For writing I am using a VectorizedRowBatch and all the vectors in org.apache.hadoop.hive.ql.exec.vector.*. For reading (in my unit tests) I am using OrcFile.createReader().
https://github.com/apache/nifi/pull/477/files#diff-e6d9b9ac010d3b7a0f50ae90a46136d2R151 The first record has a 10 for the first column (union of integer and null), the second has a null for the first column. My test right now just validates the first record. If I check isNull() for the first column of the second record, it says false. I haven't had a chance to try RC3 but plan to do so shortly. Thanks, Matt On Wed, Jun 8, 2016 at 12:28 PM, Owen O'Malley <[email protected]> wrote: > Matt, > So you were using the org.apache.orc.mapred API? Was the union value > null or was the contained value null? Can you use orc-contents to dump the > file's data? > > .. Owen > > On Thu, Jun 2, 2016 at 3:54 PM, Matt Burgess <[email protected]> wrote: > >> I am out of town through Sunday so I won't be able to give the RC a try >> before the vote closes, but I've had a good experience with the snapshot so >> far (the Java part) so I'm looking forward to the release! >> >> One thing I did see, but couldn't characterize enough to file a Jira, was >> the setting of a null value into a union column. While the isNull entry >> appeared to be set correctly in memory, when i wrote the file then read it >> back in, the isNull entry was false (when it was true right before >> writing). I will dig in when I get back; so far all observed errors have >> been my own :) >> >> Regards, >> Matt >> >> > On Jun 2, 2016, at 6:21 PM, Owen O'Malley <[email protected]> wrote: >> > >> > Ok, I addressed Prasanth's points and have rolled a new RC. >> > >> > Artifacts: http://home.apache.org/~omalley/orc-1.1.0/ >> > Git tag: https://github.com/apache/orc/releases/tag/release-1.1.0rc1 >> > >> > Voting will conclude in 72 hours. >> > >> > Thanks, >> > Owen >>
