == rn) def obsolete(repo, subset, x): """``obsolete()`` @@ -1107,7 +1107,7 @@ src = prev o = set([_firstsrc(r) for r in args]) - return baseset([r for r in subset if r in o]) + return subset.filter(lambda r: r in o) def outgoing(repo, subset, x): """``outgoing([path])`` @@ -1130,7 +1130,7 @@ repo.ui.popbuffer() cl = repo.changelog o = set([cl.rev(r) for r in outgoing.missing]) - return baseset([r for r in subset if r in o]) + return subset.filter(lambda r: r in o) def p1(repo, subset, x): """``p1([set])`` @@ -1138,7 +1138,7 @@ """ if x is None: p = repo[x].p1().rev() - return baseset([r for r in subset if r == p]) + return subset.filter(lambda r: r == p) ps = set() cl = repo.changelog @@ -1154,7 +1154,7 @@ ps = repo[x].parents() try: p = ps[1].rev() - return baseset([r for r in subset if r == p]) + return subset.filter(lambda r: r == p) except IndexError: return baseset([]) @@ -1223,7 +1223,7 @@ # i18n: "public" is a keyword getargs(x, 0, 0, _("public takes no arguments")) pc = repo._phasecache - return lazyset(subset, lambda r: pc.phase(repo, r) == phases.public) + return subset.filter(lambda r: pc.phase(repo, r) == phases.public) def remote(repo, subset, x): """``remote([id [,path]])`` @@ -1284,7 +1284,7 @@ except (TypeError, ValueError): # i18n: "rev" is a keyword raise error.ParseError(_("rev expects a number")) - return baseset([r for r in subset if r == l]) + return subset.filter(lambda r: r == l) def matching(repo, subset, x): """``matching(revision [, field])`` @@ -1396,7 +1396,7 @@ return True return False - return lazyset(subset, matches) + return subset.filter(matches) def reverse(repo, subset, x): """``reverse(set)`` @@ -1421,7 +1421,7 @@ # i18n: "secret" is a keyword getargs(x, 0, 0, _("secret takes no arguments")) pc = repo._phasecache - return lazyset(subset, lambda x: pc.phase(repo, x) == phases.secret) + return subset.filter(lambda x: pc.phase(repo, x) == phases.secret) def sort(repo, subset, x): """``sort(set[, [-]key...])``