[
https://issues.apache.org/jira/browse/HIVE-12878?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15136224#comment-15136224
]
Matt McCline commented on HIVE-12878:
-------------------------------------
Patch submitted is an experiment:
Changed the default on these environment variables (temporarily) to force
vectorization of many queries.
{code}
hive.fetch.task.conversion=none
hive.vectorized.execution.enabled=true
{code}
New environment variables are set so that all vectorized queries either use the
new vectorized versions of deserialize for LazySimple (i.e. TEXTFILE) and
LazyBinarySerDe. Or, we deserialize row-by-row to fill up VectorizedRowBatch.
{code}
hive.vectorized.use.vectorized.input.format=false
hive.vectorized.use.vector.serde.deserialize=true
hive.vectorized.use.row.serde.deserialize=true
{code}
So, MapWork tasks not vectorizing due to input file format should not happen
(except for ACID which only is permitted for vectorized input format...).
> Support Vectorization for TEXTFILE and other formats
> ----------------------------------------------------
>
> Key: HIVE-12878
> URL: https://issues.apache.org/jira/browse/HIVE-12878
> Project: Hive
> Issue Type: New Feature
> Components: Hive
> Reporter: Matt McCline
> Assignee: Matt McCline
> Priority: Critical
> Attachments: HIVE-12878.01.patch
>
>
> Support vectorizing when the input format is TEXTFILE and other formats for
> better Map Vertex performance.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
