is it possible to transform the keys in a map on just one level with specter?

for example {:a 1, :b {:c 2}} => {:aa 1, :bb {:c 2}}

so, this example: https://clojurians.slack.com/archives/clojure/p1470213401002703

(defnav KW-NS []
  (select* [this kw next-fn]
    (next-fn (namespace kw)))
  (transform* [this kw next-fn]
    (let [kw-name (name kw)
          new-ns (next-fn (namespace kw))]
      (keyword new-ns kw-name)
      )))

(setval [ALL FIRST KW-NS] "user" {:a 1 :b 2 :c 3})
;; => {:user/a 1, :user/b 2, :user/c 3}

how does it work?

so KW-NS navigates to the namespace portion of a keyword

is it the defnav portion that you're asking about in particular?

yes

there's two code paths, one for select and one for transform

select must call next-fn on the "navigated to value(s)"

in this case, it's just (next-fn (namespace kw))

@nathanmarz: what is next-fn in the example with setval?

it's the rest of the navigation + final transformation

so in this case next-fn will be (fn [v] "user")

if there was more navigation after KW-NS, it would be the transform* code paths for the subsequent navigators

@nathanmarz: specter takes constants and interprets it as a function?

setval is a thin wrapper around transform

(setval [ALL FIRST KW-NS] "user" {:a 1 :b 2 :c 3}) == (transform [ALL FIRST KW-NS] (fn [_] "user") {:a 1 :b 2 :c 3})

ok understood that part

I think I'll just have to read up on the defnav docs

here's how two navigators get composed into one navigator: https://github.com/nathanmarz/specter/blob/master/src/clj/com/rpl/specter/impl.cljx#L345

that's the core of how specter works

your linear path of navigators becomes a single nested function

I think I'm getting it, but you could also have written: (transform [ALL FIRST] (fn [kw] (keyword "user" (name kw)))) right? but your example is more abstract

correct

much more reusable and composable

cool stuff 🙂

any reason specter is still using cljx? compatibility with previous clj versions?

yea

though finally dropping support for 1.6 as of specter 0.12.0

just need to do the work now of porting to cljc