[jira] [Commented] (HIVE-7068) Integrate AccumuloStorageHandler

["Brian Femiano (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HIVE-7068?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13998937#comment-13998937 ] 

Brian Femiano commented on HIVE-7068:
-------------------------------------

These won't all get done in the next couple weeks, but some ideas:

Enhancements: 

    Support INSERT.
    Support for fixed timestamp on INSERT mutations.
    Support for configurable authorizations on SELECT.
    Support for configurable timestamp on scan.
    Optional type hints for qualifier-value mapping.
    Automatic NULL casting for key-value pairs where the type hint, or if absent the corresponding Hive column data type cannot be correctly applied to the byte[] value.
    Revisit the possibility for UDFLike predicate pushdown in the latest Hive 0.13 release.
    Revisit the possibility for disjunctive predicate pushdown in the latest Hive 0.13 release.
    Support for TinyInt, Small Int, Float, Date, Timestamp, and Binary types.
    Ability to run scans over cloned tables for isolation. 
    Investigate JOIN pushdown.
    Investigate GROUPBY pushdown.
    Support for creating views.
    Support for transactions (potentially with Conditional Mutations?) 
 
Bug fixes:
    *Merge various fixes done across forked github branches back into master.  Many of these were configuration adjustments to make it compatible with various Hadoop distributions. 
    *Issue when doing any join other than full outer.
    *Major issue with predicate constant decoding that causes incorrect results from many queries
    *Support for Hadoop 2.0/CHD4.x
    *Test JOINS involving Hive managed tables.

 

> Integrate AccumuloStorageHandler
> --------------------------------
>
>                 Key: HIVE-7068
>                 URL: https://issues.apache.org/jira/browse/HIVE-7068
>             Project: Hive
>          Issue Type: New Feature
>            Reporter: Josh Elser
>
> [Accumulo|http://accumulo.apache.org] is a BigTable-clone which is similar to HBase. Some [initial work|https://github.com/bfemiano/accumulo-hive-storage-manager] has been done to support querying an Accumulo table using Hive already. It is not a complete solution as, most notably, the current implementation presently lacks support for INSERTs.
> I would like to polish up the AccumuloStorageHandler (presently based on 0.10), implement missing basic functionality and compare it to the HBaseStorageHandler (to ensure that we follow the same general usage patterns).
> I've also been in communication with [~bfem] (the initial author) who expressed interest in working on this again. I hope to coordinate efforts with him.



--
This message was sent by Atlassian JIRA
(v6.2#6252)