[
https://issues.apache.org/jira/browse/SPARK-9644?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14701926#comment-14701926
]
Pete Robbins commented on SPARK-9644:
-------------------------------------
How much slower is the public API? We would only be creating a small byte
array on the heap then copying.
On deserialization I doubt there is any performance gain to creating an empty
BigInteger then copying in the signum and magnitude array vs simply
constructing the object passing those as parameters.
The main issue is that "fast path" for OpenJDK that you propose may only work
for the current implementation of BigInteger leaving the potential for a jdk
update to break Spark.
It would be possible to have a separate codepath for the current OpenJDK
implementation but I don't think that is desirable.
> Support update DecimalType with precision > 18 in UnsafeRow
> -----------------------------------------------------------
>
> Key: SPARK-9644
> URL: https://issues.apache.org/jira/browse/SPARK-9644
> Project: Spark
> Issue Type: Improvement
> Affects Versions: 1.5.0
> Reporter: Davies Liu
> Assignee: Davies Liu
> Priority: Critical
> Fix For: 1.5.0
>
>
> Currently, we don't support using DecimalType with precision > 18 in new
> unsafe aggregation, it's good to support it.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
