[
https://issues.apache.org/jira/browse/PIG-2395?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13162402#comment-13162402
]
Jie Li commented on PIG-2395:
-----------------------------
OK. Great. I didn't find that.
> Pig doesn't do early projection for some scripts
> ------------------------------------------------
>
> Key: PIG-2395
> URL: https://issues.apache.org/jira/browse/PIG-2395
> Project: Pig
> Issue Type: Improvement
> Affects Versions: 0.9.1
> Environment: Linux
> Reporter: Jie Li
>
> For some scripts Pig doesn't do early projection, e.g. dropping non-necessary
> fields as soon as possible. This is observed in two ways: 1) the output
> doesn't contain INFO like "ColumnPruneVisitor - Columns pruned for xxx: $0,
> $1"; 2) the job has as much or more local IO (see FILE_BYTES_READ and
> FILE_BYTES_WRITTEN) as HDFS IO (see HDFS_BYTES_READ).
> One example where Pig should figure out A's fields c~n can be dropped before
> the COGROUP.
> A = load '/tmp/A' USING PigStorage('|') as (a,b,c,d,e,f,g,h,i,j,k,l,m,n);
> B = load '/tmp/B' USING PigStorage('|') as (a);
> COG = cogroup A by a, B by a;
> out = foreach COG generate SUM(A.b) as sum;
> store out into '/tmp/out' USING PigStorage('|');
> Another similar example involves a GROUP operator.
> While Pig is able and assumed to do early projection in most cases, this
> inconsistency hurts the performance badly.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
