Re: [Pyparsing] parsing with operatorPrecedence takes too much time

[Ralf S. <sc...@gm...>]

On Thu, Apr 3, 2008 at 12:15 AM, Paul McGuire <pt...@au...> wrote:

> This performance problem has come up a couple of times in the past month.
> For a start, you could try enabling packrat parsing - change your import
> and
> following lines to:
>
>
> from pyparsing import (Word, Literal, CaselessLiteral,
>                       Combine, Optional, nums, StringEnd,
> operatorPrecedence,
>                        opAssoc,ParserElement)
> ParserElement.enablePackrat()
>
>
> This will speed up your parser about 500X (not kidding!).
>


Paul,

operatorPrecedence isn't working for me. I think something must be wrong,
it's just much too slow.
I've turned on packrat parsing. The current code can be viewed here:
http://code.pediapress.com/hg/mwlib/file/ce563c294ad1/mwlib/e.py
or downloaded here:
http://code.pediapress.com/hg/mwlib/raw-file/ce563c294ad1/mwlib/e.py

It reads expressions from stdin.

Parsing something like:
5-(((63+459.67)/1.8)<100000)-(((63+459.67)/1.8)<10000)-(((63+459.67)/1.8)<1000)-(((63+459.67)/1.8)<100)-(((63+459.67)/1.8)<10)

takes nearly 0.09 seconds. (i.e. one can only parse around 11 such
expressions, which is unbeliavably slow for a 2.4 Ghz processor).

The string representation of the expr parser element is huge:
In [2]: len(str(e.expr))
Out[2]: 4317534

For my old parser this is length is 22076.

What's going on?

- Ralf