On 7/21/19 1:49 PM, John Cowan wrote:
> Racket keywords have the form #:foo and are part of the syntax of function calls rather than a self-evaluating literal.

In newer versions of Kawa too keywords are syntax rather than self-evaluating literals.
You can quote a keyword, but in that case it just evaluates to a Keyword value, bu
a Keyword value in a argument list is *not* a keyword parameter.

See:
https://www.gnu.org/software/kawa/Primitive-expression-syntax.html#Procedure-calls
https://www.gnu.org/software/kawa/Application-and-Arguments-Lists.html

The Common Lisp keywords-are-self-evaluating model has the advantage of simplicity, but
the Kawa/Racket model has some advantages, mtoo, ost importantly static analysis
is easier:
* When calling a known procedure the compiler can statically determine if
a keyword actual argument is valid. (Not yet implemented in Kawa.)
* Easier to generate efficient code.

The Kawa calling convention involves sorting the keywords at both
the caller and callee sites.  This can be done at compile-time: For the callee,
assuming the callee doesn't allow "other keywords"; for the caller, assuming
argument-list objects aren't used.  In that case argument matching is very fast,
without requiring any form of hash table: A simple linear scan of the
caller and callee keyword lists.  In addition to being fast, it requires
less space than a hash table.
--
	--Per Bothner
xxxxxx@bothner.com   http://per.bothner.com/