Clojurians Log v2

For posterity, I created the two following aliases:

:jar     {:extra-deps {seancorfield/depstar {:mvn/version "2.0.165"}}
                     :exec-fn    hf.depstar/uberjar
                     :exec-args  {:jar        "my-lib.jar"
                                  :aot        true
                                  :main-class my-lib.core}}

:deploy  {:extra-deps    {slipset/deps-deploy {:git/url "<https://github.com/Swirrl/deps-deploy.git>"
                                                          :sha     "588e4b96b33820a6ee1e35663410450ccb876b30"}}
                     :override-deps {s3-wagon-private/s3-wagon-private {:mvn/version "1.3.2"}}
                     :exec-fn       deps-deploy.deps-deploy/deploy
                     :exec-args     {:repository {"releases" {:url "<s3p://my-s3-route/releases/>"}}
                                     :artifact   "my-lib.jar"
                                     :installer  :remote}}

can I add :jvm-opts without adding it in an alias?

I don't think that works

I don’t know the internals of tools-deps. Does that make sense? Coming from leiningen it seems like a strange omission.

-J

You can add jvm opts at the point of invocation with clj -J... , but not in deps.edn file

@peder.refsnes You can vote for it here: https://ask.clojure.org/index.php/7834/support-default-jvm-opts

I continue to feel this is weirdly at odds with using tools

and I'm going to continue not implementing it until that makes sense :)

Yeah, I don't see much of a use case for it either. I have only ever needed to set JVM options for a specific task, which means a specific alias.

But people do keep asking for it, so hopefully someone will come up with a justification for it that is convincing... 🙂

(and "leiningen supports it" isn't at all convincing to me 🙂 leiningen does a lot of things that I don't think are necessarily good ideas)

I do actually get the use case, it's just that's not the only use case

What sort of JVM options can you imagine folks wanting to set for all operations in a deps.edn project? I'm curious...

well the use case is where "all" is the two things you usually do

and you "always" want to do something like set a system property for your app or a particular garbage collector or whatever

I think in some cases "all" is really "when I run my repl I want default stuff to be set"

that is in particular an interesting case where having a "default" alias might be useful

not too different than leiningen :default profile really

My current (and only) reason for wanting this is for playing around with Dragan Djurics ML libraries. Seems it needs a workaround for java 9+. https://github.com/uncomplicate/neanderthal/blob/master/examples/hello-world/project.clj#L11-L12

Which you could put in an alias, you just need to remember to use it then

Indeed, but it’s annoying when you forget 🙂 I’m all for features carrying their weight, just thought I’d ask.

I think I would have a :run alias for running the app so it's just clojure -M:run and perhaps a :dev alias for running a REPL and just live with the duplication of :jvm-opts...

I find lein's default profiles confusing, it's always hard to tell what's enabled/on the classpath

Indeed. Leiningen's "helpfulness" is one of the things I always liked the least about it.

couldn’t you use that same reasoning to argue against having a default`:deps` key?

Although some folks think this is a bit much clj -M:repl:reflect:jedi-time:reveal:j14:classes:add-libs:dev+test 🙂

@peder.refsnes having a base set of dependencies that almost every use case builds on is very useful tho'... same with a base set of :paths...

I just have a :dev alias and have my editor always start a repl with that alias. It's pretty easy in emacs on both cider and inf-clojure. Do other editors offer something like this?

I always start my REPL separately (and leave it running for ages -- my current REPL was started on December 24th!) and then connect one or more editors or other tools to it.

@seancorfield of course, just being pedantic 🙂

That's impressive! I can't seem to go a day without some catastrophic crash that I brought on myself

how do you ensure your code is live when e.g. switching between branches?

@vlaaad We rarely have long-lived branches and for the most part the code changes across multiple branches are accretive, rather than contradictory. I almost never need to require ... :reload-all (but I do have it bound to a hot key in VS Code). I occasionally need to remove a namespace (and then load it back in), if I'm changing aliases around (and, again, I have remove-ns bound to a hot key as well).

(and I tend to develop against multiple server processes all running from inside the REPL -- we have an "everything" subproject in our monorepo that has :local/root dependencies on all the other subprojects so we can start a REPL there and have all source/test code and all dependencies available in a single REPL)