Scale Series.prod(skipna=True)

sdc/datatypes/hpat_pandas_series_functions.py

@@ -3422,7 +3422,7 @@ def hpat_pandas_series_prod_impl(self, axis=None, skipna=None, level=None, numer
             _skipna = skipna
 
         if _skipna:
-            return numpy.nanprod(self._data)
+            return numpy_like.nanprod(self._data)
         else:
             return numpy.prod(self._data)
 

sdc/functions/numpy_like.py

@@ -364,3 +364,32 @@ def impl(a):
 
 sdc_overload(nanmin)(nan_min_max_overload_factory(min))
 sdc_overload(nanmax)(nan_min_max_overload_factory(max))
+
+
+def nanprod(a):
+    pass
+
+
+@sdc_overload(nanprod)
+def np_nanprod(a):
+    """
+    Reimplemented with parfor from numba.targets.arraymath.
+    """
+    if not isinstance(a, types.Array):
+        return
+    if isinstance(a.dtype, types.Integer):
+        retty = types.intp
+    else:
+        retty = a.dtype
+    one = retty(1)
+    isnan = get_isnan(a.dtype)
+
+    def nanprod_impl(a):
+        c = one
+        for i in prange(len(a)):
+            v = a[i]
+            if not isnan(v):
+                c *= v
+        return c
+
+    return nanprod_impl

sdc/tests/test_sdc_numpy.py

@@ -240,6 +240,7 @@ def sdc_impl():
         sdc_func = self.jit(sdc_impl)
         np.testing.assert_array_equal(sdc_func(), ref_impl())
 
+
 class TestArrayReductions(TestCase):
 
     def check_reduction_basic(self, pyfunc, alt_pyfunc, all_nans=True):
@@ -281,6 +282,15 @@ def sdc_impl(a):
 
         self.check_reduction_basic(ref_impl, sdc_impl)
 
+    def test_nanprod(self):
+        def ref_impl(a):
+            return np.nanprod(a)
+
+        def sdc_impl(a):
+            return numpy_like.nanprod(a)
+
+        self.check_reduction_basic(ref_impl, sdc_impl)
+
     def test_nansum(self):
         def ref_impl(a):
             return np.nansum(a)

sdc/tests/tests_perf/test_perf_numpy.py

@@ -113,6 +113,11 @@ def _test_case(self, cases, name, total_data_length, data_num=1, input_data=test
         CE(type_='Python', code='np.nansum(data)', jitted=False),
         CE(type_='SDC', code='sdc.functions.numpy_like.nansum(data)', jitted=True),
     ], usecase_params='data'),
+    TC(name='nanprod', size=[10 ** 7], call_expr=[
+        CE(type_='Python', code='np.nanprod(data)', jitted=False),
+        CE(type_='Numba', code='np.nanprod(data)', jitted=True),
+        CE(type_='SDC', code='sdc.functions.numpy_like.nanprod(data)', jitted=True),
+    ], usecase_params='data'),
     TC(name='sum', size=[10 ** 7], call_expr=[
         CE(type_='Python', code='np.sum(data)', jitted=False),
         CE(type_='Numba', code='np.sum(data)', jitted=True),

sdc/tests/tests_perf/test_perf_series.py

@@ -115,7 +115,8 @@ def _test_case(self, pyfunc, name, total_data_length, data_num=1, input_data=tes
     TC(name='notna', size=[10 ** 7]),
     TC(name='nsmallest', size=[10 ** 6]),
     TC(name='nunique', size=[10 ** 7]),
-    TC(name='prod', size=[10 ** 8]),
+    TC(name='prod', size=[10 ** 8], params='skipna=True'),
+    TC(name='prod', size=[10 ** 8], params='skipna=False'),
     TC(name='pct_change', size=[10 ** 7], params='periods=1, limit=None, freq=None'),
     TC(name='pow', size=[10 ** 7], params='other', data_num=2),
     TC(name='quantile', size=[10 ** 8]),