Slightly complex [String] -> [UTCTime]

Question

I have a list of strings, the strings are either a unixtime or an increment from that unixtime eg.

listOfTimes :: [String]
listOfTimes = ["u1345469400","1","2","3","4","5","6","u1346427334","1","2","3","4","5","6"]

I have written functions which take a unixtime and return a UTCTime

dateStringToUTC :: [Char] -> UTCTime
dateStringToUTC a = out
  where
    asInt = read (tail a) :: Integer
    out = psUTC asInt

Or take an increment and the last unixtime and return a UTCTime

incToUTC :: [Char] -> String -> UTCTime
incToUTC a b = madeDate  
  where
    madeDate = psUTC posixOffset
    posixOffset = lastTime + incTime
    lastTime = read (tail a) :: Integer
    incTime = read b :: Integer

However I can't think of a way to write a function that I can map across the entire list that returns a [UTCTime]

Answer 1

map - change each element

fold - combine all the elements

scan - combine all the elements keeping a running "total" - this is what you need

It's going to be easier to keep everything as an Integer until the very end:

type PosixOffset = Integer

A string in your listOfTimes could be a unix time, an increment or an erroneous value. We could represent that by Maybe (Either PosixOffset Integer) but that could get annoying. Let's roll our own:

data Time = Unix PosixOffset | Inc Integer | Error String deriving Show

This allows me to be flexible about what we do later with an error: crash the program with an error, show the Error message to the user but somehow allow them to resume, or ignore the bad value.

Let's make safe version to replace read :: String -> Integer, which returns Nothing instead of crashing. We'll need to import Data.Char (isDigit)

readInteger :: String -> Maybe Integer
readInteger "" = Nothing
readInteger xs | all isDigit xs = Just (read xs)
               | otherwise = Nothing

Now we can use that to readTime with some helpful Error messages.

readTime :: String -> Time
readTime ('u':xs) = case readInteger xs of
                    Just i  -> Unix i
                    Nothing -> Error $ "readTime: there should be an integer after the u, but I got: " ++ 'u':xs
readTime [] = Error "readTime: empty time"
readTime xs = case readInteger xs of
              Just i  -> Inc i
              Nothing -> Error $ "readTime: " ++ xs ++ " is neither a unix time nor an increment."

The plan is to convert our list of Strings to a list of pairs (PosixOffset,Integer), with the last known PosixOffset from a unix time, and the current increment. We'll then need to be able to convert these pairs to a UTCTime

toUTC :: (PosixOffset,Integer) -> UTCTime
toUTC (p,i) = psUTC (p+i)

Now we need to know how to combine the running total of the Times with the next Time. We'll keep hold of the last unix time for reference.

addTime :: (PosixOffset,Integer) -> Time -> (PosixOffset,Integer)
addTime (oldunix,oldinc) time = case time of
    Unix new  -> (new,0)       -- If there's a new unix time, replace and reset the inc to 0.
    Inc inc   -> (oldunix,inc) -- If there's a new increment, replace the old one.
    Error msg -> error msg     -- If there's an error, crash showing it.

or you could use

addTimeTolerant :: (PosixOffset,Integer) -> Time -> (PosixOffset,Integer)
addTimeTolerant (oldunix,oldinc) time = case time of
    Unix new  -> (new,0)          -- If there's a new unix time, replace and reset the inc to 0.
    Inc inc   -> (oldunix,inc)    -- If there's a new increment, replace the old one.
    Error msg -> (oldunix,oldinc) -- If there's an error, ignore it and keep the time the same.

Now we can stick it together: turn the Strings into Times, then combine them into (PosixOffset,Integer) pairs by scanning with addTime, then turn all the resulting pairs into UTCTimes.

runningTotal :: [String] -> [UTCTime]
runningTotal [] = []
runningTotal xss = let (t:ts) = map readTime xss in      -- turn Strings to Times
    case t of
        Error msg -> error msg
        Inc _     -> error "runningTotal: list must start with a unix time"
        Unix po   -> map toUTC $ scanl addTime (po,0) ts -- scan the list adding times, 
                                                         -- starting with an initial unix time
                                                         -- then convert them all to UTC

or if you like the keep calm and carry on approach of addTimeTolerant, you could use

isn't_UnixTime :: Time -> Bool
isn't_UnixTime (Unix _) = False
isn't_UnixTime _        = True

runningTotalTolerant :: [String] -> [UTCTime]
runningTotalTolerant xss = 
  let ts = dropWhile isn't_UnixTime (map readTime xss) in    -- cheerily find the first unix time
    if null ts then [] else                                  -- if there wasn't one, there are no UTCTimes
       let (Unix po) = head ts in                            -- grab the first time
          map toUTC $ scanl addTimeTolerant (po,0) (tail ts) -- scan the list adding times, 
                                                             -- starting with an initial unix time
                                                             -- then convert them all to UTC

Answer 2

Another way would be to collect the times that correspond to each other into a separate list, and deal with them separately, i.e.

convertUTCs [] = []
convertUTCs (x:xs) = map (incToUTC x) increments ++ convertUTCs rest
  where
    (increments, rest) = break (\str -> head str == 'u') xs

This takes the first element (which should always be of the form "u12345") and all the increments for that time (i.e. the elements that don't start with 'u'), and then does the processing on them.

Answer 3

timesToUnixTimes :: [String] -> [UTCTime]

As ja points out, this is not a simple map. But the final step of converting a [Integer] to a [UTCTime] is a map:

timesToUnixTimes (s : ss) = map psUTC (i : is)
  where

The first element of the input list, s, had better be a unixtime:

    i = read (tail s) :: Integer

Subsequent elements, ss, may be either, so the decoding function needs access to the previous element of the output list:

    is = zipWith timeToInteger ss (i : is)

Writing timeToInteger :: String -> Integer -> Integer is left as an exercise.

Two points from this:

1. You can think of zipWith as mapping a function over two lists at a time (similarly, zipWith3 maps a function over three lists at a time, zipWith4 maps over four lists at a time, etc; there isn't a function called zipWith1 because it's called map).

2. is appears in its own definition. This works thanks to laziness non-strictness.

   1. The first element of is depends on the first element of ss and on i.
   2. The second element of is depends on the second element of ss and on the first element of is.
   3. The third element of is depends on the third element of ss and on the second element of is.
   4. Etc.

   No element of is depends on itself, or on a later element of is.

Answer 4

As ja's answer says, this is not a simple map. A general fold would work, but that's true of any list operation.

What you're trying to do here sounds more specifically like a use for scanr, which is a right fold that produces a list of each intermediate step rather than just the final result. In your case, the accumulator would be the previous time value, and at each step you'd either add an increment or replace it with a new time. The output would be a (lazy!) list of each computed time.

Answer 5

It's not a map, because you have 2 parameters to your inc function - you're using a previous list element in subsequent calls. Look into folds: foldl, foldr, etc.