JSON-Lang/grammar.py

"""
A grammar for parsing CNF files

If this module is run directly in python, it will spit out valid python which produces an
oracle table for the grammar it defines.  It's recommended that this be done using
`build_oracle.sh` instead, however, which will build a whole python module containing the
oracle table, complete with imports.
"""
from emis_funky_funktions import *

from dataclasses import dataclass
from enum import auto, IntEnum
from re import compile, Pattern

from lex import Lexeme
from parse import Action

from typing import Any, Callable, Collection, Mapping, Sequence, Tuple, TypeAlias

class Tok(IntEnum):
	"""
	All possible tokens used in the grammar
	"""
	Newline = auto()
	Whitespace = auto()
	PredicateSection = auto()
	VariablesSection = auto()
	ConstantsSection = auto()
	FunctionsSection = auto()
	ClausesSection = auto()
	Negate = auto()
	OpenP = auto()
	CloseP = auto()
	Comma = auto()
	Identifier = auto()
	Eof = auto()

	def __repr__(self):
		return self._name_

LEX_TABLE: Collection[Tuple[Pattern[str], Tok]] = [
	(compile(r"\n"), Tok.Newline),
	(compile(r"[ \t]+"), Tok.Whitespace),
	(compile("Predicates:"), Tok.PredicateSection),
	(compile("Variables:"), Tok.VariablesSection),
	(compile("Constants:"), Tok.ConstantsSection),
	(compile("Functions:"), Tok.FunctionsSection),
	(compile("Clauses:"), Tok.ClausesSection),
	(compile("!"), Tok.Negate),
	(compile(r"\("), Tok.OpenP),
	(compile(r"\)"), Tok.CloseP),
	(compile(","), Tok.Comma),
	(compile(r"\w+"), Tok.Identifier),
]
"""
A mapping of regexs to the tokens the identify

Tokens earlier on in the list should be regarded as higher priority, even if a match lower
on the list also matches.  All unicode strings should be matched by at least one token.
"""

class Variable(IntEnum):
	Start = auto()
	Idents = auto()
	Clauses = auto()
	Clauses_ = auto()
	Clause = auto()
	Clause_ = auto()
	Term = auto()
	Func = auto()
	CSTerms = auto()

	def __repr__(self) -> str:
		return f'<{self._name_}>'

ASTTerm: TypeAlias = 'ASTNegated | ASTProp'

class IdentKind(IntEnum):
	Function = auto()
	Constant = auto()
	Variable = auto()
	Predicate = auto()

@dataclass(frozen=True)
class CallingNonFunc:
	term: Lexeme[Tok]
	obj_type: IdentKind
	def __str__(self):
		return f'Semantic error: Attempted to call {repr(self.term.matched_string)} (a {self.obj_type.name.lower()}) with arguments on line {self.term.line}:{self.term.col_start}-{self.term.col_end}'

@dataclass(frozen=True)
class MissingArguments:
	term: Lexeme[Tok]
	def __str__(self):
		return f'Semantic error: The function {repr(self.term.matched_string)} on line {self.term.line}:{self.term.col_start}-{self.term.col_end} is missing arguments!'

@dataclass(frozen=True)
class UnidentifiedVariable:
	term: Lexeme[Tok]
	def __str__(self):
		return f'Semantic error: Unidentified identifier {repr(self.term.matched_string)} on line {self.term.line}:{self.term.col_start}-{self.term.col_end}'

@dataclass(frozen=True)
class PropUsedInObjectContext:
	term: Lexeme[Tok]
	def __str__(self):
		return f'Semantic error: The proposition {repr(self.term.matched_string)} was used in a context where an object was expected on line {self.term.line}:{self.term.col_start}-{self.term.col_end}'

@dataclass(frozen=True)
class ObjectUsedInPropContext:
	term: Lexeme[Tok]
	obj_type: IdentKind
	def __str__(self):
		return f'Semantic error: The {self.obj_type.name.lower()} {repr(self.term.matched_string)} was used in a context where a proposition was expected on line {self.term.line}:{self.term.col_start}-{self.term.col_end}'

@dataclass(frozen=True)
class NegationOfObject:
	line: int
	col: int
	def __str__(self):
		return f'Semantic error: Attempted to use negation in a context where working on objects on line {self.line}:{self.col}'

GenIrError: TypeAlias = CallingNonFunc | MissingArguments | UnidentifiedVariable | PropUsedInObjectContext | ObjectUsedInPropContext | NegationOfObject

@dataclass(frozen=True)
class IdentBindings:
    predicate_idents: Sequence[str]
    variable_idents: Sequence[str]
    const_idents: Sequence[str]
    func_idents: Sequence[str]

@dataclass(frozen=True)
class ASTNegated:
	neg_lexeme: Lexeme[Tok]
	term: ASTTerm

	def make_ir(self, idents: IdentBindings, is_prop: bool) -> 'Result[IRTerm, GenIrError]':
		if is_prop:
			return map_res(IRNeg, self.term.make_ir(idents, True))
		else:
			return Err(NegationOfObject(self.neg_lexeme.line, self.neg_lexeme.col_start))

@dataclass(frozen=True)
class ASTProp:
	ident: Lexeme[Tok]
	arguments: Sequence[ASTTerm]

	def make_ir(self, idents: IdentBindings, is_pred: bool) -> 'Result[IRTerm, GenIrError]':
		bound_type = (
			IdentKind.Predicate
			if self.ident.matched_string in idents.predicate_idents else
			IdentKind.Variable
			if self.ident.matched_string in idents.variable_idents else
			IdentKind.Constant
			if self.ident.matched_string in idents.const_idents else
			IdentKind.Function
			if self.ident.matched_string in idents.func_idents else
			None
		)
		if bound_type is None:
			return Err(UnidentifiedVariable(self.ident))

		if is_pred:
			if bound_type != IdentKind.Predicate:
				return Err(ObjectUsedInPropContext(self.ident, bound_type))
		else:
			if bound_type == IdentKind.Function:
				if not len(self.arguments):
					return Err(MissingArguments(self.ident))
			elif bound_type == IdentKind.Predicate:
				return Err(PropUsedInObjectContext(self.ident))
			else:
				if len(self.arguments):
					return Err(CallingNonFunc(self.ident, bound_type))

		if bound_type == IdentKind.Variable:
			return Ok(IRVar(self.ident))
		else:
			arg_ir = sequence([t.make_ir(idents, False) for t in self.arguments])
			return map_res(p(IRProp, self.ident), arg_ir)

@dataclass(frozen=True)
class IRProp:
    lexeme: Lexeme[Tok]
    arguments: 'Sequence[IRTerm]'
    def __str__(self) -> str:
        return f'{self.lexeme.matched_string}({",".join(str(arg) for arg in self.arguments)})'

@dataclass(frozen=True)
class IRVar:
    lexeme: Lexeme[Tok]
    def __str__(self) -> str:
        return f'*{self.lexeme.matched_string}'

@dataclass(frozen=True)
class IRNeg:
    inner: 'IRTerm'
    def __str__(self) -> str:
        return f'¬{self.inner}'

IRTerm: TypeAlias = IRVar | IRProp | IRNeg

@cur2
def make_ir(
	idents: IdentBindings,
    clauses: Sequence[Sequence[ASTTerm]],
) -> Result[Sequence[Sequence[IRTerm]], GenIrError]:
    return sequence([sequence([term.make_ir(idents, True) for term in clause]) for clause in clauses])

def cons(stack: Sequence[Any]) -> Sequence[Any]:
    match stack:
        case [rest, head, *popped_stack]:
            return ((head, *rest), *popped_stack)
        case bad_stack:
            raise Exception("Unexpected stack state!", bad_stack)

nil: Sequence[Any] = tuple()
@cur2
def introduce(
    cons: Any,
    stack: Sequence[Any]
) -> Sequence[Any]:
    return (cons, *stack)

def f_apply(stack: Sequence[Any]) -> Sequence[Any]:
    match stack:
        case [arg, func, *popped_stack] if hasattr(func, '__call__'):
            return (func(arg), *popped_stack)
    raise Exception("Unexpected stack state!", stack)
@cur2
def call_func(func: Callable[[Any], Any], stack: Sequence[Any]) -> Sequence[Any]:
    match stack:
        case [arg, *popped_stack]:
            return (func(arg), *popped_stack)
        case bad_stack:
            raise Exception("Unexpected stack state!", bad_stack)

def drop(stack: Sequence[Any]) -> Sequence[Any]:
    return stack[1:]

GRAMMAR: Sequence[Tuple[Variable, Sequence[Variable | Tok | Action]]] = [
	(Variable.Start,
		 [ Tok.PredicateSection, drop, Variable.Idents, call_func(p(p,p,p,IdentBindings)), Tok.Newline, drop
		 , Tok.VariablesSection, drop, Variable.Idents, f_apply, Tok.Newline, drop
		 , Tok.ConstantsSection, drop, Variable.Idents, f_apply, Tok.Newline, drop
		 , Tok.FunctionsSection, drop, Variable.Idents, f_apply, call_func(make_ir), Tok.Newline, drop
		 , Tok.ClausesSection, drop, Variable.Clauses, f_apply, Tok.Eof, drop] ),

	(Variable.Idents,
		[ Tok.Identifier, call_func(lambda i: i.matched_string), Variable.Idents, cons ]),
	(Variable.Idents,
		[ introduce(nil) ]),

	(Variable.Clauses,
		[ Tok.Newline, drop, Variable.Clauses_ ]),
	(Variable.Clauses,
		[ introduce(nil) ]),

	(Variable.Clauses_,
		[ Variable.Clause, Variable.Clauses, cons ]),
	(Variable.Clauses_,
		[ introduce(nil) ]),

	(Variable.Clause,
		[ Variable.Term, Variable.Clause_, cons ]),

	(Variable.Clause_,
		[ Variable.Clause ]),
	(Variable.Clause_,
		[ introduce(nil) ]),

	(Variable.Term,
		[ Tok.Negate, call_func(cur2(ASTNegated)), Variable.Term, f_apply ]),
	(Variable.Term,
		[ Tok.Identifier, call_func(cur2(ASTProp)), Variable.Func, f_apply ]),

	(Variable.Func,
		[ Tok.OpenP, drop, Variable.Term, Variable.CSTerms, cons, Tok.CloseP, drop ]),
	(Variable.Func,
		[ introduce(nil) ]),

	(Variable.CSTerms,
		[ Tok.Comma, drop, Variable.Term, Variable.CSTerms, cons ]),
	(Variable.CSTerms,
		[ introduce(nil) ]),
]
"""
Implements the following grammar:

Start	 := PredicateSection <Idents> Newline
			VariablesSection <Idents> Newline
			ConstantsSection <Idents> Newline
			FunctionsSection <Idents> Newline
			ClausesSection <Clauses> Eof

Idents	 := Identifier <Idents>
		 := ε

Clauses  := Newline <Clauses'>
		 := ε

Clauses' := <Clause> <Clauses>
		 := ε

Clause	 := <Term> <Clause'>

Clause'  := <Clause>
		 := ε

Term	 := Negate <Term>
		 := Identifier <Func?>

Func?	 := OpenP <Term> <CSTerms> CloseP
		 := ε

CSTerms  := Comma <Term> <CSTerms>
		 := ε
"""

def lex_and_parse(input: str) -> Result[Result[Sequence[Sequence[IRTerm]], GenIrError], Tuple[Lexeme[Tok], Collection[Tok]]]:
	lexemes = unwrap_r(tokenize(LEX_TABLE, [Tok.Whitespace], Tok.Eof, input))

	oracle_table_ = oracle_table(p_instance(Tok), p_instance(Variable), GRAMMAR) #type:ignore
	parser_ = parser(oracle_table_, flip(cur2(getattr))('token'), Variable.Start)
	match parser_(lexemes):
		case Ok([Ok(ir)]):
			return Ok(Ok(ir))
		case Ok([Err(err)]):
			return Ok(Err(err))
		case Ok(huh):
			raise Exception('Unexpected end result: ', huh)
		case Err(e_tup):
			return Err(e_tup) #type:ignore
	raise Exception('Unreachable')

if __name__ == '__main__':
	# from emis_funky_funktions import cur2, flip
	# from build_oracle import print_oracle_table_enum, oracle_table
	# print(print_oracle_table_enum(oracle_table(flip(cur2(isinstance))(Tok), GRAMMAR))) #type: ignore
	from build_oracle import oracle_table
	from parse import parser
	from lex import tokenize
	import sys

	if len(sys.argv) == 2:
		with open(sys.argv[1]) as file:
			match lex_and_parse(file.read()):
				case Ok(Ok(ir)):
					print('\n'.join(' or '.join(str(t) for t in c) for c in ir))
				case Ok(Err(err)):
					print(err)
					exit(102)
				case Err((Lexeme(token, text, line, col_start, col_end), expected)):
					print(f'Parse error!  Line {line}:{col_start}-{col_end}\n\nGot: {repr(text)}\nExpected: {expected}')
					exit(101)
	else:
		print(f'Usage: python {sys.argv[0]} <cnf-file-to-parse>')
		exit(100)