Personally I'd love this, but I agree with some of the earlier comments that
this should not be Python specific (meaning I should be able to implement a
data source in Python and then make it usable across all languages Spark
supports). I think we should find a way to make this reusable beyond P
Thanks for your feedback Martin.
However, if the primary intended purpose of this API is to provide an
interface for endpoint querying, then I find this proposal even less
convincing.
Neither the Spark execution model nor the data source API (full or
restricted as proposed here) are a good f
