Chain of transformation, how to?

[kkray]

Hello everybody.

In my project I need to apply chain of transformation for object.
For example I need, in order:
1. scale
2. rotate
3. move
4. scale (AGAIN!)
5. rotate
6. move
...
n. scale
n+1. rotate
n+2. move
n+3. draw to the screen

I need this, becuse I have an hierarchy of objects, like a tree. And if parent moves, then all of childs must move too. The same for scaling, rotating and so on.

I found sf::Matrix3 class, which can be built from transformations and can be multiplied with other matrixes, but I don't understand how to apply my result matrix to the sprite.

Can you help me? What the best way to do it?

[Laurent]

You can't, SFML is not made to allow this.

However, if you draw drawables inside other drawables (ie. in the Render function of your own class derived from sf::Drawable), the transformations are combined. You can use this to build a hierarchy of drawables.

[kkray]

Oh.. I'm sorry to hear it.
Also, I cant find a way to draw a quad from texture to quad on screen. It's impossible too?

PopCapFramework has a best draw-function I've ever seen -- DrawImageTransform( TransfromMatrix ) where TransfromMatrix can be produced from other matrixes, and matrix-class has functions like "translate","scale","rotate". Do you plan to implement functions like this?

Or, maybe, just Quad-class with tex-coords, scr-coords, texture, z-buffer for vertices, colors and so on? So that we coold fill vertext data by mulitplication point-vectors on matrix.

[kkray]

By the way, first time when I saw Sprote::Move, Rotate and Scale methods I thought that code like this:

Code: [Select]

spr.Scale( sx1,sy1 );
spr.Rotate( ang1 );
spr.Scale( sx2,sy2 );
means something like this:
Scale sprite, THEN rotate it and THEN scale it again. This would be wonderful

[Groogy]

Sprite's don't affect the original image, they are just a set of values that will be used when it's protected sf::Drawable::Render method is called. So they don't do anything until the sprite is actually being drawn.

[kkray]

Sprite's don't affect the original image, they are just a set of values that will be used when it's protected sf::Drawable::Render method is called. So they don't do anything until the sprite is actually being drawn.

Yes, I know. So what? I dont understand you.

[Groogy]

what you want are the modelview matrix transformation, but no opengl call is made until the sprite is drawn.

*EDIT*
Though now when I look at your post again... The sprite should scale by the amount that you write, rotate it by that amount of degrees, and then scale it by a new amount... You  mean that it doesn't?

[kkray]

Try to call scale, rotate and scale again. Only one scale works. Order of command also doesnt matter

spr.Rotate and then spr.Scale are not equal to "rotate and THEN scale".

[Groogy]

It should: http://www.sfml-dev.org/documentation/2.0/classsf_1_1Drawable.htm#af14b17e4f5a73d2b1457707734643550

[Laurent]

Also, I cant find a way to draw a quad from texture to quad on screen. It's impossible too?

Can't you do this with a sprite?

PopCapFramework has a best draw-function I've ever seen -- DrawImageTransform( TransfromMatrix ) where TransfromMatrix can be produced from other matrixes, and matrix-class has functions like "translate","scale","rotate". Do you plan to implement functions like this?

I don't know. Pure matrices are of course more flexible, but it is very convenient to have this "position/rotation/scale/center" API with everything separated and combined at render time. Anyway, this is something that I have to think about for SFML 2.

Or, maybe, just Quad-class with tex-coords, scr-coords, texture, z-buffer for vertices, colors and so on? So that we coold fill vertext data by mulitplication point-vectors on matrix.

Technically you'll be able to do this in SFML 2. But this is very inefficient, it's almost free to let the GPU transform the vertices with a matrix.

Though now when I look at your post again... The sprite should scale by the amount that you write, rotate it by that amount of degrees, and then scale it by a new amount... You mean that it doesn't?

No it doesn't. All transform parameters are independant. If they were not, I wouldn't be able to provide the corresponding getters. So calling Scale multiple times will combine the factors together, but it won't be impacted by the current position/rotation/center.

[kkray]

Also, I cant find a way to draw a quad from texture to quad on screen. It's impossible too?

Can't you do this with a sprite?

I cant, because I want to get rectangle from texture, and draw a quad(!), not rectangle.

(0,0),(w,0),(w,h),(0,h) --> (1,2),(3,4),(5,6),(7,8) for example. I want to get a way to render rects from texture to ANY quad on screen. This feature + vectors + matrixes = PROFIT!

And I dont want to write my own render-method, OGL is not my love

[Laurent]

Sorry, I don't understand what you want. Can you explain "get rectangle from texture"?

[kkray]

[Laurent]

Ok I get it. With SFML you'll only be able to draw it with a sprite, and apply the transformations that are provided by its API. You can't use custom geometry/transformation, unless you go with OpenGL directly.

[kkray]

Ok, thanks for answer
I'm going to write SpriteExt class with modified Move, Rotate, etc. methods, which will accumulate transformation, and in virtual draw method will use them as a chain. Most likely I will not write methods like GetPosition, becouse it will be hard work, and I dont need these methods for my task.

Should I expect some troubles between my SpriteExt and base SFMP-API ?