cast (u)int8 (u)int16 -> i64 in sum/cumsum/cumprod operations

polars/polars-core/src/series/implementations/boolean.rs

@@ -383,7 +383,7 @@ impl SeriesTrait for SeriesWrap<BooleanChunked> {
         ChunkFillNull::fill_null(&self.0, strategy).map(|ca| ca.into_series())
     }
 
-    fn sum_as_series(&self) -> Series {
+    fn _sum_as_series(&self) -> Series {
         ChunkAggSeries::sum_as_series(&self.0)
     }
     fn max_as_series(&self) -> Series {

polars/polars-core/src/series/implementations/categorical.rs

@@ -350,7 +350,7 @@ impl SeriesTrait for SeriesWrap<CategoricalChunked> {
         ChunkFillNull::fill_null(&self.0, strategy).map(|ca| ca.into_series())
     }
 
-    fn sum_as_series(&self) -> Series {
+    fn _sum_as_series(&self) -> Series {
         CategoricalChunked::full_null(self.name(), 1).into_series()
     }
     fn max_as_series(&self) -> Series {

polars/polars-core/src/series/implementations/dates_time.rs

@@ -542,7 +542,7 @@ macro_rules! impl_dyn_series {
                     .map(|ca| ca.$into_logical().into_series())
             }
 
-            fn sum_as_series(&self) -> Series {
+            fn _sum_as_series(&self) -> Series {
                 Int32Chunked::full_null(self.name(), 1)
                     .cast(self.dtype())
                     .unwrap()

polars/polars-core/src/series/implementations/floats.rs

@@ -523,7 +523,7 @@ macro_rules! impl_dyn_series {
                 ChunkFillNull::fill_null(&self.0, strategy).map(|ca| ca.into_series())
             }
 
-            fn sum_as_series(&self) -> Series {
+            fn _sum_as_series(&self) -> Series {
                 ChunkAggSeries::sum_as_series(&self.0)
             }
             fn max_as_series(&self) -> Series {

polars/polars-core/src/series/implementations/list.rs

@@ -272,7 +272,7 @@ impl SeriesTrait for SeriesWrap<ListChunked> {
         ChunkFillNull::fill_null(&self.0, strategy).map(|ca| ca.into_series())
     }
 
-    fn sum_as_series(&self) -> Series {
+    fn _sum_as_series(&self) -> Series {
         ChunkAggSeries::sum_as_series(&self.0)
     }
     fn max_as_series(&self) -> Series {

polars/polars-core/src/series/implementations/mod.rs

@@ -181,7 +181,11 @@ macro_rules! impl_dyn_series {
             }
 
             fn agg_sum(&self, groups: &[(u32, Vec<u32>)]) -> Option<Series> {
-                self.0.agg_sum(groups)
+                use DataType::*;
+                match self.dtype() {
+                    Int8 | UInt8 | Int16 | UInt16 => self.cast(&Int64).unwrap().agg_sum(groups),
+                    _ => self.0.agg_sum(groups),
+                }
             }
 
             fn agg_first(&self, groups: &[(u32, Vec<u32>)]) -> Series {
@@ -684,7 +688,7 @@ macro_rules! impl_dyn_series {
                 ChunkFillNull::fill_null(&self.0, strategy).map(|ca| ca.into_series())
             }
 
-            fn sum_as_series(&self) -> Series {
+            fn _sum_as_series(&self) -> Series {
                 ChunkAggSeries::sum_as_series(&self.0)
             }
             fn max_as_series(&self) -> Series {

polars/polars-core/src/series/implementations/object.rs

@@ -261,7 +261,7 @@ where
         &self.0
     }
 
-    fn sum_as_series(&self) -> Series {
+    fn _sum_as_series(&self) -> Series {
         ObjectChunked::<T>::full_null(self.name(), 1).into_series()
     }
     fn max_as_series(&self) -> Series {

polars/polars-core/src/series/implementations/utf8.rs

@@ -370,7 +370,7 @@ impl SeriesTrait for SeriesWrap<Utf8Chunked> {
         ChunkFillNull::fill_null(&self.0, strategy).map(|ca| ca.into_series())
     }
 
-    fn sum_as_series(&self) -> Series {
+    fn _sum_as_series(&self) -> Series {
         ChunkAggSeries::sum_as_series(&self.0)
     }
     fn max_as_series(&self) -> Series {

polars/polars-core/src/series/mod.rs

@@ -198,10 +198,16 @@ impl Series {
     pub fn cast(&self, dtype: &DataType) -> Result<Self> {
         self.0.cast(dtype)
     }
+
+    /// Compute the sum of all values in this Series.
     /// Returns `None` if the array is empty or only contains null values.
+    ///
+    /// If the [`DataType`] is one of `{Int8, UInt8, Int16, UInt16}` the `Series` is
+    /// first cast to `Int64` to prevent overflow issues.
+    ///
     /// ```
     /// # use polars_core::prelude::*;
-    /// let s = Series::new("days", [1, 2, 3].as_ref());
+    /// let s = Series::new("days", &[1, 2, 3]);
     /// assert_eq!(s.sum(), Some(6));
     /// ```
     pub fn sum<T>(&self) -> Option<T>
@@ -482,6 +488,18 @@ impl Series {
         UInt64Chunked::new_from_aligned_vec(self.name(), self.0.vec_hash(build_hasher))
     }
 
+    /// Get the sum of the Series as a new Series of length 1.
+    ///
+    /// If the [`DataType`] is one of `{Int8, UInt8, Int16, UInt16}` the `Series` is
+    /// first cast to `Int64` to prevent overflow issues.
+    pub fn sum_as_series(&self) -> Series {
+        use DataType::*;
+        match self.dtype() {
+            Int8 | UInt8 | Int16 | UInt16 => self.cast(&Int64).unwrap().sum_as_series(),
+            _ => self._sum_as_series(),
+        }
+    }
+
     /// Get an array with the cumulative max computed at every element
     #[cfg_attr(docsrs, doc(cfg(feature = "cum_agg")))]
     pub fn cummax(&self, _reverse: bool) -> Series {
@@ -509,12 +527,17 @@ impl Series {
     }
 
     /// Get an array with the cumulative sum computed at every element
+    ///
+    /// If the [`DataType`] is one of `{Int8, UInt8, Int16, UInt16}` the `Series` is
+    /// first cast to `Int64` to prevent overflow issues.
     #[cfg_attr(docsrs, doc(cfg(feature = "cum_agg")))]
     pub fn cumsum(&self, _reverse: bool) -> Series {
         #[cfg(feature = "cum_agg")]
         {
+            use DataType::*;
             match self.dtype() {
-                DataType::Boolean => self.cast(&DataType::UInt32).unwrap()._cumsum(_reverse),
+                Boolean => self.cast(&DataType::UInt32).unwrap()._cumsum(_reverse),
+                Int8 | UInt8 | Int16 | UInt16 => self.cast(&Int64).unwrap()._cumsum(_reverse),
                 _ => self._cumsum(_reverse),
             }
         }
@@ -525,12 +548,17 @@ impl Series {
     }
 
     /// Get an array with the cumulative product computed at every element
+    ///
+    /// If the [`DataType`] is one of `{Int8, UInt8, Int16, UInt16}` the `Series` is
+    /// first cast to `Int64` to prevent overflow issues.
     #[cfg_attr(docsrs, doc(cfg(feature = "cum_agg")))]
     pub fn cumprod(&self, _reverse: bool) -> Series {
         #[cfg(feature = "cum_agg")]
         {
+            use DataType::*;
             match self.dtype() {
-                DataType::Boolean => self.cast(&DataType::UInt32).unwrap()._cumprod(_reverse),
+                Boolean => self.cast(&UInt32).unwrap()._cumprod(_reverse),
+                Int8 | UInt8 | Int16 | UInt16 => self.cast(&Int64).unwrap()._cumprod(_reverse),
                 _ => self._cumprod(_reverse),
             }
         }

polars/polars-core/src/series/series_trait.rs

@@ -165,6 +165,8 @@ pub(crate) mod private {
         fn agg_max(&self, _groups: &[(u32, Vec<u32>)]) -> Option<Series> {
             None
         }
+        /// If the [`DataType`] is one of `{Int8, UInt8, Int16, UInt16}` the `Series` is
+        /// first cast to `Int64` to prevent overflow issues.
         fn agg_sum(&self, _groups: &[(u32, Vec<u32>)]) -> Option<Series> {
             None
         }
@@ -818,7 +820,10 @@ pub trait SeriesTrait:
     }
 
     /// Get the sum of the Series as a new Series of length 1.
-    fn sum_as_series(&self) -> Series {
+    ///
+    /// If the [`DataType`] is one of `{Int8, UInt8, Int16, UInt16}` the `Series` is
+    /// first cast to `Int64` to prevent overflow issues.
+    fn _sum_as_series(&self) -> Series {
         invalid_operation_panic!(self)
     }
     /// Get the max of the Series as a new Series of length 1.

polars/polars-lazy/src/tests/queries.rs

@@ -1779,7 +1779,7 @@ fn test_groupby_small_ints() -> Result<()> {
         .sort("foo", true)
         .collect()?;
 
-    assert_eq!(Vec::from(out.column("foo")?.i16()?), &[Some(2), Some(1)]);
+    assert_eq!(Vec::from(out.column("foo")?.i64()?), &[Some(2), Some(1)]);
     Ok(())
 }
 

py-polars/polars/internals/expr.py

@@ -893,6 +893,9 @@ def min(self) -> "Expr":
     def sum(self) -> "Expr":
         """
         Get sum value.
+
+        Note that dtypes in {Int8, UInt8, Int16, UInt16} are cast to
+        Int64 before summing to prevent overflow issues.
         """
         return wrap_expr(self._pyexpr.sum())
 

py-polars/polars/internals/series.py

@@ -700,6 +700,9 @@ def sum(self) -> Union[int, float]:
         """
         Reduce this Series to the sum value.
 
+        Note that dtypes in {Int8, UInt8, Int16, UInt16} are cast to
+        Int64 before summing to prevent overflow issues.
+
         Examples
         --------
         >>> s = pl.Series("a", [1, 2, 3])