Re: Continuations

On Sun, 14 Jul 2024 22:22:41 +0300, Niklas Holsti wrote:

I think of continuation passing as always ending a subprogram by calling
some other subprogram, passing it one or more "continuation" subprograms
as parameters, and expecting that callee to do something and then to
"continue" by further calling one of those "continuations" (with further
"continuation" parameters) and NOT to return to the original caller,
ever.

That’s the most general case. But it makes sense to have simpler
conventions for common cases, like subroutine call/return. So a subroutine
call implicitly passes a continuation that, when invoked, resumes the
caller. And the subroutine block can end with an implicit invocation of
this continuation. And such continuation objects are so common that it
makes sense, as an optimization, to have a special place to allocate them
in LIFO fashion--call it a “stack”.

You may have heard some people talk about “tree-like stacks”. This is
starting to generalize from simple linear stacks to fully general
continuations.