`fn?` spec for resolver functions is too tight. What about `ifn?`

[orestis]

After upgrading to Lacinia 0.33, my schema fails to validate. The reason is that I'm using vars instead of functions as my resolvers, thus allowing me to just reevaluate my resolver in the REPL, without having to stop/start my entire system.

Since vars are callable like functions, this used to work perfectly in 0.32, but now the spec is too tight. I think using ifn? as the spec predicate would allow me to use vars as resolvers while still capturing the intent of the validation.

[hlship]

fn? is too loose, but I think we may want to keep it very strict: (or (fn? %) (var? %)) because otherwise, a keyword (which returns true from ifn?) works, and that would mean that we would no longer catch when keywords were not replaced with actual resolvers.

[hlship]

But absolutely, vars should be allowed, to support REPL oriented development.

[orestis]

Thanks for the quick response! I noticed that multimethods also fail the fn? check -- though I'm not sure exactly how a multimethod could practically work as a resolver.

[hlship]

What would you dispatch on, especially at the root? But if you really want a multi-method, you could wrap the multi-method with a plain function, then use that plain function all over your schema, if you like. I don't see the point, myself.

[orestis]

Fair point. I was just trying to think of other cases were fn? was not applicable.

24 Ιουν 2019, 6:23 μμ, ο χρήστης «Howard M. Lewis Ship notifications@github.com» έγραψε:

What would you dispatch on, especially at the root? But if you really want a multi-method, you could wrap the multi-method with a plain function, then use that plain function all over your schema, if you like. I don't see the point, myself.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or mute the thread.