Files
linguist/samples/Frege/Sudoku.fr
mmhelloworld bc923bb6b1 Add Frege language
What is Frege?
-------------
Frege is a non-strict, pure functional programming language in the spirit of Haskell for the JVM.
It enjoys a strong static type system with type inference.
Higher rank types are supported, though type annotations are required for that.

Frege programs are compiled to Java and run in a JVM.
Existing Java Classes and Methods can be used seamlessly from Frege.

The Frege programming language is named after and in honor of Gottlob Frege.

Project State:
-------------
The compiler, an Eclipse plugin and a provisional version of the documentation can be downloaded
from here https://github.com/Frege/frege/releases.

The REPL can be downloaded from here
https://github.com/Frege/frege-repl/releases.

An online REPL is running here
http://try.frege-lang.org/.

Examples:
--------
1) Command Line Clock: https://github.com/Frege/frege/blob/master/examples/CommandLineClock.fr
2) Brainfuck: https://github.com/Frege/frege/blob/master/examples/Brainfuck.fr
3) Concurrency: https://github.com/Frege/frege/blob/master/examples/Concurrent.fr
4) Sudoku: https://github.com/Frege/frege/blob/master/examples/Sudoku.fr
5) Java Swing examples: https://github.com/Frege/frege/blob/master/examples/SwingExamples.fr
2013-12-10 23:36:05 -05:00

562 lines
23 KiB
Plaintext

package examples.Sudoku where
import Data.TreeMap (Tree, keys)
import Data.List as DL hiding (find, union)
type Element = Int -- 1,2,3,4,5,6,7,8,9
type Zelle = [Element] -- set of candidates
type Position = Int -- 0..80
type Feld = (Position, Zelle)
type Brett = [Feld]
--- data type for assumptions and conclusions
data Assumption =
!ISNOT Position Element
| !IS Position Element
derive Eq Assumption
derive Ord Assumption
instance Show Assumption where
show (IS p e) = pname p ++ "=" ++ e.show
show (ISNOT p e) = pname p ++ "/" ++ e.show
showcs cs = joined " " (map Assumption.show cs)
elements :: [Element] -- all possible elements
elements = [1 .. 9]
{-
a b c d e f g h i
0 1 2 | 3 4 5 | 6 7 8 1
9 10 11 |12 13 14 |15 16 17 2
18 19 20 |21 22 23 |24 25 26 3
---------|---------|--------
27 28 29 |30 31 32 |33 34 35 4
36 37 38 |39 40 41 |42 43 44 5
45 46 47 |48 49 50 |51 52 53 6
---------|---------|--------
54 55 56 |57 58 59 |60 61 62 7
63 64 65 |66 67 68 |69 70 71 8
72 73 74 |75 76 77 |78 79 80 9
-}
positions :: [Position] -- all possible positions
positions = [0..80]
rowstarts :: [Position] -- all positions where a row is starting
rowstarts = [0,9,18,27,36,45,54,63,72]
colstarts :: [Position] -- all positions where a column is starting
colstarts = [0,1,2,3,4,5,6,7,8]
boxstarts :: [Position] -- all positions where a box is starting
boxstarts = [0,3,6,27,30,33,54,57,60]
boxmuster :: [Position] -- pattern for a box, by adding upper left position results in real box
boxmuster = [0,1,2,9,10,11,18,19,20]
--- extract field for position
getf :: Brett -> Position -> Feld
getf (f:fs) p
| fst f == p = f
| otherwise = getf fs p
getf [] p = (p,[])
--- extract cell for position
getc :: Brett -> Position -> Zelle
getc b p = snd (getf b p)
--- compute the list of all positions that belong to the same row as a given position
row :: Position -> [Position]
row p = [z..(z+8)] where z = (p `quot` 9) * 9
--- compute the list of all positions that belong to the same col as a given position
col :: Position -> [Position]
col p = map (c+) rowstarts where c = p `mod` 9
--- compute the list of all positions that belong to the same box as a given position
box :: Position -> [Position]
box p = map (z+) boxmuster where
ri = p `div` 27 * 27 -- 0, 27 or 54, depending on row
ci = p `mod` 9 -- column index 0..8, 0,1,2 is left, 3,4,5 is middle, 6,7,8 is right
cs = ci `div` 3 * 3 -- 0, 3 or 6
z = ri + cs
--- check if candidate set has exactly one member, i.e. field has been solved
single :: Zelle -> Bool
single [_] = true
single _ = false
unsolved :: Zelle -> Bool
unsolved [_] = false
unsolved _ = true
-- list of rows, cols, boxes
allrows = map row rowstarts
allcols = map col colstarts
allboxs = map box boxstarts
allrcb = zip (repeat "row") allrows
++ zip (repeat "col") allcols
++ zip (repeat "box") allboxs
containers :: [(Position -> [Position], String)]
containers = [(row, "row"), (col, "col"), (box, "box")]
-- ----------------- PRINTING ------------------------------------
-- printable coordinate of field, upper left is a1, lower right is i9
pname p = packed [chr (ord 'a' + p `mod` 9), chr (ord '1' + p `div` 9)]
-- print board
printb b = mapM_ p1line allrows >> println ""
where
p1line row = do
print (joined "" (map pfld line))
where line = map (getc b) row
-- print field (brief)
-- ? = no candidate
-- 5 = field is 5
-- . = some candidates
pfld [] = "?"
pfld [x] = show x
pfld zs = "0"
-- print initial/final board
result msg b = do
println ("Result: " ++ msg)
print ("Board: ")
printb b
return b
res012 b = case concatMap (getc b) [0,1,2] of
[a,b,c] -> a*100+b*10+c
_ -> 9999999
-- -------------------------- BOARD ALTERATION ACTIONS ---------------------------------
-- print a message about what is done to the board and return the new board
turnoff1 :: Position -> Zelle -> Brett -> IO Brett
turnoff1 i off b
| single nc = do
-- print (pname i)
-- print ": set to "
-- print (head nc)
-- println " (naked single)"
return newb
| otherwise = return newb
where
cell = getc b i
nc = filter (`notElem` off) cell
newb = (i, nc) : [ f | f <- b, fst f != i ]
turnoff :: Int -> Zelle -> String -> Brett -> IO Brett
turnoff i off msg b = do
-- print (pname i)
-- print ": set to "
-- print nc
-- print " by clearing "
-- print off
-- print " "
-- println msg
return newb
where
cell = getc b i
nc = filter (`notElem` off) cell
newb = (i, nc) : [ f | f <- b, fst f != i ]
turnoffh ps off msg b = foldM toh b ps
where
toh b p = turnoff p off msg b
setto :: Position -> Element -> String -> Brett -> IO Brett
setto i n cname b = do
-- print (pname i)
-- print ": set to "
-- print n
-- print " (hidden single in "
-- print cname
-- println ")"
return newb
where
nf = [n]
newb = (i, nf) : [ f | f <- b, fst f != i ]
-- ----------------------------- SOLVING STRATEGIES ---------------------------------------------
-- reduce candidate sets that contains numbers already in same row, col or box
-- This finds (and logs) NAKED SINGLEs in passing.
reduce b = [ turnoff1 p sss | (p,cell) <- b, -- for each field
unsolved cell, -- with more than 1 candidate
-- single fields in containers that are candidates of that field
sss = [ s | (rcb, _) <- containers, [s] <- map (getc b) (rcb p), s `elem` cell],
sss != [] ] -- collect field index, elements to remove from candidate set
-- look for a number that appears in exactly 1 candidate set of a container
-- this number can go in no other place (HIDDEN SINGLE)
hiddenSingle b = [ setto i n cname | -- select index, number, containername
(cname, rcb) <- allrcb, -- FOR rcb IN allrcb
n <- elements, -- FOR n IN elements
fs = filter (unsolved • snd) (map (getf b) rcb),
occurs = filter ((n `elem`) • snd) fs,
length occurs == 1,
(i, _) <- occurs ]
-- look for NAKED PAIRS, TRIPLES, QUADS
nakedPair n b = [ turnoff p t ("(naked tuple in " ++ nm ++ ")") | -- SELECT pos, tuple, name
-- n <- [2,3,4], // FOR n IN [2,3,4]
(nm, rcb) <- allrcb, -- FOR rcb IN containers
fs = map (getf b) rcb, -- let fs = fields for rcb positions
u = (fold union [] . filter unsolved . map snd) fs, -- let u = union of non single candidates
t <- n `outof` u, -- FOR t IN n-tuples
hit = (filter ((`subset` t) . snd) . filter (unsolved . snd)) fs,
length hit == n,
(p, cell) <- fs,
p `notElem` map fst hit,
any (`elem` cell) t
]
-- look for HIDDEN PAIRS, TRIPLES or QUADS
hiddenPair n b = [ turnoff p off ("(hidden " ++ show t ++ " in " ++ nm ++ ")") | -- SELECT pos, tuple, name
-- n <- [2,3,4], // FOR n IN [2,3,4]
(nm, rcb) <- allrcb, -- FOR rcb IN containers
fs = map (getf b) rcb, -- let fs = fields for rcb positions
u = (fold union [] . filter ((>1) . length) . map snd) fs, -- let u = union of non single candidates
t <- n `outof` u, -- FOR t IN n-tuples
hit = (filter (any ( `elem` t) . snd) . filter (unsolved . snd)) fs,
length hit == n,
off = (fold union [] . map snd) hit `minus` t,
off != [],
(p, cell) <- hit,
! (cell `subset` t)
]
a `subset` b = all (`elem` b) a
a `union` b = uniq (sort (a ++ b))
a `minus` b = filter (`notElem` b) a
a `common` b = filter (`elem` b) a
n `outof` as
| length as < n = []
| [] <- as = []
| 1 >= n = map (:[]) as
| (a:bs) <- as = map (a:) ((n-1) `outof` bs) ++ (n `outof` bs)
| otherwise = undefined -- cannot happen because either as is empty or not
same f a b = b `elem` f a
intersectionlist = [(allboxs, row, "box/row intersection"), (allboxs, col, "box/col intersection"),
(allrows ++ allcols, box, "line/box intersection")]
intersections b = [
turnoff pos [c] reason | -- SELECT position, candidate, reson
(from, container, reason) <- intersectionlist,
rcb <- from,
fs = (filter (unsolved . snd) . map (getf b)) rcb, -- fs = fields in from with more than 1 candidate
c <- (fold union [] • map snd) fs, -- FOR c IN union of candidates
cpos = (map fst • filter ((c `elem`) • snd)) fs, -- cpos = positions where c occurs
cpos != [], -- WHERE cpos is not empty
all (same container (head cpos)) (tail cpos), -- WHERE all positions are in the intersection
-- we can remove all occurences of c that are in container, but not in from
(pos, cell) <- map (getf b) (container (head cpos)),
c `elem` cell,
pos `notElem` rcb ]
-- look for an XY Wing
-- - there exists a cell A with candidates X and Y
-- - there exists a cell B with candidates X and Z that shares a container with A
-- - there exists a cell C with candidates Y and Z that shares a container with A
-- reasoning
-- - if A is X, B will be Z
-- - if A is Y, C will be Z
-- - since A will indeed be X or Y -> B or C will be Z
-- - thus, no cell that can see B and C can be Z
xyWing board = [ turnoff p [z] ("xy wing " ++ pname b ++ " " ++ pname c ++ " because of " ++ pname a) |
(a, [x,y]) <- board, -- there exists a cell a with candidates x and y
rcba = map (getf board) (row a ++ col a ++ box a), -- rcba = all fields that share a container with a
(b, [b1, b2]) <- rcba,
b != a,
b1 == x && b2 != y || b2 == x && b1 != y, -- there exists a cell B with candidates x and z
z = if b1 == x then b2 else b1,
(c, [c1, c2]) <- rcba,
c != a, c!= b,
c1 == y && c2 == z || c1 == z && c2 == y, -- there exists a cell C with candidates y and z
ps = (uniq . sort) ((row b ++ col b ++ box b) `common` (row c ++ col c ++ box c)),
-- remove z in ps
(p, cs) <- map (getf board) ps,
p != b, p != c,
z `elem` cs ]
-- look for a N-Fish (2: X-Wing, 3: Swordfish, 4: Jellyfish)
-- When all candidates for a particular digit in N rows are located
-- in only N columns, we can eliminate all candidates from those N columns
-- which are not located on those N rows
fish n board = fish "row" allrows row col ++ fish "col" allcols col row where
fishname 2 = "X-Wing"
fishname 3 = "Swordfish"
fishname 4 = "Jellyfish"
fishname _ = "unknown fish"
fish nm allrows row col = [ turnoff p [x] (fishname n ++ " in " ++ nm ++ " " ++ show (map (pname . head) rset)) |
rset <- n `outof` allrows, -- take n rows (or cols)
x <- elements, -- look for certain number
rflds = map (filter ((>1) . length . snd) . map (getf board)) rset, -- unsolved fields in the rowset
colss = (map (map (head . col . fst) . filter ((x `elem`) . snd)) rflds), -- where x occurs in candidates
all ((>1) . length) colss, -- x must appear in at least 2 cols
cols = fold union [] colss,
length cols == n,
cstart <- cols,
(p, cell) <- map (getf board) (col cstart),
x `elem` cell,
all (p `notElem`) rset]
-- compute immediate consequences of an assumption of the form (p `IS` e) or (p `ISNOT` e)
conseq board (IS p e) = uniq (sort ([ p `ISNOT` x | x <- getc board p, x != e ] ++
[ a `ISNOT` e |
(a,cs) <- map (getf board) (row p ++ col p ++ box p),
a != p,
e `elem` cs
]))
conseq board (ISNOT p e) = uniq (sort ([ p `IS` x | cs = getc board p, length cs == 2, x <- cs, x != e ] ++
[ a `IS` e |
cp <- [row p, box p, col p],
as = (filter ((e `elem`) . getc board) . filter (p!=)) cp,
length as == 1,
a = head as
]))
-- check if two assumptions contradict each other
contradicts (IS a x) (IS b y) = a==b && x!=y
contradicts (IS a x) (ISNOT b y) = a==b && x==y
contradicts (ISNOT a x) (IS b y) = a==b && x==y
contradicts (ISNOT _ _) (ISNOT _ _) = false
-- get the Position of an Assumption
aPos (IS p _) = p
aPos (ISNOT p _) = p
-- get List of elements that must be turned off when assumption is true/false
toClear board true (IS p x) = filter (x!=) (getc board p)
toClear board false (IS p x) = [x]
toClear board true (ISNOT p x) = [x]
toClear board false (ISNOT p x) = filter (x!=) (getc board p)
-- look for assumptions whose implications contradict themself
chain board paths = [ solution a (head cs) (reverse cs) |
(a, css) <- paths,
cs <- take 1 [ cs | cs <- css, contradicts a (head cs) ]
]
where
solution a c cs = turnoff (aPos a) (toClear board false a) reason where
reason = "Assumption " ++ show a ++ " implies " ++ show c ++ "\n\t"
++ showcs cs ++ "\n\t"
++ "Therefore, " ++ show a ++ " must be false."
-- look for an assumption that yields to contradictory implications
-- this assumption must be false
chainContra board paths = [ solution a (reverse pro) (reverse contra) |
(a, css) <- paths, -- FOR ALL assumptions "a" with list of conlusions "css"
(pro, contra) <- take 1 [ (pro, contra) |
pro <- (uniqBy (using head) . sortBy (comparing head)) css, -- FOR ALL conslusion chains "pro"
c = head pro, -- LET "c" BE the final conclusion
contra <- take 1 (filter ((contradicts c) . head) css) -- THE FIRST conclusion that contradicts c
]
]
where
solution a pro con = turnoff (aPos a) (toClear board false a) reason where
reason = ("assumption " ++ show a ++ " leads to contradictory conclusions\n\t"
++ showcs pro ++ "\n\t" ++ showcs con)
-- look for a common implication c of some assumptions ai, where at least 1 ai is true
-- so that (a0 OR a1 OR a2 OR ...) IMPLIES c
-- For all cells pi in same container that have x as candidate, we can construct (p0==x OR p1==x OR ... OR pi==x)
-- For a cell p with candidates ci, we can construct (p==c0 OR p==c1)
cellRegionChain board paths = [ solution b as (map head os) |
as <- cellas ++ regionas, -- one of as must be true
iss = filter ((`elem` as) . fst) paths, -- the implications for as
(a, ass) <- take 1 iss, -- implications for first assumption
fs <- (uniqBy (using head) . sortBy (comparing head)) ass,
b = head fs, -- final conclusions of first assumption
os = [fs] : map (take 1 . filter ((b==) . head) . snd) (tail iss), -- look for implications with same conclusion
all ([]!=) os]
where
cellas = [ map (p `IS`) candidates | (p, candidates@(_:_:_)) <- board ]
regionas = [ map (`IS` e) ps |
region <- map (map (getf board)) (allrows ++ allcols ++ allboxs),
e <- elements,
ps = map fst (filter ((e `elem`) . snd) region),
length ps > 1 ]
solution b as oss = turnoff (aPos b) (toClear board true b) reason where
reason = "all of the assumptions " ++ joined ", " (map show as) ++ " imply " ++ show b ++ "\n\t"
++ joined "\n\t" (map (showcs . reverse) oss) ++ "\n\t"
++ "One of them must be true, so " ++ show b ++ " must be true."
{-
Wir brauchen für einige Funktionen eine Datenstruktur wie
[ (Assumption, [[Assumption]]) ]
d.i. eine Liste von möglichen Annahmen samt aller Schlußketten.
Idealerweise sollte die Schlußkette in umgekehrter Reihenfolge vorliegen,
dann kann man einfach finden:
- Annahmen, die zum Selbstwiderspruch führen.
- alles, was aus einer bestimmten Annahme folgt (map (map head) [[a]])
-...
-}
--- Liste aller Annahmen für ein bestimmtes Brett
assumptions :: Brett -> [Assumption]
assumptions board = [ a |
(p, cs) <- board,
!(single cs),
a <- map (ISNOT p) cs ++ map (IS p) cs ]
consequences :: Brett -> [Assumption] -> [[Assumption]]
consequences board as = map (conseq board) as
acstree :: Brett -> Tree Assumption [Assumption]
acstree board = Tree.fromList (zip as cs)
where
as = assumptions board
cs = consequences board as
-- bypass maybe on tree lookup
find :: Tree Assumption [Assumption] -> Assumption -> [Assumption]
find t a
| Just cs <- t.lookup a = cs
| otherwise = error ("no consequences for " ++ show a)
-- for performance resons, we confine ourselves to implication chains of length 20 per assumption
mkPaths :: Tree Assumption [Assumption] -> [ (Assumption, [[Assumption]]) ]
mkPaths acst = map impl (keys acst) -- {[a1], [a2], [a3] ]
where
-- [Assumption] -> [(a, [chains, ordered by length]
impl a = (a, impls [[a]])
impls ns = (take 1000 • concat • takeUntil null • iterate expandchain) ns
-- expandchain :: [[Assumption]] -> [[Assumption]]
expandchain css = [ (n:a:as) |
(a : as) <- css, -- list of assumptions
n <- find acst a, -- consequences of a
n `notElem` as -- avoid loops
]
-- uni (a:as) = a : uni (filter ((head a !=) • head) as)
-- uni [] = empty
-- empty = []
-- ------------------ SOLVE A SUDOKU --------------------------
-- Apply all available strategies until nothing changes anymore
-- Strategy functions are supposed to return a list of
-- functions, which, when applied to a board, give a changed board.
-- When a strategy does not find anything to alter,
-- it returns [], and the next strategy can be tried.
solve b
| all (single . snd) b = result "Solved" b
| any (([]==) . snd) b = result "not solvable" b
| res@(_:_) <- reduce b = apply b res >>=solve -- compute smallest candidate sets
-- comment "candidate sets are up to date" = ()
| res@(_:_) <- hiddenSingle b = apply b res >>= solve -- find HIDDEN SINGLES
-- comment "no more hidden singles" = ()
| res@(_:_) <- intersections b = apply b res >>= solve -- find locked candidates
-- comment "no more intersections" = ()
| res@(_:_) <- nakedPair 2 b = apply b res >>= solve -- find NAKED PAIRS, TRIPLES or QUADRUPELS
-- comment "no more naked pairs" = ()
| res@(_:_) <- hiddenPair 2 b = apply b res >>= solve -- find HIDDEN PAIRS, TRIPLES or QUADRUPELS
-- comment "no more hidden pairs" = ()
-- res@(_:_) <- nakedPair 3 b = apply b res >>= solve // find NAKED PAIRS, TRIPLES or QUADRUPELS
-- | comment "no more naked triples" = ()
-- res@(_:_) <- hiddenPair 3 b = apply b res >>= solve // find HIDDEN PAIRS, TRIPLES or QUADRUPELS
-- | comment "no more hidden triples" = ()
-- res@(_:_) <- nakedPair 4 b = apply b res >>=solve // find NAKED PAIRS, TRIPLES or QUADRUPELS
-- | comment "no more naked quadruples" = ()
-- res@(_:_) <- hiddenPair 4 b = apply b res >>=solve // find HIDDEN PAIRS, TRIPLES or QUADRUPELS
-- | comment "no more hidden quadruples" = ()
| res@(_:_) <- xyWing b = apply b res >>=solve -- find XY WINGS
-- comment "no more xy wings" = ()
| res@(_:_) <- fish 2 b = apply b res >>=solve -- find 2-FISH
-- comment "no more x-wings" = ()
-- res@(_:_) <- fish 3 b = apply b res >>=solve // find 3-FISH
-- | comment "no more swordfish" = ()
-- res@(_:_) <- fish 4 b = apply b res >>=solve // find 4-FISH
-- | comment "no more jellyfish" = ()
-- | comment pcomment = ()
| res@(_:_) <- chain b paths = apply b (take 9 res) >>= solve -- find forcing chains
| res@(_:_) <- cellRegionChain b paths = apply b (take 9 res) >>= solve -- find common conclusion for true assumption
| res@(_:_) <- chainContra b paths = apply b (take 9 res) >>= solve -- find assumptions that allow to infer both a and !a
-- comment "consistent conclusions only" = ()
| otherwise = result "ambiguous" b
where
apply brd fs = foldM (\b\f -> f b) brd fs
paths = mkPaths (acstree b)
-- pcomment = show (length paths) ++ " assumptions with " ++ show (fold (+) 0 (map (length <~ snd) paths))
-- ++ " implication chains"
-- comment com = do stderr << com << "\n" for false
-- log com = do stderr << com << "\n" for true
--- turn a string into a row
mkrow :: String -> [Zelle]
mkrow s = mkrow1 xs
where
xs = s ++ "---------" -- make sure at least 9 elements
mkrow1 xs = (take 9 • filter ([]!=) • map f • unpacked) xs
f x | x >= '1' && x <= '9' = [ord x - ord '0']
| x == ' ' = [] -- ignored
| otherwise = elements
main ["-h"] = main []
main ["-help"] = main []
main [] = do
mapM_ stderr.println [
"usage: java Sudoku file ...",
" java Sudoku position",
"where position is a 81 char string consisting of digits",
"One can get such a string by going to",
"http://www.sudokuoftheday.com/pages/s-o-t-d.php",
"Right click on the puzzle and open it in new tab",
"Copy the 81 digits from the URL in the address field of your browser.",
"",
"There is also a file with hard sudokus in examples/top95.txt\n"]
return ()
main [s@#^[0-9\W]{81}$#] = solve board >> return ()
where
board = zip positions felder
felder = decode s
main files = forM_ files sudoku
where
sudoku file = do
br <- openReader file
lines <- BufferedReader.getLines br
bs <- process lines
ss <- mapM (\b -> print "Puzzle: " >> printb b >> solve b) bs
println ("Euler: " ++ show (sum (map res012 ss)))
return ()
-- "--3-" => [1..9, 1..9, [3], 1..9]
decode s = map candi (unpacked s) where
candi c | c >= '1' && c <= '9' = [(ord c - ord '0')]
| otherwise = elements
process [] = return []
process (s:ss)
| length s == 81 = consider b1
| length s == 9,
length acht == 8,
all ((9==) • length) acht = consider b2
| otherwise = do
stderr.println ("skipped line: " ++ s)
process ss
where
acht = take 8 ss
neun = fold (++) "" (s:acht)
b1 = zip positions (decode s)
b2 = zip positions (decode neun)
consider b = do
-- print "Puzzle: "
-- printb b
bs <- process ss
return (b:bs)