> On Aug. 30, 2017, 3:35 a.m., Gopal V wrote: > >
Thanks for quick turnaround. Working on the comments. > On Aug. 30, 2017, 3:35 a.m., Gopal V wrote: > > ql/src/java/org/apache/hadoop/hive/ql/parse/SemiJoinBranchInfo.java > > Lines 30 (patched) > > <https://reviews.apache.org/r/61985/diff/1/?file=1807523#file1807523line30> > > > > I prefer a Boolean object, which goes from null -> false/true, so that > > you can encode (not-set, true, false) in one field. Will do that. > On Aug. 30, 2017, 3:35 a.m., Gopal V wrote: > > ql/src/java/org/apache/hadoop/hive/ql/parse/SemiJoinBranchInfo.java > > Lines 57 (patched) > > <https://reviews.apache.org/r/61985/diff/1/?file=1807523#file1807523line57> > > > > Add a Preconditions checkState that prevents this from going from true > > -> false. Makes sense. > On Aug. 30, 2017, 3:35 a.m., Gopal V wrote: > > ql/src/java/org/apache/hadoop/hive/ql/parse/TezCompiler.java > > Lines 1371 (patched) > > <https://reviews.apache.org/r/61985/diff/1/?file=1807524#file1807524line1371> > > > > NPE issues possible? It can't happen. However, I can put this in a try catch block and assert if it happens. > On Aug. 30, 2017, 3:35 a.m., Gopal V wrote: > > ql/src/java/org/apache/hadoop/hive/ql/parse/TezCompiler.java > > Lines 1404 (patched) > > <https://reviews.apache.org/r/61985/diff/1/?file=1807524#file1807524line1404> > > > > Is that a break or a continue? > > > > If this is a break out of the loop, then what happens to the rest of > > the items in dequeue? Yes, it is a break. For a given TS, there is only one DPP Event. No need to process the rest of the graph. > On Aug. 30, 2017, 3:35 a.m., Gopal V wrote: > > ql/src/test/queries/clientpositive/dynamic_semijoin_reduction.q > > Lines 115 (patched) > > <https://reviews.apache.org/r/61985/diff/1/?file=1807525#file1807525line115> > > > > Pick a query with a non-zero result, so that we can see when it has > > false negatives (i.e loses rows it is meant to have?). Sure. - Deepak ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/61985/#review184120 ----------------------------------------------------------- On Aug. 30, 2017, 3:07 a.m., Deepak Jaiswal wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/61985/ > ----------------------------------------------------------- > > (Updated Aug. 30, 2017, 3:07 a.m.) > > > Review request for hive, Gopal V and Jason Dere. > > > Bugs: HIVE-17399 > https://issues.apache.org/jira/browse/HIVE-17399 > > > Repository: hive-git > > > Description > ------- > > Do not remove semijoin branch if it feeds to TS->DPP_EVENT > > > Diffs > ----- > > ql/src/java/org/apache/hadoop/hive/ql/parse/SemiJoinBranchInfo.java > 5d7b9e5c6d > ql/src/java/org/apache/hadoop/hive/ql/parse/TezCompiler.java 1671773d4a > ql/src/test/queries/clientpositive/dynamic_semijoin_reduction.q b22890bc9d > ql/src/test/results/clientpositive/llap/dynamic_semijoin_reduction.q.out > 3bd35bf2d8 > > > Diff: https://reviews.apache.org/r/61985/diff/1/ > > > Testing > ------- > > > Thanks, > > Deepak Jaiswal > >
