Explicit Error Types

[anthonyjoeseph]

Closes #453, Closes #635

Completes the work started by @gcanti in the poc branch, which is explained at a high level here (by the author) and in the docs

As per the notes in the POC branch, all Schemable-related code has been moved into a separate repo called io-ts-contrib. It's registered under anthonyjoeseph's GitHub account, but ownership would ideally be transferred to a maintainer of io-ts

(This PR is still a draft, pending the following)

TODO

• Update Decoder readme to explain how to add custom LeafE error types that are printable, and how to pattern match on errors (based on gcanti's explanation)
• Complete test coverage (currently only old v2.2 tests have been ported up to TaskDecoder & Codec, new Decoder2 function tests have yet to be ported)
• Figure out how to use Codec.lazy w/o @ts-expect-error
• Replace snapshot testing (compare printed string errors) (?)
• Reduce fp-ts version (?)

io-ts-contrib TODO:

• Docs
  • Guard
  • Arbitrary
  • JsonSchema
  • Type
• Complete test coverage

Open Questions

Does io-ts 3.0.0 depend on fp-ts 3.0.0 still? (I thought I saw that somewhere but I can't find it now) Why?

Any way to use Kleisli to remove duplicate code between Decoders and TaskDecoders? Maybe through the help of multiple case-specific Applicative, Chain, Semigroup and/or Alt instances? (Or is this even worth doing?)

Why do decoders store their metadata at the value level? DecodeErrors make sense, that metadata is used for walking the tree, but what's the decoder stuff for?

Why was 'refine' originally removed from Decoder2? fromRefinement? (This PR adds refinement back in)

Add Alt & Category instances for Decoder and TaskDecoder?

Decoder/TaskDecoder - input Ord instance for traversing? Or standardize "Ord.trivial"?

• this would mean we'd have to bump fp-ts version (?)

TaskDecoder - input Applicative instances? Or standardize "Task.parallel" (would probably be sufficient for e.g. graphql resolvers)?

[cdimitroulas]

hey @anthonyjoeseph, it's been a while since this was opened. Is it still the direction of travel for the io-ts POC and (hopefully) next major version of it? Do you need any help with this?

[gcanti]

@cdimitroulas The next major iteration of a io-ts-like library is being developed here https://github.com/fp-ts/schema.

I don't plan to make changes to io-ts before @fp-ts/schema reaches v1.

Once @fp-ts/schema@1.0.0 will be out, I'll think about what's the best migration path (and whether it makes sense)