Fix for union types of union types in Records. #1926
Conversation
… allowing LiteralType through.
dpvball
changed the title
… MappedTypeNodeParser.
@domoritz So I was taking a look at this. I could add this into the UnionNodeParser and modify the UnionType constructor to take in a secondary variable, but then I'd have to store this as a secondary property on the UnionType which seems a bit quirky since that creates two sources of truth for the types even though one is derived from the other. I've instead modified it so that the UnionType class has a method to dynamically take the internal types and flatten them. And I've updated the MappedTypeNodeParser to reference the getFlattenedTypes method instead of the getTypes method. Let me know your thoughts. Thanks for looking into this. |
| @@ -64,6 +64,7 @@ function assertSchema( | |||
| // skip full check if we are not encoding refs | |||
| validateFormats: config.encodeRefs === false ? undefined : true, | |||
| keywords: config.markdownDescription ? ["markdownDescription"] : undefined, | |||
| allowUnionTypes: true, | |||
|
Thanks for the issue and sending a pull request. I think I refactored it a bit in #1933 (I just worked on this code yesterday so it felt like I could combine the logic here). I think my version is also a bit more correct since it works when some of the literal unions are exported which means they are not just an alias. Does that look good to you? |
Revised version of #1926 by dpvball
Existing code was just allowing LiteralType through. The fix now puts the types in the queue and saves if it is a LiteralType. If not then it would be an AliasType or UnionType. These items are then added to the queue and processed one at a time.
The reduce function is there to remove any duplicates in case a type is referenced multiple times in varying sub types
type1 = type2 | type3;
type4 = type1 | type2;
type2 would show up twice in the processing of types. so one of the two would need to be removed.