clj-kondo VS schema

My best suggestion here would be clj-kondo with flycheck-clj-kondo in Emacs. I really can't recommend it enough and would have killed to have it when I was learning Clojure. Not only will it underline all of those references to (now) undefined vars, but it can tell you about numerous little mistakes like mixing up arguments orders in (say) sequence functions, misplaced docstrings that get discarded, style conventions, etc. It's staggering how good it is even for a language as dynamic as Clojure.

The clj-kondo linters are worth reading.

Sure!

Spec: https://github.com/edn-format/edn

Example (linter config): https://github.com/clj-kondo/clj-kondo/blob/634294183a0aa2ca...

Clj-kondo doesn't produce an AST but you could easily combine the analysis output with the AST produced by rewrite-clj by matching on location.

But if you want full support, you can implement a hook: https://github.com/clj-kondo/clj-kondo/blob/master/doc/hooks.md

Please configure and use tools like clj-kondo and kibit. Kibit will report areas where you could write idiomatic clojure instead. Eg, it should catch all those (if (condition) true false) and ask you to replace it with (condition). Or if you really need a boolean value, use boolean to coerce it.

Try clj-kondo, a Clojure linter which will tell you about arity errors and more, before you even evaluate your code.

Plumatic schema (https://github.com/plumatic/schema) , or friends I might be wrong, but I think schema might make more sense to you coming from the F# world (might be wrong)

This one? I didn't. I hear good things about it, and it's reached a point of maturity, being widely used in production.

Uhhh spec has existed for a long time and before that, schema Nowadays we also have the excellent malli. If his codebase is full of functions where the shape of the data isn’t obvious, isn’t documented and isn’t specified in a specific/schema, that’s on him and his bad coding practices and really no different from passing data in other dynamic languages. A class by itself (without additional effort) only gives you field names.

I don't understand this reputation either. There are very large systems built on other Lisps. For example, Emacs has a massive amount of Elisp. Elisp is much more primitive than Clojure, and traditionally libraries don't use e.g. data schemas [1] as runtime contracts for data.

Obviously, once a system built on top of a dynamic language grows beyond certain threshold, you need to be very disciplined as there are no static types to ensure some degree of correctness.

[1] https://github.com/plumatic/schema

Try to learn a schema library early, like Malli or Prismatic Schema. Do not mistook them as "static-typing" things - it's more for data validation and coercion than "security that things will get the right typing information". The idea to learn them early is how you'll shape future code: validating all "output data" first, them using that data inside your program without "defensive programming" like checking every time if a specific value on a map is nil, etc

In Clojure, declarative data specifications for validation and generation are also very mainstream. Schema was first out the door, Clojure Spec is the most popular library, while malli is gaining popularity fast at the moment.