On Wednesday, December 21, 2016 at 2:44:13 PM UTC, Brian Hicks wrote: > > A previous version of the `andThen` documentation actually used circles > and rectangles as the motivating example. It's since switched to versioned > objects, which are maybe a little more realistic. Did you choose circles > and rectangles as your example here for that reason, or are you really > working with shapes? (I also wrote a blog post about `andThen` > <https://www.brianthicks.com/post/2016/06/17/how-does-json-decode-andthen-work/> > > that explains that example, but beware the 0.17 syntax.) > > If you don't have any reason to separate your `Circle` and `Rectangle` > types, you might also consider defining them in `Shape` directly. That way > `CircleAsShape` would become `Circle`. >
No I am not really working with Circles and Rectangles. I am working with a data model and writing a code generator for it that automatically writes the Elm Encoder and Decoder for me. Fields can be optional, there can be recursion and there can be mutual recursion and their can be sub-typing. So I have to write these encoders and decoders fairly defensively to allow for the wide range of possible json they may encounter. I did think about using Json.Decode.oneOf, but that depends on the order of testing the various subtypes, as one may have a set of fields that is a superset of another. I would have to do a topological sort over the partial super-set relationship amongst their fields. Even that might not work as with fields being optional, or two fields with the same name but different type, their might be no practical way to tell apart two different sub-types without a discriminator field. I had not seen your helpful example at: https://www.brianthicks.com/post/2016/06/17/how-does-json-decode-andthen-work/ only one on stack overflow that advocated using the oneOf approach. -- You received this message because you are subscribed to the Google Groups "Elm Discuss" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
