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

chore: Move some utility methods to submodules of scalar_funcs #590

Merged
merged 2 commits into from
Jun 25, 2024
Merged

chore: Move some utility methods to submodules of scalar_funcs #590

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
a298162
chore: Move some utility methods to submodules of scalar_funcs
advancedxy Jun 20, 2024
d25c37c
Address comments
advancedxy Jun 21, 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
12 changes: 7 additions & 5 deletions core/benches/hash.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,8 +19,7 @@
mod common;

use arrow_array::ArrayRef;
use comet::execution::datafusion::expressions::scalar_funcs::spark_murmur3_hash;
use comet::execution::datafusion::spark_hash::create_xxhash64_hashes;
use comet::execution::datafusion::expressions::scalar_funcs::{spark_murmur3_hash, spark_xxhash64};
use comet::execution::kernels::hash;
use common::*;
use criterion::{criterion_group, criterion_main, BenchmarkId, Criterion};
Expand Down Expand Up @@ -100,12 +99,15 @@ fn criterion_benchmark(c: &mut Criterion) {
},
);
group.bench_function(BenchmarkId::new("xxhash64", BATCH_SIZE), |b| {
let input = vec![a3.clone(), a4.clone()];
let mut dst = vec![0; BATCH_SIZE];
let inputs = &[
ColumnarValue::Array(a3.clone()),
ColumnarValue::Array(a4.clone()),
ColumnarValue::Scalar(ScalarValue::Int64(Some(42i64))),
];

b.iter(|| {
for _ in 0..NUM_ITER {
create_xxhash64_hashes(&input, &mut dst).unwrap();
spark_xxhash64(inputs).unwrap();
kazuyukitanimura marked this conversation as resolved.
Show resolved Hide resolved
}
});
});
Expand Down
141 changes: 9 additions & 132 deletions core/src/execution/datafusion/expressions/scalar_funcs.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,22 +15,16 @@
// specific language governing permissions and limitations
// under the License.

use std::{
any::Any,
cmp::min,
fmt::{Debug, Write},
sync::Arc,
};
use std::{any::Any, cmp::min, fmt::Debug, sync::Arc};

use crate::execution::datafusion::spark_hash::{create_murmur3_hashes, create_xxhash64_hashes};
use arrow::{
array::{
ArrayRef, AsArray, Decimal128Builder, Float32Array, Float64Array, GenericStringArray,
Int16Array, Int32Array, Int64Array, Int64Builder, Int8Array, OffsetSizeTrait,
},
datatypes::{validate_decimal_precision, Decimal128Type, Int64Type},
};
use arrow_array::{Array, ArrowNativeTypeOp, Decimal128Array, StringArray};
use arrow_array::{Array, ArrowNativeTypeOp, Decimal128Array};
use arrow_schema::DataType;
use datafusion::{
execution::FunctionRegistry,
Expand All @@ -39,8 +33,8 @@ use datafusion::{
physical_plan::ColumnarValue,
};
use datafusion_common::{
cast::{as_binary_array, as_generic_string_array},
exec_err, internal_err, DataFusionError, Result as DataFusionResult, ScalarValue,
cast::as_generic_string_array, exec_err, internal_err, DataFusionError,
Result as DataFusionResult, ScalarValue,
};
use datafusion_expr::ScalarUDF;
use num::{
Expand All @@ -58,6 +52,11 @@ use hex::spark_hex;
mod chr;
use chr::spark_chr;

pub mod hash_expressions;
// exposed for benchmark only
use hash_expressions::wrap_digest_result_as_hex_string;
pub use hash_expressions::{spark_murmur3_hash, spark_xxhash64};

macro_rules! make_comet_scalar_udf {
($name:expr, $func:ident, $data_type:ident) => {{
let scalar_func = CometScalarFunction::new(
Expand Down Expand Up @@ -635,125 +634,3 @@ fn spark_decimal_div(
let result = result.with_data_type(DataType::Decimal128(p3, s3));
Ok(ColumnarValue::Array(Arc::new(result)))
}

pub fn spark_murmur3_hash(args: &[ColumnarValue]) -> Result<ColumnarValue, DataFusionError> {
let length = args.len();
let seed = &args[length - 1];
match seed {
ColumnarValue::Scalar(ScalarValue::Int32(Some(seed))) => {
// iterate over the arguments to find out the length of the array
let num_rows = args[0..args.len() - 1]
.iter()
.find_map(|arg| match arg {
ColumnarValue::Array(array) => Some(array.len()),
ColumnarValue::Scalar(_) => None,
})
.unwrap_or(1);
let mut hashes: Vec<u32> = vec![0_u32; num_rows];
hashes.fill(*seed as u32);
let arrays = args[0..args.len() - 1]
.iter()
.map(|arg| match arg {
ColumnarValue::Array(array) => array.clone(),
ColumnarValue::Scalar(scalar) => {
scalar.clone().to_array_of_size(num_rows).unwrap()
}
})
.collect::<Vec<ArrayRef>>();
create_murmur3_hashes(&arrays, &mut hashes)?;
if num_rows == 1 {
Ok(ColumnarValue::Scalar(ScalarValue::Int32(Some(
hashes[0] as i32,
))))
} else {
let hashes: Vec<i32> = hashes.into_iter().map(|x| x as i32).collect();
Ok(ColumnarValue::Array(Arc::new(Int32Array::from(hashes))))
}
}
_ => {
internal_err!(
"The seed of function murmur3_hash must be an Int32 scalar value, but got: {:?}.",
seed
)
}
}
}

fn spark_xxhash64(args: &[ColumnarValue]) -> Result<ColumnarValue, DataFusionError> {
let length = args.len();
let seed = &args[length - 1];
match seed {
ColumnarValue::Scalar(ScalarValue::Int64(Some(seed))) => {
// iterate over the arguments to find out the length of the array
let num_rows = args[0..args.len() - 1]
.iter()
.find_map(|arg| match arg {
ColumnarValue::Array(array) => Some(array.len()),
ColumnarValue::Scalar(_) => None,
})
.unwrap_or(1);
let mut hashes: Vec<u64> = vec![0_u64; num_rows];
hashes.fill(*seed as u64);
let arrays = args[0..args.len() - 1]
.iter()
.map(|arg| match arg {
ColumnarValue::Array(array) => array.clone(),
ColumnarValue::Scalar(scalar) => {
scalar.clone().to_array_of_size(num_rows).unwrap()
}
})
.collect::<Vec<ArrayRef>>();
create_xxhash64_hashes(&arrays, &mut hashes)?;
if num_rows == 1 {
Ok(ColumnarValue::Scalar(ScalarValue::Int64(Some(
hashes[0] as i64,
))))
} else {
let hashes: Vec<i64> = hashes.into_iter().map(|x| x as i64).collect();
Ok(ColumnarValue::Array(Arc::new(Int64Array::from(hashes))))
}
}
_ => {
internal_err!(
"The seed of function xxhash64 must be an Int64 scalar value, but got: {:?}.",
seed
)
}
}
}

#[inline]
fn hex_encode<T: AsRef<[u8]>>(data: T) -> String {
let mut s = String::with_capacity(data.as_ref().len() * 2);
for b in data.as_ref() {
// Writing to a string never errors, so we can unwrap here.
write!(&mut s, "{b:02x}").unwrap();
}
s
}

fn wrap_digest_result_as_hex_string(
args: &[ColumnarValue],
digest: ScalarFunctionImplementation,
) -> Result<ColumnarValue, DataFusionError> {
let value = digest(args)?;
match value {
ColumnarValue::Array(array) => {
let binary_array = as_binary_array(&array)?;
let string_array: StringArray = binary_array
.iter()
.map(|opt| opt.map(hex_encode::<_>))
.collect();
Ok(ColumnarValue::Array(Arc::new(string_array)))
}
ColumnarValue::Scalar(ScalarValue::Binary(opt)) => Ok(ColumnarValue::Scalar(
ScalarValue::Utf8(opt.map(hex_encode::<_>)),
)),
_ => {
exec_err!(
"digest function should return binary value, but got: {:?}",
value.data_type()
)
}
}
}
138 changes: 138 additions & 0 deletions core/src/execution/datafusion/expressions/scalar_funcs/hash_expressions.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,138 @@
// Licensed to the Apache Software Foundation (ASF) under one
// or more contributor license agreements. See the NOTICE file
// distributed with this work for additional information
// regarding copyright ownership. The ASF licenses this file
// to you under the Apache License, Version 2.0 (the
// "License"); you may not use this file except in compliance
// with the License. You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing,
// software distributed under the License is distributed on an
// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
// KIND, either express or implied. See the License for the
// specific language governing permissions and limitations
// under the License.

use crate::execution::datafusion::expressions::scalar_funcs::hex::hex_strings;
use crate::execution::datafusion::spark_hash::{create_murmur3_hashes, create_xxhash64_hashes};
use arrow_array::{ArrayRef, Int32Array, Int64Array, StringArray};
use datafusion_common::cast::as_binary_array;
use datafusion_common::{exec_err, internal_err, DataFusionError, ScalarValue};
use datafusion_expr::{ColumnarValue, ScalarFunctionImplementation};
use std::sync::Arc;

/// Spark compatible murmur3 hash in vectorized execution fashion
pub fn spark_murmur3_hash(args: &[ColumnarValue]) -> Result<ColumnarValue, DataFusionError> {
let length = args.len();
let seed = &args[length - 1];
match seed {
ColumnarValue::Scalar(ScalarValue::Int32(Some(seed))) => {
// iterate over the arguments to find out the length of the array
let num_rows = args[0..args.len() - 1]
Copy link
Contributor

Choose a reason for hiding this comment

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

any chance here to be an index out of bounds?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

I don't think so. The seed is always provided in the Spark/JVM side.

.iter()
.find_map(|arg| match arg {
ColumnarValue::Array(array) => Some(array.len()),
ColumnarValue::Scalar(_) => None,
})
.unwrap_or(1);
let mut hashes: Vec<u32> = vec![0_u32; num_rows];
hashes.fill(*seed as u32);
let arrays = args[0..args.len() - 1]
.iter()
.map(|arg| match arg {
ColumnarValue::Array(array) => array.clone(),
ColumnarValue::Scalar(scalar) => {
scalar.clone().to_array_of_size(num_rows).unwrap()
}
})
.collect::<Vec<ArrayRef>>();
create_murmur3_hashes(&arrays, &mut hashes)?;
if num_rows == 1 {
Ok(ColumnarValue::Scalar(ScalarValue::Int32(Some(
hashes[0] as i32,
))))
} else {
let hashes: Vec<i32> = hashes.into_iter().map(|x| x as i32).collect();
Ok(ColumnarValue::Array(Arc::new(Int32Array::from(hashes))))
}
}
_ => {
internal_err!(
"The seed of function murmur3_hash must be an Int32 scalar value, but got: {:?}.",
seed
)
}
}
}

/// Spark compatible xxhash64 in vectorized execution fashion
pub fn spark_xxhash64(args: &[ColumnarValue]) -> Result<ColumnarValue, DataFusionError> {
Copy link
Contributor

Choose a reason for hiding this comment

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

Its not this PR problem but we need a description to pub methods

Copy link
Contributor

Choose a reason for hiding this comment

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

Additionally, might consider limiting the scope. This function probably isn't needed outside of scalar_funcs.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

but we need a description to pub methods

I can add a description here.

Additionally, might consider limiting the scope. This function probably isn't needed outside of scalar_funcs.

Yeah, I originally limited it to pub(super). However we are accessing it in the benchmark module which needs public interface. I think we can leave it as it is and address access scope later by rewriting the benchmark code.

let length = args.len();
let seed = &args[length - 1];
match seed {
ColumnarValue::Scalar(ScalarValue::Int64(Some(seed))) => {
// iterate over the arguments to find out the length of the array
let num_rows = args[0..args.len() - 1]
.iter()
.find_map(|arg| match arg {
ColumnarValue::Array(array) => Some(array.len()),
ColumnarValue::Scalar(_) => None,
})
.unwrap_or(1);
let mut hashes: Vec<u64> = vec![0_u64; num_rows];
hashes.fill(*seed as u64);
let arrays = args[0..args.len() - 1]
.iter()
.map(|arg| match arg {
ColumnarValue::Array(array) => array.clone(),
ColumnarValue::Scalar(scalar) => {
scalar.clone().to_array_of_size(num_rows).unwrap()
}
})
.collect::<Vec<ArrayRef>>();
create_xxhash64_hashes(&arrays, &mut hashes)?;
if num_rows == 1 {
Ok(ColumnarValue::Scalar(ScalarValue::Int64(Some(
hashes[0] as i64,
))))
} else {
let hashes: Vec<i64> = hashes.into_iter().map(|x| x as i64).collect();
Ok(ColumnarValue::Array(Arc::new(Int64Array::from(hashes))))
}
}
_ => {
internal_err!(
"The seed of function xxhash64 must be an Int64 scalar value, but got: {:?}.",
seed
)
}
}
}

pub(super) fn wrap_digest_result_as_hex_string(
args: &[ColumnarValue],
digest: ScalarFunctionImplementation,
) -> Result<ColumnarValue, DataFusionError> {
let value = digest(args)?;
match value {
ColumnarValue::Array(array) => {
let binary_array = as_binary_array(&array)?;
let string_array: StringArray = binary_array
.iter()
.map(|opt| opt.map(hex_strings::<_>))
.collect();
Ok(ColumnarValue::Array(Arc::new(string_array)))
}
ColumnarValue::Scalar(ScalarValue::Binary(opt)) => Ok(ColumnarValue::Scalar(
ScalarValue::Utf8(opt.map(hex_strings::<_>)),
)),
_ => {
exec_err!(
"digest function should return binary value, but got: {:?}",
value.data_type()
)
}
}
}
34 changes: 26 additions & 8 deletions core/src/execution/datafusion/expressions/scalar_funcs/hex.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,13 +34,31 @@ fn hex_int64(num: i64) -> String {
format!("{:X}", num)
}

fn hex_bytes<T: AsRef<[u8]>>(bytes: T) -> Result<String, std::fmt::Error> {
let bytes = bytes.as_ref();
let length = bytes.len();
let mut hex_string = String::with_capacity(length * 2);
for &byte in bytes {
write!(&mut hex_string, "{:02X}", byte)?;
#[inline(always)]
fn hex_encode<T: AsRef<[u8]>>(data: T, lower_case: bool) -> String {
let mut s = String::with_capacity(data.as_ref().len() * 2);
if lower_case {
for b in data.as_ref() {
// Writing to a string never errors, so we can unwrap here.
write!(&mut s, "{b:02x}").unwrap();
}
} else {
for b in data.as_ref() {
// Writing to a string never errors, so we can unwrap here.
write!(&mut s, "{b:02X}").unwrap();
}
}
s
}

#[inline(always)]
pub(super) fn hex_strings<T: AsRef<[u8]>>(data: T) -> String {
hex_encode(data, true)
}

#[inline(always)]
fn hex_bytes<T: AsRef<[u8]>>(bytes: T) -> Result<String, std::fmt::Error> {
let hex_string = hex_encode(bytes, false);
Ok(hex_string)
}

Expand Down Expand Up @@ -246,14 +264,14 @@ mod test {
fn test_dictionary_hex_binary() {
let mut input_builder = BinaryDictionaryBuilder::<Int32Type>::new();
input_builder.append_value("1");
input_builder.append_value("1");
advancedxy marked this conversation as resolved.
Show resolved Hide resolved
input_builder.append_value("j");
input_builder.append_null();
input_builder.append_value("3");
let input = input_builder.finish();

let mut expected_builder = StringBuilder::new();
expected_builder.append_value("31");
expected_builder.append_value("31");
expected_builder.append_value("6A");
expected_builder.append_null();
expected_builder.append_value("33");
let expected = expected_builder.finish();
Expand Down
Loading