[jira] [Updated] (PIG-5272) BagToString Output Schema

["Rohini Palaniswamy (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/PIG-5272?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Rohini Palaniswamy updated PIG-5272:
------------------------------------

[~daijy],
   It is a valid issue. Schema of the tuple output by BagToTuple cannot be determined at compile time. It depends on how many entries are there in the bag.

[~juen1jp],
   Can you upload the patch?

> BagToString Output Schema
> -------------------------
>
>                 Key: PIG-5272
>                 URL: https://issues.apache.org/jira/browse/PIG-5272
>             Project: Pig
>          Issue Type: Improvement
>            Reporter: Joshua Juen
>            Priority: Minor
>
> The output schema from BagToTuple is nonsensical causing problems using the tuple later in the same script. 
> For example: Given a bag: { data:chararray }, calling BagToTuple yields the schema: ( data:chararray )
> But, this makes no sense since if the above bag contains: {data1, data2, data3} entries, the output tuple from BagToTuple will be:
> (data1:chararray, data2:chararray, data3:chararray) != (data:chararray),the declared output schema from the UDF.
> Unfortunately, the schema of the tuple cannot be known during the initial validation phase. Thus, I believe the output schema from the UDF should be modified to be type tuple without the number of fields being fixed to the number of columns in the input bag. 
> Under the current way, the elements in the tuple cannot be accessed in the script after calling BagToTuple without getting an incompatible type error. We have modified the UDF in our internal UDF jars to work around the issue. Let me know if this sounds reasonable and I can generate the patch.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)