[
https://issues.apache.org/jira/browse/SPARK-26879?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jash Gala updated SPARK-26879:
------------------------------
Description:
In the Spark SQL functions definitions, `inline` uses col1, col2, etc. (i.e.
1-indexed columns), while `stack` uses col0, col1, col2, etc. (i.e. 0-indexed
columns).
{code:title=spark-shell|borderStyle=solid}
scala> spark.sql("SELECT stack(2, 1, 2, 3)").show
+----+----+
|col0|col1|
+----+----+
| 1| 2|
| 3|null|
+----+----+
scala> spark.sql("SELECT inline_outer(array(struct(1, 'a'), struct(2,
'b')))").show
+----+----+
|col1|col2|
+----+----+
| 1| a|
| 2| b|
+----+----+
{code}
This feels like an issue with consistency. As discussed on [PR
#23748|https://github.com/apache/spark/pull/23748], it might be a good idea to
standardize this to something specific (like zero-based indexing) for these and
other similar functions.
was:
In the Spark SQL functions definitions, `inline` uses col1, col2, etc. (i.e.
1-indexed columns), while `stack` uses col0, col1, col2, etc. (i.e. 0-indexed
columns).
{code:title=spark-shell|borderStyle=solid}
scala> spark.sql("SELECT stack(2, 1, 2, 3)").show
| col0 | col1 |
| 1 | 2 |
| 3 | null |
scala> spark.sql("SELECT inline_outer(array(struct(1, 'a'), struct(2,
'b')))").show
| col1 | col2 |
| 1 | a |
| 2 | b |
{code}
This feels like an issue with consistency. As discussed on [PR
#23748|https://github.com/apache/spark/pull/23748], it might be a good idea to
standardize this to something specific (like zero-based indexing) for these and
other similar functions.
> Inconsistency in default column names for functions like inline and stack
> -------------------------------------------------------------------------
>
> Key: SPARK-26879
> URL: https://issues.apache.org/jira/browse/SPARK-26879
> Project: Spark
> Issue Type: Improvement
> Components: SQL
> Affects Versions: 2.4.0
> Reporter: Jash Gala
> Priority: Minor
>
> In the Spark SQL functions definitions, `inline` uses col1, col2, etc. (i.e.
> 1-indexed columns), while `stack` uses col0, col1, col2, etc. (i.e. 0-indexed
> columns).
> {code:title=spark-shell|borderStyle=solid}
> scala> spark.sql("SELECT stack(2, 1, 2, 3)").show
> +----+----+
> |col0|col1|
> +----+----+
> | 1| 2|
> | 3|null|
> +----+----+
> scala> spark.sql("SELECT inline_outer(array(struct(1, 'a'), struct(2,
> 'b')))").show
> +----+----+
> |col1|col2|
> +----+----+
> | 1| a|
> | 2| b|
> +----+----+
> {code}
> This feels like an issue with consistency. As discussed on [PR
> #23748|https://github.com/apache/spark/pull/23748], it might be a good idea
> to standardize this to something specific (like zero-based indexing) for
> these and other similar functions.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
