[
https://issues.apache.org/jira/browse/HIVE-14158?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15365369#comment-15365369
]
Ashutosh Chauhan commented on HIVE-14158:
-----------------------------------------
Choice of using _c1, _c2 and so on for unnamed columns looked little
unintuitive to me, so I did little experimentation.
I tried following on mysql:
{code}
create table t1 (a int, b char(10));
insert into t1 values(10, 'fg');
create view v1 as select a,10,10 from t1;
describe v1;
COLUMN_NAME
a
10
My_exp_10
{code}
and then tried following on postgres
{code}
create table t1 (a int, b char(10));
insert into t1 values(10, 'fg');
create view v1 as select a,10,20 as a2 from t1;
drop view v1;
create view v1 as select a,10,20 from t1;
ERROR: column "?column?" specified more than once
{code}
>From above it seems mysql uses column's constant value as column name until
>there is a collision and then picks name as per its own convention. Postgres
>uses '?column?' as column name and then throws exception when there are
>multiple unnamed columns. So, choice of _cN by Hive doesn't look that
>arbitrary given what others are doing : ) I don't have oracle handy.
>Pengcheng, can you see what Oracle does in this case.
> deal with derived column names
> ------------------------------
>
> Key: HIVE-14158
> URL: https://issues.apache.org/jira/browse/HIVE-14158
> Project: Hive
> Issue Type: Sub-task
> Components: Security
> Reporter: Pengcheng Xiong
> Assignee: Pengcheng Xiong
> Fix For: 2.1.0
>
> Attachments: HIVE-14158.01.patch, HIVE-14158.02.patch,
> HIVE-14158.03.patch
>
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
