view mercurial/treediscovery.py @ 29861:2f6d5c60f6fc stable

annotate: pre-calculate the "needed" dictionary (issue5360)

The "needed" dict is used as a reference counter to free items in the giant
"hist" dict. However, currently it is not very accurate and can lead to
dropping "hist" items unnecessarily, for example, with the following DAG,

       -3-
      /   \
  0--1--2--4--

The current algorithm will visit and calculate rev 1 twice, undesired. And
it tries to be smart by clearing rev 1's parents: "pcache[1] = []" at the
time hist[1] being accessed (note: hist[1] needs to be used twice, by rev 2
and rev 3). It can result in incorrect results if p1 of rev 4 deletes chunks
belonging to rev 0.

However, simply removing "needed" is not okay, because it will consume 10x
memory:

  # without any change
  % HGRCPATH= lrun ./hg annotate mercurial/commands.py -r d130a38 3>&2 [1]
  MEMORY   49074176
  CPUTIME  9.213
  REALTIME 9.270

  # with "needed" removed
  MEMORY   637673472
  CPUTIME  8.164
  REALTIME 8.249

This patch moves "needed" (and "pcache") calculation to a separate DFS to
address the issue. It improves perf and fixes issue5360 by correctly reusing
hist, while maintaining low memory usage. Some additional attempt has been
made to further reduce memory usage, like changing "pcache[f] = []" to "del
pcache[f]". Therefore the result can be both faster and lower memory usage:

  # with this patch applied
  MEMORY   47575040
  CPUTIME  7.870
  REALTIME 7.926

[1]: lrun is a lightweight sandbox built on Linux cgroup and namespace. It's
     used to measure CPU and memory usage here. Source code is available at
     github.com/quark-zju/lrun.

# 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,
)

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:
        heads = remote.heads()

    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

    # 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)
    unknown = collections.deque(remote.branches(unknown))
    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
            repo.ui.progress(_('searching'), reqcnt, unit=_('queries'))
            repo.ui.debug("request %d: %s\n" %
                        (reqcnt, " ".join(map(short, r))))
            for p in xrange(0, len(r), 10):
                for b in remote.branches(r[p:p + 10]):
                    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
        repo.ui.progress(_('searching'), reqcnt, unit=_('queries'))
        for n, l in zip(search, remote.between(search)):
            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")

    repo.ui.progress(_('searching'), None)
    repo.ui.debug("%d total queries\n" % reqcnt)

    return base, list(fetch), heads