feat: added aggregation operators (#29)

* feat: added aggregation operators

* fix: fixed c code for agg

* fix: removed zip json

Author: jlerat

src/hydrodiy/data/c_dutils.c

@@ -2,11 +2,13 @@
 
 /**
 * Aggregate inputs based on the aggindex
-* The operator applied to the aggregation is defined by op:
-* oper = 0 : sum
-* oper = 1 : mean
+* The operatorator applied to the aggregation is defined by op:
+* operator = 0 : sum
+* operator = 1 : mean
+* operator = 2 : max
+* operator = 3 : tail
 **/
-int c_aggregate(int nval, int oper, int maxnan, int * aggindex,
+int c_aggregate(int nval, int operator, int maxnan, int * aggindex,
     double * inputs, double * outputs, int * iend)
 {
     int i, nagg, nagg_nan, count, ia, iaprev;
@@ -35,7 +37,7 @@ int c_aggregate(int nval, int oper, int maxnan, int * aggindex,
         if(ia != iaprev)
         {
             /* Mean instead of agg */
-            if(oper == 1 && nagg>0)
+            if(operator == 1 && nagg>0)
                 agg/=nagg;
 
             /* Store outputs */
@@ -65,11 +67,19 @@ int c_aggregate(int nval, int oper, int maxnan, int * aggindex,
         } else
             nagg ++;
 
-        agg += inp;
+        if(operator<=1) {
+            agg += inp;
+        }
+        else if (operator == 2){
+            agg = inp > agg ? inp : agg;
+        }
+        else if (operator == 3){
+            agg = inp;
+        }
     }
 
     /* Final step */
-    if(oper == 1 && nagg>0)
+    if(operator == 1 && nagg>0)
         agg/=nagg;
 
     if(nagg_nan > maxnan)

src/hydrodiy/data/dutils.py

@@ -163,6 +163,8 @@ def aggregate(aggindex, inputs, operator=0, maxnan=0):
         Aggregation operator:
         0 = sum
         1 = mean
+        2 = max
+        3 = tail (last valid value)
     maxnan : int
         Maximum number of nan in inputs for each
         aggregation index

src/hydrodiy/data/tests/test_hydata_dutils.py

@@ -23,13 +23,17 @@
 # Utility function to aggregate data
 # using various version of pandas
 def agg_d2m(x, fun="mean"):
-    assert fun in ["mean", "sum"]
+    assert fun in ["mean", "sum", "max", "tail"]
 
     # Define aggregation function
     if fun == "mean":
         aggfun = lambda y: np.mean(y.values)
-    else:
+    elif fun == "sum":
         aggfun = lambda y: np.sum(y.values)
+    elif fun == "max":
+        aggfun = lambda y: np.nanmax(y.values)
+    else:
+        aggfun = lambda y: y.values[~np.isnan(y.values)][-1]
 
     # Run aggregation
     try:
@@ -191,6 +195,14 @@ def test_aggregate(allclose):
     obsm2 = dutils.aggregate(aggindex, obs.values, operator=1)
     assert allclose(obsm.values, obsm2)
 
+    obsm = agg_d2m(obs, fun="max")
+    obsm2 = dutils.aggregate(aggindex, obs.values, operator=2)
+    assert allclose(obsm.values, obsm2)
+
+    obsm = agg_d2m(obs, fun="tail")
+    obsm2 = dutils.aggregate(aggindex, obs.values, operator=3)
+    assert allclose(obsm.values, obsm2)
+
     kk = np.random.choice(range(nval), nval//10, replace=False)
     obs[kk] = np.nan
     obsm = obs.resample("MS").apply(lambda x: np.sum(x.values))
@@ -209,9 +221,8 @@ def test_aggregate(allclose):
 def test_aggregate_error(allclose):
     dt = pd.date_range("1990-01-01", "2000-12-31")
     nval = len(dt)
-    obs = pd.Series(np.random.uniform(0, 1, nval), \
-            index=dt)
-
+    obs = pd.Series(np.random.uniform(0, 1, nval),
+                    index=dt)
     aggindex = dt.year * 100 + dt.month
 
     try: