Re: [Haskell-beginners] how to skip pattern match error when applying a mapM_

17 Jan 2017

      Hello

In fact I realize that my real problem is during the 'values' generation of my example.

I have a class like this

class Frame t where
  len :: t -> IO (Maybe Int)
  row :: t -> Int -> IO (Maybe (DifTomoFrame DIM1))

And I create an instance for my dataframe comming from an hdf5 file.
some time there is Nan values returned by the get_position method.
I decided to return a Maybe Double and Nan -> Nothing

instance Frame DataFrameH5 where
  len d =  lenH5Dataspace (h5delta d)

  row d idx = do
    Just n <- len d
    let eof = n - 1 == idx
    let nxs' = h5nxs d
    let mu = 0.0
    let komega = 0.0
    let kappa = 0.0
    let kphi = 0.0
    Just gamma <- get_position' (h5gamma d) 0
    Just delta <- get_position' (h5delta d) idx
    Just wavelength <- get_position' (h5wavelength d) 0
    let source = Source (head wavelength *~ nano meter)
    let positions = concat [mu, komega, kappa, kphi, gamma, delta]
    -- print positions
    let geometry =  Geometry K6c source positions Nothing
    let detector = ZeroD
    m <- geometryDetectorRotationGet geometry detector
    poniext <- ponigen d (MyMatrix HklB m) idx
    return $ Just DifTomoFrame { difTomoFrameNxs = nxs'
                               , difTomoFrameIdx = idx
                               , difTomoFrameEOF = eof
                               , difTomoFrameGeometry = geometry
                               , difTomoFramePoniExt = poniext
                               }
        where
          get_position' a b = do
                             v <- get_position a b
                             return $ if any isNaN v
                                      then Nothing
                                      else Just v

I iterate for each idx of my dataframe
So I would like row to return Nothing as soon as the get_position' return Nothing

but when I use this code, I get the error and it stop my program instead of skipping the point.

________________________________________
De : Beginners [beginners-bounces@haskell.org] de la part de Francesco Ariis [fa-ml@ariis.it]
Envoyé : mardi 17 janvier 2017 15:19
À : The Haskell-Beginners Mailing List - Discussion of primarily beginner-level topics related to Haskell
Objet : Re: [Haskell-beginners] how to skip pattern match error when applying a mapM_

On Tue, Jan 17, 2017 at 02:03:02PM +0000, PICCA Frederic-Emmanuel wrote:
...
Hello,
Here a reduction of my problem
values :: IO [IO (Maybe Int)]
values = do
  let v = [Just 1, Just 2, Just 3, Nothing, Just 5, Nothing, Just 7] :: [Maybe Int]
  return $ map return v
main :: IO ()
main = do
  vs <- values
  nvs <- mapM_ go vs
  print nvs
    where
      go :: IO (Maybe Int) -> IO Int
      go v' = do
         Just v <- v'
         return v
Hello Frédéric,
    `Just v <- v'` doesn't silently skip Nothing values, but it's
a full fledged pattern match (and one reason why I dislike `do
notation` as a syntactic sugar).

A way to solve the problem is to take advantage of `sequence`
and `catMaybes` (from `Data.Maybe`).

    λ> :t sequence
    sequence :: (Monad m) => [m a] -> m [a]
        -- I cheated a bit on the signature, but the gist
        -- of it is: from a list of monadic actions, to
        -- one monadic action returning a list of results.

    λ> :t catMaybes
    catMaybes :: [Maybe a] -> [a]

With that your main gets simpler:

    main :: IO ()
    main = do vs <- values           -- vs :: [IO (Maybe Int)]
              sv <- sequence vs      -- sequence vs :: IO [Maybe Int]
                                     -- sv :: [Maybe Int]
              print (M.catMaybes sv)

Does this help?
_______________________________________________
Beginners mailing list
Beginners@haskell.org
http://mail.haskell.org/cgi-bin/mailman/listinfo/beginners