[Self-interest] Block Question

David Ungar ungar at mac.com
Thu Nov 11 02:53:14 UTC 2021


Thank you, Jecel, for explaining it so well.

- David

> On Nov 10, 2021, at 5:11 AM, Jecel Assumpcao Jr <jecel at merlintec.com> wrote:
> 
> ´╗┐Luke McNeil wrote on Sat, 6 Nov 2021 18:15:37 -0400
>> I'm trying to figure out why the code "(| x = [1] |) x value" throws this error "cannot run a block after its enclosing method has returned."
>> What object exactly is the block's enclosing method, and when is it created?
>> My best guess would be that when "[1]" is evaluated to be placed in slot x, it is actually evaluated in the context of a new empty method activation object.
>> I realize that this code isn't really useful (it throws an error after all), but I am just trying to better understand what Self is doing.
> 
> If in Self or Smalltalk you are executing message A and send message B
> and then that sends message C, eventually C will return and even later B
> will return and you will be back in A. This is the traditional "stack
> discipline" and is very efficient to implement.
> 
> Blocks complicate this simple picture. If a block is created inside B it
> will need to use stuff B keeps in its stack frame when it runs. That
> will happen when someone sends the "value" message to that block. But
> what if B has already returned to A at that point in time?
> 
> One solution, adopted by nearly all Smalltalks, is to declare that B's
> stack frame will have to stick around. We can hold it in a regular
> object and let the garbage collector take care of eventualy getting rid
> of it. Quite a bit o the complexity of high performance Smalltalk
> virtual machine implementations is in dealing with this problem.
> 
> An alternative, adopted by Self, is to simply declare this to be an
> error - the one you saw. You are allowed to use blocks created in B as
> long as its stack frame is still around. When B returns we just throw
> all its stuff away (pop the stack) and too bad if any blocks needed it.
> Not having these "non LIFO blocks" in Self doesn't hurt like it would in
> other Smalltalks because it is so simple to create new kinds of objects.
> So if we want some data and code fragment to be usable after the method
> that created has finished we simply use an object instead of a block. It
> is just a matter of parenthesis instead of square brackets (not counting
> that blocks have lexical scoping and objects don't).
> 
> Note that even people who made this design decision sometimes regret it.
> It was certainly a key obstacle in Mario's getting GNU Smalltalk code to
> run in Self.
> 
> -- Jecel
> _______________________________________________
> Self-interest mailing list
> Self-interest at lists.selflanguage.org
> http://lists.selflanguage.org/mailman/listinfo/self-interest


More information about the Self-interest mailing list