You are viewing a plain text version of this content. The canonical link for it is here.
Posted to jira@arrow.apache.org by "Andrew Lamb (Jira)" <ji...@apache.org> on 2021/03/05 21:51:00 UTC

[jira] [Created] (ARROW-11882) [Rust] Implement Debug printing "kernel"

Andrew Lamb created ARROW-11882:
-----------------------------------

             Summary: [Rust] Implement Debug printing "kernel"
                 Key: ARROW-11882
                 URL: https://issues.apache.org/jira/browse/ARROW-11882
             Project: Apache Arrow
          Issue Type: Improvement
          Components: Rust
            Reporter: Andrew Lamb


[~jorgecarleitao] offered a great way to improve the Debug/Display implementations for various Array implementations on https://github.com/apache/arrow/pull/9624#issuecomment-790976766 

The only reason we are implementing to_isize/to_usize on NativeType is because we have a function to represent an array (for Display) that accepts a generic physical type T, and then tries to convert it to a isize depending on a logical type (DataType::Date). However, there is already a Many to one relationship between logical and physical types.

Thus, a solution for this is to have the `Debug` function branch off depending on the (logical) datatype, implementing the custom string representation depending on it, instead of having a loop of native type T and then branching off according to the DataType inside the loop.

I.e. instead of

{code}
for i in ... {
   match data_type {
         DataType::Date32 => represent array[i] as date
         DataType::Int32 => represent array[i] as int
   }
}
{code}

imo we should have

{code}
match data_type {
     DataType::Date32 => for i in ... {represent array[i] as date}
     DataType::Int32 => for i in ... {represent array[i] as int}
}
{code}

i.e. treat the Display as any other "kernel", where behavior is logical, not physical, type-dependent.



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