Incorrect mpld3 plotting in log scale

[sheldonpark]

There appears to be an problem plotting in log scale using mpld3 (v 1.3.1). This problem was reported on stackflow (http://stackoverflow.com/questions/25188903/incorrect-mpld3-plotting-in-log-scale?noredirect=1#comment39225554_25188903) and confirmed by another person.

Thanks.
-Sheldon

[jakevdp]

This does look like a bug, and I'm not sure why it's happening.

Here's a minimal snippet of code that reproduces the bug:

import matplotlib.pyplot as plt
import mpld3

x = [1, 10, 100, 1000]
y = [1, 10, 100, 1000]
ax = plt.subplot(xscale='log', yscale='log')
ax.scatter(x, y)
mpld3.show()

I think it has something to do with the translation/offset framework of path collections, but I'm not sure.

Thanks for reporting this!

[kpretz]

The panning function doesn't also doesn't work for scatterplots when either of the axes is in log scale. This seems to be a problem only for scatter, and not plot. One work around is to use plot instead of scatter.

ax.plot(x, y, marker='o', linestyle='')

rather than

ax.scatter(x, y)

Finally, the axis labels are a bit weird, for instance:

[DanielAndreasen]

It's a poor workaround though, if one wants to color the points.

[kpretz]

You can set the point colors using the args 'markeredgecolor' and 'markerfacecolor' - so I think it's still a fine workaround. You have nearly all the same control over plot style with plot compared to scatter.

plot docs here: https://matplotlib.org/api/_as_gen/matplotlib.axes.Axes.plot.html

[DanielAndreasen]

Forgot about that, but not exactly what I mean.

1. It was not a critique against your suggestion. That suggestion is better than nothing.
2. I mean to color each point in a different color (according to some value) like: plt.scatter(x, y, c=z), and have a colorbar as well. Not sure if plot can do this.

Anyway, thanks for your respond.

[kpretz]

Good point. Thanks for the clarification. Obviously, being able to use both scatter and plot as designed would be ideal.

[nanls]

The workaround worked fine for me. 🎉

For those who are using mpld3.plugins.PointLabelTooltip to add labels when hovering a point of the scatter plot with something like this:

    import mpld3
    fig, ax = plt.subplots()
    ax.set_xscale('log', basex=10) 
    # because of the log axis, the scatter plot will not work fine
    scatter = ax.scatter(x, y)
    tooltips = mpld3.plugins.PointLabelTooltip(scatter, labels=labels)
    mpld3.plugins.connect(fig, tooltips)

The labels will not work if you only replace scatter by plot. You also need to change something here:

    tooltips = mpld3.plugins.PointLabelTooltip(scatter[0], labels=labels)
    # see here ----------------------------------------^

Hope that helps. :)