Where did the name "moon" come from?
The first incarnation of the name was a misspelling of monorepo (= moonrepo). This is where the domain moonrepo.dev came from, and our official company, moonrepo LLC.
However, moonrepo is quite a long name with many syllables, and as someone who prefers short 1 syllable words, moon was perfect. The word moon also has great symmetry, as you can see in our logo!
But that's not all... moon is also an acronym. It originally stood for monorepo, organization, orchestration, and notification tool. But since moon can also be used for polyrepos, we replaced monorepo with management (as shown on the homepage). This is a great acronym, as it embraces what moon is trying to solve:
- Manage repos, projects, and tasks with ease.
- Organize projects and the repo to scale.
- Orchestrate tasks as efficiently as possible.
- Notify developers and systems about important events.
With that being said, for languages not supported in our toolchain, you can still execute them within at task by setting the type to system.
Will moon support continuous deployment?
Yes! We plan to integrate CD with the current build and CI system, but we are focusing on the latter 2 for the time being. Why not start using moon today so that you can easily adopt CD when it's ready?
What should be considered the "source of truth"?
If you're a frontend developer, you'll assume that a
package.json is the source of truth for a
project, as it defines scripts, dependencies, and repo-local relations. While true, this breaks down
with additional tooling, like TypeScript project references, as now you must maintain
tsconfig.json as well as
package.json. The risk of these falling out of sync is high.
This problem is further exacerbated by more tooling, or additional programming languages. What if
your frontend project is dependent on a backend project? This isn't easily modeled in
package.json. What if the backend project needs to be built and ran before running the frontend
project? Again, while not impossible, it's quite cumbersome to model in
package.json scripts. So
on and so forth.
moon aims to solve this with a different approach, by standardizing all projects in the workspace on
moon.yml. With this, the
moon.yml is the source of truth for each project,
and provides us with the following:
- The configuration is language agnostic. All projects are configured in a similar manner.
- Tasks can reference other tasks easily. For example, npm scripts referencing rake tasks, and vice verse, is a non-ideal experience.
- Dependencies defined with
dependsOnuse moon project names, and not language specific semantics. This field also easily populates the dependency/project graphs.
By using moon as the source of truth, we can ensure a healthy repository, by accurately keeping everything in sync, and modifying project/language configuration to operate effectively.
With all that being said, moon supports implicit dependency scanning, if you'd prefer to continue utilizing language specific functionality, instead of migrating entirely to moon.
When we run a task, we generate a dependency graph of nodes, known as actions. These labels are the actions in the graph, and break down as follows:
SetupNodeToolchain- Sets up the Node.js toolchain. Required for all other actions, so acts like a root node.
SetupSystemToolchain- Currently a no operation.
InstallNodeDeps- Installs Node.js dependencies (
node_modules) in the workspace root.
InstallSystemDeps- Currently a no operation.
SyncNodeProject- When a project's
SyncSystemProject- Currently a no operation. Is created when a project's
languagedoes not require syncing.
RunTarget- The common action. This runs the target task within a project.
Projects & tasks
How to pipe or redirect tasks?
|) or redirecting (
>) the output of a task (maybe to another task), whether via stdin or
inputs, is currently not possible. However, you could execute a custom bash script that
handles this for you, for example:
moon project foo --json | jq ...
command: 'bash ./scripts/pipe.sh'
How to run multiple commands within a task?
A task in moon is a single command, representing its inputs and outputs, and as such, cannot run
multiple commands with
&& within a single task declaration. However, we do suggest the following
- Create a task for each command, and link them as dependencies with
deps. This allows for each command to be incrementally cached (when enabled).
- Create a task for each command, and run them concurrently with
- Wrap all commands in an external file, and execute that file instead. Refer to the piping question above for an example.
How to run tasks in a shell?
By default, all tasks do not run in a shell, as we interact with and execute a tool's binary directly within the toolchain. We also provide no built-in support for shells, but that doesn't mean you can't use them. Since tasks can execute any commands available on your system, you can implement the shell manually like so:
command: 'bash -c some-command'
command: 'cmd.exe /d /s /c some-command'
command: 'pwsh.exe -c some-command'
Can we run other languages?
languages within tasks by setting their
type to "system". System tasks
are an escape hatch that will use any command available on the current machine.
command: 'phpunit tests'
However, because these languages are not supported directly within our toolchain, they will not receive the benefits of the toolchain. Some of which are:
- Automatic installation of the language. System tasks expect the command to already exist in the environment, which requires the user to manually install.
- Consistent language version across all machines.
- Built-in cpu and heap profiling (language specific).
- Automatic dependency installs when the lockfile changes.
- And many more.
Can we use
We encourage everyone to define tasks in a
moon.yml file, as it allows
for additional metadata like
options, and more. However, if you'd like to
package.json scripts, enable the
View the official documentation for more information on this approach, including risks, disadvantages, and caveats.
Can moon version/publish packages?
At this time, no, as we're focusing on the build and test aspect of development. With that being said, this is something we'd like to support first-class in the future, but until then, we suggest the following popular tools: