You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@hive.apache.org by "Mohammad Kamrul Islam (JIRA)" <ji...@apache.org> on 2015/05/21 18:35:17 UTC

[jira] [Created] (HIVE-10787) MatchPath misses the last matched row from the final result set

Mohammad Kamrul Islam created HIVE-10787:
--------------------------------------------

             Summary: MatchPath misses the last matched row from the final result set
                 Key: HIVE-10787
                 URL: https://issues.apache.org/jira/browse/HIVE-10787
             Project: Hive
          Issue Type: Bug
          Components: UDF
    Affects Versions: 1.2.0
            Reporter: Mohammad Kamrul Islam
            Assignee: Mohammad Kamrul Islam


For example, if you have a STAR(*) pattern at the end, the current code misses the last row from the final result.  For example, if I have pattern like (LATE.EARLY*), the matched rows are :
1. LATE
2. EARLY
In the current implementation, the final 'tpath' missed the last "EARLY" and returns only LATE . Ideally it should return LATE and EARLY.

The following code snippets shows the bug.
{noformat}
0. SymbolFunctionResult rowResult = symbolFn.match(row, pItr);
1. while (rowResult.matches && pItr.hasNext())
2.    {
3.      row = pItr.next();
4.        rowResult = symbolFn.match(row, pItr);
5.      }
6.
7.      result.nextRow = pItr.getIndex() - 1;
{noformat}

Line 7 of the code always moves the row index by one. If ,in some cases, loop (line 1)  is never executed (due to pItr.hasNext() being 'false'), the code still moves the row pointer back by one. Although the line 0 found the first match and the iterator reaches to the end.

I'm uploading a patch which I already tested.
  



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)