[
https://issues.apache.org/jira/browse/IMPALA-6374?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Tim Armstrong updated IMPALA-6374:
----------------------------------
Summary: test tpcds-q98.test has some incorrect data (was: 3)
> test tpcds-q98.test has some incorrect data
> --------------------------------------------
>
> Key: IMPALA-6374
> URL: https://issues.apache.org/jira/browse/IMPALA-6374
> Project: IMPALA
> Issue Type: Test
> Components: Infrastructure
> Affects Versions: Impala 2.9.0
> Reporter: Stephen Carlin
> Assignee: Tim Armstrong
> Priority: Critical
>
> I happened to look through the unit tests and it looks like tpcds-q98.test
> has some bad data in it, but it is verifying correctly.
> One example (among maybe 12 or so) is on line 469:
> line 468: 'AAAAAAAAEKGDAAAA','Houses should
> ','Books','mystery',1.77,3341.80,1.96
> line 469: 'AAAAAAAAFFDDAAAA',','Books','mystery',2.79,4237.23,2.49
> Note that the 2nd field for line 468 looks normal, but line 469 has just a
> single quote.
> I believe this is happening on all strings that end with a comma for this
> test. The correct result for this line (I believe) should be (note the comma
> after Poor):
> 'AAAAAAAAFFDDAAAA','French, civil hours must report essential values.
> Reasonable, complete judges vary clearly homes; often pleasant women would
> watch. Poor,','Books','mystery',2.79,4237.23,2.48
> My guess as to why this is happening is some code in test_result_verifier.py,
> specifically in the part that says:
> for col_val in row_string.split(','):
> # This is a bit tricky because we need to handle the case where a comma
> may be in
> # the middle of a string. We detect this by finding a split that starts
> with an
> # opening string character but that doesn't end in a string character.
> It is
> # possible for the first character to be a single-quote, so handle that
> case
> if (col_val.startswith("'") and not col_val.endswith("'")) or (col_val
> == "'"):
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
