186 replies to this topic

[~Dannyboy~]

the fastest script in gml is this

with(argument2) depth = point_distance(x,y,argument0,argument1);

Since you're only doing this to sort them by distance, it should be significantly faster to do without a squareroot... I don't have time to test it right now, but if I've understood you correctly, something like this should really be faster and achieve the same effect:

var dx, dy;
with(argument2) {
    dx = x - argument0;
    dy = y - argument1;
    depth = dx * dx + dy * dy;
}

That said, I wouldn't be surprised if it's slower in GML... Which brings me to a feature request!

Could we please have a point_distance_squared(x1, y1, x2, y2) function? Often in cases like this the square is all you need, and it's much faster to compute, it would be nice to have a nice fast built-in version.

EDIT: And if I may be so bold as to bring my feature request up to 15 minutes, point_distance_3d(x1, y1, z1, x2, y2, z2) and point_distance_squared_3d(x1, y1, z1, x2, y2, z2) would also be nice!

Edited by ~Dannyboy~, 20 March 2011 - 04:55 AM.

[Gamer3D]

What about overloading of extension functions?

Currently, when defining an extension function, we must define:

• NAME
• EXTERNAL_NAME
• ARGUMENT_COUNT
  [*}ARGUMENT_TYPE[ARGUMENT_COUNT]

I would like to be able to define two functions with the same NAME, but with different EXTERNAL_NAMEs and either different ARGUMENT_COUNTs or ARGUMENT_TYPEs. Current result(yes, I tested it): one function (always the same) gets called whether string or real is put in.

Why this would be useful:
As an example, DLLs for data structures. If the user wants to call add("blah") and add(1), then with the current system, the original programmer must have written this script for add:

if (is_real(argument0))
    return add_real(argument0);
else
    return add_string(argument0);

This adds significant overhead when the DLL itself is fast. Also, the complexity of the script required increases to 2^n when n are variables able to be either string OR real. With overloading, the entire burden would be shifted onto significantly faster compiled code, and would not require the script.

Potential benefits: Massive in some cases, none in others. Potential problems: Not backwards-compatible with GM7.
I also considered requesting this for DLLs, but because those are done with what is effectively a function pointer, it would be impossible.

Edited by Gamer3D, 20 March 2011 - 07:24 AM.

[Rani_sputnik]

EDIT: And if I may be so bold as to bring my feature request up to 15 minutes, point_distance_3d(x1, y1, z1, x2, y2, z2) and point_distance_squared_3d(x1, y1, z1, x2, y2, z2) would also be nice!

Agree! but I think that would end in a whirlpool situation like Mike discussed earlier, there are so many 3d math functions that would just be awesome to have added!
@ Mike, if 3d math is branching into a pool of math that's far to big, could you add a dot(x1,y1,x2,y2) I know it wouldn't be used very often so are probably low priority but I use it it's useful for 2d verlet physics and fake 3d engines. I use both every now and then and I don't imagine it'd be too hard. I can understand if you don't though, they are seldom used for most people I guess *shrug*

EDIT: Oh I am a fool, if point_distance_squared was implimented you could do 2d dot products *face palm* nevermind

Edited by Rani_sputnik, 20 March 2011 - 08:37 AM.

[sabriath]

#define taking too long to add? Joke?

Would you like me to add it for you?

Yes, it is more than a 10 minute fix...only 1 person on these forums can boast about their programming abilities, and that's me, and there's no way I am that fast. You require the following BNF/EBNF addition:

<preproc> := "#define" <ws>+ <ident> ['(' <ws>* <ident> [',' <ws>* <ident>]* ')' <ws>*] <notlf>* <lf>
<stmt> := <preproc>

Even if 'ws' and 'ident' parsing functionality is complete, that still leaves all the other things to parse, gen, verify, and test. Not to mention that anything that is done for the compiler, must also be done for the runner's runtime compilation part as well (including testing).

[TheSnidr]

There's a script called angle_difference. Wouldn't it be great if this was a function?

[Manuel777]

There's a script called angle_difference. Wouldn't it be great if this was a function?

A few pages before Mike said it was too complex to add new functions and test them (so, hopepefully this suggestions will be added to GM9)

Speaking of functions, is it too complex to expand the max number of arguments?

PS: why in hell GML is limited in the number of arguments??