{-# OPTIONS_GHC -fno-warn-orphans #-}
{-# LANGUAGE OverloadedStrings #-}
{-# LANGUAGE TupleSections #-}
{-# LANGUAGE LambdaCase #-}
{-# LANGUAGE CPP #-}

-- |
-- Module       : Data.TPTP.Parse.Combinators
-- Description  : Parser combinators for the TPTP language.
-- Copyright    : (c) Evgenii Kotelnikov, 2019
-- License      : GPL-3
-- Maintainer   : evgeny.kotelnikov@gmail.com
-- Stability    : experimental
--

module Data.TPTP.Parse.Combinators (
  -- * Whitespace
  skipWhitespace,
  input,

  -- * Names
  atom,
  var,
  distinctObject,
  function,
  predicate,

  -- * Sorts and types
  sort,
  tff1Sort,
  type_,

  -- * First-order logic
  number,
  term,
  literal,
  clause,
  unsortedFirstOrder,
  sortedFirstOrder,
  monomorphicFirstOrder,
  polymorphicFirstOrder,

  -- * Units
  unit,
  tptp,
  tstp,

  -- * Annotations
  szs,
  intro,
  parent,
  source,
  info
) where

#if MIN_VERSION_base(4, 8, 0)
import Prelude hiding (pure, (<$>), (<*>), (*>), (<*))
#endif

import Control.Applicative (pure, (<*>), (*>), (<*), (<|>), optional, empty, many)
import Data.Char (isAscii, isAsciiLower, isAsciiUpper, isDigit, isPrint)
import Data.Function (on)
import Data.Functor ((<$>), ($>))
import Data.Maybe (fromMaybe)
#if !MIN_VERSION_base(4, 8, 0)
import Data.Monoid (Monoid(..))
#endif
import Data.List (sortBy, genericLength)
import Data.List.NonEmpty (NonEmpty)
import qualified Data.List.NonEmpty as NEL (fromList, toList)
#if !MIN_VERSION_base(4, 11, 0)
import Data.Semigroup (Semigroup(..), sconcat)
#else
import Data.Semigroup (sconcat)
#endif

import qualified Data.Scientific as Sci (base10Exponent, coefficient)

import Data.Text (Text)
import qualified Data.Text as Text (pack, unpack, cons)

import Data.Attoparsec.Text as Atto (
    Parser,
    (<?>), char, string, decimal, scientific, signed, isEndOfLine, endOfLine,
    satisfy, option, eitherP, choice, manyTill, takeWhile, skip, skipSpace,
    skipMany, skipWhile, endOfInput, sepBy, sepBy1
  )

import Data.TPTP hiding (name, clause)
import qualified Data.TPTP as TPTP (name)


-- * Helper functions

-- | Consume all character until the end of line.
skipLine :: Parser ()
skipLine = skipWhile (not . isEndOfLine)
{-# INLINE skipLine #-}

-- | Consume the first character of a single line comment - @%@ or @#@.
-- The grammar of the TPTP language only defines @%@,
-- but some theorem provers in addition use @#@.
skipBeginComment :: Parser ()
skipBeginComment = skip (\c -> c == '%' || c == '#')
{-# INLINE skipBeginComment #-}

-- | Parse the contents of a single line comment.
commented :: Parser p -> Parser p
commented p =  skipBeginComment *> p <* skipLine <* (endOfLine <|> endOfInput)
           <?> "commented"

-- | Consume a single line comment - characters between @%@ or @#@ and newline.
skipComment :: Parser ()
skipComment = commented (pure ()) <?> "comment"
{-# INLINE skipComment #-}

-- | Consume a block comment - characters between /* and */.
skipBlockComment :: Parser ()
skipBlockComment = string "/*" *> bc <?> "block comment"
  where
    bc = skipWhile (/= '*') *> (string "*/" $> () <|> bc)

-- | Consume white space and trailing comments.
skipWhitespace :: Parser ()
skipWhitespace =  skipSpace
               *> skipMany ((skipComment <|> skipBlockComment) *> skipSpace)
              <?> "whitespace"

-- | @lexeme@ makes a given parser consume trailing whitespace. This function is
-- needed because off-the-shelf attoparsec parsers do not do it.
lexeme :: Parser a -> Parser a
lexeme p = p <* skipWhitespace
{-# INLINE lexeme #-}

-- | @input@ runs a given parser skipping leading whitespace. The function
-- succeeds if the parser consumes the entire input.
input :: Parser a -> Parser a
input p = skipWhitespace *> p <* endOfInput <?> "input"
{-# INLINE input #-}

-- | Parse an unsigned integer.
integer :: Parser Integer
integer = lexeme decimal <?> "integer"
{-# INLINE integer #-}

token :: Text -> Parser Text
token t = lexeme (string t) <?> "token " ++ Text.unpack t
{-# INLINE token #-}

op :: Char -> Parser Char
op c = lexeme (char c) <?> "operator " ++ [c]
{-# INLINE op #-}

parens :: Parser a -> Parser a
parens p = op '(' *> p <* op ')' <?> "parens"
{-# INLINE parens #-}

optionalParens :: Parser a -> Parser a
optionalParens p = p <|> parens (optionalParens p)
{-# INLINE optionalParens #-}

brackets :: Parser a -> Parser a
brackets p = op '[' *> p <* op ']' <?> "brackets"
{-# INLINE brackets #-}

bracketList :: Parser a -> Parser [a]
bracketList p = brackets (p `sepBy` op ',') <?> "bracket list"
{-# INLINE bracketList #-}

bracketList1 :: Parser a -> Parser (NonEmpty a)
bracketList1 p =  NEL.fromList <$> brackets (p `sepBy1` op ',')
              <?> "bracket list 1"
{-# INLINE bracketList1 #-}

application :: Parser f -> Parser a -> Parser (f, [a])
application f a = (,) <$> f <*> option [] (parens (optionalParens a `sepBy1` op ','))
{-# INLINE application #-}

labeled :: Text -> Parser a -> Parser a
labeled l p = token l *> parens p
{-# INLINE labeled #-}

comma :: Parser a -> Parser a
comma p = op ',' *> p
{-# INLINE comma #-}

maybeP :: Parser a -> Parser (Maybe a)
maybeP = optional . comma
{-# INLINE maybeP #-}

named :: (Named a, Enum a, Bounded a) => Parser a
named = choice
      $ fmap (\(n, c) -> string n $> c <?> "named " ++ Text.unpack n)
      $ sortBy (flip compare `on` fst)
      $ fmap (\c -> (TPTP.name c, c)) [minBound..]

enum :: (Named a, Enum a, Bounded a) => Parser a
enum = lexeme named
{-# INLINE enum #-}


-- * Parser combinators

-- ** Names

isAlphaNumeric :: Char -> Bool
isAlphaNumeric c = isAsciiLower c || isAsciiUpper c || isDigit c || c == '_'

isAsciiPrint :: Char -> Bool
isAsciiPrint c = isAscii c && isPrint c

lowerWord, upperWord :: Parser Text
lowerWord = Text.cons <$> satisfy isAsciiLower <*> Atto.takeWhile isAlphaNumeric
upperWord = Text.cons <$> satisfy isAsciiUpper <*> Atto.takeWhile isAlphaNumeric

quoted :: Char -> Parser Text
quoted q =  Text.pack <$> (char q *> manyTill escaped (char q))
        <?> "quoted " ++ [q]
  where
    escaped =  char '\\' *> (char q $> q <|> char '\\' $> '\\')
           <|> satisfy isAsciiPrint

-- | Parse an atomic word. Single-quoted atoms are parsed without the single
-- quotes and with the characters @'@ and @\\@ unescaped.
atom :: Parser Atom
atom = Atom <$> lexeme (quoted '\'' <|> lowerWord) <?> "atom"
{-# INLINE atom #-}

-- | Parse a variable.
var :: Parser Var
var = Var <$> lexeme upperWord <?> "var"
{-# INLINE var #-}

-- | Parse a distinct object. Double quotes are not preserved and the characters
-- @'@ and @\\@ are unescaped.
distinctObject :: Parser DistinctObject
distinctObject = DistinctObject <$> lexeme (quoted '"') <?> "distinct object"
{-# INLINE distinctObject #-}

-- | Parse a reserved word.
reserved :: (Named a, Enum a, Bounded a) => Parser (Reserved a)
reserved = extended <$> lexeme lowerWord <?> "reserved"
{-# INLINE reserved #-}

name :: (Named a, Enum a, Bounded a) => Parser (Name a)
name =  Reserved <$> (char '$' *> reserved)
    <|> Defined  <$> atom
    <?> "name"

-- | Parse a function name.
function :: Parser (Name Function)
function = name <?> "function"
{-# INLINE function #-}

-- | Parse a predicate name.
predicate :: Parser (Name Predicate)
predicate = name <?> "predicate"
{-# INLINE predicate #-}


-- ** Sorts and typess

-- | Parse a sort.
sort :: Parser (Name Sort)
sort = name <?> "sort"
{-# INLINE sort #-}

-- | Parse a sort in sorted polymorphic logic.
tff1Sort :: Parser TFF1Sort
tff1Sort =  SortVariable <$> var
        <|> uncurry TFF1Sort <$> application sort tff1Sort
        <?> "tff1 sort"

mapping :: Parser a -> Parser ([a], a)
mapping s = (,) <$> option [] (args <* op '>') <*> s
  where
    args = fmap (:[]) s <|> parens (s `sepBy1` op '*')

-- | Parse a type.
type_ :: Parser Type
type_ =  uncurry . tff1Type
     <$> (maybe [] NEL.toList <$> optional prefix) <*> matrix
     <?> "type"
  where
    prefix = token "!>" *> bracketList1 sortVar <* op ':'
    sortVar = var <* op ':' <* optionalParens (token "$tType")
    matrix = optionalParens (mapping (optionalParens tff1Sort))


-- ** First-order logic

-- | Parse a number.
number :: Parser Number
number =  RationalConstant <$> signed integer <* char '/' <*> integer
      <|> real <$> lexeme scientific
      <?> "number"
  where
    real n
      | Sci.base10Exponent n == 0 = IntegerConstant (Sci.coefficient n)
      | otherwise = RealConstant n

-- | Parse a term.
--
-- @term@ supports parsing superfluous parenthesis around arguments
-- of the function application, which are not present in the TPTP grammar.
term :: Parser Term
term =  uncurry Function <$> application function term
    <|> Variable         <$> var
    <|> Number           <$> number
    <|> DistinctTerm     <$> distinctObject
    <?> "term"

-- | Parse the equality and the inequality sign.
eq :: Parser Sign
eq = enum <?> "eq"
{-# INLINE eq #-}

-- | Parse a literal.
--
-- @literal@ supports parsing superfluous parenthesis around arguments
-- of the predicate application, which are not present in the TPTP grammar.
literal :: Parser Literal
literal =  Equality <$> optionalParens term <*> eq <*> optionalParens term
       <|> uncurry Predicate <$> application predicate term
       <?> "literal"

-- | Parse a signed literal.
--
-- @signedLiteral@ supports parsing superfluous parenthesis around the literal
-- under the negation sign, which are not present in the TPTP grammar.
signedLiteral :: Parser (Sign, Literal)
signedLiteral =  fmap (Negative,) (op '~' *> optionalParens literal)
             <|> fmap (Positive,) literal
             <?> "signed literal"
{-# INLINE signedLiteral #-}

-- | Parse a clause.
--
-- @clause@ supports parsing superfluous parenthesis around any of the
-- subclauses of the clause, which are not present in the TPTP grammar.
clause :: Parser Clause
clause = sconcat . NEL.fromList <$> subclause `sepBy1` op '|'
  where
    subclause =  unitClause <$> signedLiteral
             <|> parens clause
             <?> "subclause"

-- | Parse a quantifier.
quantifier :: Parser Quantifier
quantifier = enum <?> "quantifier"
{-# INLINE quantifier #-}

-- | Parse a logical connective.
connective :: Parser Connective
connective = enum <?> "connective"
{-# INLINE connective #-}

-- | Given a parser for sort annotations, parse a formula in first-order logic.
firstOrder :: Parser s -> Parser (FirstOrder s)
firstOrder p = do
  f <- unitary
  option f (Connected f <$> connective <*> firstOrder p)
  where
    unitary =  Atomic     <$> literal
           <|> Negated    <$> (op '~' *> unitary)
           <|> Quantified <$> quantifier <*> vs <* op ':' <*> unitary
           <|> parens (firstOrder p)
           <?> "unitary first order"

    vs = bracketList1 $ (,) <$> var <*> p

-- | Parse a formula in unsorted first-order logic.
unsortedFirstOrder :: Parser UnsortedFirstOrder
unsortedFirstOrder = firstOrder unsorted
  where unsorted = pure (Unsorted ()) <?> "unsorted"

sorted :: Parser s -> Parser (Sorted s)
sorted s = Sorted <$> optional (op ':' *> optionalParens s) <?> "sorted"

-- | An alias for 'monomorphicFirstOrder'.
sortedFirstOrder :: Parser SortedFirstOrder
sortedFirstOrder = monomorphicFirstOrder

-- | Parse a formula in sorted monomorphic first-order logic.
monomorphicFirstOrder :: Parser MonomorphicFirstOrder
monomorphicFirstOrder = firstOrder (sorted sort) <?> "tff0"

quantifiedSort :: Parser QuantifiedSort
quantifiedSort = token "$tType" $> QuantifiedSort ()

-- | Parse a formula in sorted polymorphic first-order logic.
polymorphicFirstOrder :: Parser PolymorphicFirstOrder
polymorphicFirstOrder =  firstOrder (sorted (eitherP quantifiedSort tff1Sort))
                     <?> "tff1"


-- ** Units

-- | Parse a formula in a given TPTP language.
formula :: Language -> Parser Formula
formula = \case
  CNF_ -> CNF <$> clause <?> "cnf"
  FOF_ -> FOF <$> unsortedFirstOrder <?> "fof"
  TFF_ -> tff <$> polymorphicFirstOrder <?> "tff"
  where
    tff f = case monomorphizeFirstOrder f of
     Just f' -> TFF0 f'
     Nothing -> TFF1 f

-- | Parse a formula role.
role :: Parser (Reserved Role)
role = reserved <?> "role"
{-# INLINE role #-}

-- | Parse the name of a TPTP language.
language :: Parser Language
language = enum <?> "language"
{-# INLINE language #-}

-- | Parse a TPTP declaration in a given language.
declaration :: Language -> Parser Declaration
declaration l =  token "type" *> comma (optionalParens typeDeclaration)
             <|> Formula <$> role <*> comma (formula l)
             <?> "declaration"

-- | Parse a declaration with the @type@ role - either a typing relation or
-- a sort declaration.
typeDeclaration :: Parser Declaration
typeDeclaration =  Sort   <$> atom <* op ':' <*> optionalParens arity
               <|> Typing <$> atom <* op ':' <*> optionalParens type_
               <?> "type declaration"
  where
    arity = genericLength . fst <$> mapping (optionalParens (token "$tType"))

-- | Parse a unit name.
unitName :: Parser (Either Atom Integer)
unitName = eitherP atom (signed integer) <?> "unit name"
{-# INLINE unitName #-}

-- | Parse a list of unit names.
unitNames :: Parser (NonEmpty UnitName)
unitNames = bracketList1 unitName <?> "unit names"
{-# INLINE unitNames #-}

-- | Parse an @include@ statement.
include :: Parser Unit
include =  labeled "include" (Include <$> atom <*> maybeP unitNames) <* op '.'
       <?> "include"

-- | Parse an annotated unit.
annotatedUnit :: Parser Unit
annotatedUnit = do
  l <- language
  let n = unitName
  let d = declaration l
  let a = maybeP annotation
  parens (Unit <$> n <*> comma d <*> a) <* op '.'
  <?> "annotated unit"

-- | Parse a TPTP unit.
unit :: Parser Unit
unit = include <|> annotatedUnit <?> "unit"

-- | Parse a TPTP input.
tptp :: Parser TPTP
tptp = TPTP <$> manyTill unit endOfInput <?> "tptp"

-- | Parse a TSTP input.
tstp :: Parser TSTP
tstp =  skipSpace *> (TSTP <$> szs <*> manyTill unit endOfInput) <* endOfInput
    <?> "tstp"


-- ** Annotations

instance Semigroup SZS where
  SZS s d <> SZS s' d' = SZS (s <|> s') (d <|> d')

instance Monoid SZS where
  mempty = SZS empty empty
  mappend = (<>)

-- | Parse the SZS ontology information of a TSTP output inside a comment.
szs :: Parser SZS
szs = fromMaybe mempty . mconcat <$> many szsComment

szsComment :: Parser (Maybe SZS)
szsComment =  commented (skipSpace *> optional szsAnnotation) <* skipSpace
          <?> "szs comment"

szsAnnotation :: Parser SZS
szsAnnotation =  string "SZS" *> skipSpace *> (szsStatus <|> szsDataform)
             <?> "szs annotation"

szsStatus :: Parser SZS
szsStatus =  string "status" *> skipSpace
          *> (fromStatus <$> status)
         <?> "status"
  where
    fromStatus s = SZS (Just s) Nothing
    status = eitherP (unwrapSZSOntology <$> named)
                     (unwrapSZSOntology <$> named)

szsDataform :: Parser SZS
szsDataform =  string "output" *> skipSpace
            *> string "start"  *> skipSpace
            *> (fromDataform <$> dataform)
           <?> "dataform"
  where
    fromDataform d = SZS Nothing (Just d)
    dataform = unwrapSZSOntology <$> named

-- | Parse an introduction marking.
intro :: Parser Intro
intro = enum <?> "intro"
{-# INLINE intro #-}

-- | Parse a unit of information about a formula.
info :: Parser Info
info =  labeled "description" (Description <$> atom)
    <|> labeled "iquote"      (Iquote      <$> atom)
    <|> labeled "status"      (Status      <$> reserved)
    <|> labeled "assumptions" (Assumptions <$> unitNames)
    <|> labeled "refutation"  (Refutation  <$> atom)
    <|> labeled "new_symbols" (NewSymbols  <$> atom <*> comma symbols)
    <|> labeled "bind"        (Bind <$> var <*> comma expr)
    <|> Expression <$> expr
    <|> uncurry Application <$> application atom info
    <|> InfoNumber <$> number
    <|> Infos <$> infos
  where
    symbols = bracketList (eitherP var atom)

infos :: Parser [Info]
infos = bracketList info <?> "infos"
{-# INLINE infos #-}

-- | Parse and expression
expr :: Parser Expression
expr =  char '$' *> (labeled "fot" (Term <$> optionalParens term)
                <|>  Logical <$> (language >>= parens . optionalParens . formula))
    <?> "expression"

-- | Parse a parent.
parent :: Parser Parent
parent = Parent <$> source <*> option [] (op ':' *> infos) <?> "parent"

-- | Parse the source of a unit.
source :: Parser Source
source =  token "unknown" $> UnknownSource
      <|> labeled "file"       (File       <$> atom     <*> maybeP unitName)
      <|> labeled "theory"     (Theory     <$> atom     <*> maybeP infos)
      <|> labeled "creator"    (Creator    <$> atom     <*> maybeP infos)
      <|> labeled "introduced" (Introduced <$> reserved <*> maybeP infos)
      <|> labeled "inference"  (Inference  <$> atom     <*> comma  infos
                                           <*> comma (bracketList parent))
      <|> UnitSource <$> unitName
      <?> "source"

-- | Parse an annotation.
annotation :: Parser Annotation
annotation = (,) <$> source <*> maybeP infos <?> "annotation"