The ULR used a bunch of code forked from the DirectRunner but I don't think it currently shares anything. And if it does share any code that I don't know about I expect that the dependency is one-way, i.e. removing the ULR shouldn't affect the DirectRunner. The only shared code I know of is between the ULR and other portable runners, particularly Flink, but I don't think that would be difficult to isolate.
I'm in support of disabling the ULR tests and ok with removing the ULR as long as we make sure it can be revived if we want, like with Mikhail's suggestion of tagging the commit. I can help with the removal of the ULR code since I know specifics about the codebase. On Thu, Mar 14, 2019 at 2:25 PM Kenneth Knowles <[email protected]> wrote: > I know the Java DirectRunner shares a lot of code with the ULR. I'm a bit > unclear on the delta and how independent they are. > > Kenn > > On Thu, Mar 14, 2019 at 2:10 PM Mikhail Gryzykhin <[email protected]> > wrote: > >> @Kenneth >> If we disable tests, I'd call Java ULR a dead code. >> >> One of the better compromises: >> 1. disable tests. >> 2. Add tag to the last commit where Java ULR existed. >> 3. Remove Java ULR from head. >> >> Keeping history, no extra dead code at head. >> >> --Mikhail >> >> Have feedback <http://go/migryz-feedback>? >> >> >> On Thu, Mar 14, 2019 at 1:02 PM Ankur Goenka <[email protected]> wrote: >> >>> On that note, we should also think about adding PVR for python reference >>> runners. Jira: https://issues.apache.org/jira/browse/BEAM-6837 >>> >>> >>> On Thu, Mar 14, 2019 at 12:57 PM Kenneth Knowles <[email protected]> >>> wrote: >>> >>>> How about this compromise: >>>> >>>> 1. disable the test since clearly no one is relying on the >>>> functionality that is broken >>>> 2. leave the Java ULR as-is for now, and a volunteer can pick it up and >>>> make it work if they want >>>> >>>> Kenn >>>> >>>> On Thu, Mar 14, 2019 at 11:41 AM Mikhail Gryzykhin <[email protected]> >>>> wrote: >>>> >>>>> Hi everyone, >>>>> >>>>> We have Python PVR Reference post-commit tests failing for quite some >>>>> time now. These are tests for java reference runner. >>>>> >>>>> According to this thread >>>>> <https://lists.apache.org/thread.html/b235f8ee55a737ea399756edd80b1218ed34d3439f7b0ed59bfa8e40@%3Cdev.beam.apache.org%3E>, >>>>> we are deciding what to do with java reference runner and might want to >>>>> remove it from code base. >>>>> >>>>> My question is: do we want to a) invest time in fixing python PVR >>>>> tests, or b) disable this test and start cleaning up code? >>>>> >>>>> a) Is worth it if we want to invest into java reference runner in the >>>>> future. >>>>> b) Is worth if we want to invest into Python and forfeit java >>>>> reference runner. >>>>> >>>>> Option b) seem more reasonable to me atm, since most people lean >>>>> towards going forward with Python reference runner. >>>>> >>>>> Please, share your thoughts. >>>>> >>>>> Regards, >>>>> --Mikhail >>>>> >>>>> Have feedback <http://go/migryz-feedback>? >>>>> >>>>
