Mailing-List: contact dev-help@hive.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@hive.apache.org
Date: Thu, 5 Dec 2013 21:28:35 +0000 (UTC)
From: "Lenni Kuff (JIRA)" <jira@apache.org>
To: hive-dev@hadoop.apache.org
Message-ID: <JIRA.12682998.1386278839301.80587.1386278915040@arcas>
In-Reply-To: <JIRA.12682998.1386278839301@arcas>
References: <JIRA.12682998.1386278839301@arcas>
Subject: [jira] [Created] (HIVE-5968) Assign (and expose to the client)
 unique object IDs for each metastore object
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit

Lenni Kuff created HIVE-5968:
--------------------------------

             Summary: Assign (and expose to the client) unique object IDs for each metastore object
                 Key: HIVE-5968
                 URL: https://issues.apache.org/jira/browse/HIVE-5968
             Project: Hive
          Issue Type: New Feature
          Components: Database/Schema, Metastore
    Affects Versions: 0.12.0
            Reporter: Lenni Kuff


The Hive Metastore should assign a unique ID to every metastore object - Database, Table, Partition,  etc. These IDs should also be exposed on each of the corresponding thrift structs. There are many cases where this would be useful, one simple case is the following:
hive1> CREATE TABLE Foo;
hive2> DROP TABLE Foo;
hive3> CREATE Table Foo;

Without object ID, there is no good way for the client to differentiate the table created in step1 versus the table created in step3. In general, working with object IDs is much more robust (especially with concurrent operations) than only the object names. With an ID the client can call get_table(object_id) and ensure the table they get back is exactly the one they expect.


--
This message was sent by Atlassian JIRA
(v6.1#6144)