Re: marshalling merges



On Wed, 22 Jul 1998, Tim Janik wrote:

> hm, i night get around to implement an automatic marshaller lookup in
> the next few days.  but still, why wouldn't you want to use that in gtk
> itself?  it would impose an extra hashtable lookup upon class_init()'s
> sgnal_new()  execution, and i'm pretty convinced that these lookups are
> not going to be noticable. 

This is "creeping featurism" and IMHO it's bad :) If we kept adding little
unnoticeable features, it would eventually result in a badly bloated lib.
(Worst part is that there's no one culprit you can pin the blame on, so
it's a nightmare to slim back down.)

Automatic marshaller lookup does makes it somewhat more convenient for the
programmer to use. However, it really isn't that hard to type in
'gtk_marshal_RETURNTYPE__PARAM_PARAM...' instead of 'NULL', especially
when you have to do it exactly once for each gtk_signal_new call. 
Approximately 25% of the widgets in gtk+ itself use gtk_signal_new. 

My point is that exerting very little extra effort will decrease startup
time slightly, and cut memory usage as well.

Little improvements count,
-- Elliot
Do you ever just feel thankful that you know me and have access to my
dementia? Explain. Be prepared to discuss in class.



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]