Mercurial > botc / file revision

#include "Expression.h"
#include "DataBuffer.h"
#include "Lexer.h"
#include "Variables.h"

struct OperatorInfo
{
	EToken		token;
	int			priority;
	int			numoperands;
	EDataHeader	header;
};

static const OperatorInfo gOperators[] =
{
	{ tkExclamationMark,	0,		1,	dhNegateLogical,	},
	{ tkMinus,				0,		1,	dhUnaryMinus,		},
	{ tkMultiply,			10,		2,	dhMultiply,			},
	{ tkDivide,				10,		2,	dhDivide,			},
	{ tkModulus,			10,		2,	dhModulus,			},
	{ tkPlus,				20,		2,	dhAdd,				},
	{ tkMinus,				20,		2,	dhSubtract,			},
	{ tkLeftShift,			30,		2,	dhLeftShift,		},
	{ tkRightShift,			30,		2,	dhRightShift,		},
	{ tkLesser,				40,		2,	dhLessThan,			},
	{ tkGreater,			40,		2,	dhGreaterThan,		},
	{ tkAtLeast,			40,		2,	dhAtLeast,			},
	{ tkAtMost,				40,		2,	dhAtMost,			},
	{ tkEquals,				50,		2,	dhEquals			},
	{ tkNotEquals,			50,		2,	dhNotEquals			},
	{ tkAmperstand,			60,		2,	dhAndBitwise		},
	{ tkCaret,				70,		2,	dhEorBitwise		},
	{ tkBar,				80,		2,	dhOrBitwise			},
	{ tkDoubleAmperstand,	90,		2,	dhAndLogical		},
	{ tkDoubleBar,			100,	2,	dhOrLogical			},
	{ tkQuestionMark,		110,	3,	(EDataHeader) 0		},
};

/*
	// There isn't a dataheader for ternary operator. Instead, we use dhIfNotGoto
	// to create an "if-block" inside an expression.
	// Behold, big block of writing madness! :P
	ByteMark* mark1 = retbuf->AddMark (""); // start of "else" case
	ByteMark* mark2 = retbuf->AddMark (""); // end of expression
	retbuf->WriteDWord (dhIfNotGoto); // if the first operand (condition)
	retbuf->AddReference (mark1); // didn't eval true, jump into mark1
	retbuf->MergeAndDestroy (rb); // otherwise, perform second operand (true case)
	retbuf->WriteDWord (dhGoto); // afterwards, jump to the end, which is
	retbuf->AddReference (mark2); // marked by mark2.
	retbuf->AdjustMark (mark1); // move mark1 at the end of the true case
	retbuf->MergeAndDestroy (tb); // perform third operand (false case)
	retbuf->AdjustMark (mark2); // move the ending mark2 here
*/

// =============================================================================
//
Expression::Expression (BotscriptParser* parser, Lexer* lx, EType reqtype) :
	mParser (parser),
	mLexer (lx),
	mType (reqtype),
	mResult (null)
{
	ExpressionSymbol* sym;

	while ((sym = ParseSymbol()) != null)
		mSymbols << sym;

	if (mSymbols.IsEmpty())
		Error ("Expected expression");

	AdjustOperators();

	for (ExpressionSymbol* sym : mSymbols)
	{
		switch (sym->GetType())
		{
			case eOperatorSymbol:
			{
				Print ("\t- Operator: %1\n", static_cast<ExpressionOperator*> (sym)->GetID());
				break;
			}

			case eValueSymbol:
			{
				ExpressionValue* val = static_cast<ExpressionValue*> (sym);

				if (val->IsConstexpr())
					Print ("\t- Constant expression value: %1\n", val->GetValue());
				else
				{
					Print ("\t- Databuffer value:\n");
					val->GetBuffer()->Dump();
				}
				break;
			}

			case eColonSymbol:
			{
				Print ("\t- Colon");
				break;
			}
		}
	}

	Verify();
	exit (0);

	/*
	mResult = Evaluate();
	*/
}

// =============================================================================
//
Expression::~Expression()
{
	for (ExpressionSymbol* sym : mSymbols)
		delete sym;

	delete mResult;
}

// =============================================================================
//
// Try to parse an expression symbol (i.e. an operator or operand or a colon)
// from the lexer.
//
ExpressionSymbol* Expression::ParseSymbol()
{
	int pos = mLexer->GetPosition();
	ExpressionValue* op = null;
	enum ELocalException { failed };

	try
	{
		ScriptVariable* globalvar;
		mLexer->MustGetNext();

		Print ("Token type: %1\n", mLexer->DescribeTokenType (mLexer->GetTokenType()));

		if (mLexer->GetTokenType() == tkColon)
			return new ExpressionColon;

		// Check for operator
		for (const OperatorInfo& op : gOperators)
			if (mLexer->GetTokenType() == op.token)
				return new ExpressionOperator ((EOperator) (&op - &gOperators[0]));

		// Check sub-expression
		if (mLexer->GetTokenType() == tkParenStart)
		{
			Expression expr (mParser, mLexer, mType);
			mLexer->MustGetNext (tkParenEnd);
			return expr.GetResult();
		}

		op = new ExpressionValue (mType);

		// Check function
		if (CommandInfo* comm = FindCommandByName (GetTokenString()))
		{
			if (mType != EUnknownType && comm->returnvalue != mType)
				Error ("%1 returns an incompatible data type", comm->name);

			op->SetBuffer (mParser->ParseCommand (comm));
			return op;
		}

		// Check constant
		if (ConstantInfo* constant = mParser->FindConstant (GetTokenString()))
		{
			if (mType != constant->type)
				Error ("constant `%1` is %2, expression requires %3\n",
					constant->name, GetTypeName (constant->type),
						GetTypeName (mType));

			switch (constant->type)
			{
				case EBoolType:
				case EIntType:
					op->SetValue (constant->val.ToLong());
					break;

				case EStringType:
					op->SetValue (GetStringTableIndex (constant->val));
					break;

				case EVoidType:
				case EUnknownType:
					break;
			}

			return op;
		}

		// Check global variable
		if ((globalvar = FindGlobalVariable (GetTokenString())))
		{
			DataBuffer* buf = new DataBuffer (8);
			buf->WriteDWord (dhPushGlobalVar);
			buf->WriteDWord (globalvar->index);
			op->SetBuffer (buf);
			return op;
		}

		EToken tt;

		// Check for literal
		switch (mType)
		{
			case EVoidType:
			case EUnknownType:
			{
				Error ("unknown identifier `%1` (expected keyword, function or variable)", GetTokenString());
				break;
			}

			case EBoolType:
			{
				if ((tt = mLexer->GetTokenType()) == tkTrue || tt == tkFalse)
				{
					op->SetValue (tt == tkTrue ? 1 : 0);
					return op;
				}
			}
			case EIntType:
			{
				if (mLexer->GetTokenType() != tkNumber)
					throw failed;

				op->SetValue (GetTokenString().ToLong());
				return op;
			}

			case EStringType:
			{
				if (mLexer->GetTokenType() != tkString)
					throw failed;

				op->SetValue (GetStringTableIndex (GetTokenString()));
				return op;
			}
		}

		assert (false);
		throw failed;
	}
	catch (ELocalException&)
	{
		// We use a local enum here since catch(...) would catch Error() calls.
		mLexer->SetPosition (pos);
		delete op;
		return null;
	}

	assert (false);
	return null;
}

// =============================================================================
//
// The symbol parsing process only does token-based checking for operators. Thus
// ALL minus operators are actually unary minuses simply because both have
// tkMinus as their token and the unary minus is prior to the binary minus in
// the operator table. Now that we have all symbols present, we can correct
// cases like this.
//
void Expression::AdjustOperators()
{
	for (auto it = mSymbols.begin() + 1; it != mSymbols.end(); ++it)
	{
		if ((*it)->GetType() != eOperatorSymbol)
			continue;

		ExpressionOperator* op = static_cast<ExpressionOperator*> (*it);

		// Unary minus with a value as the previous symbol cannot really be
		// unary; replace with binary minus.
		if (op->GetID() == opUnaryMinus && (*(it - 1))->GetType() == eValueSymbol)
		{
			Print ("Changing symbol operator #%1 from %2 to %3\n",
				it - mSymbols.begin(), op->GetID(), opSubtraction);
			op->SetID (opSubtraction);
		}
	}
}

// =============================================================================
//
// Verifies a single value. Helper function for Expression::Verify.
//
void Expression::TryVerifyValue (bool* verified, SymbolList::Iterator it)
{
	int i = it - mSymbols.begin();

	// Ensure it's an actual value
	if ((*it)->GetType() != eValueSymbol)
		Error ("malformed expression (symbol #%1 is not a value)", i);

	verified[i] = true;
}

// =============================================================================
//
// Ensures the expression is valid and well-formed and not OMGWTFBBQ. Throws an
// error if this is not the case.
//
void Expression::Verify()
{
	if (mSymbols.Size() == 1)
	{
		if (mSymbols[0]->GetType() != eValueSymbol)
			Error ("bad expression");

		Print ("Expression speedy-verified (1 expr symbol)");
		return;
	}

	bool* verified = new bool[mSymbols.Size()];
	memset (verified, 0, mSymbols.Size() * sizeof (decltype (*verified)));
	const auto last = mSymbols.end() - 1;
	const auto first = mSymbols.begin();

	for (auto it = mSymbols.begin(); it != mSymbols.end(); ++it)
	{
		int i = (it - first);

		if ((*it)->GetType() != eOperatorSymbol)
			continue;

		ExpressionOperator* op = static_cast<ExpressionOperator*> (*it);
		int numoperands = gOperators[op->GetID()].numoperands;

		switch (numoperands)
		{
			case 1:
			{
				// Ensure that:
				// -	unary operator is not the last symbol
				// -	unary operator is succeeded by a value symbol
				// -	neither symbol overlaps with something already verified
				TryVerifyValue (verified, it + 1);

				if (it == last || verified[i] == true)
					Error ("malformed expression");

				verified[i] = true;
				break;
			}

			case 2:
			{
				// Ensure that:
				// -	binary operator is not the first or last symbol
				// -	is preceded and succeeded by values
				// -	none of the three tokens are already verified
				//
				// Basically similar logic as above.
				if (it == first || it == last || verified[i] == true)
					Error ("malformed expression");

				TryVerifyValue (verified, it + 1);
				TryVerifyValue (verified, it - 1);
				verified[i] = true;
				break;
			}

			case 3:
			{
				// Ternary operator case. This goes a bit nuts.
				// This time we have the following:
				//
				// (VALUE) ? (VALUE) : (VALUE)
				//         ^
				// --------/ we are here
				//
				// Check that the:
				// -	questionmark operator is not misplaced (first or last)
				// -	the value behind the operator (-1) is valid
				// -	the value after the operator (+1) is valid
				// -	the value after the colon (+3) is valid
				// -	none of the five tokens are verified
				//
				TryVerifyValue (verified, it - 1);
				TryVerifyValue (verified, it + 1);
				TryVerifyValue (verified, it + 3);

				if (it == first ||
					it >= mSymbols.end() - 3 ||
					verified[i] == true ||
					verified[i + 2] == true ||
					(*(it + 2))->GetType() != eColonSymbol)
				{
					Error ("malformed expression");
				}

				verified[i] = true;
				verified[i + 2] = true;
				break;
			}

			default:
				Error ("WTF operator with %1 operands", numoperands);
		}
	}

	for (int i = 0; i < mSymbols.Size(); ++i)
		if (verified[i] == false)
			Error ("malformed expression: expr symbol #%1 is was left unverified", i);

	Print ("Expression verified.\n");
}

// =============================================================================
//
ExpressionValue* Expression::Evaluate()
{
	
}

// =============================================================================
//
ExpressionValue* Expression::GetResult()
{
	return mResult;
}

// =============================================================================
//
String Expression::GetTokenString()
{
	return mLexer->GetToken()->text;
}

// =============================================================================
//
ExpressionOperator::ExpressionOperator (EOperator id) :
	ExpressionSymbol (Expression::eOperatorSymbol),
	mID (id) {}

// =============================================================================
//
ExpressionValue::ExpressionValue (EType valuetype) :
	ExpressionSymbol (Expression::eValueSymbol),
	mBuffer (null),
	mValueType (valuetype) {}

// =============================================================================
//
ExpressionValue::~ExpressionValue()
{
	delete mBuffer;
}

// =============================================================================
//
void ExpressionValue::ConvertToBuffer()
{
	if (IsConstexpr() == false)
		return;

	SetBuffer (new DataBuffer);

	switch (mValueType)
	{
		case EBoolType:
		case EIntType:
			GetBuffer()->WriteDWord (dhPushNumber);
			GetBuffer()->WriteDWord (abs (mValue));

			if (mValue < 0)
				GetBuffer()->WriteDWord (dhUnaryMinus);
			break;

		case EStringType:
			GetBuffer()->WriteDWord (dhPushStringIndex);
			GetBuffer()->WriteDWord (mValue);
			break;

		case EVoidType:
		case EUnknownType:
			assert (false);
			break;
	}
}