Randomness is a bit of a strange concept in Haskell.
As I discovered writing about Haskell and Randomness a few weeks ago, the best way to avoid doing everything inside the IO monad is to write functions that always return the random generator.
The crux of the matter is that random generators are long mathematical sequences. If you keep using the same index, you'll keep getting the same "random" value. You have to make sure the random generator progresses.
Increasing a number by a random value would look something like this:
import System.RandomaddRand::(RandomGen g) => g -> Int -> (g, Int)addRand gen x = let (a, gen') = randomR (0, 20) genin (gen', x+a)main = dogenerator <- newStdGenprint $ snd $ addRand generator 5-- prints 15
While this approach is great for changing a single value, it breaks down when you try to randomly change a whole list:
print $ map (snd . (addRand generator)) [1,1,1,1]-- prints [11,11,11,11]
Bummer! We applied a random function to every list member. And got a list of identical values!
The problem is that every time addRand gets called, it uses the same random generator. This means we're always using the same number in the random generator series.
We need a way to perform a fold and a map at the same time. A map, so the function is applied to every list item and we get a new list, and a fold where the generator is used as the accumulator.
We might be tempted to write something like this monstrosity:
addRand'::(RandomGen g) => g -> [Int] -> [(g, Int)]addRand' gen [x] = [addRand gen x]addRand' gen (x:xs) = let (gen', x') = addRand gen xin (gen', x'):(addRand' gen' xs)-- and then in mainprint $ map (snd) $ addRand' generator [1,1,1,1]-- prints [11,3,13,14]
This works. Every number in the list is different!
But the code is somewhat terrible and difficult to reason about.
Luckily, Haskell has got us covered and comes with a standard function called mapAccumL, which is a combination of a fold and a map:
print $ snd $ mapAccumL addRand generator [1,1,1,1]-- prints [11,3,13,14]
Cleaner, less work, more obvious what's going on. And it just happens to be perfect for writing evolutionary algorithms in Haskell. But more on that later, hopefully as early as Friday.
Here's the whole code:
import System.Randomimport Data.ListaddRand::(RandomGen g) => g -> Int -> (g, Int)addRand gen x = let (a, gen') = randomR (0, 20) genin (gen', x+a)addRand'::(RandomGen g) => g -> [Int] -> [(g, Int)]addRand' gen [x] = [addRand gen x]addRand' gen (x:xs) = let (gen', x') = addRand gen xin (gen', x'):(addRand' gen' xs)main = dogenerator <- newStdGenprint $ snd $ addRand generator 5-- prints 15print $ map (snd . (addRand generator)) [1,1,1,1]-- prints [11,11,11,11]print $ map (snd) $ addRand' generator [1,1,1,1]-- prints [11,3,13,14]print $ snd $ mapAccumL addRand generator [1,1,1,1]-- prints [11,3,13,14]
- Haskell and randomness
- What was your "ah ha" moment with Haskell?
- On Randomness
- Random Number Generators for Dummies - Poker Blog
Here's how it works 👇
Then get thoughtful letters 💌 on mindsets, tactics, and technical skills for your career.
"Man, love your simple writing! Yours is the only email I open from marketers and only blog that I give a fuck to read & scroll till the end. And wow always take away lessons with me. Inspiring! And very relatable. 👌"
Ready to Stop copy pasting D3 examples and create data visualizations of your own? Learn how to build scalable dataviz components your whole team can understand with React for Data Visualization
Curious about Serverless and the modern backend? Check out Serverless Handbook, modern backend for the frontend engineer.
Ready to learn how it all fits together and build a modern webapp from scratch? Learn how to launch a webapp and make your first 💰 on the side with ServerlessReact.Dev
By the way, just in case no one has told you it yet today: I love and appreciate you for who you are ❤️