Re: The dynamic stack

[Andy Wingo <wingo@pobox.com>]

Hi Noah,

On Wed 07 Mar 2012 07:05, Noah Lavine <noah.b.lavine@gmail.com> writes:

>>> The “dynwind stack” actually (I misread it the first time.)
>>
>> Yes, it did have this name before.  (More often, "the wind list".)  But
>> since "dynwind" is overloaded so much (dynamic-wind operator, <dynwind>,
>> scm_dynwind_*), and the dynamic stack can have other things on it like
>> prompts, I thought it best to give it a new name.
>
> I see your point about the old name, but I find the new name
> confusing. What about the stack that holds regular program variables
> and temporaries?  That is also dynamic, so at first I thought you were
> referring to that. I didn't know what was going on until this email.

It's a good question!

The essential difference is that lookup on the VM stack is static --
things are always at known locations.  That's what lexical scoping gives
us.  Lookup on the dynamic stack is dynamic: e.g. "what's the nearest
prompt with tag FOO", "what are the dynamic-wind expressions within that
extent", etc.

In an ideal world, we'd probably use the same stack for both purposes,
as C++ does (AFAIU).  The reason that we don't is that the scm_dynwind_*
API manipulates the dynwind stack from outside of the VM.

Cheers,

Andy
-- 
http://wingolog.org/