Re: How to implement lexical closures?

Gene <gene.ressler@gmail.com> wrote:
(snip)


> A closure is a pair consisting of code and an activation record. The
> activation record is an array of slots for values of parameters (and
> local variables if your language has them). In lanuages like yours
> with lexical nesting, there will also be a frame pointer to the
> activation record of the lexically enclosing scope.


I never actually tried writing code like that, so I am not
so sure I understand closures. I do know the problem that
comes up in PL/I with ENTRY variables. (Similar to C function
pointers, but PL/I allows internal procedures. In the case
of recursion, for example, you get the activation record at
the time of the call. Because of this complication, Fortran
(as of 2003, and I believe 2008) doesn't allow the use of an
internal procedure name as an actual argument.


But maybe I am still not so sure. As you say, "consisting
of code and an activation record." Is "pointer to code"
good enough for that definition, or is it more than that?


I have never seen that PL/I ENTRY variable refered to as
a closure.


An even more interesting feature in PL/I is the use of
a LABEL variable pointing to a label in a different procedure
or a different activation level in the same procedure.


-- glen

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