[
https://issues.apache.org/jira/browse/SPARK-38130?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Steven Aerts updated SPARK-38130:
---------------------------------
Description:
{{array_sort}} has check to see if the entries it has to sort are orderable.
I think this check should be removed. Because even entries which are not
orderable can have a lambda function which makes them orderable.
{code:java}
Seq((Array[Map[String, Int]](Map("a" -> 1), Map()), "x")).toDF("a",
"b").selectExpr("array_sort(a, (x,y) -> cardinality(x) - cardinality(y))"){code}
fails with:
{code:java}
org.apache.spark.sql.AnalysisException: cannot resolve 'array_sort(`a`,
lambdafunction((cardinality(namedlambdavariable()) -
cardinality(namedlambdavariable())), namedlambdavariable(),
namedlambdavariable()))' due to data type mismatch: array_sort does not support
sorting array of type map<string,int> which is not orderable {code}
While the case where this check is relevant, fails with a different error which
is triggered earlier in the code path:
{code:java}
> Seq((Array[Map[String, Int]](Map("a" -> 1), Map()), "x")).toDF("a",
> "b").selectExpr("array_sort(a)"){code}
Fails with:
{code:java}
org.apache.spark.sql.AnalysisException: cannot resolve '(namedlambdavariable()
< namedlambdavariable())' due to data type mismatch: LessThan does not support
ordering on type map<string,int>; line 1 pos 0;
{code}
was:
{{array_sort}} has check to see if the entries it has to sort are orderable.
I think this check should be removed. Because even entries which are not
orderable can have a lambda function which makes them orderable.
{code:java}
> Seq((Array[Map[String, Int]](Map("a" -> 1), Map()), "x")).toDF("a",
> "b").selectExpr("array_sort(a, (x,y) -> cardinality(x) -
> cardinality(y))")org.apache.spark.sql.AnalysisException: cannot resolve
> 'array_sort(`a`, lambdafunction((cardinality(namedlambdavariable()) -
> cardinality(namedlambdavariable())), namedlambdavariable(),
> namedlambdavariable()))' due to data type mismatch: array_sort does not
> support sorting array of type map<string,int> which is not orderable {code}
> array_sort does not allow non-orderable datatypes
> -------------------------------------------------
>
> Key: SPARK-38130
> URL: https://issues.apache.org/jira/browse/SPARK-38130
> Project: Spark
> Issue Type: Bug
> Components: SQL
> Affects Versions: 3.2.1
> Environment:
> Reporter: Steven Aerts
> Priority: Major
>
> {{array_sort}} has check to see if the entries it has to sort are orderable.
> I think this check should be removed. Because even entries which are not
> orderable can have a lambda function which makes them orderable.
> {code:java}
> Seq((Array[Map[String, Int]](Map("a" -> 1), Map()), "x")).toDF("a",
> "b").selectExpr("array_sort(a, (x,y) -> cardinality(x) -
> cardinality(y))"){code}
> fails with:
> {code:java}
> org.apache.spark.sql.AnalysisException: cannot resolve 'array_sort(`a`,
> lambdafunction((cardinality(namedlambdavariable()) -
> cardinality(namedlambdavariable())), namedlambdavariable(),
> namedlambdavariable()))' due to data type mismatch: array_sort does not
> support sorting array of type map<string,int> which is not orderable {code}
> While the case where this check is relevant, fails with a different error
> which is triggered earlier in the code path:
> {code:java}
> > Seq((Array[Map[String, Int]](Map("a" -> 1), Map()), "x")).toDF("a",
> > "b").selectExpr("array_sort(a)"){code}
> Fails with:
> {code:java}
> org.apache.spark.sql.AnalysisException: cannot resolve
> '(namedlambdavariable() < namedlambdavariable())' due to data type mismatch:
> LessThan does not support ordering on type map<string,int>; line 1 pos 0;
> {code}
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
