No schema for process graphs?

Open-EO / openeo-api

The openEO API specification

http://api.openeo.org

Apache License 2.0

91 stars 11 forks source link

No schema for process graphs? #452

Closed LukeWeidenwalker closed 2 years ago

LukeWeidenwalker commented 2 years ago

While working on process graph parsing recently I realized that the process-graph as an object doesn't actually have a complete schema in the API spec. Now I have a couple of questions!

Are there plans for this to become part of the spec at some point?
If not, how can I know what a valid process graph (e.g. as produced by one of the clients) will be structured like?
I know of https://github.com/Open-EO/openeo-api/blob/master/assets/pg-schema.json, but the README of that repo says: The [assets](https://github.com/Open-EO/openeo-api/blob/master/assets) folder contains some useful additional files such as examples or schemas. All of these are non-binding additions. The source of truth are the top-level specification files. Would you recommend for this schema to be used in backends? Is this actually still up-to-date with what clients produce currently?

Thanks a lot for any clarifications!

m-mohr commented 2 years ago

There's a complete OpenAPI schema in the API spec and additionally, there's the conversion to JSON Schema for only the process graph. What are you missing exactly?

The OpenAPI one is the source of truth and complete. With (good) OpenAPI validators you should be able to check process graphs. If you have better access to JSON schema validators, you can at least validate the graph portion using that. What you use depends on the use case, I can't decide that without additional context.

I'm not 100% sure how to help here?

LukeWeidenwalker commented 2 years ago

Sorry, I will try to clarify what I'm after:

https://github.com/Open-EO/openeo-processes/blob/master/meta/subtype-schemas.json#L250 defines the process-graph subtype, but this subtype only tells me that an object with subtype process-graph has a process_graph object, which in turn has the process_id and arguments fields.
Is there any schema for the process arguments? I know this depends on which process it is, but there are common fields like from_parameter and from_node? Are these schematised anywhere? I guess I expected the schema to include a process_node type, that wraps the actual process definition to make fields like from_parameter parsable even if the actual process definition hasn't been loaded yet.
Does this make sense? Sorry if I'm being unclear, I'm still learning the spec - happy to jump on a call sometime if that would help!

LukeWeidenwalker commented 2 years ago

Sorry, I just realized that this whole question was based on a misunderstanding on my part and doesn't really make sense - will close!