Gracefully handle unexpected syntax

[vemv]

Brief

defn-spec could be a superset of defn, but it isn't, and also not in a very informative manner:

orchestra.expound-test=> (defn-spec foo [])

IllegalArgumentException find not supported on type: clojure.lang.Keyword  clojure.lang.RT.find (RT.java:863)
orchestra.expound-test=> (defn-spec foo [a])

IllegalArgumentException find not supported on type: clojure.lang.Keyword  clojure.lang.RT.find (RT.java:863)
orchestra.expound-test=> (defn-spec foo [a any?])

IllegalArgumentException find not supported on type: clojure.lang.Keyword  clojure.lang.RT.find (RT.java:863)
orchestra.expound-test=> (defn-spec foo any? [a any?])
orchestra.expound-test/foo

IOW, find not supported on type: clojure.lang.Keyword is not the best message one can find when writing a defn-spec using "muscle memory" from defn.

Fix

I'd suggest at least one of the following:

• If specs are omitted, they are assumed to be any?, possibly emitting a warning. All of this could be configurable behavior.
• Specs are required, otherwise the situation is precisely detected and informed of, in plain English:
  • No :ret spec supplied. Did you forget to add a spec after the function name?
  • No :args spec supplied for the last parameter. Did you forget to add a spec after the last parameter?

WDYT?

Cheers - Victor

[vemv]

As a random thought, defn-spec being a superset of defn would open some interesting doors.

One could refactor codebases to bulk-rename defn to defn-spec, resulting in a more homogeneous codebase (seeing interspersed defn/defn-spec calls is ugly), and inviting one to spec more stuff (assuming a warning approach as outlined above).