Re: specifying semantics, was Formatting of Language LRMs

Stefan Monnier schrieb:
>> What is conventionally called "syntax" is an artifact of the compiler
>> technology (or maybe of specification technology, see below):
>
> I disagree here. It is no accident. Same as the split between
> lexical and syntactic analysis, the division between syntax and
> (static) semantics is an engineering issue: specifying a language is a
> fairly large amount of work, so you want to split it into simpler
> parts.


I don't know of any *general* formal description of lexical analysis,
that would allow to e.g. describe Fortran and Algol (or C) lexing at the
same time. In an experiment with an scannerless parser a problem with
whitespace popped up, where whitespace sometimes is required to separate
keywords from identifiers, sometimes is optional:


Fortran: D O I = ...
Pascal: if x ...
C: int i ...


Just in the Fortran example neither a longest nor a shortest match lexer
is a solution. Do we have to exclude Fortran (and C) as context
sensitive languages then? But what would be the benefit of an
description formalism, that only is applicable to new languages, whose
syntax and semantics can (must!) be described using just that formalims?


In the case of the scannerless parser experiment the grammar size
exploded, when the handling of whitespace was added. Also the well known
"dangling else" problem can be solved only at the expense of a bigger
grammar, which then may still be readable by a compiler generator, but
not by humans any more.




Such problems suggest to me that the design of a language starts with
the design of an appropriate description formalism. But which formalism
could be used to describe *that* formalism?


DoDi


[You can scan Fortran with longest match, but the things you're
matching for are very context dependent. I've done it. I agree that
it's unlikely we'll find a general model for different languages. -John]

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