coderfender commented on code in PR #20594: URL: https://github.com/apache/datafusion/pull/20594#discussion_r2926681609
########## datafusion/spark/src/function/math/floor.rs: ########## @@ -0,0 +1,263 @@ +// 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. + +use std::any::Any; +use std::sync::Arc; + +use arrow::array::{AsArray, Decimal128Array}; +use arrow::compute::cast; +use arrow::datatypes::{DataType, Decimal128Type, Float32Type, Float64Type, Int64Type}; +use datafusion_common::utils::take_function_args; +use datafusion_common::{Result, exec_err}; +use datafusion_expr::{ + ColumnarValue, ScalarFunctionArgs, ScalarUDFImpl, Signature, Volatility, +}; + +/// Spark-compatible `floor` expression +/// <https://spark.apache.org/docs/latest/api/sql/index.html#floor> +/// +/// Differences with DataFusion floor: +/// - Spark's floor returns Int64 for float and integer inputs; DataFusion preserves +/// the input type (Float32→Float32, Float64→Float64, integers coerced to Float64) +/// - Spark's floor on Decimal128(p, s) returns Decimal128(p−s+1, 0), reducing scale +/// to 0; DataFusion preserves the original precision and scale +/// - Spark only supports Decimal128; DataFusion also supports Decimal32/64/256 +/// - Spark does not check for decimal overflow; DataFusion errors on overflow +#[derive(Debug, PartialEq, Eq, Hash)] +pub struct SparkFloor { + signature: Signature, +} + +impl Default for SparkFloor { + fn default() -> Self { + Self::new() + } +} + +impl SparkFloor { + pub fn new() -> Self { + Self { + signature: Signature::numeric(1, Volatility::Immutable), + } + } +} + +impl ScalarUDFImpl for SparkFloor { + fn as_any(&self) -> &dyn Any { + self + } + + fn name(&self) -> &str { + "floor" + } + + fn signature(&self) -> &Signature { + &self.signature + } + + fn return_type(&self, arg_types: &[DataType]) -> Result<DataType> { + match &arg_types[0] { + DataType::Decimal128(p, s) if *s > 0 => { + let new_p = ((*p as i64) - (*s as i64) + 1).clamp(1, 38) as u8; Review Comment: I dont think we need i64 casting here. i8 or even u8 could do here ########## datafusion/spark/src/function/math/floor.rs: ########## @@ -0,0 +1,263 @@ +// 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. + +use std::any::Any; +use std::sync::Arc; + +use arrow::array::{AsArray, Decimal128Array}; +use arrow::compute::cast; +use arrow::datatypes::{DataType, Decimal128Type, Float32Type, Float64Type, Int64Type}; +use datafusion_common::utils::take_function_args; +use datafusion_common::{Result, exec_err}; +use datafusion_expr::{ + ColumnarValue, ScalarFunctionArgs, ScalarUDFImpl, Signature, Volatility, +}; + +/// Spark-compatible `floor` expression +/// <https://spark.apache.org/docs/latest/api/sql/index.html#floor> +/// +/// Differences with DataFusion floor: +/// - Spark's floor returns Int64 for float and integer inputs; DataFusion preserves +/// the input type (Float32→Float32, Float64→Float64, integers coerced to Float64) +/// - Spark's floor on Decimal128(p, s) returns Decimal128(p−s+1, 0), reducing scale +/// to 0; DataFusion preserves the original precision and scale +/// - Spark only supports Decimal128; DataFusion also supports Decimal32/64/256 +/// - Spark does not check for decimal overflow; DataFusion errors on overflow +#[derive(Debug, PartialEq, Eq, Hash)] +pub struct SparkFloor { + signature: Signature, +} + +impl Default for SparkFloor { + fn default() -> Self { + Self::new() + } +} + +impl SparkFloor { + pub fn new() -> Self { + Self { + signature: Signature::numeric(1, Volatility::Immutable), + } + } +} + +impl ScalarUDFImpl for SparkFloor { + fn as_any(&self) -> &dyn Any { + self + } + + fn name(&self) -> &str { + "floor" + } + + fn signature(&self) -> &Signature { + &self.signature + } + + fn return_type(&self, arg_types: &[DataType]) -> Result<DataType> { + match &arg_types[0] { + DataType::Decimal128(p, s) if *s > 0 => { + let new_p = ((*p as i64) - (*s as i64) + 1).clamp(1, 38) as u8; + Ok(DataType::Decimal128(new_p, 0)) + } + DataType::Decimal128(p, s) => Ok(DataType::Decimal128(*p, *s)), + _ => Ok(DataType::Int64), + } + } + + fn invoke_with_args(&self, args: ScalarFunctionArgs) -> Result<ColumnarValue> { + let return_type = args.return_type().clone(); + spark_floor(&args.args, &return_type) + } +} + +fn spark_floor(args: &[ColumnarValue], return_type: &DataType) -> Result<ColumnarValue> { + let input = match take_function_args("floor", args)? { + [ColumnarValue::Scalar(value)] => value.to_array()?, + [ColumnarValue::Array(arr)] => Arc::clone(arr), + }; + + let result = match input.data_type() { + DataType::Float32 => Arc::new( + input + .as_primitive::<Float32Type>() + .unary::<_, Int64Type>(|x| x.floor() as i64), + ) as _, + DataType::Float64 => Arc::new( + input + .as_primitive::<Float64Type>() + .unary::<_, Int64Type>(|x| x.floor() as i64), + ) as _, + dt if dt.is_integer() => cast(&input, &DataType::Int64)?, + DataType::Decimal128(_, s) if *s > 0 => { + let div = 10_i128.pow(*s as u32); + let result: Decimal128Array = + input.as_primitive::<Decimal128Type>().unary(|x| { + let d = x / div; + let r = x % div; + if r < 0 { d - 1 } else { d } + }); + Arc::new(result.with_data_type(return_type.clone())) + } + DataType::Decimal128(_, _) => input, + other => return exec_err!("Unsupported data type {other:?} for function floor"), + }; + + Ok(ColumnarValue::Array(result)) +} + +#[cfg(test)] +mod tests { + use super::*; + use arrow::array::{Decimal128Array, Float32Array, Float64Array, Int64Array}; + use datafusion_common::ScalarValue; + + #[test] + fn test_floor_float64() { + let input = Float64Array::from(vec![ Review Comment: Could we also test `Nan / Infinity` inputs for float types ? ########## datafusion/spark/src/function/math/floor.rs: ########## @@ -0,0 +1,263 @@ +// 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. + +use std::any::Any; +use std::sync::Arc; + +use arrow::array::{AsArray, Decimal128Array}; +use arrow::compute::cast; +use arrow::datatypes::{DataType, Decimal128Type, Float32Type, Float64Type, Int64Type}; +use datafusion_common::utils::take_function_args; +use datafusion_common::{Result, exec_err}; +use datafusion_expr::{ + ColumnarValue, ScalarFunctionArgs, ScalarUDFImpl, Signature, Volatility, +}; + +/// Spark-compatible `floor` expression +/// <https://spark.apache.org/docs/latest/api/sql/index.html#floor> +/// +/// Differences with DataFusion floor: +/// - Spark's floor returns Int64 for float and integer inputs; DataFusion preserves +/// the input type (Float32→Float32, Float64→Float64, integers coerced to Float64) +/// - Spark's floor on Decimal128(p, s) returns Decimal128(p−s+1, 0), reducing scale +/// to 0; DataFusion preserves the original precision and scale +/// - Spark only supports Decimal128; DataFusion also supports Decimal32/64/256 +/// - Spark does not check for decimal overflow; DataFusion errors on overflow +#[derive(Debug, PartialEq, Eq, Hash)] +pub struct SparkFloor { + signature: Signature, +} + +impl Default for SparkFloor { + fn default() -> Self { + Self::new() + } +} + +impl SparkFloor { + pub fn new() -> Self { + Self { + signature: Signature::numeric(1, Volatility::Immutable), Review Comment: @shivbhatia10 you could probably use `logical_integer` / decimal / float and coerce inputs to reject unwanted inputs during planning phase ########## datafusion/spark/src/function/math/floor.rs: ########## @@ -0,0 +1,263 @@ +// 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. + +use std::any::Any; +use std::sync::Arc; + +use arrow::array::{AsArray, Decimal128Array}; +use arrow::compute::cast; +use arrow::datatypes::{DataType, Decimal128Type, Float32Type, Float64Type, Int64Type}; +use datafusion_common::utils::take_function_args; +use datafusion_common::{Result, exec_err}; +use datafusion_expr::{ + ColumnarValue, ScalarFunctionArgs, ScalarUDFImpl, Signature, Volatility, +}; + +/// Spark-compatible `floor` expression +/// <https://spark.apache.org/docs/latest/api/sql/index.html#floor> +/// +/// Differences with DataFusion floor: +/// - Spark's floor returns Int64 for float and integer inputs; DataFusion preserves +/// the input type (Float32→Float32, Float64→Float64, integers coerced to Float64) +/// - Spark's floor on Decimal128(p, s) returns Decimal128(p−s+1, 0), reducing scale +/// to 0; DataFusion preserves the original precision and scale +/// - Spark only supports Decimal128; DataFusion also supports Decimal32/64/256 +/// - Spark does not check for decimal overflow; DataFusion errors on overflow +#[derive(Debug, PartialEq, Eq, Hash)] +pub struct SparkFloor { + signature: Signature, +} + +impl Default for SparkFloor { + fn default() -> Self { + Self::new() + } +} + +impl SparkFloor { + pub fn new() -> Self { + Self { + signature: Signature::numeric(1, Volatility::Immutable), + } + } +} + +impl ScalarUDFImpl for SparkFloor { + fn as_any(&self) -> &dyn Any { + self + } + + fn name(&self) -> &str { + "floor" + } + + fn signature(&self) -> &Signature { + &self.signature + } + + fn return_type(&self, arg_types: &[DataType]) -> Result<DataType> { + match &arg_types[0] { + DataType::Decimal128(p, s) if *s > 0 => { + let new_p = ((*p as i64) - (*s as i64) + 1).clamp(1, 38) as u8; + Ok(DataType::Decimal128(new_p, 0)) + } + DataType::Decimal128(p, s) => Ok(DataType::Decimal128(*p, *s)), + _ => Ok(DataType::Int64), + } + } + + fn invoke_with_args(&self, args: ScalarFunctionArgs) -> Result<ColumnarValue> { + let return_type = args.return_type().clone(); + spark_floor(&args.args, &return_type) + } +} + +fn spark_floor(args: &[ColumnarValue], return_type: &DataType) -> Result<ColumnarValue> { + let input = match take_function_args("floor", args)? { + [ColumnarValue::Scalar(value)] => value.to_array()?, Review Comment: @shivbhatia10 , please take a look at my floor function for reference (I implemented macros to not convert scalars -> array and back) : https://github.com/apache/datafusion/pull/20860 ########## datafusion/spark/src/function/math/floor.rs: ########## @@ -0,0 +1,263 @@ +// 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. + +use std::any::Any; +use std::sync::Arc; + +use arrow::array::{AsArray, Decimal128Array}; +use arrow::compute::cast; +use arrow::datatypes::{DataType, Decimal128Type, Float32Type, Float64Type, Int64Type}; +use datafusion_common::utils::take_function_args; +use datafusion_common::{Result, exec_err}; +use datafusion_expr::{ + ColumnarValue, ScalarFunctionArgs, ScalarUDFImpl, Signature, Volatility, +}; + +/// Spark-compatible `floor` expression +/// <https://spark.apache.org/docs/latest/api/sql/index.html#floor> +/// +/// Differences with DataFusion floor: +/// - Spark's floor returns Int64 for float and integer inputs; DataFusion preserves +/// the input type (Float32→Float32, Float64→Float64, integers coerced to Float64) +/// - Spark's floor on Decimal128(p, s) returns Decimal128(p−s+1, 0), reducing scale +/// to 0; DataFusion preserves the original precision and scale +/// - Spark only supports Decimal128; DataFusion also supports Decimal32/64/256 +/// - Spark does not check for decimal overflow; DataFusion errors on overflow +#[derive(Debug, PartialEq, Eq, Hash)] +pub struct SparkFloor { + signature: Signature, +} + +impl Default for SparkFloor { + fn default() -> Self { + Self::new() + } +} + +impl SparkFloor { + pub fn new() -> Self { + Self { + signature: Signature::numeric(1, Volatility::Immutable), + } + } +} + +impl ScalarUDFImpl for SparkFloor { + fn as_any(&self) -> &dyn Any { + self + } + + fn name(&self) -> &str { + "floor" + } + + fn signature(&self) -> &Signature { + &self.signature + } + + fn return_type(&self, arg_types: &[DataType]) -> Result<DataType> { + match &arg_types[0] { + DataType::Decimal128(p, s) if *s > 0 => { + let new_p = ((*p as i64) - (*s as i64) + 1).clamp(1, 38) as u8; + Ok(DataType::Decimal128(new_p, 0)) + } + DataType::Decimal128(p, s) => Ok(DataType::Decimal128(*p, *s)), + _ => Ok(DataType::Int64), + } + } + + fn invoke_with_args(&self, args: ScalarFunctionArgs) -> Result<ColumnarValue> { + let return_type = args.return_type().clone(); + spark_floor(&args.args, &return_type) + } +} + +fn spark_floor(args: &[ColumnarValue], return_type: &DataType) -> Result<ColumnarValue> { + let input = match take_function_args("floor", args)? { + [ColumnarValue::Scalar(value)] => value.to_array()?, + [ColumnarValue::Array(arr)] => Arc::clone(arr), + }; + + let result = match input.data_type() { + DataType::Float32 => Arc::new( + input + .as_primitive::<Float32Type>() + .unary::<_, Int64Type>(|x| x.floor() as i64), + ) as _, + DataType::Float64 => Arc::new( + input + .as_primitive::<Float64Type>() + .unary::<_, Int64Type>(|x| x.floor() as i64), + ) as _, + dt if dt.is_integer() => cast(&input, &DataType::Int64)?, Review Comment: Could simply cast to i64 to avoid unnecessary stack frames from cast function ########## datafusion/spark/src/function/math/floor.rs: ########## @@ -0,0 +1,263 @@ +// 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. + +use std::any::Any; +use std::sync::Arc; + +use arrow::array::{AsArray, Decimal128Array}; +use arrow::compute::cast; +use arrow::datatypes::{DataType, Decimal128Type, Float32Type, Float64Type, Int64Type}; +use datafusion_common::utils::take_function_args; +use datafusion_common::{Result, exec_err}; +use datafusion_expr::{ + ColumnarValue, ScalarFunctionArgs, ScalarUDFImpl, Signature, Volatility, +}; + +/// Spark-compatible `floor` expression +/// <https://spark.apache.org/docs/latest/api/sql/index.html#floor> +/// +/// Differences with DataFusion floor: +/// - Spark's floor returns Int64 for float and integer inputs; DataFusion preserves +/// the input type (Float32→Float32, Float64→Float64, integers coerced to Float64) +/// - Spark's floor on Decimal128(p, s) returns Decimal128(p−s+1, 0), reducing scale +/// to 0; DataFusion preserves the original precision and scale +/// - Spark only supports Decimal128; DataFusion also supports Decimal32/64/256 +/// - Spark does not check for decimal overflow; DataFusion errors on overflow +#[derive(Debug, PartialEq, Eq, Hash)] +pub struct SparkFloor { + signature: Signature, +} + +impl Default for SparkFloor { + fn default() -> Self { + Self::new() + } +} + +impl SparkFloor { + pub fn new() -> Self { + Self { + signature: Signature::numeric(1, Volatility::Immutable), + } + } +} + +impl ScalarUDFImpl for SparkFloor { + fn as_any(&self) -> &dyn Any { + self + } + + fn name(&self) -> &str { + "floor" + } + + fn signature(&self) -> &Signature { + &self.signature + } + + fn return_type(&self, arg_types: &[DataType]) -> Result<DataType> { + match &arg_types[0] { + DataType::Decimal128(p, s) if *s > 0 => { + let new_p = ((*p as i64) - (*s as i64) + 1).clamp(1, 38) as u8; + Ok(DataType::Decimal128(new_p, 0)) + } + DataType::Decimal128(p, s) => Ok(DataType::Decimal128(*p, *s)), + _ => Ok(DataType::Int64), + } + } + + fn invoke_with_args(&self, args: ScalarFunctionArgs) -> Result<ColumnarValue> { + let return_type = args.return_type().clone(); + spark_floor(&args.args, &return_type) + } +} + +fn spark_floor(args: &[ColumnarValue], return_type: &DataType) -> Result<ColumnarValue> { + let input = match take_function_args("floor", args)? { + [ColumnarValue::Scalar(value)] => value.to_array()?, + [ColumnarValue::Array(arr)] => Arc::clone(arr), + }; + + let result = match input.data_type() { + DataType::Float32 => Arc::new( + input + .as_primitive::<Float32Type>() + .unary::<_, Int64Type>(|x| x.floor() as i64), + ) as _, + DataType::Float64 => Arc::new( + input + .as_primitive::<Float64Type>() + .unary::<_, Int64Type>(|x| x.floor() as i64), + ) as _, + dt if dt.is_integer() => cast(&input, &DataType::Int64)?, + DataType::Decimal128(_, s) if *s > 0 => { + let div = 10_i128.pow(*s as u32); + let result: Decimal128Array = + input.as_primitive::<Decimal128Type>().unary(|x| { + let d = x / div; + let r = x % div; + if r < 0 { d - 1 } else { d } + }); + Arc::new(result.with_data_type(return_type.clone())) + } + DataType::Decimal128(_, _) => input, + other => return exec_err!("Unsupported data type {other:?} for function floor"), + }; + + Ok(ColumnarValue::Array(result)) +} + +#[cfg(test)] +mod tests { + use super::*; + use arrow::array::{Decimal128Array, Float32Array, Float64Array, Int64Array}; + use datafusion_common::ScalarValue; + + #[test] + fn test_floor_float64() { + let input = Float64Array::from(vec![ + Some(125.9345), + Some(15.9999), + Some(0.9), + Some(-0.1), + Some(-1.999), + Some(123.0), + None, + ]); + let args = vec![ColumnarValue::Array(Arc::new(input))]; + let result = spark_floor(&args, &DataType::Int64).unwrap(); + let result = match result { + ColumnarValue::Array(arr) => arr, + _ => panic!("Expected array"), + }; + let result = result.as_primitive::<Int64Type>(); + assert_eq!( + result, + &Int64Array::from(vec![ + Some(125), + Some(15), + Some(0), + Some(-1), + Some(-2), + Some(123), + None, + ]) + ); + } + + #[test] + fn test_floor_float32() { + let input = Float32Array::from(vec![ + Some(125.9345f32), + Some(15.9999f32), + Some(0.9f32), + Some(-0.1f32), + Some(-1.999f32), + Some(123.0f32), + None, + ]); + let args = vec![ColumnarValue::Array(Arc::new(input))]; + let result = spark_floor(&args, &DataType::Int64).unwrap(); + let result = match result { + ColumnarValue::Array(arr) => arr, + _ => panic!("Expected array"), + }; + let result = result.as_primitive::<Int64Type>(); + assert_eq!( + result, + &Int64Array::from(vec![ + Some(125), + Some(15), + Some(0), + Some(-1), + Some(-2), + Some(123), + None, + ]) + ); + } + + #[test] + fn test_floor_int64() { + let input = Int64Array::from(vec![Some(1), Some(-1), None]); Review Comment: Should we test with all possible integer types here ? -- 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]
