Re: Why separate Lexical & Parser Generators

morrison@hal.cs.uiuc.edu (Vance Morrison) writes:
>Another problem is dealing with comments because they can occur ANYWHERE.
>This can be solved by simply restricting where comments can go (which may
>not be a bad idea in any case).


hagerman@ece.cmu.edu (John Hagerman) writes:
>I disagree, for what is a gross but realistic reason. When comments can
>go anywhere it is really easy to extend the language. Obviously, such
>extension should not occur in any production environment (smiley goes here
>-- c.f. Pascal 6000). But when doing *research* it is nice to (1) have a
>well-established language with a real compiler etc, and (2) still be able
>to extend the language. Pragmas are not sufficient, since the language
>designer can't predict what kind of extensions *I* may want. If it
>weren't for free-ranging comments in the language I am currently working
>with, my research would have been *much* harder.


You ignore reflexive languages like Forth and CLOS. Unless you can
predict every use to which your language will be put you should
always allow its extension. Indeed you can hardly prevent it.
The user can always modify your syntax with a preprocessor and
extend your semantics by calling assembler subroutines. You might
as well cooperate with users instead of fighting them.




--

Return to the comp.compilers page.
Search the comp.compilers archives again.