Flow + tap = flow-tap, my first npm module

July 31, 2017 - -

Although I use Javascript a lot these days, I miss having strong types. Linters like ESLint ameliorate the problem by catching tons of bugs, but it doesn’t get everything. Lately, Flow and Typescript have been garnering lots of attention for adding types to Javascript. I don’t know which one is better; they are syntactically similar and both seem pretty great. There are plenty of articles comparing them. I’ve been using Flow because I already had a largish Javascript project, and I wanted to gradually introduce static types in over time. If I was starting from scratch, I would have strongly considered Typescript since it’s actually a language, but I digress.

I use tap for my unit tests. It doesn’t seem to be quite as popular as MochaJS, but I appreciate the simplicity. As I started using Flow, my tests would complain about all the foreign characters from my type annotations. I found babel-tap which worked great, but it slowed my tests down significantly. I wanted to use something like flow-remove-types, but there was no handy integration with tap. So, I built and published my first npm module: flow-tap!

It’s inspired by babel-tap and extremely simple and fast. The primary function is that it will remove all your flow type annotations and allow tap to Just Work™. After installation, replace tap with flow-tap in your package.json file like so:

"scripts": {
	"test": "flow-tap test/**/*.js"

That’s it! Run npm run test and you’re good to go.


Now that I’m using flow annotations more, I’m actually able to eliminate a bunch of tests! I would always have some tests that passed null, undefined, or some other inappropriate argument (like a string where a number is expected, etc) to my test target. But now that Flow catches any unexpected types being passed to a method, I don’t need tests like these. Strong typing eliminates the need for a lot of tests. The same sort of thing happens in Java or C# or any other strongly typed language; the compiler protects you.


The module is simple. It’s essentially a few lines in one file named cli.js

#!/usr/bin/env node
'use strict';
process.execPath = require.resolve('.bin/flow-node');

Overwriting process.execPath is probably a bit of a hack, but it works well. If there’s a more kosher way of doing this, please let me know. Also, require.resolve is a neat way to get the path of a module.

The package.json file has a handy “bin” property for executable files which was perfect for my cli.js

Choosing a license.

Before publishing, I wanted to choose an open source license. I really don’t know a lot about this, but choosealicense.com was really handy. I went with a common MIT license.

Publishing to npm

Publishing to npm is incredibly easy. At first I was hesitant to publish because I wanted a staging area to make sure everything worked, and I didn’t want to increase the version number a bunch of times, etc. Well, it worked as I had planned, and I had to keep increasing the version anyways as I made documentation improvements. Admittedly, I do find it a tad annoying that I have to increase the version number when I update the readme, but I’ll live. Fortunately, npm version patch handily increases the version number for me.

On the occasions I’ve thought about publishing an npm module, I usually discover something similar that already exists, or bureaucratic red tape dissuades me from open sourcing, so I’m happy I had the opportunity to publish this. Hopefully it won’t be the last one! Oh, and let me know if flow-tap comes in handy for you.

EDIT: It turns out this may have been for naught, as @izs pointed out you can do roughly the same thing with node-tap’s node-arg flag, which I was unaware of. Oh well! Still a good learning experience nonetheless.