[
https://issues.apache.org/jira/browse/HIVE-73?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12649208#action_12649208
]
Raghotham Murthy commented on HIVE-73:
--------------------------------------
@Joy
1. The thrift server is stateful. You can make multiple stateful calls if the
transport is not closed between calls.
2. Right now I am just looking at driver and metastore. I guess the client side
session variables can be via the 'set' command which can be run as another
query. And if the transport is not closed the session variables should persist.
@Zheng
We are planning on supporting multiple fetch calls in the client - fetchOne,
fetchN and fetchAll. Since the server is stateful, we should be able to loop
through result rows one at a time if required. However, we are still debating
if server should also support these methods. We were thinking of just
supporting fetchOne and fetchN in the server and then implement fetchAll in the
client by calling fetchN underneath. That way the server will not be overloaded
in the case of fetchAll.
> Thrift Server and Client for Hive
> ---------------------------------
>
> Key: HIVE-73
> URL: https://issues.apache.org/jira/browse/HIVE-73
> Project: Hadoop Hive
> Issue Type: New Feature
> Affects Versions: 0.19.0
> Reporter: Raghotham Murthy
> Fix For: 0.20.0
>
>
> Currently the hive cli directly calls the driver code. We need to be able to
> run a stand alone hive server that multiple clients can connect to. The hive
> server will allow clients to run queries as well as make meta data calls (by
> inheriting from the thrift metastore server)
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
