Add a new descendants api as well as deeper element queries to ElementHandle

[tronical]

No description provided.

[ogoffart]

Should visit_element be removed. It is a bit redundent since you can do HasElementHandle::root_element(component).visit_descendants

(the only difference is that it visits the root)

[ogoffart]

Will it? All Clip elements? Why is that?

[tronical]

Maybe I'm missing something (I tried to cover that also in the test I added in search_api.rs), but the old code did this:

    pub fn is_visible(&self) -> bool {
        let item = self.borrow();
        let is_clipping = crate::item_rendering::is_clipping_item(item);
        let geometry = self.geometry();
        if is_clipping && (geometry.width() <= 0.01 as _ || geometry.height() <= 0.01 as _) {
            return false;
        }
        if let Some(parent) = self.parent_item() {
            parent.is_visible()
        } else {
            true
        }
    }

That suggests to me that is_clipping_item() returns true for Clip elements that are enabled for clipping (clip true) and the visible pass gives them the null width/height. So I think that'll return false.

The new code should return an empty clip rectangle from the call to absolute_clip_rect_and_geometry because of this:

      if crate::item_rendering::is_clipping_item(item) {
            clip = geometry.intersection(&clip).unwrap_or_default();
        }

And then the intersection of itself should be empty.

Should I add a direct unit test in here perhaps?

[ogoffart]

A test for that would be nice.

So if you have a

outer := Rectangle { 
   clip: true;  
   inner := Rectangle { x: parent.width; background: blue; } 
}

Then the Clip item for that outer rectangle would still be considered visible.
But the inner Rectangle would not;

I think either behavior is fine. It's just that the documention is a bit strange.
I think what you mean is that if an item has a size of 0x0 it will not be considered visible (because the intersection is empty)

[tronical]

Then the Clip item for that outer rectangle would still be considered visible.
But the inner Rectangle would not;

The Clip item would not be considered visible (the function returns false). AFAICS the old code did that as per this condition:

        if is_clipping && (geometry.width() <= 0.01 as _ || geometry.height() <= 0.01 as _) {  
            return false;  
        }  

and the new code should do the same by intersection.

[tronical]

Should visit_element be removed. It is a bit redundent since you can do HasElementHandle::root_element(component).visit_descendants

(the only difference is that it visits the root)

Yes, I'm in favor of that - if you're okay with the rest of this API. I'd make that a separate change though.

[tronical]

Going back into draft mode. I'm removing the children bits for now and want to proceed with the matching API that includes descendants but also the other criteria.

[ogoffart]

Naming: We often use the _impl prefix for this kind of helper function, instead of _internal

[tronical]

I assume you meant suffix. Changed :)

[ogoffart]

A test for that would be nice.

So if you have a

outer := Rectangle { 
   clip: true;  
   inner := Rectangle { x: parent.width; background: blue; } 
}

Then the Clip item for that outer rectangle would still be considered visible.
But the inner Rectangle would not;

I think either behavior is fine. It's just that the documention is a bit strange.
I think what you mean is that if an item has a size of 0x0 it will not be considered visible (because the intersection is empty)

[ogoffart]

why is this commented out?

[tronical]

This is a test that assumes the availability of children traversal, which I removed again. I can remove the commented out test, but I want to bring it back when adding children support.