On 14.05.2009, at 12:55, Steve Wirts wrote:
> Hi Folks,
>
> Would it be inappropriate to request that an effort be made at some
> point to try to provide accessors for all instance/class instance
> variables, and try to get rid of inappropriate direct variable
> accessing where possible?
>
> The "chasing variables" function is handy but seems more cumbersome
> than just putting a halt in a setter.
>
> I know this would be a lot of work but maybe it could be automated.
>
> Just tell me to shut up and go away if I'm being a nuisance.
In general I am a fan of self-style way of not destinguishing between
sends and instvar access.
(like NewSpeak, Gilad has some nice blog-posts on why you want message
sends for everything).
But one thing can be quite problematic: performance. In Morphic, some
years ago somebody did this
refactoring. *all* accesses where replaced by messages sends. The
problem: performance. Especially
the path when accessing morphic properties added *lots* of message
sends, even for the case when
it was supposed to fail fast (e.g. morph-extension nil).
Just removing those got us a lot of performance back when comparing
3.9 with 3.8 (these changes where
later added to the OLPC image, too).
Of course, this can be optimized easily in a JITed environment, but
just doing it for everything in Pharo
now would not be wise. In addition, one would need to think about
syntax.... implicit self, yes or no?
The question about debugging of course can be solved, too, by making
variables real objects that one can,
for example, put a break-point on...
Marcus
--
Marcus Denker - http://marcusdenker.de
PLEIAD Lab - Computer Science Department (DCC) - University of Chile
_______________________________________________
Pharo-project mailing list
[email protected]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
