ENH: Get rid of float cast in masked reduction ops (#50833)

Author: phofl

pandas/core/arrays/masked.py

@@ -1081,12 +1081,7 @@ def _reduce(self, name: str, *, skipna: bool = True, **kwargs):
         data = self._data
         mask = self._mask
 
-        # coerce to a nan-aware float if needed
-        # (we explicitly use NaN within reductions)
-        if self._hasna:
-            data = self.to_numpy("float64", na_value=np.nan)
-
-        # median, skew, kurt, idxmin, idxmax
+        # median, skew, kurt, sem
         op = getattr(nanops, f"nan{name}")
         result = op(data, axis=0, skipna=skipna, mask=mask, **kwargs)
 

pandas/core/nanops.py

@@ -761,15 +761,15 @@ def get_median(x, _mask=None):
             res = np.nanmedian(x[_mask])
         return res
 
-    values, mask, dtype, _, _ = _get_values(values, skipna, mask=mask)
+    values, mask, dtype, _, _ = _get_values(values, skipna, mask=mask, fill_value=0)
     if not is_float_dtype(values.dtype):
         try:
             values = values.astype("f8")
         except ValueError as err:
             # e.g. "could not convert string to float: 'a'"
             raise TypeError(str(err)) from err
-        if mask is not None:
-            values[mask] = np.nan
+    if mask is not None:
+        values[mask] = np.nan
 
     notempty = values.size
 
@@ -1043,8 +1043,11 @@ def nansem(
     if not is_float_dtype(values.dtype):
         values = values.astype("f8")
 
+    if not skipna and mask is not None and mask.any():
+        return np.nan
+
     count, _ = _get_counts_nanvar(values.shape, mask, axis, ddof, values.dtype)
-    var = nanvar(values, axis=axis, skipna=skipna, ddof=ddof)
+    var = nanvar(values, axis=axis, skipna=skipna, ddof=ddof, mask=mask)
 
     return np.sqrt(var) / np.sqrt(count)
 
@@ -1225,6 +1228,8 @@ def nanskew(
     if skipna and mask is not None:
         values = values.copy()
         np.putmask(values, mask, 0)
+    elif not skipna and mask is not None and mask.any():
+        return np.nan
 
     mean = values.sum(axis, dtype=np.float64) / count
     if axis is not None:
@@ -1313,6 +1318,8 @@ def nankurt(
     if skipna and mask is not None:
         values = values.copy()
         np.putmask(values, mask, 0)
+    elif not skipna and mask is not None and mask.any():
+        return np.nan
 
     mean = values.sum(axis, dtype=np.float64) / count
     if axis is not None: