view hgext/share.py @ 26380:56a640b0f656

revlog: don't flush data file after every added revision The current behavior of revlogs is to flush the data file when writing data to it. Tracing system calls revealed that changegroup processing incurred numerous write(2) calls for values much smaller than the default buffer size (Python defaults to 4096, but it can be adjusted based on detected block size at run time by CPython). The reason we flush revlogs is so readers have all data available. For example, the current code in revlog.py will re-open the revlog file (instead of seeking an existing file handle) to read the text of a revision. This happens when starting a new delta chain when adding several revisions from changegroups, for example. Yes, this is likely sub-optimal (we should probably be sharing file descriptors between readers and writers to avoid the flushing and associated overhead of re-opening files). While flushing revlogs is necessary, it appears all callers are diligent about flushing files before a read is performed (see buildtext() in _addrevision()), making the flush in _writeentry() redundant and unncessary. So, we remove it. In practice, this means we incur a write(2) a) when the buffer is full (typically 4096 bytes) b) when a new delta chain is created rather than after every added revision. This applies to every revlog, but by volume it mostly impacts filelogs. Removing the redundant flush from _writeentry() significantly reduces the number of write(2) calls during changegroup processing on my Linux machine. When applying a changegroup of the hg repo based on my local repo, the total number of write(2) calls during application of the mercurial/localrepo.py revlogs dropped from 1,320 to 217 with this patch applied. Total I/O related system calls dropped from 1,577 to 474. When unbundling a mozilla-central gzipped bundle (264,403 changesets with 1,492,215 changes to 222,507 files), total write(2) calls dropped from 1,252,881 to 827,106 and total system calls dropped from 3,601,259 to 3,178,636 - a reduction of 425,775! While the system call reduction is significant, it appears to have no impact on wall time on my Linux and Windows machines. Still, fewer syscalls is fewer syscalls. Surely this can't hurt. If nothing else, it makes examining remaining system call usage simpler and opens the door to experimenting with the performance impact of different buffer sizes.
author Gregory Szorc <gregory.szorc@gmail.com>
date Sat, 26 Sep 2015 21:43:13 -0700
parents bf3d10f0c34a
children 56b2bcea2529
line wrap: on
line source

# Copyright 2006, 2007 Matt Mackall <mpm@selenic.com>
#
# This software may be used and distributed according to the terms of the
# GNU General Public License version 2 or any later version.

'''share a common history between several working directories

Automatic Pooled Storage for Clones
-----------------------------------

When this extension is active, :hg:`clone` can be configured to
automatically share/pool storage across multiple clones. This
mode effectively converts :hg:`clone` to :hg:`clone` + :hg:`share`.
The benefit of using this mode is the automatic management of
store paths and intelligent pooling of related repositories.

The following ``share.`` config options influence this feature:

``share.pool``
    Filesystem path where shared repository data will be stored. When
    defined, :hg:`clone` will automatically use shared repository
    storage instead of creating a store inside each clone.

``share.poolnaming``
    How directory names in ``share.pool`` are constructed.

    "identity" means the name is derived from the first changeset in the
    repository. In this mode, different remotes share storage if their
    root/initial changeset is identical. In this mode, the local shared
    repository is an aggregate of all encountered remote repositories.

    "remote" means the name is derived from the source repository's
    path or URL. In this mode, storage is only shared if the path or URL
    requested in the :hg:`clone` command matches exactly to a repository
    that was cloned before.

    The default naming mode is "identity."
'''

from mercurial.i18n import _
from mercurial import cmdutil, commands, hg, util, extensions, bookmarks
from mercurial.hg import repository, parseurl
import errno

cmdtable = {}
command = cmdutil.command(cmdtable)
# Note for extension authors: ONLY specify testedwith = 'internal' for
# extensions which SHIP WITH MERCURIAL. Non-mainline extensions should
# be specifying the version(s) of Mercurial they are tested with, or
# leave the attribute unspecified.
testedwith = 'internal'

@command('share',
    [('U', 'noupdate', None, _('do not create a working directory')),
     ('B', 'bookmarks', None, _('also share bookmarks'))],
    _('[-U] [-B] SOURCE [DEST]'),
    norepo=True)
def share(ui, source, dest=None, noupdate=False, bookmarks=False):
    """create a new shared repository

    Initialize a new repository and working directory that shares its
    history (and optionally bookmarks) with another repository.

    .. note::

       using rollback or extensions that destroy/modify history (mq,
       rebase, etc.) can cause considerable confusion with shared
       clones. In particular, if two shared clones are both updated to
       the same changeset, and one of them destroys that changeset
       with rollback, the other clone will suddenly stop working: all
       operations will fail with "abort: working directory has unknown
       parent". The only known workaround is to use debugsetparents on
       the broken clone to reset it to a changeset that still exists.
    """

    return hg.share(ui, source, dest, not noupdate, bookmarks)

@command('unshare', [], '')
def unshare(ui, repo):
    """convert a shared repository to a normal one

    Copy the store data to the repo and remove the sharedpath data.
    """

    if not repo.shared():
        raise util.Abort(_("this is not a shared repo"))

    destlock = lock = None
    lock = repo.lock()
    try:
        # we use locks here because if we race with commit, we
        # can end up with extra data in the cloned revlogs that's
        # not pointed to by changesets, thus causing verify to
        # fail

        destlock = hg.copystore(ui, repo, repo.path)

        sharefile = repo.join('sharedpath')
        util.rename(sharefile, sharefile + '.old')

        repo.requirements.discard('sharedpath')
        repo._writerequirements()
    finally:
        destlock and destlock.release()
        lock and lock.release()

    # update store, spath, svfs and sjoin of repo
    repo.unfiltered().__init__(repo.baseui, repo.root)

# Wrap clone command to pass auto share options.
def clone(orig, ui, source, *args, **opts):
    pool = ui.config('share', 'pool', None)
    if pool:
        pool = util.expandpath(pool)

    opts['shareopts'] = dict(
        pool=pool,
        mode=ui.config('share', 'poolnaming', 'identity'),
    )

    return orig(ui, source, *args, **opts)

def extsetup(ui):
    extensions.wrapfunction(bookmarks.bmstore, 'getbkfile', getbkfile)
    extensions.wrapfunction(bookmarks.bmstore, 'recordchange', recordchange)
    extensions.wrapfunction(bookmarks.bmstore, 'write', write)
    extensions.wrapcommand(commands.table, 'clone', clone)

def _hassharedbookmarks(repo):
    """Returns whether this repo has shared bookmarks"""
    try:
        shared = repo.vfs.read('shared').splitlines()
    except IOError as inst:
        if inst.errno != errno.ENOENT:
            raise
        return False
    return 'bookmarks' in shared

def _getsrcrepo(repo):
    """
    Returns the source repository object for a given shared repository.
    If repo is not a shared repository, return None.
    """
    if repo.sharedpath == repo.path:
        return None

    # the sharedpath always ends in the .hg; we want the path to the repo
    source = repo.vfs.split(repo.sharedpath)[0]
    srcurl, branches = parseurl(source)
    return repository(repo.ui, srcurl)

def getbkfile(orig, self, repo):
    if _hassharedbookmarks(repo):
        srcrepo = _getsrcrepo(repo)
        if srcrepo is not None:
            repo = srcrepo
    return orig(self, repo)

def recordchange(orig, self, tr):
    # Continue with write to local bookmarks file as usual
    orig(self, tr)

    if _hassharedbookmarks(self._repo):
        srcrepo = _getsrcrepo(self._repo)
        if srcrepo is not None:
            category = 'share-bookmarks'
            tr.addpostclose(category, lambda tr: self._writerepo(srcrepo))

def write(orig, self):
    # First write local bookmarks file in case we ever unshare
    orig(self)
    if _hassharedbookmarks(self._repo):
        srcrepo = _getsrcrepo(self._repo)
        if srcrepo is not None:
            self._writerepo(srcrepo)