Hello
I use Hive on Spark and have an issue with assigning several aliases to the
output (several return values) of an UDF. I ran in several issues and ended
up with a workaround (described at the end of this message).
- Is assigning several aliases to the output of an UDF not supported by
Spark SQL yet?
- Is there a smarter solution than the one I ended up with finally - see
(3)?
1)
The query with following syntax is rejected due to the assigment of multiple
aliases.
Query
--------
SELECT my_function(param_one, param_two) AS (return_one, return_two,
return_three)
FROM my_table;
Error
--------
Unsupported language features in query: SELECT my_function(param_one,
param_two) AS (return_one, return_two, return_three)
FROM my_table;
TOK_QUERY
TOK_FROM
TOK_TABREF
TOK_TABNAME
my_table
TOK_SELECT
TOK_SELEXPR
TOK_FUNCTION
my_function
TOK_TABLE_OR_COL
param_one
TOK_TABLE_OR_COL
param_two
return_one
return_two
return_three
2)
Because of this error I searched for a way to avoid assigning multiple
aliases to the UDF. I ended up having the following query and encountered
another error/issue.
Note: This error only occurs when having "c_0" in select clause. Only
selecting "c_1" and "c_2" works fine.
Query
--------
SELECT return.c_0 AS return_one, return.c_1 AS return_two, return.c_2 AS
return_three FROM (SELECT my_function(param_one, param_two) FROM my_table)
return;
Error
--------
java.lang.RuntimeException: Couldn't find c_0#504 in
[c_0#521L,c_1#522,c_2#523]
3)
My final (working) workaround is wrapping the actual query (the one with the
UDF) with an additional select statement.
Query
--------
SELECT result.c_0 AS return_one, result.c_1 AS return_two, result.c_2 AS
return_three FROM(SELECT * FROM (SELECT my_function(param_one, param_two)
FROM my_table) return) result;
Error
--------
No error :)
Kind regards
Max
--
View this message in context:
http://apache-spark-user-list.1001560.n3.nabble.com/Spark-SQL-Assigning-several-aliases-to-the-output-several-return-values-of-an-UDF-tp21236.html
Sent from the Apache Spark User List mailing list archive at Nabble.com.
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
