Getting the base directory of an sbt project
This is a command you can run in a shell script to print the base directory of an sbt project:
$ sbt --batch -error "project $PROJECT" "print baseDirectory" | tr -d "\n"
$ sbt --batch -error "project ingests_api" "print baseDirectory" | tr -d "\n" /Users/alexwlchan/repos/storage-service/ingests/ingests_api⏎
We use this at work in some of our larger repositories, where we have lots of Scala apps in a single repo. The sbt base directory for each app/project includes some non-Scala files alongside the code, like a README and a Dockerfile. We have scripts that use this command to ask sbt where to find these non-Scala files, so they can be used by other build steps.
How it works
Normally I wouldn’t write a blog post for something this short, but the sbt documentation made this really hard to work out. Lots of stuff is either missing or confusingly explained, and I only got this working after I read several long GitHub issue threads.
Here’s how it works:
--batchflag tells sbt to “disable interactive mode” (according to
sbt -help). In this case, interactive mode means sbt asking for your input if something goes wrong:
Project loading failed: (r)etry, (q)uit, (l)ast, or (i)gnore?
Disabling interactive mode will cause sbt to skip asking for input, and fail immediately.
I can’t find this flag in the sbt docs, which describes an “interactive mode” which seems more akin to a console or REPL. I only found out about it from Stack Overflow posts, and to understand it I had to read the sbt source code, where it disables input by getting input from
-errorflag tells sbt to set the log level to errors-only. This skips the startup logs you get whenever you run sbt, like:
[info] welcome to sbt 1.4.1 (Homebrew Java 16.0.2) [info] loading global plugins from /Users/alexwlchan/.sbt/1.0/plugins [info] loading project definition from /Users/alexwlchan/repos/storage-service/project/project
The only place I can find this documented is in the sbt 1.0 changelog:
-debugare added as shorthand for
Seeing it alongside the related flags might help you realise this is something to do with log levels – but if you saw a script with
-errorin isolation, it’s less obvious. I also can’t find any explanation of the
earlycommand for which this is apparently a shorthand.
"project $PROJECT"selects a particular project from our multi-project build. We use multi-project builds to keep related projects in the same repository.
print baseDirectoryuses the print command to print the value of the
baseDirectorysetting. This includes a trailing newline, which you can see if you inspect the output in Python:
$ sbt --batch -error "project ingests_api" "print baseDirectory" > out.txt $ python3 -c 'output = open("out.txt", "rb").read(); print(repr(output))' b'/Users/alexwlchan/repos/storage-service/ingests/ingests_api\n'
I did try to find a flag for suppressing the newline, but I couldn’t find one if it exists.
tr -d "\n"uses the Unix tr command to delete newline characters. This gives me a string which is just the path, which I can then pass to other tools and scripts.