##// END OF EJS Templates
bdiff: replace hash algorithm...
bdiff: replace hash algorithm This patch replaces lyhash with the hash algorithm used by diffutils. The algorithm has its origins in Git commit 2e9d1410, which is all the way back from 1992. The license header in the code at that revision in GPL v2. I have not performed an extensive analysis of the distribution (and therefore buckets) of hash output. However, `hg perfbdiff` gives some clear wins. I'd like to think that if it is good enough for diffutils it is good enough for us? From the mozilla-unified repository: $ perfbdiff -m 3041e4d59df2 ! wall 0.053271 comb 0.060000 user 0.060000 sys 0.000000 (best of 100) ! wall 0.035827 comb 0.040000 user 0.040000 sys 0.000000 (best of 100) $ perfbdiff 0e9928989e9c --alldata --count 100 ! wall 6.204277 comb 6.200000 user 6.200000 sys 0.000000 (best of 3) ! wall 4.309710 comb 4.300000 user 4.300000 sys 0.000000 (best of 3) From the hg repo: $ perfbdiff 35000 --alldata --count 1000 ! wall 0.660358 comb 0.660000 user 0.660000 sys 0.000000 (best of 15) ! wall 0.534092 comb 0.530000 user 0.530000 sys 0.000000 (best of 19) Looking at the generated assembly and statistical profiler output from the kernel level, I believe there is room to make this function even faster. Namely, we're still consuming data character by character instead of at the word level. This translates to more loop iterations and more instructions. At this juncture though, the real performance killer is that we're hashing every line. We should get a significant speedup if we change the algorithm to find the longest prefix, longest suffix, treat those as single "lines" and then only do the line splitting and hashing on the parts that are different. That will require a lot of C code, however. I'm optimistic this approach could result in a ~2x speedup.

File last commit:

r30233:3afde791 stable
r30318:e1d6aa0e default
Show More
reposetup.py
389 lines | 17.1 KiB | text/x-python | PythonLexer
# Copyright 2009-2010 Gregory P. Ward
# Copyright 2009-2010 Intelerad Medical Systems Incorporated
# Copyright 2010-2011 Fog Creek Software
# Copyright 2010-2011 Unity Technologies
#
# This software may be used and distributed according to the terms of the
# GNU General Public License version 2 or any later version.
'''setup for largefiles repositories: reposetup'''
from __future__ import absolute_import
import copy
from mercurial.i18n import _
from mercurial import (
error,
localrepo,
match as matchmod,
scmutil,
)
from . import (
lfcommands,
lfutil,
)
def reposetup(ui, repo):
# wire repositories should be given new wireproto functions
# by "proto.wirereposetup()" via "hg.wirepeersetupfuncs"
if not repo.local():
return
class lfilesrepo(repo.__class__):
# the mark to examine whether "repo" object enables largefiles or not
_largefilesenabled = True
lfstatus = False
def status_nolfiles(self, *args, **kwargs):
return super(lfilesrepo, self).status(*args, **kwargs)
# When lfstatus is set, return a context that gives the names
# of largefiles instead of their corresponding standins and
# identifies the largefiles as always binary, regardless of
# their actual contents.
def __getitem__(self, changeid):
ctx = super(lfilesrepo, self).__getitem__(changeid)
if self.lfstatus:
class lfilesctx(ctx.__class__):
def files(self):
filenames = super(lfilesctx, self).files()
return [lfutil.splitstandin(f) or f for f in filenames]
def manifest(self):
man1 = super(lfilesctx, self).manifest()
class lfilesmanifest(man1.__class__):
def __contains__(self, filename):
orig = super(lfilesmanifest, self).__contains__
return (orig(filename) or
orig(lfutil.standin(filename)))
man1.__class__ = lfilesmanifest
return man1
def filectx(self, path, fileid=None, filelog=None):
orig = super(lfilesctx, self).filectx
try:
if filelog is not None:
result = orig(path, fileid, filelog)
else:
result = orig(path, fileid)
except error.LookupError:
# Adding a null character will cause Mercurial to
# identify this as a binary file.
if filelog is not None:
result = orig(lfutil.standin(path), fileid,
filelog)
else:
result = orig(lfutil.standin(path), fileid)
olddata = result.data
result.data = lambda: olddata() + '\0'
return result
ctx.__class__ = lfilesctx
return ctx
# Figure out the status of big files and insert them into the
# appropriate list in the result. Also removes standin files
# from the listing. Revert to the original status if
# self.lfstatus is False.
# XXX large file status is buggy when used on repo proxy.
# XXX this needs to be investigated.
@localrepo.unfilteredmethod
def status(self, node1='.', node2=None, match=None, ignored=False,
clean=False, unknown=False, listsubrepos=False):
listignored, listclean, listunknown = ignored, clean, unknown
orig = super(lfilesrepo, self).status
if not self.lfstatus:
return orig(node1, node2, match, listignored, listclean,
listunknown, listsubrepos)
# some calls in this function rely on the old version of status
self.lfstatus = False
ctx1 = self[node1]
ctx2 = self[node2]
working = ctx2.rev() is None
parentworking = working and ctx1 == self['.']
if match is None:
match = matchmod.always(self.root, self.getcwd())
wlock = None
try:
try:
# updating the dirstate is optional
# so we don't wait on the lock
wlock = self.wlock(False)
except error.LockError:
pass
# First check if paths or patterns were specified on the
# command line. If there were, and they don't match any
# largefiles, we should just bail here and let super
# handle it -- thus gaining a big performance boost.
lfdirstate = lfutil.openlfdirstate(ui, self)
if not match.always():
for f in lfdirstate:
if match(f):
break
else:
return orig(node1, node2, match, listignored, listclean,
listunknown, listsubrepos)
# Create a copy of match that matches standins instead
# of largefiles.
def tostandins(files):
if not working:
return files
newfiles = []
dirstate = self.dirstate
for f in files:
sf = lfutil.standin(f)
if sf in dirstate:
newfiles.append(sf)
elif sf in dirstate.dirs():
# Directory entries could be regular or
# standin, check both
newfiles.extend((f, sf))
else:
newfiles.append(f)
return newfiles
m = copy.copy(match)
m._files = tostandins(m._files)
result = orig(node1, node2, m, ignored, clean, unknown,
listsubrepos)
if working:
def sfindirstate(f):
sf = lfutil.standin(f)
dirstate = self.dirstate
return sf in dirstate or sf in dirstate.dirs()
match._files = [f for f in match._files
if sfindirstate(f)]
# Don't waste time getting the ignored and unknown
# files from lfdirstate
unsure, s = lfdirstate.status(match, [], False, listclean,
False)
(modified, added, removed, deleted, clean) = (
s.modified, s.added, s.removed, s.deleted, s.clean)
if parentworking:
for lfile in unsure:
standin = lfutil.standin(lfile)
if standin not in ctx1:
# from second parent
modified.append(lfile)
elif ctx1[standin].data().strip() \
!= lfutil.hashfile(self.wjoin(lfile)):
modified.append(lfile)
else:
if listclean:
clean.append(lfile)
lfdirstate.normal(lfile)
else:
tocheck = unsure + modified + added + clean
modified, added, clean = [], [], []
checkexec = self.dirstate._checkexec
for lfile in tocheck:
standin = lfutil.standin(lfile)
if standin in ctx1:
abslfile = self.wjoin(lfile)
if ((ctx1[standin].data().strip() !=
lfutil.hashfile(abslfile)) or
(checkexec and
('x' in ctx1.flags(standin)) !=
bool(lfutil.getexecutable(abslfile)))):
modified.append(lfile)
elif listclean:
clean.append(lfile)
else:
added.append(lfile)
# at this point, 'removed' contains largefiles
# marked as 'R' in the working context.
# then, largefiles not managed also in the target
# context should be excluded from 'removed'.
removed = [lfile for lfile in removed
if lfutil.standin(lfile) in ctx1]
# Standins no longer found in lfdirstate have been deleted
for standin in ctx1.walk(lfutil.getstandinmatcher(self)):
lfile = lfutil.splitstandin(standin)
if not match(lfile):
continue
if lfile not in lfdirstate:
deleted.append(lfile)
# Sync "largefile has been removed" back to the
# standin. Removing a file as a side effect of
# running status is gross, but the alternatives (if
# any) are worse.
self.wvfs.unlinkpath(standin, ignoremissing=True)
# Filter result lists
result = list(result)
# Largefiles are not really removed when they're
# still in the normal dirstate. Likewise, normal
# files are not really removed if they are still in
# lfdirstate. This happens in merges where files
# change type.
removed = [f for f in removed
if f not in self.dirstate]
result[2] = [f for f in result[2]
if f not in lfdirstate]
lfiles = set(lfdirstate._map)
# Unknown files
result[4] = set(result[4]).difference(lfiles)
# Ignored files
result[5] = set(result[5]).difference(lfiles)
# combine normal files and largefiles
normals = [[fn for fn in filelist
if not lfutil.isstandin(fn)]
for filelist in result]
lfstatus = (modified, added, removed, deleted, [], [],
clean)
result = [sorted(list1 + list2)
for (list1, list2) in zip(normals, lfstatus)]
else: # not against working directory
result = [[lfutil.splitstandin(f) or f for f in items]
for items in result]
if wlock:
lfdirstate.write()
finally:
if wlock:
wlock.release()
self.lfstatus = True
return scmutil.status(*result)
def commitctx(self, ctx, *args, **kwargs):
node = super(lfilesrepo, self).commitctx(ctx, *args, **kwargs)
class lfilesctx(ctx.__class__):
def markcommitted(self, node):
orig = super(lfilesctx, self).markcommitted
return lfutil.markcommitted(orig, self, node)
ctx.__class__ = lfilesctx
return node
# Before commit, largefile standins have not had their
# contents updated to reflect the hash of their largefile.
# Do that here.
def commit(self, text="", user=None, date=None, match=None,
force=False, editor=False, extra={}):
orig = super(lfilesrepo, self).commit
with self.wlock():
lfcommithook = self._lfcommithooks[-1]
match = lfcommithook(self, match)
result = orig(text=text, user=user, date=date, match=match,
force=force, editor=editor, extra=extra)
return result
def push(self, remote, force=False, revs=None, newbranch=False):
if remote.local():
missing = set(self.requirements) - remote.local().supported
if missing:
msg = _("required features are not"
" supported in the destination:"
" %s") % (', '.join(sorted(missing)))
raise error.Abort(msg)
return super(lfilesrepo, self).push(remote, force=force, revs=revs,
newbranch=newbranch)
# TODO: _subdirlfs should be moved into "lfutil.py", because
# it is referred only from "lfutil.updatestandinsbymatch"
def _subdirlfs(self, files, lfiles):
'''
Adjust matched file list
If we pass a directory to commit whose only committable files
are largefiles, the core commit code aborts before finding
the largefiles.
So we do the following:
For directories that only have largefiles as matches,
we explicitly add the largefiles to the match list and remove
the directory.
In other cases, we leave the match list unmodified.
'''
actualfiles = []
dirs = []
regulars = []
for f in files:
if lfutil.isstandin(f + '/'):
raise error.Abort(
_('file "%s" is a largefile standin') % f,
hint=('commit the largefile itself instead'))
# Scan directories
if self.wvfs.isdir(f):
dirs.append(f)
else:
regulars.append(f)
for f in dirs:
matcheddir = False
d = self.dirstate.normalize(f) + '/'
# Check for matched normal files
for mf in regulars:
if self.dirstate.normalize(mf).startswith(d):
actualfiles.append(f)
matcheddir = True
break
if not matcheddir:
# If no normal match, manually append
# any matching largefiles
for lf in lfiles:
if self.dirstate.normalize(lf).startswith(d):
actualfiles.append(lf)
if not matcheddir:
# There may still be normal files in the dir, so
# add a directory to the list, which
# forces status/dirstate to walk all files and
# call the match function on the matcher, even
# on case sensitive filesystems.
actualfiles.append('.')
matcheddir = True
# Nothing in dir, so readd it
# and let commit reject it
if not matcheddir:
actualfiles.append(f)
# Always add normal files
actualfiles += regulars
return actualfiles
repo.__class__ = lfilesrepo
# stack of hooks being executed before committing.
# only last element ("_lfcommithooks[-1]") is used for each committing.
repo._lfcommithooks = [lfutil.updatestandinsbymatch]
# Stack of status writer functions taking "*msg, **opts" arguments
# like "ui.status()". Only last element ("_lfstatuswriters[-1]")
# is used to write status out.
repo._lfstatuswriters = [ui.status]
def prepushoutgoinghook(pushop):
"""Push largefiles for pushop before pushing revisions."""
lfrevs = pushop.lfrevs
if lfrevs is None:
lfrevs = pushop.outgoing.missing
if lfrevs:
toupload = set()
addfunc = lambda fn, lfhash: toupload.add(lfhash)
lfutil.getlfilestoupload(pushop.repo, lfrevs,
addfunc)
lfcommands.uploadlfiles(ui, pushop.repo, pushop.remote, toupload)
repo.prepushoutgoinghooks.add("largefiles", prepushoutgoinghook)
def checkrequireslfiles(ui, repo, **kwargs):
if 'largefiles' not in repo.requirements and any(
lfutil.shortname+'/' in f[0] for f in repo.store.datafiles()):
repo.requirements.add('largefiles')
repo._writerequirements()
ui.setconfig('hooks', 'changegroup.lfiles', checkrequireslfiles,
'largefiles')
ui.setconfig('hooks', 'commit.lfiles', checkrequireslfiles, 'largefiles')