[Question] What is the reason for changing schema validation error type from Set of Error to List of Error?

[mdanish98]

I noticed that schema validation errors are now returned as a List of Error instead of a Set of Error in v2.0.0. I couldn’t find documentation explaining this change; the only reference I found was the statement: “Errors are returned as a List instead of a Set.”

Could you share the rationale or requirements behind switching from Set to List? In particular:

• Is error ordering now meaningful, and should consumers rely on it?
• Can duplicates occur, or is the list guaranteed to be unique?
• Are there migration or compatibility considerations we should be aware of?

[justin-tay]

It is a list of Error and not a list of String.

This was largely changed because of performance. Previously with a Set the hash code needed to be computed when added.

• The order is not generally meaningful but would follow how the schema was evaluated.
• Duplicates shouldn't occur because for a particular keyword, instance location, schema location, evaluation path there should only be a single error generated.

Note that the message now does not prepend the instance location. If you want that then Error.toString() does prepend instance location if present. Returning just the message is not likely meaningful as the location information context would be missing. If this is an API it might make sense to serialize the error as a JSON with the keyword, instance location, schema location, evaluation path and message fields.