manifest: move clearcaches to manifestlog

This is part of removing all functionality from manifest.manifest so we can
delete the class entirely.

contrib/perf.py

@@ -1,1162 +1,1162
 # perf.py - performance test routines
 '''helper extension to measure performance'''
 
 # "historical portability" policy of perf.py:
 #
 # We have to do:
 # - make perf.py "loadable" with as wide Mercurial version as possible
 #   This doesn't mean that perf commands work correctly with that Mercurial.
 #   BTW, perf.py itself has been available since 1.1 (or eb240755386d).
 # - make historical perf command work correctly with as wide Mercurial
 #   version as possible
 #
 # We have to do, if possible with reasonable cost:
 # - make recent perf command for historical feature work correctly
 #   with early Mercurial
 #
 # We don't have to do:
 # - make perf command for recent feature work correctly with early
 #   Mercurial
 
 from __future__ import absolute_import
 import functools
 import os
 import random
 import sys
 import time
 from mercurial import (
     bdiff,
     changegroup,
     cmdutil,
     commands,
     copies,
     error,
     extensions,
     mdiff,
     merge,
     revlog,
     util,
 )
 
 # for "historical portability":
 # try to import modules separately (in dict order), and ignore
 # failure, because these aren't available with early Mercurial
 try:
     from mercurial import branchmap # since 2.5 (or bcee63733aad)
 except ImportError:
     pass
 try:
     from mercurial import obsolete # since 2.3 (or ad0d6c2b3279)
 except ImportError:
     pass
 try:
     from mercurial import repoview # since 2.5 (or 3a6ddacb7198)
 except ImportError:
     pass
 try:
     from mercurial import scmutil # since 1.9 (or 8b252e826c68)
 except ImportError:
     pass
 
 # for "historical portability":
 # define util.safehasattr forcibly, because util.safehasattr has been
 # available since 1.9.3 (or 94b200a11cf7)
 _undefined = object()
 def safehasattr(thing, attr):
     return getattr(thing, attr, _undefined) is not _undefined
 setattr(util, 'safehasattr', safehasattr)
 
 # for "historical portability":
 # use locally defined empty option list, if formatteropts isn't
 # available, because commands.formatteropts has been available since
 # 3.2 (or 7a7eed5176a4), even though formatting itself has been
 # available since 2.2 (or ae5f92e154d3)
 formatteropts = getattr(commands, "formatteropts", [])
 
 # for "historical portability":
 # use locally defined option list, if debugrevlogopts isn't available,
 # because commands.debugrevlogopts has been available since 3.7 (or
 # 5606f7d0d063), even though cmdutil.openrevlog() has been available
 # since 1.9 (or a79fea6b3e77).
 revlogopts = getattr(commands, "debugrevlogopts", [
         ('c', 'changelog', False, ('open changelog')),
         ('m', 'manifest', False, ('open manifest')),
         ('', 'dir', False, ('open directory manifest')),
         ])
 
 cmdtable = {}
 
 # for "historical portability":
 # define parsealiases locally, because cmdutil.parsealiases has been
 # available since 1.5 (or 6252852b4332)
 def parsealiases(cmd):
     return cmd.lstrip("^").split("|")
 
 if safehasattr(cmdutil, 'command'):
     import inspect
     command = cmdutil.command(cmdtable)
     if 'norepo' not in inspect.getargspec(command)[0]:
         # for "historical portability":
         # wrap original cmdutil.command, because "norepo" option has
         # been available since 3.1 (or 75a96326cecb)
         _command = command
         def command(name, options=(), synopsis=None, norepo=False):
             if norepo:
                 commands.norepo += ' %s' % ' '.join(parsealiases(name))
             return _command(name, list(options), synopsis)
 else:
     # for "historical portability":
     # define "@command" annotation locally, because cmdutil.command
     # has been available since 1.9 (or 2daa5179e73f)
     def command(name, options=(), synopsis=None, norepo=False):
         def decorator(func):
             if synopsis:
                 cmdtable[name] = func, list(options), synopsis
             else:
                 cmdtable[name] = func, list(options)
             if norepo:
                 commands.norepo += ' %s' % ' '.join(parsealiases(name))
             return func
         return decorator
 
 def getlen(ui):
     if ui.configbool("perf", "stub"):
         return lambda x: 1
     return len
 
 def gettimer(ui, opts=None):
     """return a timer function and formatter: (timer, formatter)
 
     This function exists to gather the creation of formatter in a single
     place instead of duplicating it in all performance commands."""
 
     # enforce an idle period before execution to counteract power management
     # experimental config: perf.presleep
     time.sleep(getint(ui, "perf", "presleep", 1))
 
     if opts is None:
         opts = {}
     # redirect all to stderr
     ui = ui.copy()
     uifout = safeattrsetter(ui, 'fout', ignoremissing=True)
     if uifout:
         # for "historical portability":
         # ui.fout/ferr have been available since 1.9 (or 4e1ccd4c2b6d)
         uifout.set(ui.ferr)
 
     # get a formatter
     uiformatter = getattr(ui, 'formatter', None)
     if uiformatter:
         fm = uiformatter('perf', opts)
     else:
         # for "historical portability":
         # define formatter locally, because ui.formatter has been
         # available since 2.2 (or ae5f92e154d3)
         from mercurial import node
         class defaultformatter(object):
             """Minimized composition of baseformatter and plainformatter
             """
             def __init__(self, ui, topic, opts):
                 self._ui = ui
                 if ui.debugflag:
                     self.hexfunc = node.hex
                 else:
                     self.hexfunc = node.short
             def __nonzero__(self):
                 return False
             def startitem(self):
                 pass
             def data(self, **data):
                 pass
             def write(self, fields, deftext, *fielddata, **opts):
                 self._ui.write(deftext % fielddata, **opts)
             def condwrite(self, cond, fields, deftext, *fielddata, **opts):
                 if cond:
                     self._ui.write(deftext % fielddata, **opts)
             def plain(self, text, **opts):
                 self._ui.write(text, **opts)
             def end(self):
                 pass
         fm = defaultformatter(ui, 'perf', opts)
 
     # stub function, runs code only once instead of in a loop
     # experimental config: perf.stub
     if ui.configbool("perf", "stub"):
         return functools.partial(stub_timer, fm), fm
     return functools.partial(_timer, fm), fm
 
 def stub_timer(fm, func, title=None):
     func()
 
 def _timer(fm, func, title=None):
     results = []
     begin = time.time()
     count = 0
     while True:
         ostart = os.times()
         cstart = time.time()
         r = func()
         cstop = time.time()
         ostop = os.times()
         count += 1
         a, b = ostart, ostop
         results.append((cstop - cstart, b[0] - a[0], b[1]-a[1]))
         if cstop - begin > 3 and count >= 100:
             break
         if cstop - begin > 10 and count >= 3:
             break
 
     fm.startitem()
 
     if title:
         fm.write('title', '! %s\n', title)
     if r:
         fm.write('result', '! result: %s\n', r)
     m = min(results)
     fm.plain('!')
     fm.write('wall', ' wall %f', m[0])
     fm.write('comb', ' comb %f', m[1] + m[2])
     fm.write('user', ' user %f', m[1])
     fm.write('sys',  ' sys %f', m[2])
     fm.write('count',  ' (best of %d)', count)
     fm.plain('\n')
 
 # utilities for historical portability
 
 def getint(ui, section, name, default):
     # for "historical portability":
     # ui.configint has been available since 1.9 (or fa2b596db182)
     v = ui.config(section, name, None)
     if v is None:
         return default
     try:
         return int(v)
     except ValueError:
         raise error.ConfigError(("%s.%s is not an integer ('%s')")
                                 % (section, name, v))
 
 def safeattrsetter(obj, name, ignoremissing=False):
     """Ensure that 'obj' has 'name' attribute before subsequent setattr
 
     This function is aborted, if 'obj' doesn't have 'name' attribute
     at runtime. This avoids overlooking removal of an attribute, which
     breaks assumption of performance measurement, in the future.
 
     This function returns the object to (1) assign a new value, and
     (2) restore an original value to the attribute.
 
     If 'ignoremissing' is true, missing 'name' attribute doesn't cause
     abortion, and this function returns None. This is useful to
     examine an attribute, which isn't ensured in all Mercurial
     versions.
     """
     if not util.safehasattr(obj, name):
         if ignoremissing:
             return None
         raise error.Abort(("missing attribute %s of %s might break assumption"
                            " of performance measurement") % (name, obj))
 
     origvalue = getattr(obj, name)
     class attrutil(object):
         def set(self, newvalue):
             setattr(obj, name, newvalue)
         def restore(self):
             setattr(obj, name, origvalue)
 
     return attrutil()
 
 # utilities to examine each internal API changes
 
 def getbranchmapsubsettable():
     # for "historical portability":
     # subsettable is defined in:
     # - branchmap since 2.9 (or 175c6fd8cacc)
     # - repoview since 2.5 (or 59a9f18d4587)
     for mod in (branchmap, repoview):
         subsettable = getattr(mod, 'subsettable', None)
         if subsettable:
             return subsettable
 
     # bisecting in bcee63733aad::59a9f18d4587 can reach here (both
     # branchmap and repoview modules exist, but subsettable attribute
     # doesn't)
     raise error.Abort(("perfbranchmap not available with this Mercurial"),
                       hint="use 2.5 or later")
 
 def getsvfs(repo):
     """Return appropriate object to access files under .hg/store
     """
     # for "historical portability":
     # repo.svfs has been available since 2.3 (or 7034365089bf)
     svfs = getattr(repo, 'svfs', None)
     if svfs:
         return svfs
     else:
         return getattr(repo, 'sopener')
 
 def getvfs(repo):
     """Return appropriate object to access files under .hg
     """
     # for "historical portability":
     # repo.vfs has been available since 2.3 (or 7034365089bf)
     vfs = getattr(repo, 'vfs', None)
     if vfs:
         return vfs
     else:
         return getattr(repo, 'opener')
 
 def repocleartagscachefunc(repo):
     """Return the function to clear tags cache according to repo internal API
     """
     if util.safehasattr(repo, '_tagscache'): # since 2.0 (or 9dca7653b525)
         # in this case, setattr(repo, '_tagscache', None) or so isn't
         # correct way to clear tags cache, because existing code paths
         # expect _tagscache to be a structured object.
         def clearcache():
             # _tagscache has been filteredpropertycache since 2.5 (or
             # 98c867ac1330), and delattr() can't work in such case
             if '_tagscache' in vars(repo):
                 del repo.__dict__['_tagscache']
         return clearcache
 
     repotags = safeattrsetter(repo, '_tags', ignoremissing=True)
     if repotags: # since 1.4 (or 5614a628d173)
         return lambda : repotags.set(None)
 
     repotagscache = safeattrsetter(repo, 'tagscache', ignoremissing=True)
     if repotagscache: # since 0.6 (or d7df759d0e97)
         return lambda : repotagscache.set(None)
 
     # Mercurial earlier than 0.6 (or d7df759d0e97) logically reaches
     # this point, but it isn't so problematic, because:
     # - repo.tags of such Mercurial isn't "callable", and repo.tags()
     #   in perftags() causes failure soon
     # - perf.py itself has been available since 1.1 (or eb240755386d)
     raise error.Abort(("tags API of this hg command is unknown"))
 
 # perf commands
 
 @command('perfwalk', formatteropts)
 def perfwalk(ui, repo, *pats, **opts):
     timer, fm = gettimer(ui, opts)
     try:
         m = scmutil.match(repo[None], pats, {})
         timer(lambda: len(list(repo.dirstate.walk(m, [], True, False))))
     except Exception:
         try:
             m = scmutil.match(repo[None], pats, {})
             timer(lambda: len([b for a, b, c in repo.dirstate.statwalk([], m)]))
         except Exception:
             timer(lambda: len(list(cmdutil.walk(repo, pats, {}))))
     fm.end()
 
 @command('perfannotate', formatteropts)
 def perfannotate(ui, repo, f, **opts):
     timer, fm = gettimer(ui, opts)
     fc = repo['.'][f]
     timer(lambda: len(fc.annotate(True)))
     fm.end()
 
 @command('perfstatus',
          [('u', 'unknown', False,
            'ask status to look for unknown files')] + formatteropts)
 def perfstatus(ui, repo, **opts):
     #m = match.always(repo.root, repo.getcwd())
     #timer(lambda: sum(map(len, repo.dirstate.status(m, [], False, False,
     #                                                False))))
     timer, fm = gettimer(ui, opts)
     timer(lambda: sum(map(len, repo.status(unknown=opts['unknown']))))
     fm.end()
 
 @command('perfaddremove', formatteropts)
 def perfaddremove(ui, repo, **opts):
     timer, fm = gettimer(ui, opts)
     try:
         oldquiet = repo.ui.quiet
         repo.ui.quiet = True
         matcher = scmutil.match(repo[None])
         timer(lambda: scmutil.addremove(repo, matcher, "", dry_run=True))
     finally:
         repo.ui.quiet = oldquiet
         fm.end()
 
 def clearcaches(cl):
     # behave somewhat consistently across internal API changes
     if util.safehasattr(cl, 'clearcaches'):
         cl.clearcaches()
     elif util.safehasattr(cl, '_nodecache'):
         from mercurial.node import nullid, nullrev
         cl._nodecache = {nullid: nullrev}
         cl._nodepos = None
 
 @command('perfheads', formatteropts)
 def perfheads(ui, repo, **opts):
     timer, fm = gettimer(ui, opts)
     cl = repo.changelog
     def d():
         len(cl.headrevs())
         clearcaches(cl)
     timer(d)
     fm.end()
 
 @command('perftags', formatteropts)
 def perftags(ui, repo, **opts):
     import mercurial.changelog
     import mercurial.manifest
     timer, fm = gettimer(ui, opts)
     svfs = getsvfs(repo)
     repocleartagscache = repocleartagscachefunc(repo)
     def t():
         repo.changelog = mercurial.changelog.changelog(svfs)
         repo.manifestlog = mercurial.manifest.manifestlog(svfs, repo)
         repocleartagscache()
         return len(repo.tags())
     timer(t)
     fm.end()
 
 @command('perfancestors', formatteropts)
 def perfancestors(ui, repo, **opts):
     timer, fm = gettimer(ui, opts)
     heads = repo.changelog.headrevs()
     def d():
         for a in repo.changelog.ancestors(heads):
             pass
     timer(d)
     fm.end()
 
 @command('perfancestorset', formatteropts)
 def perfancestorset(ui, repo, revset, **opts):
     timer, fm = gettimer(ui, opts)
     revs = repo.revs(revset)
     heads = repo.changelog.headrevs()
     def d():
         s = repo.changelog.ancestors(heads)
         for rev in revs:
             rev in s
     timer(d)
     fm.end()
 
 @command('perfchangegroupchangelog', formatteropts +
          [('', 'version', '02', 'changegroup version'),
           ('r', 'rev', '', 'revisions to add to changegroup')])
 def perfchangegroupchangelog(ui, repo, version='02', rev=None, **opts):
     """Benchmark producing a changelog group for a changegroup.
 
     This measures the time spent processing the changelog during a
     bundle operation. This occurs during `hg bundle` and on a server
     processing a `getbundle` wire protocol request (handles clones
     and pull requests).
 
     By default, all revisions are added to the changegroup.
     """
     cl = repo.changelog
     revs = [cl.lookup(r) for r in repo.revs(rev or 'all()')]
     bundler = changegroup.getbundler(version, repo)
 
     def lookup(node):
         # The real bundler reads the revision in order to access the
         # manifest node and files list. Do that here.
         cl.read(node)
         return node
 
     def d():
         for chunk in bundler.group(revs, cl, lookup):
             pass
 
     timer, fm = gettimer(ui, opts)
     timer(d)
     fm.end()
 
 @command('perfdirs', formatteropts)
 def perfdirs(ui, repo, **opts):
     timer, fm = gettimer(ui, opts)
     dirstate = repo.dirstate
     'a' in dirstate
     def d():
         dirstate.dirs()
         del dirstate._dirs
     timer(d)
     fm.end()
 
 @command('perfdirstate', formatteropts)
 def perfdirstate(ui, repo, **opts):
     timer, fm = gettimer(ui, opts)
     "a" in repo.dirstate
     def d():
         repo.dirstate.invalidate()
         "a" in repo.dirstate
     timer(d)
     fm.end()
 
 @command('perfdirstatedirs', formatteropts)
 def perfdirstatedirs(ui, repo, **opts):
     timer, fm = gettimer(ui, opts)
     "a" in repo.dirstate
     def d():
         "a" in repo.dirstate._dirs
         del repo.dirstate._dirs
     timer(d)
     fm.end()
 
 @command('perfdirstatefoldmap', formatteropts)
 def perfdirstatefoldmap(ui, repo, **opts):
     timer, fm = gettimer(ui, opts)
     dirstate = repo.dirstate
     'a' in dirstate
     def d():
         dirstate._filefoldmap.get('a')
         del dirstate._filefoldmap
     timer(d)
     fm.end()
 
 @command('perfdirfoldmap', formatteropts)
 def perfdirfoldmap(ui, repo, **opts):
     timer, fm = gettimer(ui, opts)
     dirstate = repo.dirstate
     'a' in dirstate
     def d():
         dirstate._dirfoldmap.get('a')
         del dirstate._dirfoldmap
         del dirstate._dirs
     timer(d)
     fm.end()
 
 @command('perfdirstatewrite', formatteropts)
 def perfdirstatewrite(ui, repo, **opts):
     timer, fm = gettimer(ui, opts)
     ds = repo.dirstate
     "a" in ds
     def d():
         ds._dirty = True
         ds.write(repo.currenttransaction())
     timer(d)
     fm.end()
 
 @command('perfmergecalculate',
          [('r', 'rev', '.', 'rev to merge against')] + formatteropts)
 def perfmergecalculate(ui, repo, rev, **opts):
     timer, fm = gettimer(ui, opts)
     wctx = repo[None]
     rctx = scmutil.revsingle(repo, rev, rev)
     ancestor = wctx.ancestor(rctx)
     # we don't want working dir files to be stat'd in the benchmark, so prime
     # that cache
     wctx.dirty()
     def d():
         # acceptremote is True because we don't want prompts in the middle of
         # our benchmark
         merge.calculateupdates(repo, wctx, rctx, [ancestor], False, False,
                                acceptremote=True, followcopies=True)
     timer(d)
     fm.end()
 
 @command('perfpathcopies', [], "REV REV")
 def perfpathcopies(ui, repo, rev1, rev2, **opts):
     timer, fm = gettimer(ui, opts)
     ctx1 = scmutil.revsingle(repo, rev1, rev1)
     ctx2 = scmutil.revsingle(repo, rev2, rev2)
     def d():
         copies.pathcopies(ctx1, ctx2)
     timer(d)
     fm.end()
 
 @command('perfmanifest', [], 'REV')
 def perfmanifest(ui, repo, rev, **opts):
     timer, fm = gettimer(ui, opts)
     ctx = scmutil.revsingle(repo, rev, rev)
     t = ctx.manifestnode()
     def d():
-        repo.manifest.clearcaches()
+        repo.manifestlog.clearcaches()
         repo.manifestlog[t].read()
     timer(d)
     fm.end()
 
 @command('perfchangeset', formatteropts)
 def perfchangeset(ui, repo, rev, **opts):
     timer, fm = gettimer(ui, opts)
     n = repo[rev].node()
     def d():
         repo.changelog.read(n)
         #repo.changelog._cache = None
     timer(d)
     fm.end()
 
 @command('perfindex', formatteropts)
 def perfindex(ui, repo, **opts):
     import mercurial.revlog
     timer, fm = gettimer(ui, opts)
     mercurial.revlog._prereadsize = 2**24 # disable lazy parser in old hg
     n = repo["tip"].node()
     svfs = getsvfs(repo)
     def d():
         cl = mercurial.revlog.revlog(svfs, "00changelog.i")
         cl.rev(n)
     timer(d)
     fm.end()
 
 @command('perfstartup', formatteropts)
 def perfstartup(ui, repo, **opts):
     timer, fm = gettimer(ui, opts)
     cmd = sys.argv[0]
     def d():
         if os.name != 'nt':
             os.system("HGRCPATH= %s version -q > /dev/null" % cmd)
         else:
             os.environ['HGRCPATH'] = ''
             os.system("%s version -q > NUL" % cmd)
     timer(d)
     fm.end()
 
 @command('perfparents', formatteropts)
 def perfparents(ui, repo, **opts):
     timer, fm = gettimer(ui, opts)
     # control the number of commits perfparents iterates over
     # experimental config: perf.parentscount
     count = getint(ui, "perf", "parentscount", 1000)
     if len(repo.changelog) < count:
         raise error.Abort("repo needs %d commits for this test" % count)
     repo = repo.unfiltered()
     nl = [repo.changelog.node(i) for i in xrange(count)]
     def d():
         for n in nl:
             repo.changelog.parents(n)
     timer(d)
     fm.end()
 
 @command('perfctxfiles', formatteropts)
 def perfctxfiles(ui, repo, x, **opts):
     x = int(x)
     timer, fm = gettimer(ui, opts)
     def d():
         len(repo[x].files())
     timer(d)
     fm.end()
 
 @command('perfrawfiles', formatteropts)
 def perfrawfiles(ui, repo, x, **opts):
     x = int(x)
     timer, fm = gettimer(ui, opts)
     cl = repo.changelog
     def d():
         len(cl.read(x)[3])
     timer(d)
     fm.end()
 
 @command('perflookup', formatteropts)
 def perflookup(ui, repo, rev, **opts):
     timer, fm = gettimer(ui, opts)
     timer(lambda: len(repo.lookup(rev)))
     fm.end()
 
 @command('perfrevrange', formatteropts)
 def perfrevrange(ui, repo, *specs, **opts):
     timer, fm = gettimer(ui, opts)
     revrange = scmutil.revrange
     timer(lambda: len(revrange(repo, specs)))
     fm.end()
 
 @command('perfnodelookup', formatteropts)
 def perfnodelookup(ui, repo, rev, **opts):
     timer, fm = gettimer(ui, opts)
     import mercurial.revlog
     mercurial.revlog._prereadsize = 2**24 # disable lazy parser in old hg
     n = repo[rev].node()
     cl = mercurial.revlog.revlog(getsvfs(repo), "00changelog.i")
     def d():
         cl.rev(n)
         clearcaches(cl)
     timer(d)
     fm.end()
 
 @command('perflog',
          [('', 'rename', False, 'ask log to follow renames')] + formatteropts)
 def perflog(ui, repo, rev=None, **opts):
     if rev is None:
         rev=[]
     timer, fm = gettimer(ui, opts)
     ui.pushbuffer()
     timer(lambda: commands.log(ui, repo, rev=rev, date='', user='',
                                copies=opts.get('rename')))
     ui.popbuffer()
     fm.end()
 
 @command('perfmoonwalk', formatteropts)
 def perfmoonwalk(ui, repo, **opts):
     """benchmark walking the changelog backwards
 
     This also loads the changelog data for each revision in the changelog.
     """
     timer, fm = gettimer(ui, opts)
     def moonwalk():
         for i in xrange(len(repo), -1, -1):
             ctx = repo[i]
             ctx.branch() # read changelog data (in addition to the index)
     timer(moonwalk)
     fm.end()
 
 @command('perftemplating', formatteropts)
 def perftemplating(ui, repo, rev=None, **opts):
     if rev is None:
         rev=[]
     timer, fm = gettimer(ui, opts)
     ui.pushbuffer()
     timer(lambda: commands.log(ui, repo, rev=rev, date='', user='',
                                template='{date|shortdate} [{rev}:{node|short}]'
                                ' {author|person}: {desc|firstline}\n'))
     ui.popbuffer()
     fm.end()
 
 @command('perfcca', formatteropts)
 def perfcca(ui, repo, **opts):
     timer, fm = gettimer(ui, opts)
     timer(lambda: scmutil.casecollisionauditor(ui, False, repo.dirstate))
     fm.end()
 
 @command('perffncacheload', formatteropts)
 def perffncacheload(ui, repo, **opts):
     timer, fm = gettimer(ui, opts)
     s = repo.store
     def d():
         s.fncache._load()
     timer(d)
     fm.end()
 
 @command('perffncachewrite', formatteropts)
 def perffncachewrite(ui, repo, **opts):
     timer, fm = gettimer(ui, opts)
     s = repo.store
     s.fncache._load()
     lock = repo.lock()
     tr = repo.transaction('perffncachewrite')
     def d():
         s.fncache._dirty = True
         s.fncache.write(tr)
     timer(d)
     tr.close()
     lock.release()
     fm.end()
 
 @command('perffncacheencode', formatteropts)
 def perffncacheencode(ui, repo, **opts):
     timer, fm = gettimer(ui, opts)
     s = repo.store
     s.fncache._load()
     def d():
         for p in s.fncache.entries:
             s.encode(p)
     timer(d)
     fm.end()
 
 @command('perfbdiff', revlogopts + formatteropts + [
     ('', 'count', 1, 'number of revisions to test (when using --startrev)'),
     ('', 'alldata', False, 'test bdiffs for all associated revisions')],
     '-c|-m|FILE REV')
 def perfbdiff(ui, repo, file_, rev=None, count=None, **opts):
     """benchmark a bdiff between revisions
 
     By default, benchmark a bdiff between its delta parent and itself.
 
     With ``--count``, benchmark bdiffs between delta parents and self for N
     revisions starting at the specified revision.
 
     With ``--alldata``, assume the requested revision is a changeset and
     measure bdiffs for all changes related to that changeset (manifest
     and filelogs).
     """
     if opts['alldata']:
         opts['changelog'] = True
 
     if opts.get('changelog') or opts.get('manifest'):
         file_, rev = None, file_
     elif rev is None:
         raise error.CommandError('perfbdiff', 'invalid arguments')
 
     textpairs = []
 
     r = cmdutil.openrevlog(repo, 'perfbdiff', file_, opts)
 
     startrev = r.rev(r.lookup(rev))
     for rev in range(startrev, min(startrev + count, len(r) - 1)):
         if opts['alldata']:
             # Load revisions associated with changeset.
             ctx = repo[rev]
             mtext = repo.manifest.revision(ctx.manifestnode())
             for pctx in ctx.parents():
                 pman = repo.manifest.revision(pctx.manifestnode())
                 textpairs.append((pman, mtext))
 
             # Load filelog revisions by iterating manifest delta.
             man = ctx.manifest()
             pman = ctx.p1().manifest()
             for filename, change in pman.diff(man).items():
                 fctx = repo.file(filename)
                 f1 = fctx.revision(change[0][0] or -1)
                 f2 = fctx.revision(change[1][0] or -1)
                 textpairs.append((f1, f2))
         else:
             dp = r.deltaparent(rev)
             textpairs.append((r.revision(dp), r.revision(rev)))
 
     def d():
         for pair in textpairs:
             bdiff.bdiff(*pair)
 
     timer, fm = gettimer(ui, opts)
     timer(d)
     fm.end()
 
 @command('perfdiffwd', formatteropts)
 def perfdiffwd(ui, repo, **opts):
     """Profile diff of working directory changes"""
     timer, fm = gettimer(ui, opts)
     options = {
         'w': 'ignore_all_space',
         'b': 'ignore_space_change',
         'B': 'ignore_blank_lines',
         }
 
     for diffopt in ('', 'w', 'b', 'B', 'wB'):
         opts = dict((options[c], '1') for c in diffopt)
         def d():
             ui.pushbuffer()
             commands.diff(ui, repo, **opts)
             ui.popbuffer()
         title = 'diffopts: %s' % (diffopt and ('-' + diffopt) or 'none')
         timer(d, title)
     fm.end()
 
 @command('perfrevlog', revlogopts + formatteropts +
          [('d', 'dist', 100, 'distance between the revisions'),
           ('s', 'startrev', 0, 'revision to start reading at'),
           ('', 'reverse', False, 'read in reverse')],
          '-c|-m|FILE')
 def perfrevlog(ui, repo, file_=None, startrev=0, reverse=False, **opts):
     """Benchmark reading a series of revisions from a revlog.
 
     By default, we read every ``-d/--dist`` revision from 0 to tip of
     the specified revlog.
 
     The start revision can be defined via ``-s/--startrev``.
     """
     timer, fm = gettimer(ui, opts)
     _len = getlen(ui)
 
     def d():
         r = cmdutil.openrevlog(repo, 'perfrevlog', file_, opts)
 
         startrev = 0
         endrev = _len(r)
         dist = opts['dist']
 
         if reverse:
             startrev, endrev = endrev, startrev
             dist = -1 * dist
 
         for x in xrange(startrev, endrev, dist):
             r.revision(r.node(x))
 
     timer(d)
     fm.end()
 
 @command('perfrevlogrevision', revlogopts + formatteropts +
          [('', 'cache', False, 'use caches instead of clearing')],
          '-c|-m|FILE REV')
 def perfrevlogrevision(ui, repo, file_, rev=None, cache=None, **opts):
     """Benchmark obtaining a revlog revision.
 
     Obtaining a revlog revision consists of roughly the following steps:
 
     1. Compute the delta chain
     2. Obtain the raw chunks for that delta chain
     3. Decompress each raw chunk
     4. Apply binary patches to obtain fulltext
     5. Verify hash of fulltext
 
     This command measures the time spent in each of these phases.
     """
     if opts.get('changelog') or opts.get('manifest'):
         file_, rev = None, file_
     elif rev is None:
         raise error.CommandError('perfrevlogrevision', 'invalid arguments')
 
     r = cmdutil.openrevlog(repo, 'perfrevlogrevision', file_, opts)
     node = r.lookup(rev)
     rev = r.rev(node)
 
     def dodeltachain(rev):
         if not cache:
             r.clearcaches()
         r._deltachain(rev)
 
     def doread(chain):
         if not cache:
             r.clearcaches()
         r._chunkraw(chain[0], chain[-1])
 
     def dodecompress(data, chain):
         if not cache:
             r.clearcaches()
 
         start = r.start
         length = r.length
         inline = r._inline
         iosize = r._io.size
         buffer = util.buffer
         offset = start(chain[0])
 
         for rev in chain:
             chunkstart = start(rev)
             if inline:
                 chunkstart += (rev + 1) * iosize
             chunklength = length(rev)
             b = buffer(data, chunkstart - offset, chunklength)
             revlog.decompress(b)
 
     def dopatch(text, bins):
         if not cache:
             r.clearcaches()
         mdiff.patches(text, bins)
 
     def dohash(text):
         if not cache:
             r.clearcaches()
         r._checkhash(text, node, rev)
 
     def dorevision():
         if not cache:
             r.clearcaches()
         r.revision(node)
 
     chain = r._deltachain(rev)[0]
     data = r._chunkraw(chain[0], chain[-1])[1]
     bins = r._chunks(chain)
     text = str(bins[0])
     bins = bins[1:]
     text = mdiff.patches(text, bins)
 
     benches = [
         (lambda: dorevision(), 'full'),
         (lambda: dodeltachain(rev), 'deltachain'),
         (lambda: doread(chain), 'read'),
         (lambda: dodecompress(data, chain), 'decompress'),
         (lambda: dopatch(text, bins), 'patch'),
         (lambda: dohash(text), 'hash'),
     ]
 
     for fn, title in benches:
         timer, fm = gettimer(ui, opts)
         timer(fn, title=title)
         fm.end()
 
 @command('perfrevset',
          [('C', 'clear', False, 'clear volatile cache between each call.'),
           ('', 'contexts', False, 'obtain changectx for each revision')]
          + formatteropts, "REVSET")
 def perfrevset(ui, repo, expr, clear=False, contexts=False, **opts):
     """benchmark the execution time of a revset
 
     Use the --clean option if need to evaluate the impact of build volatile
     revisions set cache on the revset execution. Volatile cache hold filtered
     and obsolete related cache."""
     timer, fm = gettimer(ui, opts)
     def d():
         if clear:
             repo.invalidatevolatilesets()
         if contexts:
             for ctx in repo.set(expr): pass
         else:
             for r in repo.revs(expr): pass
     timer(d)
     fm.end()
 
 @command('perfvolatilesets', formatteropts)
 def perfvolatilesets(ui, repo, *names, **opts):
     """benchmark the computation of various volatile set
 
     Volatile set computes element related to filtering and obsolescence."""
     timer, fm = gettimer(ui, opts)
     repo = repo.unfiltered()
 
     def getobs(name):
         def d():
             repo.invalidatevolatilesets()
             obsolete.getrevs(repo, name)
         return d
 
     allobs = sorted(obsolete.cachefuncs)
     if names:
         allobs = [n for n in allobs if n in names]
 
     for name in allobs:
         timer(getobs(name), title=name)
 
     def getfiltered(name):
         def d():
             repo.invalidatevolatilesets()
             repoview.filterrevs(repo, name)
         return d
 
     allfilter = sorted(repoview.filtertable)
     if names:
         allfilter = [n for n in allfilter if n in names]
 
     for name in allfilter:
         timer(getfiltered(name), title=name)
     fm.end()
 
 @command('perfbranchmap',
          [('f', 'full', False,
            'Includes build time of subset'),
          ] + formatteropts)
 def perfbranchmap(ui, repo, full=False, **opts):
     """benchmark the update of a branchmap
 
     This benchmarks the full repo.branchmap() call with read and write disabled
     """
     timer, fm = gettimer(ui, opts)
     def getbranchmap(filtername):
         """generate a benchmark function for the filtername"""
         if filtername is None:
             view = repo
         else:
             view = repo.filtered(filtername)
         def d():
             if full:
                 view._branchcaches.clear()
             else:
                 view._branchcaches.pop(filtername, None)
             view.branchmap()
         return d
     # add filter in smaller subset to bigger subset
     possiblefilters = set(repoview.filtertable)
     subsettable = getbranchmapsubsettable()
     allfilters = []
     while possiblefilters:
         for name in possiblefilters:
             subset = subsettable.get(name)
             if subset not in possiblefilters:
                 break
         else:
             assert False, 'subset cycle %s!' % possiblefilters
         allfilters.append(name)
         possiblefilters.remove(name)
 
     # warm the cache
     if not full:
         for name in allfilters:
             repo.filtered(name).branchmap()
     # add unfiltered
     allfilters.append(None)
 
     branchcacheread = safeattrsetter(branchmap, 'read')
     branchcachewrite = safeattrsetter(branchmap.branchcache, 'write')
     branchcacheread.set(lambda repo: None)
     branchcachewrite.set(lambda bc, repo: None)
     try:
         for name in allfilters:
             timer(getbranchmap(name), title=str(name))
     finally:
         branchcacheread.restore()
         branchcachewrite.restore()
     fm.end()
 
 @command('perfloadmarkers')
 def perfloadmarkers(ui, repo):
     """benchmark the time to parse the on-disk markers for a repo
 
     Result is the number of markers in the repo."""
     timer, fm = gettimer(ui)
     svfs = getsvfs(repo)
     timer(lambda: len(obsolete.obsstore(svfs)))
     fm.end()
 
 @command('perflrucachedict', formatteropts +
     [('', 'size', 4, 'size of cache'),
      ('', 'gets', 10000, 'number of key lookups'),
      ('', 'sets', 10000, 'number of key sets'),
      ('', 'mixed', 10000, 'number of mixed mode operations'),
      ('', 'mixedgetfreq', 50, 'frequency of get vs set ops in mixed mode')],
     norepo=True)
 def perflrucache(ui, size=4, gets=10000, sets=10000, mixed=10000,
                  mixedgetfreq=50, **opts):
     def doinit():
         for i in xrange(10000):
             util.lrucachedict(size)
 
     values = []
     for i in xrange(size):
         values.append(random.randint(0, sys.maxint))
 
     # Get mode fills the cache and tests raw lookup performance with no
     # eviction.
     getseq = []
     for i in xrange(gets):
         getseq.append(random.choice(values))
 
     def dogets():
         d = util.lrucachedict(size)
         for v in values:
             d[v] = v
         for key in getseq:
             value = d[key]
             value # silence pyflakes warning
 
     # Set mode tests insertion speed with cache eviction.
     setseq = []
     for i in xrange(sets):
         setseq.append(random.randint(0, sys.maxint))
 
     def dosets():
         d = util.lrucachedict(size)
         for v in setseq:
             d[v] = v
 
     # Mixed mode randomly performs gets and sets with eviction.
     mixedops = []
     for i in xrange(mixed):
         r = random.randint(0, 100)
         if r < mixedgetfreq:
             op = 0
         else:
             op = 1
 
         mixedops.append((op, random.randint(0, size * 2)))
 
     def domixed():
         d = util.lrucachedict(size)
 
         for op, v in mixedops:
             if op == 0:
                 try:
                     d[v]
                 except KeyError:
                     pass
             else:
                 d[v] = v
 
     benches = [
         (doinit, 'init'),
         (dogets, 'gets'),
         (dosets, 'sets'),
         (domixed, 'mixed')
     ]
 
     for fn, title in benches:
         timer, fm = gettimer(ui, opts)
         timer(fn, title=title)
         fm.end()
 
 def uisetup(ui):
     if (util.safehasattr(cmdutil, 'openrevlog') and
         not util.safehasattr(commands, 'debugrevlogopts')):
         # for "historical portability":
         # In this case, Mercurial should be 1.9 (or a79fea6b3e77) -
         # 3.7 (or 5606f7d0d063). Therefore, '--dir' option for
         # openrevlog() should cause failure, because it has been
         # available since 3.5 (or 49c583ca48c4).
         def openrevlog(orig, repo, cmd, file_, opts):
             if opts.get('dir') and not util.safehasattr(repo, 'dirlog'):
                 raise error.Abort("This version doesn't support --dir option",
                                   hint="use 3.5 or later")
             return orig(repo, cmd, file_, opts)
         extensions.wrapfunction(cmdutil, 'openrevlog', openrevlog)

mercurial/manifest.py

@@ -1,1599 +1,1599
 # manifest.py - manifest revision class for mercurial
 #
 # Copyright 2005-2007 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 array
 import heapq
 import os
 import struct
 
 from .i18n import _
 from . import (
     error,
     mdiff,
     parsers,
     revlog,
     util,
 )
 
 propertycache = util.propertycache
 
 def _parsev1(data):
     # This method does a little bit of excessive-looking
     # precondition checking. This is so that the behavior of this
     # class exactly matches its C counterpart to try and help
     # prevent surprise breakage for anyone that develops against
     # the pure version.
     if data and data[-1] != '\n':
         raise ValueError('Manifest did not end in a newline.')
     prev = None
     for l in data.splitlines():
         if prev is not None and prev > l:
             raise ValueError('Manifest lines not in sorted order.')
         prev = l
         f, n = l.split('\0')
         if len(n) > 40:
             yield f, revlog.bin(n[:40]), n[40:]
         else:
             yield f, revlog.bin(n), ''
 
 def _parsev2(data):
     metadataend = data.find('\n')
     # Just ignore metadata for now
     pos = metadataend + 1
     prevf = ''
     while pos < len(data):
         end = data.find('\n', pos + 1) # +1 to skip stem length byte
         if end == -1:
             raise ValueError('Manifest ended with incomplete file entry.')
         stemlen = ord(data[pos])
         items = data[pos + 1:end].split('\0')
         f = prevf[:stemlen] + items[0]
         if prevf > f:
             raise ValueError('Manifest entries not in sorted order.')
         fl = items[1]
         # Just ignore metadata (items[2:] for now)
         n = data[end + 1:end + 21]
         yield f, n, fl
         pos = end + 22
         prevf = f
 
 def _parse(data):
     """Generates (path, node, flags) tuples from a manifest text"""
     if data.startswith('\0'):
         return iter(_parsev2(data))
     else:
         return iter(_parsev1(data))
 
 def _text(it, usemanifestv2):
     """Given an iterator over (path, node, flags) tuples, returns a manifest
     text"""
     if usemanifestv2:
         return _textv2(it)
     else:
         return _textv1(it)
 
 def _textv1(it):
     files = []
     lines = []
     _hex = revlog.hex
     for f, n, fl in it:
         files.append(f)
         # if this is changed to support newlines in filenames,
         # be sure to check the templates/ dir again (especially *-raw.tmpl)
         lines.append("%s\0%s%s\n" % (f, _hex(n), fl))
 
     _checkforbidden(files)
     return ''.join(lines)
 
 def _textv2(it):
     files = []
     lines = ['\0\n']
     prevf = ''
     for f, n, fl in it:
         files.append(f)
         stem = os.path.commonprefix([prevf, f])
         stemlen = min(len(stem), 255)
         lines.append("%c%s\0%s\n%s\n" % (stemlen, f[stemlen:], fl, n))
         prevf = f
     _checkforbidden(files)
     return ''.join(lines)
 
 class lazymanifestiter(object):
     def __init__(self, lm):
         self.pos = 0
         self.lm = lm
 
     def __iter__(self):
         return self
 
     def next(self):
         try:
             data, pos = self.lm._get(self.pos)
         except IndexError:
             raise StopIteration
         if pos == -1:
             self.pos += 1
             return data[0]
         self.pos += 1
         zeropos = data.find('\x00', pos)
         return data[pos:zeropos]
 
 class lazymanifestiterentries(object):
     def __init__(self, lm):
         self.lm = lm
         self.pos = 0
 
     def __iter__(self):
         return self
 
     def next(self):
         try:
             data, pos = self.lm._get(self.pos)
         except IndexError:
             raise StopIteration
         if pos == -1:
             self.pos += 1
             return data
         zeropos = data.find('\x00', pos)
         hashval = unhexlify(data, self.lm.extrainfo[self.pos],
                             zeropos + 1, 40)
         flags = self.lm._getflags(data, self.pos, zeropos)
         self.pos += 1
         return (data[pos:zeropos], hashval, flags)
 
 def unhexlify(data, extra, pos, length):
     s = data[pos:pos + length].decode('hex')
     if extra:
         s += chr(extra & 0xff)
     return s
 
 def _cmp(a, b):
     return (a > b) - (a < b)
 
 class _lazymanifest(object):
     def __init__(self, data, positions=None, extrainfo=None, extradata=None):
         if positions is None:
             self.positions = self.findlines(data)
             self.extrainfo = [0] * len(self.positions)
             self.data = data
             self.extradata = []
         else:
             self.positions = positions[:]
             self.extrainfo = extrainfo[:]
             self.extradata = extradata[:]
             self.data = data
 
     def findlines(self, data):
         if not data:
             return []
         pos = data.find("\n")
         if pos == -1 or data[-1] != '\n':
             raise ValueError("Manifest did not end in a newline.")
         positions = [0]
         prev = data[:data.find('\x00')]
         while pos < len(data) - 1 and pos != -1:
             positions.append(pos + 1)
             nexts = data[pos + 1:data.find('\x00', pos + 1)]
             if nexts < prev:
                 raise ValueError("Manifest lines not in sorted order.")
             prev = nexts
             pos = data.find("\n", pos + 1)
         return positions
 
     def _get(self, index):
         # get the position encoded in pos:
         #   positive number is an index in 'data'
         #   negative number is in extrapieces
         pos = self.positions[index]
         if pos >= 0:
             return self.data, pos
         return self.extradata[-pos - 1], -1
 
     def _getkey(self, pos):
         if pos >= 0:
             return self.data[pos:self.data.find('\x00', pos + 1)]
         return self.extradata[-pos - 1][0]
 
     def bsearch(self, key):
         first = 0
         last = len(self.positions) - 1
 
         while first <= last:
             midpoint = (first + last)//2
             nextpos = self.positions[midpoint]
             candidate = self._getkey(nextpos)
             r = _cmp(key, candidate)
             if r == 0:
                 return midpoint
             else:
                 if r < 0:
                     last = midpoint - 1
                 else:
                     first = midpoint + 1
         return -1
 
     def bsearch2(self, key):
         # same as the above, but will always return the position
         # done for performance reasons
         first = 0
         last = len(self.positions) - 1
 
         while first <= last:
             midpoint = (first + last)//2
             nextpos = self.positions[midpoint]
             candidate = self._getkey(nextpos)
             r = _cmp(key, candidate)
             if r == 0:
                 return (midpoint, True)
             else:
                 if r < 0:
                     last = midpoint - 1
                 else:
                     first = midpoint + 1
         return (first, False)
 
     def __contains__(self, key):
         return self.bsearch(key) != -1
 
     def _getflags(self, data, needle, pos):
         start = pos + 41
         end = data.find("\n", start)
         if end == -1:
             end = len(data) - 1
         if start == end:
             return ''
         return self.data[start:end]
 
     def __getitem__(self, key):
         if not isinstance(key, str):
             raise TypeError("getitem: manifest keys must be a string.")
         needle = self.bsearch(key)
         if needle == -1:
             raise KeyError
         data, pos = self._get(needle)
         if pos == -1:
             return (data[1], data[2])
         zeropos = data.find('\x00', pos)
         assert 0 <= needle <= len(self.positions)
         assert len(self.extrainfo) == len(self.positions)
         hashval = unhexlify(data, self.extrainfo[needle], zeropos + 1, 40)
         flags = self._getflags(data, needle, zeropos)
         return (hashval, flags)
 
     def __delitem__(self, key):
         needle, found = self.bsearch2(key)
         if not found:
             raise KeyError
         cur = self.positions[needle]
         self.positions = self.positions[:needle] + self.positions[needle + 1:]
         self.extrainfo = self.extrainfo[:needle] + self.extrainfo[needle + 1:]
         if cur >= 0:
             self.data = self.data[:cur] + '\x00' + self.data[cur + 1:]
 
     def __setitem__(self, key, value):
         if not isinstance(key, str):
             raise TypeError("setitem: manifest keys must be a string.")
         if not isinstance(value, tuple) or len(value) != 2:
             raise TypeError("Manifest values must be a tuple of (node, flags).")
         hashval = value[0]
         if not isinstance(hashval, str) or not 20 <= len(hashval) <= 22:
             raise TypeError("node must be a 20-byte string")
         flags = value[1]
         if len(hashval) == 22:
             hashval = hashval[:-1]
         if not isinstance(flags, str) or len(flags) > 1:
             raise TypeError("flags must a 0 or 1 byte string, got %r", flags)
         needle, found = self.bsearch2(key)
         if found:
             # put the item
             pos = self.positions[needle]
             if pos < 0:
                 self.extradata[-pos - 1] = (key, hashval, value[1])
             else:
                 # just don't bother
                 self.extradata.append((key, hashval, value[1]))
                 self.positions[needle] = -len(self.extradata)
         else:
             # not found, put it in with extra positions
             self.extradata.append((key, hashval, value[1]))
             self.positions = (self.positions[:needle] + [-len(self.extradata)]
                               + self.positions[needle:])
             self.extrainfo = (self.extrainfo[:needle] + [0] +
                               self.extrainfo[needle:])
 
     def copy(self):
         # XXX call _compact like in C?
         return _lazymanifest(self.data, self.positions, self.extrainfo,
             self.extradata)
 
     def _compact(self):
         # hopefully not called TOO often
         if len(self.extradata) == 0:
             return
         l = []
         last_cut = 0
         i = 0
         offset = 0
         self.extrainfo = [0] * len(self.positions)
         while i < len(self.positions):
             if self.positions[i] >= 0:
                 cur = self.positions[i]
                 last_cut = cur
                 while True:
                     self.positions[i] = offset
                     i += 1
                     if i == len(self.positions) or self.positions[i] < 0:
                         break
                     offset += self.positions[i] - cur
                     cur = self.positions[i]
                 end_cut = self.data.find('\n', cur)
                 if end_cut != -1:
                     end_cut += 1
                 offset += end_cut - cur
                 l.append(self.data[last_cut:end_cut])
             else:
                 while i < len(self.positions) and self.positions[i] < 0:
                     cur = self.positions[i]
                     t = self.extradata[-cur - 1]
                     l.append(self._pack(t))
                     self.positions[i] = offset
                     if len(t[1]) > 20:
                         self.extrainfo[i] = ord(t[1][21])
                     offset += len(l[-1])
                     i += 1
         self.data = ''.join(l)
         self.extradata = []
 
     def _pack(self, d):
         return d[0] + '\x00' + d[1][:20].encode('hex') + d[2] + '\n'
 
     def text(self):
         self._compact()
         return self.data
 
     def diff(self, m2, clean=False):
         '''Finds changes between the current manifest and m2.'''
         # XXX think whether efficiency matters here
         diff = {}
 
         for fn, e1, flags in self.iterentries():
             if fn not in m2:
                 diff[fn] = (e1, flags), (None, '')
             else:
                 e2 = m2[fn]
                 if (e1, flags) != e2:
                     diff[fn] = (e1, flags), e2
                 elif clean:
                     diff[fn] = None
 
         for fn, e2, flags in m2.iterentries():
             if fn not in self:
                 diff[fn] = (None, ''), (e2, flags)
 
         return diff
 
     def iterentries(self):
         return lazymanifestiterentries(self)
 
     def iterkeys(self):
         return lazymanifestiter(self)
 
     def __iter__(self):
         return lazymanifestiter(self)
 
     def __len__(self):
         return len(self.positions)
 
     def filtercopy(self, filterfn):
         # XXX should be optimized
         c = _lazymanifest('')
         for f, n, fl in self.iterentries():
             if filterfn(f):
                 c[f] = n, fl
         return c
 
 try:
     _lazymanifest = parsers.lazymanifest
 except AttributeError:
     pass
 
 class manifestdict(object):
     def __init__(self, data=''):
         if data.startswith('\0'):
             #_lazymanifest can not parse v2
             self._lm = _lazymanifest('')
             for f, n, fl in _parsev2(data):
                 self._lm[f] = n, fl
         else:
             self._lm = _lazymanifest(data)
 
     def __getitem__(self, key):
         return self._lm[key][0]
 
     def find(self, key):
         return self._lm[key]
 
     def __len__(self):
         return len(self._lm)
 
     def __nonzero__(self):
         # nonzero is covered by the __len__ function, but implementing it here
         # makes it easier for extensions to override.
         return len(self._lm) != 0
 
     def __setitem__(self, key, node):
         self._lm[key] = node, self.flags(key, '')
 
     def __contains__(self, key):
         return key in self._lm
 
     def __delitem__(self, key):
         del self._lm[key]
 
     def __iter__(self):
         return self._lm.__iter__()
 
     def iterkeys(self):
         return self._lm.iterkeys()
 
     def keys(self):
         return list(self.iterkeys())
 
     def filesnotin(self, m2):
         '''Set of files in this manifest that are not in the other'''
         diff = self.diff(m2)
         files = set(filepath
                     for filepath, hashflags in diff.iteritems()
                     if hashflags[1][0] is None)
         return files
 
     @propertycache
     def _dirs(self):
         return util.dirs(self)
 
     def dirs(self):
         return self._dirs
 
     def hasdir(self, dir):
         return dir in self._dirs
 
     def _filesfastpath(self, match):
         '''Checks whether we can correctly and quickly iterate over matcher
         files instead of over manifest files.'''
         files = match.files()
         return (len(files) < 100 and (match.isexact() or
             (match.prefix() and all(fn in self for fn in files))))
 
     def walk(self, match):
         '''Generates matching file names.
 
         Equivalent to manifest.matches(match).iterkeys(), but without creating
         an entirely new manifest.
 
         It also reports nonexistent files by marking them bad with match.bad().
         '''
         if match.always():
             for f in iter(self):
                 yield f
             return
 
         fset = set(match.files())
 
         # avoid the entire walk if we're only looking for specific files
         if self._filesfastpath(match):
             for fn in sorted(fset):
                 yield fn
             return
 
         for fn in self:
             if fn in fset:
                 # specified pattern is the exact name
                 fset.remove(fn)
             if match(fn):
                 yield fn
 
         # for dirstate.walk, files=['.'] means "walk the whole tree".
         # follow that here, too
         fset.discard('.')
 
         for fn in sorted(fset):
             if not self.hasdir(fn):
                 match.bad(fn, None)
 
     def matches(self, match):
         '''generate a new manifest filtered by the match argument'''
         if match.always():
             return self.copy()
 
         if self._filesfastpath(match):
             m = manifestdict()
             lm = self._lm
             for fn in match.files():
                 if fn in lm:
                     m._lm[fn] = lm[fn]
             return m
 
         m = manifestdict()
         m._lm = self._lm.filtercopy(match)
         return m
 
     def diff(self, m2, clean=False):
         '''Finds changes between the current manifest and m2.
 
         Args:
           m2: the manifest to which this manifest should be compared.
           clean: if true, include files unchanged between these manifests
                  with a None value in the returned dictionary.
 
         The result is returned as a dict with filename as key and
         values of the form ((n1,fl1),(n2,fl2)), where n1/n2 is the
         nodeid in the current/other manifest and fl1/fl2 is the flag
         in the current/other manifest. Where the file does not exist,
         the nodeid will be None and the flags will be the empty
         string.
         '''
         return self._lm.diff(m2._lm, clean)
 
     def setflag(self, key, flag):
         self._lm[key] = self[key], flag
 
     def get(self, key, default=None):
         try:
             return self._lm[key][0]
         except KeyError:
             return default
 
     def flags(self, key, default=''):
         try:
             return self._lm[key][1]
         except KeyError:
             return default
 
     def copy(self):
         c = manifestdict()
         c._lm = self._lm.copy()
         return c
 
     def iteritems(self):
         return (x[:2] for x in self._lm.iterentries())
 
     def iterentries(self):
         return self._lm.iterentries()
 
     def text(self, usemanifestv2=False):
         if usemanifestv2:
             return _textv2(self._lm.iterentries())
         else:
             # use (probably) native version for v1
             return self._lm.text()
 
     def fastdelta(self, base, changes):
         """Given a base manifest text as an array.array and a list of changes
         relative to that text, compute a delta that can be used by revlog.
         """
         delta = []
         dstart = None
         dend = None
         dline = [""]
         start = 0
         # zero copy representation of base as a buffer
         addbuf = util.buffer(base)
 
         changes = list(changes)
         if len(changes) < 1000:
             # start with a readonly loop that finds the offset of
             # each line and creates the deltas
             for f, todelete in changes:
                 # bs will either be the index of the item or the insert point
                 start, end = _msearch(addbuf, f, start)
                 if not todelete:
                     h, fl = self._lm[f]
                     l = "%s\0%s%s\n" % (f, revlog.hex(h), fl)
                 else:
                     if start == end:
                         # item we want to delete was not found, error out
                         raise AssertionError(
                                 _("failed to remove %s from manifest") % f)
                     l = ""
                 if dstart is not None and dstart <= start and dend >= start:
                     if dend < end:
                         dend = end
                     if l:
                         dline.append(l)
                 else:
                     if dstart is not None:
                         delta.append([dstart, dend, "".join(dline)])
                     dstart = start
                     dend = end
                     dline = [l]
 
             if dstart is not None:
                 delta.append([dstart, dend, "".join(dline)])
             # apply the delta to the base, and get a delta for addrevision
             deltatext, arraytext = _addlistdelta(base, delta)
         else:
             # For large changes, it's much cheaper to just build the text and
             # diff it.
             arraytext = array.array('c', self.text())
             deltatext = mdiff.textdiff(base, arraytext)
 
         return arraytext, deltatext
 
 def _msearch(m, s, lo=0, hi=None):
     '''return a tuple (start, end) that says where to find s within m.
 
     If the string is found m[start:end] are the line containing
     that string.  If start == end the string was not found and
     they indicate the proper sorted insertion point.
 
     m should be a buffer or a string
     s is a string'''
     def advance(i, c):
         while i < lenm and m[i] != c:
             i += 1
         return i
     if not s:
         return (lo, lo)
     lenm = len(m)
     if not hi:
         hi = lenm
     while lo < hi:
         mid = (lo + hi) // 2
         start = mid
         while start > 0 and m[start - 1] != '\n':
             start -= 1
         end = advance(start, '\0')
         if m[start:end] < s:
             # we know that after the null there are 40 bytes of sha1
             # this translates to the bisect lo = mid + 1
             lo = advance(end + 40, '\n') + 1
         else:
             # this translates to the bisect hi = mid
             hi = start
     end = advance(lo, '\0')
     found = m[lo:end]
     if s == found:
         # we know that after the null there are 40 bytes of sha1
         end = advance(end + 40, '\n')
         return (lo, end + 1)
     else:
         return (lo, lo)
 
 def _checkforbidden(l):
     """Check filenames for illegal characters."""
     for f in l:
         if '\n' in f or '\r' in f:
             raise error.RevlogError(
                 _("'\\n' and '\\r' disallowed in filenames: %r") % f)
 
 
 # apply the changes collected during the bisect loop to our addlist
 # return a delta suitable for addrevision
 def _addlistdelta(addlist, x):
     # for large addlist arrays, building a new array is cheaper
     # than repeatedly modifying the existing one
     currentposition = 0
     newaddlist = array.array('c')
 
     for start, end, content in x:
         newaddlist += addlist[currentposition:start]
         if content:
             newaddlist += array.array('c', content)
 
         currentposition = end
 
     newaddlist += addlist[currentposition:]
 
     deltatext = "".join(struct.pack(">lll", start, end, len(content))
                    + content for start, end, content in x)
     return deltatext, newaddlist
 
 def _splittopdir(f):
     if '/' in f:
         dir, subpath = f.split('/', 1)
         return dir + '/', subpath
     else:
         return '', f
 
 _noop = lambda s: None
 
 class treemanifest(object):
     def __init__(self, dir='', text=''):
         self._dir = dir
         self._node = revlog.nullid
         self._loadfunc = _noop
         self._copyfunc = _noop
         self._dirty = False
         self._dirs = {}
         # Using _lazymanifest here is a little slower than plain old dicts
         self._files = {}
         self._flags = {}
         if text:
             def readsubtree(subdir, subm):
                 raise AssertionError('treemanifest constructor only accepts '
                                      'flat manifests')
             self.parse(text, readsubtree)
             self._dirty = True # Mark flat manifest dirty after parsing
 
     def _subpath(self, path):
         return self._dir + path
 
     def __len__(self):
         self._load()
         size = len(self._files)
         for m in self._dirs.values():
             size += m.__len__()
         return size
 
     def _isempty(self):
         self._load() # for consistency; already loaded by all callers
         return (not self._files and (not self._dirs or
                 all(m._isempty() for m in self._dirs.values())))
 
     def __repr__(self):
         return ('<treemanifest dir=%s, node=%s, loaded=%s, dirty=%s at 0x%x>' %
                 (self._dir, revlog.hex(self._node),
                  bool(self._loadfunc is _noop),
                  self._dirty, id(self)))
 
     def dir(self):
         '''The directory that this tree manifest represents, including a
         trailing '/'. Empty string for the repo root directory.'''
         return self._dir
 
     def node(self):
         '''This node of this instance. nullid for unsaved instances. Should
         be updated when the instance is read or written from a revlog.
         '''
         assert not self._dirty
         return self._node
 
     def setnode(self, node):
         self._node = node
         self._dirty = False
 
     def iterentries(self):
         self._load()
         for p, n in sorted(self._dirs.items() + self._files.items()):
             if p in self._files:
                 yield self._subpath(p), n, self._flags.get(p, '')
             else:
                 for x in n.iterentries():
                     yield x
 
     def iteritems(self):
         self._load()
         for p, n in sorted(self._dirs.items() + self._files.items()):
             if p in self._files:
                 yield self._subpath(p), n
             else:
                 for f, sn in n.iteritems():
                     yield f, sn
 
     def iterkeys(self):
         self._load()
         for p in sorted(self._dirs.keys() + self._files.keys()):
             if p in self._files:
                 yield self._subpath(p)
             else:
                 for f in self._dirs[p].iterkeys():
                     yield f
 
     def keys(self):
         return list(self.iterkeys())
 
     def __iter__(self):
         return self.iterkeys()
 
     def __contains__(self, f):
         if f is None:
             return False
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             if dir not in self._dirs:
                 return False
             return self._dirs[dir].__contains__(subpath)
         else:
             return f in self._files
 
     def get(self, f, default=None):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             if dir not in self._dirs:
                 return default
             return self._dirs[dir].get(subpath, default)
         else:
             return self._files.get(f, default)
 
     def __getitem__(self, f):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             return self._dirs[dir].__getitem__(subpath)
         else:
             return self._files[f]
 
     def flags(self, f):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             if dir not in self._dirs:
                 return ''
             return self._dirs[dir].flags(subpath)
         else:
             if f in self._dirs:
                 return ''
             return self._flags.get(f, '')
 
     def find(self, f):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             return self._dirs[dir].find(subpath)
         else:
             return self._files[f], self._flags.get(f, '')
 
     def __delitem__(self, f):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             self._dirs[dir].__delitem__(subpath)
             # If the directory is now empty, remove it
             if self._dirs[dir]._isempty():
                 del self._dirs[dir]
         else:
             del self._files[f]
             if f in self._flags:
                 del self._flags[f]
         self._dirty = True
 
     def __setitem__(self, f, n):
         assert n is not None
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             if dir not in self._dirs:
                 self._dirs[dir] = treemanifest(self._subpath(dir))
             self._dirs[dir].__setitem__(subpath, n)
         else:
             self._files[f] = n[:21] # to match manifestdict's behavior
         self._dirty = True
 
     def _load(self):
         if self._loadfunc is not _noop:
             lf, self._loadfunc = self._loadfunc, _noop
             lf(self)
         elif self._copyfunc is not _noop:
             cf, self._copyfunc = self._copyfunc, _noop
             cf(self)
 
     def setflag(self, f, flags):
         """Set the flags (symlink, executable) for path f."""
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             if dir not in self._dirs:
                 self._dirs[dir] = treemanifest(self._subpath(dir))
             self._dirs[dir].setflag(subpath, flags)
         else:
             self._flags[f] = flags
         self._dirty = True
 
     def copy(self):
         copy = treemanifest(self._dir)
         copy._node = self._node
         copy._dirty = self._dirty
         if self._copyfunc is _noop:
             def _copyfunc(s):
                 self._load()
                 for d in self._dirs:
                     s._dirs[d] = self._dirs[d].copy()
                 s._files = dict.copy(self._files)
                 s._flags = dict.copy(self._flags)
             if self._loadfunc is _noop:
                 _copyfunc(copy)
             else:
                 copy._copyfunc = _copyfunc
         else:
             copy._copyfunc = self._copyfunc
         return copy
 
     def filesnotin(self, m2):
         '''Set of files in this manifest that are not in the other'''
         files = set()
         def _filesnotin(t1, t2):
             if t1._node == t2._node and not t1._dirty and not t2._dirty:
                 return
             t1._load()
             t2._load()
             for d, m1 in t1._dirs.iteritems():
                 if d in t2._dirs:
                     m2 = t2._dirs[d]
                     _filesnotin(m1, m2)
                 else:
                     files.update(m1.iterkeys())
 
             for fn in t1._files.iterkeys():
                 if fn not in t2._files:
                     files.add(t1._subpath(fn))
 
         _filesnotin(self, m2)
         return files
 
     @propertycache
     def _alldirs(self):
         return util.dirs(self)
 
     def dirs(self):
         return self._alldirs
 
     def hasdir(self, dir):
         self._load()
         topdir, subdir = _splittopdir(dir)
         if topdir:
             if topdir in self._dirs:
                 return self._dirs[topdir].hasdir(subdir)
             return False
         return (dir + '/') in self._dirs
 
     def walk(self, match):
         '''Generates matching file names.
 
         Equivalent to manifest.matches(match).iterkeys(), but without creating
         an entirely new manifest.
 
         It also reports nonexistent files by marking them bad with match.bad().
         '''
         if match.always():
             for f in iter(self):
                 yield f
             return
 
         fset = set(match.files())
 
         for fn in self._walk(match):
             if fn in fset:
                 # specified pattern is the exact name
                 fset.remove(fn)
             yield fn
 
         # for dirstate.walk, files=['.'] means "walk the whole tree".
         # follow that here, too
         fset.discard('.')
 
         for fn in sorted(fset):
             if not self.hasdir(fn):
                 match.bad(fn, None)
 
     def _walk(self, match):
         '''Recursively generates matching file names for walk().'''
         if not match.visitdir(self._dir[:-1] or '.'):
             return
 
         # yield this dir's files and walk its submanifests
         self._load()
         for p in sorted(self._dirs.keys() + self._files.keys()):
             if p in self._files:
                 fullp = self._subpath(p)
                 if match(fullp):
                     yield fullp
             else:
                 for f in self._dirs[p]._walk(match):
                     yield f
 
     def matches(self, match):
         '''generate a new manifest filtered by the match argument'''
         if match.always():
             return self.copy()
 
         return self._matches(match)
 
     def _matches(self, match):
         '''recursively generate a new manifest filtered by the match argument.
         '''
 
         visit = match.visitdir(self._dir[:-1] or '.')
         if visit == 'all':
             return self.copy()
         ret = treemanifest(self._dir)
         if not visit:
             return ret
 
         self._load()
         for fn in self._files:
             fullp = self._subpath(fn)
             if not match(fullp):
                 continue
             ret._files[fn] = self._files[fn]
             if fn in self._flags:
                 ret._flags[fn] = self._flags[fn]
 
         for dir, subm in self._dirs.iteritems():
             m = subm._matches(match)
             if not m._isempty():
                 ret._dirs[dir] = m
 
         if not ret._isempty():
             ret._dirty = True
         return ret
 
     def diff(self, m2, clean=False):
         '''Finds changes between the current manifest and m2.
 
         Args:
           m2: the manifest to which this manifest should be compared.
           clean: if true, include files unchanged between these manifests
                  with a None value in the returned dictionary.
 
         The result is returned as a dict with filename as key and
         values of the form ((n1,fl1),(n2,fl2)), where n1/n2 is the
         nodeid in the current/other manifest and fl1/fl2 is the flag
         in the current/other manifest. Where the file does not exist,
         the nodeid will be None and the flags will be the empty
         string.
         '''
         result = {}
         emptytree = treemanifest()
         def _diff(t1, t2):
             if t1._node == t2._node and not t1._dirty and not t2._dirty:
                 return
             t1._load()
             t2._load()
             for d, m1 in t1._dirs.iteritems():
                 m2 = t2._dirs.get(d, emptytree)
                 _diff(m1, m2)
 
             for d, m2 in t2._dirs.iteritems():
                 if d not in t1._dirs:
                     _diff(emptytree, m2)
 
             for fn, n1 in t1._files.iteritems():
                 fl1 = t1._flags.get(fn, '')
                 n2 = t2._files.get(fn, None)
                 fl2 = t2._flags.get(fn, '')
                 if n1 != n2 or fl1 != fl2:
                     result[t1._subpath(fn)] = ((n1, fl1), (n2, fl2))
                 elif clean:
                     result[t1._subpath(fn)] = None
 
             for fn, n2 in t2._files.iteritems():
                 if fn not in t1._files:
                     fl2 = t2._flags.get(fn, '')
                     result[t2._subpath(fn)] = ((None, ''), (n2, fl2))
 
         _diff(self, m2)
         return result
 
     def unmodifiedsince(self, m2):
         return not self._dirty and not m2._dirty and self._node == m2._node
 
     def parse(self, text, readsubtree):
         for f, n, fl in _parse(text):
             if fl == 't':
                 f = f + '/'
                 self._dirs[f] = readsubtree(self._subpath(f), n)
             elif '/' in f:
                 # This is a flat manifest, so use __setitem__ and setflag rather
                 # than assigning directly to _files and _flags, so we can
                 # assign a path in a subdirectory, and to mark dirty (compared
                 # to nullid).
                 self[f] = n
                 if fl:
                     self.setflag(f, fl)
             else:
                 # Assigning to _files and _flags avoids marking as dirty,
                 # and should be a little faster.
                 self._files[f] = n
                 if fl:
                     self._flags[f] = fl
 
     def text(self, usemanifestv2=False):
         """Get the full data of this manifest as a bytestring."""
         self._load()
         return _text(self.iterentries(), usemanifestv2)
 
     def dirtext(self, usemanifestv2=False):
         """Get the full data of this directory as a bytestring. Make sure that
         any submanifests have been written first, so their nodeids are correct.
         """
         self._load()
         flags = self.flags
         dirs = [(d[:-1], self._dirs[d]._node, 't') for d in self._dirs]
         files = [(f, self._files[f], flags(f)) for f in self._files]
         return _text(sorted(dirs + files), usemanifestv2)
 
     def read(self, gettext, readsubtree):
         def _load_for_read(s):
             s.parse(gettext(), readsubtree)
             s._dirty = False
         self._loadfunc = _load_for_read
 
     def writesubtrees(self, m1, m2, writesubtree):
         self._load() # for consistency; should never have any effect here
         m1._load()
         m2._load()
         emptytree = treemanifest()
         for d, subm in self._dirs.iteritems():
             subp1 = m1._dirs.get(d, emptytree)._node
             subp2 = m2._dirs.get(d, emptytree)._node
             if subp1 == revlog.nullid:
                 subp1, subp2 = subp2, subp1
             writesubtree(subm, subp1, subp2)
 
 class manifestrevlog(revlog.revlog):
     '''A revlog that stores manifest texts. This is responsible for caching the
     full-text manifest contents.
     '''
     def __init__(self, opener, dir='', dirlogcache=None):
         # During normal operations, we expect to deal with not more than four
         # revs at a time (such as during commit --amend). When rebasing large
         # stacks of commits, the number can go up, hence the config knob below.
         cachesize = 4
         usetreemanifest = False
         usemanifestv2 = False
         opts = getattr(opener, 'options', None)
         if opts is not None:
             cachesize = opts.get('manifestcachesize', cachesize)
             usetreemanifest = opts.get('treemanifest', usetreemanifest)
             usemanifestv2 = opts.get('manifestv2', usemanifestv2)
 
         self._treeondisk = usetreemanifest
         self._usemanifestv2 = usemanifestv2
 
         self._fulltextcache = util.lrucachedict(cachesize)
 
         indexfile = "00manifest.i"
         if dir:
             assert self._treeondisk, 'opts is %r' % opts
             if not dir.endswith('/'):
                 dir = dir + '/'
             indexfile = "meta/" + dir + "00manifest.i"
         self._dir = dir
         # The dirlogcache is kept on the root manifest log
         if dir:
             self._dirlogcache = dirlogcache
         else:
             self._dirlogcache = {'': self}
 
         super(manifestrevlog, self).__init__(opener, indexfile,
                                              checkambig=bool(dir))
 
     @property
     def fulltextcache(self):
         return self._fulltextcache
 
     def clearcaches(self):
         super(manifestrevlog, self).clearcaches()
         self._fulltextcache.clear()
         self._dirlogcache = {'': self}
 
     def dirlog(self, dir):
         if dir:
             assert self._treeondisk
         if dir not in self._dirlogcache:
             self._dirlogcache[dir] = manifestrevlog(self.opener, dir,
                                                     self._dirlogcache)
         return self._dirlogcache[dir]
 
     def add(self, m, transaction, link, p1, p2, added, removed, readtree=None):
         if (p1 in self.fulltextcache and util.safehasattr(m, 'fastdelta')
             and not self._usemanifestv2):
             # If our first parent is in the manifest cache, we can
             # compute a delta here using properties we know about the
             # manifest up-front, which may save time later for the
             # revlog layer.
 
             _checkforbidden(added)
             # combine the changed lists into one sorted iterator
             work = heapq.merge([(x, False) for x in added],
                                [(x, True) for x in removed])
 
             arraytext, deltatext = m.fastdelta(self.fulltextcache[p1], work)
             cachedelta = self.rev(p1), deltatext
             text = util.buffer(arraytext)
             n = self.addrevision(text, transaction, link, p1, p2, cachedelta)
         else:
             # The first parent manifest isn't already loaded, so we'll
             # just encode a fulltext of the manifest and pass that
             # through to the revlog layer, and let it handle the delta
             # process.
             if self._treeondisk:
                 assert readtree, "readtree must be set for treemanifest writes"
                 m1 = readtree(self._dir, p1)
                 m2 = readtree(self._dir, p2)
                 n = self._addtree(m, transaction, link, m1, m2, readtree)
                 arraytext = None
             else:
                 text = m.text(self._usemanifestv2)
                 n = self.addrevision(text, transaction, link, p1, p2)
                 arraytext = array.array('c', text)
 
         if arraytext is not None:
             self.fulltextcache[n] = arraytext
 
         return n
 
     def _addtree(self, m, transaction, link, m1, m2, readtree):
         # If the manifest is unchanged compared to one parent,
         # don't write a new revision
         if m.unmodifiedsince(m1) or m.unmodifiedsince(m2):
             return m.node()
         def writesubtree(subm, subp1, subp2):
             sublog = self.dirlog(subm.dir())
             sublog.add(subm, transaction, link, subp1, subp2, None, None,
                        readtree=readtree)
         m.writesubtrees(m1, m2, writesubtree)
         text = m.dirtext(self._usemanifestv2)
         # Double-check whether contents are unchanged to one parent
         if text == m1.dirtext(self._usemanifestv2):
             n = m1.node()
         elif text == m2.dirtext(self._usemanifestv2):
             n = m2.node()
         else:
             n = self.addrevision(text, transaction, link, m1.node(), m2.node())
         # Save nodeid so parent manifest can calculate its nodeid
         m.setnode(n)
         return n
 
 class manifestlog(object):
     """A collection class representing the collection of manifest snapshots
     referenced by commits in the repository.
 
     In this situation, 'manifest' refers to the abstract concept of a snapshot
     of the list of files in the given commit. Consumers of the output of this
     class do not care about the implementation details of the actual manifests
     they receive (i.e. tree or flat or lazily loaded, etc)."""
     def __init__(self, opener, repo):
         self._repo = repo
 
         usetreemanifest = False
 
         opts = getattr(opener, 'options', None)
         if opts is not None:
             usetreemanifest = opts.get('treemanifest', usetreemanifest)
         self._treeinmem = usetreemanifest
 
         self._oldmanifest = repo._constructmanifest()
         self._revlog = self._oldmanifest
 
         # A cache of the manifestctx or treemanifestctx for each directory
         self._dirmancache = {}
 
         # We'll separate this into it's own cache once oldmanifest is no longer
         # used
         self._mancache = self._oldmanifest._mancache
         self._dirmancache[''] = self._mancache
 
         # A future patch makes this use the same config value as the existing
         # mancache
         self.cachesize = 4
 
     def __getitem__(self, node):
         """Retrieves the manifest instance for the given node. Throws a
         LookupError if not found.
         """
         return self.get('', node)
 
     def get(self, dir, node):
         """Retrieves the manifest instance for the given node. Throws a
         LookupError if not found.
         """
         if node in self._dirmancache.get(dir, ()):
             cachemf = self._dirmancache[dir][node]
             # The old manifest may put non-ctx manifests in the cache, so
             # skip those since they don't implement the full api.
             if (isinstance(cachemf, manifestctx) or
                 isinstance(cachemf, treemanifestctx)):
                 return cachemf
 
         if dir:
             if self._revlog._treeondisk:
                 dirlog = self._revlog.dirlog(dir)
                 if node not in dirlog.nodemap:
                     raise LookupError(node, dirlog.indexfile,
                                       _('no node'))
                 m = treemanifestctx(self._repo, dir, node)
             else:
                 raise error.Abort(
                         _("cannot ask for manifest directory '%s' in a flat "
                           "manifest") % dir)
         else:
             if node not in self._revlog.nodemap:
                 raise LookupError(node, self._revlog.indexfile,
                                   _('no node'))
             if self._treeinmem:
                 m = treemanifestctx(self._repo, '', node)
             else:
                 m = manifestctx(self._repo, node)
 
         if node != revlog.nullid:
             mancache = self._dirmancache.get(dir)
             if not mancache:
                 mancache = util.lrucachedict(self.cachesize)
                 self._dirmancache[dir] = mancache
             mancache[node] = m
         return m
 
+    def clearcaches(self):
+        self._dirmancache.clear()
+        self._revlog.clearcaches()
+
 class memmanifestctx(object):
     def __init__(self, repo):
         self._repo = repo
         self._manifestdict = manifestdict()
 
     def _revlog(self):
         return self._repo.manifestlog._revlog
 
     def new(self):
         return memmanifestctx(self._repo)
 
     def copy(self):
         memmf = memmanifestctx(self._repo)
         memmf._manifestdict = self.read().copy()
         return memmf
 
     def read(self):
         return self._manifestdict
 
     def write(self, transaction, link, p1, p2, added, removed):
         return self._revlog().add(self._manifestdict, transaction, link, p1, p2,
                                   added, removed)
 
 class manifestctx(object):
     """A class representing a single revision of a manifest, including its
     contents, its parent revs, and its linkrev.
     """
     def __init__(self, repo, node):
         self._repo = repo
         self._data = None
 
         self._node = node
 
         # TODO: We eventually want p1, p2, and linkrev exposed on this class,
         # but let's add it later when something needs it and we can load it
         # lazily.
         #self.p1, self.p2 = revlog.parents(node)
         #rev = revlog.rev(node)
         #self.linkrev = revlog.linkrev(rev)
 
     def _revlog(self):
         return self._repo.manifestlog._revlog
 
     def node(self):
         return self._node
 
     def new(self):
         return memmanifestctx(self._repo)
 
     def copy(self):
         memmf = memmanifestctx(self._repo)
         memmf._manifestdict = self.read().copy()
         return memmf
 
     def read(self):
         if not self._data:
             if self._node == revlog.nullid:
                 self._data = manifestdict()
             else:
                 rl = self._revlog()
                 text = rl.revision(self._node)
                 arraytext = array.array('c', text)
                 rl._fulltextcache[self._node] = arraytext
                 self._data = manifestdict(text)
         return self._data
 
     def readfast(self, shallow=False):
         '''Calls either readdelta or read, based on which would be less work.
         readdelta is called if the delta is against the p1, and therefore can be
         read quickly.
 
         If `shallow` is True, nothing changes since this is a flat manifest.
         '''
         rl = self._revlog()
         r = rl.rev(self._node)
         deltaparent = rl.deltaparent(r)
         if deltaparent != revlog.nullrev and deltaparent in rl.parentrevs(r):
             return self.readdelta()
         return self.read()
 
     def readdelta(self, shallow=False):
         '''Returns a manifest containing just the entries that are present
         in this manifest, but not in its p1 manifest. This is efficient to read
         if the revlog delta is already p1.
 
         Changing the value of `shallow` has no effect on flat manifests.
         '''
         revlog = self._revlog()
         if revlog._usemanifestv2:
             # Need to perform a slow delta
             r0 = revlog.deltaparent(revlog.rev(self._node))
             m0 = manifestctx(self._repo, revlog.node(r0)).read()
             m1 = self.read()
             md = manifestdict()
             for f, ((n0, fl0), (n1, fl1)) in m0.diff(m1).iteritems():
                 if n1:
                     md[f] = n1
                     if fl1:
                         md.setflag(f, fl1)
             return md
 
         r = revlog.rev(self._node)
         d = mdiff.patchtext(revlog.revdiff(revlog.deltaparent(r), r))
         return manifestdict(d)
 
     def find(self, key):
         return self.read().find(key)
 
 class memtreemanifestctx(object):
     def __init__(self, repo, dir=''):
         self._repo = repo
         self._dir = dir
         self._treemanifest = treemanifest()
 
     def _revlog(self):
         return self._repo.manifestlog._revlog
 
     def new(self, dir=''):
         return memtreemanifestctx(self._repo, dir=dir)
 
     def copy(self):
         memmf = memtreemanifestctx(self._repo, dir=self._dir)
         memmf._treemanifest = self._treemanifest.copy()
         return memmf
 
     def read(self):
         return self._treemanifest
 
     def write(self, transaction, link, p1, p2, added, removed):
         def readtree(dir, node):
             return self._repo.manifestlog.get(dir, node).read()
         return self._revlog().add(self._treemanifest, transaction, link, p1, p2,
                                   added, removed, readtree=readtree)
 
 class treemanifestctx(object):
     def __init__(self, repo, dir, node):
         self._repo = repo
         self._dir = dir
         self._data = None
 
         self._node = node
 
         # TODO: Load p1/p2/linkrev lazily. They need to be lazily loaded so that
         # we can instantiate treemanifestctx objects for directories we don't
         # have on disk.
         #self.p1, self.p2 = revlog.parents(node)
         #rev = revlog.rev(node)
         #self.linkrev = revlog.linkrev(rev)
 
     def _revlog(self):
         return self._repo.manifestlog._revlog.dirlog(self._dir)
 
     def read(self):
         if not self._data:
             rl = self._revlog()
             if self._node == revlog.nullid:
                 self._data = treemanifest()
             elif rl._treeondisk:
                 m = treemanifest(dir=self._dir)
                 def gettext():
                     return rl.revision(self._node)
                 def readsubtree(dir, subm):
                     return treemanifestctx(self._repo, dir, subm).read()
                 m.read(gettext, readsubtree)
                 m.setnode(self._node)
                 self._data = m
             else:
                 text = rl.revision(self._node)
                 arraytext = array.array('c', text)
                 rl.fulltextcache[self._node] = arraytext
                 self._data = treemanifest(dir=self._dir, text=text)
 
         return self._data
 
     def node(self):
         return self._node
 
     def new(self, dir=''):
         return memtreemanifestctx(self._repo, dir=dir)
 
     def copy(self):
         memmf = memtreemanifestctx(self._repo, dir=self._dir)
         memmf._treemanifest = self.read().copy()
         return memmf
 
     def readdelta(self, shallow=False):
         '''Returns a manifest containing just the entries that are present
         in this manifest, but not in its p1 manifest. This is efficient to read
         if the revlog delta is already p1.
 
         If `shallow` is True, this will read the delta for this directory,
         without recursively reading subdirectory manifests. Instead, any
         subdirectory entry will be reported as it appears in the manifest, i.e.
         the subdirectory will be reported among files and distinguished only by
         its 't' flag.
         '''
         revlog = self._revlog()
         if shallow and not revlog._usemanifestv2:
             r = revlog.rev(self._node)
             d = mdiff.patchtext(revlog.revdiff(revlog.deltaparent(r), r))
             return manifestdict(d)
         else:
             # Need to perform a slow delta
             r0 = revlog.deltaparent(revlog.rev(self._node))
             m0 = treemanifestctx(self._repo, self._dir, revlog.node(r0)).read()
             m1 = self.read()
             md = treemanifest(dir=self._dir)
             for f, ((n0, fl0), (n1, fl1)) in m0.diff(m1).iteritems():
                 if n1:
                     md[f] = n1
                     if fl1:
                         md.setflag(f, fl1)
             return md
 
     def readfast(self, shallow=False):
         '''Calls either readdelta or read, based on which would be less work.
         readdelta is called if the delta is against the p1, and therefore can be
         read quickly.
 
         If `shallow` is True, it only returns the entries from this manifest,
         and not any submanifests.
         '''
         rl = self._revlog()
         r = rl.rev(self._node)
         deltaparent = rl.deltaparent(r)
         if (deltaparent != revlog.nullrev and
             deltaparent in rl.parentrevs(r)):
             return self.readdelta(shallow=shallow)
 
         if shallow:
             return manifestdict(rl.revision(self._node))
         else:
             return self.read()
 
     def find(self, key):
         return self.read().find(key)
 
 class manifest(manifestrevlog):
     def __init__(self, opener, dir='', dirlogcache=None):
         '''The 'dir' and 'dirlogcache' arguments are for internal use by
         manifest.manifest only. External users should create a root manifest
         log with manifest.manifest(opener) and call dirlog() on it.
         '''
         # During normal operations, we expect to deal with not more than four
         # revs at a time (such as during commit --amend). When rebasing large
         # stacks of commits, the number can go up, hence the config knob below.
         cachesize = 4
         usetreemanifest = False
         opts = getattr(opener, 'options', None)
         if opts is not None:
             cachesize = opts.get('manifestcachesize', cachesize)
             usetreemanifest = opts.get('treemanifest', usetreemanifest)
         self._mancache = util.lrucachedict(cachesize)
         self._treeinmem = usetreemanifest
         super(manifest, self).__init__(opener, dir=dir, dirlogcache=dirlogcache)
 
     def _newmanifest(self, data=''):
         if self._treeinmem:
             return treemanifest(self._dir, data)
         return manifestdict(data)
 
     def dirlog(self, dir):
         """This overrides the base revlog implementation to allow construction
         'manifest' types instead of manifestrevlog types. This is only needed
         until we migrate off the 'manifest' type."""
         if dir:
             assert self._treeondisk
         if dir not in self._dirlogcache:
             self._dirlogcache[dir] = manifest(self.opener, dir,
                                               self._dirlogcache)
         return self._dirlogcache[dir]
-
-    def clearcaches(self):
-        super(manifest, self).clearcaches()
-        self._mancache.clear()