[
https://issues.apache.org/jira/browse/PIG-492?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12639484#action_12639484
]
Pradeep Kamath commented on PIG-492:
------------------------------------
To add more to the initial description:
Here is a scenario showing the need for this: Consider a loader which samples
the first 100 records (say) to determine the schema and returns a schema which
is {(long, chararray)}. However at runtime while constructing the tuple, the
getNext() code might need to know that determineSchema() had returned {(long,
chararray)} at parse time and hence construct the tuple accordingly to consists
of Long and chararray (String) fields. To be able to do this, the loader should
somehow have access to the schema which was returned on the initial
determineSchema() call.
> There should be a way for Loader to refer to the output of determineSchema()
> in the backend
> -------------------------------------------------------------------------------------------
>
> Key: PIG-492
> URL: https://issues.apache.org/jira/browse/PIG-492
> Project: Pig
> Issue Type: Bug
> Affects Versions: types_branch
> Reporter: Pradeep Kamath
> Fix For: types_branch
>
>
> Currently LoadFunc.determineSchema() is only called from LOLoad() at parse
> time in the front end. If the loader.getNext() needs to know what the output
> of determineSchema() was there is no way to get to it in the backend - there
> should be some way to get to it.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
