Add nan integer (ARM-DOE#842)

* Checking data type befor adding in NaN since will need the data type to be float. Upconverting if data is type integer.

* Testing for upconverting data in add_in_nan() when data is not type float.

* PEP8 for Black Linting.

act/utils/data_utils.py

@@ -455,8 +455,14 @@ def add_in_nan(time, data):
             mode = stats.mode(diff, keepdims=True).mode[0]
         except TypeError:
             mode = stats.mode(diff).mode[0]
+
         index = np.where(diff > (2.0 * mode))
 
+        # If the data is not float time and we try to insert a NaN it will
+        # not auto upconvert the data. Need to convert before inserting NaN.
+        if len(index) > 0 and np.issubdtype(data.dtype, np.integer):
+            data = data.astype('float32')
+
         offset = 0
         for i in index[0]:
             corr_i = i + offset

tests/plotting/test_timeseriesdisplay.py

@@ -606,6 +606,34 @@ def test_add_nan_line():
         matplotlib.pyplot.close(display.fig)
 
 
+@pytest.mark.mpl_image_compare(tolerance=10)
+def test_add_nan_line_integer():
+    data = np.arange(100, dtype=np.int32)
+    time = np.array('2019-11-01T00:00:00', dtype='datetime64[m]') + np.arange(data.size)
+    time = time.astype('datetime64[ns]')  # Only done to stop a warning appearing
+
+    # Remove data to produce a gap
+    data = np.delete(data, np.arange(50, 60), axis=0)
+    time = np.delete(time, np.arange(50, 60), axis=0)
+    data = np.delete(data, np.arange(70, 75), axis=0)
+    time = np.delete(time, np.arange(70, 75), axis=0)
+
+    ds = xr.Dataset(
+        data_vars={'data': ('time', data, {'long_name': 'Data values', 'units': 'degC'})},
+        coords={'time': ('time', time, {'long_name': 'Time in UTC'})},
+    )
+
+    display = TimeSeriesDisplay({'test_datastream': ds}, figsize=(15, 10), subplot_shape=(1,))
+    display.plot('data', subplot_index=(0,), add_nan=True, marker='.', markersize=20, linewidth=5)
+
+    assert np.issubdtype(ds['data'].dtype, np.integer)
+
+    try:
+        return display.fig
+    finally:
+        matplotlib.pyplot.close(display.fig)
+
+
 @pytest.mark.mpl_image_compare(tolerance=10)
 def test_timeseries_invert():
     ds = act.io.arm.read_arm_netcdf(sample_files.EXAMPLE_IRT25m20s)