drill-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Paul Rogers (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (DRILL-5488) Useless code in VectorTrimmer, fixed-width vector setValueCount()
Date Mon, 08 May 2017 23:08:04 GMT

    [ https://issues.apache.org/jira/browse/DRILL-5488?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16001746#comment-16001746
] 

Paul Rogers commented on DRILL-5488:
------------------------------------

A similar analysis can be done for the variable-length vectors, such as {{VarCharVector}}.

> Useless code in VectorTrimmer, fixed-width vector setValueCount()
> -----------------------------------------------------------------
>
>                 Key: DRILL-5488
>                 URL: https://issues.apache.org/jira/browse/DRILL-5488
>             Project: Apache Drill
>          Issue Type: Bug
>    Affects Versions: 1.10.0
>            Reporter: Paul Rogers
>            Priority: Trivial
>
> Consider this code in a generated fixed-width vector, such as UInt4Vector:
> {code}
>     @Override
>     public void setValueCount(int valueCount) {
>       ...
>       final int idx = (VALUE_WIDTH * valueCount);
>       ...
>       VectorTrimmer.trim(data, idx);
>       data.writerIndex(valueCount * VALUE_WIDTH);
>     }
> {code}
> Consider the {{trim()}} method:
> {code}
> public class VectorTrimmer {
>   ...
>   public static void trim(ByteBuf data, int idx) {
>     data.writerIndex(idx);
>     if (data instanceof DrillBuf) {
>       // data.capacity(idx);
>       data.writerIndex(idx);
>     }
>   }
> }
> {code}
> This method is called {{trim}}, but it actually sets the writer index in the buffer (though
we never use that index.) Since all buffers we use are {{DrillBuf}}, the if-statement is a
no-op: we simply set the writer index twice.
> But, notice that the {{setValueCount()}} method itself calls the same {{writerIndex()}}
method, so it is actually being called three times.
> It seems this code can simply be discarded: it is called from only two places; neither
of which end up using the writer index.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

Mime
View raw message