Unfortunately there appears to be no nice way of doing this. I've seen others achieve a work around by UNIONing with a table of the same schema, containing a single row of the header names, and then finally sorting by a synthesised rank column (see: http://stackoverflow.com/a/25214480/74772).
I believe headers should really be an option on INSERT OVERWRITE DIRECTORY as it is often a far neater way of generating reports for third party consumption. Additionally it is not always possible to elegantly capture the headers generated by the CLI option in some environments such as Oozie scheduled tasks. There are arguments against including headers in such Hive generated datasets. However, these datasets are often reports at the end of a pipeline whose format is mandated by the intended third party recipient. It seems a shame to introduce yet another tool into the pipeline purely to introduce a header row, we'd rather just do so from within Hive. There is some background information on the current header implementation here: https://issues.apache.org/jira/browse/HIVE-138 Cheers - Elliot. On Wednesday, 13 January 2016, Sreenath <sreenaths1...@gmail.com> wrote: > Hey, > > This will work but lets say i want to write the output to an HDFS location > using INSERT OVERWRITE DIRECTORY '<Query>' , in this case even if we set > *hive.cli.print.header=true *the headers doesn't get written . Is there a > way to write the headers in this case > > On 13 January 2016 at 12:04, Ankit Bhatnagar <ank...@yahoo-inc.com> wrote: > >> r u looking for >> >> hive -e "*set hive.cli.print.header=true*; < query> " > output >> >> >> On Tuesday, January 12, 2016 10:14 PM, Sreenath <sreenaths1...@gmail.com> >> wrote: >> >> >> Hi All, >> >> Is there a way we can write the hive column headers also along with the >> output when we are overwriting a query's output to an HDFS or local >> directory ? >> >> >> -- >> Sreenath S Kamath >> Bangalore >> Ph No:+91-9590989106 >> >> >> > > > -- > Sreenath S Kamath > Bangalore > Ph No:+91-9590989106 >
