view mercurial/treediscovery.py @ 42121:6578654916ae

branchcache: lazily validate nodes from the branchmap

On my personal hg-repository with 365 entries in .hg/cache/branch2, following
are the numbers for perfbranchmapload.

Before this patch:

! wall 0.000866 comb 0.000000 user 0.000000 sys 0.000000 (best of 2680)
! wall 0.001525 comb 0.000000 user 0.000000 sys 0.000000 (max of 2680)
! wall 0.001107 comb 0.001097 user 0.001086 sys 0.000011 (avg of 2680)
! wall 0.001104 comb 0.000000 user 0.000000 sys 0.000000 (median of 2680)

With this patch:

! wall 0.000530 comb 0.000000 user 0.000000 sys 0.000000 (best of 4240)
! wall 0.001078 comb 0.000000 user 0.000000 sys 0.000000 (max of 4240)
! wall 0.000696 comb 0.000693 user 0.000677 sys 0.000017 (avg of 4240)
! wall 0.000690 comb 0.000000 user 0.000000 sys 0.000000 (median of 4240)

On our internal repository with ~20k entries in branchcache, I see improvement
from 0.125 sec to 0.066 sec which is 47% speed up.

The above are the numbers of perfbranchmapload which shows how much time we
saved by not validating the nodes. But we need to validate some nodes. Following
are timings of some mercurial operations which have speed up because of this
lazy validation of nodes:

No-op `hg update` on our internal repository (Avg on 4 runs):

Before: 0.540 secs
After: 0.430 secs

Setting a branch name which already exists without --force (Avg of 4 runs):

Before: 0.510 secs
After: 0.250 secs

I ran the ASV performance suite and was unable to see any improvements except
there was improvement of perfdirstatewrite() on netbeans which I think was not
related.

I looked into the commit code, the command which I am trying to speedup, it
looks like it uses revbranchcache to update the branchcache.

Differential Revision: https://phab.mercurial-scm.org/D6208

# discovery.py - protocol changeset discovery functions
#
# Copyright 2010 Matt Mackall <[e-mail]>
#
# This software may be used and distributed according to the terms of the
# GNU General Public License version 2 or any later version.

from __future__ import absolute_import

import collections

from .i18n import _
from .node import (
    nullid,
    short,
)
from . import (
    error,
    pycompat,
)

def findcommonincoming(repo, remote, heads=None, force=False):
    """Return a tuple (common, fetch, heads) used to identify the common
    subset of nodes between repo and remote.

    "common" is a list of (at least) the heads of the common subset.
    "fetch" is a list of roots of the nodes that would be incoming, to be
      supplied to changegroupsubset.
    "heads" is either the supplied heads, or else the remote's heads.
    """

    knownnode = repo.changelog.hasnode
    search = []
    fetch = set()
    seen = set()
    seenbranch = set()
    base = set()

    if not heads:
        with remote.commandexecutor() as e:
            heads = e.callcommand('heads', {}).result()

    if repo.changelog.tip() == nullid:
        base.add(nullid)
        if heads != [nullid]:
            return [nullid], [nullid], list(heads)
        return [nullid], [], heads

    # assume we're closer to the tip than the root
    # and start by examining the heads
    repo.ui.status(_("searching for changes\n"))

    unknown = []
    for h in heads:
        if not knownnode(h):
            unknown.append(h)
        else:
            base.add(h)

    if not unknown:
        return list(base), [], list(heads)

    req = set(unknown)
    reqcnt = 0
    progress = repo.ui.makeprogress(_('searching'), unit=_('queries'))

    # search through remote branches
    # a 'branch' here is a linear segment of history, with four parts:
    # head, root, first parent, second parent
    # (a branch always has two parents (or none) by definition)
    with remote.commandexecutor() as e:
        branches = e.callcommand('branches', {'nodes': unknown}).result()

    unknown = collections.deque(branches)
    while unknown:
        r = []
        while unknown:
            n = unknown.popleft()
            if n[0] in seen:
                continue

            repo.ui.debug("examining %s:%s\n"
                          % (short(n[0]), short(n[1])))
            if n[0] == nullid: # found the end of the branch
                pass
            elif n in seenbranch:
                repo.ui.debug("branch already found\n")
                continue
            elif n[1] and knownnode(n[1]): # do we know the base?
                repo.ui.debug("found incomplete branch %s:%s\n"
                              % (short(n[0]), short(n[1])))
                search.append(n[0:2]) # schedule branch range for scanning
                seenbranch.add(n)
            else:
                if n[1] not in seen and n[1] not in fetch:
                    if knownnode(n[2]) and knownnode(n[3]):
                        repo.ui.debug("found new changeset %s\n" %
                                      short(n[1]))
                        fetch.add(n[1]) # earliest unknown
                    for p in n[2:4]:
                        if knownnode(p):
                            base.add(p) # latest known

                for p in n[2:4]:
                    if p not in req and not knownnode(p):
                        r.append(p)
                        req.add(p)
            seen.add(n[0])

        if r:
            reqcnt += 1
            progress.increment()
            repo.ui.debug("request %d: %s\n" %
                        (reqcnt, " ".join(map(short, r))))
            for p in pycompat.xrange(0, len(r), 10):
                with remote.commandexecutor() as e:
                    branches = e.callcommand('branches', {
                        'nodes': r[p:p + 10],
                    }).result()

                for b in branches:
                    repo.ui.debug("received %s:%s\n" %
                                  (short(b[0]), short(b[1])))
                    unknown.append(b)

    # do binary search on the branches we found
    while search:
        newsearch = []
        reqcnt += 1
        progress.increment()

        with remote.commandexecutor() as e:
            between = e.callcommand('between', {'pairs': search}).result()

        for n, l in zip(search, between):
            l.append(n[1])
            p = n[0]
            f = 1
            for i in l:
                repo.ui.debug("narrowing %d:%d %s\n" % (f, len(l), short(i)))
                if knownnode(i):
                    if f <= 2:
                        repo.ui.debug("found new branch changeset %s\n" %
                                          short(p))
                        fetch.add(p)
                        base.add(i)
                    else:
                        repo.ui.debug("narrowed branch search to %s:%s\n"
                                      % (short(p), short(i)))
                        newsearch.append((p, i))
                    break
                p, f = i, f * 2
            search = newsearch

    # sanity check our fetch list
    for f in fetch:
        if knownnode(f):
            raise error.RepoError(_("already have changeset ")
                                  + short(f[:4]))

    base = list(base)
    if base == [nullid]:
        if force:
            repo.ui.warn(_("warning: repository is unrelated\n"))
        else:
            raise error.Abort(_("repository is unrelated"))

    repo.ui.debug("found new changesets starting at " +
                 " ".join([short(f) for f in fetch]) + "\n")

    progress.complete()
    repo.ui.debug("%d total queries\n" % reqcnt)

    return base, list(fetch), heads