You could also look at the OrcSerde and how it works. https://github.com/apache/hive/blob/trunk/ql/src/java/org/apache/hadoop/hive/ql/io/orc/OrcSerde.java
Basically, OrcSerde on "serialize" just wraps the row and object inspector in a fake writable. That is passed down to the OutputFormat. On "deserialize" it does the reverse and just passes back the object from the InputFormat. -- Owen On Mon, May 13, 2013 at 6:54 AM, Edward Capriolo <edlinuxg...@gmail.com>wrote: > You need to use a combination of output format and serde, this might allow > you to do something like present struct objects to the input format rather > then Text objects. > > You may want to take a look at the protobuf input format we use: > https://github.com/edwardcapriolo/hive-protobuf/ > > You could reverse the logic here and design an output format. > > > On Mon, May 13, 2013 at 8:14 AM, Rui Martins <ruibmart...@gmail.com>wrote: > >> Hi guys, >> >> I'm currently writing my on HiveOutputFormat as I would like to write the >> output of hive queries into a specific protobuf format my team is using. >> I have managed to do this however, the Writable object I get from Hive as >> a result of a SELECT query is of type Text. This means that I have to split >> the string to find my fields but that's very error prone, specially if some >> fields are strings that may contain spaces. >> >> My question is: >> 1) How do I get a Hive Writable that gives me each field of each result >> row? >> >> Thank you, >> rui >> > >
