Re-frame subscriptions and events are like very loosely defined function calls. There's a contract there between the sender and the receiver, but it's implicit. The sender makes sure there are 3 items in the vector. The receiver destructures naively, expecting exactly 3 items.
(rf/reg-event-db ::do-something
(fn [db [_ first second {:keys [some-key]}]]
....))
(rf/dispatch [::do-something 1 2 {:some-key 3}])
Keeping these in sync requires continuous effort, and it's easy to miss a case or two, even in less complex applications. Because the errors often involve incompatible use of core clojure language features, like vector and map destructuring on things that cannot be destructured, they are quite often incomprehensible and hard to trace.
Lately, I've been searching for some gradual typing to ease the pain. Most of the available tools based on spec or other schema libraries are capable of solving the above problem. But only one of them check all my points:
speced.def is to me a pretty optimal solution for re-frame vectors. You can annotate as much or little as you want, which is really important to avoid too much boilerplate. Here's the same example from above:
(s/def ::int int?)
(s/def ::valid-db #(:some-crucial-key %))
(rf/reg-event-db ::do-something
(d/fn ^::valid-db [db [_ ^::int first
^number? second
{:keys [^:int some-key]}]]
....))
The first thing you should notice is that I replaced clojure's own fn with a version that checks its own args and return value. The syntax is still completely standard, and should be easily adaptable in your IDE. I'm using Cursive, and it can be told to treat the d/fn as if it was clojure.core/fn
As you can see, I annotated the first 2 args with specs, and we're even able to verify the type of the one map key we are interested in, in a pretty compact way. I also add a validity check for the updated app db returned from the event.
As I mentioned, other spec libraries can do this too, but they require you to write a spec for the entire args vector, which to me is way too much boilerplate to be maintainable in the long term. And it's no fun.
With this system in place, and if you are lucky enough to have expound running, you are entering a totally different world of error messages and early failures for the many mistakes you are bound to make.