If you provide both bargap and scattergap but they're different, traces in the same alignment and offset group do not properly align. One way or another, ensuring these traces align is more important than honoring both gaps. Here, the orange and green scatter points should be centered on the bars of the same color, but they are not:
If there are no bar traces, bargap will not be defined, so scattergap should be honored.
Even if there are bar traces but only scattergap is explicitly given, seems like it should be honored and used as the bargap value as well.
There are also boxgap and violingap - in principle you could have ALL of these traces in the same alignment and offset groups, do we define a hierarchy, something like bar beats scatter beats box beats violin? Or do we take the gap from the first trace in the group, and use that for all traces in that group? Or the maximum gap that applies to any trace in the group? I have no strong opinions here, so long as the result is always proper alignment.
All but scatter also have *groupgap attributes. There's no problem allowing these to be different.
If you provide both
bargapandscattergapbut they're different, traces in the same alignment and offset group do not properly align. One way or another, ensuring these traces align is more important than honoring both gaps. Here, the orange and green scatter points should be centered on the bars of the same color, but they are not:https://codepen.io/alexcjohnson/pen/yLdEZvo?editors=0010
This is a little tricky though:
bargapwill not be defined, soscattergapshould be honored.scattergapis explicitly given, seems like it should be honored and used as thebargapvalue as well.boxgapandviolingap- in principle you could have ALL of these traces in the same alignment and offset groups, do we define a hierarchy, something like bar beats scatter beats box beats violin? Or do we take the gap from the first trace in the group, and use that for all traces in that group? Or the maximum gap that applies to any trace in the group? I have no strong opinions here, so long as the result is always proper alignment.*groupgapattributes. There's no problem allowing these to be different.