Alex Behm has posted comments on this change. Change subject: IMPALA-4470: Avoid creating a NumericLiteral from NaN/infinity/-0. ......................................................................
Patch Set 2: (4 comments) http://gerrit.cloudera.org:8080/#/c/5050/2/fe/src/main/java/org/apache/impala/analysis/NumericLiteral.java File fe/src/main/java/org/apache/impala/analysis/NumericLiteral.java: PS2, Line 39: BigDecmial > typo: BigDecimal Done PS2, Line 39: A NumericLiteral is backed by a BigDecmial which cannot represent special float values : * like NaN, infinity, or negative zero. > I think you can move this to the comment section for value_. Done PS2, Line 123: 1.0 / v < 0 > Isn't that NEGATIVE_INFINITY? Seems like a stronger check. Yes, but the first part of the condition is v == 0, so I think this will only match negative zero. http://gerrit.cloudera.org:8080/#/c/5050/2/testdata/workloads/functional-planner/queries/PlannerTest/hdfs.test File testdata/workloads/functional-planner/queries/PlannerTest/hdfs.test: PS2, Line 1012: 1 / 0 --> Infinity > Add something with -0? Exercising that code path today is actually surprisingly difficult. We need to get into a path where we evaluate a constant float/double expr from the FE. I could add a separate test, but that one would likely become obsolete very soon when adding constant folding. -- To view, visit http://gerrit.cloudera.org:8080/5050 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: I8243b2ee9fa9c470d078b385583f2f48b606a230 Gerrit-PatchSet: 2 Gerrit-Project: Impala-ASF Gerrit-Branch: master Gerrit-Owner: Alex Behm <[email protected]> Gerrit-Reviewer: Alex Behm <[email protected]> Gerrit-Reviewer: Dimitris Tsirogiannis <[email protected]> Gerrit-Reviewer: Marcel Kornacker <[email protected]> Gerrit-HasComments: Yes
