I don't think that removing the shading is a good idea. Jackson is a very common dependency and pulling in projects that use different versions has caused a lot of headache. Why go backward and make Parquet vulnerable to those problems? I don't see a good justification for it.
On Mon, Feb 18, 2019 at 8:29 AM Jacques Nadeau <[email protected]> wrote: > I haven't looked at the usage but would wonder if the core modules truly > need jackson. I don't think most of the systems that read Parquet use the > jackson part (?). If so, maybe the code could be refactored to remove the > dependency and it be moved to an optional component. We want to do the same > thing with Jackson in Arrow (and did it recently for Guava). > > On Mon, Feb 18, 2019 at 3:09 AM Driesprong, Fokko <[email protected]> > wrote: > > > Hi all, > > > > Recently I've opened a PR to move from Jackson 1.x to Jackson 2.9 > > <https://github.com/apache/parquet-mr/pull/616>. I've also removed the > > shading project since most libraries are up to date with Jackson 2.x. > Gabor > > suggested having a discussion on the mailing list to discuss the removal > of > > the shading of Jackson. > > > > Spark 2.x is at 2.6, Spark 3.0 at 2.9.6, Hadoop at 2.9.x, Flink at 2.7.9, > > but that one is shaded anyway :-) One problem might be Apache Avro which > is > > still using Jackson 1.x (codehause), until we release Avro 1.9. > > > > What are the thoughts on this subject, should we still shade Jackson, or > > not? > > > > Cheers, Fokko > > > -- Ryan Blue Software Engineer Netflix
