[jira] [Commented] (IMPALA-3170) Large literal exponents cause many seconds of delay

["Tim Armstrong (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/IMPALA-3170?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16508691#comment-16508691 ] 

Tim Armstrong commented on IMPALA-3170:
---------------------------------------

I can reproduce this on Ubuntu14.04/OpenJDK7/2.x but not Ubuntu16.04/OpenJDK8/master. The decimal_v2 setting has no impact. I'm guess something got fixed in BigDecimal.

> Large literal exponents cause many seconds of delay
> ---------------------------------------------------
>
>                 Key: IMPALA-3170
>                 URL: https://issues.apache.org/jira/browse/IMPALA-3170
>             Project: IMPALA
>          Issue Type: Bug
>          Components: Frontend
>    Affects Versions: Impala 2.5.0
>            Reporter: Jim Apple
>            Priority: Minor
>
> {noformat}
> for foo in 0 10 210 3210 43210 543210; do time -p impala-shell.sh -q "select 1e${foo}"; done
> {noformat}
> {{0}}, {{10}}, {{210}} succeed.
> {{3210}} and {{43210}} fail with a sensible error message in less than half a second.
> {{543210}} takes 28 seconds to return that error message.
> My first guess is the {{BigDecimal}} constructor in {{NumericLiteral(String value, Type t)}}.
> This is one of those, "well, don't do that, then" type of bugs, but I'd rather that malformed user input fail fast than cause impalad to spin, especially since some user input is generated programatically.



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

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-all-unsubscribe@impala.apache.org
For additional commands, e-mail: issues-all-help@impala.apache.org