view mercurial/dagparser.py @ 42745:4d20b1fe8a72

rust-discovery: using from Python code

As previously done in other topics, the Rust version is used if it's been
built.

The version fully in Rust of the partialdiscovery class has the performance
advantage over the Python version (actually using the Rust MissingAncestor) if
the undecided set is big enough. Otherwise no sampling occurs, and the
discovery is reasonably fast anyway.

Note: it's hard to predict the size of the initial undecided set, it can
depend on the kind of topological changes between the local and remote graphs.
The point of the Rust version is to make the bad cases acceptable.

More specifically, the performance advantages are:

- faster sampling, especially takefullsample()
- much faster addmissings() in almost all cases (see commit message in
  grandparent of the present changeset)
- no conversion cost of the undecided set at the interface between Rust and
  Python

== Measurements with big undecided sets

For an extreme example, discovery between mozilla-try and mozilla-unified
(over one million undecided revisions, same case as in dbd0fcca6dfc), we
get roughly a x2.5/x3 better performance:

Growing sample size (5% starting with 200): time goes down from
210 to 72 seconds.
Constant sample size of 200: time down from 1853 to 659 seconds.

With a sample size computed from number of roots and heads of the
undecided set (`respectsize` is `False`), here are perfdiscovery results:

Before ! wall 9.358729 comb 9.360000 user 9.310000 sys 0.050000 (median of 50)
After  ! wall 3.793819 comb 3.790000 user 3.750000 sys 0.040000 (median of 50)

In that later case, the sample sizes are routinely in the hundreds of
thousands of revisions. While still faster, the Rust iteration in
addmissings has less of an advantage than with smaller sample sizes, but
one sees addcommons becoming faster, probably a consequence of not having
to copy big sets back and forth.

This example is not a goal in itself, but it showcases several different
areas in which the process can become slow, due to different factors, and
how this full Rust version can help.

== Measurements with small undecided sets

In cases the undecided set is small enough than no sampling occurs,
the Rust version has a disadvantage at init if `targetheads` is really big
(some time is lost in the translation to Rust data structures),
and that is compensated by the faster `addmissings()`.

On a private repository with over one million commits, we still get a minor
improvement, of 6.8%:

  Before ! wall 0.593585 comb 0.590000 user 0.550000 sys 0.040000 (median of 50)
  After  ! wall 0.553035 comb 0.550000 user 0.520000 sys 0.030000 (median of 50)

What's interesting in that case is the first addinfo() at 180ms for Rust and
233ms for Python+C, mostly due to add_missings and the children cache
computation being done in less than 0.2ms on the Rust side vs over 40ms on the
Python side.

The worst case we have on hand is with mozilla-try, prepared with
discovery-helper.sh for 10 heads and depth 10, time goes up 2.2% on the median.
In this case `targetheads` is really huge with 165842 server heads.

  Before ! wall 0.823884 comb 0.810000 user 0.790000 sys 0.020000 (median of 50)
  After  ! wall 0.842607 comb 0.840000 user 0.800000 sys 0.040000 (median of 50)

If that would be considered a problem, more adjustments can be made, which are
prematurate at this stage: cooking special variants of methods of the inner
MissingAncestors object, retrieving local heads directly from Rust to avoid
the cost of conversion. Effort would probably be better spent at this point
improving the surroundings if needed.

Here's another data point with a smaller repository, pypy, where performance
is almost identical

 Before ! wall 0.015121 comb 0.030000 user 0.020000 sys 0.010000 (median of 186)
 After  ! wall 0.015009 comb 0.010000 user 0.010000 sys 0.000000 (median of 184)

Differential Revision: https://phab.mercurial-scm.org/D6430

# dagparser.py - parser and generator for concise description of DAGs
#
# Copyright 2010 Peter Arrenbrecht <[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 re
import string

from .i18n import _
from . import (
    error,
    pycompat,
)
from .utils import (
    stringutil,
)

def parsedag(desc):
    '''parses a DAG from a concise textual description; generates events

    "+n" is a linear run of n nodes based on the current default parent
    "." is a single node based on the current default parent
    "$" resets the default parent to -1 (implied at the start);
        otherwise the default parent is always the last node created
    "<p" sets the default parent to the backref p
    "*p" is a fork at parent p, where p is a backref
    "*p1/p2/.../pn" is a merge of parents p1..pn, where the pi are backrefs
    "/p2/.../pn" is a merge of the preceding node and p2..pn
    ":name" defines a label for the preceding node; labels can be redefined
    "@text" emits an annotation event for text
    "!command" emits an action event for the current node
    "!!my command\n" is like "!", but to the end of the line
    "#...\n" is a comment up to the end of the line

    Whitespace between the above elements is ignored.

    A backref is either
     * a number n, which references the node curr-n, where curr is the current
       node, or
     * the name of a label you placed earlier using ":name", or
     * empty to denote the default parent.

    All string valued-elements are either strictly alphanumeric, or must
    be enclosed in double quotes ("..."), with "\" as escape character.

    Generates sequence of

      ('n', (id, [parentids])) for node creation
      ('l', (id, labelname)) for labels on nodes
      ('a', text) for annotations
      ('c', command) for actions (!)
      ('C', command) for line actions (!!)

    Examples
    --------

    Example of a complex graph (output not shown for brevity):

        >>> len(list(parsedag(b"""
        ...
        ... +3         # 3 nodes in linear run
        ... :forkhere  # a label for the last of the 3 nodes from above
        ... +5         # 5 more nodes on one branch
        ... :mergethis # label again
        ... <forkhere  # set default parent to labeled fork node
        ... +10        # 10 more nodes on a parallel branch
        ... @stable    # following nodes will be annotated as "stable"
        ... +5         # 5 nodes in stable
        ... !addfile   # custom command; could trigger new file in next node
        ... +2         # two more nodes
        ... /mergethis # merge last node with labeled node
        ... +4         # 4 more nodes descending from merge node
        ...
        ... """)))
        34

    Empty list:

        >>> list(parsedag(b""))
        []

    A simple linear run:

        >>> list(parsedag(b"+3"))
        [('n', (0, [-1])), ('n', (1, [0])), ('n', (2, [1]))]

    Some non-standard ways to define such runs:

        >>> list(parsedag(b"+1+2"))
        [('n', (0, [-1])), ('n', (1, [0])), ('n', (2, [1]))]

        >>> list(parsedag(b"+1*1*"))
        [('n', (0, [-1])), ('n', (1, [0])), ('n', (2, [1]))]

        >>> list(parsedag(b"*"))
        [('n', (0, [-1]))]

        >>> list(parsedag(b"..."))
        [('n', (0, [-1])), ('n', (1, [0])), ('n', (2, [1]))]

    A fork and a join, using numeric back references:

        >>> list(parsedag(b"+2*2*/2"))
        [('n', (0, [-1])), ('n', (1, [0])), ('n', (2, [0])), ('n', (3, [2, 1]))]

        >>> list(parsedag(b"+2<2+1/2"))
        [('n', (0, [-1])), ('n', (1, [0])), ('n', (2, [0])), ('n', (3, [2, 1]))]

    Placing a label:

        >>> list(parsedag(b"+1 :mylabel +1"))
        [('n', (0, [-1])), ('l', (0, 'mylabel')), ('n', (1, [0]))]

    An empty label (silly, really):

        >>> list(parsedag(b"+1:+1"))
        [('n', (0, [-1])), ('l', (0, '')), ('n', (1, [0]))]

    Fork and join, but with labels instead of numeric back references:

        >>> list(parsedag(b"+1:f +1:p2 *f */p2"))
        [('n', (0, [-1])), ('l', (0, 'f')), ('n', (1, [0])), ('l', (1, 'p2')),
         ('n', (2, [0])), ('n', (3, [2, 1]))]

        >>> list(parsedag(b"+1:f +1:p2 <f +1 /p2"))
        [('n', (0, [-1])), ('l', (0, 'f')), ('n', (1, [0])), ('l', (1, 'p2')),
         ('n', (2, [0])), ('n', (3, [2, 1]))]

    Restarting from the root:

        >>> list(parsedag(b"+1 $ +1"))
        [('n', (0, [-1])), ('n', (1, [-1]))]

    Annotations, which are meant to introduce sticky state for subsequent nodes:

        >>> list(parsedag(b"+1 @ann +1"))
        [('n', (0, [-1])), ('a', 'ann'), ('n', (1, [0]))]

        >>> list(parsedag(b'+1 @"my annotation" +1'))
        [('n', (0, [-1])), ('a', 'my annotation'), ('n', (1, [0]))]

    Commands, which are meant to operate on the most recently created node:

        >>> list(parsedag(b"+1 !cmd +1"))
        [('n', (0, [-1])), ('c', 'cmd'), ('n', (1, [0]))]

        >>> list(parsedag(b'+1 !"my command" +1'))
        [('n', (0, [-1])), ('c', 'my command'), ('n', (1, [0]))]

        >>> list(parsedag(b'+1 !!my command line\\n +1'))
        [('n', (0, [-1])), ('C', 'my command line'), ('n', (1, [0]))]

    Comments, which extend to the end of the line:

        >>> list(parsedag(b'+1 # comment\\n+1'))
        [('n', (0, [-1])), ('n', (1, [0]))]

    Error:

        >>> try: list(parsedag(b'+1 bad'))
        ... except Exception as e: print(pycompat.sysstr(bytes(e)))
        invalid character in dag description: bad...

    '''
    if not desc:
        return

    wordchars = pycompat.bytestr(string.ascii_letters + string.digits)

    labels = {}
    p1 = -1
    r = 0

    def resolve(ref):
        if not ref:
            return p1
        elif ref[0] in pycompat.bytestr(string.digits):
            return r - int(ref)
        else:
            return labels[ref]

    chiter = pycompat.iterbytestr(desc)

    def nextch():
        return next(chiter, '\0')

    def nextrun(c, allow):
        s = ''
        while c in allow:
            s += c
            c = nextch()
        return c, s

    def nextdelimited(c, limit, escape):
        s = ''
        while c != limit:
            if c == escape:
                c = nextch()
            s += c
            c = nextch()
        return nextch(), s

    def nextstring(c):
        if c == '"':
            return nextdelimited(nextch(), '"', '\\')
        else:
            return nextrun(c, wordchars)

    c = nextch()
    while c != '\0':
        while c in pycompat.bytestr(string.whitespace):
            c = nextch()
        if c == '.':
            yield 'n', (r, [p1])
            p1 = r
            r += 1
            c = nextch()
        elif c == '+':
            c, digs = nextrun(nextch(), pycompat.bytestr(string.digits))
            n = int(digs)
            for i in pycompat.xrange(0, n):
                yield 'n', (r, [p1])
                p1 = r
                r += 1
        elif c in '*/':
            if c == '*':
                c = nextch()
            c, pref = nextstring(c)
            prefs = [pref]
            while c == '/':
                c, pref = nextstring(nextch())
                prefs.append(pref)
            ps = [resolve(ref) for ref in prefs]
            yield 'n', (r, ps)
            p1 = r
            r += 1
        elif c == '<':
            c, ref = nextstring(nextch())
            p1 = resolve(ref)
        elif c == ':':
            c, name = nextstring(nextch())
            labels[name] = p1
            yield 'l', (p1, name)
        elif c == '@':
            c, text = nextstring(nextch())
            yield 'a', text
        elif c == '!':
            c = nextch()
            if c == '!':
                cmd = ''
                c = nextch()
                while c not in '\n\r\0':
                    cmd += c
                    c = nextch()
                yield 'C', cmd
            else:
                c, cmd = nextstring(c)
                yield 'c', cmd
        elif c == '#':
            while c not in '\n\r\0':
                c = nextch()
        elif c == '$':
            p1 = -1
            c = nextch()
        elif c == '\0':
            return # in case it was preceded by whitespace
        else:
            s = ''
            i = 0
            while c != '\0' and i < 10:
                s += c
                i += 1
                c = nextch()
            raise error.Abort(_('invalid character in dag description: '
                               '%s...') % s)

def dagtextlines(events,
                 addspaces=True,
                 wraplabels=False,
                 wrapannotations=False,
                 wrapcommands=False,
                 wrapnonlinear=False,
                 usedots=False,
                 maxlinewidth=70):
    '''generates single lines for dagtext()'''

    def wrapstring(text):
        if re.match("^[0-9a-z]*$", text):
            return text
        return '"' + text.replace('\\', '\\\\').replace('"', '\"') + '"'

    def gen():
        labels = {}
        run = 0
        wantr = 0
        needroot = False
        for kind, data in events:
            if kind == 'n':
                r, ps = data

                # sanity check
                if r != wantr:
                    raise error.Abort(_("expected id %i, got %i") % (wantr, r))
                if not ps:
                    ps = [-1]
                else:
                    for p in ps:
                        if p >= r:
                            raise error.Abort(_("parent id %i is larger than "
                                               "current id %i") % (p, r))
                wantr += 1

                # new root?
                p1 = r - 1
                if len(ps) == 1 and ps[0] == -1:
                    if needroot:
                        if run:
                            yield '+%d' % run
                            run = 0
                        if wrapnonlinear:
                            yield '\n'
                        yield '$'
                        p1 = -1
                    else:
                        needroot = True
                if len(ps) == 1 and ps[0] == p1:
                    if usedots:
                        yield "."
                    else:
                        run += 1
                else:
                    if run:
                        yield '+%d' % run
                        run = 0
                    if wrapnonlinear:
                        yield '\n'
                    prefs = []
                    for p in ps:
                        if p == p1:
                            prefs.append('')
                        elif p in labels:
                            prefs.append(labels[p])
                        else:
                            prefs.append('%d' % (r - p))
                    yield '*' + '/'.join(prefs)
            else:
                if run:
                    yield '+%d' % run
                    run = 0
                if kind == 'l':
                    rid, name = data
                    labels[rid] = name
                    yield ':' + name
                    if wraplabels:
                        yield '\n'
                elif kind == 'c':
                    yield '!' + wrapstring(data)
                    if wrapcommands:
                        yield '\n'
                elif kind == 'C':
                    yield '!!' + data
                    yield '\n'
                elif kind == 'a':
                    if wrapannotations:
                        yield '\n'
                    yield '@' + wrapstring(data)
                elif kind == '#':
                    yield '#' + data
                    yield '\n'
                else:
                    raise error.Abort(_("invalid event type in dag: "
                                        "('%s', '%s')")
                                      % (stringutil.escapestr(kind),
                                         stringutil.escapestr(data)))
        if run:
            yield '+%d' % run

    line = ''
    for part in gen():
        if part == '\n':
            if line:
                yield line
                line = ''
        else:
            if len(line) + len(part) >= maxlinewidth:
                yield line
                line = ''
            elif addspaces and line and part != '.':
                line += ' '
            line += part
    if line:
        yield line

def dagtext(dag,
            addspaces=True,
            wraplabels=False,
            wrapannotations=False,
            wrapcommands=False,
            wrapnonlinear=False,
            usedots=False,
            maxlinewidth=70):
    '''generates lines of a textual representation for a dag event stream

    events should generate what parsedag() does, so:

      ('n', (id, [parentids])) for node creation
      ('l', (id, labelname)) for labels on nodes
      ('a', text) for annotations
      ('c', text) for commands
      ('C', text) for line commands ('!!')
      ('#', text) for comment lines

    Parent nodes must come before child nodes.

    Examples
    --------

    Linear run:

        >>> dagtext([(b'n', (0, [-1])), (b'n', (1, [0]))])
        '+2'

    Two roots:

        >>> dagtext([(b'n', (0, [-1])), (b'n', (1, [-1]))])
        '+1 $ +1'

    Fork and join:

        >>> dagtext([(b'n', (0, [-1])), (b'n', (1, [0])), (b'n', (2, [0])),
        ...          (b'n', (3, [2, 1]))])
        '+2 *2 */2'

    Fork and join with labels:

        >>> dagtext([(b'n', (0, [-1])), (b'l', (0, b'f')), (b'n', (1, [0])),
        ...          (b'l', (1, b'p2')), (b'n', (2, [0])), (b'n', (3, [2, 1]))])
        '+1 :f +1 :p2 *f */p2'

    Annotations:

        >>> dagtext([(b'n', (0, [-1])), (b'a', b'ann'), (b'n', (1, [0]))])
        '+1 @ann +1'

        >>> dagtext([(b'n', (0, [-1])),
        ...          (b'a', b'my annotation'),
        ...          (b'n', (1, [0]))])
        '+1 @"my annotation" +1'

    Commands:

        >>> dagtext([(b'n', (0, [-1])), (b'c', b'cmd'), (b'n', (1, [0]))])
        '+1 !cmd +1'

        >>> dagtext([(b'n', (0, [-1])),
        ...          (b'c', b'my command'),
        ...          (b'n', (1, [0]))])
        '+1 !"my command" +1'

        >>> dagtext([(b'n', (0, [-1])),
        ...          (b'C', b'my command line'),
        ...          (b'n', (1, [0]))])
        '+1 !!my command line\\n+1'

    Comments:

        >>> dagtext([(b'n', (0, [-1])), (b'#', b' comment'), (b'n', (1, [0]))])
        '+1 # comment\\n+1'

        >>> dagtext([])
        ''

    Combining parsedag and dagtext:

        >>> dagtext(parsedag(b'+1 :f +1 :p2 *f */p2'))
        '+1 :f +1 :p2 *f */p2'

    '''
    return "\n".join(dagtextlines(dag,
                                  addspaces,
                                  wraplabels,
                                  wrapannotations,
                                  wrapcommands,
                                  wrapnonlinear,
                                  usedots,
                                  maxlinewidth))