hive-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Raghotham Murthy (JIRA)" <>
Subject [jira] Commented: (HIVE-590) Pass type information in genFileSinkPlan
Date Thu, 02 Jul 2009 17:51:47 GMT


Raghotham Murthy commented on HIVE-590:

In most cases, HiveServer is called to get information about SELECT queries which dont create
tables. Also, these transient query results dont have partitionkeys, bucketcols, sortcols
etc. So a Table object seems to be an overkill. 

I do agree that we should support a getTableObject or some such metastore api call which returns
the Table object for a table in the warehouse. This call can then be used instead of doing
a 'describe extended' query on a table and parsing that blob that gets returned.

> Pass type information in genFileSinkPlan
> ----------------------------------------
>                 Key: HIVE-590
>                 URL:
>             Project: Hadoop Hive
>          Issue Type: Bug
>          Components: Query Processor
>            Reporter: Raghotham Murthy
>            Assignee: Namit Jain
>             Fix For: 0.4.0
>         Attachments: hive.590.1.patch, hive.590.2.patch
> Right now only column names are being passed between semanticanalyzer and fetchtask.
Once type information is passed, we can use LazySerDe to serialize the data (into json) in
> Driver.getSchema() should then return a new thrift type ResultSchema instead of String:
> {code}
> struct ResultSchema {
>   // column names, types, comments
>  1: list<hive_metastore.FieldSchema> fieldSchemas,
>  // delimiters etc
>  2: map<string, string> properties
> }
> {code}
> Once this is done, the jdbc client can instantiate a simplified serde from the ResultSchema
and parse the query results.

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

View raw message