You can try concat_ws(' ', map_keys(UNION_MAP(MAP(your_column, 'dummy'))))
as mentioned in https://issues.apache.org/jira/browse/HIVE-707On Thu, Jan 16, 2014 at 9:13 AM, Sambit Tripathy (RBEI/PJ-NBS) < [email protected]> wrote: > Hi Chalcy, > > > > I am using the group_concat function in my query and that actually puts > all columns in the memory and I am afraid Hive does not have this feature. > > > > > > Regards, > > Sambit. > > > > *From:* Chalcy [mailto:[email protected]] > *Sent:* Thursday, January 16, 2014 7:19 PM > *To:* [email protected] > > *Subject:* Re: Joins in Sqoop > > > > Hi Sambit, > > > > I would import all the relevant tables into hive and then do the join > there if you have enough space in the hadoop cluster. > > > > Hope this helps, > > Chalcy > > > > On Thu, Jan 16, 2014 at 8:20 AM, Sambit Tripathy (RBEI/PJ-NBS) < > [email protected]> wrote: > > Hi, > > > > I have written query which has 5 Join clauses and I am passing this query > in Sqoop import. > > > > Problem: This produces a large temp file in the MySQL server temp > directory and throws back an error saying No Space left on the device. Yes > this can be fixed by increasing the size of the temp directory in the MySQL > server, but what if you actually don’t have any space left on MySQL server. > Are there any workarounds for this? I mean something like a batch import > which does not create a big temp file in the server. > > > > > > Regards, > > Sambit. > > > > > -- -- JChan
