Seems to me that returning null makes sense, when Drill is dealing with schema-less input.
Some rows may have the 12-th column, while some rows do not. In such case, user might expect Drill to return the 12-th column if it does exist, or null if not. On Wed, Jul 2, 2014 at 11:35 AM, Alexander Zarei (JIRA) <[email protected]> wrote: > Alexander Zarei created DRILL-1100: > -------------------------------------- > > Summary: Returning NULL instead of array-index-out-of-bound > error > Key: DRILL-1100 > URL: https://issues.apache.org/jira/browse/DRILL-1100 > Project: Apache Drill > Issue Type: Bug > Components: Client - JDBC, Client - ODBC > Environment: SqlLine in Linux and C# in Windows > Reporter: Alexander Zarei > Priority: Minor > > > When an out of bound column is queried, for example, > > SELECT columns[12] as column_12 FROM `dfs`.`root`.`./sneaky.csv`; > > the result is a column with all null values: > > +------------+ > | column_12 | > +------------+ > | null | > | null | > | null | > | null | > | null | > +------------+ > > However, it makes more sense to return an error instead. > > The file content: > > > someDirectoryFoo,true,false,false,100,root,root,"2013-10-11T14:31:58.000-0700","2013-10-11T14:31:58.000-0700","-rw-r--r--." > > someDirectoryBar,true,false,false,100,root,root,"2013-10-11T14:31:58.000-0700","2013-10-11T14:31:58.000-0700","-rw-r--r--." > > aFile,false,true,false,1234,root,root,"2013-10-11T14:31:58.000-0700","2013-10-11T14:31:58.000-0700","-rw-r--r--." > > anotherFile,false,true,false,2345,root,root,"2012-11-11T14:31:58.000-0700","2012-11-11T14:31:58.000-0700","-rw-r--r--." > > yetAnotherFile,false,true,false,3456,root,root,"2011-11-11T14:31:58.000-0700","2012-11-11T14:31:58.000-0700","-rw-r--r--." > > > > -- > This message was sent by Atlassian JIRA > (v6.2#6252) >
