Last summer I moved boodle from Figwheel to Figwheel Main. I was pretty satisfied with this setup, but the rise of shadow-cljs in the Clojure community has been tickling my curiosity over and over again. While configuring shadow-cljs I took the chance to set lein aside and finally explore clj and deps.edn, another thing on my to-do list that has been waiting for me for some time.

Let’s start with deps.edn:

{:deps
{
;; … Clojure dependencies …
}

:paths ["src/clj" "resources"]

:aliases {:run {:main-opts ["-m" "boodle.core"]}
:test {:extra-paths ["test/clj"]
:extra-deps {lambdaisland/kaocha {:mvn/version "0.0-266"}}}
:outdated {:extra-deps {olical/depot {:mvn/version "1.4.0"}}
:main-opts ["-m" "depot.outdated.main"]}}}


One thing worth of notice here is :paths. I included the resources directory in it because I serve it via compojure/http-kit and that’s where I put the static assets used in my main index page. This solution works when I run mount.core/start via REPL and when I use the :run alias from the command line.

Speaking of aliases, they are invaluable. With :run I can use clj -A:run instead of lein run, and with clj -A:outdated -a outdated I can easily check for outdated dependencies in my deps.edn thanks to depot. Unit tests are now handled by kaocha, and as per its instructions I have a bin/kaocha file which picks up the :test alias and runs all my tests.

I love the simplicity of deps.edn, it feels like a de-cluttered project.clj.

Configuring shadow-cljs was a bit more complicated. It took me a while to understand how to properly integrate it with CIDER. However, the manual is great and the fine people on Slack’s #shadow-cljs (thanks Ryan Haywood and Thomas Heller!) made it all nice and smooth.

This is my shadow-cljs.edn:

{:source-paths ["src/cljs"]
:nrepl {:port 8777
:middleware [refactor-nrepl.middleware/wrap-refactor]}
:dependencies [
;; … ClojureScript dependencies …
]
:builds {:boodle {:target :browser
:output-dir "resources/public/js"
:asset-path "/js"
:modules {:main {:entries [boodle.core]}}}}}


Basic setup, straight from the manual pages. Notice that :output-dir refers to the same resources directory configured in deps.edn. This tells shadow-cljs to place the build output exactly where I need.

The other important change is in the CLJS libraries I use. shadow-cljs does not support CLJSJS, so I had to install pikaday and moment via yarn.

$yarn add pikaday moment  Since boodle relies on re-frame, I had to install three other libraries to make shadow-cljs build my code happily: $ yarn add react react-dom create-react-class


In the namespace where I use them, the :require had to be changed:

(:require ;; … other requires …
["react-dom" :refer [findDOMNode]]


(let [default-opts {:field (findDOMNode this)
;; … other options …
}
;; … other ClojureScript code …
)


However, a project setup is not ready unless I can work on it via Emacs. These are the steps I made to have a working CLJ/CLJS development environment. When this issue will be fixed some of it may become unnecessary.

First, I added a couple of dependencies in shadow-cljs.edn:

:dependencies [;; … other dependencies …
[cider/cider-nrepl "0.18.0"]
[refactor-nrepl "2.4.0"]]


Then I removed everything Figwheel Main related from the .dir-locals.el file located in the root directory of boodle:

((nil
(cider-pprint-fn . zprint)
(cider-preferred-build-tool . clojure-cli)
(cider-known-endpoints . (("hathaway" "localhost" "8777")))
(cider-ns-refresh-after-fn . "mount.core/start")
(cider-ns-refresh-before-fn . "mount.core/stop"))
(emacs-lisp-mode
(flycheck-disabled-checkers . "emacs-lisp-checkdoc")))


I ran cider-jack-in-clj to have a CLJ REPL and check I didn’t break something on the Clojure side of boodle.

From a terminal, I used shadow-cljs watch boodle to have shadow-cljs build boodle and offer me a nice nREPL server listening on port 8777, the one I specified in shadow-cljs.edn.

Back in CIDER, cider-connect-cljs brought up a REPL connected to shadow-cljs nREPL server, and let me select the build I want it operating on (default is dev, I just typed in boodle). Connecting to boodle via the browser finalized the connection between my code and the CLJS environment.

A great journey. Once again, boodle proved fertile ground to understand the Clojure ecosystem and play with it to learn something new. If you want more details, the project is on my GitHub.