JarCastingAmmonite 
This is where the code for the Ammonite project lives:
- Ammonite-REPL, the improved Scala REPL
- Ammonite-Ops, the Scala file-system library
- Ammonite-Shell, the Bash-replacement system shell
If you use Ammonite and like it, you will probably enjoy the following book by the Author:
Hands-on Scala has uses Ammonite extensively throughout the book, using Ammonite to accomplish a lot of useful tasks: exploring third party libraries or writing parallel web-crawlers in the REPL, and implementing entire programming languages in Ammonite Scala Scripts. Hands-on Scala is a great way to level up your skills in Scala in general and Ammonite in particular.
If you want to learn more about Ammonite or how to use it, check out the links above, or ask on the Gitter Channel. The remainder of this document is developer-docs for people who want to work on the Ammonite source code itself.
If you are interested in living more "on the edge", we also publish artifacts and the doc-site every commit; the doc-site is available at
And contains instructions on how to download the per-commit executable or depend on the per-commit maven artifact.
If you use Ammonite and like it, please support us by donating to our Patreon:
Developer Docs
The layout of the repository is roughly:
ops/is Ammonite-Opsamm/is Ammonite's core, REPL and script runnershell/is Ammonite-Shellterminal/is the JLine re-implementation used by Ammonite-REPL to provide syntax highlighting and multiline editingreadme/is the source code for the Documentation, written in Scalatex.published/is a synthetic project used for publishing, excluding the readme and integration tests
For more detailed information, check out the internals documentation for high-level overviews of some interesting facets of the codebase
Common Commands
Manual Testing
Although most features should be unit tested, it's still useful to fire up a REPL from the current codebase to see things work (or not). There are a variety of shells you can spin up for testing different things:
-
mill -i -w terminal[2.12.6].test.runis useful for manual testing the terminal interaction; it basically contains a minimal echo-anything terminal, with multiline input based on the count of open- and closed-parentheses. This lets you test all terminal interactions without all the complexity of the Scala compiler, classloaders, etc. that comes inrepl/ -
mill -i -w amm[2.12.6].runbrings up the Ammonite-REPL using the source code in the repository, and automatically restarts it on-exit if you have made a change to the code. Useful for manual testing both ofamm/as well asops/, since you can justimport ammonite.ops._and start using them. Note that this does not bring in filesystem utilities like thewdvariable,cd!command. You can also pass in the path to a.scfile to run it using Ammonite's script runner -
mill -i -w shell[2.12.6].test.runbrings up a fully-loaded shell with all filesystem utilities included:wd,cd!, autocomplete for filesystem paths, and more. This usesreadme/resources/example-predef.scalainstead of your default predef, for easier experimentation and development. -
sbt -i -w integration[2.12.6].test.runruns the trivial main method in theintegrationsubproject, letting you manually test running Ammonite programmatically, whether throughrunordebug -
mill -i amm[2.12.6].assemblycreates an assembly atout/amm/2.12.6/assembly/dest/out.jarthat you can then use to test: start a REPL, run scripts, etc. in a standalone environment without being wrapped in Mill build tool -
mill -i amm[2.12.6].launchercreates a launcher script atout/amm/2.12.6/launcher/dest/runthat can also be used to run Ammonite outside of the Mill build tool.
Automated Testing
While working on an arbitrary xyz subproject, sbt ~xyz/test runs tests after every change. amm/test can be a bit slow because of the amount of code it compiles, so you may want to specify the test manually via amm/test-only -- ammonite.TestObject.path.to.test.
ops/testtests the filesystem operations, without any REPL presentamm/testtests the Ammonite-REPL/Script-runner, without filesystem-shell integration.terminal/testtests the readline re-implementation: keyboard navigation, shortcuts, editing, without any filesystem/scala-repl logicshell/testtests the integration between the standaloneops/andamm/projects: features likecd!/wd, path-completion, ops-related pretty-printing and toolsintegration/testkicks off the integration tests, which bundleamm/andshell/into their respective jars and invoke them as subprocesses. Somewhat slow, but exercises all the command-line-parsing stuff that the other unit tests do not exercise, and makes sure that everything works when run from.jars instead of loose class-files
Publishing
-
Publishing is automatic, controlled by scripts in the
ci/folder. -
Every commit that lands in master will publish a new unstable version, that you can already use and download. This includes publishing the unstable version to maven central to the snapshot-commit-uploads tag, and updating the documentation-site so it's Unstable Version download instructions to point to it, though the "main" download/install instructions in the doc-site will not be changed.
-
Every commit that lands in master with a tag will re-publish a stable version to maven central and upload a new versioned release (using the tag as the version) and the doc-site is updated so the main download/install instructions point at the new published stable version.
In general, if you land a change in master, once CI completes (1-2hrs) you should be able to download it via the Unstable Version instructions and make use of your changes standalone or in an SBT project.
Occasionally, the CI job building and publishing one of the above steps flakes and fails, and needs to be re-run through the travis web interface.
Issue Tags
I've started tagging open issues in the issue tracker to try and keep things neat. This is what the various tags mean:
Each issue should only have one of these:
bug: this behavior clearly wrong, and needs to be fixedenhancement: something relatively speccable, that can be worked on, finished, and will make Ammonite betterwishlist: could be totally awesome, but we're uncertain if it is worth doing at all, what it would look like, or if it will ever reach a "finished" state.
And possibly:
help wanted: I don't have context, hardware, or for some other reason am unlikely to ever do this. But I know people out there care, so one of you should step up and fix it.
Contribution Guidelines
- All code PRs should come with: a meaningful description, inline-comments for important things, unit tests (positive and negative), and a green build in CI
- Try to keep lines below 80 characters width, with a hard limit of 100 characters.
- PRs for features should generally come with something added to the Documentation, so people can discover that it exists
- Be prepared to discuss/argue-for your changes if you want them merged! You will probably need to refactor so your changes fit into the larger codebase
- If your code is hard to unit test, and you don't want to unit test it, that's ok. But be prepared to argue why that's the case!
- It's entirely possible your changes won't be merged, or will get ripped out later. This is also the case for my changes, as the Author!
- Even a rejected/reverted PR is valuable! It helps to explore the solution space, and know what works and what doesn't. For every line in the repo, at least three lines were tried, committed, and reverted/refactored, and more than 10 were tried without committing.
- Feel free to send Proof-Of-Concept PRs that you don't intend to get merged.