fix(rust, python): treat null columns as zero in sum_horizontal (#1…

…3880)

crates/polars-core/src/frame/mod.rs

@@ -2464,32 +2464,47 @@ impl DataFrame {
 
     /// Aggregate the column horizontally to their sum values.
     pub fn sum_horizontal(&self, null_strategy: NullStrategy) -> PolarsResult<Option<Series>> {
-        let sum_fn =
-            |acc: &Series, s: &Series, null_strategy: NullStrategy| -> PolarsResult<Series> {
-                let mut acc = acc.clone();
-                let mut s = s.clone();
+        let apply_null_strategy =
+            |s: &Series, null_strategy: NullStrategy| -> PolarsResult<Series> {
                 if let NullStrategy::Ignore = null_strategy {
                     // if has nulls
-                    if acc.has_validity() {
-                        acc = acc.fill_null(FillNullStrategy::Zero)?;
-                    }
                     if s.has_validity() {
-                        s = s.fill_null(FillNullStrategy::Zero)?;
+                        return s.fill_null(FillNullStrategy::Zero);
                     }
                 }
+                Ok(s.clone())
+            };
+
+        let sum_fn =
+            |acc: &Series, s: &Series, null_strategy: NullStrategy| -> PolarsResult<Series> {
+                let acc: Series = apply_null_strategy(acc, null_strategy)?;
+                let s = apply_null_strategy(s, null_strategy)?;
                 Ok(&acc + &s)
             };
 
-        match self.columns.len() {
-            0 => Ok(None),
-            1 => Ok(Some(self.columns[0].clone())),
-            2 => sum_fn(&self.columns[0], &self.columns[1], null_strategy).map(Some),
+        let non_null_cols = self
+            .columns
+            .iter()
+            .filter(|x| x.dtype() != &DataType::Null)
+            .collect::<Vec<_>>();
+
+        match non_null_cols.len() {
+            0 => {
+                if self.columns.is_empty() {
+                    Ok(None)
+                } else {
+                    // all columns are null dtype, so result is null dtype
+                    Ok(Some(self.columns[0].clone()))
+                }
+            },
+            1 => Ok(Some(apply_null_strategy(non_null_cols[0], null_strategy)?)),
+            2 => sum_fn(non_null_cols[0], non_null_cols[1], null_strategy).map(Some),
             _ => {
                 // the try_reduce_with is a bit slower in parallelism,
                 // but I don't think it matters here as we parallelize over columns, not over elements
                 POOL.install(|| {
-                    self.columns
-                        .par_iter()
+                    non_null_cols
+                        .into_par_iter()
                         .map(|s| Ok(Cow::Borrowed(s)))
                         .try_reduce_with(|l, r| sum_fn(&l, &r, null_strategy).map(Cow::Owned))
                         // we can unwrap the option, because we are certain there is a column

py-polars/tests/unit/functions/aggregation/test_horizontal.py

@@ -248,6 +248,41 @@ def test_str_sum_horizontal() -> None:
     assert_series_equal(out["A"], pl.Series("A", ["af", "bg", "h", "c", ""]))
 
 
+def test_sum_null_dtype() -> None:
+    df = pl.DataFrame(
+        {
+            "A": [5, None, 3, 2, 1],
+            "B": [5, 3, None, 2, 1],
+            "C": [None, None, None, None, None],
+        }
+    )
+
+    assert_series_equal(
+        df.select(pl.sum_horizontal("A", "B", "C")).to_series(),
+        pl.Series("A", [10, 3, 3, 4, 2]),
+    )
+    assert_series_equal(
+        df.select(pl.sum_horizontal("C", "B")).to_series(),
+        pl.Series("C", [5, 3, 0, 2, 1]),
+    )
+    assert_series_equal(
+        df.select(pl.sum_horizontal("C", "C")).to_series(),
+        pl.Series("C", [None, None, None, None, None]),
+    )
+
+
+def test_sum_single_col() -> None:
+    df = pl.DataFrame(
+        {
+            "A": [5, None, 3, None, 1],
+        }
+    )
+
+    assert_series_equal(
+        df.select(pl.sum_horizontal("A")).to_series(), pl.Series("A", [5, 0, 3, 0, 1])
+    )
+
+
 def test_cum_sum_horizontal() -> None:
     df = pl.DataFrame(
         {