[
https://issues.apache.org/jira/browse/HIVE-8488?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14175147#comment-14175147
]
Ashutosh Chauhan commented on HIVE-8488:
----------------------------------------
One thing to keep in mind is effect of this on join/group-by etc. e.g.,
{code}
create table t1 (a string);
create table t2 (b char(500));
create table t3 (c varchar(500));
select * from t1 join t2 on t1.a = t2.b join t3 on t2.b = t3.c;
{code}
Now, columns should match and generate rows. You may want to check with mysql
and/or postgres behavior for this. If this is what we want, you may want to add
testcase for this.
{code}
select a , count (a) from t1 group by a;
alter table t1 change column a a char(500);
select a , count (a) from t1 group by a;
alter table t1 change column a a varchar(500);
select a , count (a) from t1 group by a;
{code}
All of this should return same result. You may want to add tests for these as
well.
Of course, if column length is > 500, than answers should be different but that
would be expected.
> hash() doesn't match between string and char/varchar
> ----------------------------------------------------
>
> Key: HIVE-8488
> URL: https://issues.apache.org/jira/browse/HIVE-8488
> Project: Hive
> Issue Type: Bug
> Components: UDF
> Reporter: Jason Dere
> Assignee: Jason Dere
> Attachments: HIVE-8488.1.patch
>
>
> {noformat}
> hive> select * from tab1;
> OK
> val_484 val_484 val_484
> hive> select hash(c1), hash(c2), hash(c3) from tab1;
> OK
> 230901778 1973712113 1973712113
> {noformat}
> This may throw off users expecting string/varchar/char types to be fairly
> interchangeable.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
