[lazarus] My opinions

Shane Miller SMiller1 at stvgb.org
Tue Oct 26 09:33:02 EDT 1999


This looks like a very interesting method of doing this.  I am a bit confused on how we will set it up to use the right gtkgraphint, or win32graphint, or qtgraphint.  How will the compiler know which one to use?  It seemed to me that the ???graphint.pp was the lowest class (child) so there wasn't anything above it that yuou could do a defines on.

Please just explain that to me.  I think once you explain it I'll probably agree with your method and we can push towards that eventually.

Shane


>>> Florian Klaempfl <Florian.Klaempfl at gmx.de> 10/26/99 07:29AM >>>
Shane Miller wrote:
> 
> The graphics unit was mine.
> I'll agree to you modifying it assuming that all you are really going to do it cause TCanvas to call DrawLine or DrawRect instead of the SendMessage procedure.  Also, I would like to look at your idea about using abstract classes, but I'm not sure we can implement it real easy without changing a lot of stuff.  If you think we can (and I haven't really put time into thinking about it) try it and let me take a look at it.
> 
> I don't mind using someone elses idea, as long as it's better than mine...
> 
> Shane
> 
> 
> >>> Florian Klaempfl <Florian.Klaempfl at gmx.de> 10/25/99 02:59PM >>>
> Shane Miller wrote:
> >
> > Florian brings up a good point. We COULD call the functions directly in SOME cases, like drawpixel as long as we made sure that each ???object.pp file has those functions in them.  This certainly would speed things up.
> 
> This should be no problem, you make simple an abstract class where all procedures are
> declared as abstract.
> 
> >
> > I like the send message function because then we only need to have one real public function in the object interface
> 
> What's the problem with having more public functions ? I would even propose don't putting
> everything in
> gtkinterface class, but is a matter of taste I think, but it would lead to a little bit
> more structured
> class library.
> 
> I don't know who is currently working on the graphics unit, if you want me I can implement
> the graphics
> stuff at least for win32/gtk as I propose.
> 
> > (gtkinterface class) but if we were able to call the functions directgly, we would aviod the large CASE statement but then we would have to make sure that we add a function called DRAWPIXEL in the win32 interface, the QT interface, etc.  I think in this case that would work though....

I attached a first study to clearify the ideas.
The important files are graphint and gtkgraphint, these routines
are called by the implementation of tcanvas methods.






More information about the Lazarus mailing list