The primary issue being that RUN curl -sL | bash - no longer works to install node in a FROM clojure dockerfile

Are you building your artifacts in the docker build, or are you just running build artifacts in docker?

Docker compose is for running docker images. It sounds like you're trying to build them.

I use Node inside Docker but outside of the Clojure image (inside Docker - for safety reasons because screw Node and its mistakes) and then simply ptu the built JS bundle into the Clojure container. Couldn't be happier with the setup.

Multi stage dockerfile is extremly helpful with this problem:

FROM node:latest AS builder

# Install clojure related tools
RUN ...

# Build artifacts
RUN ...

FROM clojure:latest

COPY --from=builder /path/to/compiled/artifacts /path/to/resources

@UEENNMX0T I use the clojure:temurin-20-alpine base image and then install npm with apk. It installs node as an npm dependency. That may or may not work depending on if you need Nodejs 18 specifically.

the docker container currently runs mongo, lein repl, and then two calls to npm to watch changes to the cljs (shadow-cljs), and to watch for changes to the css. that's why i can't just using COPY --from=builder ...

in the past, i'd run these in a separate tab of my console, but i'm using a new machine and don't feel like setting up all of the extra stuff to get it to work. other devs on our team (open source project) rely on it too

you still can run extra commands from separate tab: docker compose exec container -ti npm ...

i'm looking to be able to run docker compose up -d and then have everything running behind the scenes so all i have to do is jack into the repl

this is going against docker principles but you can start container with a chain of commands:

npm ... & \
npm ... & \
lein repl

lol clever

I'm using clojure:temurin-17-tools-deps-bullseye-slim and then I followed the instructions here for Debian: https://github.com/nodesource/distributions

there are examples of uber images different tech stacks in the same container (for development) - e.g. https://github.com/penpot/penpot/tree/develop/docker/devenv

If you just want a straight hacked up dockerfile here is mine lol: https://github.com/chase-lambert/clojure-app-template/blob/main/Dockerfile

oh that's very nice, i'll try that out

Can you share what doesn’t work with data.xml?

it's not the namespace that's the problem

it's a core.match macro being expanded into something wild

it is because the name of the class file for the generated code is too long, so the file cannot be written to disk because it exceeds the file name length limit of whatever filesystem you are using

for the classes the compiler generates the namespace name is usually part of, but not the whole class file name, and usually the namespace name ends up being part of the directory tree the class gets written to, but the name itself is something generated by the compiler

the names generated by the compiler typically reflect the structure of the code, for example a "$fn" will get tacked on for every level of anonymous functions

so particularly hairy very nested expressions can result in long class names

core.match can generate some rather hairy nested expressions

additionally sometimes for certain expressions the compiler will insert immediately invoked functions, for example E becomes ((fn [] E)) , which can add nesting where none may be immediately apparent

I see… is there a way to write match expressions to not nest as much? My expression is (to my eyes) pretty simple:

(def almuzara "ALMUZARA")

(defn record->publisher [{:keys [Proveedor Imprint Tema] :as record}]
  (match [Proveedor Imprint Tema]
    ["editorial almuzara" _ _] almuzara
    ;; 27 more lines like the previous
    :else (throw (Exception. (str "Found an unexpected proveedor/imprint/tema combo:" (select-keys record [:Proveedor :Imprint :Tema]))))))

it is kind of occupational hazard of core.match's implementation, and likely why its use isn't as wide spread as you might expect

I have a little pattern matching library that never really got a release https://github.com/hiredman/match2/tree/master which is much more naive about how it compiles patterns, and so results in less nesting hairy stuff

I will check that out, thank you for the link and for the explanation!

Wow. This just scared me away from core.match.

There is in 1.12! sync-deps to the rescue

:alpha      {:extra-deps {org.clojure/clojure {:mvn/version "1.12.0-alpha5"}}}

Whoa, I just tried (sync-deps :aliases [:bench]) and it worked beautifully. I was so excited about add-lib thing that I completely missed awesomeness of sync-deps. Thank you Alex!

I have a hot key bound in Calva that prompts for aliases and then does sync-deps 🙃 1.12 is awesome-sauce!

Wow, @U04V70XH6, is VS Code your daily driver? If so, I rely on the same text editor as Sean Corfield and that's pretty cool. (Good to be in good company :)

> I have a hot key bound in Calva Oh, that'd be nice to have in Emacs. What's the best way of figuring out all available aliases? Including ~/.clojure/deps.edn, etc.? Also, is it possible to (relatively easily to) determine which aliases were loaded? Ideally, it should only prompt with aliases that either not yet loaded, or modified.

clj -X:deps aliases provides alias info, happy to consider changes if needed

@U02U1T66REZ All my config is here: https://github.com/seancorfield/vscode-calva-setup (key bindings, settings, Calva REPL snippets, Joyride scripts) plus I use https://github.com/seancorfield/dot-clojure for aliases and for starting my REPL. Here's the alias I use at work for starting a REPL:

:vscode-calva-jack-in
  {:extra-deps {com.datomic/dev.datafy {:git/sha "4a9dffb"
                                        :git/tag "v0.1"
                                        :git/url ""}
                djblue/portal {:mvn/version "0.49.1"
                               #_#_:local/root "/home/sean/oss/portal"}
                io.github.seancorfield/dot-clojure
                {:git/tag "v1.0.2"
                 :git/sha "6a3f903"}
                io.github.stuarthalloway/reflector
                {:git/sha "93a0c19b4526c1180959e940202928d35e5c3cef"}
                jedi-time/jedi-time {:mvn/version "RELEASE"}
                party.donut/dbxray {:mvn/version "RELEASE"}}

{
  "calva.replConnectSequences": [
    {
      "name": "World Singles Backend (Jack-In)",
      "projectType": "deps.edn",
      "cljsType": "none",
      "autoSelectForJackIn": true,
      "projectRootPath": ["."],
      "menuSelections": {
        "cljAliases": ["build", "dev", "+default", "test", "vscode-calva-jack-in"]
      }
    },
    {
      "name": "World Singles Backend (Connect)",
      "projectType": "deps.edn",
      "cljsType": "none",
      "autoSelectForConnect": true,
      "projectRootPath": ["."]
    }
  ],
  "calva.jackInEnv": {
    "LOG4J_CONFIGURATION_FILE": "log4j2-test.properties"
  },
  "calva.autoConnectRepl": true,
  "workbench.colorTheme": "Default Dark Modern",
  "workbench.preferredDarkColorTheme": "Default Dark Modern",
  "workbench.iconTheme": "vs-seti",
}

I'm just surprised all the wizards here don't all exclusively use emacs haha

I used to. I used Emacs way back in the 17.x days through the early 19.x days. I used Together/J for my Java years (late '90s onward for a while). I picked Emacs back up when I came to Clojure (2010) but never really settled back into it, and after trying tons of configs, I switched to "modern" editors and used a bunch, settling on Atom until I switched to VS Code -- and with cljs customization via Joyride, it's programmable like Emacs, while having great integrations with everything else I use (like Jira/BitBucket at work!). And Calva is awesome, with LSP/clj-kondo for static analysis too.

Then we junior devs are lucky beneficiaries! ... I started in Atom and adopted VS Code + Calva after struggling with Proto REPL. Very happy to be here! 🔵:large_green_circle:

What does clj -X:deps aliases do under the hood? What's the equivalent for this in the REPL? I'm thinking, if I already have a connected repl, wouldn't be faster/straightforward to send something to the REPL instead?

@U0G75ARHC Source: https://github.com/clojure/tools.deps.cli/blob/main/src/main/clojure/clojure/tools/deps/cli/api.clj#L139

create-edn-maps is here https://github.com/clojure/tools.deps/blob/master/src/main/clojure/clojure/tools/deps.clj#L774

@U02U1T66REZ I really liked Atom + ProtoREPL but it stopped being maintained and updates to Atom broke things. That's when I switched to Chlorine (and a Socket REPL). And then when it looked like Atom would also stop being maintained, I switched to VS Code + Clover (a port of Chlorine), and then Clover sort of stopped being maintained(!) so I switched to Calva... and it happened to coincide with the VS Code version of Portal appearing (which I'd switched to from Reveal, which I'd switched to from REBL 🙂 ).

Okay, so I think the lesson for me here is that change is the only constant, there is no perfect tool and to stay nimble & tinker often! I like it!

I need to add a version that works even if org.clojure/tools.deps.cli is not in the classpath

I would expect it to not be on the classpath in general...

Even the new-in-1.12 stuff that provides access to the CLI tooling stuff does it by shelling out to the actual CLI.

I'v fixed the snippet. Now it works even if it's not on the classpath. First, it tries it anyway, since that would be slightly faster