[parenscript-devel] Lexical scoping question

[Vladimir Sedach]

> If this were really lexically scoped, the a in the alert form would be
> undefined. But ps generates:
>
> "var a = 123;
> blah(a);
> alert(a);"
>
> Does this reflect what is intended? If so, is there a clear explanation of
> just what "lexical scoping" is now offered by PS, i.e. what the rules are,
> what we should expect, and what we should not expect?

Currently PS implements lexical scoping by taking advantage of JS's
lexical scoping. Since JS does not have a special form to introduce
arbitrary lexical scopes (aside from "wrap everything in a lambda"),
this is accomplished by "pushing up" the actual variable definition to
the nearest enclosing JS lexical scope. This may either be the nearest
enclosing function definition (defun or lambda), or if there is none,
the JS toplevel. In the case of that example it happens to be the
toplevel. That's where name collisions will probably become a problem
and the mechanism breaks down. Which is not a terrible thing as it's
basically doing what PS used to do before, and that worked ok.

When the code is inside a function, you get a real lexical scope when
you declare a variable. The variable renaming introduced recently is
an implementation technique for doing arbitrary nested lexical scopes
using a single flat scope, but it depends on that scope being there.

So basically any form that introduces a lexical binding inside a
defun/lambda should correspond exactly to CL semantics (if not, it's a
bug!). Anything that's in the toplevel will either introduce a new
top-level lexical variable, or change the value of an existing one.

I guess one thing that can be done is to wrap any let forms in the
toplevel into a lambda that's called right then and there. I
personally don't like the toplevel/non-toplevel dimorphism in the
generated code so I'm not going to do it unless a compelling reason is
found.

Vladimir

> Daniel
>
> _______________________________________________
> parenscript-devel mailing list
> parenscript-devel at common-lisp.net
> http://common-lisp.net/cgi-bin/mailman/listinfo/parenscript-devel
>
>