I love using npm script as my build tool. By doing so, we could make sure the scripts are not sooooo complicated and easy to read (comparing to gulp/grunt in my opinion).
But as the project grows larger, there are more and more scripts that are added to package.json
and things become less and less maintainable.
You probably tried adding empty lines in between scripts, but figure npm install
will just remove your empty lines and leave you with only disappointment.
Here is a little trick which allow us to, kind of, group our scripts into categories.
{
"scripts": {
"test": "nyc mocha --recursive",
"preversion": "git checkout master && git merge --squash dev && npm test",
"version": "git add -A",
"postversion": "git push && git push --tags && git checkout -",
"\n# TESTING SCRIPTS:": "",
"createTestPages": "node ./scripts/createTestPages.js",
"test:watch": "mocha --recursive --watch",
"coverage": "nyc report --reporter=lcov > coverage.lcov && codecov",
"lint:js": "eslint .",
"\n# HUSKY GIT HOOKS:": "",
"precommit": "lint-staged"
}
}
The trick here is to define some category keys. By adding \n
to the front of the key, we would get a pretty darn neat npm run
result.
> npm run
Lifecycle scripts included in grab-lyrics:
test
nyc mocha --recursive
preversion
git checkout master && git merge --squash dev && npm test
version
git add -A
postversion
git push && git push --tags && git checkout -
available via `npm run-script`:
# TESTING SCRIPTS:
createTestPages
node ./scripts/createTestPages.js
test:watch
mocha --recursive --watch
coverage
nyc report --reporter=lcov > coverage.lcov && codecov
lint:js
eslint .
# HUSKY GIT HOOKS:
precommit
lint-staged
One thing to keep in mind is to always put NPM lifecycle hooks at the top of the script because NPM will just grab them and put them at the top.
What do you think about this trick?
Top comments (5)
I see the reasoning behind your approach, but – at least for a production-ready project – it feels a bit hack-y to me!
An alternative organizational method might be to move your NPM script definitions in to their own dedicated shell files within a
build/
directory. This way, you can organize your script files without having to pepper yourpackage.json
with unnecessary entries. Though it won't give you the prettynpm run
section headers you demonstrated (obviously), the directory structure gives you a similar kind of at-a-glance legibility. And using dedicate shell scripts will definitely simplify long-term package maintenance...Anyway, just my two-cents; thanks for thinking!
I definitely agree with you. A
build/
directory is definitely better for production projects.But, I think this is a hack that wouldn't really hurt to add. It is not at all contradictory to having
build/
. And the good thing about it is the prettynpm run
that will give a nicer guidelines to, maybe new developers who wants to find out the building instructions available.true enough. especially re: those “new developers who want to find... instructions”, a file system abstraction is probably suboptimal!
This is a cool trick :)
Makes me wish Node/NPM had used a file format that supports comments, though - Rust uses TOML, and it's hard to go back to JSON for config files after using that :(
Yes! Although it isn't really a solution, it does the trick if we want a neater
npm run
output.I am happy you find this trick useful.