pig-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jonathan Coveney (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (PIG-3093) Self join + realias results in schema errors
Date Thu, 13 Dec 2012 00:17:21 GMT

    [ https://issues.apache.org/jira/browse/PIG-3093?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13530527#comment-13530527
] 

Jonathan Coveney commented on PIG-3093:
---------------------------------------

One thing also to look into (once the initial patch is done) is to make sure that the data
is correct.
                
> Self join + realias results in schema errors
> --------------------------------------------
>
>                 Key: PIG-3093
>                 URL: https://issues.apache.org/jira/browse/PIG-3093
>             Project: Pig
>          Issue Type: Bug
>    Affects Versions: 0.11, 0.12
>            Reporter: Jonathan Coveney
>            Assignee: Jonathan Coveney
>            Priority: Critical
>             Fix For: 0.12
>
>
> So this one took a while to isolate, but is pretty crazy.
> {code}
> A = load 'a' as (field1:chararray);
> B = foreach A generate *;
> C = join A by field1, B by field1;
> D = foreach C generate A::field1 as field2, B::field1;
> describe D;
> /*
> D: {
>     field2: chararray,
>     B::field1: chararray
> }
> */
> E = foreach D generate field2, field1;
> describe E;
> /*
> E: {
>     B::field1: chararray,
>     B::field1: chararray
> }
> */
> F = foreach E generate field2;
> store F into 'fail';
> -- <file cristian_simpler.pig, line 20, column 4> Invalid field projection. Projected
field [field2] does not exist in schema: B::field1:chararray,B::field1:chararray.
> {code}
> If you take a look at that code snippet, that is pretty nuts! Since the 2 fields come
from the same original table, renaming one causes issues with both. WUT. The even weirder
part is not that they both get renamed, but that they both become the unrenamed value.
> Interestingly, flipping the value of the projection changes the order of the output,
so it looks like it's whatever the final reference is. ie
> {code}
> A = load 'a' as (field1:chararray);
> B = foreach A generate *;
> C = join A by field1, B by field1;
> D = foreach C generate B::field1, A::field1 as field2;
> describe D;
> E = foreach D generate field2, field1;
> describe E;
> F = foreach E generate field2;
> store F into 'fail';
> {code}
> results in
> {code}
> D: {
>     B::field1: chararray,
>     field2: chararray
> }
> E: {
>     field2: chararray,
>     field2: chararray
> }
> 2012-12-13 00:13:10,045 [main] ERROR org.apache.pig.tools.grunt.Grunt - ERROR 1025: 
> <file simplest.pig, line 8, column 23> Invalid field projection. Projected field
[field2] does not exist in schema: field2:chararray,field2:chararray.
> {code}
> This seems to imply the solution: make copies of the Schema. I added a test and will
hopefully have a patch soon.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message