The question is mostly if the underlying engine in GLMakie allows this and it is just not exposed or if there are other reasons why this is not available.
Meshscatter uses instancing, so the markers need to be the same to get a speed up out of it…
I guess, from an API standpoint, it may still be desirable to allow different markers…
We should likely implement it as a single big mesh with something like:
big_mesh = merge(marker_meshes...)
big_mesh = merge(marker_meshes...)
Could that work with per-element scales, rotations and positions? Also wouldn’t that result in unnecessarily large meshes when someone has some combination of two meshes as markers? (I.e. markers = rand([m1, m2], 1000))
I would say we rather want something that splits one meshscatter with multiple markers into multiple meshscatter calls with unique markers.