----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/19165/#review37151 -----------------------------------------------------------
ql/src/java/org/apache/hadoop/hive/ql/optimizer/physical/CrossProductCheck.java <https://reviews.apache.org/r/19165/#comment68541> this should say CrossProductCheck.class ql/src/java/org/apache/hadoop/hive/ql/optimizer/physical/CrossProductCheck.java <https://reviews.apache.org/r/19165/#comment68543> I think you'll always have at least one element - but that's not super obvious. a check for empty might be the defensive choice. ql/src/java/org/apache/hadoop/hive/ql/optimizer/physical/CrossProductCheck.java <https://reviews.apache.org/r/19165/#comment68544> wondering if we should just simplify this and not go hunt for bigtable at all. ql/src/java/org/apache/hadoop/hive/ql/optimizer/physical/CrossProductCheck.java <https://reviews.apache.org/r/19165/#comment68545> not sure if the coding guidelines require uppercase/camelcase inner classes... ql/src/java/org/apache/hadoop/hive/ql/optimizer/physical/PhysicalOptimizer.java <https://reviews.apache.org/r/19165/#comment68546> i'm fine with this. but most folks put bug defense flag check in "optimizers". might help if this turns out to be expensive or if it fails for some reason. (definitely should be on by default tho). - Gunther Hagleitner On March 12, 2014, 11:58 p.m., Harish Butani wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/19165/ > ----------------------------------------------------------- > > (Updated March 12, 2014, 11:58 p.m.) > > > Review request for hive and Gunther Hagleitner. > > > Bugs: HIVE-6643 > https://issues.apache.org/jira/browse/HIVE-6643 > > > Repository: hive-git > > > Description > ------- > > Now that we support old style join syntax, it is easy to write queries that > generate a plan with a cross product. > For e.g. say you have A join B join C join D on A.x = B.x and A.y = D.y and > C.z = D.z > So the JoinTree is: > A — B > __ D — C > Since we don't reorder join graphs, we will end up with a cross product > between (A join B) and C > > > Diffs > ----- > > itests/qtest/pom.xml f8b81a2 > > ql/src/java/org/apache/hadoop/hive/ql/optimizer/physical/CrossProductCheck.java > PRE-CREATION > > ql/src/java/org/apache/hadoop/hive/ql/optimizer/physical/PhysicalOptimizer.java > d593d08 > ql/src/java/org/apache/hadoop/hive/ql/parse/TezCompiler.java a5e6cbf > ql/src/test/queries/clientpositive/cross_product_check_1.q PRE-CREATION > ql/src/test/queries/clientpositive/cross_product_check_2.q PRE-CREATION > ql/src/test/results/clientpositive/cross_product_check_1.q.out PRE-CREATION > ql/src/test/results/clientpositive/cross_product_check_2.q.out PRE-CREATION > ql/src/test/results/clientpositive/tez/cross_product_check_1.q.out > PRE-CREATION > ql/src/test/results/clientpositive/tez/cross_product_check_2.q.out > PRE-CREATION > > Diff: https://reviews.apache.org/r/19165/diff/ > > > Testing > ------- > > added new tests > > > Thanks, > > Harish Butani > >
