[
https://issues.apache.org/jira/browse/SPARK-10501?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15559271#comment-15559271
]
Russell Spitzer commented on SPARK-10501:
-----------------------------------------
It's not that we need it as a unique identifier. It's already a datatype in the
Cassandra database but there is no direct translation to a spark sql type so a
conversion to string must be done. In addition TimeUUIDs require a custom
non-bytewise comparator so a greater than or less than lexical comparison of
them is always incorrect.
https://datastax-oss.atlassian.net/browse/SPARKC-405
> support UUID as an atomic type
> ------------------------------
>
> Key: SPARK-10501
> URL: https://issues.apache.org/jira/browse/SPARK-10501
> Project: Spark
> Issue Type: Improvement
> Components: SQL
> Reporter: Jon Haddad
> Priority: Minor
>
> It's pretty common to use UUIDs instead of integers in order to avoid
> distributed counters.
> I've added this, which at least lets me load dataframes that use UUIDs that I
> can cast to strings:
> {code}
> class UUIDType(AtomicType):
> pass
> _type_mappings[UUID] = UUIDType
> _atomic_types.append(UUIDType)
> {code}
> But if I try to do anything else with the UUIDs, like this:
> {code}
> ratings.select("userid").distinct().collect()
> {code}
> I get this pile of fun:
> {code}
> scala.MatchError: UUIDType (of class
> org.apache.spark.sql.cassandra.types.UUIDType$)
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
