🚀 go-pugleaf

In article <8903250115.AA08564@yorkville.csri.toronto.edu>, norvell@csri.toronto.edu (Theodore Stevens Norvell) writes:
> The real question is what do you want to call syntax and what do you
> want to call static semantics.  In the absence of any definitive definition,
> this is a matter of taste and convenience to the implementor.

For a new approach to this, look at:

"The Revised Report on the Algorithmic Language Algol 68".
- van Wijngaarden et al

This approached this problem by defining the grammar using another (context-free)
grammar (hence the term 'two-level grammar'). This meant that things like not using
a variable before it was defined were specified in the *grammar* - very nifty.
To do this, the second level grammar had a non-terminal called NEST which expanded
to sequences like "int a char b" and so on. Productions of the first level grammar
had NESTs scattered in them (translating to C):

    TYPE NEST variable : TYPE NEST IDENTIFIER, where TYPE IDENTIFIER declared in NEST .

(comma means "and"). The production

    where TYPE IDENTIFIER declared in NEST

expands to "empty" if the variable was declared, and to a nonterminal with no
production rule if it isn't - thus the grammar can't generate a program with an
undeclared variable. (Algol 68 had user-definable types, like C).

They did all kinds of things with this - type checking of assignments and operator
arguments, and context sensitive automatic type conversion come to mind.

Pity it was so hard to parse :-)
-- 
Clive D.W. Feather           clive@ixi.uucp
IXI Limited                  ...!mcvax!ukc!acorn!ixi!clive (untested)
                             +44 223 462 131