PS C:\Users\mmer> clj
Clojure 1.11.1
user=> (System/getenv "HOME")
nil
user=> (System/getenv "USERPROFILE")
"C:\\Users\\mmer"
user=> (System/getenv "userprofile")
"C:\\Users\\mmer"
user=>

are env variables case insensitive on Windows?

the clj PowerScript script checks for HOME in Windows first, but AFAIK this is not a Windows convention at all. Do you know the reason for this, @U064X3EF3?

Thanks @U4C3ZU6KX

Not a windows convention but its a common user convention to set it, tools like Emacs require it iirc

Cygwin might set it too

ok

https://github.com/JeremS/prose

a quick google search turned up https://github.com/JeremS/prose! It seems to mention sci. But from what I can tell, JeremS/prose is not an attempt to create a fast-starting bb script that one can use to preprocess.

@U04V15CAJ you got it first 😄

I didn't have to search, because I knew it existed ;)

don't know if the code is bb-compatible but perhaps can be made so

from the readme it appears to already support sci. So perhaps an argument for supporting sci from Babashka scripts (which I believe has been discussed before). > don’t know if the code is bb-compatible but perhaps can be made so Using the eval evaluation strategy could potentially work from bb! I’ve gotta try this.

sci is already available in bb scripts

Not quite pollen syntax, but my https://github.com/fabricate-site/fabricate also features the ability to embed arbitrary Clojure expressions. It uses a somewhat different evaluation model than prose. Not (yet) bb-compatible though!

I definitely need to put more work into thinking about an extensible markdown-like syntax that: • Allows embedding Clojure expressions • Can be formally specified using EBNF I think there's potential for something that can be specified using a tree-sitter grammar, which would allow for support across tools (fabricate's notation currently relies heavily on an emacs package to work well in an editor)

Right -- providing a good editor experience is a challenge here because we're no longer editing pure markdown, or pure clojure.

I'll add it to bb, but perhaps they can work around it using this for now:

#?(:clj 
  (defmacro if-exists [x y]
    (if (resolve x)
      x y)))

they were trying to catch it, which makes that harder

ok, do the same but use if-exists in catch instead

once the exception type becomes available in bb it will automatically use the more specific type

and it will work with older bb's as well

added on master

hot damn, thank you!

hmm, this doesn't work

(throw new (if-exists Foo Exception) "dude")

then just go for exception then in bb

and make a note that from 1.3.186 on the other type will work, and in a year time we can probably shift to that one

so put a date in the comment ;)

or we can just shift to the new type once 136 comes out, people can still use the old malli version

👋 I tried this:

#?(:clj
   (defmacro if-exists [x y]
    (if (resolve x)
      x y))
   (defn -string->uri [x]
     (if (string? x)
       (try
         (URI. x)
         (catch (if-exists URISyntaxException Exception) _
           x))
       x)))

So maybe sticking with a plain Exception for now, and then revisit once 136 comes out?

i think that's smart

@U03N9E40Q2F yeah, let's do that.

I'll put a comment that it can be tightened up.

(fs/list-dir "/tmp/" "**/*" is also a no-go

you probably want glob, list-dir is not recursive

(fs/glob "." "**")

the docstring made me think to use the “glob” style argument to fs/list-dir rather than the accept function style

i didn’t realize it meant fs/glob

babashka.fs/list-dir
([dir] [dir glob-or-accept])
  Returns all paths in dir as vector. For descending into subdirectories use `glob.`
     - `glob-or-accept` - a glob string such as "*.edn" or a (fn accept [^java.nio.file.Path p]) -> truthy

it doesn't mean fs/glob

but list-dir isn't recursive, glob is

list-dir does accept a glob pattern, like "*.clj"

i get it now. but the reading made me think that (and i wasn’t aware of fs/glob)

ah gotcha

but fs/glob returns an empty vector for me, for anything other that "."

e.g. (fs/glob "/tmp" "*") -> []

(fs/glob "/tmp" "**") -> []

@U0250GGJGAE You're hitting exactly a bug which has been fixed yesterday, namely when you query a root directory

oh lol

another isue may be that /tmp is a symlink and fs/glob doesn't follow symlinks (yet)

(fs/glob (fs/canonicalize "/tmp") "**")

Gotcha - yeah the canonicalize makes it work for me too. Thank you sir 👍

ah good

did you really want to query /tmp or did you try it as an example?

Was trying it as an example

I see now, that regular glob does work for non top-level

actually I need canonicalize for my use-case anyhow, so thanks for pointing that one out

can you explain why you need it?

Well I don’t need it. I just have some things mapped with soft-links and I prefer to refer to them that way.

(which then works fine if I canonicalize the target directory, which is a soft-link to somewhere else)

cool