Re: Re: Registries [2]

At 14:29 2003-08-05, David Abrahams wrote:
> >>I honestly have no clue.  How would you know which other cast graph
> >>to link up to?
> >
> > What's wrong with explicit importing? It's the duplicated storage that's
> > bad, no?
>
>Yes, I suppose.  Of course polymorphic classes have edges in both
>directions, so importing converters and defining new conversions might
>have an effect on the imported graph.  Then there's a cache which
>records reachability and casting shortcuts, and that couldn't be
>shared because different information might be visible on either side
>of the extension boundary.

This shouldn't complicate things much though, adding edges on the imported
graph doesn't seem to have any issues, and the cache can just remain
local for the extension.

>Of course, I haven't proven that the cache is really important to
>speed -- it's another premature optimization -- but I'm guessing
>you'll want it.  Certainly some people had problems with the speed of
>stepwise up/downcasting in v1.

Yeah, I want the cache. :)

> > How about revisiting hierarchical implicit importing? Didn't we only
> > abandon that because we thought local registries wouldn't work
> > on unix platforms?
>
>Argh.  I don't remember.  However, it seems like it could only
>complicate the cast graph issue.

Maybe, but it seems to me that if we have explicit importing that
modifies the cast graph's, implicit importing doesn't complicate
things much. I could be wrong though..

---
Daniel Wallin



-------------------------------------------------------
This SF.Net email sponsored by: Free pre-built ASP.NET sites including
Data Reports, E-commerce, Portals, and Forums are available now.
Download today and enter to win an XBOX or Visual Studio .NET.
http://aspnet.click-url.com/go/psa00100003ave/direct;at.aspnet_072303_01/01