I want to sort a list of String first by the length of the strings, and if the length is the same then it should sort lexically. I thought I could use the Data.List library and write my own compare function that does that. So the compare function should take a list of String as the argument and compare all the the elements (which are Strings). A compare function for Strings would look like this
comp a b
| length a > length b = GT
| length a < length b = LT
How could I address all the list elements with such a function?
First of all, your cmp function does not handle the case where the lengths are equal: you need to add that. Otherwise you'll get an runtime pattern match error:
comp a b
| length a > length b = GT
| length a < length b = LT
| otherwise = undefined -- TODO
also, note that this implementation sometimes computes the length twice, but it's likely that GHC optimizes this one away on its own, and we'll get to solving this later on more fundamentally anyway.
Then, once you've fixed your comp, all you need to do is pass it to Data.List.sortBy together with the list of strings you want to sort. An ipmplementation like that is provided below (<$> is the operator alias of fmap which works the same as map does on lists).
However, there's a better solution where you first compute the length of all elements in the list, by mapping each of the elements into a pair where the first member is the original string and the second one is its length. You then use a modified comp function that takes 2 pairs instead of just 2 strings, but otherwise behaves the same as your original comp. However, you then need to map the intermediate list back to just containing the strings (which is what the fst <$> is for, which is equivalent to map fst but, again, uses the, IMO nicer looking, <$> opetator).
So the somewhat naive solution would be:
sortByLenOrLex :: [String] -> [String]
sortByLenOrLex as = sortBy cmp as where
cmp a b | n > m = GT
| n < m = LT
| otherwise = compare a b
where n = length a
m = length b
and the more efficient one, as leftaroundabout points out, would be:
sortByLenOrLex' :: [String] -> [String]
sortByLenOrLex' as = fst <$> sortBy cmp (addLen <$> as) where
cmp (a,n) (b,m) | n > m = GT
| n < m = LT
| otherwise = compare a b
addLen x = (x, length x)
where the list is first amended with the lengths of each of its elements, so as to avoid duplicate, expensive length calls.
EDIT: please see chi's answer for a much nicer implementation of this algorithm!
Furthermore:
You can make your functions generic by making them operate on lists of lists of Ord:
sortByLenOrLex'' :: Ord a => [[a]] -> [[a]]
sortByLenOrLex'' as = fst <$> sortBy cmp (addLen <$> as) where
cmp (a,n) (b,m) | n > m = GT
| n < m = LT
| otherwise = compare a b
addLen x = (x, length x)
this gives you:
*Main> sortByLenOrLex'' [[1,2], [1,3], [1,2,3]]
[[1,2],[1,3],[1,2,3]]
...and if you want to make it as generic as possible, you can sort lists of Foldable of Ord:
sortByLenOrLex''' :: (Foldable f, Ord a) => [f a] -> [f a]
sortByLenOrLex''' as = unamend <$> sortBy cmp (amend <$> as) where
cmp (a,n,a') (b,m,b') | n > m = GT
| n < m = LT
| otherwise = compare a' b'
amend x = (x, length x, toList x)
unamend (x,_,_) = x
this gives you:
*Main> sortByLenOrLex''' [Just 3, Just 4, Just 3, Nothing]
[Nothing,Just 3,Just 3,Just 4]
*Main> sortByLenOrLex''' [(4,1),(1,1),(1,2),(1,1),(3,1)]
[(4,1),(1,1),(1,1),(3,1),(1,2)]
*Main> sortByLenOrLex''' [Left "bla", Right "foo", Right "foo", Right "baz"]
[Left "bla",Right "baz",Right "foo",Right "foo"]
*Main> sortByLenOrLex''' [(3,"hello"),(2,"goodbye"),(1,"hello")]
[(2,"goodbye"),(3,"hello"),(1,"hello")]
A variant of #Erik's solution, using some combinators from the library:
import Data.List
import Control.Arrow
sortByLen = map snd . sort . map (length &&& id)
This is essentially a Schwartzian transform.
Related
I just started learning Haskell. I am trying to get the list of all common letters between two words, for example, for "hello" and "llama" that would be [ 'l', 'l' ], for "happy" and "pay", [ 'a', 'p', 'y' ].
I tried using intersect but I have trouble with duplicates, "happy" and "pay" result in [ 'a', 'p', 'p', 'y' ]. I can't just remove duplicates cause they can exist, as in the first example.
I would be grateful for any suggestions. Thanks!
You can use the multiset package:
Data.MultiSet> fromList "hello" `intersection` fromList "llama"
fromOccurList [('l',2)]
Data.MultiSet> fromList "happy" `intersection` fromList "pay"
fromOccurList [('a',1),('p',1),('y',1)]
The data-ordlist package also offers this functionality:
Data.List Data.List.Ordered> sort "hello" `isect` sort "llama"
"ll"
Data.List Data.List.Ordered> sort "happy" `isect` sort "pay"
"apy"
Here's a nice technique that's worth learning. Suppose you have two sorted lists:
[1,1,5,10,15,15,18]
[2,5,8,10,15,20]
and you want to merge them together into a single sorted list. In Haskell, there's a very elegant way to write this algorithm using pattern matching and guards:
merge (x:xs) (y:ys) | x < y = x : merge xs (y:ys)
| otherwise = y : merge (x:xs) ys
merge xs [] = xs
merge [] ys = ys
so that:
> merge [1,1,5,10,15,15,18] [2,5,8,10,15,20]
[1,1,2,5,5,8,10,10,15,15,15,18,20]
>
In a nutshell, when both lists are non-empty, it compares the heads of both lists and outputs the smallest head; then it uses recursion to output "the rest".
It could also have been written with the three cases (less, greater, and equal) all made explicit:
merge (x:xs) (y:ys) | x < y = x : merge xs (y:ys)
| x > y = y : merge (x:xs) ys
| otherwise = y : merge (x:xs) ys
merge xs [] = xs
merge [] ys = ys
and this general template can be used to implement a number of interesting algorithms on sorted lists. Here's one that removes common elements, for example:
uncommon (x:xs) (y:ys) | x < y = x : uncommon xs (y:ys)
| x > y = y : uncommon (x:xs) ys
| otherwise = uncommon xs ys
uncommon xs [] = xs
uncommon [] ys = ys
so that:
> uncommon [1,1,5,10,15,15,18] [2,5,8,10,15,20]
[1,1,2,8,15,18,20]
>
You might want to try modifying the uncommon function to create a diff function that outputs the result of removing the elements of the second list from the first. It will require modifying one of the first three guarded cases, and you'll also need to adjust one of the two "empty list" pattern matches:
> diff [1,1,5,10,15,15,18] [2,5,8,10,15,20]
[1,1,15,18]
>
Once you've figured this out, you'll find it easy to create a common function that outputs the shared elements of the two sorted lists to give:
> common [1,1,5,10,15,15,18] [2,5,8,10,15,20]
[5,10,15]
>
Since strings are just lists of characters, this would work for your problem, too, using sort from Data.List to pre-sort the lists:
> import Data.List
> common (sort "hello") (sort "llama")
"ll"
> common (sort "happy") (sort "pay")
"apy"
>
I think this is an ideal case to use Data.Map. I would implement this as follows;
import qualified Data.Map.Lazy as M
sharedLetters :: String -> String -> String
sharedLetters s1 s2 = let cm = foldr (checkMap (\(x,y) -> (x,y+1))) charMap s2
where checkMap f c m = if M.member c m then M.adjust f c m
else M.insert c (f (0,0)) m
charMap = foldr (checkMap (\(x,y) -> (x+1,y))) M.empty s1
in M.foldlWithKey (\r k (v1,v2) -> r ++ replicate (minimum [v1,v2]) k) "" cm
main :: IO String
main = do
putStr "Enter first string :"
s1 <- getLine
putStr "Enter second string :"
s2 <- getLine
return $ sharedLetters s1 s2
Enter first string :happy
Enter second string :pay
"apy"
Enter first string :pay
Enter second string :happy
"apy"
Enter first string :hello
Enter second string :llama
"ll"
Enter first string :llama
Enter second string :hello
"ll"
How about exploiting the fact that every letter shared between the words (allowing duplicates) shows up as a pair of that letter in the set formed from the union of those words? You can find such pairs efficiently by sorting the union set and picking out duplicates -
let find_dups ([]) = []; find_dups (x:y:xs) | x == y = x:find_dups(xs); find_dups (x:xs) = find_dups(xs)
let common_letters word1 word2 = find_dups (sort (word1 ++ word2))
> common_letters "hello" "fellows"
"ello"
Suppose we have a simple grammar specification. There is a way to enumerate terms of that grammar that guarantees that any finite term will have a finite position, by iterating it diagonally. For example, for the following grammar:
S ::= add
add ::= mul | add + mul
mul ::= term | mul * term
term ::= number | ( S )
number ::= digit | digit number
digit ::= 0 | 1 | ... | 9
You can enumerate terms like that:
0
1
0+0
0*0
0+1
(0)
1+0
0*1
0+0*0
00
... etc
My question is: is there a way to do the opposite? That is, to take a valid term of that grammar, say, 0+0*0, and find its position on such enumeration - in that case, 9?
For this specific problem, we can cook up something fairly simple, if we allow ourselves to choose a different enumeration ordering. The idea is basically the one in Every Bit Counts, which I also mentioned in the comments. First, some preliminaries: some imports/extensions, a data type representing the grammar, and a pretty-printer. For the sake of simplicity, my digits only go up to 2 (big enough to not be binary any more, but small enough not to wear out my fingers and your eyes).
{-# LANGUAGE TypeSynonymInstances #-}
import Control.Applicative
import Data.Universe.Helpers
type S = Add
data Add = Mul Mul | Add :+ Mul deriving (Eq, Ord, Show, Read)
data Mul = Term Term | Mul :* Term deriving (Eq, Ord, Show, Read)
data Term = Number Number | Parentheses S deriving (Eq, Ord, Show, Read)
data Number = Digit Digit | Digit ::: Number deriving (Eq, Ord, Show, Read)
data Digit = D0 | D1 | D2 deriving (Eq, Ord, Show, Read, Bounded, Enum)
class PP a where pp :: a -> String
instance PP Add where
pp (Mul m) = pp m
pp (a :+ m) = pp a ++ "+" ++ pp m
instance PP Mul where
pp (Term t) = pp t
pp (m :* t) = pp m ++ "*" ++ pp t
instance PP Term where
pp (Number n) = pp n
pp (Parentheses s) = "(" ++ pp s ++ ")"
instance PP Number where
pp (Digit d) = pp d
pp (d ::: n) = pp d ++ pp n
instance PP Digit where pp = show . fromEnum
Now let's define the enumeration order. We'll use two basic combinators, +++ for interleaving two lists (mnemonic: the middle character is a sum, so we're taking elements from either the first argument or the second) and +*+ for the diagonalization (mnemonic: the middle character is a product, so we're taking elements from both the first and second arguments). More information on these in the universe documentation. One invariant we'll maintain is that our lists -- with the exception of digits -- are always infinite. This will be important later.
ss = adds
adds = (Mul <$> muls ) +++ (uncurry (:+) <$> adds +*+ muls)
muls = (Term <$> terms ) +++ (uncurry (:*) <$> muls +*+ terms)
terms = (Number <$> numbers) +++ (Parentheses <$> ss)
numbers = (Digit <$> digits) ++ interleave [[d ::: n | n <- numbers] | d <- digits]
digits = [D0, D1, D2]
Let's see a few terms:
*Main> mapM_ (putStrLn . pp) (take 15 ss)
0
0+0
0*0
0+0*0
(0)
0+0+0
0*(0)
0+(0)
1
0+0+0*0
0*0*0
0*0+0
(0+0)
0+0*(0)
0*1
Okay, now let's get to the good bit. Let's assume we have two infinite lists a and b. There's two things to notice. First, in a +++ b, all the even indices come from a, and all the odd indices come from b. So we can look at the last bit of an index to see which list to look in, and the remaining bits to pick an index in that list. Second, in a +*+ b, we can use the standard bijection between pairs of numbers and single numbers to translate between indices in the big list and pairs of indices in the a and b lists. Nice! Let's get to it. We'll define a class for Godel-able things that can be translated back and forth between numbers -- indices into the infinite list of inhabitants. Later we'll check that this translation matches the enumeration we defined above.
type Nat = Integer -- bear with me here
class Godel a where
to :: a -> Nat
from :: Nat -> a
instance Godel Nat where to = id; from = id
instance (Godel a, Godel b) => Godel (a, b) where
to (m_, n_) = (m + n) * (m + n + 1) `quot` 2 + m where
m = to m_
n = to n_
from p = (from m, from n) where
isqrt = floor . sqrt . fromIntegral
base = (isqrt (1 + 8 * p) - 1) `quot` 2
triangle = base * (base + 1) `quot` 2
m = p - triangle
n = base - m
The instance for pairs here is the standard Cantor diagonal. It's just a bit of algebra: use the triangle numbers to figure out where you're going/coming from. Now building up instances for this class is a breeze. Numbers are just represented in base 3:
-- this instance is a lie! there aren't infinitely many Digits
-- but we'll be careful about how we use it
instance Godel Digit where
to = fromIntegral . fromEnum
from = toEnum . fromIntegral
instance Godel Number where
to (Digit d) = to d
to (d ::: n) = 3 + to d + 3 * to n
from n
| n < 3 = Digit (from n)
| otherwise = let (q, r) = quotRem (n-3) 3 in from r ::: from q
For the remaining three types, we will, as suggested above, check the tag bit to decide which constructor to emit, and use the remaining bits as indices into a diagonalized list. All three instances necessarily look very similar.
instance Godel Term where
to (Number n) = 2 * to n
to (Parentheses s) = 1 + 2 * to s
from n = case quotRem n 2 of
(q, 0) -> Number (from q)
(q, 1) -> Parentheses (from q)
instance Godel Mul where
to (Term t) = 2 * to t
to (m :* t) = 1 + 2 * to (m, t)
from n = case quotRem n 2 of
(q, 0) -> Term (from q)
(q, 1) -> uncurry (:*) (from q)
instance Godel Add where
to (Mul m) = 2 * to m
to (m :+ t) = 1 + 2 * to (m, t)
from n = case quotRem n 2 of
(q, 0) -> Mul (from q)
(q, 1) -> uncurry (:+) (from q)
And that's it! We can now "efficiently" translate back and forth between parse trees and their Godel numbering for this grammar. Moreover, this translation matches the above enumeration, as you can verify:
*Main> map from [0..29] == take 30 ss
True
We did abuse many nice properties of this particular grammar -- non-ambiguity, the fact that almost all the nonterminals had infinitely many derivations -- but variations on this technique can get you quite far, especially if you are not too strict on requiring every number to be associated with something unique.
Also, by the way, you might notice that, except for the instance for (Nat, Nat), these Godel numberings are particularly nice in that they look at/produce one bit (or trit) at a time. So you could imagine doing some streaming. But the (Nat, Nat) one is pretty nasty: you have to know the whole number ahead of time to compute the sqrt. You actually can turn this into a streaming guy, too, without losing the property of being dense (every Nat being associated with a unique (Nat, Nat)), but that's a topic for another answer...
How can I efficiently represent the list [0..] \\ [t+0*p, t+1*p ..]?
I have defined:
Prelude> let factors p t = [t+0*p, t+1*p ..]
I want to efficiently represent an infinite list that is the difference of [0..] and factors p t, but using \\ from Data.List requires too much memory for even medium-sized lists:
Prelude Data.List> [0..10000] \\ (factors 5 0)
<interactive>: out of memory
I know that I can represent the values between t+0*p and t+1*p with:
Prelude> let innerList p1 p2 t = [t+p1+1, t+p1+2 .. t+p2-1]
Prelude> innerList 0 5 0
[1,2,3,4]
However, repeatedly calculating and concatenating innerList for increasing intervals seems clumsy.
Can I efficiently represent [0..] \\ (factors p t) without calculating rem or mod for each element?
For the infinite list [0..] \\ [t,t+p..],
yourlist t p = [0..t-1] ++ [i | m <- [0,p..], i <- [t+m+1..t+m+p-1]]
Of course this approach doesn't scale, at all, if you'd want to remove some other factors, like
[0..] \\ [t,t+p..] \\ [s,s+q..] \\ ...
in which case you'll have to remove them in sequence with minus, mentioned in Daniel Fischer's answer. There is no magic bullet here.
But there's also a union, with which the above becomes
[0..] \\ ( [t,t+p..] `union` [s,s+q..] `union` ... )
the advantage is, we can arrange the unions in a tree, and get algorithmic improvement.
You can't use (\\) for that, because
(\\) :: (Eq a) => [a] -> [a] -> [a]
(\\) = foldl (flip delete)
the list of elements you want to remove is infinite, and a left fold never terminates when the list it folds over is infinite.
If you rather want to use something already written than write it yourself, you can use minus from the data-ordlist package.
The performance should be adequate.
Otherwise,
minus :: Ord a => [a] -> [a] -> [a]
minus xxs#(x:xs) yys#(y:ys)
| x < y = x : minus xs yys
| x == y = minus xs ys
| otherwise = minus xss ys
minus xs _ = xs
You can use a list comprehesion with a predicate, using rem:
>>> let t = 0
>>> let p = 5
>>> take 40 $ [ x | x <- [1..], x `rem` p /= t ]
[1,2,3,4,6,7,8,9,11,12,13,14,16,17,18,19,21,22,23,24,26,27,28,29,31,32,33,34,36,37,38,39,41,42,43,44,46,47,48,49]
If you want efficiency, why does your solution have to use list comprehension syntax?
Why not something like this?
gen' n i p | i == p = gen' (n + p) 1 p
gen' n i p = (n+i) : gen' n (i+1) p
gen = gen' 0 1
and then do
gen 5
Because you have ascending lists, you can simply lazily merge them:
nums = [1..]
nogos = factors p t
result = merge nums (dropWhile (<head nums) nogos) where
merge (a:as) (b:bs)
| a < b = a : merge as (b:bs)
| a == b = merge as bs
| otherwise = error "should not happen"
Writing this in a general way so that we have a function that builds the difference of two infinite lists, provided only that they are in ascending order, is left as exercise. In the end, the following should be possible
[1..] `infiniteDifference` primes `infiniteDifference` squares
For this, make it a left associative operator.
Given an arbitrary set holding an arbitrary number of elements of arbitrary type, e.g.
mySet1 = Set.fromList [1,2,3,4]
or
mySet2 = Set.fromList ["a","b","c","d"]
or
mySet3 = Set.fromList [A, B, C, D]
for some data constructors A, B, C, D, ...
What is the computationally most efficient way to generate the set of all unordered pairs of elements is the given set? I.e.
setPairs mySet1 == Set.fromList [(1,2),(1,3),(1,4),(2,3),(2,4),(3,4)]
or
setPairs mySet2 == fromList [ ("a","b")
, ("a","c")
, ("a","d")
, ("b","c")
, ("b","d")
, ("c","d") ]
or
setPairs mySet2 == fromList [ (A,B)
, (A,C)
, (A,D)
, (B,C)
, (B,D)
, (C,D) ]
My initial naive guess would be:
setPairs s = fst $ Set.fold
(\e (pairAcc, elementsLeft) ->
( Set.fold
(\e2 pairAcc2 ->
Set.insert (e2, e) pairAcc2
) pairAcc $ Set.delete e elementsLeft
, Set.delete e elementsLeft )
) (Set.empty, s) s
but surely that cannot be the best solution?
Benchmarking might prove me wrong, but my suspicion is that there's no win in staying in the set representation. You're going to need O(n^2) regardless, because that's the size of the output. The key advantage would be producing your list such that you could use a call to S.fromDistinctAscList such that it only costs O(n) to build the set itself.
The following is pretty clean, preserves a fair amount of sharing, and is generally the simplest, most straightforward and intuitive solution I can imagine.
pairs s = S.fromDistinctAscList . concat $ zipWith zip (map (cycle . take 1) ts) (drop 1 ts)
where ts = tails $ S.toList s
Edit
Shorter/clearer (not sure performancewise, but probably as good/better):
pairs s = S.fromDistinctAscList [(x,y) | (x:xt) <- tails (S.toList s), y <- xt]
At first, you need to generate all sets. replicateM from Control.Monad helps with it.
λ> replicateM 2 [1..4]
[[1,1],[1,2],[1,3],[1,4],[2,1],[2,2],[2,3],[2,4],[3,1],[3,2],[3,3],[3,4],[4,1],[4,2],[4,3],[4,4]]
Then you need to filter pairs, where second element is greater than first
λ> filter (\[x,y] -> x < y) $ replicateM 2 [1 .. 4]
[[1,2],[1,3],[1,4],[2,3],[2,4],[3,4]]
Finally, you need to convert every list in a tuple
λ> map (\[x,y] -> (x,y)) $ filter (\[x,y] -> x < y) $ replicateM 2 [1 .. 4]
[(1,2),(1,3),(1,4),(2,3),(2,4),(3,4)]
Then we can formulate it into function pairs:
import Data.Set
import Control.Monad
import Data.List
mySet = Data.Set.fromList [1,2,3,4]
--setOfPairs = Data.Set.fromList [(1,2),(1,3),(1,4),(2,3),(2,4),(3,4)]
setOfPairs = Data.Set.fromList $ pairs mySet
pairs :: Ord a => Set a -> [(a,a)]
pairs x = Data.List.map (\[x,y] -> (x,y)) $ Data.List.filter (\[x,y] -> x < y) $ replicateM 2 $ toList x
So, if I got you question right, you can use pairs mySet, where pairs generate the list of all unordered pairs of mySet.
Is it what you want?
UPD:
List comprehension could be more clear and fast technique to create such sublists, so here is another instance of pairs:
pairs :: Ord a => Set a -> [(a,a)]
pairs set = [(x,y) | let list = toList set, x <- list, y <- list, x < y]
So here is a first stab at a solution using conversion back and forth to a list. Again, I am not sure this is the fastest way to do this but I do know that iteration over sets it's not terribly efficient.
import Data.List
import qualified Data.Set as S
pairs :: S.Set String -> S.Set (String,String)
pairs s = S.fromList $ foldl' (\st e -> (zip l e) ++ st) [] ls
where (l:ls) = tails $ S.toList s
By folding zip over the tails, you get a nice and efficient way to create the set of unordered pairs. However, instinct encourages me that there may be a monadic filterM or foldM solution that's even more elegant. I will keep thinking.
[EDIT]
So here is what should be [but is not on account of the size of the powerset] a faster solution that does not require a toList.
import Data.List
import qualified Data.Set as S
import qualified Data.Foldable as F
pairs :: (Ord a) => S.Set a -> S.Set (a,a)
pairs s = S.fromList $ foldl two [] $ F.foldlM (\st e -> [[e]++st,st]) [] s
where two st (x:xa:[]) = (x,xa) : st
two st _ = st
Uses the power-set solution over monadic lists to build the powerset and then filter out the pairs. I can go into more detail if necessary.
So basically, if I have a (finite or infinite) list of (finite or infinite) lists of strings, is it possible to sort the list by length first and then by lexicographic order, excluding duplicates? A sample input/output would be:
Input:
[["a", "b",...], ["a", "aa", "aaa"], ["b", "bb", "bbb",...], ...]
Output:
["a", "b", "aa", "bb", "aaa", "bbb", ...]
I know that the input list is not a valid haskell expression but suppose that there is an input like that. I tried using merge algorithm but it tends to hang on the inputs that I give it. Can somebody explain and show a decent sorting function that can do this? If there isn't any function like that, can you explain why?
In case somebody didn't understand what I meant by the sorting order, I meant that shortest length strings are sorted first AND if one or more strings are of same length then they are sorted using < operator.
Thanks!
Ultimately, you can't sort an infinite list, because items at the tail of the list could percolate all the way to the front of the result, so you can't finish sorting an infinite list until you've seen the last item, but your list is infinite, so you'll never get there.
The only way that you could even try to sort an infinite list would require constraints on the inhabitants of the list. If the values of the list items comes from a well-founded set and the contents of the list are unique then you could at least make some progress in returning elements the initial elements of the list. For instance if the list was of distinct natural numbers, you could return the first 0 you see, then the first 1, etc. but you couldn't make any headway in the result until you saw 2, no matter how far down the list you went. Ultimately, if you ever skipped an element in the set because it wasn't present in the source, you'd cease to produce new output elements until you had the entire input in hand.
You can do the same thing with strings, because they are well founded, but that is only even remotely viable if you plan on returning all possible strings.
In short, if you need this, you're going about solving the problem you have in the wrong way. This isn't a tractable path to any solution you will want to use.
As yairchu noted, merging a finite number of sorted infinite lists works fine though.
In general it is impossible to sort infinite lists. Because the smallest item could be at infinite position and we must find it before we output it.
Merging infinite sorted lists is possible.
In general, merging an infinite list of sorted lists is impossible. For same reason that sorting them is.
Merging an infinite list of sorted lists, which are sorted by heads (forall i j. i < j => head (lists !! i) <= head (lists !! j)), is possible.
So I'm guessing that what you really want is to merge a sorted infinite list of sorted lists. It's even a task that makes some sense. There's even some existing code that uses it, implemented for monadic lists there - kinda ugly syntax-wise etc. So here's a version for plain lists:
mergeOnSortedHeads :: Ord b => (a -> b) -> [[a]] -> [a]
mergeOnSortedHeads _ [] = []
mergeOnSortedHeads f ([]:xs) = mergeOnSortedHeads f xs
mergeOnSortedHeads f ((x:xs):ys) =
x : mergeOnSortedHeads f (bury xs ys)
where
bury [] ks = ks
bury js [] = [js]
bury js ([]:ks) = bury js ks
bury jj#(j:js) ll#(kk#(k:ks):ls)
| f j <= f k = jj : ll
| otherwise = kk : bury jj ls
ghci> take 20 $ mergeOnSortedHeads id $ [[0,4,6], [2,3,9], [3,5..], [8]] ++ map repeat [12..]
[0,2,3,3,4,5,6,7,8,9,9,11,12,12,12,12,12,12,12,12]
btw: what do you need this for?
Well, I'm going to ignore your request for sorting infinite data.
To sort by length of the sublists, then by lexicographic order, we can do this pretty easily. Oh, and you want duplicates removed.
We'll start with a sample:
> s
[["a","b"],["a","aa","aaa"],["b","bb","bbb"]]
And then build the program incrementally.
First sorting on length (using Data.Ord.comparing to build the sort body):
> sortBy (comparing length) s
[["a","b"],["a","aa","aaa"],["b","bb","bbb"]]
Ok. That looks reasonable. So let's just concat, and sortBy length then alpha:
> sortBy (comparing length) . nub . concat $ s
["a","b","aa","bb","aaa","bbb"]
If your input is sorted. Otherwise you'll need a sligtly different body to sortBy.
Thanks to everyone for their inputs and sorry for the late reply. Turns out I was just approaching the problem in a wrong way. I was trying to do what Yairchu showed but I was using the built in function length to do the merging but length doesnt work on an infinite list for obvious reasons. Anyways, I solved my problem by sorting as I created the list on the go, not on the end result. I wonder what other languages offer infinite lists? Such a weird but useful concept.
Here is an algorithm that let you online sort:
it is not efficient, but it is lazy enough to let you goto different sort generations, even if you sort infinite lists. It is a nice gimmick, but not very usable. For example sorting the infinite list [10,9..]:
*Main> take 10 $ sortingStream [10,9..] !! 0
[9,8,7,6,5,4,3,2,1,0]
*Main> take 10 $ sortingStream [10,9..] !! 1
[8,7,6,5,4,3,2,1,0,-1]
*Main> take 10 $ sortingStream [10,9..] !! 2
[7,6,5,4,3,2,1,0,-1,-2]
*Main> take 10 $ sortingStream [10,9..] !! 3
[6,5,4,3,2,1,0,-1,-2,-3]
*Main> take 10 $ sortingStream [10,9..] !! 4
[5,4,3,2,1,0,-1,-2,-3,-4]
*Main> take 10 $ sortingStream [10,9..] !! 1000
[-991,-992,-993,-994,-995,-996,-997,-998,-999,-1000]
As you can see the sorting improves each generation. The code:
produce :: ([a] -> [a]) -> [a] -> [[a]]
produce f xs = f xs : (produce f (f xs))
sortingStream :: (Ord a) => [a] -> [[a]]
sortingStream = produce ss
ss :: (Ord a) => [a] -> [a]
ss [] = []
ss [x] = [x]
ss [x,y] | x <= y = [x,y]
| otherwise = [y,x]
ss (x:y:xs) | x <= y = x: (ss (y:xs))
| otherwise = y:(ss (x:xs))
Whether it can be done depends very much on the nature of your input data. If you can 'stop looking' for lists of a certain length when you've seen a longer one and there are only a finite number of lists of each length, then you can go through the lengths in ascending order, sort those and concatenate the results. Something like this should work:
listsUptoLength n xss = takeWhile (\xs -> length xs <= n) $ xss
listsUptoLength' n [] = []
listsUptoLength' n (xss:xsss) = case listsUptoLength n xss of
[] -> []
xss' -> xss' : listsUptoLength' n xsss
listsOfLength n xsss = concatMap (\xss -> (filter (\xs -> length xs == n) xss)) (listsUptoLength' n xsss)
sortInfinite xsss = concatMap (\n -> sort . nub $ (listsOfLength n xsss)) [0..]
f xs y = [xs ++ replicate n y | n <- [1..]]
test = [ map (\x -> [x]) ['a'..'e'], f "" 'a', f "" 'b', f "b" 'a', f "a" 'b' ] ++ [f start 'c' | start <- f "" 'a']
(The names could probably be chosen to be more illuminating :)
I'm guessing you're working with regular expressions, so I think something like this could be made to work; I repeat the request for more background!