Subj : Re: File Syntax checking: need logical help
To   : comp.programming
From : Rob Thorpe
Date : Mon Aug 01 2005 12:45 pm

Abbacabba wrote:
> I'm trying to create an efficient and manageable way to define a file
> format and check it during processing.  Our file formats are both fixed
> record length, and variable record/delimited.
>
> I would like to be able to define a format, and have the program
> generate the applicable rules.
>
> Given a generic data structure definition: (example only)
>
> A - manditory - A structure ends with H record
> B - manditory - B structure ends with G Record
> C - manditory - C Structure ends with D Record
> C2 - Optional
> C3 - Optional
> D - Optional
> E - Manditory
> E1 - conditional - (E "packet" must have an E1, E1/E2, E2 or E3)
> E2 - conditional - (E "packet" must have an E1, E1/E2, E2 or E3)
> E3 - conditional - (E "packet" must have an E1, E1/E2, E2 or E3)
> F - manditory
> G - Manditory
> H - Manditory
>
> What type of logic structure would be the best to use?
>
> I've looked at State machines and trees, but don't feel either(or my
> implementation of either) comes across as flexable or efficient.
>
>
> Example of logic:
>
> If the program has just finished with an "A" record the only possible
> correct record type to process would be a "B".
>
> If the program has just finshed with a "C1" record, possible records
> would be both a "C" or a "D" record.
>
>
> I'm just needing some general ideas for a direction to pursue, not an
> actual solution(although that would be most helpful).

If your creating a file format that general it's probably worth parsing
it formally, as though it were a computer language.

That is, write a formal grammar for the language.  Then either:
- enter it into bison or yacc
- build a recursive-descent parser to parse it.
(see compiler textbooks for ways to do this, it's quite simple)

It probably worth splitting it into two tasks:-
1. Understanding the file format and entering it into data structures
to use (parsing)
2. Checking if a valid pattern of data structures is present at the
end. (semantic checking)

.