dosync Archive Pages Categories Tags

The Old Way, The New Way

31 December 2014

UPDATE: This post now contains obsolete information. Please read the new ClojureScript Quick Start instead

This post will walk you through using ClojureScript from master, using a contrib SNAPSHOT release, and getting a productive fast REPL setup using the new Node.js REPL support.

Imagine that you want to try out the new ClojureScript support in test.check 0.6.3-SNAPSHOT. You would really like to play around with the API at the REPL.

For kicks let’s use ClojureScript master:

git clone
cd clojurescript

We’re going to build ClojureScript from source and install into our local Maven cache:


You’ll see a bunch of stuff scroll by, the most important is this:

[INFO] ------------------------------------------------------------------------
[INFO] Building ClojureScript 0.0-2606
[INFO] ------------------------------------------------------------------------

This tells us the version number we’ll need to use in our project.clj. It will probably NOT be 0.0-2606 for you.

Let switch to the directory where our project will live and create it:

lein new mies test-fun
cd test-fun

We need to make several modifications to our project.clj file. It should look like this:

(defproject test-fun "0.1.0-SNAPSHOT"
  ;; so we can get SNAPSHOT releases
  {"sonatype-oss-public" ""}

  ;; replace NNNN with whatever you saw above
  :dependencies [[org.clojure/clojure "1.6.0"]
                 [org.clojure/clojurescript "0.0-NNNN"]
                 [org.clojure/test.check "0.6.3-SNAPSHOT"]]

  :plugins [[lein-cljsbuild "1.0.4"]]

  :source-paths ["src" "target/classes"]

  :clean-targets ["out/test_fun" "test_fun.js" "test_fun.min.js"]

  :cljsbuild {
    :builds [{:id "dev"
              :source-paths ["src"]
              :compiler {
                :output-to "test_fun.js"
                :output-dir "out"
                :optimizations :none
                :cache-analysis true                
                :source-map true}}]})

Let’s set things up so our ClojureScript experience is a bit snappier. It’s assumed that you’ve already got LEIN_FAST_TRAMPOLINE configured.

Let’s compile the ClojureScript bits we’re going to need once and for all:

lein trampoline run -m clojure.main
user=> (compile 'cljs.repl.rhino)
user=> (compile 'cljs.repl.node)
user=> (compile 'cljs.core)

Quit the REPL with CTRL-D.

Now we’re ready to start a REPL. This will demonstrate how bad the REPL experience was prior to enhancements to ClojureScript REPL support.

lein trampoline cljsbuild repl-rhino

On my 2010 Macbook Pro this takes 20 to 30 seconds to start even for subsequent runs.

This would be enough for me to walk away from my computer in utter disgust, all enthusiasm lost for exploring test.check.

This is because prior to the last ClojureScript changes REPLs did not cache compilation and analysis to disk. Every REPL invocation recompiles everything in memory again and streams it to the JS process.

Now let’s try the new Node.js REPL. Make a new file node_repl.clj, it’s contents should like the following:

(require '[cljs.repl :as repl] 
         '[cljs.repl.node :as node]) 

(repl/repl* (node/repl-env) 
  {:output-dir ".cljs_node_repl" 
   :cache-analysis true 
   :source-map true})"

Run the following:

lein trampoline run -m clojure.main node_repl.clj

The first time will be slow as we cache everything. Quit the REPL and try this again.

On my 2010 laptop the REPL launches in a couple of seconds. On newer hardware this should be quite a bit faster.

It’s time for test.check fun, require the generators namespace:

To quit, type: :cljs/quit
ClojureScript Node.js REPL server listening on 5001
ClojureScript:cljs.user> (require '[cljs.test.check.generators :as gen])

Now generate some data!

ClojureScript:cljs.user> (gen/sample (gen/vector gen/int))
([] [] [] [] [4 0 -4] [3 -4 3] [-4 -5 -3 -5 6 5] [-1 4] ...)

Check out some docstrings:

ClojureScript:cljs.user> (doc gen/sample)
([generator] [generator num-samples])
  Return a sequence of `num-samples` (default 10)
  realized values from `generator`.

The old REPL model provided an inferior development experience. By embracing the same caching infrastructure as provided to ClojureScript builds we can deliver a REPL experience far more in the line with the one enjoyed in Clojure today.