[
https://issues.apache.org/jira/browse/BEAM-881?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Francis UPTON updated BEAM-881:
-------------------------------
Comment: was deleted
(was: That's fine; though I don't understand why this is considered an invasive
change. Each IO can be accompanied by a PTransform class that does the
conversion. The name of this class can be based on the name of the I/O class,
and it can be located in the same place. That's it; there is no change to Beam
itself. It would just be a convention for I/O developers. No one would be
forced to use these PTransforms, but they would be available to anyone who
wanted interoperability.
Avro (and other like formats) are good at this sort of thing, so I don't
understand why this would not be considered "a really useful tuple type". Can
you explain what addition things you would like to see (or not see)?)
> Provide a PTransform in IOs providing a "standard" Avro IndexedRecord
> ---------------------------------------------------------------------
>
> Key: BEAM-881
> URL: https://issues.apache.org/jira/browse/BEAM-881
> Project: Beam
> Issue Type: New Feature
> Components: sdk-java-extensions
> Reporter: Jean-Baptiste Onofré
> Assignee: Jean-Baptiste Onofré
>
> Now, each IO is using a different data format. For instance, the
> {{JmsIO.Read}} provides a {{PCollection}} of {{JmsRecord}} (and
> {{JmsIO.Write}} expects also a {{JmsRecord}}), {{KafkaIO.Read}} provides a
> {{PCollection}} of {{KafkaRecord}}.
> It could appear a bit "complex" for users to manipulate such kind of data
> format: some users may expect kind of standard format.
> Without modifying the existing IO, we could add a {{PTransform}} (as part of
> the IO) that an user can optionally use. This transform will convert the IO
> data format (let say {{JmsRecord}} for instance) to a standard Avro
> {{IndexedRecord}}.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
