Stop drawing pixels out of region

[kubawal]

I'm drawing lots of rotated sprites.
And i have to draw some of this sprites partly - that ones, that are partly out of given region.
Briefly, I want do draw only that pixels of sprites that are in specified region.
Effect have to be like using sf::View but on single drawable object

PS I do it in my own Drawable component so i can't use sf::View::setViewport().

Any ideas?

[Gambit]

You will probably have to use an OpenGL stencil for this sort of operation, I'm not sure that SFML supports partially drawing sprites or textures. Alternatively you could probably use a shader but that might be a little more work.

[kubawal]

Shaders isn't my best side

[Laurent]

Why can't you use sf::View? It's your best option to implement clipping without doing OpenGL stuff.

[kubawal]

So i have another question: does view transformations applies on end of frame or for each draw()?
I'm asking it because i have idea:

virtual void draw(RenderTarget& target, RenderStates rs)
{
rs.transform *= getTransform();

View oldV = target.getView();

View newV;
// setting parameters of new view
target.setView(newV);

// draw my stuff

target.setView(oldV);
 

Does my stuff draw with newV wiew or oldV view?
Can I apply view on only a few of objects?

[eXpl0it3r]

Once you set the view, all objects that get drawn after, will be drawn with the set view.

Keep in mind though that the draw function from sf::Drawable is const, so you can't change objects states in there.

[kubawal]

It doesn't work or I do something wrong.
The whole problem is because I'm rendering a physics simulation that have its own coordinates.
And I want to render specified rect of physics simulation to specified rect on window (not on whole window)

// some useful inlines that are missing in SFML
inline Vector2f size(const FloatRect& rect)
{
    return Vector2f(rect.width, rect.height);
}

inline Vector2f pos(const FloatRect& rect)
{
    return Vector2f(rect.left, rect.top);
}

inline Vector2f localCenter(const FloatRect& rect)
{
    return size(rect) * 0.5f;
}

inline Vector2f globalCenter(const FloatRect& rect)
{
    return pos(rect) + localCenter(rect);
}

void WorldRender::draw(RenderTarget &target, RenderStates states) const
{
    //states.transform *= getTransform(); // i don't need Transformable transform because i use the same transform in view

// vp is source region of simulation

    // view set
    View oldV = target.getView();
    View newV;
newV.setRotation(getRotation());
    newV.setSize(size(vp));
    newV.setCenter(globalCenter(vp));
    newV.setViewport(FloatRect(getPosition(), Vector2f(size(vp).x * getScale().x, size(vp).y * getScale().y)));
    target.setView(newV);

// only for example    
CircleShape cs(50);
    cs.setPosition(0, 0);
    cs.setFillColor(Color::Black);
    target.draw(cs, states);

//...

target.setView(oldV); // restore old view
}
 

And in this case NOTHING renders. Even simple circle  (cs in my code).
But out of this function everything works properly.

[Laurent]

The viewport is defined in normalized coordinates. Please read the documentation carefully.

[kubawal]

Thanks, it works.
Here's my code. Maybe it will help someone.

View newV;
newV.setSize(size(vp));
newV.setCenter(globalCenter(vp));
newV.setRotation(getRotation());
Vector2f vpPos(getPosition().x / target.getSize().x, getPosition().y / target.getSize().y);
Vector2f vpSz((size(vp).x * getScale().x) / target.getSize().x, (size(vp).y * getScale().y) / target.getSize().y);
newV.setViewport(FloatRect(vpPos, vpSz));
target.setView(newV);
 

[Nexus]

I don't know how expensive these function calls are, but as a general rule, I would store results in variables instead of calling the same function multiple times. It's also more readable. That concerns getPosition(), getScale() and getSize().

Alternatively, you can use thor::cwiseProduct() or thor::cwiseQuotient() to multiply/divide vectors component-wise, see here.