Fix read_ecmwf_macc test failure for cftime>=1.6.0 (#1656)

* use 'nearest' instead of 'before' and 'after'

* whatsnew

Author: kandersolar

docs/sphinx/source/whatsnew/v0.9.5.rst

@@ -25,6 +25,9 @@ Bug fixes
 * Fixed incorrect mapping of requested parameters names when using the ``get_psm3``
   function. Also fixed the random reordering of the dataframe columns.
   (:issue:`1629`, :issue:`1647`, :pull:`1648`)
+* When using ``utc_time_range`` with :py:func:`pvlib.iotools.read_ecmwf_macc`,
+  the time index subset is now selected with ``nearest`` instead of ``before``
+  and ``after`` for consistency with ``cftime>=1.6.0``. (:issue:`1609`, :pull:`1656`)
 
 
 Testing
@@ -56,3 +59,4 @@ Contributors
 * Mark Mikofski (:ghuser:`mikofski`)
 * Anton Driesse (:ghuser:`adriesse`)
 * Adam R. Jensen (:ghuser:`AdamRJensen`)
+* Michael Deceglie (:ghuser:`mdeceglie`)

pvlib/iotools/ecmwf_macc.py

@@ -294,9 +294,9 @@ def read_ecmwf_macc(filename, latitude, longitude, utc_time_range=None):
         nctime = ecmwf_macc.data['time']
         if utc_time_range:
             start_idx = netCDF4.date2index(
-                utc_time_range[0], nctime, select='before')
+                utc_time_range[0], nctime, select='nearest')
             end_idx = netCDF4.date2index(
-                utc_time_range[-1], nctime, select='after')
+                utc_time_range[-1], nctime, select='nearest')
             time_slice = slice(start_idx, end_idx + 1)
         else:
             time_slice = slice(0, ecmwf_macc.time_size)