Inside a directory with compose.yml (later - "compose dir") run docker compose up -d, this will run our 2 containers in detached mode. Now you have two containers named sbt and spark.
When you're done you can run docker compose down to remove the containers.
Note: If you want to clear SBT system cache (downloaded libs, etc.) you can run docker compose down -v, this will remove the containers and also remove volumes connected to SBT.

Build your source code into jar files:

# log into your sbt container, will start in /app
docker exec -it sbt /bin/bash

# clean build cache
sbt clean

# create Jar files without dependencies
sbt package

# create Uber Jar
sbt assembly

# you can chain these commands also
sbt clean package assembly

Your source code and target dirs are mounted to this container, so all of the output will be present on your system too.

You can run Spark two ways: interactive shell and spark-submit.

docker exec -it -u root spark /opt/spark/bin/spark-shell

Now you can enter command-by-command to evaluate your expression/program.

This method requires you to write your Scala source code using Spark API, compile and package it into .jar files.
Important note: If you use any of the external dependencies that are not Spark-related, then you need to assemble your code like so sbt assembly.

# use your name for --class "<YOUR_CLASS>" and your filepath
docker exec -it -u root spark /opt/spark/bin/spark-submit --class "App" --master "local[*]" /app/target/scala-2.12/app-assembly.jar

This will run one instance of Spark that will output the result of a program and exit.