I noticed some differences in behavior on reading vs writing with JsonbTypeAdapters. They appear to have very different code paths and some caching issues that can alter behavior if reads are done after writes.
These test attempt to verify: - behavior is the same for both reading and writing - consistent order of preference (overriding) First round of tests cover JsonbTypeAdapter, JsonbTypeDeserializer and JsonbTypeSerializer. Tests that fail are ignored so this can be a passing PR that's merged now. - https://github.com/apache/johnzon/pull/131 Discoveries: - Reads fail with String JsonbTypeAdapter on deserialized class (writes succeed) https://issues.apache.org/jira/browse/JOHNZON-414 - Reads fail with Array JsonbTypeAdapter on deserialized class (writes succeed) https://issues.apache.org/jira/browse/JOHNZON-415 - Read/write fail with Map JsonbTypeAdapter on deserialized class (writes succeed) https://issues.apache.org/jira/browse/JOHNZON-416 - Read fail with Map or List JsonbTypeAdapter specified via JsonbConfig (writes succeed, string reads succeed) https://issues.apache.org/jira/browse/JOHNZON-417 - Inconsistent Precedence with JsonbTypeAdapter https://issues.apache.org/jira/browse/JOHNZON-418 Issues 414 to 417 could likely be fixed in 2.0.x. Fixing 418 would likely require a 2.1.x. Don't have time now, but would like to return to this at some point in the future to cover JsonbDateFormat, JsonbNumberFormat and the four flavors of JohnzonConverter. I would also like to test meta-annotation versions of the above. Any thoughts on fixes are also very welcome. -David
