表达式,最简单也最困难。各种算术、逻辑、赋值埋同函数调用,想想都头大如斗转星移山填海枯石烂。
废话有云,根据Yacc规则,表达式由操作数及操作符构成。操作数有立即数、变量、函数调用及另一个表达式。操作符有一元、二元及三元操作符。惜乎SharpC不支持三元表达式。
有两类特殊的操作数:指针及指针指示(Pointer indiction应该怎么翻译?)。
有一个特殊的操作符:sizeof。说其特殊是因为偶将其归类于一元操作符,且不像其它操作符般需要操作数,sizeof也可以类型名称作为操作数。
先看看操作符优先级定义:

public enum OperatorPriority{    // 赋值操作最低    Assign,    Logic,    Bitwise,    // 算术运算有两级优先级:+,—是一级,*,/,%是高一级    ArthmeticLow,    ArthmeticHigh,    // 拔高了移位操作符优先级于算术之上    Shift,    Address,    Unary,    Parenthese}

先看看表达式求值过程:

public Operand.Operand Evaluate(Context ctx){    if (Token is Operand.Operand)    {        if (Token is Operand.ValueOfFunctionCalling)        {            Operand.ValueOfFunctionCalling valOfFuncCalling = Token as Operand.ValueOfFunctionCalling;            return valOfFuncCalling.GetValue(ctx);        }        else            return Token as Operand.Operand;    }    else    {        List<Operand.Operand> operands = new List<Operand.Operand>();        if (LeftNode != null)            operands.Add(LeftNode.Evaluate(ctx));        if (RightNode != null)            operands.Add(RightNode.Evaluate(ctx));        Operand.Operand res = (Token as Operator.Operator).Evaluate(ctx, operands);        Debug.WriteLine(string.Format("\tExp: [{0} {1} {2}] Result: [{3}]",             operands.First().ToString(),            Token.ToString(),            operands.Count > 1 ? operands.Last().ToString() : "",             res.ToString()));        return res;    }}

表达式求值过程:

  1. 如果是操作数,则将之作为结果返回。
  2. 如果是操作符,先后对左、右节点求值,再将左、右节点求值的结果作为操作数,调用操作符的求值方法,并将其结果返回。

操作符最终的计算还是基于两种类型:int和float。



因为没有严格区分signed和unsigned,涉及到两种计算的求值过程可能会有错误。

因此有以下两个方法:

public class BinaryOperator : Operator{    public BinaryOperator(OperatorPriority prio)        : base(prio)    {    }    public Operand.Operand LeftOperand;    public Operand.Operand RightOperand;    public virtual int ComputeScalar(int left, int right)    {        throw new NotImplementedException();    }    public virtual float ComputeFloat(float left, float right)    {        throw new NotImplementedException();    }

一元操作符还多了一个Compute方法:

public class UnaryOperator : Operator{    public UnaryOperator()        : base(Operator.OperatorPriority.Unary)    {    }    public virtual int ComputeScalar(int val)    {        throw new NotImplementedException();    }    public virtual float ComputeFloat(float val)    {        throw new NotImplementedException();    }    public virtual Operand.Value Compute(Context ctx, Operand.Operand operand)    {

二元操作符的运算比较复杂,需要考虑种种情况:

public override Operand.Operand Evaluate(Context ctx, List<Operand.Operand> operands){    Debug.Assert(operands.Count() == 2);    LeftOperand = operands.First();    RightOperand = operands.Last();    DataTypeInfo leftDti = LeftOperand.GetTypeInfo(ctx);    DataTypeInfo rightDti = RightOperand.GetTypeInfo(ctx);    if (leftDti.IsPointer ||  rightDti.IsPointer)    {        int leftAddr = leftDti.IsPointer ? Context.Memory.GetInt(LeftOperand.GetValue(ctx).AsInt) : LeftOperand.GetValue(ctx).AsInt;        int rightAddr = rightDti.IsPointer ? Context.Memory.GetInt(RightOperand.GetValue(ctx).AsInt) : RightOperand.GetValue(ctx).AsInt;        Operand.ValueOfPointer valOfPointer = new Operand.ValueOfPointer() {            Address = ComputeScalar(leftAddr, rightAddr)        };        if (leftDti.IsPointer)        {            if (rightDti.IsPointer)                valOfPointer.TypeInfo = leftDti.BaseType > rightDti.BaseType ? leftDti : rightDti;            else                valOfPointer.TypeInfo = leftDti;        }        else        {            if (rightDti.IsPointer)                valOfPointer.TypeInfo = rightDti;            else                valOfPointer.TypeInfo = leftDti.BaseType > rightDti.BaseType ? leftDti : rightDti;        }        return valOfPointer;    }    else    {        Operand.Value res = new Operand.Value()        {            DataType = leftDti.BaseType > rightDti.BaseType ? leftDti.BaseType : rightDti.BaseType        };        switch (res.DataType)        {            case PrimitiveDataType.CharType:                {                    switch (rightDti.BaseType)                    {                        case PrimitiveDataType.CharType: res.DataField = (byte)ComputeScalar(LeftOperand.GetValue(ctx).AsChar, RightOperand.GetValue(ctx).AsChar); break;                        case PrimitiveDataType.ShortType: res.DataField = (byte)ComputeScalar(LeftOperand.GetValue(ctx).AsChar, RightOperand.GetValue(ctx).AsShort); break;                        case PrimitiveDataType.IntType: res.DataField = (byte)ComputeScalar(LeftOperand.GetValue(ctx).AsChar, RightOperand.GetValue(ctx).AsInt); break;                        case PrimitiveDataType.FloatType: res.DataField = (byte)ComputeFloat(LeftOperand.GetValue(ctx).AsChar, RightOperand.GetValue(ctx).AsFloat); break;                        default: throw new RuntimeException(string.Format("Operator \"{0}\" doesn't supported type \"{1}\".", this, DataTypeInfo.StringFromType(rightDti.BaseType)));                    }                }                break;            case PrimitiveDataType.ShortType:                {                    switch (rightDti.BaseType)                    {                        case PrimitiveDataType.CharType: res.DataField = (short)ComputeScalar(LeftOperand.GetValue(ctx).AsShort, RightOperand.GetValue(ctx).AsChar); break;                        case PrimitiveDataType.ShortType: res.DataField = (short)ComputeScalar(LeftOperand.GetValue(ctx).AsShort, RightOperand.GetValue(ctx).AsShort); break;                        case PrimitiveDataType.IntType: res.DataField = (short)ComputeScalar(LeftOperand.GetValue(ctx).AsShort, RightOperand.GetValue(ctx).AsInt); break;                        case PrimitiveDataType.FloatType: res.DataField = (short)ComputeFloat(LeftOperand.GetValue(ctx).AsShort, RightOperand.GetValue(ctx).AsFloat); break;                        default: throw new RuntimeException(string.Format("Operator \"{0}\" doesn't supported type \"{1}\".", this, DataTypeInfo.StringFromType(rightDti.BaseType)));                    }                }                break;            case PrimitiveDataType.IntType:                {                    switch (rightDti.BaseType)                    {                        case PrimitiveDataType.CharType: res.DataField = (int)ComputeScalar(LeftOperand.GetValue(ctx).AsInt, RightOperand.GetValue(ctx).AsChar); break;                        case PrimitiveDataType.ShortType: res.DataField = (int)ComputeScalar(LeftOperand.GetValue(ctx).AsInt, RightOperand.GetValue(ctx).AsShort); break;                        case PrimitiveDataType.IntType: res.DataField = (int)ComputeScalar(LeftOperand.GetValue(ctx).AsInt, RightOperand.GetValue(ctx).AsInt); break;                        case PrimitiveDataType.FloatType: res.DataField = (int)ComputeFloat(LeftOperand.GetValue(ctx).AsInt, RightOperand.GetValue(ctx).AsFloat); break;                        default: throw new RuntimeException(string.Format("Operator \"{0}\" doesn't supported type \"{1}\".", this, DataTypeInfo.StringFromType(rightDti.BaseType)));                    }                }                break;            case PrimitiveDataType.FloatType:                {                    switch (rightDti.BaseType)                    {                        case PrimitiveDataType.CharType: res.DataField = (float)ComputeFloat(LeftOperand.GetValue(ctx).AsFloat, RightOperand.GetValue(ctx).AsChar); break;                        case PrimitiveDataType.ShortType: res.DataField = (float)ComputeFloat(LeftOperand.GetValue(ctx).AsFloat, RightOperand.GetValue(ctx).AsShort); break;                        case PrimitiveDataType.IntType: res.DataField = (float)ComputeFloat(LeftOperand.GetValue(ctx).AsFloat, RightOperand.GetValue(ctx).AsInt); break;                        case PrimitiveDataType.FloatType: res.DataField = (float)ComputeFloat(LeftOperand.GetValue(ctx).AsFloat, RightOperand.GetValue(ctx).AsFloat); break;                        default: throw new RuntimeException(string.Format("Operator \"{0}\" doesn't supported type \"{1}\".", this, DataTypeInfo.StringFromType(rightDti.BaseType)));                    }                }                break;            default:                throw new RuntimeException(string.Format("Operator \"{0}\" doesn't supported type \"{1}\".", this, DataTypeInfo.StringFromType(res.DataType)));        } // switch        return res;    }}

一元操作符就比较简单,毕竟只有一个操作数:

public virtual Operand.Value Compute(Context ctx, Operand.Operand operand){    DataTypeInfo leftType = operand.GetTypeInfo(ctx);    if (leftType.IsPointer)    {        return new Operand.Value()        {            DataType = PrimitiveDataType.IntType,            DataField = (int)ComputeScalar(Context.Memory.GetInt(operand.GetValue(ctx).AsInt))        };    }    else    {        Operand.Value res = new Operand.Value()        {            DataType = leftType.BaseType        };        switch (leftType.BaseType)        {            case PrimitiveDataType.CharType: res.DataField = (byte)ComputeScalar(operand.GetValue(ctx).AsChar); break;            case PrimitiveDataType.ShortType: res.DataField = (short)ComputeScalar(operand.GetValue(ctx).AsShort); break;            case PrimitiveDataType.IntType: res.DataField = (int)ComputeScalar(operand.GetValue(ctx).AsInt); break;            case PrimitiveDataType.FloatType: res.DataField = (float)ComputeFloat(operand.GetValue(ctx).AsFloat); break;            default:                throw new RuntimeException(string.Format("\"{0}\" doesn't support type \"{1}\".", this, leftType.ToString()));        } // switch        return res;    }}public override Operand.Operand Evaluate(Context ctx, List<Operand.Operand> operands){    Debug.Assert(operands.Count > 0);    return Compute(ctx, operands.First());}

具体到某一表达式,以加法为例,其实现如下:

public class Add : LowLevelArithmetic{    public override string ToString()    {        return "+";    }    public override int ComputeScalar(int left, int right)    {        return left + right;    }    public override float ComputeFloat(float left, float right)    {        return left + right;    }}

逻辑运算以and为例:

public class And : LogicOperator{    public override string ToString()    {        return "&&";    }    public override int ComputeScalar(int left, int right)    {        return left != 0 && right != 0 ? 1 : 0;    }    public override float ComputeFloat(float left, float right)    {        return left != 0 && right != 0 ? 1 : 0;    }}

一元操作符,以简单点的取负为例:

public class Minus : UnaryOperator{    public override string ToString()    {        return "-";    }    public override int ComputeScalar(int val)    {        return - val;    }    public override float ComputeFloat(float val)    {        return -val;    }}

复杂的点的如自增,以后自增为例:

public class PostfixIncrease : UnaryOperator{    public override string ToString()    {        return "++";    }    public override int ComputeScalar(int val)    {        return val + 1;    }    public override Operand.Operand Evaluate(Context ctx, List<Operand.Operand> operands)    {        Operand.Operand theOperand = operands.First();        if (theOperand is Operand.Value || theOperand is Operand.ValueOfVariableReference)        {            Operand.Operand res = theOperand.GetValue(ctx);            Operand.Operand computerRres = base.Evaluate(ctx, operands);            if (theOperand is Operand.ValueOfVariableReference)            {                (theOperand as Operand.ValueOfVariableReference).GetReferencedVariable(ctx).Assign(computerRres.GetValue(ctx));                }            return res;        }        if (theOperand is Operand.ValueOfPointerIndiction)        {            Operand.Value val = theOperand.GetValue(ctx);            switch (theOperand.GetTypeInfo(ctx).BaseType)            {                case PrimitiveDataType.CharType: (theOperand as Operand.ValueOfPointerIndiction).Address++; break;                case PrimitiveDataType.ShortType: (theOperand as Operand.ValueOfPointerIndiction).Address += 2; break;                default: (theOperand as Operand.ValueOfPointerIndiction).Address += 4; break;            }            return val;        }        throw new RuntimeException(string.Format("\"{0}\" doesn't support type \"{1}\".", this, theOperand.GetTypeInfo(ctx).ToString()));    }}

自增需要考虑返回值是运算之前的值还是运算之后。
特殊的取地址操作其实比较简单,因为限制其只能作用于变量:

public class GetAddress : UnaryOperator{    public override string ToString()    {        return "&";    }    public override Operand.Value Compute(Context ctx, Operand.Operand operand)    {        if (operand is Operand.ValueOfVariableReference)        {            return new Operand.Value()            {                DataType = PrimitiveDataType.IntType,                DataField = (operand as Operand.ValueOfVariableReference).GetReferencedVariable(ctx).Address             };        }        else        {            throw new RuntimeException(string.Format("\"{0}\" only support variable.", this));        }    }}

括号操作符不属于一元、二元之列,单独一类:

public class Parentheses : Operator{    public Parentheses()        : base(OperatorPriority.Parenthese)    {    }    public override string ToString()    {        return "(";    }    public override Operand.Operand Evaluate(Context ctx, List<Operand.Operand> operands)    {        return operands.First();    }}