[
https://issues.apache.org/jira/browse/HIVE-8576?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14181749#comment-14181749
]
Prasanth J commented on HIVE-8576:
----------------------------------
My bad. Thats my mistake. There are q file tests for stats annotation but even
with that its hard to exactly reproduce many cases. This recently happened with
HIVE-8549 as well. These issues crops up when running TPCDS queries and with
higher scale factors (200,1000, 3000). Since operator tree will be different
when stats are different (ex: mapjoin vs shuffle join) its hard to replicate
these test cases with small datasets we have. I have a TPCDS scale 1 factor on
my localbox but even with that many issues weren't reproducible. But yeah this
if condition is my mistake.
+1
> Guaranteed NPE in StatsRulesProcFactory
> ---------------------------------------
>
> Key: HIVE-8576
> URL: https://issues.apache.org/jira/browse/HIVE-8576
> Project: Hive
> Issue Type: Bug
> Reporter: Lars Francke
> Assignee: Lars Francke
> Attachments: HIVE-8576.1.patch
>
>
> Code looks like this:
> {code}
> private List<Integer> getPrimaryKeyCandidates(List<Operator<? extends
> OperatorDesc>> ops) {
> List<Integer> result = Lists.newArrayList();
> if (ops != null || !ops.isEmpty()) {
> for (int i = 0; i < ops.size(); i++) {
> {code}
> The {{||}} should be {{&&}} as the current {{if}} statement can be rewritten
> as {{!(ops == null && ops.isEmpty())}} which doesn't make sense.
> Or am I missing something?
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
