Scale var/std

sdc/datatypes/hpat_pandas_series_functions.py

@@ -1497,7 +1497,7 @@ def hpat_pandas_series_var_impl(self, axis=None, skipna=None, level=None, ddof=1
             if valuable_length <= ddof:
                 return numpy.nan
 
-            return numpy.nanvar(self._data) * valuable_length / (valuable_length - ddof)
+            return numpy_like.nanvar(self._data) * valuable_length / (valuable_length - ddof)
 
         if len(self._data) <= ddof:
             return numpy.nan

sdc/functions/numpy_like.py

@@ -497,3 +497,32 @@ def nanmean_impl(a):
         return np.divide(c, count)
 
     return nanmean_impl
+
+
+def nanvar(a):
+    pass
+
+
+@sdc_overload(nanvar)
+def np_nanvar(a):
+    if not isinstance(a, types.Array):
+        return
+    isnan = get_isnan(a.dtype)
+
+    def nanvar_impl(a):
+        # Compute the mean
+        m = nanmean(a)
+
+        # Compute the sum of square diffs
+        ssd = 0.0
+        count = 0
+        for i in prange(len(a)):
+            v = a[i]
+            if not isnan(v):
+                val = (v.item() - m)
+                ssd += np.real(val * np.conj(val))
+                count += 1
+        # np.divide() doesn't raise ZeroDivisionError
+        return np.divide(ssd, count)
+
+    return nanvar_impl

sdc/tests/test_sdc_numpy.py

@@ -243,7 +243,10 @@ def sdc_impl():
 
 class TestArrayReductions(TestCase):
 
-    def check_reduction_basic(self, pyfunc, alt_pyfunc, all_nans=True):
+    def check_reduction_basic(self, pyfunc, alt_pyfunc, all_nans=True, comparator=None):
+        if not comparator:
+            comparator = np.testing.assert_array_equal
+
         alt_cfunc = self.jit(alt_pyfunc)
 
         def cases():
@@ -262,7 +265,7 @@ def cases():
 
         for case in cases():
             with self.subTest(data=case):
-                np.testing.assert_array_equal(alt_cfunc(case), pyfunc(case))
+                comparator(alt_cfunc(case), pyfunc(case))
 
     def test_nanmean(self):
         def ref_impl(a):
@@ -309,6 +312,16 @@ def sdc_impl(a):
 
         self.check_reduction_basic(ref_impl, sdc_impl)
 
+    def test_nanvar(self):
+        def ref_impl(a):
+            return np.nanvar(a)
+
+        def sdc_impl(a):
+            return numpy_like.nanvar(a)
+
+        self.check_reduction_basic(ref_impl, sdc_impl,
+                                   comparator=np.testing.assert_array_almost_equal)
+
     def test_sum(self):
         def ref_impl(a):
             return np.sum(a)

sdc/tests/tests_perf/test_perf_numpy.py

@@ -109,6 +109,11 @@ def _test_case(self, cases, name, total_data_length, data_num=1, input_data=test
         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='nanvar', size=[10 ** 7], call_expr=[
+        CE(type_='Python', code='np.nanvar(data)', jitted=False),
+        CE(type_='Numba', code='np.nanvar(data)', jitted=True),
+        CE(type_='SDC', code='sdc.functions.numpy_like.nanvar(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

@@ -127,14 +127,16 @@ def _test_case(self, pyfunc, name, total_data_length, data_num=1, input_data=tes
     TC(name='shift', size=[10 ** 8]),
     TC(name='size', size=[10 ** 7], call_expr='data.size', usecase_params='data'),
     TC(name='sort_values', size=[10 ** 5]),
-    TC(name='std', size=[10 ** 7]),
+    TC(name='std', size=[10 ** 7], params='skipna=True'),
+    TC(name='std', size=[10 ** 7], params='skipna=False'),
     TC(name='sub', size=[10 ** 7], params='other', data_num=2),
     TC(name='sum', size=[10 ** 8]),
     TC(name='take', size=[10 ** 7], call_expr='data.take([0])', usecase_params='data'),
     TC(name='truediv', size=[10 ** 7], params='other', data_num=2),
     TC(name='values', size=[10 ** 7], call_expr='data.values', usecase_params='data'),
     TC(name='value_counts', size=[10 ** 6]),
-    TC(name='var', size=[10 ** 8]),
+    TC(name='var', size=[10 ** 8], params='skipna=True'),
+    TC(name='var', size=[10 ** 8], params='skipna=False'),
     TC(name='unique', size=[10 ** 5]),
 ]