hadoop-pig-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Pradeep Kamath (JIRA)" <j...@apache.org>
Subject [jira] Updated: (PIG-1371) Pig should handle deep casting of complex types
Date Mon, 26 Apr 2010 22:41:32 GMT

     [ https://issues.apache.org/jira/browse/PIG-1371?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Pradeep Kamath updated PIG-1371:
--------------------------------

    Attachment: PIG-1371-partial.patch

partial patch - attaching here for future reference

> Pig should handle deep casting of complex types 
> ------------------------------------------------
>
>                 Key: PIG-1371
>                 URL: https://issues.apache.org/jira/browse/PIG-1371
>             Project: Pig
>          Issue Type: Bug
>            Reporter: Pradeep Kamath
>             Fix For: 0.8.0
>
>         Attachments: PIG-1371-partial.patch
>
>
> Consider input data in BinStorage format which has a field of bag type - bg:{t:(i:int)}.
In the load statement if the schema specified has the type for this field specified as bg:{t:(c:chararray)},
the current behavior is that Pig thinks of the field to be of type specified in the load statement
(bg:{t:(c:chararray)}) but no deep cast from bag of int (the real data) to bag of chararray
(the user specified schema) is made.
> There are two issues currently:
> 1) The TypeCastInserter only considers the byte 'type' between the loader presented schema
and user specified schema to decided whether to introduce a cast or not. In the above case
since both schema have the type "bag" no cast is inserted. This check has to be extended to
consider the full FieldSchema (with inner subschema) in order to decide whether a cast is
needed.
> 2) POCast should be changed to handle casting a complex type to the type specified the
user supplied FieldSchema. Here is there is one issue to be considered - if the user specified
the cast type to be bg:{t:(i:int, j:int)} and the real data had only one field what should
the result of the cast be:
>  * A bag with two fields - the int field and a null? - In this approach pig is assuming
the lone field in the data is the first field which might be incorrect if it in fact is the
second field.
>  * A null bag to indicate that the bag is of unknown value - this is the one I personally
prefer
>  * The cast throws an IncompatibleCastException

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message