Zoltán Borók-Nagy created IMPALA-8249:
-----------------------------------------

             Summary: End-to-end test framework doesn't read aggregated 
counters properly
                 Key: IMPALA-8249
                 URL: https://issues.apache.org/jira/browse/IMPALA-8249
             Project: IMPALA
          Issue Type: Bug
            Reporter: Zoltán Borók-Nagy


The test framework doesn't always read the correct value of counters from the 
runtime profile. In the .test files we can have a RUNTIME_PROFILE section where 
we can test our expectations against runtime profile data. We can even 
calculate aggregates of runtime data, currently only SUM is supported over 
integer data, e.g.:

 
{code:java}
---- RUNTIME_PROFILE
aggregation(SUM, RowsReturned): 2142543
{code}
 

However, the counters are pretty-printed in the runtime profile, which means 
that if they are greater than 1000, a shortened version is printed first, then 
the accurate number comes in parenthesis , e.g.:

 
{code:java}
RowsReturned: 2.14M (2142543){code}
 

When the test framework parses the value of an aggregated counter, it wrongly 
tries to parse the short version as a number, which returns a wrong value (2 
instead of 2142543 in the example).



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to