Skip to content

perf: optimize spark_hex dictionary path by avoiding dictionary expansion #19832

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

Merged
Jefffrey merged 6 commits into from
Jan 18, 2026
Merged

perf: optimize spark_hex dictionary path by avoiding dictionary expansion #19832

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
3b465c3
optimize spark_hex dictionary path
lyne7-sc Jan 15, 2026
8b5087e
cargo fmt
lyne7-sc Jan 15, 2026
62d2912
Merge branch 'apache:main' into perf/spark_hex_dictionary
lyne7-sc Jan 16, 2026
ba8d1ee
refactor hex_encode_bytes/hex_encode_int64 && add datatype for dictio…
lyne7-sc Jan 16, 2026
0946546
Merge branch 'apache:main' into perf/spark_hex_dictionary
lyne7-sc Jan 17, 2026
dcde0f8
update slt to dictionary(binary)
lyne7-sc Jan 17, 2026
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
91 changes: 53 additions & 38 deletions datafusion/spark/src/function/math/hex.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,7 @@ use std::any::Any;
use std::str::from_utf8_unchecked;
use std::sync::Arc;

use arrow::array::{Array, BinaryArray, Int64Array, StringArray, StringBuilder};
use arrow::array::{Array, ArrayRef, StringBuilder};
use arrow::datatypes::DataType;
use arrow::{
array::{as_dictionary_array, as_largestring_array, as_string_array},
Expand Down Expand Up @@ -92,11 +92,13 @@ impl ScalarUDFImpl for SparkHex {
&self.signature
}

fn return_type(
&self,
_arg_types: &[DataType],
) -> datafusion_common::Result<DataType> {
Ok(DataType::Utf8)
fn return_type(&self, arg_types: &[DataType]) -> datafusion_common::Result<DataType> {
Ok(match &arg_types[0] {
DataType::Dictionary(key_type, _) => {
DataType::Dictionary(key_type.clone(), Box::new(DataType::Utf8))
}
_ => DataType::Utf8,
})
}

fn invoke_with_args(
Expand Down Expand Up @@ -241,29 +243,38 @@ pub fn compute_hex(
let array = as_fixed_size_binary_array(array)?;
hex_encode_bytes(array.iter(), lowercase, array.len())
}
DataType::Dictionary(_, value_type) => {
DataType::Dictionary(_, _) => {
let dict = as_dictionary_array::<Int32Type>(&array);
Copy link
Contributor

@Jefffrey Jefffrey Jan 16, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: we should have some check that the dictionary has i32 key type, otherwise this will panic

let dict_values = dict.values();

match **value_type {
let encoded_values: ColumnarValue = match dict_values.data_type() {
Copy link
Contributor

@Jefffrey Jefffrey Jan 16, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We might want to consider arms for LargeUtf, views, etc.

DataType::Int64 => {
let arr = dict.downcast_dict::<Int64Array>().unwrap();
hex_encode_int64(arr.into_iter(), dict.len())
let arr = as_int64_array(dict_values)?;
hex_encode_int64(arr.iter(), arr.len())?
}
DataType::Utf8 => {
let arr = dict.downcast_dict::<StringArray>().unwrap();
hex_encode_bytes(arr.into_iter(), lowercase, dict.len())
let arr = as_string_array(dict_values);
hex_encode_bytes(arr.iter(), lowercase, arr.len())?
}
DataType::Binary => {
let arr = dict.downcast_dict::<BinaryArray>().unwrap();
hex_encode_bytes(arr.into_iter(), lowercase, dict.len())
let arr = as_binary_array(dict_values)?;
hex_encode_bytes(arr.iter(), lowercase, arr.len())?
}
_ => {
exec_err!(
return exec_err!(
"hex got an unexpected argument type: {}",
array.data_type()
)
dict_values.data_type()
);
}
}
};

let encoded_values_array: ArrayRef = match encoded_values {
ColumnarValue::Array(a) => a,
ColumnarValue::Scalar(s) => Arc::new(s.to_array()?),
};
Copy link
Contributor

@Jefffrey Jefffrey Jan 16, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We should probably refactor hex_encode_bytes and hex_encode_int64 to return arrays only, as their signature say they return ColumnarValue but they never return the scalar variant, forcing handling like this


let new_dict = dict.with_values(encoded_values_array);
Ok(ColumnarValue::Array(Arc::new(new_dict)))
}
_ => exec_err!("hex got an unexpected argument type: {}", array.data_type()),
},
Expand All @@ -279,11 +290,12 @@ mod test {
use arrow::array::{DictionaryArray, Int32Array, Int64Array, StringArray};
use arrow::{
array::{
BinaryDictionaryBuilder, PrimitiveDictionaryBuilder, StringBuilder,
StringDictionaryBuilder, as_string_array,
BinaryDictionaryBuilder, PrimitiveDictionaryBuilder, StringDictionaryBuilder,
as_string_array,
},
datatypes::{Int32Type, Int64Type},
};
use datafusion_common::cast::as_dictionary_array;
use datafusion_expr::ColumnarValue;

#[test]
Expand All @@ -295,12 +307,12 @@ mod test {
input_builder.append_value("rust");
let input = input_builder.finish();

let mut string_builder = StringBuilder::new();
string_builder.append_value("6869");
string_builder.append_value("627965");
string_builder.append_null();
string_builder.append_value("72757374");
let expected = string_builder.finish();
let mut expected_builder = StringDictionaryBuilder::<Int32Type>::new();
expected_builder.append_value("6869");
expected_builder.append_value("627965");
expected_builder.append_null();
expected_builder.append_value("72757374");
let expected = expected_builder.finish();

let columnar_value = ColumnarValue::Array(Arc::new(input));
let result = super::spark_hex(&[columnar_value]).unwrap();
Expand All @@ -310,7 +322,7 @@ mod test {
_ => panic!("Expected array"),
};

let result = as_string_array(&result);
let result = as_dictionary_array(&result).unwrap();

assert_eq!(result, &expected);
}
Expand All @@ -324,12 +336,12 @@ mod test {
input_builder.append_value(3);
let input = input_builder.finish();

let mut string_builder = StringBuilder::new();
string_builder.append_value("1");
string_builder.append_value("2");
string_builder.append_null();
string_builder.append_value("3");
let expected = string_builder.finish();
let mut expected_builder = StringDictionaryBuilder::<Int32Type>::new();
expected_builder.append_value("1");
expected_builder.append_value("2");
expected_builder.append_null();
expected_builder.append_value("3");
let expected = expected_builder.finish();

let columnar_value = ColumnarValue::Array(Arc::new(input));
let result = super::spark_hex(&[columnar_value]).unwrap();
Expand All @@ -339,7 +351,7 @@ mod test {
_ => panic!("Expected array"),
};

let result = as_string_array(&result);
let result = as_dictionary_array(&result).unwrap();

assert_eq!(result, &expected);
}
Expand All @@ -353,7 +365,7 @@ mod test {
input_builder.append_value("3");
let input = input_builder.finish();

let mut expected_builder = StringBuilder::new();
let mut expected_builder = StringDictionaryBuilder::<Int32Type>::new();
expected_builder.append_value("31");
expected_builder.append_value("6A");
expected_builder.append_null();
Expand All @@ -368,7 +380,7 @@ mod test {
_ => panic!("Expected array"),
};

let result = as_string_array(&result);
let result = as_dictionary_array(&result).unwrap();

assert_eq!(result, &expected);
}
Expand Down Expand Up @@ -425,8 +437,11 @@ mod test {
_ => panic!("Expected array"),
};

let result = as_string_array(&result);
let expected = StringArray::from(vec![Some("20"), None, None]);
let result = as_dictionary_array(&result).unwrap();

let keys = Int32Array::from(vec![Some(0), None, Some(1)]);
let vals = StringArray::from(vec![Some("20"), None]);
let expected = DictionaryArray::new(keys, Arc::new(vals));

assert_eq!(&expected, result);
}
Expand Down
15 changes: 15 additions & 0 deletions datafusion/sqllogictest/test_files/spark/math/hex.slt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -63,3 +63,18 @@ query T
SELECT hex(arrow_cast('test', 'LargeBinary')) as lar_b;
----
74657374

statement ok
CREATE TABLE t_dict_utf8 AS
SELECT arrow_cast(column1, 'Dictionary(Int32, Utf8)') as dict_col
FROM VALUES ('foo'), ('bar'), ('foo'), (NULL), ('baz'), ('bar');

query T
SELECT hex(dict_col) FROM t_dict_utf8;
Copy link
Contributor

@Jefffrey Jefffrey Jan 16, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can we check the output type here with arrow_typeof to ensure they are dictionaries

Copy link
Contributor Author

@lyne7-sc lyne7-sc Jan 16, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

After running the SLT tests for hex, it seems the planner might unpack dictionary-encoded inputs like Dictionary(Int32, Utf8) or Dictionary(Int32, Int64) into their underlying types (Utf8View or Int64) before calling the function. However, Dictionary(Binary) appears to stay as a dictionary

logical_plan
01)Projection: arrow_typeof(hex(CAST(t_dict_utf8.dict_col AS Utf8View)))
physical_plan
01)ProjectionExec: expr=[arrow_typeof(hex(CAST(dict_col@0 AS Utf8View)))]

logical_plan
01)Projection: arrow_typeof(hex(t_dict_binary.dict_col))
physical_plan
01)ProjectionExec: expr=[arrow_typeof(hex(dict_col@0))]

Copy link
Contributor

@Jefffrey Jefffrey Jan 17, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I guess it's related to this issue

We can still push through with this PR even though it only works for binary (we can change the tests to binary here)

Copy link
Contributor Author

@lyne7-sc lyne7-sc Jan 17, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thank you for the clarification. the tests have been updated to use dictionary(binary).

----
666F6F
626172
666F6F
NULL
62617A
626172