view mercurial/filesetlang.py @ 42043:1fac9b931d46

compression: introduce a `storage.revlog.zlib.level` configuration This option control the zlib compression level used when compression revlog chunk. This is also a good excuse to pave the way for a similar configuration option for the zstd compression engine. Having a dedicated option for each compression algorithm is useful because they don't support the same range of values. Using a higher zlib compression impact CPU consumption at compression time, but does not directly affected decompression time. However dealing with small compressed chunk can directly help decompression and indirectly help other revlog logic. I ran some basic test on repositories using different level. I am using the mercurial, pypy, netbeans and mozilla-central clone from our benchmark suite. All tested repository use sparse-revlog and got all their delta recomputed. The different compression level has a small effect on the repository size (about 10% variation in the total range). My quick analysis is that revlog mostly store small delta, that are not affected by the compression level much. So the variation probably mostly comes from better compression of the snapshots revisions, and snapshot revision only represent a small portion of the repository content. I also made some basic timings measurements. The "read" timings are gathered using simple run of `hg perfrevlogrevisions`, the "write" timings using `hg perfrevlogwrite` (restricted to the last 5000 revisions for netbeans and mozilla central). The timings are gathered on a generic machine, (not one of our performance locked machine), so small variation might not be meaningful. However large trend remains relevant. Keep in mind that these numbers are not pure compression/decompression time. They also involve the full revlog logic. In particular the difference in chunk size has an impact on the delta chain structure, affecting performance when writing or reading them. On read/write performance, the compression level has a bigger impact. Counter-intuitively, the higher compression levels improve "write" performance for the large repositories in our tested setting. Maybe because the last 5000 delta chain end up having a very different shape in this specific spot? Or maybe because of a more general trend of better delta chains thanks to the smaller chunk and snapshot. This series does not intend to change the default compression level. However, these result call for a deeper analysis of this performance difference in the future. Full data ========= repo level .hg/store size 00manifest.d read write ---------------------------------------------------------------- mercurial 1 49,402,813 5,963,475 0.170159 53.250304 mercurial 6 47,197,397 5,875,730 0.182820 56.264320 mercurial 9 47,121,596 5,849,781 0.189219 56.293612 pypy 1 370,830,572 28,462,425 2.679217 460.721984 pypy 6 340,112,317 27,648,747 2.768691 467.537158 pypy 9 338,360,736 27,639,003 2.763495 476.589918 netbeans 1 1,281,847,810 165,495,457 122.477027 520.560316 netbeans 6 1,205,284,353 159,161,207 139.876147 715.930400 netbeans 9 1,197,135,671 155,034,586 141.620281 678.297064 mozilla 1 2,775,497,186 298,527,987 147.867662 751.263721 mozilla 6 2,596,856,420 286,597,671 170.572118 987.056093 mozilla 9 2,587,542,494 287,018,264 163.622338 739.803002
author Pierre-Yves David <pierre-yves.david@octobus.net>
date Wed, 27 Mar 2019 18:35:27 +0100
parents e79a69af1593
children 2372284d9457
line wrap: on
line source

# filesetlang.py - parser, tokenizer and utility for file set language
#
# Copyright 2010 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.

from __future__ import absolute_import

from .i18n import _
from . import (
    error,
    parser,
    pycompat,
)

# common weight constants for static optimization
# (see registrar.filesetpredicate for details)
WEIGHT_CHECK_FILENAME = 0.5
WEIGHT_READ_CONTENTS = 30
WEIGHT_STATUS = 10
WEIGHT_STATUS_THOROUGH = 50

elements = {
    # token-type: binding-strength, primary, prefix, infix, suffix
    "(": (20, None, ("group", 1, ")"), ("func", 1, ")"), None),
    ":": (15, None, None, ("kindpat", 15), None),
    "-": (5, None, ("negate", 19), ("minus", 5), None),
    "not": (10, None, ("not", 10), None, None),
    "!": (10, None, ("not", 10), None, None),
    "and": (5, None, None, ("and", 5), None),
    "&": (5, None, None, ("and", 5), None),
    "or": (4, None, None, ("or", 4), None),
    "|": (4, None, None, ("or", 4), None),
    "+": (4, None, None, ("or", 4), None),
    ",": (2, None, None, ("list", 2), None),
    ")": (0, None, None, None, None),
    "symbol": (0, "symbol", None, None, None),
    "string": (0, "string", None, None, None),
    "end": (0, None, None, None, None),
}

keywords = {'and', 'or', 'not'}

symbols = {}

globchars = ".*{}[]?/\\_"

def tokenize(program):
    pos, l = 0, len(program)
    program = pycompat.bytestr(program)
    while pos < l:
        c = program[pos]
        if c.isspace(): # skip inter-token whitespace
            pass
        elif c in "(),-:|&+!": # handle simple operators
            yield (c, None, pos)
        elif (c in '"\'' or c == 'r' and
              program[pos:pos + 2] in ("r'", 'r"')): # handle quoted strings
            if c == 'r':
                pos += 1
                c = program[pos]
                decode = lambda x: x
            else:
                decode = parser.unescapestr
            pos += 1
            s = pos
            while pos < l: # find closing quote
                d = program[pos]
                if d == '\\': # skip over escaped characters
                    pos += 2
                    continue
                if d == c:
                    yield ('string', decode(program[s:pos]), s)
                    break
                pos += 1
            else:
                raise error.ParseError(_("unterminated string"), s)
        elif c.isalnum() or c in globchars or ord(c) > 127:
            # gather up a symbol/keyword
            s = pos
            pos += 1
            while pos < l: # find end of symbol
                d = program[pos]
                if not (d.isalnum() or d in globchars or ord(d) > 127):
                    break
                pos += 1
            sym = program[s:pos]
            if sym in keywords: # operator keywords
                yield (sym, None, s)
            else:
                yield ('symbol', sym, s)
            pos -= 1
        else:
            raise error.ParseError(_("syntax error"), pos)
        pos += 1
    yield ('end', None, pos)

def parse(expr):
    p = parser.parser(elements)
    tree, pos = p.parse(tokenize(expr))
    if pos != len(expr):
        raise error.ParseError(_("invalid token"), pos)
    return parser.simplifyinfixops(tree, {'list', 'or'})

def getsymbol(x):
    if x and x[0] == 'symbol':
        return x[1]
    raise error.ParseError(_('not a symbol'))

def getstring(x, err):
    if x and (x[0] == 'string' or x[0] == 'symbol'):
        return x[1]
    raise error.ParseError(err)

def getkindpat(x, y, allkinds, err):
    kind = getsymbol(x)
    pat = getstring(y, err)
    if kind not in allkinds:
        raise error.ParseError(_("invalid pattern kind: %s") % kind)
    return '%s:%s' % (kind, pat)

def getpattern(x, allkinds, err):
    if x and x[0] == 'kindpat':
        return getkindpat(x[1], x[2], allkinds, err)
    return getstring(x, err)

def getlist(x):
    if not x:
        return []
    if x[0] == 'list':
        return list(x[1:])
    return [x]

def getargs(x, min, max, err):
    l = getlist(x)
    if len(l) < min or len(l) > max:
        raise error.ParseError(err)
    return l

def _analyze(x):
    if x is None:
        return x

    op = x[0]
    if op in {'string', 'symbol'}:
        return x
    if op == 'kindpat':
        getsymbol(x[1])  # kind must be a symbol
        t = _analyze(x[2])
        return (op, x[1], t)
    if op == 'group':
        return _analyze(x[1])
    if op == 'negate':
        raise error.ParseError(_("can't use negate operator in this context"))
    if op == 'not':
        t = _analyze(x[1])
        return (op, t)
    if op == 'and':
        ta = _analyze(x[1])
        tb = _analyze(x[2])
        return (op, ta, tb)
    if op == 'minus':
        return _analyze(('and', x[1], ('not', x[2])))
    if op in {'list', 'or'}:
        ts = tuple(_analyze(y) for y in x[1:])
        return (op,) + ts
    if op == 'func':
        getsymbol(x[1])  # function name must be a symbol
        ta = _analyze(x[2])
        return (op, x[1], ta)
    raise error.ProgrammingError('invalid operator %r' % op)

def _insertstatushints(x):
    """Insert hint nodes where status should be calculated (first path)

    This works in bottom-up way, summing up status names and inserting hint
    nodes at 'and' and 'or' as needed. Thus redundant hint nodes may be left.

    Returns (status-names, new-tree) at the given subtree, where status-names
    is a sum of status names referenced in the given subtree.
    """
    if x is None:
        return (), x

    op = x[0]
    if op in {'string', 'symbol', 'kindpat'}:
        return (), x
    if op == 'not':
        h, t = _insertstatushints(x[1])
        return h, (op, t)
    if op == 'and':
        ha, ta = _insertstatushints(x[1])
        hb, tb = _insertstatushints(x[2])
        hr = ha + hb
        if ha and hb:
            return hr, ('withstatus', (op, ta, tb), ('string', ' '.join(hr)))
        return hr, (op, ta, tb)
    if op == 'or':
        hs, ts = zip(*(_insertstatushints(y) for y in x[1:]))
        hr = sum(hs, ())
        if sum(bool(h) for h in hs) > 1:
            return hr, ('withstatus', (op,) + ts, ('string', ' '.join(hr)))
        return hr, (op,) + ts
    if op == 'list':
        hs, ts = zip(*(_insertstatushints(y) for y in x[1:]))
        return sum(hs, ()), (op,) + ts
    if op == 'func':
        f = getsymbol(x[1])
        # don't propagate 'ha' crossing a function boundary
        ha, ta = _insertstatushints(x[2])
        if getattr(symbols.get(f), '_callstatus', False):
            return (f,), ('withstatus', (op, x[1], ta), ('string', f))
        return (), (op, x[1], ta)
    raise error.ProgrammingError('invalid operator %r' % op)

def _mergestatushints(x, instatus):
    """Remove redundant status hint nodes (second path)

    This is the top-down path to eliminate inner hint nodes.
    """
    if x is None:
        return x

    op = x[0]
    if op == 'withstatus':
        if instatus:
            # drop redundant hint node
            return _mergestatushints(x[1], instatus)
        t = _mergestatushints(x[1], instatus=True)
        return (op, t, x[2])
    if op in {'string', 'symbol', 'kindpat'}:
        return x
    if op == 'not':
        t = _mergestatushints(x[1], instatus)
        return (op, t)
    if op == 'and':
        ta = _mergestatushints(x[1], instatus)
        tb = _mergestatushints(x[2], instatus)
        return (op, ta, tb)
    if op in {'list', 'or'}:
        ts = tuple(_mergestatushints(y, instatus) for y in x[1:])
        return (op,) + ts
    if op == 'func':
        # don't propagate 'instatus' crossing a function boundary
        ta = _mergestatushints(x[2], instatus=False)
        return (op, x[1], ta)
    raise error.ProgrammingError('invalid operator %r' % op)

def analyze(x):
    """Transform raw parsed tree to evaluatable tree which can be fed to
    optimize() or getmatch()

    All pseudo operations should be mapped to real operations or functions
    defined in methods or symbols table respectively.
    """
    t = _analyze(x)
    _h, t = _insertstatushints(t)
    return _mergestatushints(t, instatus=False)

def _optimizeandops(op, ta, tb):
    if tb is not None and tb[0] == 'not':
        return ('minus', ta, tb[1])
    return (op, ta, tb)

def _optimizeunion(xs):
    # collect string patterns so they can be compiled into a single regexp
    ws, ts, ss = [], [], []
    for x in xs:
        w, t = _optimize(x)
        if t is not None and t[0] in {'string', 'symbol', 'kindpat'}:
            ss.append(t)
            continue
        ws.append(w)
        ts.append(t)
    if ss:
        ws.append(WEIGHT_CHECK_FILENAME)
        ts.append(('patterns',) + tuple(ss))
    return ws, ts

def _optimize(x):
    if x is None:
        return 0, x

    op = x[0]
    if op == 'withstatus':
        w, t = _optimize(x[1])
        return w, (op, t, x[2])
    if op in {'string', 'symbol'}:
        return WEIGHT_CHECK_FILENAME, x
    if op == 'kindpat':
        w, t = _optimize(x[2])
        return w, (op, x[1], t)
    if op == 'not':
        w, t = _optimize(x[1])
        return w, (op, t)
    if op == 'and':
        wa, ta = _optimize(x[1])
        wb, tb = _optimize(x[2])
        if wa <= wb:
            return wa, _optimizeandops(op, ta, tb)
        else:
            return wb, _optimizeandops(op, tb, ta)
    if op == 'or':
        ws, ts = _optimizeunion(x[1:])
        if len(ts) == 1:
            return ws[0], ts[0] # 'or' operation is fully optimized out
        ts = tuple(it[1] for it in sorted(enumerate(ts),
                                          key=lambda it: ws[it[0]]))
        return max(ws), (op,) + ts
    if op == 'list':
        ws, ts = zip(*(_optimize(y) for y in x[1:]))
        return sum(ws), (op,) + ts
    if op == 'func':
        f = getsymbol(x[1])
        w = getattr(symbols.get(f), '_weight', 1)
        wa, ta = _optimize(x[2])
        return w + wa, (op, x[1], ta)
    raise error.ProgrammingError('invalid operator %r' % op)

def optimize(x):
    """Reorder/rewrite evaluatable tree for optimization

    All pseudo operations should be transformed beforehand.
    """
    _w, t = _optimize(x)
    return t

def prettyformat(tree):
    return parser.prettyformat(tree, ('string', 'symbol'))