Was the motivation for this the bug that was found with PARQUET-1414? How did we catch this?
On Thu, Feb 21, 2019 at 4:56 AM Gabor Szadovszky <[email protected]> wrote: > HI All, > > I'm planning to push the revert tomorrow if there are no objections. > > Cheers, > Gabor > > On Tue, Feb 19, 2019 at 6:02 PM Gabor Szadovszky <[email protected]> wrote: > > > Sorry, wrong PR. So, see PARQUET-1381 > > <https://issues.apache.org/jira/browse/PARQUET-1381> and PR #621 > > <https://github.com/apache/parquet-mr/pull/621>. > > > > On Tue, Feb 19, 2019 at 5:44 PM Gabor Szadovszky <[email protected]> > wrote: > > > >> Hi All, > >> > >> During working on a fix I've discovered that the recently added (since > >> 1.10.0) feature PARQUET-1381 is not properly implemented and causes some > >> unit test failures with my independent fix. After a more deep > investigation > >> I think the design of this feature is conceptionally incompatible with > our > >> current API. See the comments of the jira for more details. > >> > >> Because I don't feel this feature very important for the current release > >> I think the most easier and painless solution is to revert it. Created > the > >> PR #620 for it. > >> > >> We usually don't do reverts especially for commits that are sitting in > >> master for a while. I would like to ask your opinions about it. Also > feel > >> free to comment in the jira (about the design/implementation of the > >> feature) or the PR (about the revert). > >> > >> Thanks a lot, > >> Gabor > >> > > > -- Ryan Blue Software Engineer Netflix
