ramda/ramda

Practical functional Javascript

Alex Schenkman

@alesch

@CrossEye Would you help me understand your suggestion, please?

f = R.assoc('id')
g = R.zipWith( f );
R.converge(g, R.keys, R.values);

I understand that we are partially applying arguments.
assoc takes 3 arguments. The last one is the list itself.
But what happens with the second one?

Martin Algesten

@algesten

@alesch R.assoc('id') expects two more arguments 1. the value for id and 2. an object to shallow copy and put {id:arg1} into.

R.zipWith with a function that accepts two arguments (1. and 2. above) will then be expecting one array of 1. and one array of 2.

R.keys gives you the array of 1.s and R.values an array of 2.s

Scott Sauyet

@CrossEye

I guess I'm just a quiche-eater. I love pipe! I feel a little dirty when I have to resort to converge or useWith. I'm glad we have them, but I prefer nice pipelines of functions.

Martin Algesten

@algesten

hm. ok. if you say so. perhaps i try compose, to see whether backwards feels better ;)

Scott Sauyet

@CrossEye

Thank you @algesten. I struggle to explain these things succinctly.

Martin Algesten

@algesten

not sure i managed to. but i wanted to take a stab.

Aldwin Vlasblom

@Avaq

@alesch I also found CrossEyes answer pretty astounding, so I picked it apart in the REPL. Let's start with converge:

//converge(f, a, b)(x) === f(a(x), b(x))
//so:
R.converge(R.zipWith(R.assoc('id')), R.keys, R.values)(x) //===
R.zipWith(R.assoc('id'))(R.keys(x), R.values(x)) //===
R.zipWith(R.assoc('id'), R.keys(x), R.values(x))

Now R.keys and R.values both produce lists. One containing the ID's (keys) and the other containing objects (values). And as it happens, R.zipWith(f) wants two lists, and it'll pass a pair of items (taken from both lists) into f so that f may produce a single item for the returned list.

In this case we use R.assoc('id') as our f (zipper function), which will receive the ID from the first list, and the object from the second and return a new object on which the "id" property will have been set.

Scott Sauyet

@CrossEye

An important point here is the parallel between keys and values. They are ordered in parallel. The fourth entry in the list of values is the value of the property of the object at the fourth entry in the list of keys. Without that, this would not work.

Scott Sauyet

@CrossEye

A version of the same technique where that association is not so important would be R.compose(R.map(R.apply(R.assoc('id'))), R.toPairs). The tradeoff is having to call R.apply. But this might be a little cleaner.

Martin Algesten

@algesten

well done avoiding R.pipe there ;)

Scott Sauyet

@CrossEye

Just for you. I wrote it as pipe because that's how I think, and reversed it to compose :smile:

Martin Algesten

@algesten

haha! :D

just out of curiosity. does Object.keys guarantee an order?

or that multiple invocations will have the same order as a result.

Scott Sauyet

@CrossEye

Very big point of contention now. #1067.

Scott Sauyet

@CrossEye

It is not really consisyent with our notion of equals. Two things we think of as equal could have different order of keys. I'm proposing a change, but there's significant opposition.

Martin Algesten

@algesten

I'm reading. just now at the bit where jdalton appear to be wilfully obtuse regarding the natural order of a list.

Scott Sauyet

@CrossEye

He's not just trolling, even if it sometimes feels that way. He does have serious points to make. And decisions here are clearly not as obvious to everyone as I would have thought. Lots of smart people with lots of different opinions.

John-David Dalton

@jdalton

:hear_no_evil:

Martin Algesten

@algesten

:blush:

Scott Sauyet

@CrossEye

... and jdalton is one of them :angel:

Raine Virta

@raine

I think the arrow syntax is a bit too loose

let foo = (a) => ... // ok
let bar = () => ... // ok
let baz = a => ... // ok
let xyz = => ... // not ok

however in case of map(x => x*x) I can understand without parens it's more readable

John-David Dalton

@jdalton

Arrow funcs are nice

I donno why skinny arrow wasn't done (no lexical bind)

like CS. There was prolly a ton of discussion (as per usual)

Alex Schenkman

@alesch

@algesten @Avaq @CrossEye Thank you all for the explanations. I’ll have to study them more.

@CrossEye You said you used compose but you thought of it with pipe. How would it be with pipe? In case it becomes easier to read for me.

Scott Sauyet

@CrossEye

R.pipe(R.toPairs, R.map(R.apply(R.assoc('id')))) -- just reversing the parameter order.

Aldwin Vlasblom

@Avaq

@alesch pipe is just compose but reversed: pipe(a, b, c) === compose(c, b, a)

I think compose is meant to resemble nested-function calling, like so: compose(a, b, c)(x) looks like a(b(c(x)))

And pipe is meant to resemble piping, like so: pipe(a, b, c)(x) looks like x > a | b | c

Or that's how I think of them anyway. ;)

John-David Dalton

@jdalton

trying to come up with a better name for useWith and converge. It looks like useWith may be created by way of converge.

naming these things is hhaaaarrrdd grumble grumble

John-David Dalton

@jdalton

modArgs and distArgs woo

John-David Dalton

@jdalton

still not great

but betterish

Raine Virta

@raine

mapArgs

John-David Dalton

@jdalton

that makes me think of 1 function to map over args

Raine Virta

@raine

yeah that's true

John-David Dalton

@jdalton

it could be maybe

converge -> modArgs and useWith -> modPerArg

John-David Dalton

@jdalton

oh snao

so I created a createModArgs

that takes a resolver

the resolver accepts arg, index, arguments (like value, index, array)

it's used to then create a function that passes the transforms args based on the resolves of the resolver

so for useWith the resolver would be identity and for converge it would be function(value, index, args) { return args }

could be baseModArgs(func, transforms, resolver)

Jethro Larson

@jethrolarson

About half the time I use useWith I actually want the W combinator

Where communities thrive

People

Repo info

Activity