Hi folks, can anyone recommend a good article on flatMap vs map? I'm finding that i'm almost always ...

# arrow

[tim]

Hi folks, can anyone recommend a good article on flatMap vs map? I'm finding that i'm almost always using map in my code and so suspect I'm missing something as I learn FP

[simon.vergauwen]

They have different signatures and thus different powers.

While map takes a function of the shape (A) -> B, flatMap takes a function of the shape (A) -> Kind<F, B>.

If we look at this for List that means.

listOf(1, 2, 3).map { it + 1 }
 // [2, 3, 4]

listOf(1, 2, 3).flatMap { listOf(it, it + 1) }
 // [1, 2, 2, 3, 3, 4, 4, 5]

With flatMap here we can return 2 items instead of 1 since we can return the result inside List.

The same is true for all data types.
For example with Option that means using `map`` I always return a value, but in case of flatMap I could also return Option.None.

[aballano]

also -> flatmap = map + flatten

[tim]

Ahh okay that makes sense ... so I use flatMap when the lambda returns a nesting ...

So, for example, if the lambda returned an Either I can use flat map to unpack the inner Ether so I don't end up with nested Eithers?

[simon.vergauwen]

Exactly!

[tim]

great thanks :)

[tim]

... and a few minutes later I needed flatMap 🙂

[stojan]

I recommend this talk on the subject: https://fsharpforfunandprofit.com/rop/

[tim]

thanks will add it to my list

[Attila Domokos]

You could also build your own Option and Either type. It's really simple, and looking at the implementation tells you how it works. This is the code for `Option`:

sealed class Option<A> {
    abstract val value: A

    abstract fun <B> map(f: (A) -> B): Option<B>
    abstract fun <B> flatMap(f: (A) -> Option<B>): Option<B>

    class None<A>: Option<A>() {
        override fun <B> map(f: (A) -> B): Option<B> = None()
        override fun <B> flatMap(f: (A) -> Option<B>): Option<B> = None()
        override fun equals(other: Any?): Boolean {
            if (this === other) return true
            if (javaClass != other?.javaClass) return false
            return true
        }

        override val value: A
            get() = TODO("Not yet implemented")
    }

    data class Some<A>(val valueInput: A): Option<A>() {
        override fun <B> map(f: (A) -> B): Option<B> = Some(f(valueInput))
        override fun <B> flatMap(f: (A) -> Option<B>): Option<B> = f(valueInput)

        override val value: A
            get() = valueInput
    }

    companion object {
        operator fun <A> invoke(a: A? = null): Option<A> =
                when(a) {
                    null -> None()
                    else -> Some(a)
                }
    }
}

[Attila Domokos]

Imagine you don't have flatMap and you tried to use map for a function with (A) -> Option<A>, you get Option<Option<A>>. You need to flatten that with flatMap to get Option<A> . Seeing a basic implementation will tell you a lot about its behavior.

[tim]

Thanks for this. Ya I need to spend more time in the source code. Option makes a lot of sense. I'll dig into either tomorrow.

[Attila Domokos]

Arrow's source code could be a lot more complex, I have not looked.
Anybody who wants to understand the concept of Arrow should read the book "The Joy of Kotlin", it's a fantastic intro book showing you how to build these types, understanding Lazy, what a traverse is.
I am using ideas from this book to teach core FP concepts, and once the team feels comfortable, I'll pull out Arrow, as a great library that does (and many other things) for you.

[tim]

I was about to ask if anyone had a book recommendation 😂

[Attila Domokos]

☝️

[Attila Domokos]

I am halfway through the book, but it's eye opening. Here are the examples I have completed so far. I've built my own Maybe and Either in Haskell (for learning purposes) and the fact that you can do it in Kotlin with ease tells you a lot about the language.

I wish more book examples would use TDD...

[tim]

It's it a "pure" FP book?

[tim]

One area that's unclear to me, is how mixing different paradigms is going to turn out!

[Attila Domokos]

I am not sure about "pureness", but it's the best book I've found so far describing FP principles.

[tim]

I have lots of code that follows a more OO approach, and currently I can't imagine doing away with OO completely

[tim]

From what you've read so far, does the book touch on that?

[Attila Domokos]

Whatever fits your team I think. The folks I am working with are scared of FP and the "complexity" of it. I am thinking of writing certain components (like API and dataaccess) in more object oriented style, but the "orchestration" logic is functional.

[Attila Domokos]

Tim, that's why included the examples from my GitHub account.

[tim]

Oh perfect I'm on my phone currently but will check it out tomorrow

[Attila Domokos]

Take a look at the "Table of Contents", look at the examples.

[Attila Domokos]

Nothing teaches you better than building a traverse function on an Either type yourself.

[Attila Domokos]

It's a demanding book, you must type in the examples in your editor and run them to understand them.

[tim]

Thanks, that's useful to know

[tim]

Okay just bought the book