User talk:Mimoso
Revision as of 22:13, 5 April 2011 by Mimoso (talk | contribs) (→Othello (Reversi). Manuel Hernández, April 2011.)
Othello (Reversi). Manuel Hernández, April 2011.
import Random
import List
data Element = O | X | E | L deriving (Eq,Show)
data ArbolG = T Board [ArbolG] deriving Show
type Board = [Element]
data ArbolG1 = T1 (Element,Integer,Board,Int,Int) [ArbolG1] deriving Show
--Bug? Hugs does not accept "vectors" from size > 5 to show
-- T1 (Int,Int,Int,Int,Int) is not shown...
data ArbolG3 = T3 (Element, -- Player
Integer, -- Mov
Board, -- Position
(Int,Int,Int) -- (num Xs, num Os, num Movs)
) [ArbolG3] deriving Show
data ArbolG4 = T4 (Element, -- Player
[Integer], -- Mov
(Int,Int,Int) -- (num Xs, num Os, num Movs)
) [ArbolG4] deriving Show
data MvVal = MvVal {mov::Integer, xs :: Int, os :: Int} deriving Show
lcoords = concat [[(x+1)..(x+8)]| x<-[10,20..80]]
coords pos = zipWith (\x y -> (x,y)) lcoords pos
expandx pos ((ini,X),(cand, O)) =
let (delta,av) = (cand-ini,cand+delta)
es = findCoords av (coords pos)
in if null es then [] else
let h = snd (head es) in
case h of
X -> []
E -> [(cand,O),(av,E)]
O -> let rest = expandoX pos delta (av, O) in
if null rest then [] else (cand, O):rest
expandoX pos delta (av, O) =
let av1 = av + delta
es = findCoords av1 (coords pos)
in if null es then [] else
let h = snd (head es) in
case h of
X -> []
E -> [(av, O),(av1, E)]
O -> let rest= expandoX pos delta (av1, O) in
if null rest then [] else (av, O):rest
expando pos ((ini,O),(cand, X)) =
let
(delta,av) = (cand-ini,cand+delta)
es = findCoords av (coords pos)
in
if null es then [] else
let h = snd (head es) in
case h of
O -> []
E -> [(cand, X),(av, E)]
X -> let rest = expandxO pos delta (av, X) in
if null rest then [] else (cand, X):rest
expandxO pos delta (av, X) = let
av1 = av + delta
es = findCoords av1 (coords pos) in
if null es then [] else
let h = snd (head es) in
case h of
O -> []
E -> [(av, X),(av1, E)]
X -> let rest= expandxO pos delta (av1, X) in
if null rest then [] else (av, X):rest
allNum player pos = (nub . sort)
(map (fst) (validMoves player pos))
validMoves player pos = map last (allMoves player pos)
allMoves player pos = filter (/=[]) (movs player pos)
movs player pos = if player==X then
map (expandx pos) (concat (onlyNBos X pos))
else
map (expando pos) (concat (onlyNBos O pos))
candidates (init, ls) = zip (repeat init) ls
findF e coors = (filter (\x -> (snd x)==e) coors)
findCoords m coors = (filter (\x -> (fst x) == m) coors)
neighbs e1 pos (n, e) = ((n, e),(only e1
(concat
[findCoords x (coords pos)| x<-dirs n])))
-- To find the neighborhoods...
onlyNBos player pos = map (candidates . (neighbs (change player) pos))
(findF player (coords pos))
only e ls = filter (\x -> (snd x)==e) ls
dirs n | elem n ([22..27]++[32..37]++[42..47]++[52..57]++[62..67]++[72..77])
= map (+n) [-11,-10,-9,-1,1,9,10,11]
| elem n [12..17] = map (+n) [-1,1,9,10,11]
| elem n [82..87] = map (+n) [-1,1,-9,-10,-11]
| elem n [21,31..71] = map (+n) [-10,-9,1,10,11]
| elem n [28,38..78] = map (+n) [-11,-10,-1,9,10]
| elem n [11] = [12,22,21]
| elem n [88] = [87,77,78]
| elem n [81] = [71,72,82]
| elem n [18] = [17,27,28]
numTTT::Int
numTTT= 64
part8 [] = []
part8 (a:bs) = (take 8 (a:bs)):(part8 (drop 8 (a:bs)))
posIni::[Element]
posIni = (take 24 (repeat E))++[E,E,E,X,O,E,E,E,
E,E,E,O,X,E,E,E]++(take 24 (repeat E))
to a = snd (head (filter (\x -> a==fst x) (zip lcoords [1..64])))
maxV (MvVal n1 a1 b1) (MvVal n2 a2 b2) | a1<=a2 = MvVal n2 a2 b2
| otherwise = MvVal n1 a1 b1
findMaxV ls = foldr (maxV) (MvVal 0 (-1000) 0) ls
showB [] = ""
showB (a:bs) = (show a)++"\n"++(showB bs)
showBoard pos = putStr (" _ _ _ _ _ _ _ _ \n"
++[xchange x|x<-(showB (part8 pos))])
xchange x | x==',' = '|'
| x=='E' = '_'
| otherwise = x
count player pos = length (filter (==player) pos)
validCoord player pos = nub (strip (validMoves player pos))
strip [] = []
strip ((n, x):ls) = fst (n, x):strip ls
------------------------------begin wrt X---------------------
allVBasic player g pos =
T1 (player,g,pos,count X pos,count O pos) ls
where
ls = [T1 (mMoveVirtual player k pos) [] | k <- (allNum player pos)]
barrer player (T1 (p,m,pos,n1,n2) []) = allVBasic player m pos
barrer player (T1 (p,m,pos,n1,n2) (c:cs)) =
T1 (player,m,pos,count X pos,count O pos)
(map (barrer (change player)) (c:cs))
genTree player pos n = take n (iterate (barrer player) (T1 (X,0,pos,2,2) []))
--Realmente sólo se utilizan jugador=X y posición=pos
mMoveVirtual player n pos =
let newpos = (applyMove
(apply player (nub (concat
(filter (\x->fst (head x)==n)
(map reverse
(allMoves player pos)))))) pos)
in (player, n, newpos, count X newpos, count O newpos)
--Dato: (player, movement, position, howmanyX, howmanyO)
sortby [] = []
sortby ((a1,b1):bs) = sortby [x | x<- bs, snd x < b1]++[(a1,b1)]++
sortby [x | x<- bs, snd x >= b1]
newPos player n pos =
applyMove
(apply player (nub (concat
(filter (\x->fst (head x)==n)
(map reverse
(allMoves player pos)))))) pos
mMvVirtual player n pos =
let newpos = (applyMove
(apply player (nub (concat
(filter (\x->fst (head x)==n)
(map reverse
(allMoves player pos)))))) pos)
in MvVal n (count X newpos) (count O newpos)
transTree (T1 (n,mov,pos,xs,os) []) = T4 (n,[mov],(xs-os,os,nm)) []
where
nm = length (allNum n pos)
transTree (T1 (n,mov,pos,xs,os) (a:bs)) = T4 (n,[mov],(xs,os,nm)) ls
where
ls = (map transTree (a:bs))
nm = length (allNum n pos)
--Simple minimax: alpha-beta pruning, see below...
minimax (T4 (n,[mov],(numberOfxs,numberOfos,nm)) [])
|nm==0 = ([mov],-70) -- -70 o 70?
minimax (T4 (n,[mov],(numberOfxs,numberOfos,nm)) [])
|nm>0 = ([mov],-numberOfxs) -- positive, it is "greedy"
minimax (T4 (n,[mov],(numberOfxs,numberOfos,nm)) (a:bs)) = (ms,n)
where
ls = (negP (minList' (map minimax (a:bs))))
(mvT,val) = ls
ms = (mov:mvT) -- ++[mov]
n = val
app ls (ms,t) = (ls++ms,t)
negP (a,b) = (a,-b)
bestMv player pos n =
minimax (transTree (last (genTree player pos n)))
minList ls = foldr (min) (1000) ls
minP (a1,b1) (a2,b2) = if b1<b2 then (a1,b1) else (a2,b2)
minList' ls = foldr (minP) ([],1000) ls
mMove player n pos = showBoard (applyMove
(apply player (nub (concat
(filter (\x->fst (head x)==n)
(map reverse
(allMoves player pos)))))) pos)
---------------------------end wrt X-----------------------------
apply player [] = []
apply player ((n,e):ls) = (n,player):apply player ls
applyMove [] pos = pos
applyMove ((n,player):ls) pos = applyMove ls (sustn player (to n) pos)
sustn :: (Num a, Ord a) => b -> a -> [b] -> [b]
sustn a 1 (c:cs) = (a:cs)
sustn a n (c:cs) | n>1 = c:(sustn a (n-1) cs)
change X = O
change O = X
-- The strength of playing depends on the eval function.
calcMov :: Board -> IO()
calcMov pos = do
--Report winner..., missing
let
bm = head (tail (fst (bestMv X pos 3))) -- Empty list..., missing
newpos1 = newPos X bm pos
mMove X bm pos
putStr $ show bm
putStr $ "\n"
putStrLn $ "Black: " ++ (show (count X newpos1))
putStrLn $ "White: " ++ (show (count O newpos1))
putStrLn $ (show (allNum O newpos1))++"\n"
putStr "Your move: "
input <- getLine
let square = (read input) :: Integer
-- putStr (show square)
let
newpos2 = newPos O square newpos1
mMove O square newpos1
putStrLn $ "Black: "++ (show (count X newpos2))
putStrLn $ "White: "++ (show (count O newpos2))
calcMov (newPos O square newpos2)
main = calcMov posIni
--Variants:
- Three players (or more)
- Scattering pieces over the board
- Boards with obstacles (squares, or diamonds, for example)
- Boards with distinct geometrical forms.
- Boards with distinct square geometry.
- Random static token
- Factor number betrayed
- ¿Dimensions? I am thinking... 3D Othello.
- Special turns (like to put a token over an arbitrary square)
- Hexa "squares"
The alpha-beta pruning process is:
bmx alpha beta (T4 (n,[mov],(numberOfxs,numberOfos,nm)) []) =
([mov],max alpha (min numberOfxs beta))
bmx alpha beta (T4 (n,ls,(numberOfxs,numberOfos,nm)) (b:bs)) =
(ls++mvT,val)
where
(mvT,val) = cmx alpha beta (b:bs)
-- ...following Bird and Wadler...
cmx alpha beta [] = ([],alpha)
cmx alpha beta (b:bs)
| alpha'== beta = ([],alpha')
| otherwise = (ls++mvs,val)
where
(ls,val) = cmx alpha' beta bs
(mvs,alpha') = negP (bmx (-beta) (-alpha) b)
-- Use the following "best move" instead of bestMv
bestMv' player pos level =
bmx (-70) 70 (transTree (last (genTree player pos level)))
I suggest to use OpenGL to make an interface.
For future editing: Wikipedia