Merge pull request #362 from NREL/pandas_200

Fixes for pandas 2.0.0 compatibility

.github/workflows/pytest.yaml

@@ -43,4 +43,4 @@ jobs:
         pip install ${{ matrix.env }}
     - name: Test with pytest ${{ matrix.env }}
       run: |
-        pytest
+        pytest

docs/sphinx/source/changelog.rst

@@ -1,5 +1,6 @@
 RdTools Change Log
 ==================
+.. include:: changelog/v2.1.5.rst
 .. include:: changelog/v2.1.4.rst
 .. include:: changelog/v2.1.3.rst
 .. include:: changelog/v2.1.2.rst

docs/sphinx/source/changelog/v2.1.5.rst

@@ -0,0 +1,13 @@
+*************************
+v2.1.5 (May 16, 2023)
+*************************
+
+Bug Fixes
+---------
+* Add support for pandas 2.0 (:issue:`361`, :pull:`362`)
+
+
+Contributors
+------------
+* Kevin Anderson (:ghuser:`kanderso-nrel`)
+* Michael Deceglie (:ghuser:`mdeceglie`)

rdtools/degradation.py

@@ -36,7 +36,7 @@ def degradation_ols(energy_normalized, confidence_level=68.2):
 
     # calculate a years column as x value for regression, ignoring leap years
     day_diffs = (df.index - df.index[0])
-    df['days'] = day_diffs.astype('timedelta64[s]') / (60 * 60 * 24)
+    df['days'] = day_diffs / pd.Timedelta('1d')
     df['years'] = df.days / 365.0
 
     # add intercept-constant to the exogeneous variable
@@ -121,7 +121,7 @@ def degradation_classical_decomposition(energy_normalized,
 
     # calculate a years column as x value for regression, ignoring leap years
     day_diffs = (df.index - df.index[0])
-    df['days'] = day_diffs.astype('timedelta64[s]') / (60 * 60 * 24)
+    df['days'] = day_diffs / pd.Timedelta('1d')
     df['years'] = df.days / 365.0
 
     # Compute yearly rolling mean to isolate trend component using
@@ -266,7 +266,7 @@ def degradation_year_on_year(energy_normalized, recenter=True,
                        tolerance=pd.Timedelta('8D')
                        )
 
-    df['time_diff_years'] = (df.dt - df.dt_right).astype('timedelta64[h]') / 8760.0
+    df['time_diff_years'] = (df.dt - df.dt_right) / pd.Timedelta('365d')
     df['yoy'] = 100.0 * (df.energy - df.energy_right) / (df.time_diff_years)
     df.index = df.dt
 

rdtools/filtering.py

@@ -424,8 +424,9 @@ def logic_clip_filter(power_ac,
     # series sampling frequency is less than 95% consistent.
     _check_data_sampling_frequency(power_ac)
     # Get the sampling frequency of the time series
-    time_series_sampling_frequency = power_ac.index.to_series().diff()\
-        .astype('timedelta64[m]').mode()[0]
+    time_series_sampling_frequency = (
+        power_ac.index.to_series().diff() / pd.Timedelta('60s')
+    ).mode()[0]
     # Make copies of the original inputs for the cases that the data is
     # changes for clipping evaluation
     original_time_series_sampling_frequency = time_series_sampling_frequency
@@ -651,8 +652,7 @@ def xgboost_clip_filter(power_ac,
     # series sampling frequency is less than 95% consistent.
     _check_data_sampling_frequency(power_ac)
     # Get the most common sampling frequency
-    sampling_frequency = int(power_ac.index.to_series().diff()
-                             .astype('timedelta64[m]').mode()[0])
+    sampling_frequency = int((power_ac.index.to_series().diff() / pd.Timedelta('60s')).mode()[0])
     freq_string = str(sampling_frequency) + "T"
     # Min-max normalize
     # Resample the series based on the most common sampling frequency

rdtools/test/availability_test.py

@@ -162,7 +162,7 @@ def difficult_data():
 
     # generate a plausible clear-sky power signal
     times = pd.date_range('2019-01-01', '2019-01-06', freq='15min',
-                          tz='US/Eastern', closed='left')
+                          tz='US/Eastern')
     location = pvlib.location.Location(40, -80)
     clearsky = location.get_clearsky(times, model='haurwitz')
     # just scale GHI to power for simplicity

rdtools/test/degradation_test.py

@@ -34,7 +34,7 @@ def get_corr_energy(cls, rd, input_freq):
             freq = input_freq
 
         x = pd.date_range(start=start, end=end, freq=freq)
-        day_deltas = (x - x[0]).astype('timedelta64[s]') / (60.0 * 60.0 * 24)
+        day_deltas = (x - x[0]) / pd.Timedelta('1d')
         noise = (np.random.rand(len(day_deltas)) - 0.5) / 1e3
 
         y = 1 + daily_rd * day_deltas + noise

setup.py

@@ -35,7 +35,7 @@
     'pytest >= 3.6.3',
     'coverage',
     'flake8',
-    'nbval',
+    'nbval==0.9.6',  # https://github.com/computationalmodelling/nbval/issues/194
     'pytest-mock',
 ]