Ok mapreduce is the engine, but I am talking about the file format. Even with Tez or Spark as an engine the fileformat will always be built on mapped.* and/or mapreduce.* Aside from hive this is also the case in Spark and Flink to leverage the benefits to access different storages (S3, HDFS etc) and file formats transparently.
> On 13. Sep 2017, at 20:53, Alan Gates <alanfga...@gmail.com> wrote: > > I’m not aware of any plans in Hive to do any more work that uses Map Reduce > as the execution engine, so I expect Hive will continue to use mapred. > > Alan. > >> On Wed, Sep 13, 2017 at 4:25 AM, Jörn Franke <zuinn...@gmail.com> wrote: >> >> Dear all, >> >> I have developed several custom input formats (e.g. for the Bitcoin >> blockchain) including a HiveSerde, which are open source. >> I plan to develop for my HadoopOffice inputformat also a HiveSerde, but I >> wonder if I should continue to use mapred.* apis or if i should use >> mapreduce.* >> >> My inputformats support both APIs, but it seems that Hive is one of the >> last (please correct me here) to use the mapred.* API >> >> Personally, i have no preference, since I support both. >> >> Thank you. >> >> All the best >>
