Converts yarn.lock
files into nix expression.
- Make yarn and yarn2nix available in your shell.
cd $GIT_REPO nix-env -i yarn2nix -f . nix-env -i yarn -f .
- Go to your project dir
- If you have not generated a yarn.lock file before, run
yarn install
- Create a
yarn.nix
via:yarn2nix > yarn.nix
- Create a
default.nix
to build your application (see the example below)
Make sure to generate the lock file with yarn >= 1.10.1
For example, for the front-end
of weave's microservice reference application:
with (import <nixpkgs> {});
with (import /home/maarten/code/nixos/yarn2nix { inherit pkgs; });
rec {
weave-front-end = mkYarnPackage {
name = "weave-front-end";
src = ./.;
packageJSON = ./package.json;
yarnLock = ./yarn.lock;
# NOTE: this is optional and generated dynamically if omitted
yarnNix = ./yarn.nix;
};
}
note: you must modify /home/maarten/code/nixos/yarn2nix
To make this work nicely, I exposed the express server in server.js
as a binary:
- Add a
bin
entry topackages.json
with the valueserver.js
- Add
#!/usr/bin/env node
at the top of the file chmod +x server.js
- Run
nix-build
In thefront-end
directory. Copy the result path. - Create an isolated environment
cd /tmp; nix-shell --pure -p bash
. /nix/store/some-path-to-frontend/bin/weave-demo-frontend
./update-yarn-nix.sh
./tests/no-import-from-derivation/update-yarn-nix.sh
./run-tests.sh
yarn2nix
runs on the beginning of build to ensure that yarn.lock
is correct (and to collect nix-expressions if you haven't provided them). So, if there is something wrong, yarn2nix tries to change the lockfile, but fails and aborts, as it is run with --no-patch
flag (as we can't let changes inside of nix-build
).
It can sometimes give false-positives, so the solution at the moment is running yarn2nix --no-nix
to patch the lockfile in place (the flag is just to avoid config spam, unless you're up to write yarn.nix
).
yarn2nix
is released under the terms of the GPL-3.0 license.