[MNT]: MultiCursor should not take canvas as first parameter #21496
Comments
|
I guess you need to do it more or less by hand, taking |
|
If we are going to mess up with the signature, perhaps we can also consider changing, at the same time, the somewhat unsightly horizOn/vertOn kwargs to both MultiCursor and Cursor? Currently some widgets use "orientation" and others use "direction", but I think standardizing on "orientation" is better ("direction" is also used for "in"/"out" for ticks, whereas colorbar uses "orientation" for horizontal/vertical). |
Summary
During the sprint, we were looking at widgets, and I noticed that the
MultiCursorconstructor takes aFigureCanvasBaseas first parameter. This probably exists to say where to connect signals, but this does not match how any of the other widgets work, which infer this from the suppliedAxes. Theoretically, I don't see any reason why theAxesneed to be on a singleFigureforMultiCursorto work, but that's implied by it only using a single canvas.Proposed fix
The canvas should be found from the
Axes, and the constructor should assert that all suppliedAxesare on the sameFigure(if that is indeed required.)However, I don't know how to deprecate this. If we use
_api.delete_parameter, then the remaining parameters become keyword-only, but we don't want to do that to theAxesparameter, which should have been the first positional parameter.The text was updated successfully, but these errors were encountered: