VSCode and Scala....how hard can that be?

This week I am setting up my new laptop - moving off Mac finally and back onto Linux - YAY.

So one of the things I’m getting working is Visual Code Studio and Scala. Sounds easy doesn’t it?

Step 2

Scalameta is the extension for Scala on vscode. Install it; but wait, that’s not all - there’s more.

Of course, you went for Step 1 which was “aaah, this is easy, just download the extension and you’re done. How hard is that?". It didn’t work, but at least you’ve completed Step 1. We’re starting at Step 2 because step 1 was a fail.

Have you got the correct version of Java installed? I installed Java from the Ubuntu repo and while it was 11.0.7-10, it seemed it was giving issues. Despite futsing around with various Metal options, it seemed to not be loading Bloop.

Bloop. To the layman (yup, that’s me), this is another thing that flummoxed me. It seems it’s job is to automagically download the libraries from Maven.

Step 3

Uninstall java from the Ubuntu repo’s and install it via SdkMan. That’s what the sbt recommends. Do it:

Install sdkman

curl -s “https://get.sdkman.io” | bash

Install Java

What I don’t understand is that this installs as your user, not system-wide; which I can see is going to bite me later!

sdk install java 11.0.7.hs-adpt (no, I don’t know what these (adpt and hs) things mean yet)

Test java: java -version

Add JAVA_HOME to your .bashrc

Remember to re-run your .bashrc (. ~/.bashrc)

Step 4

When you open your .scala or .sbt files, Metal should create two directories namely .bloop and .metals. If .bloop is not created, my experience was that this is a Java problem and bloop is not starting up.

Check by running:

ps -elf|grep bloop.Server|sed ‘s/-classpath [^ ]*//g’

Yeah, that classpath stuff just complicates life.

If bloop is not running, revisit the Java version. For this tut, I installed java 11.0.7.hs-adpt as above.

Final step

Now bloop is running, you’re still getting this error:

INFO [error] sbt.librarymanagement.ResolveException: Error downloading org.apache.spark:spark-graphx:2.4.4 INFO [error] Not found INFO [error] Not found

What I found is that copying the sbt path from Maven didn’t work. What Maven suggested I put in my build.sbt was this (snippet):

libraryDependencies += “org.apache.spark” % “spark-core” % “2.4.4”, libraryDependencies += “org.apache.spark” % “spark-sql” % “2.4.4"

Which is not a path because it’s being searched for here:

INFO [error] not found: /home/hamish/.ivy2/local/org.apache.spark/spark-sql/2.4.4/ivys/ivy.xml INFO [error] not found: https://repo1.maven.org/maven2/org/apache/spark/spark-sql/2.4.4/spark-sql-2.4.4.pom

Which, as you can see, is not found.

A bit of _curl_ing did the trick:

curl https://repo1.maven.org/maven2/org/apache/spark

showed me that the path was in fact not as I’ve put it in the dependencies, but rather

<a href="spark-core_2.10/" title="spark-core_2.10/">spark-core_2.10/</a>
<a href="spark-core_2.11/" title="spark-core_2.11/">spark-core_2.11/</a>
<a href="spark-core_2.12/" title="spark-core_2.12/">spark-core_2.12/</a>

So changing the path in the build.sbt to:

libraryDependencies += “org.apache.spark” % “spark-core_2.12” % “2.4.4”, libraryDependencies += “org.apache.spark” % “spark-sql_2.12” % “2.4.4"

fixed this.

Test by writing some Scala and autocomplete should work.

Time spent figuring this out: half-a-day + 2 hours