view mercurial/parser.py @ 26379:39d643252b9f

revlog: use existing file handle when reading during _addrevision

_addrevision() may need to read from revlogs as part of computing
deltas. Previously, we would flush existing file handles and open
a new, short-lived file handle to perform the reading.

If we have an existing file handle, it seems logical to reuse it
for reading instead of opening a new file handle. This patch
makes that the new behavior.

After this patch, revlog files are only reopened when adding
revisions if the revlog is switched from inline to non-inline.

On Linux when unbundling a bundle of the mozilla-central repo, this
patch has the following impact on system call counts:

Call     Before     After       Delta
write    827,639    673,390   -154,249
open     700,103    684,089    -16,014
read      74,489     74,489          0
fstat    493,924    461,896    -32,028
close    249,131    233,117    -16,014
stat     242,001    242,001          0
lstat     18,676     18,676          0
lseek     20,268     20,268          0
ioctl     14,652     13,173     -1,479
TOTAL  3,180,758  2,930,679   -250,079

It's worth noting that many of the open() calls fail due to missing
files. That's why there are many more open() calls than close().

Despite the significant system call reduction, this change does not
seem to have a significant performance impact on Linux.

On Windows 10 (not a VM, on a SSD), this patch appears to reduce
unbundle time for mozilla-central from ~960s to ~920s. This isn't
as significant as I was hoping. But a decrease it is nonetheless.
Still, Windows unbundle performance is still >2x slower than Linux.

Despite the lack of significant gains, fewer system calls is fewer
system calls. If nothing else, this will narrow the focus of potential
areas to optimize in the future.

# parser.py - simple top-down operator precedence parser for mercurial
#
# Copyright 2010 Matt Mackall <[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.

# see http://effbot.org/zone/simple-top-down-parsing.htm and
# http://eli.thegreenplace.net/2010/01/02/top-down-operator-precedence-parsing/
# for background

# takes a tokenizer and elements
# tokenizer is an iterator that returns (type, value, pos) tuples
# elements is a mapping of types to binding strength, primary, prefix, infix
# and suffix actions
# an action is a tree node name, a tree label, and an optional match
# __call__(program) parses program into a labeled tree

from __future__ import absolute_import

from .i18n import _
from . import error

class parser(object):
    def __init__(self, elements, methods=None):
        self._elements = elements
        self._methods = methods
        self.current = None
    def _advance(self):
        'advance the tokenizer'
        t = self.current
        self.current = next(self._iter, None)
        return t
    def _hasnewterm(self):
        'True if next token may start new term'
        return any(self._elements[self.current[0]][1:3])
    def _match(self, m):
        'make sure the tokenizer matches an end condition'
        if self.current[0] != m:
            raise error.ParseError(_("unexpected token: %s") % self.current[0],
                                   self.current[2])
        self._advance()
    def _parseoperand(self, bind, m=None):
        'gather right-hand-side operand until an end condition or binding met'
        if m and self.current[0] == m:
            expr = None
        else:
            expr = self._parse(bind)
        if m:
            self._match(m)
        return expr
    def _parse(self, bind=0):
        token, value, pos = self._advance()
        # handle prefix rules on current token, take as primary if unambiguous
        primary, prefix = self._elements[token][1:3]
        if primary and not (prefix and self._hasnewterm()):
            expr = (primary, value)
        elif prefix:
            expr = (prefix[0], self._parseoperand(*prefix[1:]))
        else:
            raise error.ParseError(_("not a prefix: %s") % token, pos)
        # gather tokens until we meet a lower binding strength
        while bind < self._elements[self.current[0]][0]:
            token, value, pos = self._advance()
            # handle infix rules, take as suffix if unambiguous
            infix, suffix = self._elements[token][3:]
            if suffix and not (infix and self._hasnewterm()):
                expr = (suffix[0], expr)
            elif infix:
                expr = (infix[0], expr, self._parseoperand(*infix[1:]))
            else:
                raise error.ParseError(_("not an infix: %s") % token, pos)
        return expr
    def parse(self, tokeniter):
        'generate a parse tree from tokens'
        self._iter = tokeniter
        self._advance()
        res = self._parse()
        token, value, pos = self.current
        return res, pos
    def eval(self, tree):
        'recursively evaluate a parse tree using node methods'
        if not isinstance(tree, tuple):
            return tree
        return self._methods[tree[0]](*[self.eval(t) for t in tree[1:]])
    def __call__(self, tokeniter):
        'parse tokens into a parse tree and evaluate if methods given'
        t = self.parse(tokeniter)
        if self._methods:
            return self.eval(t)
        return t

def buildargsdict(trees, funcname, keys, keyvaluenode, keynode):
    """Build dict from list containing positional and keyword arguments

    Invalid keywords or too many positional arguments are rejected, but
    missing arguments are just omitted.
    """
    if len(trees) > len(keys):
        raise error.ParseError(_("%(func)s takes at most %(nargs)d arguments")
                               % {'func': funcname, 'nargs': len(keys)})
    args = {}
    # consume positional arguments
    for k, x in zip(keys, trees):
        if x[0] == keyvaluenode:
            break
        args[k] = x
    # remainder should be keyword arguments
    for x in trees[len(args):]:
        if x[0] != keyvaluenode or x[1][0] != keynode:
            raise error.ParseError(_("%(func)s got an invalid argument")
                                   % {'func': funcname})
        k = x[1][1]
        if k not in keys:
            raise error.ParseError(_("%(func)s got an unexpected keyword "
                                     "argument '%(key)s'")
                                   % {'func': funcname, 'key': k})
        if k in args:
            raise error.ParseError(_("%(func)s got multiple values for keyword "
                                     "argument '%(key)s'")
                                   % {'func': funcname, 'key': k})
        args[k] = x[2]
    return args

def unescapestr(s):
    try:
        return s.decode("string_escape")
    except ValueError as e:
        # mangle Python's exception into our format
        raise error.ParseError(str(e).lower())

def _prettyformat(tree, leafnodes, level, lines):
    if not isinstance(tree, tuple) or tree[0] in leafnodes:
        lines.append((level, str(tree)))
    else:
        lines.append((level, '(%s' % tree[0]))
        for s in tree[1:]:
            _prettyformat(s, leafnodes, level + 1, lines)
        lines[-1:] = [(lines[-1][0], lines[-1][1] + ')')]

def prettyformat(tree, leafnodes):
    lines = []
    _prettyformat(tree, leafnodes, 0, lines)
    output = '\n'.join(('  ' * l + s) for l, s in lines)
    return output

def simplifyinfixops(tree, targetnodes):
    """Flatten chained infix operations to reduce usage of Python stack

    >>> def f(tree):
    ...     print prettyformat(simplifyinfixops(tree, ('or',)), ('symbol',))
    >>> f(('or',
    ...     ('or',
    ...       ('symbol', '1'),
    ...       ('symbol', '2')),
    ...     ('symbol', '3')))
    (or
      ('symbol', '1')
      ('symbol', '2')
      ('symbol', '3'))
    >>> f(('func',
    ...     ('symbol', 'p1'),
    ...     ('or',
    ...       ('or',
    ...         ('func',
    ...           ('symbol', 'sort'),
    ...           ('list',
    ...             ('or',
    ...               ('or',
    ...                 ('symbol', '1'),
    ...                 ('symbol', '2')),
    ...               ('symbol', '3')),
    ...             ('negate',
    ...               ('symbol', 'rev')))),
    ...         ('and',
    ...           ('symbol', '4'),
    ...           ('group',
    ...             ('or',
    ...               ('or',
    ...                 ('symbol', '5'),
    ...                 ('symbol', '6')),
    ...               ('symbol', '7'))))),
    ...       ('symbol', '8'))))
    (func
      ('symbol', 'p1')
      (or
        (func
          ('symbol', 'sort')
          (list
            (or
              ('symbol', '1')
              ('symbol', '2')
              ('symbol', '3'))
            (negate
              ('symbol', 'rev'))))
        (and
          ('symbol', '4')
          (group
            (or
              ('symbol', '5')
              ('symbol', '6')
              ('symbol', '7'))))
        ('symbol', '8')))
    """
    if not isinstance(tree, tuple):
        return tree
    op = tree[0]
    if op not in targetnodes:
        return (op,) + tuple(simplifyinfixops(x, targetnodes) for x in tree[1:])

    # walk down left nodes taking each right node. no recursion to left nodes
    # because infix operators are left-associative, i.e. left tree is deep.
    # e.g. '1 + 2 + 3' -> (+ (+ 1 2) 3) -> (+ 1 2 3)
    simplified = []
    x = tree
    while x[0] == op:
        l, r = x[1:]
        simplified.append(simplifyinfixops(r, targetnodes))
        x = l
    simplified.append(simplifyinfixops(x, targetnodes))
    simplified.append(op)
    return tuple(reversed(simplified))