this is an unfortunate issue with hive we only support mapred api not mapreduce. If you look at the hbase and cassandra support you see that sonetimes you can wrap a mapreduce input format by extending the new interface and implementing the old one. So it is possible to use a newer api but its ugly code wise. Cleaner would be to produce two distinct input formats.
On Tuesday, December 4, 2012, Mark Grover <grover.markgro...@gmail.com> wrote: > Yong, > In my opinion, #2 wouldn't be that hard to do. > Of course, upgrading to the new API (albeit by updating Hive) would the right thing to do long term. > Mark > On Mon, Dec 3, 2012 at 7:42 AM, java8964 java8964 <java8...@hotmail.com> wrote: >> >> Hi, >> Our company current is using CDH3 release, which comes with Hive 0.7.1. >> Right now, I have the data coming from another team, which also provides the custom InputFormat and RecorderReader, but using the new mapreduce API. >> I am trying to build a hive table on these data, and hope I can reuse the existing InputFormat. But I got a problem, that in HIVE 0.7.x, only old API is supported. >> I have 3 options now: >> 1) Upgrade whole hadoop system,which is not a real option as right now. >> 2) Ask the other team to provide the custom InputFormat/RecorderReader under old mapred API. I am working on this, but not sure if it is possible or how long. >> 3) Writable a wrapper class, providing old API based on the new API class. My question is, is the 3rd optional valid? Does anyone do the similar things before? >> >> Thanks >> Yong >
