Getting a random emoji/character from a unicode string

Question

My goal is to get a random emoticon, from a list, in F#.

I started with this:

let pickOne (icons: string) : char = icons.[Helpers.random.Next(icons.Length)]
let happySymbols = "🔥😂😊😁🙏😎💪😋😇🎉🙌🤘👍🤑🤩🤪🤠🥳😌🤤😍😀"
let sadSymbols   = "😭😔😒😩😢🤦🤷😱👎🤨😑😬🙄🤮😵🤯🧐😕😟😤😡🤬"

that doesn't work because:

"🔥😂😊😁🙏😎💪😋😇🎉🙌🤘👍🤑🤩🤪🤠🥳😌🤤😍😀".Length

is returning 44 as length returns the number of chars in a string, which is not working well with unicode characters. I can't just divide by 2 because I may add some single byte characters in the string at some point.

Indexing doesn't work either:

let a = "🔥😂😊😁🙏😎💪😋😇🎉🙌🤘👍🤑🤩🤪🤠🥳😌🤤😍😀"
a.[0]

will not return 🔥 but I get some unknown character symbol.

so, plan B was: let's make this an array instead of a string:

let a = [| '🔥'; '😂'; '😊'; '😁'; '🙏'; '😎'; '💪'; '😋'; '😇'; '🎉'; '🙌'; '🤘'; '👍'; '🤑'; '🤩'; '🤪'; '🤠'; '🥳'; '😌'; '🤤'; '😍'; '😀' |]

this is not compiling, I'm getting:

Parse error Unexpected quote symbol in binding. Expected '|]' or other token.

why is that?

anyhow, I can make a list of strings and get it to work, but I'm curious: is there a "proper" way to make the first one work and take a random unicode character from a unicode string?

Answer 1

Asti's answer works for your purpose, but I wasn't too happy about where we landed on this. I guess I got hung up in the word "proper" in the answer. After a lot of research in various places, I got curious about the method String.EnumerateRunes, which again lead me to the type Rune. The documentation for that type is particularly enlightening about proper string handling, and what's in a Unicode UTF-8 string in .NET. I also experimented in LINQPad, and got this.

let dump x = x.Dump()
let runes = "abcABCæøåÆØÅ😂😊😁₅茨茧茦茥".EnumerateRunes().ToArray()
runes.Length |> dump
// 20
runes |> Array.iter (fun rune -> dump (string rune))
// a b c A B C æ ø å Æ Ø Å 😂 😊 😁 ₅ 茨 茧 茦 茥
dump runes
// see screenshot
let smiley = runes.[13].ToString()
dump smiley
// 😊

Answer 2

All strings in .NET are 16-bit unicode strings. That's the definition of char:

Represents a character as a UTF-16 code unit.

All characters take up the minimum encoding size (2 bytes for UTF-16), up to as many bytes as required. Emojis don't fit in 2 bytes, so they align to 4 bytes, or 2 chars.

So what's the solution? align(4) all the things! (insert GCC joke here).

First we convert everything into UTF32:

let utf32 (source: string) =
    Encoding.Convert(Encoding.Unicode, Encoding.UTF32, Encoding.Unicode.GetBytes(source))

Then we can pick and choose any "character":

let pick (arr: byte[]) index = 
    Encoding.UTF32.GetString(arr, index * 4, 4)

Test:

let happySymbols = "🔥😂😊😁🙏😎💪😋😇🎉🙌🤘👍🤑🤩🤪🤠🥳😌🤤😍😀YTHO"

pick (utf32 happySymbols) 0;;
val it : string = "🔥"

> pick (utf32 happySymbols) 22;;
val it : string = "Y"

For the actual length, just div by 4.

let surpriseMe arr =
    let rnd = Random()
    pick arr (rnd.Next(0, arr.Length / 4))

Hmmm

> surpriseMe (utf32 happySymbols);;
val it : string = "😍"