tests/drawdag.py
author Gregory Szorc <gregory.szorc@gmail.com>
Fri, 13 Jan 2017 20:16:56 -0800
changeset 30818 4c0a5a256ae8
parent 30458 a31634336471
child 31676 d761ef24d6e1
permissions -rw-r--r--
localrepo: experimental support for non-zlib revlog compression The final part of integrating the compression manager APIs into revlog storage is the plumbing for repositories to advertise they are using non-zlib storage and for revlogs to instantiate a non-zlib compression engine. The main intent of the compression manager work was to zstd all of the things. Adding zstd to revlogs has proved to be more involved than other places because revlogs are... special. Very small inputs and the use of delta chains (which are themselves a form of compression) are a completely different use case from streaming compression, which bundles and the wire protocol employ. I've conducted numerous experiments with zstd in revlogs and have yet to formalize compression settings and a storage architecture that I'm confident I won't regret later. In other words, I'm not yet ready to commit to a new mechanism for using zstd - or any other compression format - in revlogs. That being said, having some support for zstd (and other compression formats) in revlogs in core is beneficial. It can allow others to conduct experiments. This patch introduces *highly experimental* support for non-zlib compression formats in revlogs. Introduced is a config option to control which compression engine to use. Also introduced is a namespace of "exp-compression-*" requirements to denote support for non-zlib compression in revlogs. I've prefixed the namespace with "exp-" (short for "experimental") because I'm not confident of the requirements "schema" and in no way want to give the illusion of supporting these requirements in the future. I fully intend to drop support for these requirements once we figure out what we're doing with zstd in revlogs. A good portion of the patch is teaching the requirements system about registered compression engines and passing the requested compression engine as an opener option so revlogs can instantiate the proper compression engine for new operations. That's a verbose way of saying "we can now use zstd in revlogs!" On an `hg pull` conversion of the mozilla-unified repo with no extra redelta settings (like aggressivemergedeltas), we can see the impact of zstd vs zlib in revlogs: $ hg perfrevlogchunks -c ! chunk ! wall 2.032052 comb 2.040000 user 1.990000 sys 0.050000 (best of 5) ! wall 1.866360 comb 1.860000 user 1.820000 sys 0.040000 (best of 6) ! chunk batch ! wall 1.877261 comb 1.870000 user 1.860000 sys 0.010000 (best of 6) ! wall 1.705410 comb 1.710000 user 1.690000 sys 0.020000 (best of 6) $ hg perfrevlogchunks -m ! chunk ! wall 2.721427 comb 2.720000 user 2.640000 sys 0.080000 (best of 4) ! wall 2.035076 comb 2.030000 user 1.950000 sys 0.080000 (best of 5) ! chunk batch ! wall 2.614561 comb 2.620000 user 2.580000 sys 0.040000 (best of 4) ! wall 1.910252 comb 1.910000 user 1.880000 sys 0.030000 (best of 6) $ hg perfrevlog -c -d 1 ! wall 4.812885 comb 4.820000 user 4.800000 sys 0.020000 (best of 3) ! wall 4.699621 comb 4.710000 user 4.700000 sys 0.010000 (best of 3) $ hg perfrevlog -m -d 1000 ! wall 34.252800 comb 34.250000 user 33.730000 sys 0.520000 (best of 3) ! wall 24.094999 comb 24.090000 user 23.320000 sys 0.770000 (best of 3) Only modest wins for the changelog. But manifest reading is significantly faster. What's going on? One reason might be data volume. zstd decompresses faster. So given more bytes, it will put more distance between it and zlib. Another reason is size. In the current design, zstd revlogs are *larger*: debugcreatestreamclonebundle (size in bytes) zlib: 1,638,852,492 zstd: 1,680,601,332 I haven't investigated this fully, but I reckon a significant cause of larger revlogs is that the zstd frame/header has more bytes than zlib's. For very small inputs or data that doesn't compress well, we'll tend to store more uncompressed chunks than with zlib (because the compressed size isn't smaller than original). This will make revlog reading faster because it is doing less decompression. Moving on to bundle performance: $ hg bundle -a -t none-v2 (total CPU time) zlib: 102.79s zstd: 97.75s So, marginal CPU decrease for reading all chunks in all revlogs (this is somewhat disappointing). $ hg bundle -a -t <engine>-v2 (total CPU time) zlib: 191.59s zstd: 115.36s This last test effectively measures the difference between zlib->zlib and zstd->zstd for revlogs to bundle. This is a rough approximation of what a server does during `hg clone`. There are some promising results for zstd. But not enough for me to feel comfortable advertising it to users. We'll get there...

# drawdag.py - convert ASCII revision DAG to actual changesets
#
# Copyright 2016 Facebook, Inc.
#
# This software may be used and distributed according to the terms of the
# GNU General Public License version 2 or any later version.
"""
create changesets from an ASCII graph for testing purpose.

For example, given the following input::

    c d
    |/
    b
    |
    a

4 changesets and 4 local tags will be created.
`hg log -G -T "{rev} {desc} (tag: {tags})"` will output::

    o  3 d (tag: d tip)
    |
    | o  2 c (tag: c)
    |/
    o  1 b (tag: b)
    |
    o  0 a (tag: a)

For root nodes (nodes without parents) in the graph, they can be revsets
pointing to existing nodes.  The ASCII graph could also have disconnected
components with same names referring to the same changeset.

Therefore, given the repo having the 4 changesets (and tags) above, with the
following ASCII graph as input::

    foo    bar       bar  foo
     |     /          |    |
    ancestor(c,d)     a   baz

The result (`hg log -G -T "{desc}"`) will look like::

    o    foo
    |\
    +---o  bar
    | | |
    | o |  baz
    |  /
    +---o  d
    | |
    +---o  c
    | |
    o |  b
    |/
    o  a

Note that if you take the above `hg log` output directly as input. It will work
as expected - the result would be an isomorphic graph::

    o    foo
    |\
    | | o  d
    | |/
    | | o  c
    | |/
    | | o  bar
    | |/|
    | o |  b
    | |/
    o /  baz
     /
    o  a

This is because 'o' is specially handled in the input: instead of using 'o' as
the node name, the word to the right will be used.
"""
from __future__ import absolute_import, print_function

import collections
import itertools

from mercurial.i18n import _
from mercurial import (
    cmdutil,
    context,
    error,
    node,
    scmutil,
)

cmdtable = {}
command = cmdutil.command(cmdtable)

_pipechars = '\\/+-|'
_nonpipechars = ''.join(chr(i) for i in xrange(33, 127)
                        if chr(i) not in _pipechars)

def _isname(ch):
    """char -> bool. return True if ch looks like part of a name, False
    otherwise"""
    return ch in _nonpipechars

def _parseasciigraph(text):
    """str -> {str : [str]}. convert the ASCII graph to edges"""
    lines = text.splitlines()
    edges = collections.defaultdict(list)  # {node: []}

    def get(y, x):
        """(int, int) -> char. give a coordinate, return the char. return a
        space for anything out of range"""
        if x < 0 or y < 0:
            return ' '
        try:
            return lines[y][x]
        except IndexError:
            return ' '

    def getname(y, x):
        """(int, int) -> str. like get(y, x) but concatenate left and right
        parts. if name is an 'o', try to replace it to the right"""
        result = ''
        for i in itertools.count(0):
            ch = get(y, x - i)
            if not _isname(ch):
                break
            result = ch + result
        for i in itertools.count(1):
            ch = get(y, x + i)
            if not _isname(ch):
                break
            result += ch
        if result == 'o':
            # special handling, find the name to the right
            result = ''
            for i in itertools.count(2):
                ch = get(y, x + i)
                if ch == ' ' or ch in _pipechars:
                    if result or x + i >= len(lines[y]):
                        break
                else:
                    result += ch
            return result or 'o'
        return result

    def parents(y, x):
        """(int, int) -> [str]. follow the ASCII edges at given position,
        return a list of parents"""
        visited = set([(y, x)])
        visit = []
        result = []

        def follow(y, x, expected):
            """conditionally append (y, x) to visit array, if it's a char
            in excepted. 'o' in expected means an '_isname' test.
            if '-' (or '+') is not in excepted, and get(y, x) is '-' (or '+'),
            the next line (y + 1, x) will be checked instead."""
            ch = get(y, x)
            if any(ch == c and c not in expected for c in '-+'):
                y += 1
                return follow(y + 1, x, expected)
            if ch in expected or ('o' in expected and _isname(ch)):
                visit.append((y, x))

        #  -o-  # starting point:
        #  /|\ # follow '-' (horizontally), and '/|\' (to the bottom)
        follow(y + 1, x, '|')
        follow(y + 1, x - 1, '/')
        follow(y + 1, x + 1, '\\')
        follow(y, x - 1, '-')
        follow(y, x + 1, '-')

        while visit:
            y, x = visit.pop()
            if (y, x) in visited:
                continue
            visited.add((y, x))
            ch = get(y, x)
            if _isname(ch):
                result.append(getname(y, x))
                continue
            elif ch == '|':
                follow(y + 1, x, '/|o')
                follow(y + 1, x - 1, '/')
                follow(y + 1, x + 1, '\\')
            elif ch == '+':
                follow(y, x - 1, '-')
                follow(y, x + 1, '-')
                follow(y + 1, x - 1, '/')
                follow(y + 1, x + 1, '\\')
                follow(y + 1, x, '|')
            elif ch == '\\':
                follow(y + 1, x + 1, '\\|o')
            elif ch == '/':
                follow(y + 1, x - 1, '/|o')
            elif ch == '-':
                follow(y, x - 1, '-+o')
                follow(y, x + 1, '-+o')
        return result

    for y, line in enumerate(lines):
        for x, ch in enumerate(line):
            if ch == '#':  # comment
                break
            if _isname(ch):
                edges[getname(y, x)] += parents(y, x)

    return dict(edges)

class simplefilectx(object):
    def __init__(self, path, data):
        self._data = data
        self._path = path

    def data(self):
        return self._data

    def path(self):
        return self._path

    def renamed(self):
        return None

    def flags(self):
        return ''

class simplecommitctx(context.committablectx):
    def __init__(self, repo, name, parentctxs, added=None):
        opts = {
            'changes': scmutil.status([], added or [], [], [], [], [], []),
            'date': '0 0',
            'extra': {'branch': 'default'},
        }
        super(simplecommitctx, self).__init__(self, name, **opts)
        self._repo = repo
        self._name = name
        self._parents = parentctxs
        self._parents.sort(key=lambda c: c.node())
        while len(self._parents) < 2:
            self._parents.append(repo[node.nullid])

    def filectx(self, key):
        return simplefilectx(key, self._name)

    def commit(self):
        return self._repo.commitctx(self)

def _walkgraph(edges):
    """yield node, parents in topologically order"""
    visible = set(edges.keys())
    remaining = {}  # {str: [str]}
    for k, vs in edges.iteritems():
        for v in vs:
            if v not in remaining:
                remaining[v] = []
        remaining[k] = vs[:]
    while remaining:
        leafs = [k for k, v in remaining.items() if not v]
        if not leafs:
            raise error.Abort(_('the graph has cycles'))
        for leaf in sorted(leafs):
            if leaf in visible:
                yield leaf, edges[leaf]
            del remaining[leaf]
            for k, v in remaining.iteritems():
                if leaf in v:
                    v.remove(leaf)

@command('debugdrawdag', [])
def debugdrawdag(ui, repo, **opts):
    """read an ASCII graph from stdin and create changesets

    The ASCII graph is like what :hg:`log -G` outputs, with each `o` replaced
    to the name of the node. The command will create dummy changesets and local
    tags with those names to make the dummy changesets easier to be referred
    to.

    If the name of a node is a single character 'o', It will be replaced by the
    word to the right. This makes it easier to reuse
    :hg:`log -G -T '{desc}'` outputs.

    For root (no parents) nodes, revset can be used to query existing repo.
    Note that the revset cannot have confusing characters which can be seen as
    the part of the graph edges, like `|/+-\`.
    """
    text = ui.fin.read()

    # parse the graph and make sure len(parents) <= 2 for each node
    edges = _parseasciigraph(text)
    for k, v in edges.iteritems():
        if len(v) > 2:
            raise error.Abort(_('%s: too many parents: %s')
                              % (k, ' '.join(v)))

    committed = {None: node.nullid}  # {name: node}

    # for leaf nodes, try to find existing nodes in repo
    for name, parents in edges.iteritems():
        if len(parents) == 0:
            try:
                committed[name] = scmutil.revsingle(repo, name)
            except error.RepoLookupError:
                pass

    # commit in topological order
    for name, parents in _walkgraph(edges):
        if name in committed:
            continue
        pctxs = [repo[committed[n]] for n in parents]
        ctx = simplecommitctx(repo, name, pctxs, [name])
        n = ctx.commit()
        committed[name] = n
        repo.tag(name, n, message=None, user=None, date=None, local=True)