[datafusion-spark] [SQL] [TEST] IntervalMonthDayNano(0,0,0) give line blank #17455

Activity

1. davidlghellin

   added

   bugSomething isn't workingSomething isn't working

   on Sep 6, 2025
2. Jefffrey commented on Sep 17, 2025

   

   Jefffrey

   on Sep 17, 2025

   Last edited by Jefffrey

   Contributor

   More actions

   In DataFusion itself, it seems an empty MonthDayNano interval is displayed as an empty string.

   e.g. this code generates this output:

       let a = ctx.sql("select make_interval()").await?.collect().await?;
       dbg!(a);
       ctx.sql("select make_interval()").await?.show().await?;

   • Using this PR branch feat(spark): implement Spark make_interval function #17424 to get an empty MonthDayNano interval

   Output:

   [datafusion-examples/examples/regexp.rs:42:5] a = [
       RecordBatch {
           schema: Schema {
               fields: [
                   Field {
                       name: "make_interval()",
                       data_type: Interval(
                           MonthDayNano,
                       ),
                       nullable: false,
                       dict_id: 0,
                       dict_is_ordered: false,
                       metadata: {},
                   },
               ],
               metadata: {},
           },
           columns: [
               PrimitiveArray<Interval(MonthDayNano)>
               [
                 IntervalMonthDayNano {
                   months: 0,
                   days: 0,
                   nanoseconds: 0,
               },
               ],
           ],
           row_count: 1,
       },
   ]
   +-----------------+
   | make_interval() |
   +-----------------+
   |                 |
   +-----------------+
   

   This seems to be because in arrow-rs, if all values are 0 then nothing is written, see:

   https://github.com/apache/arrow-rs/blob/2ec77b548d9242bc85ca36f04761d116ceab8963/arrow-cast/src/display.rs#L733-L758

   • Related issue: Improve the display of interval types arrow-rs#5914

   I'll assume this is an arrow-rs issue and mark as waiting upstream, unless the SLT tests do output formatting of values elsewhere
3. Jefffrey

   added

   waiting-on-upstreamPR is waiting on an upstream dependency to be updatedPR is waiting on an upstream dependency to be updated

   on Sep 17, 2025
4. github-merge-queue

   added a commit that references this issue on Feb 28, 2026

   Upgrade DataFusion to arrow-rs/parquet 58.0.0 / `object_store` 0.13.0 (…

   ...

   73fbd48
5. alamb

   closed this as completedin #19728on Feb 28, 2026
6. de-bgunter

   added a commit that references this issue on Mar 24, 2026

   Upgrade DataFusion to arrow-rs/parquet 58.0.0 / `object_store` 0.13.0 (…

   ...

   9bdc5ec