[SPARK-21212] Can't use Count(*) with Order Clause

• Type: Bug
• Status: Closed
• Priority: Minor
• Resolution: Not A Problem
• Affects Version/s: 2.1.0
• Component/s: SQL

Environment

Windows; external data provided through data source api

• Assignee: Unassigned
• Reporter: Shawn Lavelle
• Created: 26/Jun/17 14:54
• Updated: 26/Jun/17 23:49
• Resolved: 26/Jun/17 23:49

Description

I don't think this should fail the query:

_Notes: VALUE is a column of table TABLE. columns and table names redacted. I can generate a simplified test case if needed, but this is easy to reproduce. _

jdbc:hive2://user:port/> select count(*) from table where value between 1498240079000 and cast(now() as bigint)*1000 order by value;

Error: org.apache.spark.sql.AnalysisException: cannot resolve '`value`' given input columns: [count(1)]; line 1 pos 113;
'Sort ['value ASC NULLS FIRST], true
+- Aggregate [count(1) AS count(1)#718L]
   +- Filter ((value#413L >= 1498240079000) && (value#413L <= (cast(current_timestamp() as bigint) * cast(1000 as bigint))))
      +- SubqueryAlias table
         +- Relation[field1#411L,field2#412,value#413L,field3#414,field4#415,field5#416,field6#417,field7#418,field8#419,field9#420] com.redacted@16004579 (state=,code=0)

Arguably, the optimizer could ignore the "order by" clause, but I leave that to more informed minds than my own.