匹配并生成可变长度的有序向量的 Clojure 规范

A Clojure Spec that matches and generates an ordered vector of variable length

让我们从

的常规序列开始
(require '[clojure.spec     :as spec]
         '[clojure.spec.gen :as gen])
(spec/def ::cat (spec/cat :sym symbol? :str string? :kws (spec/* keyword?)))

匹配向量

(spec/conform ::cat '[af "5"])
=> {:sym af, :str "5"}
(spec/conform ::cat '[af "5" :key])
=> {:sym af, :str "5", :kws [:key]}

还有列表

(spec/conform ::cat '(af "5"))
=> {:sym af, :str "5"}
(spec/conform ::cat '(af "5" :key))
=> {:sym af, :str "5", :kws [:key]}

如果我们想限制这一点,我们可以尝试使用 spec/tuple;但遗憾的是它只匹配固定长度的向量,即它至少需要一个空列表作为元组的最后一部分:

(spec/def ::tuple (spec/tuple symbol? string? (spec/* keyword?)))
(spec/conform ::tuple '[af "5"])
=> :clojure.spec/invalid
(spec/exercise ::tuple)
=> ([[r "" ()] [r "" []]] [[kE "" (:M)] [kE "" [:M]]] ...)

我们还可以尝试使用 spec/and:

::cat 添加附加条件
(spec/def ::and-cat
  (spec/and vector? (spec/cat :sym symbol? :str string? :kws (spec/* keyword?))))

匹配得很好

(spec/conform ::and-cat '[af "5"])
=> {:sym af, :str "5"}
(spec/conform ::and-cat '[af "5" :key])
=> {:sym af, :str "5", :kws [:key]}
(spec/conform ::and-cat '(af "5" :key))
=> :clojure.spec/invalid

但遗憾的是无法生成它自己的数据,因为 spec/cat 的生成器只生成列表,当然不符合 vector? 谓词:

(spec/exercise ::and-cat)
=> Couldn't satisfy such-that predicate after 100 tries.

总结一下:如何编写既能接受又能生成像 [hi "there"] [my "dear" :friend] 这样的向量的规范?

也可以将问题改写为 "Is there an alternative to spec/cat which generates vectors instead of lists?" 或 "Is it possible to pass a :kind argument to spec/cat?" 或 "Can I attach a generator to a spec which takes the output of the original generator and casts it to a vector?"。

事实证明,从 clojure-1.9.0-alpha15 开始,没有简单的方法可以解决这个问题。一种可能的解决方案是修改生成器以将 cat 给出的序列转换为向量,如:

(spec/def ::solution
  (let [s (spec/cat :sym symbol? :str string? :kws (spec/* keyword?))]
    (spec/with-gen s #(gen/fmap vec (spec/gen s)))))

然后我们可以看到它生成并接受了正确的数据:

(spec/exercise ::solution)
=> ([[T ""] {:sym T, :str ""}]
    [[t* "Z" :g*] {:sym t*, :str "Z", :kws [:g*]}]
    [[G?8 "td" :*K/j] {:sym G?8, :str "td", :kws [:*K/j]}])

虽然有一个问题,但规范并未验证输入是否为向量,它接受列表等序列:

(spec/conform ::solution '(N-G.?8?4/- "" :G7y_.?Gx_/Oy1Dv :g!/Ooh0 :N-??h/o+cN))
=> {:sym N-G.?8?4/-, :str "", :kws [:G7y_.?Gx_/Oy1Dv :g!/Ooh0 :N-??h/o+cN]}

独立于规范创建正则表达式模式:

(require '[clojure.spec :as s] '[clojure.spec.gen :as gen])

(def pattern 
  (s/cat :sym symbol? :str string? :kws (s/* keyword?)))

(s/def ::solution
  (s/with-gen (s/and vector? pattern) 
              #(gen/fmap vec (spec/gen pattern))))

(s/valid? ::solution '(af "5" :key))  ;; false

(s/valid? ::solution ['af "5" :key])  ;; true

(gen/sample (s/gen ::solution) 4)
;; ([m ""] [. "" :Q] [- "" :?-/-9y :_7*/!] [O._7l/.?*+ "z" :**Q.tw.!_/+!gN :wGR/K :n/L])

要添加 Alex 的解决方案,这里有一个定义 vector-cat 正则表达式操作的宏:

(defmacro vcat
  "Takes key+pred pairs, e.g.

  (vcat :e even? :o odd?)

  Returns a regex op that matches vectors, returning a map containing
  the keys of each pred and the corresponding value. The attached
  generator produces vectors."
  [& key-pred-forms]
  `(spec/with-gen (spec/and vector? (spec/cat ~@key-pred-forms))
     #(gen/fmap vec (spec/gen (spec/cat ~@key-pred-forms)))))