VIS: test plots losing information on y axis

[jbrockmendel]

tests.plotting.test_frame.TestDataFramePlots.test_subplots_timeseries_y_axis issues a warning when plotting a datetime64[tz] column:

pandas/tests/plotting/test_frame.py::TestDataFramePlots::test_subplots_timeseries_y_axis
  /Users/bmendel/Desktop/pd/matwarns/pandas/plotting/_core.py:384: FutureWarning: Converting timezone-aware DatetimeArray to timezone-naive ndarray with 'datetime64[ns]' dtype. In the future, this will return an ndarray with 'object' dtype where each element is a 'pandas.Timestamp' with the correct 'tz'.
  	To accept the future behavior, pass 'dtype=object'.
  	To keep the old behavior, pass 'dtype="datetime64[ns]"'.
    numeric_data[col] = np.asarray(numeric_data[col])

The relevant code is:

data = {"numeric": np.array([1, 2, 5]),
                "timedelta": [pd.Timedelta(-10, unit="s"),
                              pd.Timedelta(10, unit="m"),
                              pd.Timedelta(10, unit="h")],
                "datetime_no_tz": [pd.to_datetime("2017-08-01 00:00:00"),
                                   pd.to_datetime("2017-08-01 02:00:00"),
                                   pd.to_datetime("2017-08-02 00:00:00")],
                "datetime_all_tz": [pd.to_datetime("2017-08-01 00:00:00",
                                                   utc=True),
                                    pd.to_datetime("2017-08-01 02:00:00",
                                                   utc=True),
                                    pd.to_datetime("2017-08-02 00:00:00",
                                                   utc=True)],
                "text": ["This", "should", "fail"]}

testdata = pd.DataFrame(data)

ax_datetime_no_tz = testdata.plot(y="datetime_no_tz")
ax_datetime_all_tz = testdata.plot(y="datetime_all_tz")

Checking the actual images generated by these last two:

The UTC is completely lost for the tz-aware case, and in both cases the dates are rendered without year and with weird-looking hour.

[dstansby]

• The axis formatting is the default Matplotlib datetime formatting, which is currently "3 significant bits", so in this case month-day hour. In this case it isn't the best, but I think it plays nicely with different zoom levels. Suggestions for improvements welcome on the Matplotlib bug tracker!
• I think the warning is being emitted by numpy? The call np.asarray() is trying to cast some timezone aware stuff to np.datetime64, which I think is dropping timezone support soon.

[jklymak]

You can also see https://matplotlib.org/3.1.0/gallery/ticks_and_spines/date_concise_formatter.html for formatting alternatives. Can't help w/ the timezone stuff!