hooks.c

这是数据挖掘算法的cubist算法的具体实现.运行在dos下.

    if ( ! Conjunct() ) FailSyn("expression");

    while ( Find("or") )
    {
	BN += 2;

	if ( ! Conjunct() ) FailSyn("expression");

	DumpOp(OP_OR, Fi);
    }

    return true;
}



Boolean Conjunct()
/*      --------  */
{
    int		Fi=BN;

    if ( ! SExpression() ) FailSyn("expression");

    while ( Find("and") )
    {
	BN += 3;

	if ( ! SExpression() ) FailSyn("expression");

	DumpOp(OP_AND, Fi);
    }

    return true;
}



String RelOps[] = {">=", "<=", "!=", "<>", ">", "<", "=", (String) 0};

Boolean SExpression()
/*      -----------  */
{
    int		o, Fi=BN;

    if ( ! AExpression() ) FailSyn("expression");

    if ( (o = FindOne(RelOps)) >= 0 )
    {
	BN += strlen(RelOps[o]);

	if ( ! AExpression() ) FailSyn("expression");

	DumpOp(( o == 0 ? OP_GE :
		 o == 1 ? OP_LE :
		 o == 4 ? OP_GT :
		 o == 5 ? OP_LT :
		 o == 2 || o == 3 ?
			( TStack[TSN-1].Type == 'S' ? OP_SNE : OP_NE ) :
			( TStack[TSN-1].Type == 'S' ? OP_SEQ : OP_EQ ) ), Fi);
    }

    return true;
}



String AddOps[] = {"+", "-", (String) 0};

Boolean AExpression()
/*      -----------  */
{
    int		o, Fi=BN;

    if ( Buff[BN] == ' ' ) BN++;

    if ( (o = FindOne(AddOps)) >= 0 )
    {
	BN += 1;
    }

    if ( ! Term() ) FailSyn("expression");

    if ( o == 1 ) DumpOp(OP_UMINUS, Fi);

    while ( (o = FindOne(AddOps)) >= 0 )
    {
	BN += 1;

	if ( ! Term() ) FailSyn("arithmetic expression");

	DumpOp((char)(OP_PLUS + o), Fi);
    }

    return true;
}



String MultOps[] = {"*", "/", "%", (String) 0};

Boolean Term()
/*      ----  */
{
    int		o, Fi=BN;

    if ( ! Factor() ) FailSyn("expression");

    while ( (o = FindOne(MultOps)) >= 0 )
    {
	BN += 1;

	if ( ! Factor() ) FailSyn("arithmetic expression");

	DumpOp((char)(OP_MULT + o), Fi);
    }

    return true;
}



Boolean Factor()
/*      ----  */
{
    int		Fi=BN;

    if ( ! Primary() ) FailSyn("value");

    while ( Find("^") )
    {
	BN += 1;

	if ( ! Primary() ) FailSyn("exponent");

	DumpOp(OP_POW, Fi);
    }

    return true;
}



Boolean Primary()
/*      -------  */
{
    if ( Atom() )
    {
	return true;
    }
    else
    if ( Find("(") )
    {
	BN++;
	if ( ! Expression() ) FailSyn("expression in parentheses");
	if ( ! Find(")") ) FailSyn("')'");
	BN++;
	return true;
    }
    else
    {
	FailSyn("attribute, value, or '('");
    }
}



String Funcs[] = {"sin", "cos", "tan", "log", "exp", "int", (String) 0};

Boolean Atom()
/*      ----  */
{
    char	*EndPtr, *Str, Date[11], Time[9];
    int		o, FirstBN, Fi=BN;
    ContValue	F;
    Attribute	Att;

    if ( Buff[BN] == ' ' ) BN++;

    if ( Buff[BN] == '"' )
    {
	FirstBN = ++BN;
	while ( Buff[BN] != '"' )
	{
	    if ( ! Buff[BN] ) FailSyn("closing '\"'");
	    BN++;
	}

	/*  Make a copy of the string without double quotes  */

	Buff[BN] = '\00';
	Str = strdup(Buff + FirstBN);

	Buff[BN++] = '"';
	Dump(OP_STR, 0, Str, Fi);
    }
    else
    if ( (Att = FindAttName()) )
    {
	BN += strlen(AttName[Att]);

	Dump(OP_ATT, 0, (String) (long) Att, Fi);
    }
    else
    if ( isdigit(Buff[BN]) )
    {
	/*  Check for date or time first  */

	if ( ( Buff[BN+4] == '/' && Buff[BN+7] == '/' ||
	       Buff[BN+4] == '-' && Buff[BN+7] == '-' )&&
	     isdigit(Buff[BN+1]) && isdigit(Buff[BN+2]) &&
		isdigit(Buff[BN+3]) &&
	     isdigit(Buff[BN+5]) && isdigit(Buff[BN+6]) &&
	     isdigit(Buff[BN+8]) && isdigit(Buff[BN+9]) )
	{
	    memcpy(Date, Buff+BN, 10);
	    Date[10] = '\00';
	    if ( (F = DateToDay(Date)) == 0 )
	    {
		Error(BADDEF1, Date, "date");
	    }

	    BN += 10;
	}
	else
	if ( Buff[BN+2] == ':' && Buff[BN+5] == ':' &&
	     isdigit(Buff[BN+1]) &&
	     isdigit(Buff[BN+3]) && isdigit(Buff[BN+4]) &&
	     isdigit(Buff[BN+6]) && isdigit(Buff[BN+7]) )
	{
	    memcpy(Time, Buff+BN, 8);
	    Time[8] = '\00';
	    if ( (F = TimeToSecs(Time)) == 0 )
	    {
		Error(BADDEF1, Time, "time");
	    }

	    BN += 8;
	}
	else
	{
	    F = strtod(Buff+BN, &EndPtr);

	    /*  Check for period after integer  */

	    if ( EndPtr > Buff+BN+1 && *(EndPtr-1) == '.' )
	    {
		EndPtr--;
	    }

	    BN = EndPtr - Buff;
	}

	Dump(OP_NUM, F, Nil, Fi);
    }
    else
    if ( (o = FindOne(Funcs)) >= 0 )
    {
	BN += 3;

	if ( ! Find("(") ) FailSyn("'(' after function name");
	BN++;

	if ( ! Expression() ) FailSyn("expression");

	if ( ! Find(")") ) FailSyn("')' after function argument");
	BN++;

	DumpOp((char)(OP_SIN + o), Fi);
    }
    else
    if ( Buff[BN] == '?' )
    {
	BN++;
	if ( TStack[TSN-1].Type == 'N' )
	{
	    Dump(OP_NUM, _UNK.cval, Nil, Fi);
	}
	else
	{
	    Dump(OP_STR, 0, Nil, Fi);
	}
    }
    else
    if ( ! memcmp(Buff+BN, "N/A", 3) )
    {
	BN += 3;
	if ( TStack[TSN-1].Type == 'N' )
	{
	    Dump(OP_NUM, _NA.cval, Nil, Fi);
	}
	else
	{
	    Dump(OP_STR, 0, strdup("N/A"), Fi);
	}
    }
    else
    {
	return false;
    }

    return true;
}



/*************************************************************************/
/*									 */
/*	Skip spaces and check for specific string			 */
/*									 */
/*************************************************************************/


Boolean Find(String S)
/*      ----  */
{
    if ( Buff[BN] == ' ' ) BN++;

    return ( ! Buff[BN] ? false : ! memcmp(Buff+BN, S, strlen(S)) );
}



/*************************************************************************/
/*									 */
/*	Find one of a zero-terminated list of alternatives		 */
/*									 */
/*************************************************************************/


int FindOne(String *Alt)
/*  -------  */
{
    int	a;

    for ( a = 0 ; Alt[a] ; a++ )
    {
	if ( Find(Alt[a]) ) return a;
    }

    return -1;
}



/*************************************************************************/
/*									 */
/*	Find an attribute name						 */
/*									 */
/*************************************************************************/


Attribute FindAttName()
/*        -----------  */
{
    Attribute	Att, LongestAtt=0;

    ForEach(Att, 1, MaxAtt-1)
    {
	if ( ! Exclude(Att) && Find(AttName[Att]) )
	{
	    if ( ! LongestAtt ||
		 strlen(AttName[Att]) > strlen(AttName[LongestAtt]) )
	    {
		LongestAtt = Att;
	    }
	}
    }

    return LongestAtt;
}



/*************************************************************************/
/*									 */
/*	Error message routines.  Syntax errors come from the		 */
/*	recursive descent parser, semantics errors from the routines	 */
/*	that build up the equivalent polish				 */
/*									 */
/*************************************************************************/


void DefSyntaxError(String Msg)
/*   --------------  */
{
    String	RestOfText;
    int		i=10;

    if ( ! PreviousError )
    {
	RestOfText = Buff + BN;

	/*  Abbreviate text if longer than 12 characters  */

	if ( CharWidth(RestOfText) > 12 )
	{
#ifdef UTF8
	    /*  Find beginning of UTF-8 character  */

	    for ( ; (RestOfText[i] & 0x80) ; i++)
		;
#endif
	    RestOfText[i] = RestOfText[i+1] = '.';
	}

	Error(BADDEF1, RestOfText, Msg);
	PreviousError = true;
    }
}



void DefSemanticsError(int Fi, String Msg, int OpCode)
/*   -----------------  */
{
    char	Exp[1000], XMsg[1000], Op[1000];

    if ( ! PreviousError )
    {
	/*  Abbreviate the input if necessary  */

	if ( BN - Fi > 23 )
	{
	    sprintf(Exp, "%.10s...%.10s", Buff+Fi, Buff+BN-10);
	}
	else
	{
	    sprintf(Exp, "%.*s", BN - Fi, Buff+Fi);
	}

	switch ( OpCode )
	{
	    case OP_AND:	sprintf(Op, "%s", "and"); break;
	    case OP_OR:		sprintf(Op, "%s", "or"); break;
	    case OP_SEQ:
	    case OP_EQ:		sprintf(Op, "%s", "="); break;
	    case OP_SNE:
	    case OP_NE:		sprintf(Op, "%s", "<>"); break;
	    case OP_GT:		sprintf(Op, "%s", ">"); break;
	    case OP_GE:		sprintf(Op, "%s", ">="); break;
	    case OP_LT:		sprintf(Op, "%s", "<"); break;
	    case OP_LE:		sprintf(Op, "%s", "<="); break;
	    case OP_PLUS:	sprintf(Op, "%s", "+"); break;
	    case OP_MINUS:	sprintf(Op, "%s", "-"); break;
	    case OP_UMINUS:	sprintf(Op, "%s", "unary -"); break;
	    case OP_MULT:	sprintf(Op, "%s", "*"); break;
	    case OP_DIV:	sprintf(Op, "%s", "/"); break;
	    case OP_MOD:	sprintf(Op, "%s", "%"); break;
	    case OP_POW:	sprintf(Op, "%s", "^"); break;
	    case OP_SIN:	sprintf(Op, "%s", "sin"); break;
	    case OP_COS:	sprintf(Op, "%s", "cos"); break;
	    case OP_TAN:	sprintf(Op, "%s", "tan"); break;
	    case OP_LOG:	sprintf(Op, "%s", "log"); break;
	    case OP_EXP:	sprintf(Op, "%s", "exp"); break;
	    case OP_INT:	sprintf(Op, "%s", "int");
	}

	sprintf(XMsg, "%s with '%s'", Msg, Op);
	Error(BADDEF2, Exp, XMsg);
	PreviousError = true;
    }
}



/*************************************************************************/
/*									 */
/*	Reverse polish routines.  These use a model of the stack	 */
/*	during expression evaluation to detect type conflicts etc	 */
/*									 */
/*************************************************************************/



void Dump(char OpCode, ContValue F, String S, int Fi)
/*   ----  */
{
    if ( Buff[Fi] == ' ' ) Fi++;

    if ( ! UpdateTStack(OpCode, F, S, Fi) ) return;

    /*  Make sure enough room for this element  */

    if ( DN >= DefSize-1 )
    {
	Realloc(AttDef[MaxAtt], DefSize += 100, DefElt);
    }

    DefOp(AttDef[MaxAtt][DN]) = OpCode;
    if ( OpCode == OP_ATT || OpCode == OP_STR )
    {
	DefSVal(AttDef[MaxAtt][DN]) = S;
    }
    else
    {
	DefNVal(AttDef[MaxAtt][DN]) = F;
    }

    DN++;
}



void DumpOp(char OpCode, int Fi)
/*   ------  */
{
    Dump(OpCode, 0, Nil, Fi);
}



Boolean UpdateTStack(char OpCode, ContValue F, String S, int Fi)
/*      ------------  */
{
    if ( TSN >= TStackSize )
    {
	Realloc(TStack, TStackSize += 50, EltRec);
    }

    switch ( OpCode )
    {
	case OP_ATT:
		TStack[TSN].Type = ( Continuous((long) S) ? 'N' : 'S' );
		break;

	case OP_NUM:
		TStack[TSN].Type = 'N';
		break;

	case OP_STR:
		TStack[TSN].Type = 'S';
		break;

	case OP_AND:
	case OP_OR:
		if ( TStack[TSN-2].Type != 'B' || TStack[TSN-1].Type != 'B' )
		{
		    FailSem("non-logical value");
		}
		TSN -= 2;
		break;

	case OP_EQ:
	case OP_NE:
		if ( TStack[TSN-2].Type != TStack[TSN-1].Type )
		{
		    FailSem("incompatible values");
		}
		TSN -= 2;
		TStack[TSN].Type = 'B';
		break;

	case OP_GT:
	case OP_GE:
	case OP_LT:
	case OP_LE:
		if ( TStack[TSN-2].Type != 'N' || TStack[TSN-1].Type != 'N' )
		{
		    FailSem("non-arithmetic value");
		}
		TSN -= 2;
		TStack[TSN].Type = 'B';
		break;

	case OP_SEQ:
	case OP_SNE:
		if ( TStack[TSN-2].Type != 'S' || TStack[TSN-1].Type != 'S' )
		{
		    FailSem("incompatible values");
		}
		TSN -= 2;
		TStack[TSN].Type = 'B';
		break;

	case OP_PLUS:
	case OP_MINUS:
	case OP_MULT:
	case OP_DIV:
	case OP_MOD:
	case OP_POW:
		if ( TStack[TSN-2].Type != 'N' || TStack[TSN-1].Type != 'N' )
		{
		    FailSem("non-arithmetic value");
		}
		TSN -= 2;
		break;

	case OP_UMINUS:
		if ( TStack[TSN-1].Type != 'N' )
		{
		    FailSem("non-arithmetic value");
		}
		TSN--;
		break;

	case OP_SIN:
	case OP_COS:
	case OP_TAN:
	case OP_LOG:
	case OP_EXP:
	case OP_INT:
		if ( TStack[TSN-1].Type != 'N' )
		{
		    FailSem("non-arithmetic argument");
		}
		TSN--;
    }

    TStack[TSN].Fi = Fi;
    TStack[TSN].Li = BN-1;
    TSN++;

    return true;
}



/*************************************************************************/
/*									 */
/*	Evaluate an implicit attribute for a case			 */
/*									 */
/*************************************************************************/

#define	Unknown(c,a)		(! DVal(c,a))
#define	CUnknownVal(AV)		(AV.cval==UNKNOWN)
#define	DUnknownVal(AV)		(! AV.dval)
#define DUNA(a)	(DUnknownVal(XStack[a]) || NotApplicVal(XStack[a]))
#define CUNA(a)	(CUnknownVal(XStack[a]) || NotApplicVal(XStack[a]))
#define	C1(x)	(CUNA(XSN-1) ? _UNK.cval : (x))
#define	C2(x)	(CUNA(XSN-1) || CUNA(XSN-2) ? _UNK.cval : (x))
#define	CD2(x)	(CUNA(XSN-1) || CUNA(XSN-2) ? 0 : (x))
#define	D2(x)	(DUNA(XSN-1) || DUNA(XSN-2) ? 0 : (x))


AttValue EvaluateDef(Definition D, Description Case)
/*       -----------  */
{
    XStackElt	XStack[100];			/* allows 100-level nesting  */
    int		XSN=0, DN, bv1, bv2, Mult;
    double	cv1, cv2;
    String	sv1, sv2;
    Attribute	Att;
    DefElt	DElt;
    AttValue	ReturnVal;

    for ( DN = 0 ; ; DN++)
    {
	switch ( DefOp((DElt = D[DN])) )
	{
	    case OP_ATT:
		    Att = (long) DefSVal(DElt);

#if defined PREDICT || defined SEE5 && defined WIN32 && ! defined _CONSOLE
		    GetValue(Att);
#endif
		    if ( Continuous(Att) )
		    {
			XStack[XSN++].cval = CVal(Case, Att);
		    }
		    else
		    {