DEV Community

We Don't Need No Stinking map() or filter()

Jason Steinhauser on September 22, 2018

Pretty clickbait-y title, I know, but hear me out! I posed a shower thinking question to some of our junior developers. If you had to choose one of...

Read full post

JavaScript Joel • Oct 17 '18

Great article!

And I could show you a real use case for this. I have created MojiScript and implemented map, filter and reduce. Both map and filter internally call reduce. Here's some examples:

const map = func => iterable => reduceWhile (null) (asyncMapReducer (func)) ([]) (iterable)

github.com/joelnet/MojiScript/blob...

const filter = predicate => iterable => reduceWhile (null) (asyncFilterReducer (predicate)) ([]) (iterable)

github.com/joelnet/MojiScript/blob...

Well technically they (and also reduce) call reduceWhile.

Jason Steinhauser • Oct 18 '18

Glad you enjoyed it! And it looks like I'm going to have to invest some time in checking out MojiScript. It looks like it could have some legs!

JavaScript Joel • Oct 18 '18

lol I would agree with you on MojiScript, but I am obviously biased ;)

Augie De Blieck Jr. • Sep 24 '18

I learned this one time when digging into the source code for the Elixir language. The Enum module there has map(), filter(), and reduce() -- but both map and filter are defined as reduce functions.

You can see it over here:

github.com/elixir-lang/elixir/blob...

I wouldn't be surprised if every other language does the same, now that I think about it.

So, yeah, write the code that reads the best for the sake of future coders, and the language will take care of simplifying the rest for you. ;-)

Brad Gilbert • Sep 25 '18

Perl 6 doesn't do it this way, because it can't do that easily.

If you tried to implement map with reduce it would hang before producing a single value when called on an infinite sequence.

my @a = (0..Inf).map( *² ); # infinite sequence of all squared values

.say() for @a.head(20); # prints first twenty squares

# sequence of only twenty squared values
my @b = (0..^20).reduce: { @_.head(*-1).Slip, @_.tail² }

.say() for @b; # prints the same twenty squares

# doesn't stop until it is killed or runs out of resources
my @c = (0..Inf).reduce: { @_.head(*-1).Slip, @_.tail² }

.say() for @c.head(20); # never gets to this line of code

You could maybe find a way to pass upwards the number of values to produce:

my @d = (0..Inf).reduce: {
  last if @_ > 20; # stop early

  @_.head(*-1).Slip, @_.tail²
}

.say() for @d;

The thing is that iteration is done using an object based protocol, so map and grep just return an object with the appropriate functionality.
(Most of the time this is abstracted away so you don't need to know how it works.)

Perl 6 is a functional language while also being a procedural language.
So it can't always make the same sort of guesses about what you mean that pure functional languages can make.

Jason Steinhauser • Sep 24 '18

I'm just now digging into Elixir/OTP, and I'm actually kind of amused to see that 😀 thanks for pointing that out!

Kushan Joshi • Sep 23 '18

While reduce can pretty much do anything, I would recommend that you also mention that not to use the same hammer for every kind of nail. We developers like to stay in our familiar territory and that’s why we avoid trying out new things, it would really suck if a fellow developer uses reduce to do everything, when they could have used other more optimised things.

Jason Steinhauser • Sep 23 '18

You are absolutely correct. I eluded to the fact that map and filter were more optimized than their reduce implementations, but I need to reiterate that at the end. Thanks for the feedback!

Raunak Ramakrishnan • Sep 23 '18

Nice article. I was recently trying out some coding problems using javascript. Array.reduce was very useful in solving them in a functional way. It is similar to fold in Haskell. This is a great article on the power of folds in Haskell.

Jason Steinhauser • Sep 23 '18

I hadn't read that particular paper yet, but now it's on my list 😊 Thanks for the recommendation!

Sergiu Mureşan • Sep 25 '18

I used filter, map and reduce almost every day at work and never thought of that! Great article!

Maybe use reduce instead when you both want to filter and map at the same time? Or maybe you want to do some other operations as well. reduce is like the swiss army knife of JS.

Alain Van Hout • Sep 27 '18

Before anyone would be inclinded to use reduce for everything, since it indeed can be used to do all the things that mapand filter(as well as someand any), keep in mind that the same could be said (and then some) about a traditional for loop.

As to using reduce when you want to combine different operations, again note that the same would be true for a forloop. Neither a for loop nor reduceshould be used in those cases, for the very same reason: you sacrifice legilibity (and the next developer that has to modify your code won't be thanking you for your cleverness).

Zohaib • Sep 23 '18

Glad to know that I have an idea in common with such an experienced developer. I had also written an article on LinkedIn regarding this approach.

Jason Steinhauser • Sep 23 '18

Nice! I had a few other functions I thought about implementing (takeWhile and every were on my list), but I opted not to in order to keep the article brief. I'm glad that you took the time to show those implementations in your article!

IsaacLeimgruber • Sep 23 '18

I remember doing the implementation of map and fikter if I remember correctly, it was in scala. Whatever, map and filter are good because they are cornerstones of the bridge between high order abstraction and intuition

Manel Escuer García • Oct 18 '18

Using reduce instead of map won't be returning a new array on each iteration? Therefore lesser performance?

Jason Steinhauser • Oct 18 '18

In JavaScript, there would be a slight performance hit if you used reduce to perform a map operation because of returning an array each time. However, immutable data structures in other languages (like Clojure's vector) would not really see a performance hit because of how they are stored. You've given me an idea on a follow up post to discuss immutability and lazy vs. eager evaluation. Thanks for the feedback!

Manel Escuer García • Oct 18 '18

Thank you!

Mihail Malo • Feb 26 '19

For the lulz you could write pushAndReturn as:

const pushAndReturn = (arr, item) => (arr.push(item), arr)