[CORE-6836] schema_registry/json: Support Internal References #22815
Conversation
Signed-off-by: Ben Pope <[email protected]>
Signed-off-by: Ben Pope <[email protected]>
Signed-off-by: Ben Pope <[email protected]>
Signed-off-by: Ben Pope <[email protected]>
| // Internal reference | ||
| if (ref.starts_with("#")) { | ||
| json::Pointer ptr{ | ||
| ref.data() + 1, | ||
| ref.length() - 1, | ||
| }; | ||
| if (auto* p = ptr.Get(_schema.doc); p) { | ||
| return *p; | ||
| } | ||
| throw as_exception(error_info{ | ||
| error_code::schema_invalid, | ||
| fmt::format("Reference not found: '{}'", ref)}); | ||
| } | ||
|
|
||
| throw as_exception(error_info{ | ||
| error_code::schema_invalid, | ||
| fmt::format("External references not supported: '{}'", ref)}); | ||
| } | ||
|
|
There was a problem hiding this comment.
By internal references, do you mean non-cross-schema-version references in general or references starting with # specifically? Looking at https://json-schema.org/understanding-json-schema/structuring#dollarref, it's possible to have "internal" references (ie. a reference to a schema within the same schema document) that don't start with #. Do we want to support those kinds of references as well (eg. "$ref": "/schemas/address")?
|
ducktape was retried in https://buildkite.com/redpanda/redpanda/builds/52710#01913784-8de4-46f8-8757-59e952a379b5 ducktape was retried in https://buildkite.com/redpanda/redpanda/builds/52710#01913784-8de0-4cac-ae47-007c20ff1dbc |
| std::string_view as_string_view(json::Value const& v) { | ||
| return {v.GetString(), v.GetStringLength()}; | ||
| } | ||
|
|
| @@ -1197,7 +1223,7 @@ bool is_not_combinator_superset( | |||
| // for not combinator, we want to check if the "not" newer subschema is | |||
| // less strict than the older subschema, because this means that newer | |||
| // validated less data than older | |||
| return is_superset(newer_it->value, older_it->value); | |||
| return is_superset(ctx, newer_it->value, older_it->value); | |||
There was a problem hiding this comment.
need to swap the order of contexts , like newer and older are subschema are swapped.
pseudo code:
| return is_superset(ctx, newer_it->value, older_it->value); | |
| return is_superset({ctx.newer, ctx.older}, newer_it->value, older_it->value); |
otherwise newer will access the context for older as older will access the context for newer
| if (v.IsObject()) { | ||
| if (auto it = v.FindMember("$ref"); it != v.MemberEnd()) { | ||
| return ctx.resolve(as_string_view(it->value)).GetObject(); |
There was a problem hiding this comment.
this should be the ok for draft4/7 ($ref overwrites the schema), but reading the spec seems like for 2019/20 the result should be merged on top of the schema.
A schema can reference another schema using the $ref keyword. The value of $ref is a URI-reference that is resolved against the schema's [Base URI](https://json-schema.org/understanding-json-schema/structuring#base-uri). When evaluating a $ref, an implementation uses the resolved identifier to retrieve the referenced schema and applies that schema to the [instance](https://json-schema.org/learn/glossary#instance).
Draft-specific info
In Draft 4-7, $ref behaves a little differently. When an object contains a $ref property, the object is considered a reference, not a schema. Therefore, any other properties you put in that object will not be treated as JSON Schema keywords and will be ignored by the validator. $ref can only be used where a schema is expected.
Then again, I suspect our reference implementation does not implement the merge behavior.
I'll check
|
Closed in favour of: #23461 |
As described in the tests, support
"$ref": "#...":{ "$ref": "#/$defs/a_ref", "$defs": { "a_ref": { "type": "string" } } }I.e., they are internal-only, qualified as starting with
#.Backports Required
Release Notes