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.

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

Implement boolean equality kernels #844

Merged
merged 4 commits into from
Oct 23, 2021
Merged
Show file tree
Hide file tree
Changes from 2 commits
Commits
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
2 changes: 1 addition & 1 deletion arrow/src/compute/kernels/boolean.rs
Original file line number Diff line number Diff line change
Expand Up @@ -175,7 +175,7 @@ where
}

/// Helper function to implement binary kernels
fn binary_boolean_kernel<F>(
pub(crate) fn binary_boolean_kernel<F>(
left: &BooleanArray,
right: &BooleanArray,
op: F,
Expand Down
165 changes: 164 additions & 1 deletion arrow/src/compute/kernels/comparison.rs
Original file line number Diff line number Diff line change
Expand Up @@ -26,7 +26,8 @@ use regex::Regex;
use std::collections::HashMap;

use crate::array::*;
use crate::buffer::{Buffer, MutableBuffer};
use crate::buffer::{bitwise_bin_op_helper, buffer_unary_not, Buffer, MutableBuffer};
use crate::compute::binary_boolean_kernel;
use crate::compute::util::combine_option_bitmap;
use crate::datatypes::{ArrowNumericType, DataType};
use crate::error::{ArrowError, Result};
Expand Down Expand Up @@ -623,6 +624,107 @@ pub fn eq_utf8_scalar<OffsetSize: StringOffsetSizeTrait>(
compare_op_scalar!(left, right, |a, b| a == b)
}

/// Perform `left == right` operation on [`BooleanArray`]
fn eq_bool(left: &BooleanArray, right: &BooleanArray) -> Result<BooleanArray> {
binary_boolean_kernel(
left,
right,
|left: &Buffer,
left_offset_in_bits: usize,
right: &Buffer,
right_offset_in_bits: usize,
len_in_bits: usize| {
bitwise_bin_op_helper(
left,
left_offset_in_bits,
right,
right_offset_in_bits,
len_in_bits,
|a, b| !(a ^ b),
Copy link
Contributor Author

Choose a reason for hiding this comment

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

This works per pair of u64 instead of per bool.

)
},
)
}

/// Perform `left != right` operation on [`BooleanArray`]
fn neq_bool(left: &BooleanArray, right: &BooleanArray) -> Result<BooleanArray> {
binary_boolean_kernel(
left,
right,
|left: &Buffer,
left_offset_in_bits: usize,
right: &Buffer,
right_offset_in_bits: usize,
len_in_bits: usize| {
bitwise_bin_op_helper(
Copy link
Member

Choose a reason for hiding this comment

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

How about using bitwise_bin_op_simd_helper?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

That only should be used together with a simd feature / implementation.
Certainly possible, but could be implemented as follow up work.

left,
left_offset_in_bits,
right,
right_offset_in_bits,
len_in_bits,
|a, b| (a ^ b),
)
},
)
}

/// Perform `left == right` operation on [`BooleanArray`] and a scalar
fn eq_bool_scalar(left: &BooleanArray, right: bool) -> Result<BooleanArray> {
let len = left.len();
let left_offset = left.offset();

let values = if right {
left.values().bit_slice(left_offset, len)
} else {
buffer_unary_not(left.values(), left.offset(), left.len())
};

let data = unsafe {
ArrayData::new_unchecked(
Copy link
Contributor Author

Choose a reason for hiding this comment

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

Not sure if there is a more elegant way of doing this.

DataType::Boolean,
len,
None,
left.data_ref()
.null_bitmap()
.as_ref()
.map(|b| b.bits.bit_slice(left_offset, len)),
0,
vec![values],
vec![],
)
};

Ok(BooleanArray::from(data))
}

/// Perform `left != right` operation on [`BooleanArray`] and a scalar
fn neq_bool_scalar(left: &BooleanArray, right: bool) -> Result<BooleanArray> {
let len = left.len();
Copy link
Contributor

Choose a reason for hiding this comment

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

I wonder if this could simply be a call to eq_bool_scalar(left, !right)?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Good idea!

let left_offset = left.offset();

let values = if right {
buffer_unary_not(left.values(), left.offset(), left.len())
} else {
left.values().bit_slice(left_offset, len)
};

let data = unsafe {
ArrayData::new_unchecked(
DataType::Boolean,
len,
None,
left.data_ref()
.null_bitmap()
.as_ref()
.map(|b| b.bits.bit_slice(left_offset, len)),
0,
vec![values],
vec![],
)
};

Ok(BooleanArray::from(data))
}
/// Perform `left != right` operation on [`StringArray`] / [`LargeStringArray`].
pub fn neq_utf8<OffsetSize: StringOffsetSizeTrait>(
left: &GenericStringArray<OffsetSize>,
Expand Down Expand Up @@ -1252,6 +1354,67 @@ mod tests {
);
}

#[test]
fn test_boolean_array_eq() {
Copy link
Contributor

Choose a reason for hiding this comment

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

Normally since there is special case handling for processing these at 64 bits at a time, I would suggest tests that have at least 65 entries to test the boundary conditions. However, in this case as it is using a pre-existing helper method bitwise_bin_op_helper I am not sure such extra coverage is needed

let a: BooleanArray =
vec![Some(true), Some(false), Some(false), Some(true), Some(true), None]
.into();
let b: BooleanArray =
vec![Some(true), Some(true), Some(false), Some(false), None, Some(false)]
.into();

let res: Vec<Option<bool>> = eq_bool(&a, &b).unwrap().iter().collect();

assert_eq!(
res,
vec![Some(true), Some(false), Some(true), Some(false), None, None]
)
}

#[test]
fn test_boolean_array_neq() {
let a: BooleanArray =
vec![Some(true), Some(false), Some(false), Some(true), Some(true), None]
.into();
let b: BooleanArray =
vec![Some(true), Some(true), Some(false), Some(false), None, Some(false)]
.into();

let res: Vec<Option<bool>> = neq_bool(&a, &b).unwrap().iter().collect();

assert_eq!(
res,
vec![Some(false), Some(true), Some(false), Some(true), None, None]
)
}

#[test]
fn test_boolean_array_eq_scalar() {
let a: BooleanArray = vec![Some(true), Some(false), None].into();

let res1: Vec<Option<bool>> = eq_bool_scalar(&a, false).unwrap().iter().collect();

assert_eq!(res1, vec![Some(false), Some(true), None]);

let res2: Vec<Option<bool>> = eq_bool_scalar(&a, true).unwrap().iter().collect();

assert_eq!(res2, vec![Some(true), Some(false), None]);
}

#[test]
fn test_boolean_array_neq_scalar() {
let a: BooleanArray = vec![Some(true), Some(false), None].into();

let res1: Vec<Option<bool>> =
neq_bool_scalar(&a, false).unwrap().iter().collect();

assert_eq!(res1, vec![Some(true), Some(false), None]);

let res2: Vec<Option<bool>> = neq_bool_scalar(&a, true).unwrap().iter().collect();

assert_eq!(res2, vec![Some(false), Some(true), None]);
}

#[test]
fn test_primitive_array_lt() {
cmp_i64!(
Expand Down