sbv-13.0: Data/SBV/Core/Symbolic.hs
-----------------------------------------------------------------------------
-- |
-- Module : Data.SBV.Core.Symbolic
-- Copyright : (c) Levent Erkok
-- License : BSD3
-- Maintainer: erkokl@gmail.com
-- Stability : experimental
--
-- Symbolic values
-----------------------------------------------------------------------------
{-# LANGUAGE BangPatterns #-}
{-# LANGUAGE CPP #-}
{-# LANGUAGE DefaultSignatures #-}
{-# LANGUAGE DeriveAnyClass #-}
{-# LANGUAGE DeriveDataTypeable #-}
{-# LANGUAGE DeriveFunctor #-}
{-# LANGUAGE DeriveGeneric #-}
{-# LANGUAGE DerivingStrategies #-}
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE GeneralizedNewtypeDeriving #-}
{-# LANGUAGE MultiParamTypeClasses #-}
{-# LANGUAGE NamedFieldPuns #-}
{-# LANGUAGE OverloadedStrings #-}
{-# LANGUAGE Rank2Types #-}
{-# LANGUAGE ScopedTypeVariables #-}
{-# LANGUAGE StandaloneDeriving #-}
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE UndecidableInstances #-}
{-# LANGUAGE ViewPatterns #-}
{-# OPTIONS_GHC -Wall -Werror -fno-warn-orphans #-}
module Data.SBV.Core.Symbolic
( NodeId(..)
, SV(..), swKind, trueSV, falseSV, contextOfSV
, Op(..), PBOp(..), OvOp(..), FPOp(..), NROp(..), StrOp(..), RegExOp(..), SeqOp(..), SetOp(..), SpecialRelOp(..), ADTOp(..)
, RegExp(..), regExpToSMTString, SMTLambda(..)
, Quantifier(..), needsExistentials, SBVContext(..), checkCompatibleContext, VarContext(..)
, SBVType(..), svUninterpreted, svUninterpretedNamedArgs, newUninterpreted, prefixNameToUnique
, SVal(..)
, svMkSymVar, sWordN, sWordN_, sIntN, sIntN_
, svToSV, svToSymSV, forceSVArg
, SBVExpr(..), newExpr, isCodeGenMode, isSafetyCheckingIStage, isRunIStage, isSetupIStage
, Cached, cache, uncache, modifyState, modifyIncState
, NamedSymVar(..), Name, UserInputs, Inputs(..), getSV, swNodeId, namedNodeId
, addInternInput, addUserInput
, getUserName', getUserName
, lookupInput , getSValPathCondition, extendSValPathCondition
, getTableIndex, sObserve
, SBVPgm(..), MonadSymbolic(..), SymbolicT, Symbolic, runSymbolic, mkNewState, runSymbolicInState, State(..), SMTDef(..), withNewIncState, IncState(..), incrementInternalCounter, incrementFreshNameCounter
, inSMTMode, SBVRunMode(..), IStage(..), Result(..), ResultInp(..), UICodeKind(..), UIName(..)
, registerKind, registerLabel, recordObservable
, addAssertion, addNewSMTOption, imposeConstraint, internalConstraint, newInternalVariable, lambdaVar, quantVar
, SMTLibPgm(..), SMTLibVersion(..), smtLibVersionExtension
, SolverCapabilities(..)
, extractSymbolicSimulationState, CnstMap
, OptimizeStyle(..), Objective(..), Penalty(..), objectiveName, addSValOptGoal
, MonadQuery(..), QueryT(..), Query, QueryState(..), QueryContext(..)
, SMTScript(..), Solver(..), SMTSolver(..), SMTResult(..), SMTModel(..), SMTConfig(..), TPOptions(..), SMTEngine, isEmptyModel
, validationRequested, outputSVal, ProgInfo(..), mustIgnoreVar, getRootState
) where
import Control.DeepSeq (NFData(..))
import Control.Monad (when, unless)
import Control.Monad.Except (MonadError, ExceptT)
import Control.Monad.Reader (MonadReader(..), ReaderT, runReaderT,
mapReaderT)
import Control.Monad.State.Lazy (MonadState)
import Control.Monad.Trans (MonadIO(liftIO), MonadTrans(lift))
import Control.Monad.Trans.Maybe (MaybeT)
import Control.Monad.Writer.Strict (MonadWriter)
import Data.IORef (IORef, newIORef, readIORef)
import Data.List (intercalate, sortBy, isPrefixOf)
import Data.Maybe (fromMaybe)
import Data.String (IsString(fromString))
import Data.Time (getCurrentTime, UTCTime)
import Data.Int (Int64)
import GHC.Stack
import GHC.Stack.Types
import GHC.Generics (Generic)
import qualified Control.Exception as C
import qualified Control.Monad.State.Lazy as LS
import qualified Control.Monad.State.Strict as SS
import qualified Control.Monad.Writer.Lazy as LW
import qualified Control.Monad.Writer.Strict as SW
import qualified Data.IORef as R (modifyIORef')
import qualified Data.Generics as G (Data(..))
import qualified Data.Generics.Uniplate.Data as G
import qualified Data.IntMap.Strict as IMap (IntMap, empty, lookup, insertWith)
import qualified Data.Map.Strict as Map (Map, empty, toList, lookup, insert, size, notMember)
import qualified Data.Set as Set (Set, empty, toList, insert, member)
import qualified Data.Foldable as F (toList)
import qualified Data.Sequence as S (Seq, empty, (|>), (<|), lookup, elemIndexL)
import qualified Data.Text as T
import System.Mem.StableName
import System.Random
import Data.SBV.Core.Kind
import Data.SBV.Core.Concrete
import Data.SBV.SMT.SMTLibNames
import Data.SBV.Utils.TDiff (Timing)
import Data.SBV.Utils.Lib (stringToQFS, checkObservableName, barify)
import Data.SBV.Utils.Numeric (RoundingMode)
import Data.Containers.ListUtils (nubOrd)
import Data.SBV.Control.Types
#if MIN_VERSION_base(4,11,0)
import Control.Monad.Fail as Fail
#endif
-- | Context identifier. 0 is reserved global context
newtype SBVContext = SBVContext Int64 deriving (Eq, Ord, G.Data, Show)
instance NFData SBVContext where
rnf (SBVContext i) = i `seq` ()
-- | Global context
globalSBVContext :: SBVContext
globalSBVContext = SBVContext 0
-- | Generate context. We make sure it isn't 0, i.e., the global context
-- The "hope" here is that each time we call this we get a different context number.
-- A random number doesn't necessarily have to do that, but I think the pseudo-generator
-- has a large enough period for this to go through OK.
genSBVContext :: IO SBVContext
genSBVContext = do ctx <- SBVContext <$> randomIO
if ctx == globalSBVContext -- unlikely, but possible
then genSBVContext
else pure ctx
-- | A symbolic node id
newtype NodeId = NodeId { getId :: (SBVContext, Maybe Int, Int) } -- Lambda-level, and node-id
deriving (Ord, G.Data)
-- Equality is pair-wise, except we accommodate for negative node-id; which is reserved for true/false
instance Eq NodeId where
NodeId n1@(_, _, i) == NodeId n2@(_, _, j)
| i < 0 && j < 0
= i == j
| True
= n1 == n2
-- | A symbolic word, tracking its kind and node representing it
data SV = SV !Kind !NodeId
deriving G.Data
-- | Which context are we using this var at?
contextOfSV :: SV -> SBVContext
contextOfSV (SV _ (NodeId (c, _, _))) = c
-- | For equality, we merely use the lambda-level/node-id
instance Eq SV where
SV _ n1 == SV _ n2 = n1 == n2
-- | Again, simply use the lambda-level/node-id for ordering
instance Ord SV where
SV _ n1 `compare` SV _ n2 = n1 `compare` n2
instance HasKind SV where
kindOf (SV k _) = k
instance Show SV where
show (SV _ (NodeId (_, l, n))) = case n of
-2 -> "false"
-1 -> "true"
_ -> prefix ++ 's' : show n
where prefix = case l of
Nothing -> "arg" -- top-level lambda
Just 0 -> ""
Just i -> 'l' : show i ++ "_"
-- | Kind of a symbolic word.
swKind :: SV -> Kind
swKind (SV k _) = k
-- | retrieve the node id of a symbolic word
swNodeId :: SV -> NodeId
swNodeId (SV _ nid) = nid
-- | Forcing an argument; this is a necessary evil to make sure all the arguments
-- to an uninterpreted function are evaluated before called; the semantics of uinterpreted
-- functions is necessarily strict; deviating from Haskell's
forceSVArg :: SV -> IO ()
forceSVArg (SV k n) = k `seq` n `seq` return ()
-- | Constant False as an t'SV'. Note that this value always occupies slot -2 and level 0.
falseSV :: SV
falseSV = SV KBool $ NodeId (globalSBVContext, Just 0, -2)
-- | Constant True as an t'SV'. Note that this value always occupies slot -1 and level 0.
trueSV :: SV
trueSV = SV KBool $ NodeId (globalSBVContext, Just 0, -1)
-- | Symbolic operations
data Op = Plus
| Times
| Minus
| UNeg
| Abs
| Quot
| Rem
| Equal Bool -- ^ If bool is True then this is strong (i.e., object equality). Matters for floats or structures containing them.
| Implies
| NotEqual
| LessThan
| GreaterThan
| LessEq
| GreaterEq
| Ite
| And
| Or
| XOr
| Not
| Shl
| Shr
| Rol Int
| Ror Int
| Divides Integer -- divides k n is True if k divides n. k must be > 0 constant
| Extract Int Int -- Extract i j: extract bits i to j. Least significant bit is 0 (big-endian)
| Join -- Concat two words to form a bigger one, in the order given
| ZeroExtend Int
| SignExtend Int
| LkUp (Int, Kind, Kind, Int) !SV !SV -- (table-index, arg-type, res-type, length of the table) index out-of-bounds-value
| KindCast Kind Kind
| Uninterpreted String
| QuantifiedBool String -- When we generate a forall/exists (nested etc.) boolean value. NB. This used to be "QuantifiedBool [Op] String", keeping track of Ops. That turned out to cause memory leaks. So avoid that.
| SpecialRelOp Kind SpecialRelOp -- Generate the equality to the internal operation
| Label String -- Essentially no-op; useful for code generation to emit comments.
| IEEEFP FPOp -- Floating-point ops, categorized separately
| NonLinear NROp -- Non-linear ops (mostly trigonometric), categorized separately
| OverflowOp OvOp -- Overflow-ops, categorized separately
| PseudoBoolean PBOp -- Pseudo-boolean ops, categorized separately
| RegExOp RegExOp -- RegEx operations, categorized separately
| StrOp StrOp -- String ops, categorized separately
| SeqOp SeqOp -- Sequence ops, categorized separately
| SetOp SetOp -- Set operations, categorized separately
| TupleConstructor Int -- Construct an n-tuple
| TupleAccess Int Int -- Access element i of an n-tuple; second argument is n
| RationalConstructor -- Construct a rational. Note that there's no access to numerator or denumerator, since we cannot store rationals in canonical form
| ADTOp ADTOp -- ADT access/construction/testing
| ArrayLambda SMTLambda -- An array value, created from a lambda
| ReadArray -- Reading an array value
| WriteArray -- Writing to an array
deriving (Eq, Ord, Generic, G.Data, NFData)
-- | ADT operations
data ADTOp = ADTConstructor String Kind -- Construct an ADT. Kind is the kind of the resulting ADT
| ADTTester String Kind -- Check if top-level constructor matches. Kind is the kind of the argument
| ADTAccessor String Kind -- Extract a field from an ADT value. Kind is the kind of the argument
deriving (Eq, Ord, Generic, G.Data, NFData)
-- | Special relations supported by z3
data SpecialRelOp = IsPartialOrder String
| IsLinearOrder String
| IsTreeOrder String
| IsPiecewiseLinearOrder String
deriving (Eq, Ord, G.Data, Show)
instance NFData SpecialRelOp where
rnf (IsPartialOrder n) = rnf n
rnf (IsLinearOrder n) = rnf n
rnf (IsTreeOrder n) = rnf n
rnf (IsPiecewiseLinearOrder n) = rnf n
-- | Floating point operations
data FPOp = FP_Cast Kind Kind SV -- From-Kind, To-Kind, RoundingMode. This is "value" conversion
| FP_Reinterpret Kind Kind -- From-Kind, To-Kind. This is bit-reinterpretation using IEEE-754 interchange format
| FP_Abs
| FP_Neg
| FP_Add
| FP_Sub
| FP_Mul
| FP_Div
| FP_FMA
| FP_Sqrt
| FP_Rem
| FP_RoundToIntegral
| FP_Min
| FP_Max
| FP_ObjEqual
| FP_IsNormal
| FP_IsSubnormal
| FP_IsZero
| FP_IsInfinite
| FP_IsNaN
| FP_IsNegative
| FP_IsPositive
deriving (Eq, Ord, G.Data, NFData, Generic)
-- Note that the show instance maps to the SMTLib names. We need to make sure
-- this mapping stays correct through SMTLib changes. The only exception
-- is FP_Cast; where we handle different source/origins explicitly later on.
instance Show FPOp where
show (FP_Cast f t r) = "(FP_Cast: " ++ show f ++ " -> " ++ show t ++ ", using RM [" ++ show r ++ "])"
show (FP_Reinterpret f t) = case t of
KFloat -> "(_ to_fp 8 24)"
KDouble -> "(_ to_fp 11 53)"
KFP eb sb -> "(_ to_fp " ++ show eb ++ " " ++ show sb ++ ")"
_ -> error $ "SBV.FP_Reinterpret: Unexpected conversion: " ++ show f ++ " to " ++ show t
show FP_Abs = "fp.abs"
show FP_Neg = "fp.neg"
show FP_Add = "fp.add"
show FP_Sub = "fp.sub"
show FP_Mul = "fp.mul"
show FP_Div = "fp.div"
show FP_FMA = "fp.fma"
show FP_Sqrt = "fp.sqrt"
show FP_Rem = "fp.rem"
show FP_RoundToIntegral = "fp.roundToIntegral"
show FP_Min = "fp.min"
show FP_Max = "fp.max"
show FP_ObjEqual = "="
show FP_IsNormal = "fp.isNormal"
show FP_IsSubnormal = "fp.isSubnormal"
show FP_IsZero = "fp.isZero"
show FP_IsInfinite = "fp.isInfinite"
show FP_IsNaN = "fp.isNaN"
show FP_IsNegative = "fp.isNegative"
show FP_IsPositive = "fp.isPositive"
-- | Non-linear operations
data NROp = NR_Sin
| NR_Cos
| NR_Tan
| NR_ASin
| NR_ACos
| NR_ATan
| NR_Sqrt
| NR_Sinh
| NR_Cosh
| NR_Tanh
| NR_Exp
| NR_Log
| NR_Pow
deriving (Eq, Ord, G.Data, NFData, Generic)
-- | The show instance carefully arranges for these to be printed as it can be understood by dreal
instance Show NROp where
show NR_Sin = "sin"
show NR_Cos = "cos"
show NR_Tan = "tan"
show NR_ASin = "asin"
show NR_ACos = "acos"
show NR_ATan = "atan"
show NR_Sinh = "sinh"
show NR_Cosh = "cosh"
show NR_Tanh = "tanh"
show NR_Sqrt = "sqrt"
show NR_Exp = "exp"
show NR_Log = "log"
show NR_Pow = "pow"
-- | Pseudo-boolean operations
data PBOp = PB_AtMost Int -- ^ At most k
| PB_AtLeast Int -- ^ At least k
| PB_Exactly Int -- ^ Exactly k
| PB_Le [Int] Int -- ^ At most k, with coefficients given. Generalizes PB_AtMost
| PB_Ge [Int] Int -- ^ At least k, with coefficients given. Generalizes PB_AtLeast
| PB_Eq [Int] Int -- ^ Exactly k, with coefficients given. Generalized PB_Exactly
deriving (Eq, Ord, Show, G.Data, NFData, Generic)
-- | Overflow operations
data OvOp = PlusOv Bool -- ^ Addition overflow. Bool is True if signed.
| SubOv Bool -- ^ Subtraction overflow. Bool is True if signed.
| MulOv Bool -- ^ Multiplication overflow. Bool is True if signed.
| DivOv -- ^ Division overflow. Only signed, since unsigned division does not overflow.
| NegOv -- ^ Unary negation overflow. Only signed, since unsigned negation does not overflow.
deriving (Eq, Ord, G.Data, NFData, Generic)
-- | Show instance. It's important that these follow the SMTLib names.
instance Show OvOp where
show (PlusOv signed) = "bv" ++ (if signed then "s" else "u") ++ "addo"
show (SubOv signed) = "bv" ++ (if signed then "s" else "u") ++ "subo"
show (MulOv signed) = "bv" ++ (if signed then "s" else "u") ++ "mulo"
show DivOv = "bvsdivo" -- This is confusing, the division is called bvsdivo, but negation is bvnego
show NegOv = "bvnego" -- But SMTLib's choice is deliberate: https://groups.google.com/u/0/g/smt-lib/c/J4D99wT0aKI
-- | String operations.
data StrOp = StrStrToNat -- ^ Retrieve integer encoded by string @s@ (ground rewriting only)
| StrNatToStr -- ^ Retrieve string encoded by integer @i@ (ground rewriting only)
| StrToCode -- ^ Equivalent to Haskell's ord
| StrFromCode -- ^ Equivalent to Haskell's chr
| StrInRe RegExp -- ^ Check if string is in the regular expression
deriving (Eq, Ord, G.Data, NFData, Generic)
-- | Regular-expression operators. The only thing we can do is to compare for equality/disequality.
data RegExOp = RegExEq RegExp RegExp
| RegExNEq RegExp RegExp
deriving (Eq, Ord, G.Data, NFData, Generic)
-- | Regular expressions. Note that regular expressions themselves are
-- concrete, but the 'Data.SBV.RegExp.match' function from the 'Data.SBV.RegExp.RegExpMatchable' class
-- can check membership against a symbolic string/character. Also, we
-- are preferring a datatype approach here, as opposed to coming up with
-- some string-representation; there are way too many alternatives
-- already so inventing one isn't a priority. Please get in touch if you
-- would like a parser for this type as it might be easier to use.
data RegExp = Literal String -- ^ Precisely match the given string
| All -- ^ Accept every string
| AllChar -- ^ Accept every single character
| None -- ^ Accept no strings
| Range Char Char -- ^ Accept range of characters
| Conc [RegExp] -- ^ Concatenation
| KStar RegExp -- ^ Kleene Star: Zero or more
| KPlus RegExp -- ^ Kleene Plus: One or more
| Opt RegExp -- ^ Zero or one
| Comp RegExp -- ^ Complement of regular expression
| Diff RegExp RegExp -- ^ Difference of regular expressions
| Loop Int Int RegExp -- ^ From @n@ repetitions to @m@ repetitions
| Power Int RegExp -- ^ Exactly @n@ repetitions, i.e., nth power
| Union [RegExp] -- ^ Union of regular expressions
| Inter RegExp RegExp -- ^ Intersection of regular expressions
deriving (Eq, Ord, G.Data, Generic, NFData)
-- | With overloaded strings, we can have direct literal regular expressions.
instance IsString RegExp where
fromString = Literal
-- | Regular expressions as a 'Num' instance. Note that only some operations make sense and
-- not in the most obvious way. For instance, we would typically expect @a - b@ to be the
-- same as @a + negate b@, but that equality does not hold in general. So, use the @Num@
-- instance only as constructing syntax, not doing algebraic manipulations.
instance Num RegExp where
-- flatten the concats to make them simpler
Conc xs * y = Conc (xs ++ [y])
x * Conc ys = Conc (x : ys)
x * y = Conc [x, y]
-- flatten the unions to make them simpler
Union xs + y = Union (xs ++ [y])
x + Union ys = Union (x : ys)
x + y = Union [x, y]
x - y = Diff x y
abs = error "Num.RegExp: no abs method"
signum = error "Num.RegExp: no signum method"
fromInteger x
| x == 0 = None
| x == 1 = Literal "" -- Unit for concatenation is the empty string
| True = error $ "Num.RegExp: Only 0 and 1 makes sense as a reg-exp, no meaning for: " ++ show x
negate = Comp
-- | Convert a reg-exp to a Haskell-like string
instance Show RegExp where
show = regExpToString show
-- | Convert a reg-exp to a SMT-lib acceptable representation
regExpToSMTString :: RegExp -> String
regExpToSMTString = regExpToString (\s -> '"' : stringToQFS s ++ "\"")
-- | Convert a RegExp to a string, parameterized by how strings are converted
regExpToString :: (String -> String) -> RegExp -> String
regExpToString fs (Literal s) = "(str.to.re " ++ fs s ++ ")"
regExpToString _ All = "re.all"
regExpToString _ AllChar = "re.allchar"
regExpToString _ None = "re.nostr"
regExpToString fs (Range ch1 ch2) = "(re.range " ++ fs [ch1] ++ " " ++ fs [ch2] ++ ")"
regExpToString _ (Conc []) = show (1 :: Integer)
regExpToString fs (Conc [x]) = regExpToString fs x
regExpToString fs (Conc xs) = "(re.++ " ++ unwords (map (regExpToString fs) xs) ++ ")"
regExpToString fs (KStar r) = "(re.* " ++ regExpToString fs r ++ ")"
regExpToString fs (KPlus r) = "(re.+ " ++ regExpToString fs r ++ ")"
regExpToString fs (Opt r) = "(re.opt " ++ regExpToString fs r ++ ")"
regExpToString fs (Comp r) = "(re.comp " ++ regExpToString fs r ++ ")"
regExpToString fs (Diff r1 r2) = "(re.diff " ++ regExpToString fs r1 ++ " " ++ regExpToString fs r2 ++ ")"
regExpToString fs (Loop lo hi r)
| lo >= 0, hi >= lo = "((_ re.loop " ++ show lo ++ " " ++ show hi ++ ") " ++ regExpToString fs r ++ ")"
| True = error $ "Invalid regular-expression Loop with arguments: " ++ show (lo, hi)
regExpToString fs (Power n r)
| n >= 0 = regExpToString fs (Loop n n r)
| True = error $ "Invalid regular-expression Power with arguments: " ++ show n
regExpToString fs (Inter r1 r2) = "(re.inter " ++ regExpToString fs r1 ++ " " ++ regExpToString fs r2 ++ ")"
regExpToString _ (Union []) = "re.nostr"
regExpToString fs (Union [x]) = regExpToString fs x
regExpToString fs (Union xs) = "(re.union " ++ unwords (map (regExpToString fs) xs) ++ ")"
-- | Show instance for @StrOp@. Note that the mapping here is important to match the SMTLib equivalents.
instance Show StrOp where
show StrStrToNat = "str.to.int" -- NB. SMTLib uses "int" here though only nats are supported
show StrNatToStr = "int.to.str" -- NB. SMTLib uses "int" here though only nats are supported
show StrToCode = "str.to_code"
show StrFromCode = "str.from_code"
-- Note the breakage here with respect to argument order. We fix this explicitly later.
show (StrInRe s) = "str.in_re " ++ regExpToSMTString s
-- | Show instance for @RegExOp@.
instance Show RegExOp where
show (RegExEq r1 r2) = "(= " ++ regExpToSMTString r1 ++ " " ++ regExpToSMTString r2 ++ ")"
show (RegExNEq r1 r2) = "(distinct " ++ regExpToSMTString r1 ++ " " ++ regExpToSMTString r2 ++ ")"
-- | For now, we represent lambda functions in op with their SMTLib equivalent strings.
-- This might change in the future.
newtype SMTLambda = SMTLambda String
deriving (Eq, Ord, G.Data, Generic)
deriving newtype NFData
-- | Simple show instance for SMTLambda
instance Show SMTLambda where
show (SMTLambda s) = s
-- | Sequence operations. Indexed by the element kind.
data SeqOp = SeqLen Kind
| SeqConcat Kind
| SeqNth Kind
| SeqUnit Kind
| SeqSubseq Kind
| SeqIndexOf Kind
| SeqContains Kind
| SeqPrefixOf Kind
| SeqSuffixOf Kind
| SeqReplace Kind
deriving (Eq, Ord, G.Data, NFData, Generic)
-- | Pick the correct operator
pickSeqOp :: Kind -> String -> String -> String
pickSeqOp KChar st _ = st
pickSeqOp _ _ sq = sq
-- | Show instance for SeqOp. Again, mapping is important.
instance Show SeqOp where
show (SeqLen k) = pickSeqOp k "str.len" "seq.len"
show (SeqConcat k) = pickSeqOp k "str.++" "seq.++"
show (SeqNth k) = pickSeqOp k "str.at" "seq.nth"
show (SeqUnit k) = pickSeqOp k "str.unit" "seq.unit"
show (SeqSubseq k) = pickSeqOp k "str.substr" "seq.extract"
show (SeqIndexOf k) = pickSeqOp k "str.indexof" "seq.indexof"
show (SeqContains k) = pickSeqOp k "str.contains" "seq.contains"
show (SeqPrefixOf k) = pickSeqOp k "str.prefixof" "seq.prefixof"
show (SeqSuffixOf k) = pickSeqOp k "str.suffixof" "seq.suffixof"
show (SeqReplace k) = pickSeqOp k "str.replace" "seq.replace"
-- | Set operations.
data SetOp = SetEqual
| SetMember
| SetInsert
| SetDelete
| SetIntersect
| SetUnion
| SetSubset
| SetDifference
| SetComplement
deriving (Eq, Ord, G.Data, NFData, Generic)
-- The show instance for 'SetOp' is merely for debugging, we map them separately so
-- the mapped strings are less important here.
instance Show SetOp where
show SetEqual = "=="
show SetMember = "Set.member"
show SetInsert = "Set.insert"
show SetDelete = "Set.delete"
show SetIntersect = "Set.intersect"
show SetUnion = "Set.union"
show SetSubset = "Set.subset"
show SetDifference = "Set.difference"
show SetComplement = "Set.complement"
-- Show instance for 'Op'. Note that this is largely for debugging purposes, not used
-- for being read by any tool.
instance Show Op where
show Shl = "<<"
show Shr = ">>"
show (Rol i) = "<<<" ++ show i
show (Ror i) = ">>>" ++ show i
show (Extract i j) = "choose [" ++ show i ++ ":" ++ show j ++ "]"
show (LkUp (ti, at, rt, l) i e)
= "lookup(" ++ tinfo ++ ", " ++ show i ++ ", " ++ show e ++ ")"
where tinfo = "table" ++ show ti ++ "(" ++ show at ++ " -> " ++ show rt ++ ", " ++ show l ++ ")"
show (KindCast fr to) = "cast_" ++ show fr ++ "_" ++ show to
show (Uninterpreted i) = "[uninterpreted] " ++ i
show (QuantifiedBool i) = "[quantified boolean] " ++ i
show (Label s) = "[label] " ++ s
show (IEEEFP w) = show w
show (NonLinear w) = show w
show (PseudoBoolean p) = show p
show (OverflowOp o) = show o
show (StrOp s) = show s
show (RegExOp s) = show s
show (SeqOp s) = show s
show (SetOp s) = show s
show (TupleConstructor 0) = "mkSBVTuple0"
show (TupleConstructor n) = "mkSBVTuple" ++ show n
show (TupleAccess i n) = "proj_" ++ show i ++ "_SBVTuple" ++ show n
show RationalConstructor = "SBV.Rational"
show (ArrayLambda s) = show s
show ReadArray = "select"
show WriteArray = "store"
show op
| Just s <- op `lookup` syms = s
| True = error "impossible happened; can't find op!" -- NB. Can't display the OP here! it's the show instance after all.
where syms = [ (Plus, "+"), (Times, "*"), (Minus, "-"), (UNeg, "-"), (Abs, "abs")
, (Quot, "quot")
, (Rem, "rem")
, (Equal True, "==="), (Equal False, "=="), (NotEqual, "/="), (Implies, "=>")
, (LessThan, "<"), (GreaterThan, ">"), (LessEq, "<="), (GreaterEq, ">=")
, (Ite, "if_then_else")
, (And, "&"), (Or, "|"), (XOr, "^"), (Not, "~")
, (Join, "#")
]
-- | Quantifiers: forall or exists. Note that we allow arbitrary nestings.
data Quantifier = ALL | EX deriving (Eq, G.Data)
-- | Show instance for 'Quantifier'
instance Show Quantifier where
show ALL = "Forall"
show EX = "Exists"
-- | Which context is this variable being created?
data VarContext = NonQueryVar (Maybe Quantifier) -- in this case, it can be quantified
| QueryVar -- in this case, it is always existential
-- | Are there any existential quantifiers?
needsExistentials :: [Quantifier] -> Bool
needsExistentials = (EX `elem`)
-- | A simple type for SBV computations, used mainly for uninterpreted constants.
-- We keep track of the signedness/size of the arguments. A non-function will
-- have just one entry in the list.
newtype SBVType = SBVType [Kind]
deriving (Eq, Ord, G.Data)
instance Show SBVType where
show (SBVType []) = error "SBV: internal error, empty SBVType"
show (SBVType xs) = intercalate " -> " $ map show xs
-- | A symbolic expression
data SBVExpr = SBVApp !Op ![SV]
deriving (Eq, Ord, G.Data)
-- | To improve hash-consing, take advantage of commutative operators by
-- reordering their arguments.
reorder :: SBVExpr -> SBVExpr
reorder s = case s of
SBVApp op [a, b] | isCommutative op && a > b -> SBVApp op [b, a]
_ -> s
where isCommutative :: Op -> Bool
isCommutative o = o `elem` [Plus, Times, Equal True, Equal False, NotEqual, And, Or, XOr]
-- Show instance for 'SBVExpr'. Again, only for debugging purposes.
instance Show SBVExpr where
show (SBVApp Ite [t, a, b]) = unwords ["if", show t, "then", show a, "else", show b]
show (SBVApp Shl [a, i]) = unwords [show a, "<<", show i]
show (SBVApp Shr [a, i]) = unwords [show a, ">>", show i]
show (SBVApp (Rol i) [a]) = unwords [show a, "<<<", show i]
show (SBVApp (Ror i) [a]) = unwords [show a, ">>>", show i]
show (SBVApp (PseudoBoolean pb) args) = unwords (show pb : map show args)
show (SBVApp (OverflowOp op) args) = unwords (show op : map show args)
show (SBVApp op args) | showOpInfix op, length args == 2 = unwords (map show (take 1 args) ++ show op : map show (drop 1 args))
| True = unwords (show op : map show args)
-- | Should we display this Op infix?
showOpInfix :: Op -> Bool
showOpInfix = (`elem` infixOps)
where infixOps = [ Plus, Times, Minus, Quot, Rem, Implies
, Equal True, Equal False, NotEqual, LessThan, GreaterThan, LessEq, GreaterEq
, And, Or, XOr, Join
]
-- | A program is a sequence of assignments
newtype SBVPgm = SBVPgm {pgmAssignments :: S.Seq (SV, SBVExpr)}
deriving G.Data
-- | Helper synonym for text, in case we switch to something else later.
type Name = T.Text
-- | t'NamedSymVar' pairs symbolic values and user given/automatically generated names
data NamedSymVar = NamedSymVar !SV !Name
deriving (Show, Generic, G.Data)
-- | For comparison purposes, we simply use the SV and ignore the name
instance Eq NamedSymVar where
(==) (NamedSymVar l _) (NamedSymVar r _) = l == r
instance Ord NamedSymVar where
compare (NamedSymVar l _) (NamedSymVar r _) = compare l r
-- | Convert to a named symvar, from string
toNamedSV' :: SV -> String -> NamedSymVar
toNamedSV' s = NamedSymVar s . T.pack
-- | Convert to a named symvar, from text
toNamedSV :: SV -> Name -> NamedSymVar
toNamedSV = NamedSymVar
-- | Get the node id from a named sym var
namedNodeId :: NamedSymVar -> NodeId
namedNodeId = swNodeId . getSV
-- | Get the SV from a named sym var
getSV :: NamedSymVar -> SV
getSV (NamedSymVar s _) = s
-- | Get the user-name typed value from named sym var
getUserName :: NamedSymVar -> Name
getUserName (NamedSymVar _ nm) = nm
-- | Get the string typed value from named sym var
getUserName' :: NamedSymVar -> String
getUserName' = T.unpack . getUserName
-- | Style of optimization. Note that in the pareto case the user is allowed
-- to specify a max number of fronts to query the solver for, since there might
-- potentially be an infinite number of them and there is no way to know exactly
-- how many ahead of time. If 'Nothing' is given, SBV will possibly loop forever
-- if the number is really infinite.
data OptimizeStyle = Lexicographic -- ^ Objectives are optimized in the order given, earlier objectives have higher priority.
| Independent -- ^ Each objective is optimized independently.
| Pareto (Maybe Int) -- ^ Objectives are optimized according to pareto front: That is, no objective can be made better without making some other worse.
deriving (Eq, Show)
-- | Penalty for a soft-assertion. The default penalty is @1@, with all soft-assertions belonging
-- to the same objective goal. A positive weight and an optional group can be provided by using
-- the v'Penalty' constructor.
data Penalty = DefaultPenalty -- ^ Default: Penalty of @1@ and no group attached
| Penalty Rational (Maybe String) -- ^ Penalty with a weight and an optional group
deriving Show
-- | Objective of optimization. We can minimize, maximize, or give a soft assertion with a penalty
-- for not satisfying it.
data Objective a = Minimize String a -- ^ Minimize this metric
| Maximize String a -- ^ Maximize this metric
| AssertWithPenalty String a Penalty -- ^ A soft assertion, with an associated penalty
deriving (Show, Functor)
-- | The name of the objective
objectiveName :: Objective a -> String
objectiveName (Minimize s _) = s
objectiveName (Maximize s _) = s
objectiveName (AssertWithPenalty s _ _) = s
-- | The state we keep track of as we interact with the solver
data QueryState = QueryState { queryAsk :: Maybe Int -> String -> IO String
, querySend :: Maybe Int -> String -> IO ()
, queryRetrieveResponse :: Maybe Int -> IO String
, queryConfig :: SMTConfig
, queryTerminate :: Maybe C.SomeException -> IO ()
, queryTimeOutValue :: Maybe Int
, queryAssertionStackDepth :: Int
}
-- | Computations which support query operations.
class Monad m => MonadQuery m where
queryState :: m State
default queryState :: (MonadTrans t, MonadQuery m', m ~ t m') => m State
queryState = lift queryState
instance MonadQuery m => MonadQuery (ExceptT e m)
instance MonadQuery m => MonadQuery (MaybeT m)
instance MonadQuery m => MonadQuery (ReaderT r m)
instance MonadQuery m => MonadQuery (SS.StateT s m)
instance MonadQuery m => MonadQuery (LS.StateT s m)
instance (MonadQuery m, Monoid w) => MonadQuery (SW.WriterT w m)
instance (MonadQuery m, Monoid w) => MonadQuery (LW.WriterT w m)
-- | A query is a user-guided mechanism to directly communicate and extract
-- results from the solver. A generalization of 'Data.SBV.Query'.
newtype QueryT m a = QueryT { runQueryT :: ReaderT State m a }
deriving newtype (Applicative, Functor, Monad, MonadIO, MonadTrans, MonadError e, MonadState s, MonadWriter w)
instance Monad m => MonadQuery (QueryT m) where
queryState = QueryT ask
mapQueryT :: (ReaderT State m a -> ReaderT State n b) -> QueryT m a -> QueryT n b
mapQueryT f = QueryT . f . runQueryT
{-# INLINE mapQueryT #-}
-- Have to define this one by hand, because we use ReaderT in the implementation
instance MonadReader r m => MonadReader r (QueryT m) where
ask = lift ask
local f = mapQueryT $ mapReaderT $ local f
-- | A query is a user-guided mechanism to directly communicate and extract
-- results from the solver.
type Query = QueryT IO
instance MonadSymbolic Query where
symbolicEnv = queryState
instance NFData OptimizeStyle where
rnf x = x `seq` ()
instance NFData Penalty where
rnf DefaultPenalty = ()
rnf (Penalty p mbs) = rnf p `seq` rnf mbs
instance NFData a => NFData (Objective a) where
rnf (Minimize s a) = rnf s `seq` rnf a
rnf (Maximize s a) = rnf s `seq` rnf a
rnf (AssertWithPenalty s a p) = rnf s `seq` rnf a `seq` rnf p
-- | A result can either produce something at the top or as a lambda/constraint. Distinguish by inputs
data ResultInp = ResultTopInps ([NamedSymVar], [NamedSymVar]) -- user inputs -- trackers
| ResultLamInps [(Quantifier, NamedSymVar)] -- for constraints, we can have quantifiers
deriving G.Data
instance NFData ResultInp where
rnf (ResultTopInps xs) = rnf xs
rnf (ResultLamInps xs) = rnf xs
-- | Several data about the program
data ProgInfo = ProgInfo { hasQuants :: Bool
, progSpecialRels :: [SpecialRelOp]
, progTransClosures :: [(String, String)]
}
deriving G.Data
instance NFData ProgInfo where
rnf (ProgInfo a b c) = rnf a `seq` rnf b `seq` rnf c
deriving instance G.Data CallStack
deriving instance G.Data SrcLoc
-- | Result of running a symbolic computation
data Result = Result { progInfo :: ProgInfo -- ^ various info we collect about the program
, reskinds :: Set.Set Kind -- ^ kinds used in the program
, resTraces :: [(String, CV)] -- ^ quick-check counter-example information (if any)
, resObservables :: [(String, CV -> Bool, SV)] -- ^ observable expressions (part of the model)
, resUISegs :: [(String, [String])] -- ^ uninterpeted code segments
, resParams :: ResultInp -- ^ top-inputs or lambda params
, resConsts :: (CnstMap, [(SV, CV)]) -- ^ constants
, resTables :: [((Int, Kind, Kind), [SV])] -- ^ tables (automatically constructed) (tableno, index-type, result-type) elts
, resUIConsts :: [(String, (Bool, Maybe [String], SBVType))] -- ^ uninterpreted constants
, resDefinitions :: [(String, (SMTDef, SBVType))] -- ^ definitions created via smtFunction
, resAsgns :: SBVPgm -- ^ assignments
, resConstraints :: S.Seq (Bool, [(String, String)], SV) -- ^ additional constraints (boolean)
, resAssertions :: [(String, Maybe CallStack, SV)] -- ^ assertions
, resOutputs :: [SV] -- ^ outputs
}
deriving G.Data
-- Show instance for 'Result'. Only for debugging purposes.
instance Show Result where
-- If there's nothing interesting going on, just print the constant. Note that the
-- definition of interesting here is rather subjective; but essentially if we reduced
-- the result to a single constant already, without any reference to anything.
show Result{resConsts=(_, cs), resOutputs=[r]}
| Just c <- r `lookup` cs
= show c
show (Result _ kinds _ _ cgs params (_, cs) ts uis defns xs cstrs asserts os) = intercalate "\n" $
(if null usorts then [] else "SORTS" : map (" " ++) usorts)
++ (case params of
ResultTopInps (i, t) -> "INPUTS" : map shn i ++ (if null t then [] else "TRACKER VARS" : map shn t)
ResultLamInps qs -> "LAMBDA-CONSTRAINT PARAMS" : map shq qs
)
++ ["CONSTANTS"]
++ concatMap shc cs
++ ["TABLES"]
++ map sht ts
++ ["UNINTERPRETED CONSTANTS"]
++ map shui uis
++ ["USER GIVEN CODE SEGMENTS"]
++ concatMap shcg cgs
++ ["AXIOMS-DEFINITIONS"]
++ map show defns
++ ["DEFINE"]
++ map (\(s, e) -> " " ++ shs s ++ " = " ++ show e) (F.toList (pgmAssignments xs))
++ ["CONSTRAINTS"]
++ map ((" " ++) . shCstr) (F.toList cstrs)
++ ["ASSERTIONS"]
++ map ((" "++) . shAssert) asserts
++ ["OUTPUTS"]
++ sh2 os
where sh2 :: Show a => [a] -> [String]
sh2 = map ((" "++) . show)
usorts = [s | KADT s _ _ <- filter isUninterpreted (Set.toList kinds)]
shs sv = show sv ++ " :: " ++ show (swKind sv)
sht ((i, at, rt), es) = " Table " ++ show i ++ " : " ++ show at ++ "->" ++ show rt ++ " = " ++ show es
shc (sv, cv)
| sv == falseSV || sv == trueSV
= []
| True
= [" " ++ show sv ++ " = " ++ show cv]
shcg (s, ss) = ("Variable: " ++ s) : map (" " ++) ss
shn (NamedSymVar sv nm) = " " <> ni <> " :: " ++ show (swKind sv) ++ alias
where ni = show sv
alias | ni == T.unpack nm = ""
| True = ", aliasing " ++ show nm
shq (q, v) = shn v ++ ", " ++ if q == ALL then "universal" else "existential"
shui (nm, t) = " [uninterpreted] " ++ nm ++ " :: " ++ show t
shCstr (isSoft, [], c) = soft isSoft ++ show c
shCstr (isSoft, [(":named", nm)], c) = soft isSoft ++ nm ++ ": " ++ show c
shCstr (isSoft, attrs, c) = soft isSoft ++ show c ++ " (attributes: " ++ show attrs ++ ")"
soft True = "[SOFT] "
soft False = ""
shAssert (nm, stk, p) = " -- assertion: " ++ nm ++ " " ++ maybe "[No location]"
#if MIN_VERSION_base(4,9,0)
prettyCallStack
#else
showCallStack
#endif
stk ++ ": " ++ show p
-- | Expression map, used for hash-consing
type ExprMap = Map.Map SBVExpr SV
-- | Constants are stored in a map, for hash-consing.
type CnstMap = Map.Map CV SV
-- | Kinds used in the program; used for determining the final SMT-Lib logic to pick
type KindSet = Set.Set Kind
-- | Tables generated during a symbolic run
type TableMap = Map.Map (Kind, Kind, [SV]) Int
-- | Uninterpreted-constants generated during a symbolic run
type UIMap = Map.Map String (Bool, Maybe [String], SBVType) -- If Bool is true, then this is a curried function
-- | Code-segments for Uninterpreted-constants, as given by the user
type CgMap = Map.Map String [String]
-- | Cached values, implementing sharing
type Cache a = IMap.IntMap [(StableName (State -> IO a), a)]
-- | Stage of an interactive run
data IStage = ISetup -- Before we initiate contact.
| ISafe -- In the context of a safe/safeWith call
| IRun -- After the contact is started
-- | Are we checking safety
isSafetyCheckingIStage :: IStage -> Bool
isSafetyCheckingIStage s = case s of
ISetup -> False
ISafe -> True
IRun -> False
-- | Are we in setup?
isSetupIStage :: IStage -> Bool
isSetupIStage s = case s of
ISetup -> True
ISafe -> False
IRun -> True
-- | Are we in a run?
isRunIStage :: IStage -> Bool
isRunIStage s = case s of
ISetup -> False
ISafe -> False
IRun -> True
-- | Different means of running a symbolic piece of code
data SBVRunMode = SMTMode QueryContext IStage Bool SMTConfig -- ^ In regular mode, with a stage. Bool is True if this is SAT.
| CodeGen -- ^ Code generation mode.
| LambdaGen (Maybe Int) -- ^ Inside a lambda-expression at level. If Nothing, then closed lambda.
| Concrete (Maybe (Bool, [(NamedSymVar, CV)])) -- ^ Concrete simulation mode, with given environment if any. If Nothing: Random.
-- Show instance for SBVRunMode; debugging purposes only
instance Show SBVRunMode where
show (SMTMode qc ISetup True _) = "Satisfiability setup (" ++ show qc ++ ")"
show (SMTMode qc ISafe True _) = "Safety setup (" ++ show qc ++ ")"
show (SMTMode qc IRun True _) = "Satisfiability (" ++ show qc ++ ")"
show (SMTMode qc ISetup False _) = "Proof setup (" ++ show qc ++ ")"
show (SMTMode qc ISafe False _) = error $ "ISafe-False is not an expected/supported combination for SBVRunMode! (" ++ show qc ++ ")"
show (SMTMode qc IRun False _) = "Proof (" ++ show qc ++ ")"
show CodeGen = "Code generation"
show LambdaGen{} = "Lambda generation"
show (Concrete Nothing) = "Concrete evaluation with random values"
show (Concrete (Just (True, _))) = "Concrete evaluation during model validation for sat"
show (Concrete (Just (False, _))) = "Concrete evaluation during model validation for prove"
-- | Is this a CodeGen run? (i.e., generating code)
isCodeGenMode :: State -> IO Bool
isCodeGenMode State{runMode} = do rm <- readIORef runMode
return $ case rm of
Concrete{} -> False
SMTMode{} -> False
LambdaGen{} -> False
CodeGen -> True
-- | The state in query mode, i.e., additional context
data IncState = IncState { rNewInps :: IORef [NamedSymVar] -- always existential!
, rNewKinds :: IORef KindSet
, rNewConsts :: IORef CnstMap
, rNewTbls :: IORef TableMap
, rNewUIs :: IORef UIMap
, rNewAsgns :: IORef SBVPgm
, rNewConstraints :: IORef (S.Seq (Bool, [(String, String)], SV))
}
-- | Get a new IncState
newIncState :: IO IncState
newIncState = do
is <- newIORef []
ks <- newIORef Set.empty
nc <- newIORef Map.empty
tm <- newIORef Map.empty
ui <- newIORef Map.empty
pgm <- newIORef (SBVPgm S.empty)
cstrs <- newIORef S.empty
return IncState { rNewInps = is
, rNewKinds = ks
, rNewConsts = nc
, rNewTbls = tm
, rNewUIs = ui
, rNewAsgns = pgm
, rNewConstraints = cstrs
}
-- | Get a new IncState
withNewIncState :: State -> (State -> IO a) -> IO (IncState, a)
withNewIncState st cont = do
is <- newIncState
R.modifyIORef' (rIncState st) (const is)
r <- cont st
finalIncState <- readIORef (rIncState st)
return (finalIncState, r)
-- | User defined inputs
type UserInputs = S.Seq NamedSymVar
-- | Internally declared
type InternInps = S.Seq NamedSymVar
-- | Entire set of names, for faster lookup
type AllInps = Set.Set Name
-- | Inputs as a record of maps and sets. See above type-synonyms for their roles.
data Inputs = Inputs { userInputs :: !UserInputs
, internInputs :: !InternInps
, allInputs :: !AllInps
} deriving (Eq,Show)
-- | Inputs to a lambda-abstraction. These are quantified to handle constraints
type LambdaInputs = S.Seq (Quantifier, NamedSymVar)
-- | Semigroup instance; combining according to indexes.
instance Semigroup Inputs where
(Inputs lui lii lai) <> (Inputs rui rii rai) = Inputs (lui <> rui) (lii <> rii) (lai <> rai)
-- | Monoid instance, we start with no maps.
instance Monoid Inputs where
mempty = Inputs { userInputs = mempty
, internInputs = mempty
, allInputs = mempty
}
-- | Modify the user-inputs field
onUserInputs :: (UserInputs -> UserInputs) -> Inputs -> Inputs
onUserInputs f inp@Inputs{userInputs} = inp{userInputs = f userInputs}
-- | Modify the internal-inputs field
onInternInputs :: (InternInps -> InternInps) -> Inputs -> Inputs
onInternInputs f inp@Inputs{internInputs} = inp{internInputs = f internInputs}
-- | Modify the all-inputs field
onAllInputs :: (AllInps -> AllInps) -> Inputs -> Inputs
onAllInputs f inp@Inputs{allInputs} = inp{allInputs = f allInputs}
-- | Add a new internal input
addInternInput :: SV -> Name -> Inputs -> Inputs
addInternInput sv nm = goAll . goIntern
where !new = toNamedSV sv nm
goIntern = onInternInputs (S.|> new)
goAll = onAllInputs (Set.insert nm)
-- | Add a new user input
addUserInput :: SV -> Name -> Inputs -> Inputs
addUserInput sv nm = goAll . goUser
where new = toNamedSV sv nm
goUser = onUserInputs (S.|> new) -- add to the end of the sequence
goAll = onAllInputs (Set.insert nm)
-- | Find a user-input from its SV. Note that only level-0 vars
-- can be found this way.
lookupInput :: (a -> SV) -> SV -> S.Seq a -> Maybe a
lookupInput f sv ns
| l == Just 0 = res
| True = Nothing -- l != Just 0, a lambda var, whether top-level or in a scope, so we ignore
where
(_, l, i) = getId (swNodeId sv)
svs = fmap f ns
res = case S.lookup i ns of -- Nothing on negative Int or Int > length seq
Nothing -> secondLookup
x@(Just e) -> if sv == f e then x else secondLookup
-- we try the fast lookup first, if the node ids don't match then
-- we use the more expensive O (n) to find the index and the elem
secondLookup = S.elemIndexL sv svs >>= flip S.lookup ns
-- | A defined function/value
data SMTDef = SMTDef Kind -- ^ Final kind of the definition (resulting kind, not the params)
[String] -- ^ other definitions it refers to
(Maybe String) -- ^ parameter string
(Int -> String) -- ^ Body, in SMTLib syntax, given the tab amount
deriving G.Data
-- | For debug purposes
instance Show SMTDef where
show (SMTDef fk frees p body) = unlines [ "-- User defined function:"
, "-- Final return type : " ++ show fk
, "-- Refers to : " ++ intercalate ", " frees
, "-- Parameters : " ++ fromMaybe "NONE" p
, "-- Body : "
, body 2
]
-- | NFData instance for SMTDef
instance NFData SMTDef where
rnf (SMTDef fk frees params body) = rnf fk `seq` rnf frees `seq` rnf params `seq` rnf body
-- | The state of the symbolic interpreter
data State = State { sbvContext :: SBVContext
, pathCond :: SVal -- ^ kind KBool
, stCfg :: SMTConfig
, startTime :: UTCTime
, rProgInfo :: IORef ProgInfo
, runMode :: IORef SBVRunMode
, rIncState :: IORef IncState
, rCInfo :: IORef [(String, CV)]
, rObservables :: IORef (S.Seq (Name, CV -> Bool, SV))
, rctr :: IORef Int -- Used for numbering SVs
, freshNameCtr :: IORef Int -- Used for calls to some
, rLambdaLevel :: IORef (Maybe Int) -- If Nothing, then top-level lambda
, rUsedKinds :: IORef KindSet
, rUsedLbls :: IORef (Set.Set String)
, rinps :: IORef Inputs
, rlambdaInps :: IORef LambdaInputs
, rConstraints :: IORef (S.Seq (Bool, [(String, String)], SV))
, rPartitionVars :: IORef [String]
, routs :: IORef [SV]
, rtblMap :: IORef TableMap
, spgm :: IORef SBVPgm
, rconstMap :: IORef CnstMap
, rexprMap :: IORef ExprMap
, rUIMap :: IORef UIMap
, rUserFuncs :: IORef (Set.Set String) -- Functions that the user wanted explicit code generation for
, rCgMap :: IORef CgMap
, rDefns :: IORef [(String, (SMTDef, SBVType))]
, rSMTOptions :: IORef [SMTOption]
, rOptGoals :: IORef [Objective (SV, SV)]
, rAsserts :: IORef [(String, Maybe CallStack, SV)]
, rOutstandingAsserts :: IORef Bool -- Did we send an assert after the last check-sat call?
, rSVCache :: IORef (Cache SV)
, rQueryState :: IORef (Maybe QueryState)
, parentState :: Maybe State -- Pointer to our parent if we're in a sublevel
}
-- | Chase to the root state. No infinite chains!
getRootState :: State -> State
getRootState st = maybe st getRootState (parentState st)
-- NFData is a bit of a lie, but it's sufficient, most of the content is iorefs that we don't want to touch
instance NFData State where
rnf State{} = ()
-- | Get the current path condition
getSValPathCondition :: State -> SVal
getSValPathCondition = pathCond
-- | Extend the path condition with the given test value.
extendSValPathCondition :: State -> (SVal -> SVal) -> State
extendSValPathCondition st f = st{pathCond = f (pathCond st)}
-- | Are we running in proof mode?
inSMTMode :: State -> IO Bool
inSMTMode State{runMode} = do rm <- readIORef runMode
return $ case rm of
CodeGen -> False
LambdaGen{} -> False
Concrete{} -> False
SMTMode{} -> True
-- | The "Symbolic" value. Either a constant (@Left@) or a symbolic
-- value (@Right Cached@). Note that caching is essential for making
-- sure sharing is preserved.
data SVal = SVal !Kind !(Either CV (Cached SV))
-- | Kind instance for SVal simply passes the kind out
instance HasKind SVal where
kindOf (SVal k _) = k
-- Show instance for t'SVal'. Not particularly "desirable", but will do if needed
-- NB. We do not show the type info on constant KBool values, since there's no
-- implicit "fromBoolean" applied to Booleans in Haskell; and thus a statement
-- of the form "True :: SBool" is just meaningless. (There should be a fromBoolean!)
instance Show SVal where
show (SVal KBool (Left c)) = showCV False c
show (SVal k (Left c)) = showCV False c ++ " :: " ++ show k
show (SVal k (Right _)) = "<symbolic> :: " ++ show k
-- | Things we do not support in interactive mode, at least for now!
noInteractive :: [String] -> a
noInteractive ss = error $ unlines $ ""
: "*** Data.SBV: Unsupported interactive/query mode feature."
: map ("*** " ++) ss
++ ["*** Data.SBV: Please report this as a feature request!"]
-- | Things we do not support in interactive mode, nor we ever intend to
noInteractiveEver :: [String] -> a
noInteractiveEver ss = error $ unlines $ ""
: "*** Data.SBV: Unsupported interactive/query mode feature."
: map ("*** " ++) ss
-- | Modification of the state, but carefully handling the interactive tasks.
-- Note that the state is always updated regardless of the mode, but we get
-- to also perform extra operation in interactive mode. (Typically error out, but also simply
-- ignore if it has no impact.)
modifyState :: State -> (State -> IORef a) -> (a -> a) -> IO () -> IO ()
modifyState st@State{runMode} field update interactiveUpdate = do
R.modifyIORef' (field st) update
rm <- readIORef runMode
case rm of
SMTMode _ IRun _ _ -> interactiveUpdate
_ -> return ()
-- | Modify the incremental state
modifyIncState :: State -> (IncState -> IORef a) -> (a -> a) -> IO ()
modifyIncState State{rIncState} field update = do
incState <- readIORef rIncState
R.modifyIORef' (field incState) update
-- | Add an observable
-- notice that we cons like a list, we should build at the end of the seq, but cons to preserve semantics for now
recordObservable :: State -> String -> (CV -> Bool) -> SV -> IO ()
recordObservable st (T.pack -> nm) chk sv = modifyState st rObservables ((nm, chk, sv) S.<|) (return ())
-- | Increment the variable counter
incrementInternalCounter :: State -> IO Int
incrementInternalCounter st = do ctr <- readIORef (rctr st)
modifyState st rctr (+1) (return ())
return ctr
{-# INLINE incrementInternalCounter #-}
-- | Increment the fresh-var counter
incrementFreshNameCounter :: State -> IO Int
incrementFreshNameCounter st = do ctr <- readIORef (freshNameCtr st)
modifyState st freshNameCtr (+1) (return ())
return ctr
{-# INLINE incrementFreshNameCounter #-}
-- | Kind of code we have for uninterpretation
data UICodeKind = UINone Bool -- no code. If bool is true, then curried.
| UISMT SMTDef -- SMTLib, first argument are the free-variables in it
| UICgC [String] -- Code-gen, currently only C
-- | A newtype wrapper for uninterpreted function names. We distinguish between user names and those of constructors
data UIName = UIGiven String -- ^ Full name
| UIADT ADTOp -- ^ The name of an ADT operation based on the constructor
-- | Uninterpreted constants and functions. An uninterpreted constant is
-- a value that is indexed by its name. The only property the prover assumes
-- about these values are that they are equivalent to themselves; i.e., (for
-- functions) they return the same results when applied to same arguments.
-- We support uninterpreted-functions as a general means of black-box'ing
-- operations that are /irrelevant/ for the purposes of the proof; i.e., when
-- the proofs can be performed without any knowledge about the function itself.
svUninterpreted :: Kind -> UIName -> UICodeKind -> [SVal] -> SVal
svUninterpreted k nm code args = svUninterpretedGen k nm code args Nothing
svUninterpretedNamedArgs :: Kind -> UIName -> UICodeKind -> [(SVal, String)] -> SVal
svUninterpretedNamedArgs k nm code args = svUninterpretedGen k nm code (map fst args) (Just (map snd args))
svUninterpretedGen :: Kind -> UIName -> UICodeKind -> [SVal] -> Maybe [String] -> SVal
svUninterpretedGen k nm code args mbArgNames = SVal k $ Right $ cache result
where result st = do let ty = SBVType (map kindOf args ++ [k])
op <- newUninterpreted st nm mbArgNames ty code
sws <- mapM (svToSV st) args
mapM_ forceSVArg sws
newExpr st k $ SBVApp op sws
-- | Generate a unique name for the given function based on the object's stable name
prefixNameToUnique :: State -> String -> IO String
prefixNameToUnique st pre = do
uiMap <- readIORef (rUIMap st)
let suffix 0 = pre
suffix i = pre ++ "_" ++ show i
case [cand | i <- [0::Int ..], let cand = suffix i, cand `Map.notMember` uiMap] of
(n:_) -> pure n
[] -> error $ "genUniqueName: Can't generate a unique name for prefix: " ++ pre -- can't happen
-- | Create a new value, possibly with user given code. This function might change
-- the name given, putting bars around it if needed. That's the name returned.
newUninterpreted :: State -> UIName -> Maybe [String] -> SBVType -> UICodeKind -> IO Op
newUninterpreted st uiName mbArgNames t uiCode = do
let (adtOp, candName) = case uiName of
UIGiven n -> (False, n)
UIADT o -> case o of
ADTConstructor n _ -> (True, n)
ADTTester n _ -> (True, n)
ADTAccessor n _ -> (True, n)
-- determine the final name. We leave constructors alone.
let nm = case () of
() | "__internal_sbv_" `isPrefixOf` candName -> candName -- internal names go thru
| adtOp -> candName -- ADT names go thru
| True -> barify candName -- surround with bars if not legitimate in SMTLib
extraComment = case uiName of
UIGiven n | nm /= n -> " (Given: " ++ n ++ ")"
_ -> ""
-- Check if reserved:
when (isReserved nm) $
error $ unlines [ ""
, "*** Data.SBV: User given name " ++ show nm ++ " is a reserved name in SMTLib."
, "***"
, "*** Please use a different name to avoid collisions."
]
isCurried <- case uiCode of
UINone c -> pure c
UISMT d -> do modifyState st rDefns (\defs -> (nm, (d, t)) : filter (\(onm, _) -> onm /= nm) defs)
$ noInteractive [ "Defined functions (smtFunction):"
, " Name: " ++ nm ++ extraComment
, " Type: " ++ show t
, ""
, "You should explicitly register these functions by calling"
, "the function 'registerFunction' on them before starting the query section."
]
pure True
UICgC c -> -- No need to record the code in interactive mode: CodeGen doesn't use interactive
do modifyState st rCgMap (Map.insert nm c) (return ())
pure True
let checkType :: SBVType -> r -> r
checkType t' cont
| t /= t' = error $ "Uninterpreted constant " ++ show nm ++ extraComment ++ " used at incompatible types\n"
++ " Current type : " ++ show t ++ "\n"
++ " Previously used at: " ++ show t'
| True = cont
-- If we're not a constructor, register it:
unless adtOp $ do
uiMap <- readIORef (rUIMap st)
case nm `Map.lookup` uiMap of
Just (_, _, t') -> checkType t' (return ())
Nothing -> modifyState st rUIMap (Map.insert nm (isCurried, mbArgNames, t))
$ modifyIncState st rNewUIs
(\newUIs -> case nm `Map.lookup` newUIs of
Just (_, _, t') -> checkType t' newUIs
Nothing -> Map.insert nm (isCurried, mbArgNames, t) newUIs)
pure $ case uiName of
UIGiven{} -> Uninterpreted nm
UIADT (ADTConstructor _ k) -> ADTOp (ADTConstructor nm k)
UIADT (ADTTester _ k) -> ADTOp (ADTTester nm k)
UIADT (ADTAccessor _ k) -> ADTOp (ADTAccessor nm k)
-- | Add a new sAssert based constraint
addAssertion :: State -> Maybe CallStack -> String -> SV -> IO ()
addAssertion st cs msg cond = modifyState st rAsserts ((msg, cs, cond):)
$ noInteractive [ "Named assertions (sAssert):"
, " Tag: " ++ msg
, " Loc: " ++ maybe "Unknown" show cs
]
-- | Create an internal variable, which acts as an input but isn't visible to the user.
-- Such variables are existentially quantified in a SAT context, and universally quantified
-- in a proof context.
newInternalVariable :: State -> Kind -> IO SV
newInternalVariable st k = do NamedSymVar sv nm <- newSV st k
let n = "__internal_sbv_" <> nm
v = NamedSymVar sv n
modifyState st rinps (addUserInput sv n) $ modifyIncState st rNewInps (v :)
return sv
{-# INLINE newInternalVariable #-}
-- | Create a variable to be used in a constraint-expression
quantVar :: Quantifier -> State -> Kind -> IO SV
quantVar q st k = do v@(NamedSymVar sv _) <- newSV st k
modifyState st rlambdaInps (S.|> (q, v)) (return ())
return sv
{-# INLINE quantVar #-}
-- | Create a variable to be used in a lambda-expression
lambdaVar :: State -> Kind -> IO SV
lambdaVar = quantVar ALL
{-# INLINE lambdaVar #-}
-- | Create a new SV
newSV :: State -> Kind -> IO NamedSymVar
newSV st k = do ctr <- incrementInternalCounter st
ll <- readIORef (rLambdaLevel st)
let sv = SV k (NodeId (sbvContext st, ll, ctr))
registerKind st k
return $ NamedSymVar sv $ T.pack (show sv)
{-# INLINE newSV #-}
-- | Register a new kind with the system, used for uninterpreted sorts.
-- NB: Is it safe to have new kinds in query mode? It could be that
-- the new kind might introduce a constraint that effects the logic. For
-- instance, if we're seeing 'Double' for the first time and using a BV
-- logic, then things would fall apart. But this should be rare, and hopefully
-- the success-response checking mechanism will catch the rare cases where this
-- is an issue. In either case, the user can always arrange for the right
-- logic by calling 'Data.SBV.setLogic' appropriately, so it seems safe to just
-- allow for this.
registerKind :: State -> Kind -> IO ()
registerKind st k
| KADT sortName _ _ <- k, isReserved sortName
= error $ "SBV: " ++ show sortName ++ " is a reserved sort; please use a different name."
| True
= do -- Adding a kind to the incState is tricky; we only need to add it
-- * If it's an uninterpreted sort that's not already in the general state
-- * OR If it's a tuple-sort whose cardinality isn't already in the general state
-- * OR If it's a list that's not already in the general state (so we can send the flatten commands)
existingKinds <- readIORef (rUsedKinds st)
-- For ADTs we need to make sure we haven't added it before
let adtExists = case k of
KADT s _ _ -> s `elem` [s' | KADT s' _ _ <- Set.toList existingKinds]
_ -> False
unless adtExists $
modifyState st rUsedKinds (Set.insert k) $ do
-- Why do we discriminate here? Because the incremental context is sensitive to the
-- order: In particular, if an uninterpreted kind is already in there, we don't
-- want to re-add because double-declaration would be wrong. See 'cvtInc' for details.
let needsAdding = case k of
KADT s _ _ -> s `notElem` [s' | KADT s' _ _ <- Set.toList existingKinds]
KList{} -> k `notElem` existingKinds
KTuple nks -> length nks `notElem` [length oks | KTuple oks <- Set.toList existingKinds]
_ -> False
when needsAdding $ modifyIncState st rNewKinds (Set.insert k)
-- Don't forget to register subkinds!
case k of
KVar {} -> return ()
KBool {} -> return ()
KBounded {} -> return ()
KUnbounded{} -> return ()
KReal {} -> return ()
KFloat {} -> return ()
KDouble {} -> return ()
KFP {} -> return ()
KRational {} -> return ()
KChar {} -> return ()
KString {} -> return ()
KApp _ ks -> mapM_ (registerKind st) ks
KADT _ pks cks -> mapM_ (registerKind st) (map snd pks ++ concatMap snd cks)
KList ek -> registerKind st ek
KSet ek -> registerKind st ek
KTuple eks -> mapM_ (registerKind st) eks
KArray k1 k2 -> mapM_ (registerKind st) [k1, k2]
-- | Register a new label with the system, making sure they are unique and have no '|'s in them
registerLabel :: String -> State -> String -> IO ()
registerLabel whence st nm
| isReserved nm
= err "is a reserved string; please use a different name."
| '|' `elem` nm
= err "contains the character `|', which is not allowed!"
| '\\' `elem` nm
= err "contains the character `\\', which is not allowed!"
| True
= do old <- readIORef $ rUsedLbls st
if nm `Set.member` old
then err "is used multiple times. Please do not use duplicate names!"
else modifyState st rUsedLbls (Set.insert nm) (return ())
where err w = error $ "SBV (" ++ whence ++ "): " ++ show nm ++ " " ++ w
-- | Create a new constant; hash-cons as necessary
newConst :: State -> CV -> IO SV
newConst st c = do
constMap <- readIORef (rconstMap st)
case c `Map.lookup` constMap of
-- NB. Unlike in 'newExpr', we don't have to make sure the returned sv
-- has the kind we asked for, because the constMap stores the full CV
-- which already has a kind field in it.
Just sv -> return sv
Nothing -> do (NamedSymVar sv _) <- newSV st (kindOf c)
let ins = Map.insert c sv
modifyState st rconstMap ins $ modifyIncState st rNewConsts ins
return sv
{-# INLINE newConst #-}
-- | Create a new table; hash-cons as necessary
getTableIndex :: State -> Kind -> Kind -> [SV] -> IO Int
getTableIndex st at rt elts = do
let key = (at, rt, elts)
tblMap <- readIORef (rtblMap st)
case key `Map.lookup` tblMap of
Just i -> return i
_ -> do let i = Map.size tblMap
upd = Map.insert key i
modifyState st rtblMap upd $ modifyIncState st rNewTbls upd
return i
-- | Create a new expression; hash-cons as necessary
newExpr :: State -> Kind -> SBVExpr -> IO SV
newExpr st k app = do
let e = reorder app
exprMap <- readIORef (rexprMap st)
case e `Map.lookup` exprMap of
-- NB. Check to make sure that the kind of the hash-consed value
-- is the same kind as we're requesting. This might look unnecessary,
-- at first, but `svSign` and `svUnsign` rely on this as we can
-- get the same expression but at a different type. See
-- <http://github.com/GaloisInc/cryptol/issues/566> as an example.
Just sv | kindOf sv == k -> return sv
_ -> do (NamedSymVar sv _) <- newSV st k
checkConsistent sv e
let append (SBVPgm xs) = SBVPgm (xs S.|> (sv, e))
modifyState st spgm append $ modifyIncState st rNewAsgns append
modifyState st rexprMap (Map.insert e sv) (return ())
return sv
{-# INLINE newExpr #-}
-- | In rare cases, we can get a context mismatch; so make sure the expression is well-formed.
-- This isn't a full solution, but handles the common case (hopefully!)
checkConsistent :: SV -> SBVExpr -> IO ()
checkConsistent lhs (SBVApp _ args) = mapM_ check args
where SV _ (NodeId (lhsContext, _, _)) = lhs
check (SV _ (NodeId (rhsContext, _, _)))
| lhsContext `compatibleContext` rhsContext
= pure ()
| True
= contextMismatchError lhsContext rhsContext
{-# INLINE checkConsistent #-}
-- | Are these compatible contexts? Either the same, or one of them is global
compatibleContext :: SBVContext -> SBVContext -> Bool
compatibleContext c1 c2 = c1 == c2 || c1 == globalSBVContext || c2 == globalSBVContext
{-# INLINE compatibleContext #-}
-- | Same as checkConsistent above, except in an array context
checkCompatibleContext :: SBVContext -> SBVContext -> IO ()
checkCompatibleContext ctx1 ctx2
| ctx1 `compatibleContext` ctx2
= pure ()
| True
= contextMismatchError ctx1 ctx2
{-# INLINE checkCompatibleContext #-}
-- | Convert a symbolic value to an internal SV
svToSV :: State -> SVal -> IO SV
svToSV st (SVal _ (Left c)) = newConst st c
svToSV st (SVal _ (Right f)) = uncache f st
-- | Generalization of 'Data.SBV.svToSymSV'
svToSymSV :: MonadSymbolic m => SVal -> m SV
svToSymSV sbv = do st <- symbolicEnv
liftIO $ svToSV st sbv
-------------------------------------------------------------------------
-- * Symbolic Computations
-------------------------------------------------------------------------
-- | A Symbolic computation. Represented by a reader monad carrying the
-- state of the computation, layered on top of IO for creating unique
-- references to hold onto intermediate results.
-- | Computations which support symbolic operations
class MonadIO m => MonadSymbolic m where
symbolicEnv :: m State
default symbolicEnv :: (MonadTrans t, MonadSymbolic m', m ~ t m') => m State
symbolicEnv = lift symbolicEnv
instance MonadSymbolic m => MonadSymbolic (ExceptT e m)
instance MonadSymbolic m => MonadSymbolic (MaybeT m)
instance MonadSymbolic m => MonadSymbolic (ReaderT r m)
instance MonadSymbolic m => MonadSymbolic (SS.StateT s m)
instance MonadSymbolic m => MonadSymbolic (LS.StateT s m)
instance (MonadSymbolic m, Monoid w) => MonadSymbolic (SW.WriterT w m)
instance (MonadSymbolic m, Monoid w) => MonadSymbolic (LW.WriterT w m)
-- | A generalization of 'Data.SBV.Symbolic'.
newtype SymbolicT m a = SymbolicT { runSymbolicT :: ReaderT State m a }
deriving newtype ( Applicative, Functor, Monad, MonadIO, MonadTrans
, MonadError e, MonadState s, MonadWriter w
#if MIN_VERSION_base(4,11,0)
, Fail.MonadFail
#endif
)
-- | `MonadSymbolic` instance for `SymbolicT m`
instance MonadIO m => MonadSymbolic (SymbolicT m) where
symbolicEnv = SymbolicT ask
-- | Map a computation over the symbolic transformer.
mapSymbolicT :: (ReaderT State m a -> ReaderT State n b) -> SymbolicT m a -> SymbolicT n b
mapSymbolicT f = SymbolicT . f . runSymbolicT
{-# INLINE mapSymbolicT #-}
-- Have to define this one by hand, because we use ReaderT in the implementation
instance MonadReader r m => MonadReader r (SymbolicT m) where
ask = lift ask
local f = mapSymbolicT $ mapReaderT $ local f
-- | 'Symbolic' is specialization of t'SymbolicT' to the `IO` monad. Unless you are using
-- transformers explicitly, this is the type you should prefer.
type Symbolic = SymbolicT IO
-- | Create a symbolic value, based on the quantifier we have. If an
-- explicit quantifier is given, we just use that. If not, then we
-- pick the quantifier appropriately based on the run-mode.
-- @randomCV@ is used for generating random values for this variable
-- when used for @quickCheck@ or 'Data.SBV.Tools.GenTest.genTest' purposes.
svMkSymVar :: VarContext -> Kind -> Maybe String -> State -> IO SVal
svMkSymVar = svMkSymVarGen False
-- | Create an existentially quantified tracker variable
svMkTrackerVar :: Kind -> String -> State -> IO SVal
svMkTrackerVar k nm = svMkSymVarGen True (NonQueryVar (Just EX)) k (Just nm)
-- | Generalization of 'Data.SBV.sWordN'
sWordN :: MonadSymbolic m => Int -> String -> m SVal
sWordN w nm = symbolicEnv >>= liftIO . svMkSymVar (NonQueryVar Nothing) (KBounded False w) (Just nm)
-- | Generalization of 'Data.SBV.sWordN_'
sWordN_ :: MonadSymbolic m => Int -> m SVal
sWordN_ w = symbolicEnv >>= liftIO . svMkSymVar (NonQueryVar Nothing) (KBounded False w) Nothing
-- | Generalization of 'Data.SBV.sIntN'
sIntN :: MonadSymbolic m => Int -> String -> m SVal
sIntN w nm = symbolicEnv >>= liftIO . svMkSymVar (NonQueryVar Nothing) (KBounded True w) (Just nm)
-- | Generalization of 'Data.SBV.sIntN_'
sIntN_ :: MonadSymbolic m => Int -> m SVal
sIntN_ w = symbolicEnv >>= liftIO . svMkSymVar (NonQueryVar Nothing) (KBounded True w) Nothing
-- | Create a symbolic value, based on the quantifier we have. If an
-- explicit quantifier is given, we just use that. If not, then we
-- pick the quantifier appropriately based on the run-mode.
-- @randomCV@ is used for generating random values for this variable
-- when used for @quickCheck@ or 'Data.SBV.Tools.GenTest.genTest' purposes.
svMkSymVarGen :: Bool -> VarContext -> Kind -> Maybe String -> State -> IO SVal
svMkSymVarGen isTracker varContext k mbNm st = do
registerKind st k
rm <- readIORef (runMode st)
let varInfo = case mbNm of
Nothing -> "While defining a variable of type " ++ show k
Just nm -> "While defining: " ++ nm ++ " :: " ++ show k
disallow what = error $ unlines [ "*** Data.SBV: Unsupported: " ++ what
, "***"
, "*** " ++ varInfo
, "*** "
, "*** In mode: " ++ show rm
]
(isQueryVar, mbQ) = case varContext of
NonQueryVar mq -> (False, mq)
QueryVar -> (True, Just EX)
mkS q = do (NamedSymVar sv internalName) <- newSV st k
let nm = fromMaybe (T.unpack internalName) mbNm
introduceUserName st (isQueryVar, isTracker) nm k q sv
mkC cv = do modifyState st rCInfo ((fromMaybe "_" mbNm, cv):) (return ())
return $ SVal k (Left cv)
case (mbQ, rm) of
(Just q, SMTMode{} ) -> mkS q
(Nothing, SMTMode _ _ isSAT _) -> mkS (if isSAT then EX else ALL)
(Just EX, CodeGen{}) -> disallow "Existentially quantified variables"
(_ , CodeGen) -> mkS ALL -- code generation, pick universal
(Just EX, Concrete Nothing) -> disallow "Existentially quantified variables"
(_ , Concrete Nothing) -> randomCV k >>= mkC
(Just EX, LambdaGen{}) -> disallow "Existentially quantified variables"
(_, LambdaGen{}) -> mkS ALL
-- Model validation:
(_ , Concrete (Just (_isSat, env))) -> do
let bad why conc = error $ unlines [ ""
, "*** Data.SBV: " ++ why
, "***"
, "*** To turn validation off, use `cfg{validateModel = False}`"
, "***"
, "*** " ++ conc
]
report = "Please report this as a bug in SBV!"
(NamedSymVar sv internalName) <- newSV st k
let nm = fromMaybe (T.unpack internalName) mbNm
nsv = toNamedSV' sv nm
-- Ignore the context equivalence check here. When validating, we are in a different
-- context; so they won't match
same (NamedSymVar (SV _ (NodeId (_, ll1, li1))) _)
(NamedSymVar (SV _ (NodeId (_, ll2, li2))) _) = (ll1, li1) == (ll2, li2)
cv = case [v | (nsv', v) <- env, nsv `same` nsv'] of
[] -> if isTracker
then -- The sole purpose of a tracker variable is to send the optimization
-- directive to the solver, so we can name "expressions" that are minimized
-- or maximized. There will be no constraints on these when we are doing
-- the validation; in fact they will not even be used anywhere during a
-- validation run. So, simply push a zero value that inhabits all metrics.
mkConstCV k (0::Integer)
else bad ("Cannot locate variable: " ++ show (nsv, k)) report
[c] -> c
r -> bad ( "Found multiple matching values for variable: " ++ show nsv
++ "\n*** " ++ show r) report
mkC cv
-- | Introduce a new user name. We simply append a suffix if we have seen this variable before.
introduceUserName :: State -> (Bool, Bool) -> String -> Kind -> Quantifier -> SV -> IO SVal
introduceUserName st@State{runMode} (isQueryVar, isTracker) nmOrig k q sv = do
old <- allInputs <$> readIORef (rinps st)
let nm = mkUnique (T.pack nmOrig) old
-- If this is not a query variable and we're in a query, reject it.
-- See https://github.com/LeventErkok/sbv/issues/554 for the rationale.
-- In theory, it should be possible to support this, but fixing it is
-- rather costly as we'd have to track the regular updates and sync the
-- incremental state appropriately. Instead, we issue an error message
-- and ask the user to obey the query mode rules.
rm <- readIORef runMode
case rm of
SMTMode _ IRun _ _ | not isQueryVar -> noInteractiveEver [ "Adding a new input variable in query mode: " ++ show nm
, ""
, "Hint: Use freshVar/freshVar_ for introducing new inputs in query mode."
]
_ -> pure ()
if isTracker && q == ALL
then error $ "SBV: Impossible happened! A universally quantified tracker variable is being introduced: " ++ show nm
else do let newInp olds = case q of
EX -> toNamedSV sv nm : olds
ALL -> noInteractive [ "Adding a new universally quantified variable: "
, " Name : " ++ show nm
, " Kind : " ++ show k
, " Quantifier: Universal"
, " Node : " ++ show sv
, "Only existential variables are supported in query mode."
]
if isTracker
then modifyState st rinps (addInternInput sv nm)
$ noInteractive ["Adding a new tracker variable in interactive mode: " ++ show nm]
else modifyState st rinps (addUserInput sv nm)
$ modifyIncState st rNewInps newInp
return $ SVal k $ Right $ cache (const (return sv))
where -- The following can be rather slow if we keep reusing the same prefix, but I doubt it'll be a problem in practice
-- Also, the following will fail if we span the range of integers without finding a match, but your computer would
-- die way ahead of that happening if that's the case!
mkUnique :: T.Text -> Set.Set Name -> T.Text
mkUnique prefix names = case dropWhile (`Set.member` names) (prefix : [prefix <> "_" <> T.pack (show i) | i <- [(0::Int)..]]) of
h:_ -> h
_ -> error $ "mkUnique: Impossible happened! Couldn't get a unique name for " ++ show (prefix, names)
-- | Create a new state
mkNewState :: MonadIO m => SMTConfig -> SBVRunMode -> m State
mkNewState cfg currentRunMode = liftIO $ do
currTime <- getCurrentTime
progInfo <- newIORef ProgInfo { hasQuants = False
, progSpecialRels = []
, progTransClosures = []
}
rm <- newIORef currentRunMode
ctr <- newIORef (-2) -- start from -2; False and True will always occupy the first two elements
fnctr <- newIORef 0
lambda <- newIORef $ case currentRunMode of
SMTMode{} -> Just 0
CodeGen{} -> Just 0
Concrete{} -> Just 0
LambdaGen mbi -> mbi
cInfo <- newIORef []
observes <- newIORef mempty
pgm <- newIORef (SBVPgm S.empty)
emap <- newIORef Map.empty
cmap <- newIORef Map.empty
inps <- newIORef mempty
lambdaInps <- newIORef mempty
outs <- newIORef []
tables <- newIORef Map.empty
userFuncs <- newIORef Set.empty
uis <- newIORef Map.empty
cgs <- newIORef Map.empty
defns <- newIORef []
swCache <- newIORef IMap.empty
usedKinds <- newIORef Set.empty
usedLbls <- newIORef Set.empty
cstrs <- newIORef S.empty
pvs <- newIORef []
smtOpts <- newIORef []
optGoals <- newIORef []
asserts <- newIORef []
outstandingAsserts <- newIORef False
istate <- newIORef =<< newIncState
qstate <- newIORef Nothing
ctx <- genSBVContext
pure $ State { sbvContext = ctx
, runMode = rm
, stCfg = cfg
, startTime = currTime
, rProgInfo = progInfo
, pathCond = SVal KBool (Left trueCV)
, rIncState = istate
, rCInfo = cInfo
, rObservables = observes
, rctr = ctr
, freshNameCtr = fnctr
, rLambdaLevel = lambda
, rUsedKinds = usedKinds
, rUsedLbls = usedLbls
, rinps = inps
, rlambdaInps = lambdaInps
, routs = outs
, rtblMap = tables
, spgm = pgm
, rconstMap = cmap
, rexprMap = emap
, rUserFuncs = userFuncs
, rUIMap = uis
, rCgMap = cgs
, rDefns = defns
, rSVCache = swCache
, rConstraints = cstrs
, rPartitionVars = pvs
, rSMTOptions = smtOpts
, rOptGoals = optGoals
, rAsserts = asserts
, rOutstandingAsserts = outstandingAsserts
, rQueryState = qstate
, parentState = Nothing
}
-- | Generalization of 'Data.SBV.runSymbolic'
runSymbolic :: MonadIO m => SMTConfig -> SBVRunMode -> SymbolicT m a -> m (a, Result)
runSymbolic cfg currentRunMode comp = do
st <- mkNewState cfg currentRunMode
runSymbolicInState st comp
-- | Catch the catastrophic case of context mismatch
-- NB. We're not printing _ctx1/_ctx2 here (hence the underscored variables).
-- The reason is that they can get different values; causing test-suite failures with no helpful info.
contextMismatchError :: SBVContext -> SBVContext -> a
contextMismatchError _ctx1 _ctx2 = error $ unlines [
"Data.SBV: Mismatched contexts detected."
, "***"
, "*** This happens if you call a proof-function (prove/sat/runSMT/isSatisfiable) etc."
, "*** while another one is in execution, or use results from one such call in another."
, "*** Please avoid such nested calls, all interactions should be from the same context."
, "*** See https://github.com/LeventErkok/sbv/issues/71 for several examples."
]
-- | Run a symbolic computation in a given state
runSymbolicInState :: MonadIO m => State -> SymbolicT m a -> m (a, Result)
runSymbolicInState st (SymbolicT c) = do
_ <- liftIO $ newConst st falseCV -- s(-2) == falseSV
_ <- liftIO $ newConst st trueCV -- s(-1) == trueSV
r <- runReaderT c st
res <- liftIO $ extractSymbolicSimulationState st
-- Check that the state wasn't clobbered in any way
let check ctx | ctx == sbvContext st || ctx == globalSBVContext
= pure ()
| True
= contextMismatchError (sbvContext st) ctx
mapM_ check $ nubOrd $ G.universeBi res
return (r, res)
-- | Grab the program from a running symbolic simulation state.
extractSymbolicSimulationState :: State -> IO Result
extractSymbolicSimulationState st@State{ runMode=rrm
, spgm=pgm, rinps=inps, rlambdaInps=linps, routs=outs, rtblMap=tables
, rUIMap=uis, rDefns=defns
, rAsserts=asserts, rUsedKinds=usedKinds, rCgMap=cgs, rCInfo=cInfo, rConstraints=cstrs
, rObservables=observes, rProgInfo=progInfo
} = do
SBVPgm rpgm <- readIORef pgm
rm <- readIORef rrm
inpsO <- do Inputs{userInputs, internInputs} <- readIORef inps
ls <- readIORef linps
let lambdaOnly = case rm of
SMTMode{} -> False
CodeGen{} -> False
Concrete{} -> False
LambdaGen{} -> True
topInps = (F.toList userInputs, F.toList internInputs)
lamInps = F.toList ls
if lambdaOnly
then case topInps of
([], []) -> pure $ ResultLamInps (F.toList ls)
(xs, ys) -> error $ unlines [ ""
, "*** Data.SBV: Impossible happened; saw inputs in lambda mode."
, "***"
, "*** Inps : " ++ show xs
, "*** Trackers: " ++ show ys
]
else case lamInps of
[] -> pure $ ResultTopInps topInps
_ -> error $ unlines [ ""
, "*** Data.SBV: Impossible happened; saw lambda inputs in regular mode."
, "***"
, "*** Params: " ++ show lamInps
]
outsO <- reverse <$> readIORef outs
let swap (a, b) = (b, a)
cmp (a, _) (b, _) = a `compare` b
arrange (i, (at, rt, es)) = ((i, at, rt), es)
constMap <- readIORef (rconstMap st)
let cnsts = sortBy cmp . map swap . Map.toList $ constMap
tbls <- map arrange . sortBy cmp . map swap . Map.toList <$> readIORef tables
ds <- reverse <$> readIORef defns
unint <- do unints <- Map.toList <$> readIORef uis
-- drop those that has a definition associated with it
let defineds = map fst ds
pure [ui | ui@(nm, _) <- unints, nm `notElem` defineds]
knds <- readIORef usedKinds
cgMap <- Map.toList <$> readIORef cgs
traceVals <- reverse <$> readIORef cInfo
observables <- reverse . fmap (\(n,f,sv) -> (T.unpack n, f, sv)) . F.toList
<$> readIORef observes
extraCstrs <- readIORef cstrs
assertions <- reverse <$> readIORef asserts
pinfo <- readIORef progInfo
return $ Result pinfo knds traceVals observables cgMap inpsO (constMap, cnsts) tbls unint ds (SBVPgm rpgm) extraCstrs assertions outsO
-- | Generalization of 'Data.SBV.addNewSMTOption'
addNewSMTOption :: MonadSymbolic m => SMTOption -> m ()
addNewSMTOption o = do st <- symbolicEnv
liftIO $ modifyState st rSMTOptions (o:) (return ())
-- | Generalization of 'Data.SBV.imposeConstraint'
imposeConstraint :: MonadSymbolic m => Bool -> [(String, String)] -> SVal -> m ()
imposeConstraint isSoft attrs c = do st <- symbolicEnv
rm <- liftIO $ readIORef (runMode st)
case rm of
CodeGen -> error "SBV: constraints are not allowed in code-generation"
_ -> liftIO $ do mapM_ (registerLabel "Constraint" st) [nm | (":named", nm) <- attrs]
internalConstraint st isSoft attrs c
-- | Require a boolean condition to be true in the state. Only used for internal purposes.
internalConstraint :: State -> Bool -> [(String, String)] -> SVal -> IO ()
internalConstraint st isSoft attrs b = do v <- svToSV st b
rm <- liftIO $ readIORef (runMode st)
-- Are we running validation? If so, we always want to
-- add the constraint for debug purposes. Otherwise
-- we only add it if it's interesting; i.e., not directly
-- true or has some attributes.
let isValidating = case rm of
SMTMode _ _ _ cfg -> validationRequested cfg
CodeGen -> False
LambdaGen{} -> False
Concrete Nothing -> False
Concrete (Just _) -> True -- The case when we *are* running the validation
let c = (isSoft, attrs, v)
interesting = v /= trueSV || not (null attrs)
when (isValidating || interesting) $
modifyState st rConstraints (S.|> c)
$ modifyIncState st rNewConstraints (S.|> c)
-- | Generalization of 'Data.SBV.addSValOptGoal'
addSValOptGoal :: MonadSymbolic m => Objective SVal -> m ()
addSValOptGoal obj = do st <- symbolicEnv
-- create the tracking variable here for the metric
let mkGoal nm orig = liftIO $ do origSV <- svToSV st orig
track <- svMkTrackerVar (kindOf orig) nm st
trackSV <- svToSV st track
return (origSV, trackSV)
let walk (Minimize nm v) = Minimize nm <$> mkGoal nm v
walk (Maximize nm v) = Maximize nm <$> mkGoal nm v
walk (AssertWithPenalty nm v mbP) = flip (AssertWithPenalty nm) mbP <$> mkGoal nm v
!obj' <- walk obj
liftIO $ modifyState st rOptGoals (obj' :)
$ noInteractive [ "Adding an optimization objective:"
, " Objective: " ++ show obj
]
-- | Generalization of 'Data.SBV.sObserve'
sObserve :: MonadSymbolic m => String -> SVal -> m ()
sObserve m x
| Just bad <- checkObservableName m
= error bad
| True
= do st <- symbolicEnv
liftIO $ do xsv <- svToSV st x
recordObservable st m (const True) xsv
-- | Generalization of 'Data.SBV.outputSVal'
outputSVal :: MonadSymbolic m => SVal -> m ()
outputSVal (SVal _ (Left c)) = do
st <- symbolicEnv
sv <- liftIO $ newConst st c
liftIO $ modifyState st routs (sv:) (return ())
outputSVal (SVal _ (Right f)) = do
st <- symbolicEnv
sv <- liftIO $ uncache f st
liftIO $ modifyState st routs (sv:) (return ())
---------------------------------------------------------------------------------
-- * Cached values
---------------------------------------------------------------------------------
-- | We implement a peculiar caching mechanism, applicable to the use case in
-- implementation of SBV's. Whenever we do a state based computation, we do
-- not want to keep on evaluating it in the then-current state. That will
-- produce essentially a semantically equivalent value. Thus, we want to run
-- it only once, and reuse that result, capturing the sharing at the Haskell
-- level. This is similar to the "type-safe observable sharing" work, but also
-- takes into the account of how symbolic simulation executes.
--
-- See Andy Gill's type-safe observable sharing trick for the inspiration behind
-- this technique: <http://ku-fpg.github.io/files/Gill-09-TypeSafeReification.pdf>
--
-- Note that this is *not* a general memo utility!
newtype Cached a = Cached (State -> IO a)
-- | Cache a state-based computation
cache :: (State -> IO a) -> Cached a
cache = Cached
-- | Uncache a previously cached computation
uncache :: Cached SV -> State -> IO SV
uncache = uncacheGen rSVCache
-- | Generic uncaching. Note that this is entirely safe, since we do it in the IO monad.
uncacheGen :: (State -> IORef (Cache a)) -> Cached a -> State -> IO a
uncacheGen getCache (Cached f) st = do
let rCache = getCache st
stored <- readIORef rCache
sn <- f `seq` makeStableName f
let h = hashStableName sn
case (h `IMap.lookup` stored) >>= (sn `lookup`) of
Just r -> return r
Nothing -> do r <- f st
r `seq` R.modifyIORef' rCache (IMap.insertWith (++) h [(sn, r)])
return r
-- | Representation of SMTLib Program versions. As of June 2015, we're dropping support
-- for SMTLib1, and supporting SMTLib2 only. We keep this data-type around in case
-- SMTLib3 comes along and we want to support 2 and 3 simultaneously.
data SMTLibVersion = SMTLib2
deriving (Bounded, Enum, Eq, Show)
-- | The extension associated with the version
smtLibVersionExtension :: SMTLibVersion -> String
smtLibVersionExtension SMTLib2 = "smt2"
-- | Representation of an SMT-Lib program. The second [String] are the function definitions,
-- which is *replicated* in the first one. There are cases where that we need the second part on its own.
data SMTLibPgm = SMTLibPgm SMTLibVersion [String] [String]
instance NFData SMTLibVersion where rnf a = a `seq` ()
instance NFData SMTLibPgm where rnf (SMTLibPgm v p d) = rnf v `seq` rnf p `seq` rnf d
instance Show SMTLibPgm where
show (SMTLibPgm _ pgm _) = intercalate "\n" pgm
-- Other Technicalities..
instance NFData GeneralizedCV where
rnf (ExtendedCV e) = e `seq` ()
rnf (RegularCV c) = c `seq` ()
#if MIN_VERSION_base(4,9,0)
#else
-- Can't really force this, but not a big deal
instance NFData CallStack where
rnf _ = ()
#endif
instance NFData NamedSymVar where
rnf (NamedSymVar s n) = rnf s `seq` rnf n
instance NFData Result where
rnf (Result hasQuants kindInfo qcInfo obs cgs inps consts tbls uis axs pgm cstr asserts outs)
= rnf hasQuants `seq` rnf kindInfo `seq` rnf qcInfo `seq` rnf obs `seq` rnf cgs
`seq` rnf inps `seq` rnf consts `seq` rnf tbls
`seq` rnf uis `seq` rnf axs `seq` rnf pgm
`seq` rnf cstr `seq` rnf asserts `seq` rnf outs
instance NFData SV where rnf a = seq a ()
instance NFData SBVExpr where rnf a = seq a ()
instance NFData Quantifier where rnf a = seq a ()
instance NFData SBVType where rnf a = seq a ()
instance NFData SBVPgm where rnf a = seq a ()
instance NFData (Cached a) where rnf (Cached f) = f `seq` ()
instance NFData SVal where rnf (SVal x y) = rnf x `seq` rnf y
instance NFData SMTResult where
rnf (Unsatisfiable _ m ) = rnf m
rnf (Satisfiable _ m ) = rnf m
rnf (DeltaSat _ p m ) = rnf m `seq` rnf p
rnf (SatExtField _ m ) = rnf m
rnf (Unknown _ m ) = rnf m
rnf (ProofError _ m mr) = rnf m `seq` rnf mr
instance NFData SMTModel where
rnf (SMTModel objs bndgs assocs uifuns) = rnf objs `seq` rnf bndgs `seq` rnf assocs `seq` rnf uifuns
instance NFData SMTScript where
rnf (SMTScript b m) = rnf b `seq` rnf m
-- | Translation tricks needed for specific capabilities afforded by each solver
data SolverCapabilities = SolverCapabilities {
supportsQuantifiers :: Bool -- ^ Supports SMT-Lib2 style quantifiers?
, supportsDefineFun :: Bool -- ^ Supports define-fun construct?
, supportsDistinct :: Bool -- ^ Supports calls to distinct?
, supportsBitVectors :: Bool -- ^ Supports bit-vectors?
, supportsADTs :: Bool -- ^ Supports SMT-Lib2 style uninterpreted-sorts and ADTs
, supportsUnboundedInts :: Bool -- ^ Supports unbounded integers?
, supportsReals :: Bool -- ^ Supports reals?
, supportsApproxReals :: Bool -- ^ Supports printing of approximations of reals?
, supportsDeltaSat :: Maybe String -- ^ Supports delta-satisfiability? (With given precision query)
, supportsIEEE754 :: Bool -- ^ Supports floating point numbers?
, supportsSets :: Bool -- ^ Supports set operations?
, supportsOptimization :: Bool -- ^ Supports optimization routines?
, supportsPseudoBooleans :: Bool -- ^ Supports pseudo-boolean operations?
, supportsCustomQueries :: Bool -- ^ Supports interactive queries per SMT-Lib?
, supportsGlobalDecls :: Bool -- ^ Supports global declarations? (Needed for push-pop.)
, supportsDataTypes :: Bool -- ^ Supports datatypes?
, supportsLambdas :: Bool -- ^ Does it support lambdas?
, supportsSpecialRels :: Bool -- ^ Does it support special relations (orders, transitive closure etc.)
, supportsDirectTesters :: Bool -- ^ Supports data-type testers without full ascription?
, supportsFlattenedModels :: Maybe [String] -- ^ Supports flattened model output? (With given config lines.)
}
-- | Solver configuration. See also 'Data.SBV.z3', 'Data.SBV.yices', 'Data.SBV.cvc4', 'Data.SBV.boolector', 'Data.SBV.mathSAT', etc.
-- which are instantiations of this type for those solvers, with reasonable defaults. In particular, custom configuration can be
-- created by varying those values. (Such as @z3{verbose=True}@.)
--
-- Most fields are self explanatory. The notion of precision for printing algebraic reals stems from the fact that such values does
-- not necessarily have finite decimal representations, and hence we have to stop printing at some depth. It is important to
-- emphasize that such values always have infinite precision internally. The issue is merely with how we print such an infinite
-- precision value on the screen. The field 'printRealPrec' controls the printing precision, by specifying the number of digits after
-- the decimal point. The default value is 16, but it can be set to any positive integer.
--
-- When printing, SBV will add the suffix @...@ at the end of a real-value, if the given bound is not sufficient to represent the real-value
-- exactly. Otherwise, the number will be written out in standard decimal notation. Note that SBV will always print the whole value if it
-- is precise (i.e., if it fits in a finite number of digits), regardless of the precision limit. The limit only applies if the representation
-- of the real value is not finite, i.e., if it is not rational.
--
-- The 'printBase' field can be used to print numbers in base 2, 10, or 16.
--
-- The 'crackNum' field can be used to display numbers in detail, all its bits and how they are laid out in memory. Works with all bounded number types
-- (i.e., SWord and SInt), but also with floats. It is particularly useful with floating-point numbers, as it shows you how they are laid out in
-- memory following the IEEE754 rules.
data SMTConfig = SMTConfig {
verbose :: Bool -- ^ Debug mode
, timing :: Timing -- ^ Print timing information on how long different phases took (construction, solving, etc.)
, printBase :: Int -- ^ Print integral literals in this base (2, 10, and 16 are supported.)
, printRealPrec :: Int -- ^ Print algebraic real values with this precision. (SReal, default: 16)
, crackNum :: Bool -- ^ For each numeric value, show it in detail in the model with its bits spliced out. Good for floats.
, crackNumSurfaceVals :: [(String, Integer)] -- ^ For crackNum: The surface representation of variables, if available
, satCmd :: String -- ^ Usually "(check-sat)". However, users might tweak it based on solver characteristics.
, allSatMaxModelCount :: Maybe Int -- ^ In a 'Data.SBV.allSat' call, return at most this many models. If nothing, return all.
, allSatPrintAlong :: Bool -- ^ In a 'Data.SBV.allSat' call, print models as they are found.
, allSatTrackUFs :: Bool -- ^ In a 'Data.SBV.allSat' call, should we try to extract values of uninterpreted functions?
, isNonModelVar :: String -> Bool -- ^ When constructing a model, ignore variables whose name satisfy this predicate. (Default: (const False), i.e., don't ignore anything)
, validateModel :: Bool -- ^ If set, SBV will attempt to validate the model it gets back from the solver.
, optimizeValidateConstraints :: Bool -- ^ Validate optimization results. NB: Does NOT make sure the model is optimal, just checks they satisfy the constraints.
, transcript :: Maybe FilePath -- ^ If Just, the entire interaction will be recorded as a playable file (for debugging purposes mostly)
, smtLibVersion :: SMTLibVersion -- ^ What version of SMT-lib we use for the tool
, dsatPrecision :: Maybe Double -- ^ Delta-sat precision
, solver :: SMTSolver -- ^ The actual SMT solver.
, extraArgs :: [String] -- ^ Extra command line arguments to pass to the solver.
, roundingMode :: RoundingMode -- ^ Rounding mode to use for floating-point conversions
, solverSetOptions :: [SMTOption] -- ^ Options to set as we start the solver
, ignoreExitCode :: Bool -- ^ If true, we shall ignore the exit code upon exit. Otherwise we require ExitSuccess.
, redirectVerbose :: Maybe FilePath -- ^ Redirect the verbose output to this file if given. If Nothing, stdout is implied.
, firstifyUniqueLen :: Int -- ^ Unique length used for firstified higher-order function names
, tpOptions :: TPOptions -- ^ TP specific options
}
-- | Configuration for TP
data TPOptions = TPOptions {
ribbonLength :: Int -- ^ Line length for TP proofs
, quiet :: Bool -- ^ No messages what-so-ever for successful steps. (Will print if something fails)
, printAsms :: Bool -- ^ Print assumptions as they are proven as separate steps.
, printStats :: Bool -- ^ Print time/statistics. If quiet is True, then measureTime is ignored.
, cacheProofs :: Bool -- ^ Treat lemma names as unique, and cache the results. Default: False. Note that this
-- feature is unsound unless you make sure (by some other mechanism) that your lemma names
-- are indeed unique.
}
-- | Ignore internal names and those the user told us to
mustIgnoreVar :: SMTConfig -> String -> Bool
mustIgnoreVar cfg s = "__internal_sbv" `isPrefixOf` s || isNonModelVar cfg s
-- | We show the name of the solver for the config. Arguably this is misleading, but better than nothing.
instance Show SMTConfig where
show = show . name . solver
-- | Returns true if we have to perform validation
validationRequested :: SMTConfig -> Bool
validationRequested SMTConfig{validateModel, optimizeValidateConstraints} = validateModel || optimizeValidateConstraints
-- We're just seq'ing top-level here, it shouldn't really matter. (i.e., no need to go deeper.)
instance NFData SMTConfig where
rnf SMTConfig{} = ()
-- | A model, as returned by a solver
data SMTModel = SMTModel {
modelObjectives :: [(String, GeneralizedCV)] -- ^ Mapping of symbolic values to objective values.
, modelBindings :: Maybe [(NamedSymVar, CV)] -- ^ Mapping of input variables as reported by the solver. Only collected if model validation is requested.
, modelAssocs :: [(String, CV)] -- ^ Mapping of symbolic values to constants.
, modelUIFuns :: [(String, (Bool, SBVType, Either String ([([CV], CV)], CV)))] -- ^ Mapping of uninterpreted functions to association lists in the model.
-- Note that an uninterpreted constant (function of arity 0) will be stored
-- in the 'modelAssocs' field. Left is used when the function returned is too
-- difficult for SBV to figure out what it means
}
deriving Show
-- | Is it the case that the model is really uninteresting? This is the case when there are no assocs nor ui's
isEmptyModel :: SMTModel -> Bool
isEmptyModel SMTModel{modelAssocs, modelUIFuns} = null modelAssocs && null modelUIFuns
-- | The result of an SMT solver call. Each constructor is tagged with
-- the t'SMTConfig' that created it so that further tools can inspect it
-- and build layers of results, if needed. For ordinary uses of the library,
-- this type should not be needed, instead use the accessor functions on
-- it. (Custom Show instances and model extractors.)
data SMTResult = Unsatisfiable SMTConfig (Maybe [String]) -- ^ Unsatisfiable. If unsat-cores are enabled, they will be returned in the second parameter.
| Satisfiable SMTConfig SMTModel -- ^ Satisfiable with model
| DeltaSat SMTConfig (Maybe String) SMTModel -- ^ Delta satisfiable with queried string if available and model
| SatExtField SMTConfig SMTModel -- ^ Prover returned a model, but in an extension field containing Infinite/epsilon
| Unknown SMTConfig SMTReasonUnknown -- ^ Prover returned unknown, with the given reason
| ProofError SMTConfig [String] (Maybe SMTResult) -- ^ Prover errored out, with possibly a bogus result
-- | A script, to be passed to the solver.
data SMTScript = SMTScript {
scriptBody :: String -- ^ Initial feed
, scriptModel :: [String] -- ^ Continuation script, to extract results
}
-- | An SMT engine
type SMTEngine = forall res.
SMTConfig -- ^ current configuration
-> State -- ^ the state in which to run the engine
-> String -- ^ program
-> (State -> IO res) -- ^ continuation
-> IO res
-- | Solvers that SBV is aware of
data Solver = ABC
| Boolector
| Bitwuzla
| CVC4
| CVC5
| DReal
| MathSAT
| Yices
| Z3
| OpenSMT
deriving (Show, Enum, Bounded)
-- | An SMT solver
data SMTSolver = SMTSolver {
name :: Solver -- ^ The solver in use
, executable :: String -- ^ The path to its executable
, preprocess :: String -> String -- ^ Each line sent to the solver will be passed through this function (typically id)
, options :: SMTConfig -> [String] -- ^ Options to provide to the solver
, engine :: SMTEngine -- ^ The solver engine, responsible for interpreting solver output
, capabilities :: SolverCapabilities -- ^ Various capabilities of the solver
}
-- | Query execution context
data QueryContext = QueryInternal -- ^ Triggered from inside SBV
| QueryExternal -- ^ Triggered from user code
-- | Show instance for 'QueryContext', for debugging purposes
instance Show QueryContext where
show QueryInternal = "Internal Query"
show QueryExternal = "User Query"
{- HLint ignore type FPOp "Use camelCase" -}
{- HLint ignore type PBOp "Use camelCase" -}
{- HLint ignore type OvOp "Use camelCase" -}
{- HLint ignore type NROp "Use camelCase" -}