spec-keyword-collisions.clj

;; Spec's global ns registry makes it challenging to declare multiple instances
;; of the same concept in the same ns. For example, say I want to define specs
;; for multiple continents in (ns models.geography.continent). Different continents
;; have different animal populations, so they require different animal counts:

(ns models.geography.continent
  (:require [models.animals :as animals]))

;; First, we define a spec for Antarctica:

(defn antarctica [{:keys [::animals/penguins]}])

(s/fdef antarctica
  :args (s/cat :animals ::animals))

(s/def ::animals (s/keys :req [::animals/penguins]))

;; Now, if we want to do the same for Africa, we're in a bit of trouble:

(defn africa [{:keys [::animals/lions]}])

(s/def africa
  :args (s/cat :animals ???))
;;                      ^^^
;; ::animals has already been defined.

;; One way to solve this problem is to put each continent in its own ns, e.g.

;;                            ˅
(ns models.geography.continent.antarctica)

(s/def ::animals ...)

;; This works, but it feels pretty heavy-handed, especially when the fn implementations
;; share a lot of references.

;; A different approach would involve a language extension:

(ns models.geography.continent
  (:require [models.animals :as animals]))

;; This would allow you to extend ns aliases inline:
;;       ˅
(s/def ::.africa/animals ...)

;; the above would expand to :models.geography.continent.africa/animals

;; For symmetry, it could also make sense to allow aliases to be extended this way:

::animals.penguin/weight

;; expands to :models.animals.penguin/weight

;; One downside of this language modification is that it promotes ad-hoc creation of
;; namespaces, which could theoretically become hard to reason about.

;; (The other obvious downside is... language modification.)

;; I'm curious how other people approach this problem?