stoty commented on pull request #21: URL: https://github.com/apache/phoenix-connectors/pull/21#issuecomment-661648359
> As I understand it, this is largely just copy-paste'ing from `phoenix-hive` into a `phoenix-hive3`, and then modifying the pom.xml to pull in the correct dependencies. While this approach is effective, I'm worried about the long-term maintenance of supporting multiple versions of Hive via separate copies of code. > > Maybe we can stop caring about Hive2 soon and we just drop the `phoenix-hive` module entirely? The current Apache connector is actually based on Hive 1. I don't think it'd even work with Hive 2. This patch contains significant code changes to support the V3 API instead of the V1 one. According to @richardantal's research Hive1 and Hive2 is only supported with Hadoop2, while Hive3+ is Hadoop3 only. This in turn means that 4.x can only hope to support Hive1,2, while 5.x can only support Hive 3+. Replacing phoenix-hive would mean that we'd drop hive connector support for 4.x completely. I don't know how large the Hive1 installed base is, and whether it'd be even possible for them to use a recent-ish Phoenix on such an old cluster, so it could be a good thing. While there are significant changes between between the Hive 1 and Hive 2 APIs, the Hive2 and Hive3 APIs seem similar enough that it should be possible to backport this Hive3 connector for Hive2/Phoenix 4.x, and then refactor the Hive3 connector as an extension on top of the new Hive2 connector, though that would either need copying and overwriting the source files, or putting the hive2 and hive3 connectors into different java packages. However, the Hive 2.x and 3.x connectors would need to live in their own their own Hive version specific sub-projects, so I still think that putting this into a new repo is a good idea. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
