x/rambda@v7.0.1/CONTRIBUTING.md

Faster and smaller alternative to Ramda
GitHub Stars
1342
Latest

Contribution guidelines

You need to have the code of selfrefactor/rambda-scripts repo on the same level as selfrefactor/rambda as that repo contains all of the build logic.

git clone https://github.com/selfrefactor/rambda-scripts.git

Final step for any code change is yarn out as it generates the output files.

Fix a method

  1. If the error is in R.foo then you need to write a test in source/foo.spec.js, that reproduces the bug.

  2. Change source/foo.js so the tests are passing.

  3. Uncomment // await runSingleBenchmark('foo') line in scripts/all-scripts/all-scripts.spec.js.

  4. Run yarn out

  5. Restore the state of scripts/all-scripts/all-scripts.spec.js and commit

Fix a Typescript definition

  1. You may add a new test to source/foo-spec.ts to reproduce the bug.

  2. Apply your fix to files/index.d.ts.

  3. Run yarn typings to confirm your fix.

It is known that definitions of R.compose/R.pipe are far from perfect. The issue has been previously discussed but there is no obvious solution to it.

Add new Rambda method using helper

There is a helper script to assist you, when creating a new method in Rambda/Rambdax.

  1. Run yarn new NEW_METHOD_NAME, e.g. yarn new zip.with or yarn new zipWith.

  2. Edit source/NEW_METHOD_NAME.js

  3. Edit source/NEW_METHOD_NAME.spec.js

  4. Edit source/NEW_METHOD_NAME-spec.ts

Add new method manually

The new method should have exact or very close implementation compared to the corresponding Ramda method.

Create a method file in source folder

If the new method is R.foo, then the created file will be ./source/foo.js

Write the function declaration and function's logic

function foo(fn, input){
  return fn(input)
}

Write your test cases

Create file foo.spec.js in folder source

import { foo } from './foo'

test('happy', () => {
  expect(
    foo(console.log)(1)
  ).not.toThrow()
})

Run Jest

Running jest --watch is possible but as they are many tests, I am unsure that this is the best option.

I'd suggest to temporarily change the file name in package.json

  "scripts":{
    "dev": "jest source/foo.spec.js --watch"
  }

and then run yarn dev.

Match Ramda functionality as much as possible

It is important step as Rambda shouldn't have too many differences with Ramda.

In the hypothetical case with R.foo, you should check the content of current Ramda tests:

https://raw.githubusercontent.com/ramda/ramda/master/test/foo.js

Use these cases as a base for your tests, but feel free to write your own test logic.

Later, during generation of README.md, your new method would be tested against Ramda R.foo specification as assurance that the actual Ramda tests are passing with this new Rambda method.

It is very possible that there is some Ramda specifications, that our R.foo cannot easily satisfy. Then you can simply write new entry in scripts/run-ramda-specs/allDifferences.json, where you need to specify the number of failing Ramda tests and the reason for the failure.

Once the tests are ready, now you need to take care for Typescript definitions.

Typescript definitions and method description

Edit files/index.d.ts and add new entry. Methods in this file are sorted alphabetically, so in case of R.foo you need to add the entry between R.flip and R.forEach.

Write typings and specifications:

/*

Method: foo

Explanation: It returns the result of `fn(input)`.

Example:

const fn = x => x + 1 const input = 1

const result = R.foo(fn, input) // => 2


Categories: Function

Notes:

*/
// @SINGLE_MARKER
export function foo<T, U>(fn: (x: T) => U, input: T): U;

Note the consistency of fn and input throughout the example, the actual method(source/foo.js), the explanations and the TypeScript definition.

Currying

Any method, which takes more than one argument, should be curried.

We can use the standard currying used throughout Rambda.

export function foo(x, y){
  if(arguments.length === 1) return _y => foo(x, _y)

  return x(y)
}

If R.foo accept more than two arguments, then you should use R.curry.

import {curry} from './curry'

function fooFn(x,y){
  return x(y)
}
export const foo = curry(fooFn)