Removing failing fields from JSON?

We’re interested in using a JSON schema to produce valid CSL JSON Items (citation styles language). Our CSL inputs contain many fields that are not part of the schema. In addition, some fields have values that violate the schema. Therefore we would like to perform the following two steps:

1. Removing fields not part of the specification (when additionalProperties is false). Cross-referencing SO44694835.

2. Removing fields whose values violate the schema. Here we would want to remove the lowest-level errors, such that we don’t remove entire objects because of a single deeply nested error contained within them.

Is this doable with jsonschema or should we look to an alternative solution? Any suggestions will be appreciated.

[dhimmel]

Thanks @Julian for the advice. I had looked at the code in the FAQ, but with my limited understanding of the internals of this package, it was a bit difficult. However, of course, I still think it would be a great builtin feature for your package should it be in scope sometime in the future.

For others interested in this feature, we ended up implementing something in https://github.com/greenelab/manubot/pull/49. See the relevant code here and the relevant test suite here, where we note:

The fidelity of the remove_jsonschema_errors implementation is difficult to assess theoretically, therefore its performance should be evaluated empirically with thorough test coverage.

[aj07mm]

this doesn’t work with polymorphic schemas