2
\$\begingroup\$

The following asks for 2 sets of data (with corresponding x and y values), calculates a linear equation that matches the data as closely as possible, then allows you to "play around" with the equation by substituting x values.

import Data.Char slope :: [Float] -> [Float] -> Float slope xs ys | xN /= yN = error $ "x len: " ++ show xN ++ " y len: " ++ show yN | otherwise = (xN * (sum $ coorProds xs ys) - xSum * ySum) / (xN * (sum $ squares xs) - (xSum * xSum)) where (xN,yN) = (fromIntegral $ length xs, fromIntegral $ length ys) (xSum,ySum) = (sum xs, sum ys) intercept :: [Float] -> [Float] -> Float intercept xs ys | xN /= yN = error $ "x len: " ++ show xN ++ " y len: " ++ show yN | otherwise = (ySum - (b * xSum)) / xN where b = slope xs ys (xN,yN) = (fromIntegral $ length xs, fromIntegral $ length ys) (xSum,ySum) = (sum xs, sum ys) getEquation :: [Float] -> [Float] -> ((Float -> Float),Float,Float) getEquation xs ys = (\x -> (inter + (b * x)) , inter, b) where inter = intercept xs ys b = slope xs ys coorProds :: [Float] -> [Float] -> [Float] coorProds xs ys = map (\(x,y) -> x * y) $ zip xs ys squares :: [Float] -> [Float] squares = map (\x -> x * x) cutIntoPair :: String -> (String,String) cutIntoPair = (\(x,y)-> (x,drop 1 y)) . break (== ' ') isFloat :: String -> Bool isFloat = all (\c -> isDigit c || c == '.') toFloat :: String -> Float toFloat s | isFloat s = read s :: Float | otherwise = error $ show s ++ " is not a number" getPairs :: IO ([Float],[Float]) getPairs = do i <- getLine let (x,y) = cutIntoPair i if isFloat x then do (xs,ys) <- getPairs return ((toFloat x : xs), (toFloat y : ys)) else return ([],[]) useEquation :: (Float -> Float) -> IO () useEquation e = do putStr "When x = "; x <- getLine putStrLn $ "y = " ++ (show $ e (toFloat x)) ++ "\n" useEquation e showEquation :: Float -> Float -> String showEquation i b = "y = " ++ show i ++ " + " ++ show b ++ "x" main = do putStrLn $ "Enter x and y values seperated by a space\n\t" ++ "Then type a non-number to continue" (xs,ys) <- getPairs let (linEq,i,b) = getEquation xs ys putStrLn $ showEquation i b putStrLn "Enter x values to get their calculated y values" useEquation linEq 

I'm looking for general feedback, and suggestions regarding the following points:

  1. getEquation has a couple issues.

    • I was able to write it point-free when the function only returned the equation, but now that I'm wrapping everything in a tuple, so I can also return the calculated slope and intercept values (to show the equation later), it's not behaving as I expected. I tried writing it as a section, like (inter + (b *), but that attempts to apply + to the function (b*), which obviously results in a "No instance" error. My solution was to just wrap it in a lambda, but that's quite ugly, in my opinion.
    • Is there a better way of dealing with the need to have the slope and intercept after the function is "constructed" than passing them back with the equation? The optimal solution would the ability to extract them out of the function, but I'm pretty sure that's not possible.
  2. cutIntoPair looks ugly too. It splits a string at the first space, then returns the 2 parts. Break leaves the space in the second slot though, so to remove it, I resorted to passing the result to a lambda that leaves the first part of the tuple the same, but drops the first char of the second slot. I'd like to know if there is a cleaner way to go about this task.

  3. The slope and intercept functions are slightly messy, but I had to cram a lot of math into the space. Tips on how to format it neater would be appreciated.

To use it, type x and y values separated by a space, pressing enter between sets. When you're done entering, type a non-number, like so:

2 4 4 8 12 24 k 
\$\endgroup\$
2
  • \$\begingroup\$ Regarding the paragraph I've removed, you can put it into a self-answer if no one else gets to it first. If someone did so anyway, then you would no longer be able to change the code in this post. \$\endgroup\$ Commented Sep 30, 2014 at 20:27
  • \$\begingroup\$ Thanks. It's was a minor issue anyways. I only added it as an afterthought. \$\endgroup\$ Commented Sep 30, 2014 at 21:49

1 Answer 1

3
\$\begingroup\$

I'd structure your program in a slightly different way. First, instead of passing around two arrays of float, why don't you introduce a Point data structure and work with it? It would save you from checking that you get passed two array of the same length, which looks quite neat to me.

I think you could also consider introducing a Coefficients data structure to hold your pair of slope and intercept value. If you do that it will be much clearer what your function does.

Since you eventually want to have a function you can play with, you could create a toFunction :: Coefficients -> Float ->Float. In this way you decouple the computation of the coefficients that best fit with your data and the construction of the corresponding curve. This enables you to play not only with the curves you fit from data but also with curves you can define from their coefficients.

I think that a nicer way to compute the value could be by putting as much math as possible in some temporary variables you define in the where block. That could allow you to separate the high level computation you need to perform with all the other smaller steps required to compute some other intermediate results you need.

If you apply the changes I suggested you should come up with code similar to the following:

type Point = (Float, Float) type Coefficients = (Float, Float) toFunction :: Coefficients -> Float -> Float toFunction (s, i) x = s * x + i slope :: [Point] -> Float slope [] = error "Empty list of points" slope (_ : []) = error "Single point" slope points = (n*sumCoorProds - (sumX*sumY))/ (n*sumSquares - (sumX*sumX)) where n = fromIntegral $ length points (xs, ys) = unzip points sumCoorProds = sum $ map ( \(x, y) -> x * y ) points sumSquares = sum $ map ( \(x) -> x * x ) xs sumX = sum xs sumY = sum ys intercept :: [Point] -> Float intercept points = (sumY - (s * sumX)) / n where (xs, ys) = unzip points n = fromIntegral $ length points s = slope points sumX = sum xs sumY = sum ys getCoefficients :: [Point] -> Coefficients getCoefficients points = (slope points, intercept points) 
\$\endgroup\$
0

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.