Description
For my use case, it would be ok if the setting is on the component level.
Use case:
I have a bunch of computer vision ML models that infer segmentation masks or bounding boxes for objects in an image. To compare the different checkpoints, I would like to have a flexible layout that would allow me to arrange model outputs side-by-side or layered on top of each other in z-index. The latter is to compare model results pixel-wise.
The dockview component already provides most of the UI and interaction that I want. For example, I can arrange outputs rendered in panels side-by-side where each panel contains the input image and segmentation masks or bounding boxes on top of an image (as SVG elements).
Now, for the z-index layering, it should render the input image in one tab, and then tabs on top in the same tab group should just render the segmentation masks or bounding boxes. Maybe one way to think about this is the layers view in Photoshop (see screenshot).
So, for this to work, it needs to render contents of all tabs, with each tab having a transparent background, so it shows the contents of the tabs "underneath"
Originally posted by @raedle in #397 (comment)