[
https://issues.apache.org/jira/browse/PHOENIX-3504?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15685539#comment-15685539
]
Josh Mahonin commented on PHOENIX-3504:
---------------------------------------
Hi [~sergey.soldatov]
The behaviour has been in since PHOENIX-2288 was resolved with this commit:
https://github.com/apache/phoenix/commit/9343743157f1cc03b1b8b815289b6127a30d740f
As I recall, those specific values of (38, 18) for precision and scale were
chosen since that's what Spark was using as a default.
The patch looks good, although it might be good to retain the previous bounds
check as well, e.g.:
{noformat}
if (columnInfo.getPrecision == null || columnInfo.getPrecision < 0)
{noformat}
> Spark integration doesn't work with decimal columns that are using default
> precision
> ------------------------------------------------------------------------------------
>
> Key: PHOENIX-3504
> URL: https://issues.apache.org/jira/browse/PHOENIX-3504
> Project: Phoenix
> Issue Type: Bug
> Affects Versions: 4.8.0
> Reporter: Sergey Soldatov
> Assignee: Sergey Soldatov
> Attachments: PHOENIX-3504-1.patch
>
>
> Not sure when this issue was introduced and whether this code was working
> well before, but in PhoenixRDD.phoenixTypeToCatalystType for decimal
> precision we have a check
> (columnInfo.getPrecision < 0)
> which is fail for decimal columns that were created with default precision
> and scale because precision is null in this case.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
