TODO: this needs editing and adding to the website



## Per-project setup

Currently haxelib has two ways to have project local setups.

1. Using `haxelib newrepo`
2. Using `haxelib install all`

### Using haxelib newrepo

When using `haxelib newrepo` you can have a project-local haxelib repository. This feature is quite new and a little rough around the edges.

Caveats:

- libraries get downloaded for each project
- if you mistakenly run a haxelib command in a subdirectory of your project, it will be executed on the global repo ([to be fixed](https://github.com/HaxeFoundation/haxelib/issues/292))

### Using haxelib install all

Haxe allows you to define specific versions of the libraries you want to use with `-lib <libname>:<version>`. If you make sure to use this in all your hxmls, then `haxelib install all --always` (the `--always` avoiding you being prompted for confirmation) will be able to ensure the libraries your project needs are available in the necessary versions. If in fact you run this in a checkout hook, your get to track your dependencies in your git repo (some other VCSs should allow for a similar setup), allowing you to have a well defined and replicable setup for any state (commit/branch/etc.).

Disadvantages:

- the approach requires you to define all dependencies with specific versions and then running `haxelib install all` to grab them
- with this approach, any other project that does not have specific versions defined may be affected, as under some circumstances `haxelib install all` may set the global "current" version of the libraries (to be fixed)

Advantages:

- as pointed out above, this approach allows defining a *versionable* and *replicable* state.
- you don't have to download libraries for each project, which does make a difference for heavy weights like openfl and hxcpp

#### Sidestepping haxelib git issues

Because you cannot specify git versions with `-lib` paremeters, we suggest using git submodules instead, as again they provide an adequate way of definining a *versionable* and *replicable* state.

### Combining both approaches

You can of course combine both approaches, giving you the isolation provided by the first one, and the replicability provided by the second one.

### Future solutions

A solution that combines the strengths of both approaches is in the making. Stay tuned.