>>=
similar to how ;
works in a usual imperative language. do
is a set of nested for loops do
produces a list [ ]
breaks out of a loop do
statement, where you have <-
assignment of variables to values from statements. The actual impertative language you are using, that is the way that "do" for a monad is interpreted, depends on how >>=
is defined for the monad. For example List []
- a do
over the List monad is a nested set of loops with possible breaks.
Maybe
- a do
over the Maybe monad is a straight-line sequence of operations which proceeds only if every step is a Just
value, and fails if a Nothing
value is computed at any step.
State
- A do
over the State monad is a straight-line sequence of transitions from the current to the next state, where the state is passed from transition to transition. The transition can produce a result that is bound to a variable to be used in later transiitons.
IO
- A do
over the IO monad is a sequence of input and output operations with the "outside" world. M
. You can think of a value of type M a
as a statement in the language M
that returns a value of type a
as result. s
is a function that takes the current state of the machine to the next state, that is next = s(current)
. So a sequence of statements s1 ;
s2 ;
s3
can be thought of as a function composition applied to the current state s3 (s2 (s1 (current))))
Here is a concrete example. If the program state is a pair (x,y), then a statement like x = x + 1
applied to (1,3) would be (x = x + 1)(1, 3) = (2, 3)
So this program on initial state (1, 3) x = x + 1;
y = y - 1;
x = x + y
becomes (x = x + y) (y = y - 1) (x = x + 1) (1, 3) =
(x = x + y) (y = y - 1) (2, 3) =
(x = x + y) (2, 2) =
(4, 2)
If you don't supply the initial state, then you can just think in terms of function composition to get that The key idea here is that we need a rule for how to compose statements, and that rule depends on our language. We also need a way to extract out information from a statement for later use.(x = x + y) (y = y - 1) (x = x + 1)
is the state transition function
\(x, y) -> (x+y, y-1)
M
(think imperative language M), then you can think of M a
as a statement in M that produces something of type a
that can be used in later statements. So the bind operator >>=
(>>=) :: M a -> ( a -> M b ) -> M b
tells you how to sequence the statement M a
followed by M b
, given that M b
might depend on the results computed by M a
. And in fact in a do
, where s2
might contain free occurrences of variable x
Now, we are still operating within a functional language, so you can only bind a variable once in its scope, so there can only be one instance of a variable on the left of ax <- s1;
s2
is the same as this outside of ado
s1 >>= (\x -> s2)
<-
. So any imperative aspects similar to updating the value of a storage location will have to be done with explicit operations in the monad. return
? You can think of return
as a function that constructs a statement in the monad that delivers the value supplied to the return. What happens next depends on the monad. right unit:m >>= return
is equivalent tom
left unit:return x >>= f
is equivalent tof x
associativity:(m >>= f) >>= g
is equivalent tom >>= (\x -> f x >>= g)
import Control.Monad.State |
|
{- |
List is a Monad with these operations: |
|
return x = [x] |
>>= :: [a] -> (a -> [b]) -> [b] |
xs >>= f = concat (map f xs) |
|
or using the concatMap = concat . map |
xs >>= f = concatMap f xs |
|
also note, since |
m >> n = m >>= \_ -> n |
then |
m >> n = concat (map (\_ -> n) m) |
which is length m copies of n |
|
-} |
|
{- with the definitions above, this -} |
|
ex1a = do |
b1 <- [1, 2] |
b2 <- ["a", "b"] |
return (b1, b2) |
|
{- produces |
[(1,"a"),(1,"b"),(2,"a"),(2,"b")] |
|
Why is this? In essence, a do over the list monad is a series of |
nested for loops, one for each variable on the lhs of <- |
|
To prove this, begin by converting to the use of >>= |
-} |
|
ex1b = |
[1, 2] >>= \b1 -> |
["a", "b"] >>= \b2 -> |
return (b1, b2) |
|
{- and then use the definition of >>= and return for lists |
|
This amounts to mapping a function over a list, and concatenating the |
results. This means the function being mapped must return a list. |
The function being mapped is itself a concat of a map over a list, so |
the inner function is returning a list. |
|
-} |
|
ex1c' = |
[1, 2] >>= \b1 -> |
concat ( map (\b2 -> [ (b1, b2) ]) ["a", "b"] ) |
|
ex1c = |
concat (map (\b1 -> |
concat ( map (\b2 -> |
[ (b1, b2) ]) |
["a", "b"] )) -- values of b2 |
[1, 2] ) -- values of b1 |
|
{- |
finally, this can also be written as a list comprehension, which is |
just syntactic sugar for the do, which is syntactic sugar for the >>= |
-} |
|
ex1d = [ (b1, b2) | b1 <- [1,2], b2 <- ["a", "b"] ] |
|
{- |
now here is something a bit unexpected. Every "statement" in the |
list monad program must generate a list. So we can insert a |
statement into a program as follows: |
-} |
|
ex1p s = do |
b1 <- [1, 2] |
-- observe what happens when you change s: [] [()] [1] [(), ()] |
s |
b2 <- ["a", "b"] |
return (b1, b2) |
|
{- you get this: |
|
*Main> ex1p [] |
[] |
|
*Main> ex1p [1] |
[(1,"a"),(1,"b"),(2,"a"),(2,"b")] |
|
*Main> ex1p [()] |
[(1,"a"),(1,"b"),(2,"a"),(2,"b")] |
|
*Main> ex1p [(),()] |
[(1,"a"),(1,"b"),(1,"a"),(1,"b"),(2,"a"),(2,"b"),(2,"a"),(2,"b")] |
|
The reason this happens is because >> is defined as |
m >> n = concat (map (\_ -> n) m) |
so |
> map (\_ -> 1) [] |
[] |
> map (\_ -> 1) [()] |
[1] |
|
and so a sequence m >> n can be selectively interrupted or continued |
or even duplicated. |
|
You also get interesting things like this, remembering that return |
just generates a statement in the monad. |
|
*Main> [1, 2, 3] >>= return [4] >>= return [5] |
[5,5,5] |
-} |
|
import Control.Monad.State |
|
{- |
List is a Monad with these operations: |
|
return x = [x] |
>>= :: [a] -> (a -> [b]) -> [b] |
xs >>= f = concat (map f xs) |
|
and it should satisfy these laws |
|
right unit: |
m >>= return |
is equivalent to |
m |
|
left unit: |
return x >>= f |
is equivalent to |
f x |
|
associativity: |
(m >>= f) >>= g |
is equivalent to |
m >>= (\x -> f x >>= g) |
-} |
|
{- with the definitions above, |
|
note: m results in a list, say [e0, e1, ...] |
|
m >>= return |
concat (map return m) |
concat [ (return e0), (return e1), ... ] |
concat [ [e0], [e1], ... ] |
[e0, e1, ...] |
m |
|
note: operations f,g below have to result in a list |
|
return x >> f |
concat (map f (return x)) |
concat (map f [x]) |
concat [f x] |
f x |
|
(m >> f) >>= g |
concat (map g (m >> f)) |
concat (map g (concat (map f m))) |
|
note: concat [ (map g L0), (map g L1), ... ] |
is the same as |
(map g (concat [ L0, L1, ... ])) |
|
m >> = (\x -> f x >>= g) |
m >> = (\x -> (concat (map g (f x)))) |
(concat (map (\x -> (concat (map g (f x)))) m)) |
(concat (map (\x -> (concat (map g (f x)))) [e0, e1, ...])) |
|
(concat [(concat (map g (f e0))), (concat (map g (f e1))), ...]) |
|
f results in a list, g results in a list for element of the |
list from f, so we can collect all the f results first, and |
then feed that list into g |
|
(concat (map g (concat (map f [e0, e1, ...])))) |
(concat (map g (concat (map f m)))) |
(concat (map g (m >>= f)))) |
(m >>= f) >>= g |
|
-} |
|
ex1a = do |
b1 <- [1, 2] |
b2 <- ["a", "b"] |
return (b1, b2) |
|
ex1b = |
[1, 2] >>= \b1 -> |
["a", "b"] >>= \b2 -> |
return (b1, b2) |
|
ex2b = |
[1, 2] >>= \b1 -> |
["a", "b"] >>= |
return >>= \b2 -> |
return (b1, b2) |
|
{- What if you have a sequence of monad operations you want to perform. |
For example |
|
*Main> map print [1, 2,3 ] |
|
<interactive>:80:1: |
No instance for (Show (IO ())) arising from a use of `print' |
Possible fix: add an instance declaration for (Show (IO ())) |
In a stmt of an interactive GHCi command: print it |
|
*Main> :t map print [1, 2, 3] |
map print [1, 2, 3] :: [IO ()] |
|
produces a sequence of IO operations, but does not actually perform them. |
|
To actually do them, i.e evaluate each action in the sequence from left |
to right, and collect the results, use the seqeunce operation: |
|
sequence :: Monad m => [m a] -> m [a] |
sequence ms = foldr k (return []) ms |
where |
k m m' = do { x <- m; xs <- m'; return (x:xs) } |
|
Examples: |
*Main> do sequence $ map print [1, 2, 3] |
1 |
2 |
3 |
[(),(),()] |
|
*Main> do sequence $ map print [1, 2, 3]; return () |
1 |
2 |
3 |
-} |
|
|
import Control.Monad.State |
{- |
|
A MonadPlus is a way of combining the results of two monad computations. |
What "combine" means depends on your goals and the particular monad. |
|
class Monad m => MonadPlus m where |
mzero :: m a |
mplus :: m a -> m a -> m a |
|
where for lists we have |
instance MonadPlus [] where |
mzero = [] |
mplus = (++) |
|
msum combines, under mplus, a list of monad computations |
msum :: MonadPlus m => [m a] -> m a |
msum = foldr mplus mzero |
|
guard blocks off a route |
guard :: MonadPlus m => Bool -> m () |
guard True = return () |
guard False = mzero |
|
so for lists we have |
guard True = [()] |
guard False = [] |
the guard command is like a break in a for loop |
-} |
|
{- |
*Main> msum $ map (\x -> [[x]]) [1, 2, 3] |
[[1],[2],[3]] |
|
*Main> msum $ map (\x -> (if (x==2) then [] else [[x]])) [1, 2, 3] |
[[1],[3]] |
|
*Main> sequence $ map (\x -> [[x]]) [1, 2, 3] |
[[[1],[2],[3]]] |
|
*Main> sequence $ map (\x -> (if (x==2) then [] else [[x]])) [1, 2, 3] |
[] |
} |
|
ex2a = do |
b1 <- [1, 2, 3, 4] |
guard (b1 /= 2) |
b2 <- ["a", "b", "c"] |
guard (b2 /= "b") |
return (b1, b2) |
|
ex2a' = do |
b1 <- [1, 2, 3, 4] |
if (b1 /= 2) |
then do |
b2 <- ["a", "b", "c"] |
if ( b2 /= "b") |
then do |
return (b1, b2) |
else do |
[] |
else do |
[] |
|
{- so why should this work? For lists, it has to do with how a map of a |
function that ignores its argument works: |
|
Note: |
> map (\_ -> 1) [] |
[] |
> map (\_ -> 1) [()] |
[1] |
|
So the guard returns the [] when false, so further maps are aborted. |
-} |
|
ex2b = |
[1, 2, 3, 4] >>= \b1 -> |
guard(b1 /= 2) >> |
["a", "b", "c"] >>= \b2 -> |
guard (b2 /= "b") >> |
return (b1, b2) |
|
|
ex2c' = |
concatMap (\b1 -> |
concatMap (\_ -> |
concatMap (\b2 -> |
concatMap (\_ -> [ (b1, b2) ]) |
(guard (b2 /= "b")) ) |
["a", "b", "c"] ) -- values of b2 |
(guard (b1 /= 2))) -- values of guard: [] or [()] |
[1, 2, 3, 4] -- values of b1 |
|
ex2c = |
concat (map (\b1 -> |
concat (map (\_ -> |
concat ( map (\b2 -> |
concat (map (\_ -> |
[ (b1, b2) ]) |
(guard (b2 /= "b")) )) |
["a", "b", "c"] )) -- values of b2 |
(guard (b1 /= 2)) )) -- values of guard: [] or [()] |
[1, 2, 3, 4] ) -- values of b1 |
|
|
{- Another example taken from |
http://en.wikibooks.org/wiki/Haskell/MonadPlus |
-} |
|
pythags = do |
z <- [1..] |
x <- [1..z] |
y <- [x..z] |
guard (x^2 + y^2 == z^2) |
return (x, y, z) |
|
pythags1 = |
[1..] >>= \z -> |
[1..z] >>= \x -> |
[x..z] >>= \y -> |
guard (x^2 + y^2 == z^2) >>= \_ -> |
return (x, y, z) |
|
pythags2 = |
let ret x y z = [(x, y, z)] |
gd z x y = concatMap (\_ -> ret x y z) (guard $ x^2 + y^2 == z^2) |
doY z x = concatMap (gd z x) [x..z] |
doX z = concatMap (doY z ) [1..z] |
doZ = concatMap (doX ) [1..] |
in doZ |
|
|
{- |
Given a list of tuples (Name, [ Values]) where values is a |
[ String ] that gives all the allowable values of name, generate all |
possible assignments. |
-} |
|
type Name = String |
type Value = String |
|
{- generates all possible assignments -} |
|
-- monad version |
genA :: [ (Name, [Value]) ] -> [ [ (Name, Value) ] ] |
-- genA [ ] = [ [] ] |
genA [ ] = [ ] |
genA ( (n, vals):rest ) = do |
v <- vals |
o <- genA rest |
return ( (n, v) : o ) |
|
-- list comprehension version |
genB :: [ (Name, [Value]) ] -> [ [ (Name, Value) ] ] |
genB [] = [ [] ] |
genB ( (n, vals):rest ) = [ |
(n, v) : o | |
v <- vals, |
o <- genB rest |
] |
|
a1 = [ ("a", ["0", "1"]), ("b", ["x", "y", "z"]) ] |
import Control.Monad.State |
|
{- A tree containing nodes of type a -} |
data Tree a = Leaf a | Node a (Tree a) (Tree a) deriving (Show, Eq) |
|
testTree = |
Node "A" (Node "B" (Leaf "C" ) (Node "D" (Leaf "E") (Leaf "F"))) (Leaf "G") |
|
{- |
Given a tree t, dfsLabelTree generates an isomorphic Tree Int that |
has values that are the positions that the leaf or node is visited in |
a depth first searhc. |
|
The algorithm is expressed as a state monad, in which the state of |
the monad is the next dfs number to use. The initial number is 0 |
|
Note how |
*Main> :t dfsTree testTree |
dfsTree testTree :: State Int (Tree Int) |
|
is a state monad, that is, a state machine program that is waiting to |
be run on an initial state. |
|
-} |
|
dfsTree :: Tree a -> State Int (Tree Int) |
|
dfsTree (Leaf x) = do |
n <- get |
put (n+1) |
return (Leaf n) |
|
dfsTree (Node x t1 t2) = do |
n <- get |
put (n+1) |
dt1 <- dfsTree t1 |
dt2 <- dfsTree t2 |
return (Node n dt1 dt2) |
|
{- your result depends on how you run the program resulting from dfsTree |
|
prog = dfsTree testTree |
prog :: State Int (Tree Int) |
|
runState prog 0 |
evalState prog 0 |
execState prog 0 |
|
result and the state |
*Main> runState prog 0 |
(Node 0 (Node 1 (Leaf 2) (Node 3 (Leaf 4) (Leaf 5))) (Leaf 6),7) |
|
result only |
*Main> evalState prog 0 |
Node 0 (Node 1 (Leaf 2) (Node 3 (Leaf 4) (Leaf 5))) (Leaf 6) |
|
state only |
*Main> execState prog 0 |
7 |
-} |
|
dfsLabelTree :: Tree a -> Tree Int |
dfsLabelTree t = evalState (dfsTree t) 0 |
|
{- if you want to label two trees, you can do this -} |
p1 = do |
t1 <- dfsTree testTree; |
t2 <- dfsTree testTree; |
return (t1,t2) |
|
{- and get this output |
*Main> evalState p1 0 |
(Node 0 (Node 1 (Leaf 2) (Node 3 (Leaf 4) (Leaf 5))) (Leaf 6), |
Node 7 (Node 8 (Leaf 9) (Node 10 (Leaf 11) (Leaf 12))) (Leaf 13)) |
-} |
|
|
{- |
suppose you want to label the nodes in the tree with their path from |
the root. |
-} |
|
pathTree :: Tree a -> State [ Int ] (Tree [ Int ]) |
pathTree (Leaf x) = do |
pRoot <- get |
return (Leaf pRoot) |
|
pathTree (Node x t1 t2) = do |
pRoot <- get |
put (pRoot ++ [0]) |
pt1 <- pathTree t1 |
put (pRoot ++ [1]) |
pt2 <- pathTree t1 |
return (Node pRoot pt1 pt2) |
|
path :: Tree a -> Tree [ Int ] |
path t = evalState (pathTree t) [] |
{- Constraint satisfaction solving |
|
Modifed from |
|
from http://www.haskell.org/haskellwiki/All_About_Monads#StateT_with_List |
|
Background modules: |
|
http://www.haskell.org/haskellwiki/State_Monad |
http://www.cis.upenn.edu/~bcpierce/courses/advprog/resources/base/Control.Monad.State.html |
|
http://hackage.haskell.org/package/base-4.6.0.1/docs/Control-Monad.html |
https://hackage.haskell.org/package/mtl-2.0.1.0/docs/Control-Monad-State.html |
|
|
Here is a interesting example of combining the StateT monad with |
the List monad to produce a monad for stateful nondeterministic |
computations. |
|
We will apply this powerful monad combination to the task of |
solving constraint satisfaction problems (in this case, a logic |
problem). The idea behind it is to have a number of variables that |
can take on different values and a number of predicates involving |
those variables that must be satisfied. The current variable |
assignments and the predicates make up the state of the |
computation, and the non-deterministic nature of the List monad |
allows us to easily test all combinations of variable assignments. |
|
We start by laying the groundwork we will need to represent the |
logic problem, a simple predicate language: |
|
-} |
import Data.Maybe |
import Control.Monad |
import Control.Monad.State |
import Text.Printf |
|
|
{- pretty printer for models, display using putStrLn |
Assumes that each model has variables in the same order in the |
list -} |
|
showModels :: [Variables] -> String |
showModels models = |
show models |
|
{- |
A language to express logic problems |
|
Var are variables, named with strings |
Value are values of variables, values are strings |
Predicates are functions with values are true or false |
-} |
|
type Name = String |
type Value = String |
type VarType = String |
data Predicate = |
Is Name Value -- var has specific value |
| Equal Name Name -- vars have same (unspecified) value |
| And Predicate Predicate -- both are true |
| Or Predicate Predicate -- at least one is true |
| Not Predicate -- it is not true |
deriving (Eq, Show) |
|
type Variables = [(Name,Value)] |
|
-- derived predicates |
|
-- test for a variable NOT equaling a value |
isNot :: Name -> Value -> Predicate |
isNot var value = Not (Is var value) |
|
-- if a is true, then b must also be true |
implies :: Predicate -> Predicate -> Predicate |
implies a b = Not (a `And` (Not b)) |
|
-- iff |
iff :: Predicate -> Predicate -> Predicate |
iff a b = (a `implies` b) `And` (b `implies` a) |
|
-- exclusive or |
orElse :: Predicate -> Predicate -> Predicate |
orElse a b = (a `And` (Not b)) `Or` ((Not a) `And` b) |
|
{- |
Check a predicate with the given variable bindings. |
An unbound variable causes a Nothing return value. |
otherwise you get Just True or Just False. |
|
lookup takes a list of name-value tuples, and returns the |
corresponding value Just v, or Nothing if name is not present. |
|
lookup :: Eq a => a -> [(a, b)] -> Maybe b |
-} |
|
check :: Predicate -> Variables -> Maybe Bool |
check (Is var value) vars = do val <- lookup var vars |
return (val == value) |
check (Equal v1 v2) vars = do val1 <- lookup v1 vars |
val2 <- lookup v2 vars |
return (val1 == val2) |
|
{- |
lifting takes a function (a1 -> r) and turns it into a function |
on monads (m a1 -> m r) |
|
liftM :: Monad m => (a1 -> r) -> m a1 -> m r |
liftM2 :: Monad m => (a1 -> a2 -> r) -> m a1 -> m a2 -> m rSource |
|
Promote a function to a monad, scanning the monadic arguments from |
left to right. For example, |
|
liftM2 (+) [0,1] [0,2] = [0,2,1,3] |
liftM2 (+) (Just 1) Nothing = Nothing |
-} |
|
check (And p1 p2) vars = liftM2 (&&) (check p1 vars) (check p2 vars) |
check (Or p1 p2) vars = liftM2 (||) (check p1 vars) (check p2 vars) |
check (Not p) vars = liftM (not) (check p vars) |
|
{- |
Machinery for representing and solving constraint satisfaction |
problems. This is where we will define our combined monad. |
|
ProblemState is the type of our logic problem. Note the use of |
named fields where: |
|
vars is the environment, binding variables to values. |
constraints are a collection of predicates that must be satisfied |
in order for an environment to be a solution. |
-} |
|
data ProblemState = PS {vars::Variables, constraints::[Predicate]} |
|
-- this is our monad type for non-determinstic computations with state |
type NDS a = StateT ProblemState [] a |
|
-- lookup a variable in the environment. |
getVar :: Name -> NDS (Maybe Value) |
getVar v = do vs <- gets vars |
return $ lookup v vs |
|
-- set a variable in the environment. |
setVar :: Name -> Value -> NDS () |
setVar v x = do st <- get |
vs' <- return $ filter ((v/=).fst) (vars st) |
put $ st {vars=(v,x):vs'} |
|
{- |
Check if the variable assignments satisfy all of the predicates. |
The partial argument determines the value used when a predicate |
returns Nothing because some variable it uses is not set. Setting |
this to True allows us to accept partial solutions, then we can use |
a value of False at the end to signify that all solutions should be |
complete. |
|
Note: maybe :: b -> (a -> b) -> Maybe a -> b |
(maybe x id) (Just y) = y |
(maybe x id) Nothing = x |
is the way you yield a default value |
|
Note: (maybe x id) (Just y) = y |
(maybe x id) Nothing = x |
is the way you yield a default value |
-} |
|
isConsistent :: Bool -> NDS Bool |
isConsistent partial = do cs <- gets constraints |
vs <- gets vars |
let results = map (\p->check p vs) cs |
return $ and (map (maybe partial id) results) |
|
{- |
Return only the variable bindings that are complete consistent |
solutions. getFinalVars :: NDS Variables getFinalVars = do c <- |
isConsistent False guard c gets vars |
-} |
|
{- |
|
Get the first solution to the problem, by evaluating the solver |
computation with an initial problem state and then returning the |
first solution in the result list, or Nothing if there was no |
solution. |
|
-} |
|
getSolution :: NDS a -> ProblemState -> Maybe a |
getSolution c i = listToMaybe (evalStateT c i) |
|
{- |
Get a list of all possible solutions to the problem by evaluating |
the solver computation with an initial problem state. |
-} |
|
getAllSolutions :: NDS a -> ProblemState -> [a] |
getAllSolutions c i = evalStateT c i |
|
said :: Name -> Predicate -> Predicate |
said v p = ((v `Is` "knight") `implies` p) `And` |
((v `Is` "knave") `implies` (Not p)) `And` |
((v `Is` "normal") `implies` (p `Or` (Not p))) |
|
{- lying is saying something is true when it isn't, or not saying |
anything -} |
|
lied :: Name -> Predicate -> Predicate |
lied v p = (v `said` p) `And` (Not p) |
|
{- telling truth is saying something is true when it is, or not |
saying anything. -} |
|
toldTruth :: Name -> Predicate -> Predicate |
toldTruth v p = (v `said` p) `implies` p |
|
-- Test consistency over all allowed settings of the variable. |
|
tryAllValues :: (Name,VarType) -> NDS () |
tryAllValues (var,vtype) = do |
-- try all the possible values of the type of var using |
-- lookupType :: VarType -> [ Value ] |
-- note the use of monad sum to put plus between each possibility |
msum $ map (setVar var) (lookupType vtype) |
c <- isConsistent True |
guard c |
|
{- All that remains to be done is to define the puzzle in the |
predicate language and get a solution that satisfies all of the |
predicates: -} |
|
|
{- Define the problem, try all of the variable assignments and |
print a solution. solve takes a set of constraints and returns a |
list of all the variable assignments that satisfy the constraints. |
|
Each call to tryAllValues will fork the solution space, assigning |
the named variable to be "knight" in one fork and "knave" in the |
other. The forks which produce inconsistent variable assignments |
are eliminated (using the guard function). The call to getFinalVars |
applies guard again to eliminate inconsistent variable assignments |
and returns the remaining assignments as the value of the |
computation. |
|
if only asking for a few solutions, lazyness should stop further |
search |
|
-} |
|
solve :: [(Name,VarType)] -> [Predicate] -> [Variables] |
solve varOrder constraints = do |
let variables = [] |
problem = PS variables constraints |
|
(getAllSolutions |
(do |
sequence $ map tryAllValues varOrder |
getFinalVars) |
problem) |
|
{- |
|
We introduce types for the variables using a function lookupType |
that for each type gives the possible values it can have. |
|
The we identify the variables in the problem and their types, along |
with the order in which they should be searched, with the earliest |
variables being the most frequently changed (in model checking the |
order in which variables are examined can dramatically affect |
solution time). |
|
-} |
|
lookupType :: VarType -> [ Value ] |
-- a generic person |
lookupType "Per" = ["knight", "knave", "normal" ] |
-- just knights and knaves |
lookupType "KK" = ["knight", "knave" ] |
-- paths are good or bad |
lookupType "Path" = ["good", "bad" ] |
|
varOrder = [ |
("B", "KK"), |
("A", "KK"), |
("R", "Path"), |
("L", "Path") |
] |
|
pSolve :: [Predicate] -> IO () |
pSolve constraints = putStrLn (showModels $ solve varOrder constraints) |
|
{- Puzzles from notes -} |
|
{- there are 16 possible configurations that the situation can take |
when there are just knights and knaves. |
-} |
|
pU = [ |
Not (Equal "A" "B"), |
Not (Equal "L" "R") |
] |
|
{- |
Let's simplify by assuming just knights and knaves, only one of |
each type, and a good and bad path. The solver tells you that there |
are only these possible states in the world: |
|
L R A B |
bad good knave knight |
good bad knave knight |
bad good knight knave |
good bad knight knave |
|
So, what questions could you ask that will identify the proper path |
to take when you don't know which of the 4 states above actually |
holds? |
|
For example, suppose you ask A: would you say |
"If B says 'the left is good', then the left is indeed good" |
if A answers yes, then this is true: |
L R A B |
good bad knight knave |
and we can encode the yes answer to the question as: |
("A" `said` ( |
(("B" `said` ("L" `Is` "good")) `implies` ("L" `Is` "good")) |
)), |
and we have one case of 4 covered when this problem has only one |
model. |
|
But if A answers no, then this is true of the possible models: |
L R A B |
bad good knave knight |
good bad knave knight |
bad good knight knave |
|
and you still don't know which path to take. So you need to ask |
another question. So you will have to make a series of calls to |
the solver, with different questions until you can narrow down the |
proper answer. That decision process can of course be encoded in a |
function to which you supply the answers to the questions. |
|
This question, if you ask it of A (or B, symmetrically) |
-} |
|
-- if this has a model, then go left |
p0 = [ |
-- A says "If B says 'the left is good', then the left is good" |
("A" `said` ( |
(("B" `said` ("L" `Is` "good")) `implies` ("L" `Is` "good")) |
)), |
-- consistency check |
("L" `Is` "good"), |
|
-- one is knight, one is knave |
Not (Equal "A" "B"), |
-- one is good, one is bad |
Not (Equal "L" "R") |
] |
|
-- if A said no, if no model, then must be this |
p1 = [ |
-- negation of previous, A didn't say that, |
Not ("A" `said` ( |
(("B" `said` ("L" `Is` "good")) `implies` ("L" `Is` "good")) |
)), |
("A" `Is` "knight"), |
|
-- one is knight, one is knave |
Not (Equal "A" "B"), |
-- one is good, one is bad |
Not (Equal "L" "R") |
] |
|
14. Week 12 - Mar 31 CMPUT 325 Schedule / Version 2.31 2014-04-04 |