My new ebook β€œComprehending Comprehensions” is on pre-sale and 40% off!

A comparison of many possible Python implementations of the sign function.

I challenged you πŸ†...

You delivered πŸ’ͺ!

I asked you to implement the sign function in Python 🐍.

Now I'll go over some alternatives and tell you what I like βœ… and dislike ❌ about them.

I'll also tell you which one I think is the best, most Pythonic ✨ one.

πŸ‘‡

# sign: returns 1 for positive numbers,
# 0 for 0, and -1 for negative numbers.
>>> sign(73.73)
1
>>> sign(0)
0
>>> sign(-42)
-1

By the way, for reference, here is the original challenge:

Let's start with the β€œbasic” if: ... elif: ... else: ... approach first.

Gets the job done, is simple, and is easy to read.

Funnily enough, the only one to share that solution was a very seasoned Python 🐍 programmer, author, and trainer, @dabeaz πŸ™ƒ

Now, a couple of notes:

def sign(x):
    if x > 0:
        return 1
    elif x < 0:
        return -1
    else:
        return 0

For the arg name, we don't need something VERY long like number, but don't shorten it to n.

n has a connotation for positive whole numbers, and sign accepts other numbers.

I also prefer to have the else: return the 0, and that's to preserve the symmetry of 1 and -1.

We can use this as the cornerstone for many interesting implementations.

For such a simple function, it might look like a bummer that we have to spend 6 lines implementing it.

Can we shorten it a bit, perhaps?

Using conditional expressions, we can:

def sign(x):
    if x == 0:
        return 0
    else:
        return 1 if x > 0 else -1

We can also get rid of the else, which may reinforce the sense of β€œ0 is just an edge-case”...

Depending on how you usually read Python code.

Here it is:

def sign(x):
    if x == 0:
        return 0

    return 1 if x > 0 else -1

Now, you might be thinking...

Instead of returning the conditional expression...

Can I compute Abacus the 1 or -1?

Well, you can!

And many of you submitted things like what follows.

But that gives rise to another issue...

def sign(x):
    if x == 0:
        return 0

    return int(abs(x) // x)

Do we really need the int?

Well, depends on what you want your function to return, because abs(x) // x returns a float if x is a float.

And what about floating point inaccuracies? Does int(abs(x) // x) always return 1?

To be honest with you, I am not entirely sure!

I couldn't find a value of x for which abs(x) // x evaluated to 0.99999(something).

However, to be extra safe, you could replace int with round.

Ok, and are there β€œsafer” alternatives?

There are!

Have you heard about the Truthy and Falsy values of things in Python?

That makes it so that all objects can be interpreted as Booleans when needed.

But did you know that Booleans (True and False) can be handled as integers??

Hence, you can write this

What⁉

def sign(x):
    if x == 0:
        return 0

    return (x > 0) - (x < 0)

I like the fact that it looks like a face:

(x > 0) - (x < 0)

The - is the nose, and the (...) are the eyes!

And there's even the added benefit that you can drop the if statement:

def sign(x):
    return (x > 0) - (x < 0)

In my opinion, this is very cool!

BUT it's not the β€œway to go” in Python, generally.

This looks more like a thing you'd do in an array-oriented language, like APL.

So, we steered away from Pythonic solutions.

Can we get back on track?

Yes!

But let's not, for now πŸ˜‚

I just wanna show you another interesting one, making use of the integer value of Boolean values πŸ‘‡

It looks interesting, and it is useful to understand how it works.

But I'm sure the majority won't feel this is Pythonic.

def sign(x):
    return -1 if x < 0 else int(bool(x))

I think the conditional expression wasn't that bad, right?

Maybe we could improve on it?

Well, we can try!

What if we nest two conditional expressions to handle the three cases?

Again, just a quick remark about this one:

def sign(x):
    return 1 if x > 0 else -1 if x < 0 else 0

I like to have 0 at the end, because I like the symmetry between

  • x > 0 -> 1
  • x < 0 -> -1

This might sound silly to you, but these symmetries and patterns really make my life easier!

That's why I prefer this ordering over, say, this one:

def sign(x):
    return 0 if x == 0 else 1 if x > 0 else -1

But we can still be friends if you order things your way 😁

Another thing to notice is that all the ifs with if x == 0 could be replaced by if not x.

Some might prefer it, some might not.

I have no strong feelings for neither πŸ€·β€β™‚οΈ

Is there any other way to emulate the β€œchoose one of -1, 0, or 1” behaviour without using long ifs..?

Someone submitted this, and I gotta say:

It looks odd, but it is incredibly easy to read.

Is it Pythonic? I don't think so 😒

def sign(x):
    return 0 if x == 0 else {True: 1, False: -1}[x > 0]

Instead of accessing a dictionary, we can also try to index into a list.

Now, close your eyes and skip this if you don't want to be horrified πŸ‘‡

It's not that bad πŸ˜‚

Again, not recommended style, but understanding how it works gives insights into how Python works.

def sign(x):
    return 0 if x == 0 else [-1, 1][x > 0]

This can be taken one step further.

Should you?

Nah.

Will I do it nonetheless?

Absolutely!

Here it goes πŸ‘‡ Can you understand how it works?

def sign(x):
    return [-1, 0, 1][(x >= 0) + (x > 0)]

Hmmm, 'kay.

That's all nice and fine.

What if I don't want to use ifs, but also no weird Boolean computations?

Then, maybe you could try the new pattern matching features from Python 3.10!

In case you need to learn it:

Can we use match to solve this?

Well, not directly.

Which shows that match really isn't the tool for this job.

But we can use guards and get away with it πŸ‘‡

But if you look closely, the case statements aren't doing anything!

So, can we get rid of them..?

def sign(x):
    match x:
        case x if x > 0: return 1
        case x if x < 0: return -1
        case _: return 0

Absolutely, and you end up with this πŸ‘‡

Equivalent to the if from the beginning, but condensed.

Canned if-blocks, everyone!

def sign(x):
    if x > 0: return 1
    elif x < 0: return -1
    else: return 0

Ok!

So, we used match and ended up getting away from it?

Unacceptable!

One more try:

match is supposed to be useful for structural pattern matching.

Can we match patterns, then?

Of course. Here's a solution no one submitted, but that I came up with πŸ‘‡

What do you make of it?

def sign(x):
    match x > 0, x < 0:
        case True, False: return 1
        case False, True: return -1
        case False, False: return 0

All in all, there are plenty of interesting solutions in this thread.

So, which ones do I prefer?

Well, personally, I love weird snippets of code πŸ˜‚ But I don't write them in production.

So, that means I end up with a split preference, between πŸ‘‡

and the slightly condensed if: ... elif: ... else: ... block πŸ‘‡

I don't know why, but it really bothers me that such a simple function takes 6 lines of code as a β€œfull” if: ... elif: ... else: ... block!

(Does anyone else feel like that?)

Thanks for making it so far!

I'd love if you read your thoughts on these solutions!

Here's a quick link to the beginning of the thread. Leave your comments there πŸ’¬ and retweet πŸ” it if you found value in this thread!

Finally, if you have ideas for future challenges, feel free to send them to me!

Write them down in the comments, or send me a DM!

Then, if you want to keep learning a lot about Python 🐍, follow me @mathsppblog.

I'll see you soon! πŸ‘‹

P.S. I should give credit to everyone who contributed to this thread by sending their solutions.

Many people replied, so I don't think I should mention all of them..? You can find all of them in the original challenge, linked at the beginning.

Thanks a lot for participating!

This article is an unedited reproduction of a piece of content I posted on Twitter here:

I hope you learned something new! If you did, consider following the footsteps of the readers who bought me a slice of pizza πŸ•. Your small contribution helps me produce this content for free and without spamming you with annoying ads.

Next Post

Blog Comments powered by Disqus.