Mapping a function on the values of a map in Cloju

问题:

I want to transform one map of values to another map with the same keys but with a function applied to the values. I would think there was a function for doing this in the clojure api, but I have been unable to find it.

Here's an example implementation of what I'm looking for

(defn map-function-on-map-vals [m f]
  (reduce (fn [altered-map [k v]] (assoc altered-map k (f v))) {} m))
(println (map-function-on-map-vals {:a "test" :b "testing"} #(.toUpperCase %)))
{:b TESTING, :a TEST}

Does anybody know if map-function-on-map-vals already exists? I would think it did (probably with a nicer name too).

回答1:

I like your reduce version just fine. I think it's idiomatic. Here's a version using list comprehension anyways.

(defn foo [m f]
  (into {} (for [[k v] m] [k (f v)])))

回答2:

You can use the clojure.algo.generic.functor/fmap:

user=> (use '[clojure.algo.generic.functor :only (fmap)])
nil
user=> (fmap inc {:a 1 :b 3 :c 5})
{:a 2, :b 4, :c 6}

回答3:

Here is a fairly typical way to transform a map. zipmap takes a list of keys and a list of values and "does the right thing" producing a new Clojure map. You could also put the map around the keys to change them, or both.

(zipmap (keys data) (map #(do-stuff %) (vals data)))

or to wrap it up in your function:

(defn map-function-on-map-vals [m f]
    (zipmap (keys m) (map f (vals m))))

回答4:

Taken from the Clojure Cookbook, there is reduce-kv:

(defn map-kv [m f]
  (reduce-kv #(assoc %1 %2 (f %3)) {} m))

回答5:

Here's a fairly idiomatic way to do this:

(defn map-function-on-map-vals [m f]
        (apply merge
               (map (fn [[k v]] {k (f v)})
                    m)))

Example:

user> (map-function-on-map-vals {1 1, 2 2, 3 3} inc))
{3 4, 2 3, 1 2}

回答6:

map-map, map-map-keys, and map-map-values

I know of no existing function in Clojure for this, but here’s an implementation of that function as map-map-values that you are free to copy. It comes with two closely related functions, map-map and map-map-keys, which are also missing from the standard library:

(defn map-map
    "Returns a new map with each key-value pair in `m` transformed by `f`. `f` takes the arguments `[key value]` and should return a value castable to a map entry, such as `{transformed-key transformed-value}`."
    [f m]
    (into (empty m) (map #(apply f %) m)) )

(defn map-map-keys [f m]
    (map-map (fn [key value] {(f key) value}) m) )

(defn map-map-values [f m]
    (map-map (fn [key value] {key (f value)}) m) )

Usage

You can call map-map-values like this:

(map-map-values str {:a 1 :b 2})
;;           => {:a "1", :b "2"}

And the other two functions like this:

(map-map-keys str {:a 1 :b 2})
;;         => {":a" 1, ":b" 2}
(map-map (fn [k v] {v k}) {:a 1 :b 2})
;;    => {1 :a, 2 :b}

Alternative implementations

If you only want map-map-keys or map-map-values, without the more general map-map function, you can use these implementations, which don’t rely on map-map:

(defn map-map-keys [f m]
    (into (empty m)
        (for [[key value] m]
            {(f key) value} )))

(defn map-map-values [f m]
    (into (empty m)
        (for [[key value] m]
            {key (f value)} )))

Also, here’s an alternative implementation of map-map that is based on clojure.walk/walk instead of into, if you prefer this phrasing:

(defn map-map [f m]
    (clojure.walk/walk #(apply f %) identity m) )

Parellel versions – pmap-map, etc.

There are also parallel versions of these functions if you need them. They simply use pmap instead of map.

(defn pmap-map [f m]
    (into (empty m) (pmap #(apply f %) m)) )
(defn pmap-map-keys [f m]
    (pmap-map (fn [key value] {(f key) value}) m) )
(defn pmap-map-values [f m]
    (pmap-map (fn [key value] {key (f value)}) m) )

回答7:

I'm a Clojure n00b, so there may well be much more elegant solutions. Here's mine:

(def example {:a 1 :b 2 :c 3 :d 4})
(def func #(* % %))

(prn example)

(defn remap [m f]
  (apply hash-map (mapcat #(list % (f (% m))) (keys m))))

(prn (remap example func))

The anon func makes a little 2-list from each key and its f'ed value. Mapcat runs this function over the sequence of the map's keys and concatenates the whole works into one big list. "apply hash-map" creates a new map from that sequence. The (% m) may look a little weird, it's idiomatic Clojure for applying a key to a map to look up the associated value.

Most highly recommended reading: The Clojure Cheat Sheet .

回答8:

(defn map-vals
  "Map f over every value of m.
   Returns a map with the same keys as m, where each of its values is now the result of applying f to them one by one.
   f is a function of one arg, which will be called which each value of m, and should return the new value.
   Faster then map-vals-transient on small maps (8 elements and under)"
  [f m]
  (reduce-kv (fn [m k v]
               (assoc m k (f v)))
             {} m))

(defn map-vals-transient
  "Map f over every value of m.
   Returns a map with the same keys as m, where each of its values is now the result of applying f to them one by one.
   f is a function of one arg, which will be called which each value of m, and should return the new value.
   Faster then map-vals on big maps (9 elements or more)"
  [f m]
  (persistent! (reduce-kv (fn [m k v]
                            (assoc! m k (f v)))
                          (transient {}) m)))

回答9:

I like your reduce version. With a very slight variation, it can also retain the type of records structures:

(defn map-function-on-map-vals [m f]
  (reduce (fn [altered-map [k v]] (assoc altered-map k (f v))) m m))

The {} was replaced by m. With that change, records remain records:

(defrecord Person [firstname lastname])

(def p (map->Person {}))
(class p) '=> Person

(class (map-function-on-map-vals p
  (fn [v] (str v)))) '=> Person

By starting with {}, the record loses its recordiness, which one might want to retain, if you desire the record capabilities (compact memory representation for instance).

回答10:

I'm wondering why nobody has mentioned the specter library yet. It has been written to make this kind of transform easy to code (and, even more important, the written code easy to understand), while still being very performant:

(require '[com.rpl.specter :as specter])

(defn map-vals [m f]
  (specter/transform
   [specter/ALL specter/LAST]
   f m))

(map-vals {:a "test" :b "testing"}
          #(.toUpperCase %))

Writing such a function in pure Clojure is simple, but the code gets much trickier once you move on to highly nested code composed of different data structures. And this is where specter comes in.

I recommend watching this episode on Clojure TV which explains the motivation behind and details of specter.