[DGD] DGDMP 1.1.10

[Shentino]

On Tue, Jun 9, 2009 at 3:52 PM, Felix A. Croes <felix at dworkin.nl> wrote:

> Shentino <shentino at gmail.com> wrote:
>
> > Is there any way to collapse redundancy out of this?  Perhaps with
> > indirection or with a tree instead of an array?  I bet there's a
> gazillion
> > copies of AUTO and/or SECOND_AUTO floating around in that array.
>
> There is only one reference to the auto object.  That particular
> optimization was too good to leave out.


I was also talking about the "second auto" that the klib allows.  Since it's
an LPC abstraction and all that the low level driver would likely not be
aware of.

If you replace the simple array with a more complex structure, you need
> a coordinate system for it, and each program lookup will either involve
> walking a graph or using a hash table.  That's a lot of speed you're
> giving up, probably a factor 3 or 4 in the interpreter's main loop.
>
> The best I came up with uses an extra level of indirection, with a
> moderate slowdown.  I may use it in DGDMP in the future; I'm not planning
> any more significant innovations for DGD at this point.
>

Not trying to irritate anyone, I just wanted to emphasize that my case
wasn't a random pathological case.  I only simplified it for the purpose of
illustration.  My case, as I now recall, was three layers of inheritance
mesh.  One for the variables, one for the function defs, and a third for the
function decs.  Except for the inheritance overflow error, it was about 18
files of pure neat-freak heaven.

And speaking about DGD vs DGDMP, what does DGD's future look like?

I have a vague sense that it's going to be passively EOL'ed once the MP
version is stable.