[ https://issues.apache.org/jira/browse/IMPALA-1903?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16907565#comment-16907565 ]
Tim Armstrong edited comment on IMPALA-1903 at 8/14/19 7:35 PM: ---------------------------------------------------------------- [~h-vetinari] the context is all on this JIRA. Allowing partitioning by nanosecond timestamps is just too much of a footgun for us to be inclined to support out-of-the-box without any guard-rails - it's just way too easy to explode the number of partitions and have various operational issues as a result. It might be possible to get some consensus on a change if someone came along with a patch that added the support with some guardrails to prevent that. We recently added full DATE support, including partitioning by date, which helps with a lot of use cases. I can see there are probably cases where power users might find partitioning by timestamp to be a useful tool if they're careful with truncation. was (Author: tarmstrong): [~h-vetinari] the context is all on this JIRA. Allowing partitioning by nanosecond timestamps is just too much of a footgun for us to be inclined to support out-of-the-box without any guard-rails - it's just way too easy to explode the number of partitions and have various operational issues as a result. It might be possible to get some consensus on a change if someone came along with a patch that added the support with some guardrails to prevent that. We recently added full DATE support, including partitioning by date, which helps with a lot of use cases. I can see there are probably cases where power users might find partitioning by timestamp to be a useful tool if they're careful with truncation, but > Add support for partitions by timestamp > --------------------------------------- > > Key: IMPALA-1903 > URL: https://issues.apache.org/jira/browse/IMPALA-1903 > Project: IMPALA > Issue Type: New Feature > Components: Frontend > Affects Versions: Impala 2.2 > Reporter: Grant Sohn > Assignee: Jim Apple > Priority: Critical > Labels: ramp-up, sql-language > > Timestamps or some time related parameter is a very common way data is > partitioned. At Yahoo almost all the Hadoop ETL data was partitioned this > way. This should be on par with nested data types as an important feature > for Impala to have. -- This message was sent by Atlassian JIRA (v7.6.14#76016) --------------------------------------------------------------------- To unsubscribe, e-mail: issues-all-unsubscr...@impala.apache.org For additional commands, e-mail: issues-all-h...@impala.apache.org