Conversation
89d24ac to
16b6f3a
Compare
|
What about forcing the user to be more explicit? It might help users avoid doing crazy things with In the latest push, What do you think? How could it be improved? I think this might make people think twice before doing crazy things with the partial constructor. |
|
@kschweppe do you have any thoughts on this solution? I could use a second set of eyes on this. |
working pretty encoding using json instance. holes working too fix querydata plus sign space encoding all tests passing
only ViewId, ViewAction use Encoded tests passing comments changelog changelog fix rebase rebase conflicts
93c42d9 to
5258de2
Compare
merge conflict
5258de2 to
77b3fdf
Compare
|
Sure! Sorry for the delay, I just didn't get to have a closer look over the weekend. I generally agree that it should be more explicit, and was also thinking we need a typeclass for this. Is there any drawback in providing a default implementation for numbers? I guess having number dropdowns is a common use case, for example I have dropdowns for year and month and needed to add NumInput to both. And then I'm wondering if this can/should also work with search input, which actually was my initial use case. |
See #194
Instead of using undefined to serialize a partially applied constructor, we
throw InputHole- we catch it during serialization and turn it into aHole.This inspired a refactor to remove Web.Hyperbole.Data.Param. Everything now uses ToJSON as a generic serialization format, then encodes in its own way. This changes QueryData, Session, and Encoded (ViewId, ViewAction)