Re: [Wf-gdk-wfcode] floats and attribedit

[Kevin S. <kt...@te...>]

On Sun, 2003-01-12 at 07:45, nl...@nl... wrote:
> With regard to adding float support to attribedit, should new classes
> TypeFloat32 and GUIElementFloat be added, or should the existing
> TypeFixed32/GUIElementFixed be renamed to TypeScalar32/GUIElementScalar
> with the actual type (fixed/float/double) being defined at compile time
> depending on the WF scalar type?

This is the first I have thought all of this through, so I could miss
something:

Hmmm, that is an interesting question, both answers make sense in
different contexts. From a general attribute editor point of view (not
specific to WF), AttribEdit ought to be able to edit both types
simultaneously. (or maybe not, if a system supports floats, who would
want to use the fixed type?).=20

I don't want a 'Scalar' Attribute in the oad .iff files because I want
the .iff file to be self describing (one should be able to tell the
actual type of a field by its name). So in the .iff file I think we
should add a type 'FLOT' and a type 'DOBL'.=20

Given that, it would probably make sense to have levelcon be able to
convert floats to fixed point if building a level file for the fixed
point version of the engine.=20

Given THAT, maybe AttribEdit should ALWAYS edit floats (or doubles), and
leave it to the level converted to truncate it to fixed if needed.

To do that would require updating the oad generation process to ouput
'DOBL' for the oad files (but still mark them as fixed in the in game
structures (if building fixed point)).

> The latter choice would mean that there is only ONE scalar type per
> level - no mixing of fixed, float, and double would be possible. The
> former choice would allow mixing of fixed, float, and double attributes
> in a level.

Since it is a compile time switch in the engine I agree there should
only be fixed or float in the level (and the whole game for that
matter). However, there might be some value to supporting multiple
precisions at once (float vs double) (and 32 bit fixed vs 64 bit fixed,
which we do have internally in WF).=20

I can't think of a reason to have Attribedit support fixed any longer,
just float & double. (Since the only purpose of the fixed point code was
for speed on systems which don't support floating point).

In summary I think the tools should all work only in floating point, the
engine should be buildable for fixed or float, and iff2lvl should
convert as appropriate.

so I am suggesting:
* add 'FLOT' & 'DOBL' support to iffcomp.=20
* add ability to edit 'FLOT' and 'DOBL' to AttribEdit
* add ability to read 'FLOT' and 'DOBL' chunks to iff2lvl (and output
fixed point)
* modify oad file generation to output 'FLOT' instead of 'F32' (and
consider adding a double precision scalar type which would output
'DOBL')
* finally, add a switch to iff2lvl to output floating point data in the=20
lvl file and change the .ht file generation to generate headers which
specify floats (both of these would occur based on the SCALAR_TYPE
switch).
=20
--=20
Kevin Seghetti: E-Mail: kt...@te..., HTTP: www.tenetti.org
GPG public key: http://tenetti.org/phpwiki/index.php/KevinSeghettiGPGKey
Check out www.worldfoundry.org for my GPL'ed 3D video game engine