[ 
https://issues.apache.org/jira/browse/ARROW-9503?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17159145#comment-17159145
 ] 

Chao Sun commented on ARROW-9503:
---------------------------------

I think its because 
[{{comparison.rs}}|https://github.com/apache/arrow/blob/master/rust/arrow/src/compute/kernels/comparison.rs#L61]
 uses the offset of the input slice as the result array offset:

{code}
/// Helper function to perform boolean lambda function on values from two 
arrays, this
/// version does not attempt to use SIMD.
macro_rules! compare_op {
    ($left: expr, $right:expr, $op:expr) => {{
        if $left.len() != $right.len() {
            return Err(ArrowError::ComputeError(
                "Cannot perform comparison operation on arrays of different 
length"
                    .to_string(),
            ));
        }

        let null_bit_buffer = apply_bin_op_to_option_bitmap(
            $left.data().null_bitmap(),
            $right.data().null_bitmap(),
            |a, b| a & b,
        )?;

        let mut result = BooleanBufferBuilder::new($left.len());
        for i in 0..$left.len() {
            result.append($op($left.value(i), $right.value(i)))?;
        }

        let data = ArrayData::new(
            DataType::Boolean,
            $left.len(),
            None,
            null_bit_buffer,
            $left.offset(),
            vec![result.finish()],
            vec![],
        );
        Ok(PrimitiveArray::<BooleanType>::from(Arc::new(data)))
    }};
}
{code}

Seems it should always use 0 as result array offset, as it constructs a new 
boolean array and return it here. 

> Comparison sliced arrays is wrong
> ---------------------------------
>
>                 Key: ARROW-9503
>                 URL: https://issues.apache.org/jira/browse/ARROW-9503
>             Project: Apache Arrow
>          Issue Type: Bug
>          Components: Rust
>            Reporter: Ritchie
>            Priority: Major
>
> Comparison of arrow arrays where one is sliced is dependent on the order of 
> comparison, and doesn't always yield correct results.
> The following minimal example shows the differing results.
> {code:rust}
>         use arrow::{
>             datatypes::Int32Type,
>             compute,
>             array::{
>                 Array,
>                 PrimitiveBuilder,
>                 PrimitiveArray
>             }
>         };
>         let mut builder = PrimitiveBuilder::new(10);
>         for v in 0..10 {
>             builder.append_value(v).unwrap()
>         }
>         let a: PrimitiveArray<Int32Type> = builder.finish();
>         let mut builder = PrimitiveBuilder::new(10);
>         for v in 5..10 {
>             builder.append_value(v).unwrap()
>         }
>         let b: PrimitiveArray<Int32Type> = builder.finish();
>         // returns Array trait
>         let sliced_a = a.slice(5, 5);
>         // Downcast to PrimitiveArray
>         let sliced_a = sliced_a.as_any().downcast_ref().unwrap();
>         println!("{:?}", a.slice(5, 5));
>         println!("{:?}", b);
>         println!("{:?}", compute::eq(sliced_a, &b));
>         println!("{:?}", compute::eq(&b, sliced_a))
> {code}
> This prints:
> {code:text}
> PrimitiveArray<Int32>
> [
>   5,
>   6,
>   7,
>   8,
>   9,
> ]
> PrimitiveArray<Int32>
> [
>   5,
>   6,
>   7,
>   8,
>   9,
> ]
> Ok(PrimitiveArray<Boolean>
> [
>   false,
>   false,
>   false,
>   false,
>   false,
> ])
> Ok(PrimitiveArray<Boolean>
> [
>   true,
>   true,
>   true,
>   true,
>   true,
> ])
> {code}
> I would expect both comparison arrays to evaluate to true. This same effect 
> is also occurring with utf8arrays.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

Reply via email to