[HIVE-7115] Support a mechanism for running hive locally that doesnt require having a hadoop executable.

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: Testing Infrastructure, Tests
• Assignee: Unassigned
• Reporter: jay vyas
• Created: 22/May/14 15:09
• Updated: 09/Jun/14 21:24

Description

Mapreduce has a "local" mode by default, and likewise, tools such as pig and SOLR do as well,.... maybe we can have a first class local mode for hive also.

For local integration testing of a hadoop app, it would be nice if we could fire up a local hive instance which didnt require "bin/hadoop" for running local jobs. This would allow us to maintain polyglot hadoop applications much easier by incorporating hive into the integration tests. For example:

LocalHiveInstance hive = new LocalHiveInstance();
hive.set("course","crochet")l
hive.runScript("hive_flow.ql")l

Would essentially run a local hive query which mirrors

hive -f hive_flow.ql -hiveconf course=crochet
{noformat{

It seems like thee might be a simple way to do this, at least for small data sets, by putting some kind of alternative (i.e. in memory) execution environment under hive, if one is not already underway ?

Issue Links

• relates to: BIGTOP-1270 BigPetStore: Productionize the Hive portion