This topic was raised during the code sprint:

When we are transforming from a more relaxed schema to another more restrictive, it is possible that the result will not be compliant with the output schema, as we are not validating the output.

I took a look on how this could be implemented, and here are my thoughts about it:

• The current validate command is a proxy to validate_mcf
• I understand that we want to keep validate_mcf, as it is used in the pygeoapi_plugin
• Validation can be useful on its own, not only when transforming
• The commands that accept an MCF as an input are different from the commands that accept metadata in other schemas

The path I am following is:

• Add a validate method to BaseOutputSchema
• Add can_validate to the schema_matrix elements in get_supported_schemas
• Add validate to the schemas click command from get_supported_schemas
• Modify the get_supported_schemas to filter by schemas that can validate?
• Add a validate option to the transform click command using the previous filter when calling to get_supported_schemas to build the list of choices
• If the option validate is present, call to schema_object_output.validate(content) in transform_metadata

The questions I have are:

• What to do with can_validate in get_supported_schemas for include_autodetect
• Should we add the filter by validate in get_supported_schemas?
• How to generalize the validate click command? It seems like a design decision to have differentiated the commands that accept MFC from the ones that accept metadata in other schemas, so a unification might break that principle

I will reference a branch in my fork to show all the details mentioned in this issue.