Add spec reproducing issue with custom type with and nested hashes validation

[edgenard]

This is meant to reproduce an issue where custom types in nested hashes do not take effect.
This is my first time with this project, I hope the spec is in the right place. If not I'd be happy to move it to the appropriate place.

This PR is related to this Issue: #396

I tried but could not figure out how to get it to pass.

[flash-gordon]

Thanks, I left a comment, can you also mark the spec as pending so that we can merge it in?

[edgenard]

@flash-gordon I tried to get the spec to pass but had a very hard time following the code, if you could point me in the right direction I'd be happy to try again.

For example, what file or files would need to change to make it pass.

[flash-gordon]

I didn't mean making it pass :) Just make it pending by changing specify to xspecify, it'll stop failing the CI.

[edgenard]

I didn't mean making it pass :) Just make it pending by changing specify to xspecify, it'll stop failing the CI.

I added the pending to the spec.

But I'd be happy to work on it if someone could point me in the right direction.

[solnic]

Your type crashes, that's why it doesn't work:

(byebug) Test::ExpirationDate["2021-11-11T00:00:00+00:00"]
*** Dry::Types::CoercionError Exception: undefined method `to_time' for "2021-11-11T00:00:00+00:00":String
Did you mean?  to_i

If you change it so that it does DateTime.parse(value), the spec will pass.

@flash-gordon it seems like Safe constructor rescues silently this crash though - THAT is probably a bug 🙂

[edgenard]

Your type crashes, that's why it doesn't work:

(byebug) Test::ExpirationDate["2021-11-11T00:00:00+00:00"]
*** Dry::Types::CoercionError Exception: undefined method `to_time' for "2021-11-11T00:00:00+00:00":String
Did you mean?  to_i

If you change it so that it does DateTime.parse(value), the spec will pass.

@flash-gordon it seems like Safe constructor rescues silently this crash though - THAT is probably a bug 🙂

Thank you! I think that definitely cleared it up. And yeah the silent failure is certainly confusing to users of the library.

I'll close this PR.