Simple and flexible tool to build clojure projects with Nix.
The example/ directory has a small clojure program and the nix code required to build it.
To generate/update the lockfile:
nix-shell --run clojure-nix-lockerTo build:
nix-build -A uberjarYou can generate a flake example with:
mkdir play-with-clojure-nix-locker && cd play-with-clojure-nix-locker && nix flake init -t github:bevuta/clojure-nix-lockerThe example README has some next steps.
There are two existing projects with a similar goal already, clj2nix and clj-nix. Both of these are designed to be used roughly like this:
- At lock-time, call into
clojure.tools.depsto resolve all dependencies, then generate a lockfile from this. - At nix-eval-time, use the information from the lockfile to compute the classpath.
- At build-time, invoke clojure and pass it the precomputed classpath.
By contrast, clojure-nix-locker is designed around letting classpath computation happen later, at build-time.
It works roughly like this:
- At lock-time, call arbitrary user-provided commands (like
clojure -P) to pre-populate the caches in.m2and.gitlibs, then crawl those to generate the lockfile. - At nix-eval-time, use the information from the lockfile to recreate these caches in a way that's "close enough" to the real thing.
- At build-time, invoke clojure as normal. If the prefetching was done correctly, it will resolve its dependencies just fine without hitting the network.
This approach results in a pretty simple implementation and loose coupling to the clojure tooling.
As a consequence, things like aliases "just work" without requiring clojure-nix-locker to know about them.
Of course, this has its downsides too:
- If the directory layout of these caches changes, this tool breaks.
- Whatever classpath(s) your clojure tools compute at build-time will only work for the duration of that build.
Distributed under the GNU General Public License, Version 3. See LICENSE for more details.