Note to self: That’s a very unhelpful error message…

In 356 we added a check for that we have a cljs runtime before Calva connects the build. (Previously the Calva UI had things look like the cljs REPL was connected even though there might not even be a cljs app running.) Obviously we need some error handling there… What shadow-cljs version are you using?

Please file an issue about this, @U2BDZ9JG3.

Here is the shadow-cljs version from package.json:

"shadow-cljs": "2.20.12",

https://github.com/BetterThanTomorrow/calva/issues/2200

And what’s the shadow-cljs version in your deps.edn? (Assuming you are using deps: true or similar.)

Hmm, no, we don’t use deps: true (yet). We define dependencies inside the shadow-cljs.edn. So, it should be using Shadow CLJS from the package.json. There is no other place in this project that we define a Shadow CLJS version.

Thanks. I’ll try to reproduce and get back to you.

Hi! Calva has commands for refreshing namespaces. You should find them in the command palette. Do any of those work?

The commands do work fine. Although I could not find a command that would reload the current file along with nested dependencies. e.g: Supose I have 3 namespaces: app.ns1; app.ns2; app.ns3 app.ns1 requires app.ns2 and app.ns2 requires app.ns3. if I make a change in app.ns3 to test something in app.ns1, I need to reload all three files in order to get the behaviour to update accordingly and test it. I was woundering if there is any command that simulates the behaviour of (require 'ns :reload-all)

You can give yourself that command pretty easily with custom repl command snippets. I’m on the run, but search http://calva.io and you should find.

Great! Thanks!

Actully pretty tricky to get a https://calva.io/custom-commands/ snippet to scrape the relevant data for adding :reload-all automatically to all your requires. I think I've mostly got it solved for you. But I've hit a brick wall that I think is probably a bug. Hopefully @U0ETXRFEW will have a chance to help figure out a solution. You can try out the code below if you point the symbol file-text to a string 1+ Clojure forms, with the first form being ns with the same namespace as the current file. The code evaluates that ns form after adding :reload-all to each require. Then for simplicity, it simply evaluates the whole file-text string, instead of trying to strip out the un-patched version of the ns form.

(let [update-seq-when (fn [xs pred y]
                        (map #(if (pred %) y %) xs))
      ns-form ((requiring-resolve 'clojure.edn/read-string) file-text)
      require? #(when (coll? %) (#{:require} (first %)) %)
      require-form (some require? (rest ns-form))
      deps (rest require-form)
      add-reload #(conj (vec %) :reload-all true) 
      reload-ns-form (when (pos? (count deps))
                       (->> deps
                            (map add-reload)
                            (cons :require)
                            (update-seq-when ns-form require?)))]
  (eval reload-ns-form)
  (load-string file-text))

{...

    "calva.customREPLCommandSnippets": [
        {
            "name": "Reload required deps and evaluate current namespace",
            "key": "eval-ns-reload",
            "snippet": "(let [update-seq-when (fn [xs pred y] (map #(if (pred %) y %) xs)) ns-form ((requiring-resolve 'clojure.edn/read-string) $file-text) require? #(when (coll? %) (#{:require} (first %)) %) require-form (some require? (rest ns-form)) deps (rest require-form) add-reload #(conj (vec %) :reload-all true) reload-ns-form (if-not (pos? (count deps)) ns-form (->> deps (map add-reload) (cons :require) (update-seq-when ns-form require?)))] (eval reload-ns-form)) (load-string $file-text)",
        },
    ],

...}

{...

    {
        "key": "ctrl+alt+shift+r",
        "command": "calva.runCustomREPLCommand",
        "args": "eval-ns-reload"
    }

...}

I might be misunderstanding something. I was thinking something like

{
            "key": "r",
            "name": "Require-reload current namespace",
            "snippet": "(require '$ns :reload-all)"
        }

Using your example, If I evaluate app.ns1, then open, make changes to, save, and close app.ns3, those changes will not be reflected in app.ns1 until I explicitly eval or reload app.ns3 or restart the repl. If I understand the request correctly, this is the type of scenario of concern; and the desire is to eval app.ns1 from app.ns1 in such a way that app.ns3 is automatically reloaded because it is a transitive dependency.

Again, I might have misunderstood. But the snippet I use, and pasted above, allows for: 1. Have app.ns1 require app.ns2, which require app.ns3 2. Load app.ns1 3. Evaluate something in app.ns1 which uses something in app.ns2 which in turn uses something in app.ns3 4. Update the something in app.ns3, save the file. 5. Open the editor with app.ns1, run the snippet ctrl+alt+space r 6. Evaluate something in app.ns1 which uses something in app.ns2 which in turn uses something in app.ns3 -> updated result

Cool. Way simpler than what I did. It never even occurred to me that requiring the current namespace into itself might work correctly.

Thanks a lot guys!! The combination of your answers worked perfectly! Thanks a lot!

I think most keys are the same. Anything in particular you miss?

ctrl+alt+ c and enter is not working

Does it work from the command palette? How are you starting the repl?

ctrl +al +c and cltr +alt + j

The REPL starts and connects fine? What kind of project is it? Also 😄: > Does it work from the command palette?

i just noticed that repl is not started properly, I am taking a look

I stole it from CIDER. Could date back to some Symbolic Lisp Machine, or even earlier.

oh yeah, I was thinking that CIDER probably did that before Light Table

Since this has come up, I often find that the inline display -- at the end of the line -- is not visible, since I do not have word wrap on and keep my editor window tightly-cropped (I hate wrapped lines, so this forces me to break code into multiple lines to keep lines reasonably short). I would love an option for Calva to show the inline display below the current line, essentially indented to match the start of the form evaluated.

(It's not a really high priority most of the time since I have evaluations go through Portal middleware and into one of my two Portal windows -- I have Calva's REPL output hidden -- but there are times, when don't have the Portal middleware running, e.g., in a remote server, and so I have to try to remember to use ctrl+alt+t t or ctrl+alt+t space instead of the more muscle memory ctrl+enter or alt+enter 🙂 )

fwiw the evaluation result runs off screen even with word wrap