line-and-point legends

I need a Line Chart with Points plots and there is an example. It states that altair.Chart(data).mark_line(point=True) will produce line chart with points.

Since I have multiple series to plot I specified color='kind', strokeDash='kind' (or both). I found that the legends are only circles.

• With color='kind', the circles are filled with the colors.
• With strokeDash='kind', the boundary of the circles are solid or dashed lines.

Without point=True I have lines and the effects are that lines are colored or dashed. With both point=True and {color='kind', strokeDash='kind', or both} I would expect the legends to have both lines and circles, same appearance as plot. I couldn’t figure out how to achieve the desired behavior so that I am opening this issue.

[joelostblom]

Thanks for providing an example @majiang! Currently it is not possible to combine legends of points and lines, you can follow the discussion here https://github.com/altair-viz/altair/issues/1785 and upstreams in vega-lite here https://github.com/vega/vega-lite/issues/5996

If you append .resolve_scale(color='independent', strokeDash='independent') to your chart, you can get something like this:

The title can be removed from one of the legends by setting title='' for that encoding channel.

[majiang]

Sorry for not being able to try and feedback that time.

I now think it is impossible due to vega-lite’s limitation and describe it with examples.

The desired output was something like

which I created with Excel. By “line-and-point legend” I meant line legend and point legend combined as they look.

With altair, when I opened this issue,

def sample_data():
    return pandas.DataFrame({
        "x": [0, 1, 2, 3, 0, 1, 2, 3],
        "y": [0, 1, 2, 3, 3, 2, 1, 0],
        "kind": ["a", "a", "a", "a", "b", "b", "b", "b"],
    })

altair.Chart(sample_data()).mark_line(point=True).encode(
    x="x", y="y", color="kind", strokeDash="kind")

produced

the main part of which was satisfactory, but only legend was not.

I have also tried

(
    altair.Chart(sample_data()).mark_point().encode(
        x="x", y="y", color="kind"
    )
    +
    altair.Chart(sample_data()).mark_line().encode(
        x="x", y="y", color="kind", strokeDash="kind"
    )
)

which produces similar result:

Here I don’t mind whether the circles are filled or not.

Since the appearance of legend of a Chart is determined fully automatically by vega-lite from the definition of encoding channels, we can’t produce the desired output.