[ https://issues.apache.org/jira/browse/HIVE-22771?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17023137#comment-17023137 ]
Mithun Radhakrishnan commented on HIVE-22771: --------------------------------------------- Hmm... This looks like a good catch to me. It looks like the static-partitioning case was missed/mishandled before. The dynamic-partitioning case seems to be handled better, with a more greedy regex: [https://github.com/apache/hive/blob/706c1d44f7734ed36900a02c7255c1d0ce0ad45a/hcatalog/core/src/main/java/org/apache/hive/hcatalog/mapreduce/FileOutputCommitterContainer.java#L867] [~shivam-mohan], you might want to post the patch here, for {{master}} as well, and have CI run tests. The fix seems good to me, on the face of it. > Partition location incorrectly formed in FileOutputCommitterContainer > --------------------------------------------------------------------- > > Key: HIVE-22771 > URL: https://issues.apache.org/jira/browse/HIVE-22771 > Project: Hive > Issue Type: Bug > Components: HCatalog > Affects Versions: 1.2.1 > Reporter: Shivam > Priority: Critical > Labels: pull-request-available > Time Spent: 10m > Remaining Estimate: 0h > > Class _HCatOutputFormat_ in package _org.apache.hive.hcatalog.mapreduce_ uses > function _setOutput_ to generate _idHash_ using below statement: > +In file org/apache/hive/hcatalog/mapreduce/HCatOutputFormat.java+ > *line 116: idHash = String.valueOf(Math.random());* > The output of idHash can be similar to values like this : 7.145347157239135E-4 > > And, > in class _FileOutputCommitterContainer_ in package > _org.apache.hive.hcatalog.mapreduce;_ > Uses below statement to compute final partition path: > +*In org/apache/hive/hcatalog/mapreduce/FileOutputCommitterContainer.java*+ > *line 366: String finalLocn = jobLocation.replaceAll(Path.SEPARATOR + > SCRATCH_DIR_NAME + "{color:#FF0000}\\d\\.?\\d+"{color},"");* > *line 367: partPath = new Path(finalLocn);* > > Regex used here is incorrect, since it will only remove integers after the > *SCRATCH_DIR_NAME,* and hence will append 'E-4' (for the above example) in > the final partition location. -- This message was sent by Atlassian Jira (v8.3.4#803005)