view mercurial/repoview.py @ 39506:b66ea3fc3a86

sparse-revlog: set max delta chain length to on thousand The new snapshot system used in the sparse-revlog case gave us some small size benefit so far. However its most important property is to gracefully handle harder limit on delta chainlength. Long delta chain has a very detrimental impact on read (and write) performance in revlog. Being able to shorter them provide a great boost. However, shorting delta used to result significantly lower compression ratio. The intermediate snapshots effectively suppress most of this effect (even all in some case). # Effect on the test repository The repository we use for test is not "realistic" but can still show this in action using an unreasonably low chain limit. Limiting the chain length show a sizeable increase but stay under control: +6% for limit=15; +15% for limit=10. Without the snapshot system the increase is significantly bigger: +45% for limit=15; +80% for limit=10. Even slightly larger than without delta chain limit, the resulting size is still smaller than before we started doing snapshots. Here is a table for comparison. *Since the repository is not branchy, the initial sparse-revlog version does not bring much benefit compare to the non-sparse one): chain length limit | none | limit=15 | limit=10 | without sparse-revlog | 62 818 987 | 112 664 615 | 131 222 574 | without snapshot | 74 365 490 | 108 211 410 | 133 857 764 | with snapshot | 59 230 936 | 63 002 924 | 68 415 329 | # Effect On Real Life Repositories The series provides significant benefits on all kind of repositories. Using `hg debugupgraderepo -o redeltaparent --run`, we recomputed delta chain for various repositories with different settings: - delta chain length: unlimited or 1000 limit - sparse-revlog: enabled or disabled - this series: applied or not applied We can observe multiple types of effect: - On very branchy repositories: * The delta chain limit as low impact on the repo size. * Intermediate snapshot greatly reduces manifest size: - pypy: -80% - netbeans: -95% * The delta chain limit is effective, without a size impact: - netbeans average: 613 -> 282 - private #1 average: 1 068 -> 307 - On more linear repository: * Intermediate snapshot limit the impact of delta chain limit: - mozilla: without the series: +360% with the series: +25% * The delta chain limit provides large improvement: - mozilla's average chain length: unlimited: 15 338 limited: 469 * Despite the chain length limit, the manifest size is reduced: - mercurial: -25% - mozilla: -30% It is clear that the use of chains of intermediate snapshots provide large benefits both in storage size and delta chains quality. We should now switch our effort toward making sure the write performance are acceptable. Then, `sparse-revlog` will be a suitable format for all new repository. # Raw Statistic * no-sparse: general delta repository not using sparse-revlog * no-snapshot: sparse-revlog repository not using this series * snapshot: sparse-revlog repository using this series mercurial Manifest Size: limit | none | 1000 ------------|-------------|------------ no-sparse | 8 021 373 | 8 199 366 no-snapshot | 8 103 561 | 8 259 719 snapshot | 6 137 116 | 6 126 433 Manifest Chain length data limit || none || 1000 || value || average | max || average | max || ------------||---------|---------||---------|---------|| no-sparse || 307 | 1456 || 279 | 1000 || no-snapshot || 312 | 1456 || 283 | 1000 || snapshot || 248 | 1208 || 241 | 1000 || Full Store Size limit | none | 1000 ------------|-------------|------------ no-sparse | 51 013 198 | 51 201 574 no-snapshot | 50 930 795 | 51 141 006 snapshot | 48 072 037 | 48 093 572 pypy Manifest Size: limit | none | 1000 ------------|-------------|------------ no-sparse | 193 987 784 | 193 987 784 no-snapshot | 163 171 745 | 163 312 229 snapshot | 34 605 900 | 34 600 750 Manifest Chain length data limit || none || 1000 || value || average | max || average | max || ------------||---------|---------||---------|---------|| no-sparse || 101 | 692 || 101 | 692 || no-snapshot || 151 | 1307 || 148 | 1000 || snapshot || 128 | 1309 || 125 | 1000 || Full Store Size limit | none | 1000 ------------|-------------|------------ no-sparse | 495 931 473 | 495 931 473 no-snapshot | 465 441 017 | 465 581 501 snapshot | 355 467 301 | 355 472 451 Mozilla Manifest Size: limit | none | 1000 ------------|----------------|--------------- no-sparse | 416 757 148 | 1 869 009 668 no-snapshot | 401 592 370 | 1 843 493 795 snapshot | 224 359 521 | 284 615 500 Manifest Chain length data limit || none || 1000 || value || average | max || average | max || ------------||---------|---------||---------|---------|| no-sparse || 15 333 | 58 980 || 468 | 1 000 || no-snapshot || 15 336 | 58 980 || 469 | 1 000 || snapshot || 15 338 | 58 983 || 469 | 1 000 || Full Store Size limit | none | 1000 ------------|----------------|--------------- no-sparse | 2 712 477 887 | 4 164 995 451 no-snapshot | 2 698 887 835 | 4 141 054 304 snapshot | 2 518 130 385 | 2 578 587 596 Netbeans Manifest Size: limit | none | 1000 ------------|----------------|--------------- no-sparse | 4 766 794 101 | 4 870 642 687 no-snapshot | 4 334 806 082 | 4 428 681 309 snapshot | 232 659 666 | 240 330 665 Manifest Chain length data limit || none || 1000 || value || average | max || average | max || ------------||---------|---------||---------|---------|| no-sparse || 597 | 6802 || 254 | 1 000 || no-snapshot || 648 | 6 802 || 305 | 1 000 || snapshot || 613 | 6 804 || 282 | 1 000 || Full Store Size limit | none | 1000 ------------|----------------|--------------- no-sparse | 5 807 347 998 | 5 911 196 584 no-snapshot | 5 375 398 602 | 5 469 273 829 snapshot | 1 282 519 928 | 1 290 190 927 Private repo #1 Manifest Size: limit | none | 1000 ------------|-----------------|--------------- no-sparse | 41 389 010 840 | 41 398 162 091 no-snapshot | 9 737 319 435 | 10 223 773 150 snapshot | 744 215 807 | 747 961 822 Manifest Chain length data limit || none || 1000 || value || average | max || average | max || ------------||---------|---------||---------|---------|| no-sparse || 245 | 8 885 || 81 | 1 000 || no-snapshot || 1 225 | 8 885 || 336 | 1 000 || snapshot || 1 068 | 7 909 || 307 | 1 000 || Full Store Size limit | none | 1000 ------------|----------------|--------------- no-sparse | 49 646 065 126 | 49 655 216 377 no-snapshot | 17 924 862 856 | 18 411 316 571 snapshot | 9 009 024 710 | 9 012 770 725 Private repo #2 We currently have less data available for that repository. * Before is a sparse-revlog repository without this series * After is a sparse-revlog repository with this series + 1000 chain limit Manifest Size: Before: 1 531 485 040 bytes After: 1 091 422 451 bytes Manifest Chain: Before: 2 218 avg; 6 575 Max After: 442 avg; 1 000 Max Full Store Size Before: 15 203 955 615 after: 8 207 180 693
author Boris Feld <boris.feld@octobus.net>
date Fri, 07 Sep 2018 11:18:45 -0400
parents 06c976acc581
children bc15e37ecc16
line wrap: on
line source

# repoview.py - Filtered view of a localrepo object
#
# Copyright 2012 Pierre-Yves David <pierre-yves.david@ens-lyon.org>
#                Logilab SA        <contact@logilab.fr>
#
# 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 copy
import weakref

from .node import nullrev
from . import (
    obsolete,
    phases,
    pycompat,
    tags as tagsmod,
)

def hideablerevs(repo):
    """Revision candidates to be hidden

    This is a standalone function to allow extensions to wrap it.

    Because we use the set of immutable changesets as a fallback subset in
    branchmap (see mercurial.branchmap.subsettable), you cannot set "public"
    changesets as "hideable". Doing so would break multiple code assertions and
    lead to crashes."""
    obsoletes = obsolete.getrevs(repo, 'obsolete')
    internals = repo._phasecache.getrevset(repo, phases.localhiddenphases)
    internals = frozenset(internals)
    return obsoletes | internals

def pinnedrevs(repo):
    """revisions blocking hidden changesets from being filtered
    """

    cl = repo.changelog
    pinned = set()
    pinned.update([par.rev() for par in repo[None].parents()])
    pinned.update([cl.rev(bm) for bm in repo._bookmarks.values()])

    tags = {}
    tagsmod.readlocaltags(repo.ui, repo, tags, {})
    if tags:
        rev, nodemap = cl.rev, cl.nodemap
        pinned.update(rev(t[0]) for t in tags.values() if t[0] in nodemap)
    return pinned


def _revealancestors(pfunc, hidden, revs):
    """reveals contiguous chains of hidden ancestors of 'revs' by removing them
    from 'hidden'

    - pfunc(r): a funtion returning parent of 'r',
    - hidden: the (preliminary) hidden revisions, to be updated
    - revs: iterable of revnum,

    (Ancestors are revealed exclusively, i.e. the elements in 'revs' are
    *not* revealed)
    """
    stack = list(revs)
    while stack:
        for p in pfunc(stack.pop()):
            if p != nullrev and p in hidden:
                hidden.remove(p)
                stack.append(p)

def computehidden(repo, visibilityexceptions=None):
    """compute the set of hidden revision to filter

    During most operation hidden should be filtered."""
    assert not repo.changelog.filteredrevs

    hidden = hideablerevs(repo)
    if hidden:
        hidden = set(hidden - pinnedrevs(repo))
        if visibilityexceptions:
            hidden -= visibilityexceptions
        pfunc = repo.changelog.parentrevs
        mutable = repo._phasecache.getrevset(repo, phases.mutablephases)

        visible = mutable - hidden
        _revealancestors(pfunc, hidden, visible)
    return frozenset(hidden)

def computeunserved(repo, visibilityexceptions=None):
    """compute the set of revision that should be filtered when used a server

    Secret and hidden changeset should not pretend to be here."""
    assert not repo.changelog.filteredrevs
    # fast path in simple case to avoid impact of non optimised code
    hiddens = filterrevs(repo, 'visible')
    if phases.hassecret(repo):
        secrets = repo._phasecache.getrevset(repo, phases.remotehiddenphases)
        return frozenset(hiddens | frozenset(secrets))
    else:
        return hiddens

def computemutable(repo, visibilityexceptions=None):
    assert not repo.changelog.filteredrevs
    # fast check to avoid revset call on huge repo
    if any(repo._phasecache.phaseroots[1:]):
        getphase = repo._phasecache.phase
        maymutable = filterrevs(repo, 'base')
        return frozenset(r for r in maymutable if getphase(repo, r))
    return frozenset()

def computeimpactable(repo, visibilityexceptions=None):
    """Everything impactable by mutable revision

    The immutable filter still have some chance to get invalidated. This will
    happen when:

    - you garbage collect hidden changeset,
    - public phase is moved backward,
    - something is changed in the filtering (this could be fixed)

    This filter out any mutable changeset and any public changeset that may be
    impacted by something happening to a mutable revision.

    This is achieved by filtered everything with a revision number egal or
    higher than the first mutable changeset is filtered."""
    assert not repo.changelog.filteredrevs
    cl = repo.changelog
    firstmutable = len(cl)
    for roots in repo._phasecache.phaseroots[1:]:
        if roots:
            firstmutable = min(firstmutable, min(cl.rev(r) for r in roots))
    # protect from nullrev root
    firstmutable = max(0, firstmutable)
    return frozenset(pycompat.xrange(firstmutable, len(cl)))

# function to compute filtered set
#
# When adding a new filter you MUST update the table at:
#     mercurial.branchmap.subsettable
# Otherwise your filter will have to recompute all its branches cache
# from scratch (very slow).
filtertable = {'visible': computehidden,
               'visible-hidden': computehidden,
               'served': computeunserved,
               'immutable':  computemutable,
               'base':  computeimpactable}

def filterrevs(repo, filtername, visibilityexceptions=None):
    """returns set of filtered revision for this filter name

    visibilityexceptions is a set of revs which must are exceptions for
    hidden-state and must be visible. They are dynamic and hence we should not
    cache it's result"""
    if filtername not in repo.filteredrevcache:
        func = filtertable[filtername]
        if visibilityexceptions:
            return func(repo.unfiltered, visibilityexceptions)
        repo.filteredrevcache[filtername] = func(repo.unfiltered())
    return repo.filteredrevcache[filtername]

class repoview(object):
    """Provide a read/write view of a repo through a filtered changelog

    This object is used to access a filtered version of a repository without
    altering the original repository object itself. We can not alter the
    original object for two main reasons:
    - It prevents the use of a repo with multiple filters at the same time. In
      particular when multiple threads are involved.
    - It makes scope of the filtering harder to control.

    This object behaves very closely to the original repository. All attribute
    operations are done on the original repository:
    - An access to `repoview.someattr` actually returns `repo.someattr`,
    - A write to `repoview.someattr` actually sets value of `repo.someattr`,
    - A deletion of `repoview.someattr` actually drops `someattr`
      from `repo.__dict__`.

    The only exception is the `changelog` property. It is overridden to return
    a (surface) copy of `repo.changelog` with some revisions filtered. The
    `filtername` attribute of the view control the revisions that need to be
    filtered.  (the fact the changelog is copied is an implementation detail).

    Unlike attributes, this object intercepts all method calls. This means that
    all methods are run on the `repoview` object with the filtered `changelog`
    property. For this purpose the simple `repoview` class must be mixed with
    the actual class of the repository. This ensures that the resulting
    `repoview` object have the very same methods than the repo object. This
    leads to the property below.

        repoview.method() --> repo.__class__.method(repoview)

    The inheritance has to be done dynamically because `repo` can be of any
    subclasses of `localrepo`. Eg: `bundlerepo` or `statichttprepo`.
    """

    def __init__(self, repo, filtername, visibilityexceptions=None):
        object.__setattr__(self, r'_unfilteredrepo', repo)
        object.__setattr__(self, r'filtername', filtername)
        object.__setattr__(self, r'_clcachekey', None)
        object.__setattr__(self, r'_clcache', None)
        # revs which are exceptions and must not be hidden
        object.__setattr__(self, r'_visibilityexceptions',
                           visibilityexceptions)

    # not a propertycache on purpose we shall implement a proper cache later
    @property
    def changelog(self):
        """return a filtered version of the changeset

        this changelog must not be used for writing"""
        # some cache may be implemented later
        unfi = self._unfilteredrepo
        unfichangelog = unfi.changelog
        # bypass call to changelog.method
        unfiindex = unfichangelog.index
        unfilen = len(unfiindex)
        unfinode = unfiindex[unfilen - 1][7]

        revs = filterrevs(unfi, self.filtername, self._visibilityexceptions)
        cl = self._clcache
        newkey = (unfilen, unfinode, hash(revs), unfichangelog._delayed)
        # if cl.index is not unfiindex, unfi.changelog would be
        # recreated, and our clcache refers to garbage object
        if (cl is not None and
            (cl.index is not unfiindex or newkey != self._clcachekey)):
            cl = None
        # could have been made None by the previous if
        if cl is None:
            cl = copy.copy(unfichangelog)
            cl.filteredrevs = revs
            object.__setattr__(self, r'_clcache', cl)
            object.__setattr__(self, r'_clcachekey', newkey)
        return cl

    def unfiltered(self):
        """Return an unfiltered version of a repo"""
        return self._unfilteredrepo

    def filtered(self, name, visibilityexceptions=None):
        """Return a filtered version of a repository"""
        if name == self.filtername and not visibilityexceptions:
            return self
        return self.unfiltered().filtered(name, visibilityexceptions)

    def __repr__(self):
        return r'<%s:%s %r>' % (self.__class__.__name__,
                                pycompat.sysstr(self.filtername),
                                self.unfiltered())

    # everything access are forwarded to the proxied repo
    def __getattr__(self, attr):
        return getattr(self._unfilteredrepo, attr)

    def __setattr__(self, attr, value):
        return setattr(self._unfilteredrepo, attr, value)

    def __delattr__(self, attr):
        return delattr(self._unfilteredrepo, attr)

# Python <3.4 easily leaks types via __mro__. See
# https://bugs.python.org/issue17950. We cache dynamically created types
# so they won't be leaked on every invocation of repo.filtered().
_filteredrepotypes = weakref.WeakKeyDictionary()

def newtype(base):
    """Create a new type with the repoview mixin and the given base class"""
    if base not in _filteredrepotypes:
        class filteredrepo(repoview, base):
            pass
        _filteredrepotypes[base] = filteredrepo
    return _filteredrepotypes[base]