Taco Shop Name Generator

10/1/2014
A mug with a child's drawing on it

It took me almost four years of practicing Haskell at home to be able to write moderately complex programs in the language. I’m not sure if that amount of time is representative (at least I hope it isn’t) but it did feel painfully slow and endlessly confusing. Today, however, I’d like to take a look at one of the first programs I wrote in the language, a taco shop name generator.

But Why

San Diego has a plethora of taco shops representing a diverse range of burritos and hot sauces, but the creative names they use is something I enjoy. Most famous locally is Roberto’s but there’s also Fredberto’s and Philiberto’s and even Tacobertos. These don’t sound like real names, so one day I was wondering how they come up with all of these names and that’s when it hit me: they must use a program.

Writing Our Own Taco Shop Name Generator

My first idea for writing this program was to randomly pick a vowel or consonant, and then build up a word from these random choices. I started like this:

import System.Random
import Control.Monad.State.Lazy

all_let :: [Char]
all_let = ['a'..'z']

vowels :: [Char]
vowels = "aeiouy"

consonants :: [Char]
consonants = [x | x <- all_let, not $ x `elem` vowels]

Pulling Random Things from Lists

One of the things I appreciate about Haskell (and why I stuck with it in my “beginner” state for so many years) is that it always forces me to look at things differently. For instance, pre-Haskell, it didn’t occur to me to wonder how a programming language, with its deterministic mechanisms, could possibly produce random numbers. This is because I was used to languages like Python and Javascript, where if you need a random number, you just ask for one.

In my first, early experiences with Haskell, by contrast, I remember thinking, “Why would I need IO to produce a random number?”

getRandChar' :: [Char] -> IO Char
getRandChar' xs = do
  x <- getStdRandom $ randomR (0, (length xs - 1))
  return (xs !! x)

makeRandWord :: [Int] -> IO [Char]
makeRandWord ns = sequence [if n == 0 then getRandChar' vowels else getRandChar' consonants | n <- ns]

My strategy here was to select a random character from a list of characters, but to toggle between vowels and consonants depending on the input list ns, which was a series of 0 and 1 values.

Putting It Together with LiftM2

Here’s how I wrapped up this program, by taking an input parameter using getLine with argument that expects an Int, which represents how long our output taco shop name should be.

buildVals :: Int -> [Int]
buildVals 0 = 1 : buildVals 1
buildVals 1 = 0 : buildVals 0

main = do
  putStrLn "enter total length of taco shop name"
  total_length <- getLine
  let total = (read total_length :: Int) - 6
  g <- getStdGen
  let first = fst $ randomR (0, 1) g
  let randomizer = take total $ buildVals first
  let newWord = makeRandWord randomizer
  newName <- liftM2 (++) newWord ((return "bertos") :: IO [Char])
  -- Alternate (simpler):
  -- fmap (++"bertos") newName
  putStrLn (show newName)

I know that I struggled writing this early program, in part because I can see some confusion about monadic values in my main function (let newWord = makeRandWord randomizer plus liftM2 on the next line…), not to mention the fact that I haven’t used liftM2 since then.

Redoing It

If I were reworking this program today and not changing the logic, I’d probably swap the use of Ints which are only ever 0 and 1 for Bool, which only has two possible values: True and False. Int represents way too many states for what I’d really like to express here.

In addition, in Haskell String is a type synonym for [char], which is typically a linked-list of UTF-16 codepoints. As such, it’s notoriously inefficient, so most Haskell programmers beyond the beginner stage utilize one or both of two well-known and much more performant libraries: text and bytestring.

Still, for this program, where we’re dealing with short strings constructed once and built once, I’d probably stick with String.

What I’d really like to do, though, is make sure the caller is really passing in an Int for an argument and I’d like to make sure that value is of a reasonable size, let’s say something less than maxBound, for instance (where overflow will probably return a negative number):

-- on my machine, `maxBound` comes out to this value:
-- maxBound :: Int
-- 9223372036854775807
import System.Exit -- this line has been added

printTacoShopName :: Int -> IO ()
printTacoShopName total = do
    g <- getStdGen
    let first = fst $ randomR (False, True) g
        randomizer = take total $ buildVals first
    newWord <- makeRandWord randomizer
    print $ newWord ++ "bertos"

main :: IO ()
main = do
  putStrLn "enter total length of taco shop name"
  total_length <- getLine
  let total = (read total_length :: Int) - 6
      maxI = maxBound :: Int
  if total >= maxI || total <= 0
    then die $ "Please pass in a number between 0 and " ++ show maxI
    else printTacoShopName total

I’d probably also use something like optparse-applicative even for this simple, single-argument program, because it’ll handle dynamic input for me.

Rewritten and Running

At any rate, my hastily rewritten version, would probably look something like this:

#!/usr/bin/env stack
--stack --resolver lts-12.24 runghc --package random

-- run to find out your taco shop name
import System.Exit
import System.Random

allLet :: String
allLet = ['a'..'z']

vowels :: String
vowels = "aeiouy"

consonants :: String
consonants = [x | x <- allLet, x `notElem` vowels]

getRandChar' :: String -> IO Char
getRandChar' xs = do
  x <- getStdRandom $ randomR (0, length xs - 1)
  return (xs !! x)

buildVals :: Bool -> [Bool]
buildVals False = True : buildVals True
buildVals True = False : buildVals False

makeRandWord :: [Bool] -> IO String
makeRandWord ns = sequence [if n then getRandChar' vowels else getRandChar' consonants | n <- ns]

printTacoShopName :: Int -> IO ()
printTacoShopName total = do
    g <- getStdGen
    let first = fst $ randomR (False, True) g
        randomizer = take total $ buildVals first
    newWord <- makeRandWord randomizer
    print $ newWord ++ "bertos"


main :: IO ()
main = do
  putStrLn "enter total length of taco shop name"
  total_length <- getLine
  let total = (read total_length :: Int) - 6
        maxI = maxBound :: Int
  if total >= maxI || total <= 0
    then die $ "Please pass in a number between 0 and " ++ show maxI
    else printTacoShopName total

And when I run it, it looks like this:

❯ stack taco_shop_name.hs
12
enter total length of taco shop name
"papuvobertos"

❯ stack taco_shop_name.hs
enter total length of taco shop name
123
"ahyhoqizijeluhiwuzaqubopycegatabyxysojukopexutukejesyfahosazarelodypisoxesafalazecupuhajetelisuxukyjyhuwudodukuponohybertos"

❯ stack taco_shop_name.hs
enter total length of taco shop name
912839128391829312983891923891
Please pass in a number between 0 and 9223372036854775807

Unfortunately, though, it takes way too long if you do pass in a number in the neighborhood of maxBound:

❯ stack taco_shop_name.hs
enter total length of taco shop name
922337203685477580
^C

Even one nowhere near maxBound takes so long that it has to be cancelled:

❯ stack taco_shop_name.hs
enter total length of taco shop name
92222372
^C

Knowing that, how would I write this program today? If this is an application to be in production, I’d probably use bytestring and print each character as we select it instead of trying to construct a giant String value. This could be a fun exercise, but I’m going to leave it up to the reader to implement it!

Tags: haskell