[lazarus] KCL and Open for Opinion

Michael Van Canneyt michael.vancanneyt at wisa.be
Fri May 21 05:51:23 EDT 1999




On Thu, 20 May 1999, Michael A. Hess wrote:

> Nicolas Aragon wrote:
> > 
> <snip>
> >
> > These features *need* a well designed CL, not only a working editor,
> > because they rely on components interface. Not talking about
> > streaming...
> 
> All true, but you need to walk before you can run. Also FPC can be used
> to write code that doesn't use OOP. An IDE would be nice for that as
> well. There is text based IDEs and a GUI IDE for Windows but no GUI IDE
> for doing any kind of work in Linux. Once we have a simple IDE like the
> Windows version and like the text based version then we can more easily
> work on an actual RAD IDE.

There are many GUI editors that can be used perfectly with FPC. XCoral,
Nedit, FTE. All of them support commands to compile and build, syntax
highlighting etc.

I tend to agree with Nicolas. I think that creating a 'flat' ide is a step
backwards, and should be avoided. I didn't want to influence the discussion,
hence I stayed silent till now. But something came up recently...

The main argument seems that the compiler isn't ready for OOP yet. Poppycock.
The classes have been functioning fine for ages. The compiler itself is
completely written in OOP !

The two main problems are, at the moment:
- Ansistrings. Florian just did a complete overhaul of the ansistring code,
  and it seems to be running fine (from the tests we do, of course)
- The dispatch mechanism, and related the procedure of object. In principle,
  this has been fixed too, by Peter Vreman, and Florian.

It is my firm belief that the first of these two problems has been the
principal source of problems that have been experienced by the lazarus
people. I have seen some pretty dubious code in the ansistring department when
I checked out the lazarus code so it could compile on Win32. Heck, it even didn't
compile because of the faulty string handling. 

Hence, the principal argument for making a 'flat' IDE has been rebuffed :-)
Can we get back to OOP please ? Sergio has been doing a great job with
'Opendialog', it would be a shame to loose that :)

Furthermore, and this is more serious:

There are 2 germans working on what they call KCL : A FCL descendent like Lazarus
the builds on top of QT and KDE ! (they solved somehow the C++ name mangling problems)
I have been in contact with them for some time, and they seem enthousiast in
co-operating with lazarus in the sense that we should try and build a common
GUI-independent class tree as much as possible.

They will also write a IDE in this (for the same reasons you do), but it will be
completely OOP on top of this.

They also have a signal-and-slot mechanism (as found in Qt) that could be adopted
to GTK, from what I saw. This would allow multiple slots for 1 signal etc.
(i.e. multiple event handlers get triggered for the same event).

Since they'll be using OOP from the very start (no other way with Qt :) ) I think
it should be so for Lazarus too.

I hope I have convinced you that OOP is the way to go from the start, or
at least that I have opened the discussed again...

But let's not stay on the discussion level too long. There is code to be written.
Preferably OOP code :-)

Michael V.






More information about the Lazarus mailing list