Converting image size from Maybe (Int, Int) to (Float, Float)

Question

I have a function that returns size of an image IO (Maybe (Int, Int)) It would be nice to convert it to float tuple beautifully.

I am sure there must be a better way than what I currently have:

data Vec2D = Vec2D Float Float
getImageSize :: FilePath -> IO (Maybe (Int, Int))
getImageSize path = do
 img <- readImage path
 case img of
 Left _ -> return Nothing
 Right img' -> return (go img')
 where
 go :: DynamicImage -> Maybe (Int, Int)
 go (ImageRGB8 (Image w h _)) = Just (w, h)
 go (ImageRGBA8 (Image w h _)) = Just (w, h)
 go _ = Nothing
getImageSizeVec :: FilePath -> IO Vec2D
getImageSizeVec path = do
 res <- getImageSize path
 let dims = fromMaybe (0, 0) res
 return $ Vec2D (fromIntegral (fst dims)) (fromIntegral (snd dims)) 

Answer 1

1. I do not understand what benefit you get from the conversion to float
2. I would advise you to not mix - IO with the mostly pure calculation you have

so let us look at your code (minus the IO)

getImageSize :: DynamicImage -> Maybe (Int, Int)
getImageSize (ImageRGB8 (Image w h _)) = Just (w, h)
getImageSize (ImageRGBA8 (Image w h _)) = Just (w, h)
getImageSize _ = Nothing

but readImage already has a more useful return type than Maybe it has an Either String which is giving you error messages instead of Nothing so we can adjust.

getImageSize :: DynamicImage -> Either String (Int, Int)
getImageSize (ImageRGB8 (Image w h _)) = Right (w, h)
getImageSize (ImageRGBA8 (Image w h _)) = Right (w, h)
getImageSize _ = Left "Image type neither ImageRGB8 or ImageRGBA8"

this section could be improved a bit - also working on pure data

sizeToVec :: (Int,Int) -> Vec2D
sizeToVec (x,y) = Vec2D (fromIntegral x) (fromIntegral y)

you can then use those with the many mechanisms haskell provides you:

 readImage path :: IO (Either String DynamicImage)
 getImageSize :: DynamicImage -> Either String (Int, Int)
 sizeToVec :: (Int,Int) -> Vec2D

the last two you can combine à la

fmap sizeToVec (getImageSize img) -- or fmap sizeToVec . getImageSize

and the first two

doStuff = do eitherImg <- readImg mypathtoimg
 return $ do size <- getImageSize
 sizeToVec size

or

doStuff = do eitherImg <- readImg mypathtoimg
 return (getImageSize eitherImg >>= sizeToVec)

or even shorter

(削除) doStuff = readImg mypathtoimg >>= (getImageSize >>= sizeToVec) (削除ここまで)

doStuff = readImg path >>= fmap sizeToVec . getImageSize

if I am not mistaken (I have no haskell env on this machine, therefore this code is untested - sorry)

• Zeta: sizeToVec has the wrong type for >>=. Perhaps you meant readImg path >>= fmap sizeToVec . getImageSize?