Re: parsing bigger files

[Hans Aberg]

On 27 Oct 2006, at 17:59, Joerg Hoffmann wrote:

> how do I get my bison/flex parser to parse bigger files? Is there  
> some way to control maximal sizes, e.g. the maximal number of items  
> parsed into a list, or so?
>
> I'm using Bison 1.875 and flex 2.5.4. with newer versions of bison,  
> my parser does not compile. The file I want  to parse is not   
> exaggeratedly long, only just over 1000 lines. A few years ago, I  
> did manage to configure the parser and/or the bison, on some other  
> computer, to parse much bigger files in the order of many MBytes.  
> So I know this is possible in principle. I have compared the two  
> parsers (their .y source code) and cannot see a discernible  
> difference that could account for the ability to parse larger  
> files. Does the particular Linux version used have any influence on  
> this?
>
> I don't know if this is of any help, but I attach the two .y files  
> just in case. Note that the two parsers differ in various details  
> of what exactly is parsed, so I can't try out what happens if I  
> replace them,

I compiled the first of them in Bison 2.3a (using --verbose to get  
the '.output' file), and just got a shift/reduce conflict in state  
116. You may try to remove it by setting token precedences (%left  
etc.) on the token before and after the '-' in the conflicting rules,  
i.e., NOT_TOK, CLOSE_PAREN.

state 116

   56 literal_term: OPEN_PAREN NOT_TOK atomic_formula_term .  
CLOSE_PAREN
   57             | atomic_formula_term .

    CLOSE_PAREN  shift, and go to state 136

    CLOSE_PAREN  [reduce using rule 57 (literal_term)]

  Hans Aberg