[ https://issues.apache.org/jira/browse/IMPALA-6661?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16609491#comment-16609491 ]
Michal Ostrowski commented on IMPALA-6661: ------------------------------------------ For consistency sake, I think we have to have Nan != Nan. [https://github.com/apache/impala/blob/master/docs/shared/impala_common.xml#L2714-L2718] This is turn means that we cannot collapse this group count since that would make the count/group by query inconsistent with a non-aggregating query. > Group by float results in one group per NaN value > ------------------------------------------------- > > Key: IMPALA-6661 > URL: https://issues.apache.org/jira/browse/IMPALA-6661 > Project: IMPALA > Issue Type: Bug > Components: Backend > Affects Versions: Impala 2.11.0, Impala 2.12.0 > Reporter: Tim Armstrong > Assignee: Michal Ostrowski > Priority: Major > Labels: correctness, perf, ramp-up > > I don't know if this is the desired behaviour but it could be problematic for > some users since it will blow up the number of distinct groups in an > aggregation. I suspect that it's more useful to coalesce all the NaNs into a > single group, similar to how NULL is handled in GROUP BY. > {noformat} > [localhost:21000] > select distinct * from (values(cast("nan" as float)), > (cast("nan" as float)), (sqrt(cast("-1" as float)))) v; > +----------------------+ > | cast('nan' as float) | > +----------------------+ > | NaN | > | NaN | > | NaN | > +----------------------+ > Fetched 3 row(s) in 0.11s > {noformat} > I suspect IMPALA-6069 slightly changed the behaviour here, although it would > have been broken beforehand anyway, since not all NaNs have the same bit > pattern, so Equals() and Hash() were inconsistent. > We should decided what the preferred behaviour is and tweak the behaviour of > the hash table to produce it. -- This message was sent by Atlassian JIRA (v7.6.3#76005) --------------------------------------------------------------------- To unsubscribe, e-mail: issues-all-unsubscr...@impala.apache.org For additional commands, e-mail: issues-all-h...@impala.apache.org