Now that spark-xml supports XML Schema definitions (XSD files) it would be good to use the schema information to generate the Spark Dataframe schema instead of either relying on schema-inference or having to manually create the schema.

To do this I think we need a few things:

• agreement that this is an idea worth pursuing.
• agreement that it is ok to add an import of a library like the Apache Web Services XmlSchema to parse the XSD.
• sufficient tests.

I have already build a proof of concept which correctly parses the XSD to a StructType using the XmlSchema library above (which identified issues with the DateTime and Date types #448). I have tested this against a set of ISO20022 messages.

My challenge is I only have a limited knowledge of XSD so having correct XSD and valid XML files to run against and some help on this would be very useful. I am not sure how representative the ISO20022 XSDs really are.