datafusion_comet_spark_expr/array_funcs/
array_insert.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::{
19    array::{as_primitive_array, Capacities, MutableArrayData},
20    buffer::{NullBuffer, OffsetBuffer},
21    datatypes::ArrowNativeType,
22    record_batch::RecordBatch,
23};
24use arrow_array::{make_array, Array, ArrayRef, GenericListArray, Int32Array, OffsetSizeTrait};
25use arrow_schema::{DataType, Field, Schema};
26use datafusion::logical_expr::ColumnarValue;
27use datafusion_common::{
28    cast::{as_large_list_array, as_list_array},
29    internal_err, DataFusionError, Result as DataFusionResult,
30};
31use datafusion_physical_expr::PhysicalExpr;
32use std::hash::Hash;
33use std::{
34    any::Any,
35    fmt::{Debug, Display, Formatter},
36    sync::Arc,
37};
38
39// 2147483632 == java.lang.Integer.MAX_VALUE - 15
40// It is a value of ByteArrayUtils.MAX_ROUNDED_ARRAY_LENGTH
41// https://github.com/apache/spark/blob/master/common/utils/src/main/java/org/apache/spark/unsafe/array/ByteArrayUtils.java
42const MAX_ROUNDED_ARRAY_LENGTH: usize = 2147483632;
43
44#[derive(Debug, Eq)]
45pub struct ArrayInsert {
46    src_array_expr: Arc<dyn PhysicalExpr>,
47    pos_expr: Arc<dyn PhysicalExpr>,
48    item_expr: Arc<dyn PhysicalExpr>,
49    legacy_negative_index: bool,
50}
51
52impl Hash for ArrayInsert {
53    fn hash<H: std::hash::Hasher>(&self, state: &mut H) {
54        self.src_array_expr.hash(state);
55        self.pos_expr.hash(state);
56        self.item_expr.hash(state);
57        self.legacy_negative_index.hash(state);
58    }
59}
60impl PartialEq for ArrayInsert {
61    fn eq(&self, other: &Self) -> bool {
62        self.src_array_expr.eq(&other.src_array_expr)
63            && self.pos_expr.eq(&other.pos_expr)
64            && self.item_expr.eq(&other.item_expr)
65            && self.legacy_negative_index.eq(&other.legacy_negative_index)
66    }
67}
68
69impl ArrayInsert {
70    pub fn new(
71        src_array_expr: Arc<dyn PhysicalExpr>,
72        pos_expr: Arc<dyn PhysicalExpr>,
73        item_expr: Arc<dyn PhysicalExpr>,
74        legacy_negative_index: bool,
75    ) -> Self {
76        Self {
77            src_array_expr,
78            pos_expr,
79            item_expr,
80            legacy_negative_index,
81        }
82    }
83
84    pub fn array_type(&self, data_type: &DataType) -> DataFusionResult<DataType> {
85        match data_type {
86            DataType::List(field) => Ok(DataType::List(Arc::clone(field))),
87            DataType::LargeList(field) => Ok(DataType::LargeList(Arc::clone(field))),
88            data_type => Err(DataFusionError::Internal(format!(
89                "Unexpected src array type in ArrayInsert: {:?}",
90                data_type
91            ))),
92        }
93    }
94}
95
96impl PhysicalExpr for ArrayInsert {
97    fn as_any(&self) -> &dyn Any {
98        self
99    }
100
101    fn data_type(&self, input_schema: &Schema) -> DataFusionResult<DataType> {
102        self.array_type(&self.src_array_expr.data_type(input_schema)?)
103    }
104
105    fn nullable(&self, input_schema: &Schema) -> DataFusionResult<bool> {
106        self.src_array_expr.nullable(input_schema)
107    }
108
109    fn evaluate(&self, batch: &RecordBatch) -> DataFusionResult<ColumnarValue> {
110        let pos_value = self
111            .pos_expr
112            .evaluate(batch)?
113            .into_array(batch.num_rows())?;
114
115        // Spark supports only IntegerType (Int32):
116        // https://github.com/apache/spark/blob/branch-3.5/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/collectionOperations.scala#L4737
117        if !matches!(pos_value.data_type(), DataType::Int32) {
118            return Err(DataFusionError::Internal(format!(
119                "Unexpected index data type in ArrayInsert: {:?}, expected type is Int32",
120                pos_value.data_type()
121            )));
122        }
123
124        // Check that src array is actually an array and get it's value type
125        let src_value = self
126            .src_array_expr
127            .evaluate(batch)?
128            .into_array(batch.num_rows())?;
129
130        let src_element_type = match self.array_type(src_value.data_type())? {
131            DataType::List(field) => &field.data_type().clone(),
132            DataType::LargeList(field) => &field.data_type().clone(),
133            _ => unreachable!(),
134        };
135
136        // Check that inserted value has the same type as an array
137        let item_value = self
138            .item_expr
139            .evaluate(batch)?
140            .into_array(batch.num_rows())?;
141        if item_value.data_type() != src_element_type {
142            return Err(DataFusionError::Internal(format!(
143                "Type mismatch in ArrayInsert: array type is {:?} but item type is {:?}",
144                src_element_type,
145                item_value.data_type()
146            )));
147        }
148
149        match src_value.data_type() {
150            DataType::List(_) => {
151                let list_array = as_list_array(&src_value)?;
152                array_insert(
153                    list_array,
154                    &item_value,
155                    &pos_value,
156                    self.legacy_negative_index,
157                )
158            }
159            DataType::LargeList(_) => {
160                let list_array = as_large_list_array(&src_value)?;
161                array_insert(
162                    list_array,
163                    &item_value,
164                    &pos_value,
165                    self.legacy_negative_index,
166                )
167            }
168            _ => unreachable!(), // This case is checked already
169        }
170    }
171
172    fn children(&self) -> Vec<&Arc<dyn PhysicalExpr>> {
173        vec![&self.src_array_expr, &self.pos_expr, &self.item_expr]
174    }
175
176    fn with_new_children(
177        self: Arc<Self>,
178        children: Vec<Arc<dyn PhysicalExpr>>,
179    ) -> DataFusionResult<Arc<dyn PhysicalExpr>> {
180        match children.len() {
181            3 => Ok(Arc::new(ArrayInsert::new(
182                Arc::clone(&children[0]),
183                Arc::clone(&children[1]),
184                Arc::clone(&children[2]),
185                self.legacy_negative_index,
186            ))),
187            _ => internal_err!("ArrayInsert should have exactly three childrens"),
188        }
189    }
190}
191
192fn array_insert<O: OffsetSizeTrait>(
193    list_array: &GenericListArray<O>,
194    items_array: &ArrayRef,
195    pos_array: &ArrayRef,
196    legacy_mode: bool,
197) -> DataFusionResult<ColumnarValue> {
198    // The code is based on the implementation of the array_append from the Apache DataFusion
199    // https://github.com/apache/datafusion/blob/main/datafusion/functions-nested/src/concat.rs#L513
200    //
201    // This code is also based on the implementation of the array_insert from the Apache Spark
202    // https://github.com/apache/spark/blob/branch-3.5/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/collectionOperations.scala#L4713
203
204    let values = list_array.values();
205    let offsets = list_array.offsets();
206    let values_data = values.to_data();
207    let item_data = items_array.to_data();
208    let new_capacity = Capacities::Array(values_data.len() + item_data.len());
209
210    let mut mutable_values =
211        MutableArrayData::with_capacities(vec![&values_data, &item_data], true, new_capacity);
212
213    let mut new_offsets = vec![O::usize_as(0)];
214    let mut new_nulls = Vec::<bool>::with_capacity(list_array.len());
215
216    let pos_data: &Int32Array = as_primitive_array(&pos_array); // Spark supports only i32 for positions
217
218    for (row_index, offset_window) in offsets.windows(2).enumerate() {
219        let pos = pos_data.values()[row_index];
220        let start = offset_window[0].as_usize();
221        let end = offset_window[1].as_usize();
222        let is_item_null = items_array.is_null(row_index);
223
224        if list_array.is_null(row_index) {
225            // In Spark if value of the array is NULL than nothing happens
226            mutable_values.extend_nulls(1);
227            new_offsets.push(new_offsets[row_index] + O::one());
228            new_nulls.push(false);
229            continue;
230        }
231
232        if pos == 0 {
233            return Err(DataFusionError::Internal(
234                "Position for array_insert should be greter or less than zero".to_string(),
235            ));
236        }
237
238        if (pos > 0) || ((-pos).as_usize() < (end - start + 1)) {
239            let corrected_pos = if pos > 0 {
240                (pos - 1).as_usize()
241            } else {
242                end - start - (-pos).as_usize() + if legacy_mode { 0 } else { 1 }
243            };
244            let new_array_len = std::cmp::max(end - start + 1, corrected_pos);
245            if new_array_len > MAX_ROUNDED_ARRAY_LENGTH {
246                return Err(DataFusionError::Internal(format!(
247                    "Max array length in Spark is {:?}, but got {:?}",
248                    MAX_ROUNDED_ARRAY_LENGTH, new_array_len
249                )));
250            }
251
252            if (start + corrected_pos) <= end {
253                mutable_values.extend(0, start, start + corrected_pos);
254                mutable_values.extend(1, row_index, row_index + 1);
255                mutable_values.extend(0, start + corrected_pos, end);
256                new_offsets.push(new_offsets[row_index] + O::usize_as(new_array_len));
257            } else {
258                mutable_values.extend(0, start, end);
259                mutable_values.extend_nulls(new_array_len - (end - start));
260                mutable_values.extend(1, row_index, row_index + 1);
261                // In that case spark actualy makes array longer than expected;
262                // For example, if pos is equal to 5, len is eq to 3, than resulted len will be 5
263                new_offsets.push(new_offsets[row_index] + O::usize_as(new_array_len) + O::one());
264            }
265        } else {
266            // This comment is takes from the Apache Spark source code as is:
267            // special case- if the new position is negative but larger than the current array size
268            // place the new item at start of array, place the current array contents at the end
269            // and fill the newly created array elements inbetween with a null
270            let base_offset = if legacy_mode { 1 } else { 0 };
271            let new_array_len = (-pos + base_offset).as_usize();
272            if new_array_len > MAX_ROUNDED_ARRAY_LENGTH {
273                return Err(DataFusionError::Internal(format!(
274                    "Max array length in Spark is {:?}, but got {:?}",
275                    MAX_ROUNDED_ARRAY_LENGTH, new_array_len
276                )));
277            }
278            mutable_values.extend(1, row_index, row_index + 1);
279            mutable_values.extend_nulls(new_array_len - (end - start + 1));
280            mutable_values.extend(0, start, end);
281            new_offsets.push(new_offsets[row_index] + O::usize_as(new_array_len));
282        }
283        if is_item_null {
284            if (start == end) || (values.is_null(row_index)) {
285                new_nulls.push(false)
286            } else {
287                new_nulls.push(true)
288            }
289        } else {
290            new_nulls.push(true)
291        }
292    }
293
294    let data = make_array(mutable_values.freeze());
295    let data_type = match list_array.data_type() {
296        DataType::List(field) => field.data_type(),
297        DataType::LargeList(field) => field.data_type(),
298        _ => unreachable!(),
299    };
300    let new_array = GenericListArray::<O>::try_new(
301        Arc::new(Field::new("item", data_type.clone(), true)),
302        OffsetBuffer::new(new_offsets.into()),
303        data,
304        Some(NullBuffer::new(new_nulls.into())),
305    )?;
306
307    Ok(ColumnarValue::Array(Arc::new(new_array)))
308}
309
310impl Display for ArrayInsert {
311    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
312        write!(
313            f,
314            "ArrayInsert [array: {:?}, pos: {:?}, item: {:?}]",
315            self.src_array_expr, self.pos_expr, self.item_expr
316        )
317    }
318}
319
320#[cfg(test)]
321mod test {
322    use super::*;
323    use arrow::datatypes::Int32Type;
324    use arrow_array::{Array, ArrayRef, Int32Array, ListArray};
325    use datafusion_common::Result;
326    use datafusion_expr::ColumnarValue;
327    use std::sync::Arc;
328
329    #[test]
330    fn test_array_insert() -> Result<()> {
331        // Test inserting an item into a list array
332        // Inputs and expected values are taken from the Spark results
333        let list = ListArray::from_iter_primitive::<Int32Type, _, _>(vec![
334            Some(vec![Some(1), Some(2), Some(3)]),
335            Some(vec![Some(4), Some(5)]),
336            Some(vec![None]),
337            Some(vec![Some(1), Some(2), Some(3)]),
338            Some(vec![Some(1), Some(2), Some(3)]),
339            None,
340        ]);
341
342        let positions = Int32Array::from(vec![2, 1, 1, 5, 6, 1]);
343        let items = Int32Array::from(vec![
344            Some(10),
345            Some(20),
346            Some(30),
347            Some(100),
348            Some(100),
349            Some(40),
350        ]);
351
352        let ColumnarValue::Array(result) = array_insert(
353            &list,
354            &(Arc::new(items) as ArrayRef),
355            &(Arc::new(positions) as ArrayRef),
356            false,
357        )?
358        else {
359            unreachable!()
360        };
361
362        let expected = ListArray::from_iter_primitive::<Int32Type, _, _>(vec![
363            Some(vec![Some(1), Some(10), Some(2), Some(3)]),
364            Some(vec![Some(20), Some(4), Some(5)]),
365            Some(vec![Some(30), None]),
366            Some(vec![Some(1), Some(2), Some(3), None, Some(100)]),
367            Some(vec![Some(1), Some(2), Some(3), None, None, Some(100)]),
368            None,
369        ]);
370
371        assert_eq!(&result.to_data(), &expected.to_data());
372
373        Ok(())
374    }
375
376    #[test]
377    fn test_array_insert_negative_index() -> Result<()> {
378        // Test insert with negative index
379        // Inputs and expected values are taken from the Spark results
380        let list = ListArray::from_iter_primitive::<Int32Type, _, _>(vec![
381            Some(vec![Some(1), Some(2), Some(3)]),
382            Some(vec![Some(4), Some(5)]),
383            Some(vec![Some(1)]),
384            None,
385        ]);
386
387        let positions = Int32Array::from(vec![-2, -1, -3, -1]);
388        let items = Int32Array::from(vec![Some(10), Some(20), Some(100), Some(30)]);
389
390        let ColumnarValue::Array(result) = array_insert(
391            &list,
392            &(Arc::new(items) as ArrayRef),
393            &(Arc::new(positions) as ArrayRef),
394            false,
395        )?
396        else {
397            unreachable!()
398        };
399
400        let expected = ListArray::from_iter_primitive::<Int32Type, _, _>(vec![
401            Some(vec![Some(1), Some(2), Some(10), Some(3)]),
402            Some(vec![Some(4), Some(5), Some(20)]),
403            Some(vec![Some(100), None, Some(1)]),
404            None,
405        ]);
406
407        assert_eq!(&result.to_data(), &expected.to_data());
408
409        Ok(())
410    }
411
412    #[test]
413    fn test_array_insert_legacy_mode() -> Result<()> {
414        // Test the so-called "legacy" mode exisiting in the Spark
415        let list = ListArray::from_iter_primitive::<Int32Type, _, _>(vec![
416            Some(vec![Some(1), Some(2), Some(3)]),
417            Some(vec![Some(4), Some(5)]),
418            None,
419        ]);
420
421        let positions = Int32Array::from(vec![-1, -1, -1]);
422        let items = Int32Array::from(vec![Some(10), Some(20), Some(30)]);
423
424        let ColumnarValue::Array(result) = array_insert(
425            &list,
426            &(Arc::new(items) as ArrayRef),
427            &(Arc::new(positions) as ArrayRef),
428            true,
429        )?
430        else {
431            unreachable!()
432        };
433
434        let expected = ListArray::from_iter_primitive::<Int32Type, _, _>(vec![
435            Some(vec![Some(1), Some(2), Some(10), Some(3)]),
436            Some(vec![Some(4), Some(20), Some(5)]),
437            None,
438        ]);
439
440        assert_eq!(&result.to_data(), &expected.to_data());
441
442        Ok(())
443    }
444}
datafusion_comet_spark_expr/array_funcs/array_insert.rs

datafusion_comet_spark_expr/array_funcs/
array_insert.rs
