How viable is it to use shadow's clojure APIs directly as opposed to using npx shadow-cljs ? Does the NodeJS API do a lot of magic (middleware, etc.) in addition to calling (server/start!)?

What's the minimum Clojure code to replicate the exact behavior of npx shadow-cljs server?

@zhuxun2 https://code.thheller.com/blog/shadow-cljs/2017/11/18/the-many-ways-to-use-shadow-cljs.html

to get the exact behavious of server you just have (server/start!). thats all as far as the shadow-cljs build server is concerned

however the "connect to server" part of the node shadow-cljs command cannot really be replicated

so you probably want to do all other stuff over the REPL or UI and now use further CLI command to start a watch for example

Hello, @thheller. Could you help me, please. I’m working on migrating from figwheel-main to shadow-cljs and unfortunately I can’t change the structure of the project. I have a backend module that serves my static files via /static route, and I have a frontend module that compiles js to the backend resource directory backend/resources/public/frontend2/assets/js. I don’t have a file index.html - the server generates it itself. JS files are reloaded correctly, but I have some problems with reloading resources (CSS). Any changes to CSS files are not handled. I have a webpack configuration that compiles less files to the backend resources directory /backend/resources/public/frontend2/assets/bundle.css I added the parameters :watch-dir and :watch-path, but it didn’t help.

[:link {:rel "stylesheet", :href "static/frontend2/assets/bundle.css", :type "text/css"}]
[:script {:src "static/frontend2/assets/js/app.js", :type "text/javascript"}]

:app  {:target     :browser
       :output-dir "../backend/resources/public/frontend2/assets/js"
       :asset-path "static/frontend2/assets/js"
       :modules    {:app {:init-fn frontend.core/-main}}

       :devtools   {:preloads  [devtools.preload]
                    :watch-dir "../backend/resources/public/frontend2/assets"
                    :watch-path "static/frontend2"}

@just.sultanov what is the path you include to css in the html?

https://code.thheller.com/blog/shadow-cljs/2021/05/13/paths-paths-paths.html might help

I'm guessing its :watch-dir "../backend/resources/public" but can't say exactly without known the paths

(meaning why is the path in that html relative? what is the path of the page you are calling?)

[:link {:rel "stylesheet", :href "/static/frontend2/assets/bundle.css", :type "text/css"}]?

Backend generates the following paths:

[:link {:rel "stylesheet", :href "static/frontend2/assets/bundle.css", :type "text/css"}]
[:script {:src "static/frontend2/assets/js/app.js", :type "text/javascript"}]

then its :watch-dir "../backend/resources/public/frontend2" and :watch-path "static/frontend2"

or :watch-dir "../backend/resources/public" and :watch-path "static"

the blog post explains this I hope

Yes, the path to static files is relative. This is provided by the compojure context. Well, thank you. I will try to use absolute paths

thats fine, just makes it harder to answer the question without knowing the full exact paths 😉

Hi, new to shadow-cljs and CLJS in general. I've tried out both figwheel-main and shadow-cljs in some learning projects with basic configurations, and I've noticed that when using figwheel, I can get completions for goog.dom functions in a CIDER REPL by simply typing "goog.dom/...", while I don't get auto-completion in shadow-cljs builds for vars in that namespace. Is this related to this issue https://github.com/thheller/shadow-cljs/issues/834 ?

not a clue how cider autocomplete works so can't say

is there an alternative to nuking .shadow-cljs when a https://github.com/babashka/sci/blob/133a7565749ac6cd5a8308182f0b1c7fc47e8a3d/src/sci/impl/namespaces.cljc#L53-L64 depending on the contents of an env var? Maybe a way to whitelist env vars that shadow can take into account for caching?

or is there a way to rewrite this macro so it plays well with shadows caching?

what I do in one of my projects is https://github.com/clavascript/clavascript/blob/a5a63bf19025f8535a5a3f3f4ef8c1dde2d4786c/bb/tasks.clj#L61

so, delete the cache file for that macro-related ns

https://github.com/thheller/shadow-cljs/blob/732577ea5257f6418d4926d10f5c0e3261303e8b/src/main/shadow/build/compiler.clj#L742-L746

could store it there and get it out of the (get-in @cljs.env/*compiler* [:options :external-config :whatever-you-added]) during compilation

ah nice.

shadow does a lot better in cache invalidation regarding macros than the regular cljs compiler, correct? But reading from the env will fail bust its caching strategy?

yes, it does not check the env

basically it checks all the options above + the files involved in the compilation

thats all it looks at

ok, have you considered being able to signal (maybe via metadata) on a macro var that it should not be cached or does that not make sense?

well to know if it can use the cache it must be able to derive it from only the file and compare before compilation

I guess there could be some way for a macro to signal "hey, please don't cache this file"

if cache is never written it can never be read

ah so the cache is on a file level, not on the level of vars?

@mkvlr have you considered the :cache-blockers config here? https://shadow-cljs.github.io/docs/UsersGuide.html#_compiler_cache

@isak nope, wasn’t aware of this, thanks!

that also works yes