is parsing scala-cli .bsp directory out of scope for metals?

[ag91]

Hello,

I was trying to run metals on a sample script (.sc file) whose directory was set in this way:

scala-cli setup-ide . --dep com.lihaoyi::os-lib:0.9.0

And I discovered that metals couldn't find the dependency library,
although the .bsp directory created by scala-cli contained information about the dependency.

I gave a peek into the code and I realized metals only writes .bsp data.

The workaround was to inject the dependency via directive.
I was tempted to open an issue, but I thought that maybe somebody can share the rationale behind not using a .bsp directory if present.
Would this indeed be out of scope for metals?

[tgodzik]

Thanks for the question, Metals does use .bsp but the dependency is not contained there. Those are available for Metals in .scala-build directory. Trying out your examples doesn't cause any issues. os-lib is available within Metals and can run there.

My workflow was:

mkdir test
cd test
echo "@main def run = println(os.pwd)" > main-test.scala
scala-cli setup-ide . --dep com.lihaoyi::os-lib:0.9.0
code .

and then everything worked fine in Metals.

Although I would suggest using directives for dependencies since you wouldn't need to add the dependency in the terminal each time.

[ag91]

Thanks for the explanation! You are right, the example works fine. I got mislead by an incorrect setup of my editor settings (I use Emacs and my lsp-mode workspace was pointing to the parent directory instead of the directory with the .scala-build settings and so metals couldn't find the dependencies).

Great news!