I think you’re running into https://issues.apache.org/jira/browse/CALCITE-1178 <https://issues.apache.org/jira/browse/CALCITE-1178> which was logged by a Drill committer a while ago but has not been fixed. Calcite is strongly typed internally. So if we are to allow ’aTimestamp > aDate’ then we either have to add a new overload of ‘>’ that allows (timestamp, date) arguments, or we need to insert implicit casts, so the expression becomes either ‘cast(aTimestamp AS DATE) > aDate’ or ‘aTimestamp > CAST(aDate AS TIMESTAMP)’.
I don’t know which of these 3 alternatives matches the desired semantics of ‘>’. What should the following expressions return? TIMESTAMP ‘1970-01-01 12:00:00’ > DATE ‘1970-01-01’ TIMESTAMP ‘1970-01-01 12:00:00’ = DATE ‘1970-01-01’ TIMESTAMP ‘1970-01-01 12:00:00’ >= DATE ‘1970-01-01’ Julian > On Sep 1, 2017, at 9:31 AM, Roman Kulyk <[email protected]> wrote: > > Hi, team! > > Working on upgrade of Calcite version for Drill I got class cast exceptions > like "ClassCastException: org.apache.calcite.util.DateString cannot be cast > to org.apache.calcite.util.TimestampString" in Drill. > > This issue is reproducible we have "date + interval" situation, for example: > > > > *where o.o_orderdate >= date '1996-10-01' and o.o_orderdate < date > '1996-10-01' + interval '3' month* > > In this case Drill will cast DATE to TIMESTAMP and we will get > ClassCastException when Calcite will try to compare DateString > ('1996-10-01') with TimestampString ('1996-10-01' + interval '3'). To avoid > this situation added a general class for Date/Time/Timestamp literals to > make them comparable to each other. You can see my commit in my local > branch: https://github.com/KulykRoman/incubator-calcite/commit/0656d8fec > > So do I need to create Calcite Jira and open pull request on Calcite master > or this changes should be only in Drill-Calcite branch?
