+1 to disable, It is already introduced by new type system in TimestampType. I think it is time to update document too.
Best, Jingsong Lee On Fri, Nov 22, 2019 at 6:05 PM Kurt Young <ykt...@gmail.com> wrote: > +1 to disable, we also need to highlight this in 1.10 release notes. > > Best, > Kurt > > > On Fri, Nov 22, 2019 at 5:56 PM Zhenghua Gao <doc...@gmail.com> wrote: > > > Hi, > > > > I wanted to bring up the discuss of Disable conversion between TIMESTAMP > > and Long in parameters and results of UDXs. > > > > Since FLINK-12253[1] introduce the new TimestampType and conversion from > > and > > to long is not supported, the UDXs with Long parameters should not > receive > > TIMESTAMP fields and vice versa. > > > > The current situation is we use long as internal representation of > > TIMESTAMP, the legacy planner and blink planner DO NOT DISABLE this > > conversion. Now FLINK-14599[2] would introduce a new internal > > representation of TIMESTAMP and it's time to make a decision to DISABLE > it. > > > > In addition, our document[3] recommends UDXs users use long as > > representation of SQL_TIMESTAMP, which is obsolete too. > > > > Please let me know what you think! > > > > [1] https://issues.apache.org/jira/browse/FLINK-12253 > > [2] https://issues.apache.org/jira/browse/FLINK-14599 > > [3] > > > > > https://ci.apache.org/projects/flink/flink-docs-release-1.9/dev/table/udfs.html#best-practices-for-implementing-udfs > > > > *Best Regards,* > > *Zhenghua Gao* > > > -- Best, Jingsong Lee