Shell script to parse variable definitions from Scala build file

Question

The build.sbt text file contains versions like this:

name := "happy"
scalaVersion := "2.11.8"
sparkVersion := "2.2.0"

I wrote a Bash script to parse out the PROJECT_NAME and SCALA_VERSION from the build.sbt file:

PROJECT_NAME=$(cat build.sbt | grep "name :=" | cut -f 3 -d " " | tr -d '"')
SCALA_VERSION=$(cat build.sbt | grep "scalaVersion :=" | cut -f 3 -d " " | tr -d '"')

How can I write this more elegantly / robustly? I'm ok with an awk or sed approach, but don't want to add an external dependency to the script.

Here's some more of the script to see how the variables are being used.

if [ "$SCALA_VERSION" = "" ]
 then
 echo "SCALA_VERSION variable cannot be empty"
 exit 1
fi
SCALA_BINARY_VERSION=${SCALA_VERSION%.*}
if [ "$SCALA_BINARY_VERSION" = "" ]
 then
 echo "SCALA_BINARY_VERSION variable cannot be empty"
 exit 1
fi
echo "Create a GitHub release"
JAR_PATH=target/scala-${SCALA_BINARY_VERSION}/${PROJECT_NAME}_${SCALA_BINARY_VERSION}-${SPARK_VERSION}_${PROJECT_VERSION}.jar
hub release create -a $JAR_PATH -m "Release v${PROJECT_VERSION}" v${PROJECT_VERSION}

• 200_success: Could you also show more of the Bash script, where you use the variables?
• Powers: @200_success - Sure, I updated the post to show more of the script, thanks.

Answer 1

If you don't mind spinning up sbt to extract the settings keys you could use the following:

PROJECT_NAME="$(sbt name | tail -1)"
PROJECT_NAME="${PROJECT_NAME#* }"
SCALA_VERSION="$(sbt scalaVersion | tail -1)"
SCALA_VERSION="${SCALA_VERSION#* }"
echo $PROJECT_NAME
echo $SCALA_VERSION

• Powers: sbt -no-colors name is even better to remove ANSI color codes from the text stream

Answer 2

I'd use awk to generate shell syntax, and source it in the shell:

$ source <(
 awk '
 1ドル == "name" {print "PROJECT_NAME=" $NF}
 1ドル == "scalaVersion" {print "SCALA_VERSION=" $NF}
 ' build.sbt
)
$ echo $PROJECT_NAME,$SCALA_VERSION
happy,2.11.8

If the name or version strings might contain spaces, then we need to be more specific about awk's field separator:

$ cat build.sbt
name := "hello world"
scalaVersion := "2.11.8"
sparkVersion := "2.2.0"
$ awk '
 1ドル == "name" {print "PROJECT_NAME=" $NF}
 1ドル == "scalaVersion" {print "SCALA_VERSION=" $NF}
' build.sbt
PROJECT_NAME=world" # << oops
SCALA_VERSION="2.11.8"
$ awk -F '"' '
 1ドル ~ /^name/ {printf "PROJECT_NAME=\"%s\"\n", 2ドル}
 1ドル ~ /^scalaVersion/ {printf "SCALA_VERSION=\"%s\"\n", 2ドル}
' build.sbt
PROJECT_NAME="hello world"
SCALA_VERSION="2.11.8"

Answer 3

No one else has mentioned this, so I’ll add it here. There is no need to cat file | grep pattern. I might do it in an interactive shell if I forget which parameter is which, but for a script it is needless.

In

PROJECT_NAME=$(cat build.sbt | grep "name :=" | cut -f 3 -d " " | tr -d '"')
SCALA_VERSION=$(cat build.sbt | grep "scalaVersion :=" | cut -f 3 -d " " | tr -d '"')

prefer grep file pattern usage. Grep can read the file for you; no need for cat and a pipe.

• D. Ben Knoble: Or alternately <file grep
• Sᴀᴍ Onᴇᴌᴀ: Instead of adding a comment with more context, please edit and add the additional information. Refer to the section When shouldn't I comment? on Comment everywhere.