Tim Armstrong has posted comments on this change. ( http://gerrit.cloudera.org:8080/12974 )
Change subject: IMPALA-7608: Estimate row count from file size when no stats available ...................................................................... Patch Set 10: (6 comments) Mainly comments about comments and some cleanup. One administrative thing - it would have been a little easier to review PS10 if you did the rebase in a separate patchset. The diff from PS9->PS10 was noisy because of unrelated changes in query-options.cc picked up from the rebase. http://gerrit.cloudera.org:8080/#/c/12974/10/fe/src/main/java/org/apache/impala/planner/HdfsScanNode.java File fe/src/main/java/org/apache/impala/planner/HdfsScanNode.java: http://gerrit.cloudera.org:8080/#/c/12974/10/fe/src/main/java/org/apache/impala/planner/HdfsScanNode.java@179 PS10, Line 179: private static double ESTIMATED_COMPRESSION_FACTOR_LEGACY = 3.58;// to change Can you remove these "to change" comments? I don't think they help much. http://gerrit.cloudera.org:8080/#/c/12974/10/fe/src/test/java/org/apache/impala/planner/CardinalityTest.java File fe/src/test/java/org/apache/impala/planner/CardinalityTest.java: http://gerrit.cloudera.org:8080/#/c/12974/10/fe/src/test/java/org/apache/impala/planner/CardinalityTest.java@225 PS10, Line 225: // functional.alltypesmixedformat is a table of 4 partitions, I have some nits about these test comments. I think they should be javadoc comments, just for consistency. The text is also wrapped at < 90 lines - in some cases the comment would fit on fewer lines. http://gerrit.cloudera.org:8080/#/c/12974/10/fe/src/test/java/org/apache/impala/planner/CardinalityTest.java@243 PS10, Line 243: // True cardinality of tpch_text_gzip.lineitem is 6,001,215. Thanks for these comments about the cardinality, this is actually really helpful to understand the test. http://gerrit.cloudera.org:8080/#/c/12974/10/fe/src/test/java/org/apache/impala/planner/CardinalityTest.java@516 PS10, Line 516: // Estimated cardinality of the NestedLoopJoinNode is 550,564 = 742 * 742. I guess we're not so good at estimating cardinality of tiny parquet files because of the footer? http://gerrit.cloudera.org:8080/#/c/12974/10/fe/src/test/java/org/apache/impala/planner/CardinalityTest.java@564 PS10, Line 564: //TODO: It seems that the cardinality of the SelectNode should be 1 instead Nice catch. Can you file a JIRA for this and mention in TODO? This is much better for tracking purposes - bugs that are only tracked by TODOs in the code tend to be forgotten easily. And consider fixing it in a separate commit. It seems like an oversight, I think this code should probably have a max(1, ...) to avoid setting it to 0. cardinality_ = Math.round(((double) getChild(0).cardinality_) * computeSelectivity()); Preconditions.checkState(cardinality_ >= 0); http://gerrit.cloudera.org:8080/#/c/12974/10/testdata/workloads/functional-planner/queries/PlannerTest/default-join-distr-mode-shuffle-hdfs-num-rows-est-disabled.test File testdata/workloads/functional-planner/queries/PlannerTest/default-join-distr-mode-shuffle-hdfs-num-rows-est-disabled.test: PS10: I think you can revert this file to its original name (same for the other similar cases where gerrit shows a rename). Encoding all of the options in the file name isn't really scalable, so I'd prefer default-join-distr-mode-shuffle.test for conciseness. -- To view, visit http://gerrit.cloudera.org:8080/12974 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: Impala-ASF Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: Ic414121c8df0d5222e4aeea096b5365beb04568a Gerrit-Change-Number: 12974 Gerrit-PatchSet: 10 Gerrit-Owner: Fang-Yu Rao <fangyu....@cloudera.com> Gerrit-Reviewer: Fang-Yu Rao <fangyu....@cloudera.com> Gerrit-Reviewer: Impala Public Jenkins <impala-public-jenk...@cloudera.com> Gerrit-Reviewer: Paul Rogers <prog...@cloudera.com> Gerrit-Reviewer: Tim Armstrong <tarmstr...@cloudera.com> Gerrit-Reviewer: Todd Lipcon <t...@apache.org> Gerrit-Comment-Date: Mon, 10 Jun 2019 18:07:43 +0000 Gerrit-HasComments: Yes