Basically, what I expect is this and I might extend it from there 1. One CLJ file containing several independent functions 2. Possibly using some extra Java lib, if it loads fast enough 3. Likely using some clojure libs 4. One .sh file connecting them, e.g. alias doit='bb -i '(myutils/doit *input*)' doit2() { # shell function some shell code branch out to clojure }

Now, I looked at tasks but I'm not sure if I get it, or if it doesn't apply. It seems to suggest it should replace "Makefile" style operations. So it is for building/doing stuff with your project, rather than implementing small callable functions like above? Or is tasks the right way to access the individual functions in a single utils.clj ? It just hasn't clicked for me so far.

@U0359E1F02H You can make a global tasks file in e.g. ~/tasks/bb.edn and then call it with bb --config ~/tasks/bb.edn mytask . You can make a small bash wrapper for this, so you can call tasks mytask

Of course I want the bb invocation to be as simple as possible from the shell side, but whatever is required is fine.... Another idea is to have a main function, which takes an argument and calls out to the clojure function with that name and provides the *command-line-args and input* etc.?

OK... thanks let me try that with tasks, it didn't quite click before

You are right that tasks are mainly for project-specific tasks, but it can be used as I described above too

What you can also do:

bb -cp ~/tasks -m foo/bar

and write a small bash wrapper for that

OK, yeah, the second one feels more right to me. Considering this for example with tasks: https://gist.github.com/borkdude/35bc0a20bd4c112dec2c5645f67250e3 It feels to me like another indirection level. It feels more natural to me to do bb '(clojure function)' rather than bb 'task' and have to define each "task" as a simple redirect to the actual clojure function I want to call. and defining them all inside of the { } structure just doesn't feel nice to me.

The second one seems reasonable, let me try it.

... but ... hmm, how do I pass in input ?

bb -cp ~/tasks -m foo/bar

forget about *input* , just use (slurp *in*) in scripts.

Aha!

Let me try this a while and come back.

Namespace used in foo.clj must be exactly "foo" for this to work?

yes

Got it. Yes I think this got me over the hump. Thanks @U04V15CAJ. And thanks for BB, awesome stuff! (I already wrote many of the functions 🎉 so I'm glad I got this invocation stuff clarified)

I suspected as much! Thanks anyway, I’ll suffer without it for now 😊

cool beans! The performance aspect isn’t that interesting to me at least, I was just working on a script that transformed a nested yaml file and the code ended up looking pretty horrible (`(update .. (map (update .. (map …))))`) So I figured if specter was easily available I could at least write a simple path down to the thing I want to modify 😊

I discovered some things which need to be fixed first so maybe in a couple of weeks or so

Very cool, thanks! Really loving Babashka as a tool so far 🙏 My use-case is covered for now with my terrible update-algorithm so I’m not in a hurry 😊

https://github.com/babashka/babashka/issues/1203

$ ./bb -cp src/clj -e "(use '[com.rpl.specter]) (def data {:a [1 2 3]}) (setval [:a END] [4 5] data)"
{:a [1 2 3 4 5]}

I'm surprised it works :)

@odd.andreas specter is now available as a library to be used in babashka, see Changelogs :)

You need to use my fork though, which adds a compatibility commit for babashka. Hopefully #specter will accept it back as a PR

Awesome!

All specter tests are passing