datafusion_spark/function/aggregate/
collect.rs

1// Licensed to the Apache Software Foundation (ASF) under one
2// or more contributor license agreements.  See the NOTICE file
3// distributed with this work for additional information
4// regarding copyright ownership.  The ASF licenses this file
5// to you under the Apache License, Version 2.0 (the
6// "License"); you may not use this file except in compliance
7// with the License.  You may obtain a copy of the License at
8//
9//   http://www.apache.org/licenses/LICENSE-2.0
10//
11// Unless required by applicable law or agreed to in writing,
12// software distributed under the License is distributed on an
13// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
14// KIND, either express or implied.  See the License for the
15// specific language governing permissions and limitations
16// under the License.
17
18use arrow::array::ArrayRef;
19use arrow::datatypes::{DataType, Field, FieldRef};
20use datafusion_common::utils::SingleRowListArrayBuilder;
21use datafusion_common::{Result, ScalarValue};
22use datafusion_expr::function::{AccumulatorArgs, StateFieldsArgs};
23use datafusion_expr::utils::format_state_name;
24use datafusion_expr::{Accumulator, AggregateUDFImpl, Signature, Volatility};
25use datafusion_functions_aggregate::array_agg::{
26    ArrayAggAccumulator, DistinctArrayAggAccumulator,
27};
28use std::{any::Any, sync::Arc};
29
30// Spark implementation of collect_list/collect_set aggregate function.
31// Differs from DataFusion ArrayAgg in the following ways:
32// - ignores NULL inputs
33// - returns an empty list when all inputs are NULL
34// - does not support ordering
35
36// <https://spark.apache.org/docs/latest/api/sql/index.html#collect_list>
37#[derive(Debug, PartialEq, Eq, Hash)]
38pub struct SparkCollectList {
39    signature: Signature,
40}
41
42impl Default for SparkCollectList {
43    fn default() -> Self {
44        Self::new()
45    }
46}
47
48impl SparkCollectList {
49    pub fn new() -> Self {
50        Self {
51            signature: Signature::any(1, Volatility::Immutable),
52        }
53    }
54}
55
56impl AggregateUDFImpl for SparkCollectList {
57    fn as_any(&self) -> &dyn Any {
58        self
59    }
60
61    fn name(&self) -> &str {
62        "collect_list"
63    }
64
65    fn signature(&self) -> &Signature {
66        &self.signature
67    }
68
69    fn return_type(&self, arg_types: &[DataType]) -> Result<DataType> {
70        Ok(DataType::List(Arc::new(Field::new_list_field(
71            arg_types[0].clone(),
72            true,
73        ))))
74    }
75
76    fn state_fields(&self, args: StateFieldsArgs) -> Result<Vec<FieldRef>> {
77        Ok(vec![
78            Field::new_list(
79                format_state_name(args.name, "collect_list"),
80                Field::new_list_field(args.input_fields[0].data_type().clone(), true),
81                true,
82            )
83            .into(),
84        ])
85    }
86
87    fn accumulator(&self, acc_args: AccumulatorArgs) -> Result<Box<dyn Accumulator>> {
88        let field = &acc_args.expr_fields[0];
89        let data_type = field.data_type().clone();
90        let ignore_nulls = true;
91        Ok(Box::new(NullToEmptyListAccumulator::new(
92            ArrayAggAccumulator::try_new(&data_type, ignore_nulls)?,
93            data_type,
94        )))
95    }
96}
97
98// <https://spark.apache.org/docs/latest/api/sql/index.html#collect_set>
99#[derive(Debug, PartialEq, Eq, Hash)]
100pub struct SparkCollectSet {
101    signature: Signature,
102}
103
104impl Default for SparkCollectSet {
105    fn default() -> Self {
106        Self::new()
107    }
108}
109
110impl SparkCollectSet {
111    pub fn new() -> Self {
112        Self {
113            signature: Signature::any(1, Volatility::Immutable),
114        }
115    }
116}
117
118impl AggregateUDFImpl for SparkCollectSet {
119    fn as_any(&self) -> &dyn Any {
120        self
121    }
122
123    fn name(&self) -> &str {
124        "collect_set"
125    }
126
127    fn signature(&self) -> &Signature {
128        &self.signature
129    }
130
131    fn return_type(&self, arg_types: &[DataType]) -> Result<DataType> {
132        Ok(DataType::List(Arc::new(Field::new_list_field(
133            arg_types[0].clone(),
134            true,
135        ))))
136    }
137
138    fn state_fields(&self, args: StateFieldsArgs) -> Result<Vec<FieldRef>> {
139        Ok(vec![
140            Field::new_list(
141                format_state_name(args.name, "collect_set"),
142                Field::new_list_field(args.input_fields[0].data_type().clone(), true),
143                true,
144            )
145            .into(),
146        ])
147    }
148
149    fn accumulator(&self, acc_args: AccumulatorArgs) -> Result<Box<dyn Accumulator>> {
150        let field = &acc_args.expr_fields[0];
151        let data_type = field.data_type().clone();
152        let ignore_nulls = true;
153        Ok(Box::new(NullToEmptyListAccumulator::new(
154            DistinctArrayAggAccumulator::try_new(&data_type, None, ignore_nulls)?,
155            data_type,
156        )))
157    }
158}
159
160/// Wrapper accumulator that returns an empty list instead of NULL when all inputs are NULL.
161/// This implements Spark's behavior for collect_list and collect_set.
162#[derive(Debug)]
163struct NullToEmptyListAccumulator<T: Accumulator> {
164    inner: T,
165    data_type: DataType,
166}
167
168impl<T: Accumulator> NullToEmptyListAccumulator<T> {
169    pub fn new(inner: T, data_type: DataType) -> Self {
170        Self { inner, data_type }
171    }
172}
173
174impl<T: Accumulator> Accumulator for NullToEmptyListAccumulator<T> {
175    fn update_batch(&mut self, values: &[ArrayRef]) -> Result<()> {
176        self.inner.update_batch(values)
177    }
178
179    fn merge_batch(&mut self, states: &[ArrayRef]) -> Result<()> {
180        self.inner.merge_batch(states)
181    }
182
183    fn state(&mut self) -> Result<Vec<ScalarValue>> {
184        self.inner.state()
185    }
186
187    fn evaluate(&mut self) -> Result<ScalarValue> {
188        let result = self.inner.evaluate()?;
189        if result.is_null() {
190            let empty_array = arrow::array::new_empty_array(&self.data_type);
191            Ok(SingleRowListArrayBuilder::new(empty_array).build_list_scalar())
192        } else {
193            Ok(result)
194        }
195    }
196
197    fn size(&self) -> usize {
198        self.inner.size() + self.data_type.size()
199    }
200}
datafusion_spark/function/aggregate/collect.rs

datafusion_spark/function/aggregate/
collect.rs
