/// <summary>
/// Parses ok, or throws
/// </summary>
/// <param name="terminatingCharacter">
/// Stops parsing if it sees this CodePoint in an unnested context.
/// </param>
/// <returns>
/// AbstractSyntaxTree representing parse, or throws
/// </returns>
public AbstractSyntaxTree Parse(CodePoint terminatingCharacter)
{
_operatorStack.Clear();
_operatorStack.Push(Operator.None); // make Peek () safe to do without first checking if empty
_operandStack.Clear();
return(ParseToResult(terminatingCharacter));
}
private void ReadIntoNext()
{
if (!_next.AtEOF())
{
_nextCursor++;
_next = _cps.Read();
}
}
private AbstractSyntaxTree ParseToResult(CodePoint terminatingCharacter)
{
ParseExpressionToStack(terminatingCharacter);
while (_operatorStack.Peek() != Operator.None)
{
ReduceTopOperator();
}
if (_operandStack.Count != 1)
{
throw new AssertionFailedException("operand stack should have only 1 operand");
}
return(_operandStack.Pop());
}
/// <summary>
/// Attempts to parse. Does not throw on error.
/// </summary>
/// <param name="terminatingCharacter">
/// Stops parsing if it sees this CodePoint in an unnested context.
/// </param>
/// <returns>
/// Either the result or a list of errors
/// (Could/should return list of warnings with results as well.)
/// </returns>
public ResultOrErrorList <AbstractSyntaxTree> TryParse(CodePoint terminatingCharacter)
{
_operatorStack.Clear();
_operatorStack.Push(Operator.None); // make Peek () safe to do without first checking if empty
_operandStack.Clear();
try {
return(new ResultOrErrorList <AbstractSyntaxTree> (ParseToResult(terminatingCharacter)));
} catch (Exceptions.CompilationException p) {
return(new ResultOrErrorList <AbstractSyntaxTree> (new List <Exceptions.CompilationException> ()
{
p
}));
}
}
public void Advance()
{
if (!_curr.AtEOF( ))
{
_cursor = _nextCursor;
_curr = _next;
_currLineMark = _nextLineMark;
ReadIntoNext( );
switch (_curr.Value)
{
case '\r':
_curr = new CodePoint(( byte )'\n');
if (_next.Value == '\n')
{
ReadIntoNext( );
}
goto case '\n';
case '\n':
_line++;
_nextLineMark = _nextCursor;
_indentationLevel = 0;
_nonBlankSeen = false;
HandlePPDirectivesFromNormalMode( );
break;
case ' ':
_indentationLevel++;
break;
case '\t':
if ((_indentationLevel & 7) != 0)
{
throw new CompilationException("tab after space is illegal for indentation");
}
_indentationLevel += 8; // tab level
break;
default:
_nonBlankSeen = true;
break;
}
}
}
private readonly CodePoint [] _idMores = { new CodePoint((byte)'_'), new CodePoint((byte)'.') }; //, new CodePoint ( (byte) '-' ) };
public ScanIt(CodePointStream cps, string job, System.IO.TextWriter lg)
{
_defdDict = new HashSet <ByteString> {
new ByteString("true"), new ByteString("1")
};
_cps = cps;
_job = job;
_lg = lg;
_line = 0;
_currTokenCursorStart = _cursor = _currLineMark = -2;
_nextCursor = _nextLineMark = -1;
_indentationLevel = 0;
_tokenModeStack = new Stack <TokenModeEnum> ();
PushMode(TokenModeEnum.Illegal);
PushMode(TokenModeEnum.Normal);
_curr = new CodePoint((byte)' ');
_next = new CodePoint((byte)'\n');
Advance();
Advance();
}
private void ParseExpressionToStack(CodePoint terminatingCharacter)
{
for ( ; ;) // Alternate between the States
{
// initial State: Unary State
for ( ; ;) // Unary State
{
var token = _scanner.Token();
switch (token.Value)
{
case '(':
_scanner.Advance();
_operatorStack.Push(Operator
.PrecedenceGroupingParenthesis); // nothing to reduce, as '(' has highest precedence
continue;
case ')':
_scanner.Advance();
if (_operatorStack.Peek() != Operator.FunctionCall)
{
_scanner.ErrorAtMark("unexpected ) in unary state");
}
_operatorStack.Pop();
// NB: We have a function call with no argument list, and we can either:
// provide a unary function call operator
// but our automated operator generators don't have that concept, or,
// push an empty operand onto the stack so there's two operands for it to work
_operandStack.Push(null);
BuildAndPushOperator(Operator.FunctionCall);
break;
case '!':
_scanner.Advance();
_operatorStack.Push(Operator.LogicalNot); // prefix unary operators have implicit highest precedence
continue;
case '&':
_scanner.Advance();
_operatorStack.Push(Operator.AddressOf);
continue;
case '*':
_scanner.Advance();
_operatorStack.Push(Operator.Indirection);
continue;
case '+':
_scanner.Advance();
ReduceAndPushByChoice('+', Operator.PrefixIncrement, Operator.FixPoint);
continue;
case '-':
_scanner.Advance();
ReduceAndPushByChoice('-', Operator.PrefixDecrement, Operator.Negation);
continue;
case '~':
_scanner.Advance();
_operatorStack.Push(Operator.BitwiseComplement);
continue;
case 'A':
var id = _scanner.GetIdentifier();
// we can look up the variable in a symbol table here
_operandStack.Push(_symbolTable.LookupSymbol(id));
break;
case '0':
var num = _scanner.GetNonNegativeIntegralLiteral();
_operandStack.Push(new LongIntegerTreeNode(num));
break;
case '\"':
var str = _scanner.GetStringLiteral();
_operandStack.Push(new StringTreeNode(str));
break;
case '\'':
var chStr = _scanner.GetStringLiteral();
var cps = new CodePointStream(chStr);
var cp1 = cps.Read();
if (cp1.AtEOF())
{
_scanner.ErrorAtMark("not enough characters in character literal");
}
var cp2 = cps.Read();
if (!cp2.AtEOF())
{
_scanner.ErrorAtMark("too many characters in character literal");
}
_operandStack.Push(new LongIntegerTreeNode(cp1.Value));
break;
default:
_scanner.ErrorAtMark("unrecognized token: " + (char)token.Value);
return;
}
break; // switch to Binary State
} // for (;;) { end of Unary State
for ( ; ;) // Binary State
{
var cp = _scanner.Token();
if (cp.AtEOF() || cp == terminatingCharacter && AtTopLevelAndUnblocked())
{
return;
}
switch (cp.Value)
{
case '(':
_scanner.Advance();
ReduceThenPushOperator(Operator.FunctionCall);
break;
case ',':
_scanner.Advance();
AcceptComma();
break;
case ')':
_scanner.Advance();
AcceptCloseParen();
continue; // stay in Binary State
case '[':
_scanner.Advance();
ReduceThenPushOperator(Operator.Subscript);
break;
case ']':
_scanner.Advance();
// we have a [ b ], so we reduce until we reach the '['
ReduceUntilMatch(Operator.Subscript);
// reduction pops the operator, and, leaves us with a & b on the operand stack
BuildAndPushOperator(Operator.Subscript);
continue; // stay in Binary State
case '!':
_scanner.Advance();
if (!_scanner.IfCharacter('='))
{
_scanner.ErrorAtMark("expected = for !=");
}
ReduceThenPushOperator(Operator.NotEqual);
break;
case '*':
_scanner.Advance();
ReduceAndPushByChoice('=', Operator.AssignmentMultiplication, Operator.Multiplication);
break;
case '/':
_scanner.Advance();
ReduceAndPushByChoice('=', Operator.AssignmentDivision, Operator.Division);
break;
case '%':
_scanner.Advance();
ReduceAndPushByChoice('=', Operator.AssignmentModulo, Operator.Modulo);
break;
case '=':
_scanner.Advance();
ReduceAndPushByChoice('=', Operator.EqualEqual, Operator.Assignment);
break;
case '^':
_scanner.Advance();
ReduceAndPushByChoice('=', Operator.AssignmentBitwiseXor, Operator.BitwiseXor);
break;
case '+':
_scanner.Advance();
if (_scanner.IfCharacter('+'))
{
// C/C++ postfix ++
ReduceThenPushOperator(Operator.PostfixIncrement);
continue; // stay in Binary State
}
ReduceAndPushByChoice('=', Operator.AssignmentAddition, Operator.Addition);
break;
case '-':
_scanner.Advance();
if (_scanner.IfCharacter('>'))
{
// at least ->
if (_scanner.IfCharacter('*'))
{
// C++: member pointer: ->*
// ToDo: parse rhs of ->* (which is a member name) and generate an binary node to push on the operand stack
throw new System.NotImplementedException("->*");
}
else
{
// C/C++ indirect selection: ->
var memberName = _scanner.ExpectIdentifier("member identifier");
// ToDo: parse rhs of -> (which is a member name) and generate an operand to push on
throw new System.NotImplementedException("->");
}
}
else
{
ReduceAndPushByChoice('-', Operator.PostfixDecrement, '=', Operator.AssignmentSubtraction,
Operator.Subtraction);
}
break;
case '&':
_scanner.Advance();
ReduceAndPushByChoice('&', Operator.ShortCircutAnd, '=', Operator.AssignmentBitwiseAnd, Operator.BitwiseAnd);
break;
case '|':
_scanner.Advance();
ReduceAndPushByChoice('|', Operator.ShortCircutOr, '=', Operator.AssignmentBitwiseOr, Operator.BitwiseOr);
break;
case '.':
throw new System.NotImplementedException(". operator");
case '?':
_scanner.Advance();
// Operator.TernaryTest is used for grouping of the expressions during parsing,
// This operator won't appear in the final tree (similar to Operator.GroupingParen which doesn't appear in the abstract tree either)
ReduceThenPushOperator(Operator.TernaryTest);
// Test 1003: a?b,c:d, which should parse ok as:
//// a?(b,c):d
//// -and not-
//// (a?b),(c:d)
//// as this makes no sense and thus is not helpful
break;
case ':':
_scanner.Advance();
if (_scanner.IfCharacter(':'))
{
throw new System.NotImplementedException(":: operator");
}
else
{
ReduceUntilMatch(Operator.TernaryTest);
// this will leave two operands on the stack, so for
// a ? b : -- which is what we have so far
// a and b will be on the operand stack.
// now we push the true ternary operator:
_operatorStack.Push(Operator.TernaryChoice);
// when it gets reduced later, so for
// a ? b : c;
// reducing this operator will consume a, b, and c.
}
break;
case '<':
_scanner.Advance();
ReduceAndPushByChoice('<', '=', Operator.AssignmentBitwiseLeftShift, Operator.BitwiseLeftShift, '=',
Operator.LessOrEqual, Operator.LessThan);
break;
case '>':
_scanner.Advance();
ReduceAndPushByChoice('>', '=', Operator.AssignmentBitwiseRightShift, Operator.BitwiseRightShift, '=',
Operator.GreaterOrEqual, Operator.GreaterThan);
break;
default:
return; // let caller deal with the "unexpected" input, like ;
}
break; // switch to Unary State
} // end of for (;;) { Binary State
} // end of for(;;) { Alternating between the states
}