view mercurial/simplemerge.py @ 26623:5a95fe44121d

clonebundles: support for seeding clones from pre-generated bundles

Cloning can be an expensive operation for servers because the server
generates a bundle from existing repository data at request time. For
a large repository like mozilla-central, this consumes 4+ minutes
of CPU time on the server. It also results in significant network
utilization. Multiplied by hundreds or even thousands of clients and
the ensuing load can result in difficulties scaling the Mercurial server.

Despite generation of bundles being deterministic until the next
changeset is added, the generation of bundles to service a clone request
is not cached. Each clone thus performs redundant work. This is
wasteful.

This patch introduces the "clonebundles" extension and related
client-side functionality to help alleviate this deficiency. The
client-side feature is behind an experimental flag and is not enabled by
default.

It works as follows:

1) Server operator generates a bundle and makes it available on a
   server (likely HTTP).
2) Server operator defines the URL of a bundle file in a
   .hg/clonebundles.manifest file.
3) Client `hg clone`ing sees the server is advertising bundle URLs.
4) Client fetches and applies the advertised bundle.
5) Client performs equivalent of `hg pull` to fetch changes made since
   the bundle was created.

Essentially, the server performs the expensive work of generating a
bundle once and all subsequent clones fetch a static file from
somewhere. Scaling static file serving is a much more manageable
problem than scaling a Python application like Mercurial. Assuming your
repository grows less than 1% per day, the end result is 99+% of CPU
and network load from clones is eliminated, allowing Mercurial servers
to scale more easily. Serving static files also means data can be
transferred to clients as fast as they can consume it, rather than as
fast as servers can generate it. This makes clones faster.

Mozilla has implemented similar functionality of this patch on
hg.mozilla.org using a custom extension. We are hosting bundle files in
Amazon S3 and CloudFront (a CDN) and have successfully offloaded
>1 TB/day in data transfer from hg.mozilla.org, freeing up significant
bandwidth and CPU resources. The positive impact has been stellar and
I believe it has proved its value to be included in Mercurial core. I
feel it is important for the client-side support to be enabled in core
by default because it means that clients will get faster, more reliable
clones and will enable server operators to reduce load without
requiring any client-side configuration changes (assuming clients are
up to date, of course).

The scope of this feature is narrowly and specifically tailored to
cloning, despite "serve pulls from pre-generated bundles" being a valid
and useful feature. I would eventually like for Mercurial servers to
support transferring *all* repository data via statically hosted files.
You could imagine a server that siphons all pushed data to bundle files
and instructs clients to apply a stream of bundles to reconstruct all
repository data. This feature, while useful and powerful, is
significantly more work to implement because it requires the server
component have awareness of discovery and a mapping of which changesets
are in which files. Full, clone bundles, by contrast, are much simpler.

The wire protocol command is named "clonebundles" instead of something
more generic like "staticbundles" to leave the door open for a new, more
powerful and more generic server-side component with minimal backwards
compatibility implications. The name "bundleclone" is used by Mozilla's
extension and would cause problems since there are subtle differences
in Mozilla's extension.

Mozilla's experience with this idea has taught us that some form of
"content negotiation" is required. Not all clients will support all
bundle formats or even URLs (advanced TLS requirements, etc). To ensure
the highest uptake possible, a server needs to advertise multiple
versions of bundles and clients need to be able to choose the most
appropriate from that list one. The "attributes" in each
server-advertised entry facilitate this filtering and sorting. Their
use will become apparent in subsequent patches.

Initial inspiration and credit for the idea of cloning from static files
belongs to Augie Fackler and his "lookaside clone" extension proof of
concept.

# Copyright (C) 2004, 2005 Canonical Ltd
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, see <http://www.gnu.org/licenses/>.

# mbp: "you know that thing where cvs gives you conflict markers?"
# s: "i hate that."

from __future__ import absolute_import

import os
import sys

from .i18n import _
from . import (
    error,
    mdiff,
    scmutil,
    util,
)

class CantReprocessAndShowBase(Exception):
    pass

def intersect(ra, rb):
    """Given two ranges return the range where they intersect or None.

    >>> intersect((0, 10), (0, 6))
    (0, 6)
    >>> intersect((0, 10), (5, 15))
    (5, 10)
    >>> intersect((0, 10), (10, 15))
    >>> intersect((0, 9), (10, 15))
    >>> intersect((0, 9), (7, 15))
    (7, 9)
    """
    assert ra[0] <= ra[1]
    assert rb[0] <= rb[1]

    sa = max(ra[0], rb[0])
    sb = min(ra[1], rb[1])
    if sa < sb:
        return sa, sb
    else:
        return None

def compare_range(a, astart, aend, b, bstart, bend):
    """Compare a[astart:aend] == b[bstart:bend], without slicing.
    """
    if (aend - astart) != (bend - bstart):
        return False
    for ia, ib in zip(xrange(astart, aend), xrange(bstart, bend)):
        if a[ia] != b[ib]:
            return False
    else:
        return True

class Merge3Text(object):
    """3-way merge of texts.

    Given strings BASE, OTHER, THIS, tries to produce a combined text
    incorporating the changes from both BASE->OTHER and BASE->THIS."""
    def __init__(self, basetext, atext, btext, base=None, a=None, b=None):
        self.basetext = basetext
        self.atext = atext
        self.btext = btext
        if base is None:
            base = mdiff.splitnewlines(basetext)
        if a is None:
            a = mdiff.splitnewlines(atext)
        if b is None:
            b = mdiff.splitnewlines(btext)
        self.base = base
        self.a = a
        self.b = b

    def merge_lines(self,
                    name_a=None,
                    name_b=None,
                    name_base=None,
                    start_marker='<<<<<<<',
                    mid_marker='=======',
                    end_marker='>>>>>>>',
                    base_marker=None,
                    localorother=None):
        """Return merge in cvs-like form.
        """
        self.conflicts = False
        newline = '\n'
        if len(self.a) > 0:
            if self.a[0].endswith('\r\n'):
                newline = '\r\n'
            elif self.a[0].endswith('\r'):
                newline = '\r'
        if name_a and start_marker:
            start_marker = start_marker + ' ' + name_a
        if name_b and end_marker:
            end_marker = end_marker + ' ' + name_b
        if name_base and base_marker:
            base_marker = base_marker + ' ' + name_base
        merge_regions = self.merge_regions()
        for t in merge_regions:
            what = t[0]
            if what == 'unchanged':
                for i in range(t[1], t[2]):
                    yield self.base[i]
            elif what == 'a' or what == 'same':
                for i in range(t[1], t[2]):
                    yield self.a[i]
            elif what == 'b':
                for i in range(t[1], t[2]):
                    yield self.b[i]
            elif what == 'conflict':
                if localorother == 'local':
                    for i in range(t[3], t[4]):
                        yield self.a[i]
                elif localorother == 'other':
                    for i in range(t[5], t[6]):
                        yield self.b[i]
                else:
                    self.conflicts = True
                    if start_marker is not None:
                        yield start_marker + newline
                    for i in range(t[3], t[4]):
                        yield self.a[i]
                    if base_marker is not None:
                        yield base_marker + newline
                        for i in range(t[1], t[2]):
                            yield self.base[i]
                    if mid_marker is not None:
                        yield mid_marker + newline
                    for i in range(t[5], t[6]):
                        yield self.b[i]
                    if end_marker is not None:
                        yield end_marker + newline
            else:
                raise ValueError(what)

    def merge_groups(self):
        """Yield sequence of line groups.  Each one is a tuple:

        'unchanged', lines
             Lines unchanged from base

        'a', lines
             Lines taken from a

        'same', lines
             Lines taken from a (and equal to b)

        'b', lines
             Lines taken from b

        'conflict', base_lines, a_lines, b_lines
             Lines from base were changed to either a or b and conflict.
        """
        for t in self.merge_regions():
            what = t[0]
            if what == 'unchanged':
                yield what, self.base[t[1]:t[2]]
            elif what == 'a' or what == 'same':
                yield what, self.a[t[1]:t[2]]
            elif what == 'b':
                yield what, self.b[t[1]:t[2]]
            elif what == 'conflict':
                yield (what,
                       self.base[t[1]:t[2]],
                       self.a[t[3]:t[4]],
                       self.b[t[5]:t[6]])
            else:
                raise ValueError(what)

    def merge_regions(self):
        """Return sequences of matching and conflicting regions.

        This returns tuples, where the first value says what kind we
        have:

        'unchanged', start, end
             Take a region of base[start:end]

        'same', astart, aend
             b and a are different from base but give the same result

        'a', start, end
             Non-clashing insertion from a[start:end]

        Method is as follows:

        The two sequences align only on regions which match the base
        and both descendants.  These are found by doing a two-way diff
        of each one against the base, and then finding the
        intersections between those regions.  These "sync regions"
        are by definition unchanged in both and easily dealt with.

        The regions in between can be in any of three cases:
        conflicted, or changed on only one side.
        """

        # section a[0:ia] has been disposed of, etc
        iz = ia = ib = 0

        for region in self.find_sync_regions():
            zmatch, zend, amatch, aend, bmatch, bend = region
            #print 'match base [%d:%d]' % (zmatch, zend)

            matchlen = zend - zmatch
            assert matchlen >= 0
            assert matchlen == (aend - amatch)
            assert matchlen == (bend - bmatch)

            len_a = amatch - ia
            len_b = bmatch - ib
            len_base = zmatch - iz
            assert len_a >= 0
            assert len_b >= 0
            assert len_base >= 0

            #print 'unmatched a=%d, b=%d' % (len_a, len_b)

            if len_a or len_b:
                # try to avoid actually slicing the lists
                equal_a = compare_range(self.a, ia, amatch,
                                        self.base, iz, zmatch)
                equal_b = compare_range(self.b, ib, bmatch,
                                        self.base, iz, zmatch)
                same = compare_range(self.a, ia, amatch,
                                     self.b, ib, bmatch)

                if same:
                    yield 'same', ia, amatch
                elif equal_a and not equal_b:
                    yield 'b', ib, bmatch
                elif equal_b and not equal_a:
                    yield 'a', ia, amatch
                elif not equal_a and not equal_b:
                    yield 'conflict', iz, zmatch, ia, amatch, ib, bmatch
                else:
                    raise AssertionError("can't handle a=b=base but unmatched")

                ia = amatch
                ib = bmatch
            iz = zmatch

            # if the same part of the base was deleted on both sides
            # that's OK, we can just skip it.


            if matchlen > 0:
                assert ia == amatch
                assert ib == bmatch
                assert iz == zmatch

                yield 'unchanged', zmatch, zend
                iz = zend
                ia = aend
                ib = bend

    def find_sync_regions(self):
        """Return a list of sync regions, where both descendants match the base.

        Generates a list of (base1, base2, a1, a2, b1, b2).  There is
        always a zero-length sync region at the end of all the files.
        """

        ia = ib = 0
        amatches = mdiff.get_matching_blocks(self.basetext, self.atext)
        bmatches = mdiff.get_matching_blocks(self.basetext, self.btext)
        len_a = len(amatches)
        len_b = len(bmatches)

        sl = []

        while ia < len_a and ib < len_b:
            abase, amatch, alen = amatches[ia]
            bbase, bmatch, blen = bmatches[ib]

            # there is an unconflicted block at i; how long does it
            # extend?  until whichever one ends earlier.
            i = intersect((abase, abase + alen), (bbase, bbase + blen))
            if i:
                intbase = i[0]
                intend = i[1]
                intlen = intend - intbase

                # found a match of base[i[0], i[1]]; this may be less than
                # the region that matches in either one
                assert intlen <= alen
                assert intlen <= blen
                assert abase <= intbase
                assert bbase <= intbase

                asub = amatch + (intbase - abase)
                bsub = bmatch + (intbase - bbase)
                aend = asub + intlen
                bend = bsub + intlen

                assert self.base[intbase:intend] == self.a[asub:aend], \
                       (self.base[intbase:intend], self.a[asub:aend])

                assert self.base[intbase:intend] == self.b[bsub:bend]

                sl.append((intbase, intend,
                           asub, aend,
                           bsub, bend))

            # advance whichever one ends first in the base text
            if (abase + alen) < (bbase + blen):
                ia += 1
            else:
                ib += 1

        intbase = len(self.base)
        abase = len(self.a)
        bbase = len(self.b)
        sl.append((intbase, intbase, abase, abase, bbase, bbase))

        return sl

    def find_unconflicted(self):
        """Return a list of ranges in base that are not conflicted."""
        am = mdiff.get_matching_blocks(self.basetext, self.atext)
        bm = mdiff.get_matching_blocks(self.basetext, self.btext)

        unc = []

        while am and bm:
            # there is an unconflicted block at i; how long does it
            # extend?  until whichever one ends earlier.
            a1 = am[0][0]
            a2 = a1 + am[0][2]
            b1 = bm[0][0]
            b2 = b1 + bm[0][2]
            i = intersect((a1, a2), (b1, b2))
            if i:
                unc.append(i)

            if a2 < b2:
                del am[0]
            else:
                del bm[0]

        return unc

def simplemerge(ui, local, base, other, **opts):
    def readfile(filename):
        f = open(filename, "rb")
        text = f.read()
        f.close()
        if util.binary(text):
            msg = _("%s looks like a binary file.") % filename
            if not opts.get('quiet'):
                ui.warn(_('warning: %s\n') % msg)
            if not opts.get('text'):
                raise error.Abort(msg)
        return text

    mode = opts.get('mode','merge')
    if mode == 'union':
        name_a = None
        name_b = None
        name_base = None
    else:
        name_a = local
        name_b = other
        name_base = None
        labels = opts.get('label', [])
        if len(labels) > 0:
            name_a = labels[0]
        if len(labels) > 1:
            name_b = labels[1]
        if len(labels) > 2:
            name_base = labels[2]
        if len(labels) > 3:
            raise error.Abort(_("can only specify three labels."))

    try:
        localtext = readfile(local)
        basetext = readfile(base)
        othertext = readfile(other)
    except error.Abort:
        return 1

    local = os.path.realpath(local)
    if not opts.get('print'):
        opener = scmutil.opener(os.path.dirname(local))
        out = opener(os.path.basename(local), "w", atomictemp=True)
    else:
        out = sys.stdout

    m3 = Merge3Text(basetext, localtext, othertext)
    extrakwargs = {"localorother": opts.get("localorother", None)}
    if mode == 'union':
        extrakwargs['start_marker'] = None
        extrakwargs['mid_marker'] = None
        extrakwargs['end_marker'] = None
    elif name_base is not None:
        extrakwargs['base_marker'] = '|||||||'
        extrakwargs['name_base'] = name_base
    for line in m3.merge_lines(name_a=name_a, name_b=name_b, **extrakwargs):
        out.write(line)

    if not opts.get('print'):
        out.close()

    if m3.conflicts and not mode == 'union':
        return 1