🚀 go-pugleaf

>>>>> On Tue, 21 Jul 1992 22:09:35 GMT, drraymon@watdragon.waterloo.edu (Darrell Raymond) said:

>>The BNF used to define Algol and C, and the railroad track diagrams used
>>to define Pascal (and presumably Modula 2) cannot handle the
>>context-sensitive construct DR cited (viz. X**n Y**n) any more than SGML
>>can, but the grammars are still useful for those working with those
>>languages.

>   This example doesn't work for several reasons.  First, the BNF of C 
> doesn't define C - that requires a lot of semantic information as well 
> (I haven't thought about it enough, but it may be that it even requires 
> Turing equivalence).  Second, the construct "X**n Y**n" is in fact 
> context-free.  Third, the construct "ww", which is the one I actually 
> mentioned, has no special semantics in C, or any other programming 
> language that I am aware of - it just means "do w and do w again".  So 
> recognizing it is not important.  Finally, are you confusing two 
> languages? - namely, the language consisting of the set of strings that 
> represent valid C programs (which is context free) and the language 
> accepted by any given C program (which, in general, is recursively 
> enumerable)?

The set of all strings representing valid C programs can not be
described by a context free grammar, a superset of all valid C programs
can be recognised by a context free grammar. To recognize correct C
programs you will need a semantical analysis as well. 

When using context sensitive grammars you can do some type checking
in the syntactical analysis phase.


>>SGML's most serious problems as a practical tool seem to me
>>not its lack of non-CF power but the number of bells and whistles it has.
>  	.
>  	.
>>Darrell, what on earth are you up to now?

>   Trying to establish formal arguments in favor of eliminating bells
> and whistles.  Question: could markup itself be one of them?

> -Darrell. 
--
-------------------------------------------------------------------------------
Name:	Knut-Hevard Aksnes (ECMA 94) or		Knut-Haavard Aksnes (ASCII)
Ericsson signature: HI/ETN/U/IT KNA		Phone: +47 41 19481
Email: etnkna@etn.ericsson.se (internet)  ETN.ETNKNA (memo)