open source or to work on it?

open source

So this is just bouncing in my head

There are two kinds of build tools

1. The one that just does the stuff (build.sh) 2. The one that has an abstract notion of a "task graph"

and then within 2 there are spectrums of "is this tailored to a single language/platform" and "how much of the properties of the graph is used to do caching, parallelism, etc"

There are a lot of (2)

Buck, Bazel, Gradle, (Maven, Ant), mill etc

The hypothesis is that, instead of representing the task graph in a regular language / DSL or in specialized configuration files - represent it visually

Unreal Engine, the game engine, has a visual scripting system they call "Blueprints"

That sounds a bit, at the abstraction level, like what I understood the Boot build tool to be. Maybe someone just needs to write a UI on top of it.

I could be wrong about Boot.

Boot.clj

It is a full language, in a sense

So the idea is to make something similar, but which represents the "task graph" of a build tool explicitly

Why CLJS? Well, the libraries I know of for this (was told about) are react/JS things

https://reactflow.dev/ https://github.com/bhouston/behave-graph

and I am in full realization of my inability to make this

This was what I sketched up as a concept last night

Doing stuff like making a Multi Release Jar is very hard with build tools like maven. Feels almost simple if the graph is drawn out and editable

...if that makes sense

I don't think a build tool for CLJ(s,*) is that interesting since for the most part it doesn't seem like we need them

but for java, c, etc. it feels like it would be cool

http://www.lihaoyi.com/post/SowhatswrongwithSBT.html http://www.lihaoyi.com/post/BuildToolsasPureFunctionalPrograms.html

(some related stuff i've been reading)

but yeah - free project idea

cool

ever look at pathom? dependency resolution is easy work for a backward chaining logic engine, just model dependencies as attributes, and it comes with a UI pathom-viz that gets you 90% of the way there

@U797MAJ8M I'm unclear on how pathom relates?

(my brain isn't attuned to the problems of execution so is it maybe that?)

> The hypothesis is that, instead of representing the task graph in a regular language / DSL or in specialized configuration files - represent it visually you can represent a task graph in clojure as pathom resolvers -- it is surprising how powerful this abstraction is -- and get much of the visual part for free... but reading it again I think you may just mean parsing some existing data outputted by some tool, and converting that to something a charting library like cytoscape would understand?

@U797MAJ8M https://www.youtube.com/watch?v=bY6Nl-OEhSo&t=1s

I don't expect you to watch the whole thing obv. but this is what i am imagining for the user experience

> you can represent a task graph in clojure as pathom resolvers -- it is surprising how powerful this abstraction is -- and get much of the visual part for free... The point isn't to have code that is introspect-able to make a graph, its to have users engage directly with the visual representation

(~15 min in seems a good point to jump to)

you need to require your config namespace from inside app namespace

correct, you need (ns (:require [my.config :as cfg])) and then cfg/one.

Ok, will try that . It is strange because in a project I was referring, it is using the keyword :refer, but it only refer a value in the file yet other stuff is still available. I suppose I'll be able to understand that later

if you use a fully qualified name like my.app.config/one then it will work if that ns has been loaded for some other reason in the app. but it's not good form to rely on that.

(the other reason might be eg that another namespace has required it)

That's what I thought! Thanks

(ns (:require [my.config :refer (one)])) and just one is totally fine too

avoid using full qualified names such as my.config/one in code directly though. you are much better off using alias references (e.g. :as or :refer)

There are many, many similar instances of undefined behavior.

worth me raising it as a problem on http://ask.clojure.org do you think?

not really. it all ulitmately comes down on how JS is different to Java

null + 1 is fine in JS, in the JVM it throws

indeed. this is a little more dangerous because, at least in our case, the unexpected infinite sequence crashes the browser.

For future reference, as you can see by the highlighted #C6N245JGG link, there's a channel for it so it's best to ask such questions there. But no need to move this message that's already been posted.

Sorry, I haven't used Slack for long time and thought hashtag was just to mark a topic, but it's a channel reference. Thanks for reminding!

:node-test doesn't support a REPL since the entire point is running the tests and exiting with a non-zero exit code on failing tests

you can create a :node-script build or just use the built-in shadow-cljs node-repl

then just require and run whatever from the REPL

Great. Thanks! I'll try both! Shadow-cljs is awesome! If you have a way to donate Bitcoin via Lightning, I'd be happy to support the project

LN looks interesting but I haven't done anything with it yet, so only traditional github sponsors and so on for now 😉

If you happen to be on Nostr, I could send you Zaps (SATs via lightning https://nostr.how/en/zaps) (non-custodial). Or https://tippin.me/ is easy too with Twitter (custodial)

nostr is another interesting thing I didn't have time to check out yet 😛

too much cool stuff out there

Yeah. I'm building some MVP-s of my ideas related to Bitcoin & Lightning using shadow-cljs. Loving it!

Try properly formatting your code - you'll see why.

Hint - there's an extra couple of ().

Oh, wait, never mind. While it would also fail, it would result in a different error - an attempt to call null.

Are you sure that error comes from that let and not some other? If yes, then there should be more details to that error that show exactly where the spec has failed.

The following gives an error at the let part but is able to println % properly:

(-> (js/logseq.App.getAllPages)
    (.then #(
             do(
                (println %)
                (let [pages %]
                  (println pages))))))

That should be a different error then - about calling null since you have an extra () after do - you're trying to call the result of (let ...). The macroexpand error happens during compilation and is probably some earlier error that has nothing to do with the above.

Oh, you're right, the error is not related... it was a previous error

I get this error in the app console:

VM421:2 Uncaught (in promise) TypeError: Cannot read properties of null (reading 'call')
    at eval (eval at shadow$cljs$devtools$client$browser$global_eval

Try #(let [pages %] ...) - you don't need do. And with do it would be

#(do
   (let [...] ...))

So the println called within let is in error

(println ...) always returns nil. (let [...] (println ...)) also returns nil then. (do ...) evaluates its body and returns the result of the last form. (do (...)) (that you have) calls ... as a function and returns its result. In your case, that ... is (let ...), so you're trying to call nil as a function. Hence why I keep talking about extra () which turn a plain value into a function call.

Oh I see! Took me long to see it , thanks a lot!

js doesn’t have threads or blocking. So no way for a future to run concurrently and no way for a promise to receive a value and execution to continue

so in

(let [p (promise)]
  (future (Thread/sleep 2000) (deliver p :hi))
  @p)