Bug summary

Plotting a numpy masked array against a list of datetime objects,plt.plot functions as expected, butplt.scatter still displays the masked data. Auto-axis limitsdo respect the masked data though.

Code for reproduction

importmatplotlib.pyplotaspltimportnumpy.maasmafromdatetimeimportdatetime# Create a list of datetime objectstime_unmasked= [datetime(2022,11,24,4,49,52),datetime(2022,11,24,4,49,53),datetime(2022,11,24,4,49,54),datetime(2022,11,24,4,49,55),datetime(2022,11,24,4,49,56),datetime(2022,11,24,4,49,57),datetime(2022,11,24,4,49,58)]# Create a masked array with three masked variables, one datapoint far from othersdata_masked=ma.array([15,14,50,15,15,15,16],mask=[0,0,1,1,1,0,0])# Line plot (expected behaviour)plt.plot(time_unmasked,data_masked)plt.show()# Scatter plot (plots masked points, but ignores masked values when auto-setting axis limits)plt.scatter(time_unmasked,data_masked)plt.show()

Actual outcome

plot behaves as expected.

scatter unexpectedly draws the masked datapoints, but axes limits are still respecting the masked array as the maskedy=50 value is not visible.

Expected outcome

Scatter should not plot the masked data even though all timestamps are unmasked.

Additional information

It seems as though there is a check withinmatplotlib.pyplot.scatter that is not catching. I'd expect that logic to check that if eitherx ory were masked, to not plot that point. The datetime object is able to slip through. I've also tried puttingtime_unmasked within a numpy masked array with the same mask asmasked_data (so bothx andy have the same mask), but the same unexpected behaviour occurs.

Operating system

MacOS 11.3.1 (20E241)

Matplotlib Version

3.5.3

Matplotlib Backend

module://ipykernel.pylab.backend_inline

Python version

3.8.11

Jupyter version

5.7.10

Installation

conda