Adjust Python wrapper to year counting as start of year

See also JGCRI/hector#177

pyhector/__init__.py

@@ -254,4 +254,4 @@ def run(scenario, options=None):
         global_temp = h.get_observable("temperature", "Tgav")
         start = int(options["core"]["startDate"])
         end = int(options["core"]["endDate"])
-    return pd.Series(global_temp, index=pd.Index(range(start, end + 1)))
+    return pd.Series(global_temp, index=pd.Index(range(start, end)))

tests/test.py

@@ -22,13 +22,15 @@ def read_hector_output(csv_file):
     Reads a Hector output stream csv and returns a wide DataFrame with
     Hector output data.
     """
-    # Filter out spin-up values. RCP output streams seem to start at 1746
-    # though startDate ist 1745.
+    # Filter out spin-up values. In Hector 1.x RCP output streams years are
+    # written as end of simulation year. This will change in Hector 2.x.
+    # See https://github.com/JGCRI/hector/issues/177
     start_year = 1746
     output_stream = pd.read_csv(csv_file, skiprows=1)
 
     wide = output_stream[output_stream.year >= start_year].pivot_table(
         index="year", columns="variable", values="value")
+    wide.index = wide.index - 1  # Adjust index to year only
 
     return wide
 
@@ -40,4 +42,4 @@ def test_rcps():
             os.path.join(path, "./data/outputstream_{}.csv".format(name))
         )
         output = pyhector.run(scenario)
-        assert output.loc[1746:].round(2).equals(original.Tgav.round(2))
+        assert output.round(2).equals(original.Tgav.round(2))