Schema vs core.typed

schema.md

Schema

I think of Schema as a runtime contracts library (I believe it does coercion and validation, but they seem related to me).

Pros

• small barrier to entry, thus immediately useful
  • just add a contract/coercion to a function and you're off
• can manipulate contracts as regular clojure data
• documentation
• compatible with the other amazing prismatic libraries (like fn house)

Cons

• runtime cost
• contracts often disabled in production
  • bug reports have bad errors and stack traces unrelated to the problem
  • ... or GIGO

core.typed

core.typed uses syntactic analysis to avoid certain programming errors. You could theoretically use core.typed to ensure you're using Schema correctly (core.typed needs to learn about Schemas first).

Pros

• syntactic assertions
• statically checked documentation
• no runtime cost for typed code

Cons

• type checking is slow
• way more complex than contracts
• requires a bigger investment to reap the benefits
• currently turns a blind eye to untyped-typed interaction
  • you must provide good annotations for untyped libraries, otherwise things will fail in a similar way to Schema when contracts are disabled
  • Typed Racket shows us how to do this, can be fixed (working on it)
• if you forget to actually type check the code, the only benefit is (unchecked) documentation