letmaik / monadiccp Goto Github PK

View Code? Open in Web Editor NEW

26.0 7.0 10.0 293 KB

Monadic Constraint Programming framework

Home Page: http://people.cs.kuleuven.be/~tom.schrijvers/MCP/

License: Other

Haskell 100.00%

haskell constraint-programming monads

monadiccp's Introduction

monadiccp

Acknowledgments

Thanks to Chris Mears for a patch to get the package to compile against GHC 7.4.1.
Thanks to Nicholas Tung for a patch to get the package to compile against GHC 7.2.2.

monadiccp's People

Contributors

Stargazers

Watchers

Forkers

jtiko notae franklinchen silky hanslub42 stefaj jwaldmann j-mueller saraswat

monadiccp's Issues

Exception on unnecessary inequality

-- throws Exception: Cannot process EGBoolValue BoolConst True
model :: FDModel
model = exists $ \col -> do
  [a,b] <- colList col 2
  xsum col @= 2
  a @>= 0
  b @>= 1
  2 @<= 1 -- this causes the exception
  return col

Extract Gecode runtime into a separate library

Cabal flags should not change the API of a library. For monadiccp, this rule is broken with the RuntimeGecode flag which makes available additional functionality.

The Gecode runtime modules should be packaged as a separate library which has a dependency on monadiccp, e.g. called monadiccp-gecode.

EDIT: See https://hackage.haskell.org/package/monadiccp-gecode.

Cannot process EGMod

cabal test monadiccp-vuza for 9da3818 gives

Test suite monadiccp-vuza: RUNNING...
monadiccp-vuza: Cannot process EGMod
CallStack (from HasCallStack):
  error, called at src/Control/CP/FD/FD.hs:978:28 in monadiccp-0.7.7-inplace:Control.CP.FD.FD

indeed the Mod operation is missing here: https://github.com/letmaik/monadiccp/blob/master/src/Control/CP/FD/OvertonFD/OvertonFD.hs#L300

refutable patterns require MonadFail

(with ghc-9.0.1-alpha1) when compiling

model :: ExampleModel ()
model _ = exists $ \col -> do
  [x1,x2,x3,x4,x5,x6,x7,x8,x9,x10] <- colList col 10
  ...

I am seeing

examples/Olympic.hs:40:3: error:
    • Could not deduce (MonadFail m)
        arising from a do statement 
        with the failable pattern ‘[x1,  ...
      from the context: (Show (Control.CP.FD.FD.FDIntTerm s), FDSolver s,
                         Control.CP.SearchTree.MonadTree m,
                         Control.CP.SearchTree.TreeSolver m ~ FDInstance s)
...

the suggested fix seems impossible (there is no m)

      Possible fix:
        add (MonadFail m) to the context of
          the type signature for:
            model :: ExampleModel ()

But still, someone needs to get a MonadFail instance.

Exception when using "or"

-- throws Exception: Cannot process EGLess True
model :: FDModel
model = exists $ \col -> do
  [x1,x2,x3,x4] <- colList col 4
  allin col (cte 0,cte 5)
  x1 @<= x2
  x3 @>= x4
  x3 @>= x2 @|| x4 @<= x1 -- this causes the exception
  return col

I haven't tested it in other situations, this might be a corner case.

Hi. I was searching for an Haskell API to Gecode, and found this project. I made some obvious changes so that it can be compiled with a recent ghc (8.2). But I found that there is no Gecode. E.g., it's commented out here:

src/Control/CP/FD/Solvers.hs:-- import Control.CP.FD.Gecode.CodegenSolver

Depend on official Monatron library

Since monadiccp 0.7.2, for some reason the Monatron library was directly included as source. This should be reverted.

Add straight-forward introduction example

At the moment, the existing examples all rely on the Control.CP.FD.Example module which is OK for running the examples but on the other side it is quite hard to understand and to use for creating own code with monadiccp.

The two files MonadicCpExample.hs and MonadicCpHelper.hs from the adp-multi-monadiccp project which use the OvertonFD solver might be a good starting point.

Build failure with ghc 7.8.2

The latest version of monadiccp fails to build with ghc 7.8.2 with the error message below:

[21 of 63] Compiling Control.Search.Generator ( Control/Search/Generator.hs, dist/build/Control/Search/Generator.o )

Control/Search/Generator.hs:732:70:
    Overlapping instances for Monad (t' ((:>) t2 IdT m))
      arising from a use of `mapE'
    Matching instances:
      instance [overlap ok] (Monad m, MonadT t) => Monad (t m)
        -- Defined at Control/Monatron/AutoInstances.hs:11:10
      instance Monad m => Monad (Control.Applicative.WrappedMonad m)
        -- Defined in `Control.Applicative'
    (The choice depends on the instantiation of `t2, t', m'
     To pick the first instance above, use IncoherentInstances
     when compiling the other instance declarations)
    In the first argument of `liftM', namely `(mapE (mmap L . runL))'
    In the first argument of `(.)', namely
      `liftM (mapE (mmap L . runL))'
    In the `mapper' field of a record

Control/Search/Generator.hs:741:73:
    Overlapping instances for Monad (t' ((:>) t2 (t1 :> t3) m))
      arising from a use of `mapE'
    Matching instances:
      instance [overlap ok] (Monad m, MonadT t) => Monad (t m)
        -- Defined at Control/Monatron/AutoInstances.hs:11:10
      instance Monad m => Monad (Control.Applicative.WrappedMonad m)
        -- Defined in `Control.Applicative'
    (The choice depends on the instantiation of `t2, t1, t3, t', m'
     To pick the first instance above, use IncoherentInstances
     when compiling the other instance declarations)
    In the first argument of `liftM', namely `(mapE (mmap L . runL))'
    In the first argument of `(.)', namely
      `liftM (mapE (mmap L . runL))'
    In the `mapper' field of a record

Control/Search/Generator.hs:752:44:
    Overlapping instances for Monad (t' ((:>) t0 (t1 :> t2) m))
      arising from a use of `mapE'
    Matching instances:
      instance [overlap ok] (Monad m, MonadT t) => Monad (t m)
        -- Defined at Control/Monatron/AutoInstances.hs:11:10
      instance Monad m => Monad (Control.Applicative.WrappedMonad m)
        -- Defined in `Control.Applicative'
    (The choice depends on the instantiation of `t1, t2, t0, t', m'
     To pick the first instance above, use IncoherentInstances
     when compiling the other instance declarations)
    In the first argument of `liftM', namely `(mapE (mmap L . runL))'
    In the first argument of `(.)', namely
      `liftM (mapE (mmap L . runL))'
    In the `mapper' field of a record

Control/Search/Generator.hs:752:50:
    Overlapping instances for Monad (t0 ((:>) t1 t2 m))
      arising from a use of `mmap'
    Matching instances:
      instance [overlap ok] (Monad m, MonadT t) => Monad (t m)
        -- Defined at Control/Monatron/AutoInstances.hs:11:10
      instance Monad m => Monad (Control.Applicative.WrappedMonad m)
        -- Defined in `Control.Applicative'
    (The choice depends on the instantiation of `t1, t2, t0, m'
     To pick the first instance above, use IncoherentInstances
     when compiling the other instance declarations)
    In the first argument of `(.)', namely `mmap L'
    In the first argument of `mapE', namely `(mmap L . runL)'
    In the first argument of `liftM', namely `(mapE (mmap L . runL))'

Control/Search/Generator.hs:752:82:
    Overlapping instances for Monad (t0 ((:>) t1 t2 m))
      arising from a use of `mmap'
    Matching instances:
      instance [overlap ok] (Monad m, MonadT t) => Monad (t m)
        -- Defined at Control/Monatron/AutoInstances.hs:11:10
      instance Monad m => Monad (Control.Applicative.WrappedMonad m)
        -- Defined in `Control.Applicative'
    (The choice depends on the instantiation of `t1, t2, t0, m'
     To pick the first instance above, use IncoherentInstances
     when compiling the other instance declarations)
    In the second argument of `(.)', namely `mmap runL'
    In the first argument of `mapE', namely `(L . mmap runL)'
    In the second argument of `(.)', namely `mapE (L . mmap runL)'

build with ghc-8.10.2

for 0.7.7 branch, with ghc-8.10.2 and stackage nightly-2020-10-31

monadiccp       > [ 7 of 63] Compiling Control.CP.SearchTree
monadiccp       >      
monadiccp       > /tmp/stack-3aa44fdabb3e7b1a/monadiccp-0.7.7/src/Control/CP/SearchTree.hs:203:10: error:
monadiccp       >     • Illegal nested constraint ‘Solver (TreeSolver m)’
monadiccp       >       (Use UndecidableInstances to permit this)
monadiccp       >     • In the instance declaration for ‘MonadTree (ContT r m)’
monadiccp       >     |
monadiccp       > 203 | instance (MonadTree m, Solver (TreeSolver m)) => MonadTree (ContT r m) where
monadiccp       >     |          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Language.CPP.* ?

I'm shocked to discover that this package contains a library for C++ AST and pretty-printing (... to String) https://hackage.haskell.org/package/monadiccp-0.7.7/docs/src/Language.CPP.Syntax.AST.html#CPPFile ("based on https://github.com/visq/language-c")

Well if it works, it works ..

Divergence When Using @/=

This might qualify more as a question than a bug report. I tried writing the following model, which creates two variables ranging over {1,2} and constrains them with @/= so that it is not the case that (a,b) = (1,2). Then it branches on the values of the first variable.
When I evaluate solve_not_12, the program diverges -- when compiled with GHC 7.6.3 -O2, it doesn't finish and allocates up to at least 700MB.
However, if I evaluate solve_not_12', which states the constraint in a slightly different way, it succeeds and returns (10,[[1,1]]).
Is this a bug or am I something wrong? I'm using monadiccp-0.7.6 compiled without any special flags (i.e. just cabal install monadiccp).

not_12 :: Tree (FDInstance OvertonFD) [Int]
not_12 =
    exist 2 $ \[a,b] -> do
      a @: (cte 1, cte 2)
      b @: (cte 1, cte 2)
      a @/= b - 1
      a @= 1 \/ a @= 2
      assignments [a,b]
solve_not_12 = solve dfs fs not_12

not_12' :: Tree (FDInstance OvertonFD) [Int]
not_12' =
    exist 2 $ \[a,b] -> do
      a @: (cte 1, cte 2)
      b @: (cte 1, cte 2)
      a + 1 @/= b
      a @= 1 \/ a @= 2
      assignments [a,b]
solve_not_12' = solve dfs fs not_12'

main = withArgs ["overton_run"] $ example_sat_main_void model

This "overton_run" seems to be the only solver available? https://github.com/letmaik/monadiccp/blob/master/src/Control/CP/FD/Example.hs#L79

When I try to run other examples, I am getting a variety of errors, e.g.,

stack exec -- runghc examples/Domino.hs overton_run
Domino.hs: Prelude.head: empty list

I can guess that it needs a command line argument, but then

stack exec -- runghc examples/Domino.hs overton_run 3
Domino.hs: constant collections not yet supported by overton interface

I think it would be good to

convert tests so that they also work without arguments (if they need some input,
then there should be a default) (it is impossible to specify test arguments in a cabal file)
mark those tests that are supposed to work (with the given back-end), and run them in CI