datafusion_physical_expr/expressions/
is_null.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
18//! IS NULL expression
19
20use std::hash::Hash;
21use std::{any::Any, sync::Arc};
22
23use crate::PhysicalExpr;
24use arrow::{
25    datatypes::{DataType, Schema},
26    record_batch::RecordBatch,
27};
28use datafusion_common::Result;
29use datafusion_common::ScalarValue;
30use datafusion_expr::ColumnarValue;
31
32/// IS NULL expression
33#[derive(Debug, Eq)]
34pub struct IsNullExpr {
35    /// Input expression
36    arg: Arc<dyn PhysicalExpr>,
37}
38
39// Manually derive PartialEq and Hash to work around https://github.com/rust-lang/rust/issues/78808
40impl PartialEq for IsNullExpr {
41    fn eq(&self, other: &Self) -> bool {
42        self.arg.eq(&other.arg)
43    }
44}
45
46impl Hash for IsNullExpr {
47    fn hash<H: std::hash::Hasher>(&self, state: &mut H) {
48        self.arg.hash(state);
49    }
50}
51
52impl IsNullExpr {
53    /// Create new not expression
54    pub fn new(arg: Arc<dyn PhysicalExpr>) -> Self {
55        Self { arg }
56    }
57
58    /// Get the input expression
59    pub fn arg(&self) -> &Arc<dyn PhysicalExpr> {
60        &self.arg
61    }
62}
63
64impl std::fmt::Display for IsNullExpr {
65    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
66        write!(f, "{} IS NULL", self.arg)
67    }
68}
69
70impl PhysicalExpr for IsNullExpr {
71    /// Return a reference to Any that can be used for downcasting
72    fn as_any(&self) -> &dyn Any {
73        self
74    }
75
76    fn data_type(&self, _input_schema: &Schema) -> Result<DataType> {
77        Ok(DataType::Boolean)
78    }
79
80    fn nullable(&self, _input_schema: &Schema) -> Result<bool> {
81        Ok(false)
82    }
83
84    fn evaluate(&self, batch: &RecordBatch) -> Result<ColumnarValue> {
85        let arg = self.arg.evaluate(batch)?;
86        match arg {
87            ColumnarValue::Array(array) => Ok(ColumnarValue::Array(Arc::new(
88                arrow::compute::is_null(&array)?,
89            ))),
90            ColumnarValue::Scalar(scalar) => Ok(ColumnarValue::Scalar(
91                ScalarValue::Boolean(Some(scalar.is_null())),
92            )),
93        }
94    }
95
96    fn children(&self) -> Vec<&Arc<dyn PhysicalExpr>> {
97        vec![&self.arg]
98    }
99
100    fn with_new_children(
101        self: Arc<Self>,
102        children: Vec<Arc<dyn PhysicalExpr>>,
103    ) -> Result<Arc<dyn PhysicalExpr>> {
104        Ok(Arc::new(IsNullExpr::new(Arc::clone(&children[0]))))
105    }
106
107    fn fmt_sql(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
108        self.arg.fmt_sql(f)?;
109        write!(f, " IS NULL")
110    }
111}
112
113/// Create an IS NULL expression
114pub fn is_null(arg: Arc<dyn PhysicalExpr>) -> Result<Arc<dyn PhysicalExpr>> {
115    Ok(Arc::new(IsNullExpr::new(arg)))
116}
117
118#[cfg(test)]
119mod tests {
120    use super::*;
121    use crate::expressions::col;
122    use arrow::array::{
123        Array, BooleanArray, Float64Array, Int32Array, StringArray, UnionArray,
124    };
125    use arrow::buffer::ScalarBuffer;
126    use arrow::datatypes::*;
127    use datafusion_common::cast::as_boolean_array;
128    use datafusion_physical_expr_common::physical_expr::fmt_sql;
129
130    #[test]
131    fn is_null_op() -> Result<()> {
132        let schema = Schema::new(vec![Field::new("a", DataType::Utf8, true)]);
133        let a = StringArray::from(vec![Some("foo"), None]);
134
135        // expression: "a is null"
136        let expr = is_null(col("a", &schema)?).unwrap();
137        let batch = RecordBatch::try_new(Arc::new(schema), vec![Arc::new(a)])?;
138
139        let result = expr
140            .evaluate(&batch)?
141            .into_array(batch.num_rows())
142            .expect("Failed to convert to array");
143        let result =
144            as_boolean_array(&result).expect("failed to downcast to BooleanArray");
145
146        let expected = &BooleanArray::from(vec![false, true]);
147
148        assert_eq!(expected, result);
149
150        Ok(())
151    }
152
153    fn union_fields() -> UnionFields {
154        [
155            (0, Arc::new(Field::new("A", DataType::Int32, true))),
156            (1, Arc::new(Field::new("B", DataType::Float64, true))),
157            (2, Arc::new(Field::new("C", DataType::Utf8, true))),
158        ]
159        .into_iter()
160        .collect()
161    }
162
163    #[test]
164    fn sparse_union_is_null() {
165        // union of [{A=1}, {A=}, {B=1.1}, {B=1.2}, {B=}, {C=}, {C="a"}]
166        let int_array =
167            Int32Array::from(vec![Some(1), None, None, None, None, None, None]);
168        let float_array =
169            Float64Array::from(vec![None, None, Some(1.1), Some(1.2), None, None, None]);
170        let str_array =
171            StringArray::from(vec![None, None, None, None, None, None, Some("a")]);
172        let type_ids = [0, 0, 1, 1, 1, 2, 2]
173            .into_iter()
174            .collect::<ScalarBuffer<i8>>();
175
176        let children = vec![
177            Arc::new(int_array) as Arc<dyn Array>,
178            Arc::new(float_array),
179            Arc::new(str_array),
180        ];
181
182        let array =
183            UnionArray::try_new(union_fields(), type_ids, None, children).unwrap();
184
185        let result = arrow::compute::is_null(&array).unwrap();
186
187        let expected =
188            &BooleanArray::from(vec![false, true, false, false, true, true, false]);
189        assert_eq!(expected, &result);
190    }
191
192    #[test]
193    fn dense_union_is_null() {
194        // union of [{A=1}, {A=}, {B=3.2}, {B=}, {C="a"}, {C=}]
195        let int_array = Int32Array::from(vec![Some(1), None]);
196        let float_array = Float64Array::from(vec![Some(3.2), None]);
197        let str_array = StringArray::from(vec![Some("a"), None]);
198        let type_ids = [0, 0, 1, 1, 2, 2].into_iter().collect::<ScalarBuffer<i8>>();
199        let offsets = [0, 1, 0, 1, 0, 1]
200            .into_iter()
201            .collect::<ScalarBuffer<i32>>();
202
203        let children = vec![
204            Arc::new(int_array) as Arc<dyn Array>,
205            Arc::new(float_array),
206            Arc::new(str_array),
207        ];
208
209        let array =
210            UnionArray::try_new(union_fields(), type_ids, Some(offsets), children)
211                .unwrap();
212
213        let result = arrow::compute::is_null(&array).unwrap();
214
215        let expected = &BooleanArray::from(vec![false, true, false, true, false, true]);
216        assert_eq!(expected, &result);
217    }
218
219    #[test]
220    fn test_fmt_sql() -> Result<()> {
221        let schema = Schema::new(vec![Field::new("a", DataType::Utf8, true)]);
222
223        // expression: "a is null"
224        let expr = is_null(col("a", &schema)?).unwrap();
225        let display_string = expr.to_string();
226        assert_eq!(display_string, "a@0 IS NULL");
227        let sql_string = fmt_sql(expr.as_ref()).to_string();
228        assert_eq!(sql_string, "a IS NULL");
229
230        Ok(())
231    }
232}
datafusion_physical_expr/expressions/is_null.rs

datafusion_physical_expr/expressions/
is_null.rs
