Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Support IS NULL and IS NOT NULL on Unions #11321

Merged
merged 6 commits into from
Jul 8, 2024
Merged

Support IS NULL and IS NOT NULL on Unions #11321

Show file tree
Hide file tree
Changes from 4 commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
f762dcf
Demonstrate unions can't be null
samuelcolvin Jul 7, 2024
6e2ed65
add scalar test cases
samuelcolvin Jul 7, 2024
53cc9a0
support "IS NULL" and "IS NOT NULL" on unions
samuelcolvin Jul 7, 2024
fa015ef
formatting
samuelcolvin Jul 7, 2024
db72566
fix comments from @alamb
samuelcolvin Jul 8, 2024
229255d
fix docstring
samuelcolvin Jul 8, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
34 changes: 33 additions & 1 deletion datafusion/common/src/scalar/mod.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1459,7 +1459,10 @@ impl ScalarValue {
ScalarValue::DurationMillisecond(v) => v.is_none(),
ScalarValue::DurationMicrosecond(v) => v.is_none(),
ScalarValue::DurationNanosecond(v) => v.is_none(),
ScalarValue::Union(v, _, _) => v.is_none(),
ScalarValue::Union(v, _, _) => match v {
Some((_, s)) => s.is_null(),
None => true,
},
ScalarValue::Dictionary(_, v) => v.is_null(),
}
}
Expand Down Expand Up @@ -6514,4 +6517,33 @@ mod tests {
}
intervals
}

fn union_fields() -> UnionFields {
[
(0, Arc::new(Field::new("A", DataType::Int32, true))),
(1, Arc::new(Field::new("B", DataType::Float64, true))),
]
.into_iter()
.collect()
}

#[test]
fn sparse_scalar_union_is_null() {
let sparse_scalar = ScalarValue::Union(
Some((0_i8, Box::new(ScalarValue::Int32(None)))),
union_fields(),
UnionMode::Sparse,
);
assert!(sparse_scalar.is_null());
}

#[test]
fn dense_scalar_union_is_null() {
let dense_scalar = ScalarValue::Union(
Some((0_i8, Box::new(ScalarValue::Int32(None)))),
union_fields(),
UnionMode::Dense,
);
assert!(dense_scalar.is_null());
}
}
165 changes: 163 additions & 2 deletions datafusion/core/tests/dataframe/mod.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,8 +29,9 @@ use arrow::{
},
record_batch::RecordBatch,
};
use arrow_array::Float32Array;
use arrow_schema::ArrowError;
use arrow_array::{Array, Float32Array, Float64Array, UnionArray};
use arrow_buffer::ScalarBuffer;
use arrow_schema::{ArrowError, UnionFields, UnionMode};
use datafusion_functions_aggregate::count::count_udaf;
use object_store::local::LocalFileSystem;
use std::fs;
Expand Down Expand Up @@ -2195,3 +2196,163 @@ async fn write_parquet_results() -> Result<()> {

Ok(())
}

fn union_fields() -> UnionFields {
[
(0, Arc::new(Field::new("A", DataType::Int32, true))),
(1, Arc::new(Field::new("B", DataType::Float64, true))),
(2, Arc::new(Field::new("C", DataType::Utf8, true))),
]
.into_iter()
.collect()
}

#[tokio::test]
async fn sparse_union_is_null() {
// union of [{A=1}, {A=}, {B=3.2}, {B=}, {C="a"}, {C=}]
let int_array = Int32Array::from(vec![Some(1), None, None, None, None, None]);
let float_array = Float64Array::from(vec![None, None, Some(3.2), None, None, None]);
let str_array = StringArray::from(vec![None, None, None, None, Some("a"), None]);
let type_ids = [0, 0, 1, 1, 2, 2].into_iter().collect::<ScalarBuffer<i8>>();

let children = vec![
Arc::new(int_array) as Arc<dyn Array>,
Arc::new(float_array),
Arc::new(str_array),
];

let array = UnionArray::try_new(union_fields(), type_ids, None, children).unwrap();

let field = Field::new(
"my_union",
DataType::Union(union_fields(), UnionMode::Sparse),
true,
);
let schema = Arc::new(Schema::new(vec![field]));

let batch = RecordBatch::try_new(schema, vec![Arc::new(array)]).unwrap();

let ctx = SessionContext::new();

ctx.register_batch("union_batch", batch).unwrap();

let df = ctx.table("union_batch").await.unwrap();

// view_all
let expected = [
"+----------+",
"| my_union |",
"+----------+",
"| {A=1} |",
"| {A=} |",
"| {B=3.2} |",
"| {B=} |",
"| {C=a} |",
"| {C=} |",
"+----------+",
];
assert_batches_sorted_eq!(expected, &df.clone().collect().await.unwrap());

// filter where is null
let result_df = df.clone().filter(col("my_union").is_null()).unwrap();
let expected = [
"+----------+",
"| my_union |",
"+----------+",
"| {A=} |",
"| {B=} |",
"| {C=} |",
"+----------+",
];
assert_batches_sorted_eq!(expected, &result_df.collect().await.unwrap());

// filter where is not null
let result_df = df.filter(col("my_union").is_not_null()).unwrap();
let expected = [
"+----------+",
"| my_union |",
"+----------+",
"| {A=1} |",
"| {B=3.2} |",
"| {C=a} |",
"+----------+",
];
assert_batches_sorted_eq!(expected, &result_df.collect().await.unwrap());
}

#[tokio::test]
async fn dense_union_is_null() {
// union of [{A=1}, null, {B=3.2}, {A=34}]
let int_array = Int32Array::from(vec![Some(1), None]);
let float_array = Float64Array::from(vec![Some(3.2), None]);
let str_array = StringArray::from(vec![Some("a"), None]);
let type_ids = [0, 0, 1, 1, 2, 2].into_iter().collect::<ScalarBuffer<i8>>();
let offsets = [0, 1, 0, 1, 0, 1]
.into_iter()
.collect::<ScalarBuffer<i32>>();

let children = vec![
Arc::new(int_array) as Arc<dyn Array>,
Arc::new(float_array),
Arc::new(str_array),
];

let array =
UnionArray::try_new(union_fields(), type_ids, Some(offsets), children).unwrap();

let field = Field::new(
"my_union",
DataType::Union(union_fields(), UnionMode::Dense),
true,
);
let schema = Arc::new(Schema::new(vec![field]));

let batch = RecordBatch::try_new(schema, vec![Arc::new(array)]).unwrap();

let ctx = SessionContext::new();

ctx.register_batch("union_batch", batch).unwrap();

let df = ctx.table("union_batch").await.unwrap();

// view_all
let expected = [
"+----------+",
"| my_union |",
"+----------+",
"| {A=1} |",
"| {A=} |",
"| {B=3.2} |",
"| {B=} |",
"| {C=a} |",
"| {C=} |",
"+----------+",
];
assert_batches_sorted_eq!(expected, &df.clone().collect().await.unwrap());

// filter where is null
let result_df = df.clone().filter(col("my_union").is_null()).unwrap();
let expected = [
"+----------+",
"| my_union |",
"+----------+",
"| {A=} |",
"| {B=} |",
"| {C=} |",
"+----------+",
];
assert_batches_sorted_eq!(expected, &result_df.collect().await.unwrap());

// filter where is not null
let result_df = df.filter(col("my_union").is_not_null()).unwrap();
let expected = [
"+----------+",
"| my_union |",
"+----------+",
"| {A=1} |",
"| {B=3.2} |",
"| {C=a} |",
"+----------+",
];
assert_batches_sorted_eq!(expected, &result_df.collect().await.unwrap());
}
19 changes: 16 additions & 3 deletions datafusion/physical-expr/src/expressions/is_not_null.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,6 +27,7 @@ use arrow::{
datatypes::{DataType, Schema},
record_batch::RecordBatch,
};
use arrow_array::{BooleanArray, UnionArray};
use datafusion_common::Result;
use datafusion_common::ScalarValue;
use datafusion_expr::ColumnarValue;
Expand Down Expand Up @@ -73,9 +74,16 @@ impl PhysicalExpr for IsNotNullExpr {
fn evaluate(&self, batch: &RecordBatch) -> Result<ColumnarValue> {
let arg = self.arg.evaluate(batch)?;
match arg {
ColumnarValue::Array(array) => Ok(ColumnarValue::Array(Arc::new(
compute::is_not_null(array.as_ref())?,
))),
ColumnarValue::Array(array) => {
let bool_array = if let Some(union_array) =
samuelcolvin marked this conversation as resolved.
Show resolved Hide resolved
array.as_any().downcast_ref::<UnionArray>()
{
union_is_not_null(union_array)?
} else {
compute::is_not_null(array.as_ref())?
};
Ok(ColumnarValue::Array(Arc::new(bool_array)))
}
ColumnarValue::Scalar(scalar) => Ok(ColumnarValue::Scalar(
ScalarValue::Boolean(Some(!scalar.is_null())),
)),
Expand Down Expand Up @@ -112,6 +120,11 @@ pub fn is_not_null(arg: Arc<dyn PhysicalExpr>) -> Result<Arc<dyn PhysicalExpr>>
Ok(Arc::new(IsNotNullExpr::new(arg)))
}

fn union_is_not_null(union_array: &UnionArray) -> Result<BooleanArray> {
super::is_null::union_is_null(union_array)
.map(|is_null| compute::not(&is_null).expect("Failed to compute is not null"))
}

#[cfg(test)]
mod tests {
use super::*;
Expand Down
Loading