2.a. "meta" is optional, but when "meta" is in the result set, there is no definition about it's attributes or if those are mandatory or if everything there could be custom and vendor-dependent.
2.b. "rows" when there are no results is not defined if should appear or not, like "rows": [] (IMO an empty array for no results should be mandatory because simplifies implementation and checking the number of results, if "rows" is optional, null checks should be added alongside with length checks)
2.c. found that columns.path is not really useful in the result set, because the path should be applied to a root object and that is not on the result, also to get, for instance, the name of many nodes, ends up on having the same path on different columns, again because the root object where the path is applied is not on the result, the path is relative to something that is not reported, so for automatic processing of query results, there is not enough information there, query results can only be processed if there is specific knowledge about the query that is being executed.
This abstract result set model https://specifications.openehr.org/releases/SM/latest/openehr_platform.html#_query_package need to be serialized to be able to implement the REST API, so we need the serialization rules for this defined somewhere (maybe by an XSD and a JSON Schema in the ITS component).
2.a. "meta" is optional, but when "meta" is in the result set, there is no definition about it's attributes or if those are mandatory or if everything there could be custom and vendor-dependent.
2.b. "rows" when there are no results is not defined if should appear or not, like "rows": [] (IMO an empty array for no results should be mandatory because simplifies implementation and checking the number of results, if "rows" is optional, null checks should be added alongside with length checks)
2.c. found that columns.path is not really useful in the result set, because the path should be applied to a root object and that is not on the result, also to get, for instance, the name of many nodes, ends up on having the same path on different columns, again because the root object where the path is applied is not on the result, the path is relative to something that is not reported, so for automatic processing of query results, there is not enough information there, query results can only be processed if there is specific knowledge about the query that is being executed.