Open svetakvsundhar opened 11 months ago
Rather than implementing this from scratch, should we just add a cross language wrapper to the Java version? It seems like that might both accomplish the goal of reducing the maintenance/bug surface and giving users easy access to the functionality.
+1, that'd be my first preference, but I'm not sure the effort required for that. @ahmedabu98 any thoughts here?
I'd generally prefer xlang as well
+1 to leaning on the xlang path.
Regarding the effort required, we would just have to create a SchemaTransform for the Java IO (can range from a day to a week depending on how familiar you are with the IO). There's a guide here that explains what is needed for that.
I'm currently working on a way to automatically generate Python wrappers so a well-documented SchemaTransform would be sufficient.
What needs to happen?
Currently, we are creating a PTransform for
WriteToFhir
, which allows Beam to write to a Fhir store in the Python SDK. Note that this capability exists in the Java SDK via theFhirIO
, but not in the Python SDK.Given the maintenance burden associated with adding new IOs, I propose we add PTransforms and connectors like this in the form of colab examples, rather than in the repo itself. This will decrease the risk of unmaintained code in the repo.
cc: @damccorm @johnjcasey @liferoad to get your thoughts on the above paragraph.
Issue Priority
Priority: 3 (nice-to-have improvement)
Issue Components