view python/ppci/c3/parser.py @ 311:ff665880a6b0

Added testcase for kernel and userspace
author Windel Bouwman
date Mon, 16 Dec 2013 12:49:24 +0100
parents 2e7f55319858
children 04cf4d26a3bc
line wrap: on
line source

import logging
from ppci import CompilerError
from .astnodes import Member, Literal, TypeCast, Unop, Binop
from .astnodes import Assignment, ExpressionStatement, Compound
from .astnodes import Return, While, If, Empty
from .astnodes import FunctionType, Function, FormalParameter
from .astnodes import StructureType, DefinedType, PointerType
from .astnodes import Constant, Variable
from .astnodes import StructField, Deref
from .astnodes import Package
from .astnodes import Identifier
from .astnodes import FunctionCall


class Parser:
    """ Parses sourcecode into an abstract syntax tree (AST) """
    def __init__(self, diag):
        self.logger = logging.getLogger('c3')
        self.diag = diag

    def parseSource(self, tokens):
        self.logger.info('Parsing source')
        self.tokens = tokens
        self.token = self.tokens.__next__()
        try:
            self.parsePackage()
            self.mod.ok = True  # Valid until proven wrong :)
            return self.mod
        except CompilerError as e:
            self.diag.addDiag(e)

    def Error(self, msg):
        raise CompilerError(msg, self.token.loc)

    # Lexer helpers:
    def Consume(self, typ):
        if self.Peak == typ:
            return self.NextToken()
        else:
            self.Error('Excected: "{0}", got "{1}"'.format(typ, self.Peak))

    @property
    def Peak(self):
        return self.token.typ

    @property
    def CurLoc(self):
        return self.token.loc

    def hasConsumed(self, typ):
        if self.Peak == typ:
            self.Consume(typ)
            return True
        return False

    def NextToken(self):
        t = self.token
        if t.typ != 'END':
            self.token = self.tokens.__next__()
        return t

    def addDeclaration(self, decl):
        self.currentPart.declarations.append(decl)

    def parseImport(self):
        self.Consume('import')
        name = self.Consume('ID').val
        self.mod.imports.append(name)
        self.Consume(';')

    def parsePackage(self):
        self.Consume('module')
        name = self.Consume('ID')
        self.Consume(';')
        self.mod = Package(name.val, name.loc)
        self.currentPart = self.mod
        while self.Peak != 'END':
            self.parseTopLevel()
        self.Consume('END')

    def parseTopLevel(self):
        if self.Peak == 'function':
            self.parseFunctionDef()
        elif self.Peak == 'var':
            self.parseVarDef()
            # TODO handle variable initialization
        elif self.Peak == 'const':
            self.parseConstDef()
        elif self.Peak == 'type':
            self.parseTypeDef()
        elif self.Peak == 'import':
            self.parseImport()
        else:
            self.Error('Expected function, var, const or type')

    def parseDesignator(self):
        """ A designator designates an object with a name. """
        name = self.Consume('ID')
        return Identifier(name.val, name.loc)

    def parseIdSequence(self):
        ids = [self.Consume('ID')]
        while self.hasConsumed(','):
            ids.append(self.Consume('ID'))
        return ids

    # Type system
    def parseTypeSpec(self):
        # For now, do simple type spec, just parse an ID:
        if self.Peak == 'struct':
            self.Consume('struct')
            self.Consume('{')
            mems = []
            while self.Peak != '}':
                mem_t = self.parseTypeSpec()
                for i in self.parseIdSequence():
                    mems.append(StructField(i.val, mem_t))
                self.Consume(';')
            self.Consume('}')
            theT = StructureType(mems)
        elif self.Peak == 'enum':
            # TODO)
            raise NotImplementedError()
        else:
            theT = self.PostFixExpression()
        # Check for pointer suffix:
        while self.hasConsumed('*'):
            theT = PointerType(theT)
        return theT

    def parseTypeDef(self):
        self.Consume('type')
        newtype = self.parseTypeSpec()
        typename = self.Consume('ID')
        self.Consume(';')
        df = DefinedType(typename.val, newtype, typename.loc)
        self.addDeclaration(df)

    # Variable declarations:
    def parseVarDef(self):
        self.Consume('var')
        t = self.parseTypeSpec()
        for name in self.parseIdSequence():
            v = Variable(name.val, t)
            v.loc = name.loc
            self.addDeclaration(v)
        self.Consume(';')
        return Empty()

    def parseConstDef(self):
        self.Consume('const')
        t = self.parseTypeSpec()
        while True:
            name = self.Consume('ID')
            self.Consume('=')
            val = self.Expression()
            c = Constant(name.val, t, val)
            c.loc = name.loc
            if not self.hasConsumed(','):
                break
        self.Consume(';')

    def parseFunctionDef(self):
        loc = self.Consume('function').loc
        returntype = self.parseTypeSpec()
        fname = self.Consume('ID').val
        f = Function(fname, loc)
        self.addDeclaration(f)
        savePart = self.currentPart
        self.currentPart = f
        self.Consume('(')
        parameters = []
        if not self.hasConsumed(')'):
            while True:
                typ = self.parseTypeSpec()
                name = self.Consume('ID')
                param = FormalParameter(name.val, typ)
                param.loc = name.loc
                self.addDeclaration(param)
                parameters.append(param)
                if not self.hasConsumed(','):
                    break
            self.Consume(')')
        paramtypes = [p.typ for p in parameters]
        f.typ = FunctionType(paramtypes, returntype)
        f.body = self.parseCompound()
        self.currentPart = savePart

    def parseIf(self):
        loc = self.Consume('if').loc
        self.Consume('(')
        condition = self.Expression()
        self.Consume(')')
        yes = self.Statement()
        no = self.Statement() if self.hasConsumed('else') else Empty()
        return If(condition, yes, no, loc)

    def parseWhile(self):
        loc = self.Consume('while').loc
        self.Consume('(')
        condition = self.Expression()
        self.Consume(')')
        statements = self.Statement()
        return While(condition, statements, loc)

    def parseReturn(self):
        loc = self.Consume('return').loc
        if self.Peak == ';':
            expr = Literal(0, loc)
        else:
            expr = self.Expression()
        self.Consume(';')
        return Return(expr, loc)

    def parseCompound(self):
        self.Consume('{')
        statements = []
        while not self.hasConsumed('}'):
            statements.append(self.Statement())
        return Compound(statements)

    def Statement(self):
        # Determine statement type based on the pending token:
        if self.Peak == 'if':
            return self.parseIf()
        elif self.Peak == 'while':
            return self.parseWhile()
        elif self.Peak == '{':
            return self.parseCompound()
        elif self.hasConsumed(';'):
            return Empty()
        elif self.Peak == 'var':
            return self.parseVarDef()
        elif self.Peak == 'return':
            return self.parseReturn()
        else:
            x = self.UnaryExpression()
            if self.Peak == '=':
                # We enter assignment mode here.
                loc = self.Consume('=').loc
                rhs = self.Expression()
                return Assignment(x, rhs, loc)
            else:
                return ExpressionStatement(x, x.loc)

    # Expression section:
    # We not implement these C constructs:
    # a(2), f = 2
    # and this:
    # a = 2 < x : 4 ? 1;

    def Expression(self):
        exp = self.LogicalAndExpression()
        while self.Peak == 'or':
            loc = self.Consume('or').loc
            e2 = self.LogicalAndExpression()
            exp = Binop(exp, 'or', e2, loc)
        return exp

    def LogicalAndExpression(self):
        o = self.EqualityExpression()
        while self.Peak == 'and':
            loc = self.Consume('and').loc
            o2 = self.EqualityExpression()
            o = Binop(o, 'and', o2, loc)
        return o

    def EqualityExpression(self):
        ee = self.SimpleExpression()
        while self.Peak in ['<', '==', '>', '>=', '<=', '!=']:
            op = self.Consume(self.Peak)
            ee2 = self.SimpleExpression()
            ee = Binop(ee, op.typ, ee2, op.loc)
        return ee

    def SimpleExpression(self):
        """ Shift operations before + and - ? """
        e = self.AddExpression()
        while self.Peak in ['>>', '<<']:
            op = self.Consume(self.Peak)
            e2 = self.AddExpression()
            e = Binop(e, op.typ, e2, op.loc)
        return e

    def AddExpression(self):
        e = self.Term()
        while self.Peak in ['+', '-']:
            op = self.Consume(self.Peak)
            e2 = self.Term()
            e = Binop(e, op.typ, e2, op.loc)
        return e

    def Term(self):
        t = self.BitwiseOr()
        while self.Peak in ['*', '/']:
            op = self.Consume(self.Peak)
            t2 = self.BitwiseOr()
            t = Binop(t, op.typ, t2, op.loc)
        return t

    def BitwiseOr(self):
        a = self.BitwiseAnd()
        while self.Peak == '|':
            op = self.Consume(self.Peak)
            b = self.BitwiseAnd()
            a = Binop(a, op.typ, b, op.loc)
        return a

    def BitwiseAnd(self):
        a = self.CastExpression()
        while self.Peak == '&':
            op = self.Consume(self.Peak)
            b = self.CastExpression()
            a = Binop(a, op.typ, b, op.loc)
        return a

    # Domain of unary expressions:

    def CastExpression(self):
        """
          the C-style type cast conflicts with '(' expr ')'
          so introduce extra keyword 'cast'
        """
        if self.Peak == 'cast':
            loc = self.Consume('cast').loc
            self.Consume('<')
            t = self.parseTypeSpec()
            self.Consume('>')
            self.Consume('(')
            ce = self.Expression()
            self.Consume(')')
            return TypeCast(t, ce, loc)
        else:
            return self.UnaryExpression()

    def UnaryExpression(self):
        if self.Peak in ['&', '*']:
            op = self.Consume(self.Peak)
            ce = self.CastExpression()
            if op.val == '*':
                return Deref(ce, op.loc)
            else:
                return Unop(op.typ, ce, op.loc)
        else:
            return self.PostFixExpression()

    def PostFixExpression(self):
        pfe = self.PrimaryExpression()
        while self.Peak in ['[', '.', '->', '(']:
            if self.hasConsumed('['):
                raise NotImplementedError('Array not yet implemented')
            elif self.hasConsumed('->'):
                field = self.Consume('ID')
                pfe = Deref(pfe, pfe.loc)
                pfe = Member(pfe, field.val, field.loc)
            elif self.hasConsumed('.'):
                field = self.Consume('ID')
                pfe = Member(pfe, field.val, field.loc)
            elif self.hasConsumed('('):
                # Function call
                args = []
                if not self.hasConsumed(')'):
                    args.append(self.Expression())
                    while self.hasConsumed(','):
                        args.append(self.Expression())
                    self.Consume(')')
                pfe = FunctionCall(pfe, args, pfe.loc)
            else:
                raise Exception()
        return pfe

    def PrimaryExpression(self):
        if self.hasConsumed('('):
            e = self.Expression()
            self.Consume(')')
            return e
        elif self.Peak == 'NUMBER':
            val = self.Consume('NUMBER')
            return Literal(val.val, val.loc)
        elif self.Peak == 'REAL':
            val = self.Consume('REAL')
            return Literal(val.val, val.loc)
        elif self.Peak == 'true':
            val = self.Consume('true')
            return Literal(True, val.loc)
        elif self.Peak == 'false':
            val = self.Consume('false')
            return Literal(False, val.loc)
        elif self.Peak == 'STRING':
            val = self.Consume('STRING')
            return Literal(val.val, val.loc)
        elif self.Peak == 'ID':
            return self.parseDesignator()
        self.Error('Expected NUM, ID or (expr), got {0}'.format(self.Peak))