[prev in list] [next in list] [prev in thread] [next in thread] 

List:       koffice
Subject:    Re: includes
From:       weis () stud ! uni-frankfurt ! de
Date:       1999-04-20 20:15:59
[Download RAW message or body]

Hi,

Coolo, you have a point here.

There are basically two solutions to your problem:

a) the killu approach.
   Which separates the KOffice stuff entirely.
b) wait for the mico people. With namespaces they can
   avaoid including the complete ORB headers all the time
c) Use Torbens latest idea :-) I did that in the trader.

One just encapsulates as good as possible the whole CORBA interface
with a KDE interface. That means: Map sequences to QValueList, String_var
to QString and so on. I did that in KTrader. That seems to be more or less
the same idea Stephan had. But I can tell you that it really starts to 
suck if you have to change the idl interface and map it to the
KDE interface. Everytime you change the IDL, you have to change the KDE 
interface by hand and so on. Terrible on the long run.

In a kde library it is worth the efforts I think, but in application code ?
I think I prefer waiting for the compiler a bit longer instead of hacking
a KDE<-> CORBA layer all the time.

Bye
Torben


 On Mon, 19 Apr 1999, Stephan Kulow wrote:

> Hi!
> 
> Since it's really a pain to compile koffice,
> I thought about some possible changes to make
> the compilation faster. This should be every-
> one's interest, since compiling is a waste of
> time and other resources.
> 
> I just did some measurements and a single
> file with just one line "#include <koDocument.h>"
> expands to 37000 lines of code when ran through
> the preprocessor (koffice.h gets included, which
> includes CORBA.h, ...) and this file takes my
> computer:
> 
> real    0m15.038s
> user    0m12.410s
> sys     0m1.130s 
> 
> Quite terrible time I would say. Of course don't
> take this statistic as only truth, but it shows
> the trend.
> 
> And now I would like to know if this wouldn't
> qualify to work around including koffice.h
> whereever possible and making some "derives from"
> relations to "has a" even though this may 
> uglify the code. But when it would save 10 
> seconds per file, it would be for a good
> reason. 
> 
> I'm going to test a bit more with this "has
> a" stuff and I hope I can give you numbers
> to decide on.
> 
> Greetings, Stephan
> 
> -- 
> Und sie nannten ihn, wie er selbst unterschrieb -
> Den Trojanischen Pferdedieb
> 
> 

[prev in list] [next in list] [prev in thread] [next in thread]