Jefffrey commented on code in PR #22459: URL: https://github.com/apache/datafusion/pull/22459#discussion_r3291883853
########## datafusion/functions-nested/src/array_add.rs: ########## @@ -0,0 +1,274 @@ +// Licensed to the Apache Software Foundation (ASF) under one +// or more contributor license agreements. See the NOTICE file +// distributed with this work for additional information +// regarding copyright ownership. The ASF licenses this file +// to you under the Apache License, Version 2.0 (the +// "License"); you may not use this file except in compliance +// with the License. You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, +// software distributed under the License is distributed on an +// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +// KIND, either express or implied. See the License for the +// specific language governing permissions and limitations +// under the License. + +//! [`ScalarUDFImpl`] definitions for array_add function. + +use crate::utils::make_scalar_function; +use arrow::array::{ + Array, ArrayRef, Float64Array, GenericListArray, NullBufferBuilder, + OffsetBufferBuilder, OffsetSizeTrait, +}; +use arrow::buffer::NullBuffer; +use arrow::datatypes::{ + DataType, + DataType::{FixedSizeList, LargeList, List, Null}, + Field, +}; +use datafusion_common::cast::{as_float64_array, as_generic_list_array}; +use datafusion_common::utils::{ListCoercion, coerced_type_with_base_type_only}; +use datafusion_common::{ + Result, exec_err, not_impl_err, plan_err, utils::take_function_args, +}; +use datafusion_expr::{ + ColumnarValue, Documentation, ScalarFunctionArgs, ScalarUDFImpl, Signature, + Volatility, +}; +use datafusion_macros::user_doc; +use std::sync::Arc; + +make_udf_expr_and_func!( + ArrayAdd, + array_add, + array1 array2, + "returns the element-wise sum of two numeric arrays.", + array_add_udf +); + +#[user_doc( + doc_section(label = "Array Functions"), + description = "Returns the element-wise sum of two numeric arrays of equal length, computed as `array1[i] + array2[i]` per position. NULL is propagated per element: if either input element at position `i` is NULL, the corresponding output element is NULL (positions are preserved). Returns NULL if either entire input array is NULL. Errors if the per-row lengths differ. Returns an empty array if both inputs are empty.", + syntax_example = "array_add(array1, array2)", + sql_example = r#"```sql +> select array_add([1.0, 2.0, 3.0], [10.0, 20.0, 30.0]); ++---------------------------------------------------------+ +| array_add(List([1.0,2.0,3.0]),List([10.0,20.0,30.0])) | ++---------------------------------------------------------+ +| [11.0, 22.0, 33.0] | ++---------------------------------------------------------+ +```"#, + argument( + name = "array1", + description = "Array expression. Can be a constant, column, or function, and any combination of array operators." + ), + argument( + name = "array2", + description = "Array expression. Can be a constant, column, or function, and any combination of array operators." + ) +)] +#[derive(Debug, PartialEq, Eq, Hash)] +pub struct ArrayAdd { + signature: Signature, + aliases: Vec<String>, +} + +impl Default for ArrayAdd { + fn default() -> Self { + Self::new() + } +} + +impl ArrayAdd { + pub fn new() -> Self { + Self { + signature: Signature::user_defined(Volatility::Immutable), + aliases: vec!["list_add".to_string()], + } + } +} + +impl ScalarUDFImpl for ArrayAdd { + fn name(&self) -> &str { + "array_add" + } + + fn signature(&self) -> &Signature { + &self.signature + } + + fn return_type(&self, arg_types: &[DataType]) -> Result<DataType> { + // After `coerce_types`, both args share the same List/LargeList<Float64> shape. + Ok(arg_types[0].clone()) + } + + fn coerce_types(&self, arg_types: &[DataType]) -> Result<Vec<DataType>> { + let [_, _] = take_function_args(self.name(), arg_types)?; + let coercion = Some(&ListCoercion::FixedSizedListToList); + + for arg_type in arg_types { + if !matches!(arg_type, Null | List(_) | LargeList(_) | FixedSizeList(..)) { + return plan_err!("{} does not support type {arg_type}", self.name()); + } + // Only flat lists of numeric leaves are supported. Peel exactly + // one layer so that `List<List<_>>` is rejected as non-numeric + // rather than passing through and failing opaquely in the kernel. + let element_type = match arg_type { + List(field) | LargeList(field) | FixedSizeList(field, _) => { + field.data_type() + } + other => other, + }; + if matches!( + element_type, + DataType::Decimal128(_, _) | DataType::Decimal256(_, _) + ) { + return not_impl_err!( + "{} does not yet support decimal element types ({element_type}); \ + cast to DOUBLE explicitly to opt into lossy float arithmetic", + self.name() + ); + } + if !matches!(element_type, Null) && !element_type.is_numeric() { + return plan_err!( + "{} requires numeric array elements, got list of {element_type}", + self.name() + ); + } + } + + // If either side is `LargeList`, widen both to `LargeList` so the runtime + // dispatch sees a homogeneous pair. + let any_large_list = arg_types.iter().any(|t| matches!(t, LargeList(_))); + + let coerced = arg_types + .iter() + .map(|arg_type| { + if matches!(arg_type, Null) { + let field = Arc::new(Field::new_list_field(DataType::Float64, true)); + return if any_large_list { + LargeList(field) + } else { + List(field) + }; + } + let coerced = coerced_type_with_base_type_only( + arg_type, + &DataType::Float64, + coercion, + ); + match coerced { + List(field) if any_large_list => LargeList(field), + other => other, + } + }) + .collect(); + + Ok(coerced) + } + + fn invoke_with_args(&self, args: ScalarFunctionArgs) -> Result<ColumnarValue> { + make_scalar_function(array_add_inner)(&args.args) + } + + fn aliases(&self) -> &[String] { + &self.aliases + } + + fn documentation(&self) -> Option<&Documentation> { + self.doc() + } +} + +fn array_add_inner(args: &[ArrayRef]) -> Result<ArrayRef> { + let [array1, array2] = take_function_args("array_add", args)?; + match (array1.data_type(), array2.data_type()) { + (List(_), List(_)) => general_array_add::<i32>(args), + (LargeList(_), LargeList(_)) => general_array_add::<i64>(args), + (arg_type1, arg_type2) => exec_err!( + "array_add received unexpected types after coercion: {arg_type1} and {arg_type2}" + ), + } +} + +fn general_array_add<O: OffsetSizeTrait>(arrays: &[ArrayRef]) -> Result<ArrayRef> { + let lhs = as_generic_list_array::<O>(&arrays[0])?; + let rhs = as_generic_list_array::<O>(&arrays[1])?; + + if lhs.len() != rhs.len() { Review Comment: We don't need to check this, this is already a guarantee ########## datafusion/functions-nested/src/array_add.rs: ########## @@ -0,0 +1,274 @@ +// Licensed to the Apache Software Foundation (ASF) under one +// or more contributor license agreements. See the NOTICE file +// distributed with this work for additional information +// regarding copyright ownership. The ASF licenses this file +// to you under the Apache License, Version 2.0 (the +// "License"); you may not use this file except in compliance +// with the License. You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, +// software distributed under the License is distributed on an +// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +// KIND, either express or implied. See the License for the +// specific language governing permissions and limitations +// under the License. + +//! [`ScalarUDFImpl`] definitions for array_add function. + +use crate::utils::make_scalar_function; +use arrow::array::{ + Array, ArrayRef, Float64Array, GenericListArray, NullBufferBuilder, + OffsetBufferBuilder, OffsetSizeTrait, +}; +use arrow::buffer::NullBuffer; +use arrow::datatypes::{ + DataType, + DataType::{FixedSizeList, LargeList, List, Null}, + Field, +}; +use datafusion_common::cast::{as_float64_array, as_generic_list_array}; +use datafusion_common::utils::{ListCoercion, coerced_type_with_base_type_only}; +use datafusion_common::{ + Result, exec_err, not_impl_err, plan_err, utils::take_function_args, +}; +use datafusion_expr::{ + ColumnarValue, Documentation, ScalarFunctionArgs, ScalarUDFImpl, Signature, + Volatility, +}; +use datafusion_macros::user_doc; +use std::sync::Arc; + +make_udf_expr_and_func!( + ArrayAdd, + array_add, + array1 array2, + "returns the element-wise sum of two numeric arrays.", + array_add_udf +); + +#[user_doc( + doc_section(label = "Array Functions"), + description = "Returns the element-wise sum of two numeric arrays of equal length, computed as `array1[i] + array2[i]` per position. NULL is propagated per element: if either input element at position `i` is NULL, the corresponding output element is NULL (positions are preserved). Returns NULL if either entire input array is NULL. Errors if the per-row lengths differ. Returns an empty array if both inputs are empty.", + syntax_example = "array_add(array1, array2)", + sql_example = r#"```sql +> select array_add([1.0, 2.0, 3.0], [10.0, 20.0, 30.0]); ++---------------------------------------------------------+ +| array_add(List([1.0,2.0,3.0]),List([10.0,20.0,30.0])) | ++---------------------------------------------------------+ +| [11.0, 22.0, 33.0] | ++---------------------------------------------------------+ +```"#, + argument( + name = "array1", + description = "Array expression. Can be a constant, column, or function, and any combination of array operators." + ), + argument( + name = "array2", + description = "Array expression. Can be a constant, column, or function, and any combination of array operators." + ) +)] +#[derive(Debug, PartialEq, Eq, Hash)] +pub struct ArrayAdd { + signature: Signature, + aliases: Vec<String>, +} + +impl Default for ArrayAdd { + fn default() -> Self { + Self::new() + } +} + +impl ArrayAdd { + pub fn new() -> Self { + Self { + signature: Signature::user_defined(Volatility::Immutable), + aliases: vec!["list_add".to_string()], + } + } +} + +impl ScalarUDFImpl for ArrayAdd { + fn name(&self) -> &str { + "array_add" + } + + fn signature(&self) -> &Signature { + &self.signature + } + + fn return_type(&self, arg_types: &[DataType]) -> Result<DataType> { + // After `coerce_types`, both args share the same List/LargeList<Float64> shape. + Ok(arg_types[0].clone()) + } + + fn coerce_types(&self, arg_types: &[DataType]) -> Result<Vec<DataType>> { + let [_, _] = take_function_args(self.name(), arg_types)?; + let coercion = Some(&ListCoercion::FixedSizedListToList); + + for arg_type in arg_types { + if !matches!(arg_type, Null | List(_) | LargeList(_) | FixedSizeList(..)) { + return plan_err!("{} does not support type {arg_type}", self.name()); + } + // Only flat lists of numeric leaves are supported. Peel exactly + // one layer so that `List<List<_>>` is rejected as non-numeric + // rather than passing through and failing opaquely in the kernel. + let element_type = match arg_type { + List(field) | LargeList(field) | FixedSizeList(field, _) => { + field.data_type() + } + other => other, + }; + if matches!( + element_type, + DataType::Decimal128(_, _) | DataType::Decimal256(_, _) + ) { + return not_impl_err!( + "{} does not yet support decimal element types ({element_type}); \ + cast to DOUBLE explicitly to opt into lossy float arithmetic", + self.name() + ); + } + if !matches!(element_type, Null) && !element_type.is_numeric() { + return plan_err!( + "{} requires numeric array elements, got list of {element_type}", + self.name() + ); + } + } + + // If either side is `LargeList`, widen both to `LargeList` so the runtime + // dispatch sees a homogeneous pair. + let any_large_list = arg_types.iter().any(|t| matches!(t, LargeList(_))); + + let coerced = arg_types + .iter() + .map(|arg_type| { + if matches!(arg_type, Null) { + let field = Arc::new(Field::new_list_field(DataType::Float64, true)); + return if any_large_list { + LargeList(field) + } else { + List(field) + }; + } + let coerced = coerced_type_with_base_type_only( + arg_type, + &DataType::Float64, + coercion, + ); + match coerced { + List(field) if any_large_list => LargeList(field), + other => other, + } + }) + .collect(); + + Ok(coerced) + } + + fn invoke_with_args(&self, args: ScalarFunctionArgs) -> Result<ColumnarValue> { + make_scalar_function(array_add_inner)(&args.args) + } + + fn aliases(&self) -> &[String] { + &self.aliases + } + + fn documentation(&self) -> Option<&Documentation> { + self.doc() + } +} + +fn array_add_inner(args: &[ArrayRef]) -> Result<ArrayRef> { + let [array1, array2] = take_function_args("array_add", args)?; + match (array1.data_type(), array2.data_type()) { + (List(_), List(_)) => general_array_add::<i32>(args), + (LargeList(_), LargeList(_)) => general_array_add::<i64>(args), + (arg_type1, arg_type2) => exec_err!( + "array_add received unexpected types after coercion: {arg_type1} and {arg_type2}" + ), + } +} + +fn general_array_add<O: OffsetSizeTrait>(arrays: &[ArrayRef]) -> Result<ArrayRef> { + let lhs = as_generic_list_array::<O>(&arrays[0])?; + let rhs = as_generic_list_array::<O>(&arrays[1])?; + + if lhs.len() != rhs.len() { + return exec_err!( + "array_add row counts differ ({} vs {})", + lhs.len(), + rhs.len() + ); + } + + let lhs_values = as_float64_array(lhs.values())?; + let rhs_values = as_float64_array(rhs.values())?; + let lhs_offsets = lhs.value_offsets(); + let rhs_offsets = rhs.value_offsets(); + + let mut out_values: Vec<f64> = Vec::with_capacity(lhs_values.len()); + let mut out_inner_nulls = NullBufferBuilder::new(lhs_values.len()); + let mut out_offsets = OffsetBufferBuilder::<O>::new(lhs.len()); + let mut out_row_nulls = NullBufferBuilder::new(lhs.len()); + + for row in 0..lhs.len() { + // Whole-row NULL on either side -> NULL output row, no elements. + if lhs.is_null(row) || rhs.is_null(row) { + out_row_nulls.append_null(); + out_offsets.push_length(0); + continue; + } + + let start1 = lhs_offsets[row].as_usize(); + let len1 = lhs.value_length(row).as_usize(); + let start2 = rhs_offsets[row].as_usize(); + let len2 = rhs.value_length(row).as_usize(); + + if len1 != len2 { + return exec_err!( + "array_add requires both list inputs to have the same length per row, got {len1} and {len2} at row {row}" + ); + } + + let l_slice = lhs_values.slice(start1, len1); + let r_slice = rhs_values.slice(start2, len2); + + let l_vals = l_slice.values(); + let r_vals = r_slice.values(); + + // Per-element validity: position `i` is valid iff both lhs[i] and rhs[i] + // are valid. `NullBuffer::union` returns `None` when both sides are + // entirely valid, keeping the common case branch-free. + let combined = NullBuffer::union(l_slice.nulls(), r_slice.nulls()); + + for i in 0..len1 { + // Push the value unconditionally; the null buffer is the source of + // truth for validity. + out_values.push(l_vals[i] + r_vals[i]); + match &combined { + Some(nb) if nb.is_null(i) => out_inner_nulls.append_null(), + _ => out_inner_nulls.append_non_null(), + } + } + Review Comment: ```suggestion } match &combined { Some(nb) => out_inner_nulls.append_buffer(nb), _ => out_inner_nulls.append_n_non_nulls(len1), }; ``` Can do this once outside the loop ########## datafusion/functions-nested/src/array_add.rs: ########## @@ -0,0 +1,274 @@ +// Licensed to the Apache Software Foundation (ASF) under one +// or more contributor license agreements. See the NOTICE file +// distributed with this work for additional information +// regarding copyright ownership. The ASF licenses this file +// to you under the Apache License, Version 2.0 (the +// "License"); you may not use this file except in compliance +// with the License. You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, +// software distributed under the License is distributed on an +// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +// KIND, either express or implied. See the License for the +// specific language governing permissions and limitations +// under the License. + +//! [`ScalarUDFImpl`] definitions for array_add function. + +use crate::utils::make_scalar_function; +use arrow::array::{ + Array, ArrayRef, Float64Array, GenericListArray, NullBufferBuilder, + OffsetBufferBuilder, OffsetSizeTrait, +}; +use arrow::buffer::NullBuffer; +use arrow::datatypes::{ + DataType, + DataType::{FixedSizeList, LargeList, List, Null}, + Field, +}; +use datafusion_common::cast::{as_float64_array, as_generic_list_array}; +use datafusion_common::utils::{ListCoercion, coerced_type_with_base_type_only}; +use datafusion_common::{ + Result, exec_err, not_impl_err, plan_err, utils::take_function_args, +}; +use datafusion_expr::{ + ColumnarValue, Documentation, ScalarFunctionArgs, ScalarUDFImpl, Signature, + Volatility, +}; +use datafusion_macros::user_doc; +use std::sync::Arc; + +make_udf_expr_and_func!( + ArrayAdd, + array_add, + array1 array2, + "returns the element-wise sum of two numeric arrays.", + array_add_udf +); + +#[user_doc( + doc_section(label = "Array Functions"), + description = "Returns the element-wise sum of two numeric arrays of equal length, computed as `array1[i] + array2[i]` per position. NULL is propagated per element: if either input element at position `i` is NULL, the corresponding output element is NULL (positions are preserved). Returns NULL if either entire input array is NULL. Errors if the per-row lengths differ. Returns an empty array if both inputs are empty.", + syntax_example = "array_add(array1, array2)", + sql_example = r#"```sql +> select array_add([1.0, 2.0, 3.0], [10.0, 20.0, 30.0]); ++---------------------------------------------------------+ +| array_add(List([1.0,2.0,3.0]),List([10.0,20.0,30.0])) | ++---------------------------------------------------------+ +| [11.0, 22.0, 33.0] | ++---------------------------------------------------------+ +```"#, + argument( + name = "array1", + description = "Array expression. Can be a constant, column, or function, and any combination of array operators." + ), + argument( + name = "array2", + description = "Array expression. Can be a constant, column, or function, and any combination of array operators." + ) +)] +#[derive(Debug, PartialEq, Eq, Hash)] +pub struct ArrayAdd { + signature: Signature, + aliases: Vec<String>, +} + +impl Default for ArrayAdd { + fn default() -> Self { + Self::new() + } +} + +impl ArrayAdd { + pub fn new() -> Self { + Self { + signature: Signature::user_defined(Volatility::Immutable), + aliases: vec!["list_add".to_string()], + } + } +} + +impl ScalarUDFImpl for ArrayAdd { + fn name(&self) -> &str { + "array_add" + } + + fn signature(&self) -> &Signature { + &self.signature + } + + fn return_type(&self, arg_types: &[DataType]) -> Result<DataType> { + // After `coerce_types`, both args share the same List/LargeList<Float64> shape. + Ok(arg_types[0].clone()) + } + + fn coerce_types(&self, arg_types: &[DataType]) -> Result<Vec<DataType>> { + let [_, _] = take_function_args(self.name(), arg_types)?; + let coercion = Some(&ListCoercion::FixedSizedListToList); + + for arg_type in arg_types { + if !matches!(arg_type, Null | List(_) | LargeList(_) | FixedSizeList(..)) { + return plan_err!("{} does not support type {arg_type}", self.name()); + } + // Only flat lists of numeric leaves are supported. Peel exactly + // one layer so that `List<List<_>>` is rejected as non-numeric + // rather than passing through and failing opaquely in the kernel. + let element_type = match arg_type { + List(field) | LargeList(field) | FixedSizeList(field, _) => { + field.data_type() + } + other => other, + }; + if matches!( + element_type, + DataType::Decimal128(_, _) | DataType::Decimal256(_, _) + ) { + return not_impl_err!( + "{} does not yet support decimal element types ({element_type}); \ + cast to DOUBLE explicitly to opt into lossy float arithmetic", + self.name() + ); + } + if !matches!(element_type, Null) && !element_type.is_numeric() { + return plan_err!( + "{} requires numeric array elements, got list of {element_type}", + self.name() + ); + } + } + + // If either side is `LargeList`, widen both to `LargeList` so the runtime + // dispatch sees a homogeneous pair. + let any_large_list = arg_types.iter().any(|t| matches!(t, LargeList(_))); + + let coerced = arg_types + .iter() + .map(|arg_type| { + if matches!(arg_type, Null) { + let field = Arc::new(Field::new_list_field(DataType::Float64, true)); + return if any_large_list { + LargeList(field) + } else { + List(field) + }; + } + let coerced = coerced_type_with_base_type_only( + arg_type, + &DataType::Float64, + coercion, + ); + match coerced { + List(field) if any_large_list => LargeList(field), + other => other, + } + }) + .collect(); + + Ok(coerced) + } + + fn invoke_with_args(&self, args: ScalarFunctionArgs) -> Result<ColumnarValue> { + make_scalar_function(array_add_inner)(&args.args) + } + + fn aliases(&self) -> &[String] { + &self.aliases + } + + fn documentation(&self) -> Option<&Documentation> { + self.doc() + } +} + +fn array_add_inner(args: &[ArrayRef]) -> Result<ArrayRef> { + let [array1, array2] = take_function_args("array_add", args)?; + match (array1.data_type(), array2.data_type()) { + (List(_), List(_)) => general_array_add::<i32>(args), + (LargeList(_), LargeList(_)) => general_array_add::<i64>(args), + (arg_type1, arg_type2) => exec_err!( + "array_add received unexpected types after coercion: {arg_type1} and {arg_type2}" + ), + } +} + +fn general_array_add<O: OffsetSizeTrait>(arrays: &[ArrayRef]) -> Result<ArrayRef> { + let lhs = as_generic_list_array::<O>(&arrays[0])?; + let rhs = as_generic_list_array::<O>(&arrays[1])?; + + if lhs.len() != rhs.len() { + return exec_err!( + "array_add row counts differ ({} vs {})", + lhs.len(), + rhs.len() + ); + } + + let lhs_values = as_float64_array(lhs.values())?; + let rhs_values = as_float64_array(rhs.values())?; + let lhs_offsets = lhs.value_offsets(); + let rhs_offsets = rhs.value_offsets(); + + let mut out_values: Vec<f64> = Vec::with_capacity(lhs_values.len()); + let mut out_inner_nulls = NullBufferBuilder::new(lhs_values.len()); + let mut out_offsets = OffsetBufferBuilder::<O>::new(lhs.len()); + let mut out_row_nulls = NullBufferBuilder::new(lhs.len()); Review Comment: Can use [`NullBuffer::union`](https://docs.rs/arrow/latest/arrow/buffer/struct.NullBuffer.html#method.union) to calculate these nulls upfront once (just for the list nulls, not the inner) ########## datafusion/functions-nested/src/array_add.rs: ########## @@ -0,0 +1,274 @@ +// Licensed to the Apache Software Foundation (ASF) under one +// or more contributor license agreements. See the NOTICE file +// distributed with this work for additional information +// regarding copyright ownership. The ASF licenses this file +// to you under the Apache License, Version 2.0 (the +// "License"); you may not use this file except in compliance +// with the License. You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, +// software distributed under the License is distributed on an +// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +// KIND, either express or implied. See the License for the +// specific language governing permissions and limitations +// under the License. + +//! [`ScalarUDFImpl`] definitions for array_add function. + +use crate::utils::make_scalar_function; +use arrow::array::{ + Array, ArrayRef, Float64Array, GenericListArray, NullBufferBuilder, + OffsetBufferBuilder, OffsetSizeTrait, +}; +use arrow::buffer::NullBuffer; +use arrow::datatypes::{ + DataType, + DataType::{FixedSizeList, LargeList, List, Null}, + Field, +}; +use datafusion_common::cast::{as_float64_array, as_generic_list_array}; +use datafusion_common::utils::{ListCoercion, coerced_type_with_base_type_only}; +use datafusion_common::{ + Result, exec_err, not_impl_err, plan_err, utils::take_function_args, +}; +use datafusion_expr::{ + ColumnarValue, Documentation, ScalarFunctionArgs, ScalarUDFImpl, Signature, + Volatility, +}; +use datafusion_macros::user_doc; +use std::sync::Arc; + +make_udf_expr_and_func!( + ArrayAdd, + array_add, + array1 array2, + "returns the element-wise sum of two numeric arrays.", + array_add_udf +); + +#[user_doc( + doc_section(label = "Array Functions"), + description = "Returns the element-wise sum of two numeric arrays of equal length, computed as `array1[i] + array2[i]` per position. NULL is propagated per element: if either input element at position `i` is NULL, the corresponding output element is NULL (positions are preserved). Returns NULL if either entire input array is NULL. Errors if the per-row lengths differ. Returns an empty array if both inputs are empty.", + syntax_example = "array_add(array1, array2)", + sql_example = r#"```sql +> select array_add([1.0, 2.0, 3.0], [10.0, 20.0, 30.0]); ++---------------------------------------------------------+ +| array_add(List([1.0,2.0,3.0]),List([10.0,20.0,30.0])) | ++---------------------------------------------------------+ +| [11.0, 22.0, 33.0] | ++---------------------------------------------------------+ +```"#, + argument( + name = "array1", + description = "Array expression. Can be a constant, column, or function, and any combination of array operators." + ), + argument( + name = "array2", + description = "Array expression. Can be a constant, column, or function, and any combination of array operators." + ) +)] +#[derive(Debug, PartialEq, Eq, Hash)] +pub struct ArrayAdd { + signature: Signature, + aliases: Vec<String>, +} + +impl Default for ArrayAdd { + fn default() -> Self { + Self::new() + } +} + +impl ArrayAdd { + pub fn new() -> Self { + Self { + signature: Signature::user_defined(Volatility::Immutable), + aliases: vec!["list_add".to_string()], + } + } +} + +impl ScalarUDFImpl for ArrayAdd { + fn name(&self) -> &str { + "array_add" + } + + fn signature(&self) -> &Signature { + &self.signature + } + + fn return_type(&self, arg_types: &[DataType]) -> Result<DataType> { + // After `coerce_types`, both args share the same List/LargeList<Float64> shape. + Ok(arg_types[0].clone()) + } + + fn coerce_types(&self, arg_types: &[DataType]) -> Result<Vec<DataType>> { + let [_, _] = take_function_args(self.name(), arg_types)?; + let coercion = Some(&ListCoercion::FixedSizedListToList); + + for arg_type in arg_types { + if !matches!(arg_type, Null | List(_) | LargeList(_) | FixedSizeList(..)) { + return plan_err!("{} does not support type {arg_type}", self.name()); + } + // Only flat lists of numeric leaves are supported. Peel exactly + // one layer so that `List<List<_>>` is rejected as non-numeric + // rather than passing through and failing opaquely in the kernel. + let element_type = match arg_type { + List(field) | LargeList(field) | FixedSizeList(field, _) => { + field.data_type() + } + other => other, + }; + if matches!( + element_type, + DataType::Decimal128(_, _) | DataType::Decimal256(_, _) + ) { + return not_impl_err!( + "{} does not yet support decimal element types ({element_type}); \ + cast to DOUBLE explicitly to opt into lossy float arithmetic", + self.name() + ); + } + if !matches!(element_type, Null) && !element_type.is_numeric() { + return plan_err!( + "{} requires numeric array elements, got list of {element_type}", + self.name() + ); + } + } + + // If either side is `LargeList`, widen both to `LargeList` so the runtime + // dispatch sees a homogeneous pair. + let any_large_list = arg_types.iter().any(|t| matches!(t, LargeList(_))); + + let coerced = arg_types + .iter() + .map(|arg_type| { + if matches!(arg_type, Null) { + let field = Arc::new(Field::new_list_field(DataType::Float64, true)); + return if any_large_list { + LargeList(field) + } else { + List(field) + }; + } + let coerced = coerced_type_with_base_type_only( + arg_type, + &DataType::Float64, + coercion, + ); + match coerced { + List(field) if any_large_list => LargeList(field), + other => other, + } + }) + .collect(); + + Ok(coerced) + } + + fn invoke_with_args(&self, args: ScalarFunctionArgs) -> Result<ColumnarValue> { + make_scalar_function(array_add_inner)(&args.args) + } + + fn aliases(&self) -> &[String] { + &self.aliases + } + + fn documentation(&self) -> Option<&Documentation> { + self.doc() + } +} + +fn array_add_inner(args: &[ArrayRef]) -> Result<ArrayRef> { + let [array1, array2] = take_function_args("array_add", args)?; + match (array1.data_type(), array2.data_type()) { + (List(_), List(_)) => general_array_add::<i32>(args), + (LargeList(_), LargeList(_)) => general_array_add::<i64>(args), + (arg_type1, arg_type2) => exec_err!( + "array_add received unexpected types after coercion: {arg_type1} and {arg_type2}" + ), + } +} + +fn general_array_add<O: OffsetSizeTrait>(arrays: &[ArrayRef]) -> Result<ArrayRef> { Review Comment: Pass in the arrays as separate arguments since we already use `take_function_args` to get two arrays at the callsite ########## datafusion/functions-nested/src/array_add.rs: ########## @@ -0,0 +1,274 @@ +// Licensed to the Apache Software Foundation (ASF) under one +// or more contributor license agreements. See the NOTICE file +// distributed with this work for additional information +// regarding copyright ownership. The ASF licenses this file +// to you under the Apache License, Version 2.0 (the +// "License"); you may not use this file except in compliance +// with the License. You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, +// software distributed under the License is distributed on an +// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +// KIND, either express or implied. See the License for the +// specific language governing permissions and limitations +// under the License. + +//! [`ScalarUDFImpl`] definitions for array_add function. + +use crate::utils::make_scalar_function; +use arrow::array::{ + Array, ArrayRef, Float64Array, GenericListArray, NullBufferBuilder, + OffsetBufferBuilder, OffsetSizeTrait, +}; +use arrow::buffer::NullBuffer; +use arrow::datatypes::{ + DataType, + DataType::{FixedSizeList, LargeList, List, Null}, + Field, +}; +use datafusion_common::cast::{as_float64_array, as_generic_list_array}; +use datafusion_common::utils::{ListCoercion, coerced_type_with_base_type_only}; +use datafusion_common::{ + Result, exec_err, not_impl_err, plan_err, utils::take_function_args, +}; +use datafusion_expr::{ + ColumnarValue, Documentation, ScalarFunctionArgs, ScalarUDFImpl, Signature, + Volatility, +}; +use datafusion_macros::user_doc; +use std::sync::Arc; + +make_udf_expr_and_func!( + ArrayAdd, + array_add, + array1 array2, + "returns the element-wise sum of two numeric arrays.", + array_add_udf +); + +#[user_doc( + doc_section(label = "Array Functions"), + description = "Returns the element-wise sum of two numeric arrays of equal length, computed as `array1[i] + array2[i]` per position. NULL is propagated per element: if either input element at position `i` is NULL, the corresponding output element is NULL (positions are preserved). Returns NULL if either entire input array is NULL. Errors if the per-row lengths differ. Returns an empty array if both inputs are empty.", + syntax_example = "array_add(array1, array2)", + sql_example = r#"```sql +> select array_add([1.0, 2.0, 3.0], [10.0, 20.0, 30.0]); ++---------------------------------------------------------+ +| array_add(List([1.0,2.0,3.0]),List([10.0,20.0,30.0])) | ++---------------------------------------------------------+ +| [11.0, 22.0, 33.0] | ++---------------------------------------------------------+ +```"#, + argument( + name = "array1", + description = "Array expression. Can be a constant, column, or function, and any combination of array operators." + ), + argument( + name = "array2", + description = "Array expression. Can be a constant, column, or function, and any combination of array operators." + ) +)] +#[derive(Debug, PartialEq, Eq, Hash)] +pub struct ArrayAdd { + signature: Signature, + aliases: Vec<String>, +} + +impl Default for ArrayAdd { + fn default() -> Self { + Self::new() + } +} + +impl ArrayAdd { + pub fn new() -> Self { + Self { + signature: Signature::user_defined(Volatility::Immutable), + aliases: vec!["list_add".to_string()], + } + } +} + +impl ScalarUDFImpl for ArrayAdd { + fn name(&self) -> &str { + "array_add" + } + + fn signature(&self) -> &Signature { + &self.signature + } + + fn return_type(&self, arg_types: &[DataType]) -> Result<DataType> { + // After `coerce_types`, both args share the same List/LargeList<Float64> shape. + Ok(arg_types[0].clone()) + } + + fn coerce_types(&self, arg_types: &[DataType]) -> Result<Vec<DataType>> { Review Comment: The logic in here seems to differ from what we usually have, e.g. https://github.com/apache/datafusion/blob/ba240b243fd942107733ccaff5b2221c342bd5ed/datafusion/functions-nested/src/cosine_distance.rs#L100-L139 Is there a reason for this? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
