On Mar 15, 2:14 pm, Jonah Thomas <jethom...@[EMAIL PROTECTED]
> wrote:
> I have the idea that locals would not improve this much. I would still
> have to distribute copies of all needed values to each factor. It would
> be a little easier because I wouldn't have to jockey them in place, I
> could just name them in the preferred order.
So what's the problem with using localized global variables? If,
indeed, the only problem with using localized global variables is the
tedium of writing out all the "DUP @[EMAIL PROTECTED]
>R !" and "R> !" at the end, that
is definitely something that could be automated, with a kind of:
LOCALIZE{ ... }
format where a stack frame is built with each variable's original
contents under its address in the R-stack, and a count of how many
localized globals on top of that. You'd need to redefine ``:'' to do a
``0 >R'' and ``;'' and ``EXIT'' to check the size of the stack frame
and restore the localized globals if any, but there'd be no problem at
all with factoring that code.
And no special locals namespace, just the ordinary dictionary.
|
