tclparseexpr.c

tcl是工具命令语言

/* 
 * tclParseExpr.c --
 *
 *	This file contains procedures that parse Tcl expressions. They
 *	do so in a general-purpose fashion that can be used for many
 *	different purposes, including compilation, direct execution,
 *	code analysis, etc.
 *
 * Copyright (c) 1997 Sun Microsystems, Inc.
 * Copyright (c) 1998-2000 by Scriptics Corporation.
 * Contributions from Don Porter, NIST, 2002.  (not subject to US copyright)
 *
 * See the file "license.terms" for information on usage and redistribution
 * of this file, and for a DISCLAIMER OF ALL WARRANTIES.
 *
 * RCS: @(#) $Id: tclParseExpr.c,v 1.17 2003/02/16 01:36:32 msofer Exp $
 */

#include "tclInt.h"

/*
 * The stuff below is a bit of a hack so that this file can be used in
 * environments that include no UNIX, i.e. no errno: just arrange to use
 * the errno from tclExecute.c here.
 */

#ifndef TCL_GENERIC_ONLY
#include "tclPort.h"
#else
#define NO_ERRNO_H
#endif

#ifdef NO_ERRNO_H
extern int errno;			/* Use errno from tclExecute.c. */
#define ERANGE 34
#endif

/*
 * Boolean variable that controls whether expression parse tracing
 * is enabled.
 */

#ifdef TCL_COMPILE_DEBUG
static int traceParseExpr = 0;
#endif /* TCL_COMPILE_DEBUG */

/*
 * The ParseInfo structure holds state while parsing an expression.
 * A pointer to an ParseInfo record is passed among the routines in
 * this module.
 */

typedef struct ParseInfo {
    Tcl_Parse *parsePtr;	/* Points to structure to fill in with
				 * information about the expression. */
    int lexeme;			/* Type of last lexeme scanned in expr.
				 * See below for definitions. Corresponds to
				 * size characters beginning at start. */
    CONST char *start;		/* First character in lexeme. */
    int size;			/* Number of bytes in lexeme. */
    CONST char *next;		/* Position of the next character to be
				 * scanned in the expression string. */
    CONST char *prevEnd;	/* Points to the character just after the
				 * last one in the previous lexeme. Used to
				 * compute size of subexpression tokens. */
    CONST char *originalExpr;	/* Points to the start of the expression
				 * originally passed to Tcl_ParseExpr. */
    CONST char *lastChar;	/* Points just after last byte of expr. */
} ParseInfo;

/*
 * Definitions of the different lexemes that appear in expressions. The
 * order of these must match the corresponding entries in the
 * operatorStrings array below.
 *
 * Basic lexemes:
 */

#define LITERAL		0
#define FUNC_NAME	1
#define OPEN_BRACKET	2
#define OPEN_BRACE	3
#define OPEN_PAREN	4
#define CLOSE_PAREN	5
#define DOLLAR		6
#define QUOTE		7
#define COMMA		8
#define END		9
#define UNKNOWN		10
#define UNKNOWN_CHAR	11

/*
 * Binary numeric operators:
 */

#define MULT		12
#define DIVIDE		13
#define MOD		14
#define PLUS		15
#define MINUS		16
#define LEFT_SHIFT	17
#define RIGHT_SHIFT	18
#define LESS		19
#define GREATER		20
#define LEQ		21
#define GEQ		22
#define EQUAL		23
#define NEQ		24
#define BIT_AND		25
#define BIT_XOR		26
#define BIT_OR		27
#define AND		28
#define OR		29
#define QUESTY		30
#define COLON		31

/*
 * Unary operators. Unary minus and plus are represented by the (binary)
 * lexemes MINUS and PLUS.
 */

#define NOT		32
#define BIT_NOT		33

/*
 * Binary string operators:
 */

#define STREQ		34
#define STRNEQ		35

/*
 * Mapping from lexemes to strings; used for debugging messages. These
 * entries must match the order and number of the lexeme definitions above.
 */

static char *lexemeStrings[] = {
    "LITERAL", "FUNCNAME",
    "[", "{", "(", ")", "$", "\"", ",", "END", "UNKNOWN", "UNKNOWN_CHAR",
    "*", "/", "%", "+", "-",
    "<<", ">>", "<", ">", "<=", ">=", "==", "!=",
    "&", "^", "|", "&&", "||", "?", ":",
    "!", "~", "eq", "ne",
};

/*
 * Declarations for local procedures to this file:
 */

static int		GetLexeme _ANSI_ARGS_((ParseInfo *infoPtr));
static void		LogSyntaxError _ANSI_ARGS_((ParseInfo *infoPtr,
				CONST char *extraInfo));
static int		ParseAddExpr _ANSI_ARGS_((ParseInfo *infoPtr));
static int		ParseBitAndExpr _ANSI_ARGS_((ParseInfo *infoPtr));
static int		ParseBitOrExpr _ANSI_ARGS_((ParseInfo *infoPtr));
static int		ParseBitXorExpr _ANSI_ARGS_((ParseInfo *infoPtr));
static int		ParseCondExpr _ANSI_ARGS_((ParseInfo *infoPtr));
static int		ParseEqualityExpr _ANSI_ARGS_((ParseInfo *infoPtr));
static int		ParseLandExpr _ANSI_ARGS_((ParseInfo *infoPtr));
static int		ParseLorExpr _ANSI_ARGS_((ParseInfo *infoPtr));
static int		ParseMaxDoubleLength _ANSI_ARGS_((CONST char *string,
				CONST char *end));
static int		ParseMultiplyExpr _ANSI_ARGS_((ParseInfo *infoPtr));
static int		ParsePrimaryExpr _ANSI_ARGS_((ParseInfo *infoPtr));
static int		ParseRelationalExpr _ANSI_ARGS_((ParseInfo *infoPtr));
static int		ParseShiftExpr _ANSI_ARGS_((ParseInfo *infoPtr));
static int		ParseUnaryExpr _ANSI_ARGS_((ParseInfo *infoPtr));
static void		PrependSubExprTokens _ANSI_ARGS_((CONST char *op,
				int opBytes, CONST char *src, int srcBytes,
				int firstIndex, ParseInfo *infoPtr));

/*
 * Macro used to debug the execution of the recursive descent parser used
 * to parse expressions.
 */

#ifdef TCL_COMPILE_DEBUG
#define HERE(production, level) \
    if (traceParseExpr) { \
	fprintf(stderr, "%*s%s: lexeme=%s, next=\"%.20s\"\n", \
		(level), " ", (production), \
		lexemeStrings[infoPtr->lexeme], infoPtr->next); \
    }
#else
#define HERE(production, level)
#endif /* TCL_COMPILE_DEBUG */

/*
 *----------------------------------------------------------------------
 *
 * Tcl_ParseExpr --
 *
 *	Given a string, this procedure parses the first Tcl expression
 *	in the string and returns information about the structure of
 *	the expression. This procedure is the top-level interface to the
 *	the expression parsing module.  No more that numBytes bytes will
 *	be scanned.
 *
 * Results:
 *	The return value is TCL_OK if the command was parsed successfully
 *	and TCL_ERROR otherwise. If an error occurs and interp isn't NULL
 *	then an error message is left in its result. On a successful return,
 *	parsePtr is filled in with information about the expression that 
 *	was parsed.
 *
 * Side effects:
 *	If there is insufficient space in parsePtr to hold all the
 *	information about the expression, then additional space is
 *	malloc-ed. If the procedure returns TCL_OK then the caller must
 *	eventually invoke Tcl_FreeParse to release any additional space
 *	that was allocated.
 *
 *----------------------------------------------------------------------
 */

int
Tcl_ParseExpr(interp, string, numBytes, parsePtr)
    Tcl_Interp *interp;		/* Used for error reporting. */
    CONST char *string;		/* The source string to parse. */
    int numBytes;		/* Number of bytes in string. If < 0, the
				 * string consists of all bytes up to the
				 * first null character. */
    Tcl_Parse *parsePtr;	/* Structure to fill with information about
				 * the parsed expression; any previous
				 * information in the structure is
				 * ignored. */
{
    ParseInfo info;
    int code;

    if (numBytes < 0) {
	numBytes = (string? strlen(string) : 0);
    }
#ifdef TCL_COMPILE_DEBUG
    if (traceParseExpr) {
	fprintf(stderr, "Tcl_ParseExpr: string=\"%.*s\"\n",
	        numBytes, string);
    }
#endif /* TCL_COMPILE_DEBUG */
    
    parsePtr->commentStart = NULL;
    parsePtr->commentSize = 0;
    parsePtr->commandStart = NULL;
    parsePtr->commandSize = 0;
    parsePtr->numWords = 0;
    parsePtr->tokenPtr = parsePtr->staticTokens;
    parsePtr->numTokens = 0;
    parsePtr->tokensAvailable = NUM_STATIC_TOKENS;
    parsePtr->string = string;
    parsePtr->end = (string + numBytes);
    parsePtr->interp = interp;
    parsePtr->term = string;
    parsePtr->incomplete = 0;

    /*
     * Initialize the ParseInfo structure that holds state while parsing
     * the expression.
     */

    info.parsePtr = parsePtr;
    info.lexeme = UNKNOWN;
    info.start = NULL;
    info.size = 0;
    info.next = string;
    info.prevEnd = string;
    info.originalExpr = string;
    info.lastChar = (string + numBytes); /* just after last char of expr */

    /*
     * Get the first lexeme then parse the expression.
     */

    code = GetLexeme(&info);
    if (code != TCL_OK) {
	goto error;
    }
    code = ParseCondExpr(&info);
    if (code != TCL_OK) {
	goto error;
    }
    if (info.lexeme != END) {
	LogSyntaxError(&info, "extra tokens at end of expression");
	goto error;
    }
    return TCL_OK;
    
    error:
    if (parsePtr->tokenPtr != parsePtr->staticTokens) {
	ckfree((char *) parsePtr->tokenPtr);
    }
    return TCL_ERROR;
}

/*
 *----------------------------------------------------------------------
 *
 * ParseCondExpr --
 *
 *	This procedure parses a Tcl conditional expression:
 *	condExpr ::= lorExpr ['?' condExpr ':' condExpr]
 *
 *	Note that this is the topmost recursive-descent parsing routine used
 *	by Tcl_ParseExpr to parse expressions. This avoids an extra procedure
 *	call since such a procedure would only return the result of calling
 *	ParseCondExpr. Other recursive-descent procedures that need to parse
 *	complete expressions also call ParseCondExpr.
 *
 * Results:
 *	The return value is TCL_OK on a successful parse and TCL_ERROR
 *	on failure. If TCL_ERROR is returned, then the interpreter's result
 *	contains an error message.
 *
 * Side effects:
 *	If there is insufficient space in parsePtr to hold all the
 *	information about the subexpression, then additional space is
 *	malloc-ed.
 *
 *----------------------------------------------------------------------
 */

static int
ParseCondExpr(infoPtr)
    ParseInfo *infoPtr;		/* Holds the parse state for the
				 * expression being parsed. */
{
    Tcl_Parse *parsePtr = infoPtr->parsePtr;
    Tcl_Token *tokenPtr, *firstTokenPtr, *condTokenPtr;
    int firstIndex, numToMove, code;
    CONST char *srcStart;
    
    HERE("condExpr", 1);
    srcStart = infoPtr->start;
    firstIndex = parsePtr->numTokens;
    
    code = ParseLorExpr(infoPtr);
    if (code != TCL_OK) {
	return code;
    }
    
    if (infoPtr->lexeme == QUESTY) {
	/*
	 * Emit two tokens: one TCL_TOKEN_SUB_EXPR token for the entire
	 * conditional expression, and a TCL_TOKEN_OPERATOR token for 
	 * the "?" operator. Note that these two tokens must be inserted
	 * before the LOR operand tokens generated above.
	 */

	if ((parsePtr->numTokens + 1) >= parsePtr->tokensAvailable) {
	    TclExpandTokenArray(parsePtr);
	}
	firstTokenPtr = &parsePtr->tokenPtr[firstIndex];
	tokenPtr = (firstTokenPtr + 2);
	numToMove = (parsePtr->numTokens - firstIndex);
	memmove((VOID *) tokenPtr, (VOID *) firstTokenPtr,
	        (size_t) (numToMove * sizeof(Tcl_Token)));
	parsePtr->numTokens += 2;
	
	tokenPtr = firstTokenPtr;
	tokenPtr->type = TCL_TOKEN_SUB_EXPR;
	tokenPtr->start = srcStart;
	
	tokenPtr++;
	tokenPtr->type = TCL_TOKEN_OPERATOR;
	tokenPtr->start = infoPtr->start;
	tokenPtr->size = 1;
	tokenPtr->numComponents = 0;
    
	/*
	 * Skip over the '?'.
	 */
	
	code = GetLexeme(infoPtr); 
	if (code != TCL_OK) {
	    return code;
	}

	/*
	 * Parse the "then" expression.
	 */

	code = ParseCondExpr(infoPtr);
	if (code != TCL_OK) {
	    return code;
	}
	if (infoPtr->lexeme != COLON) {
	    LogSyntaxError(infoPtr, "missing colon from ternary conditional");
	    return TCL_ERROR;
	}
	code = GetLexeme(infoPtr); /* skip over the ':' */
	if (code != TCL_OK) {
	    return code;
	}

	/*
	 * Parse the "else" expression.
	 */

	code = ParseCondExpr(infoPtr);
	if (code != TCL_OK) {
	    return code;
	}

	/*
	 * Now set the size-related fields in the '?' subexpression token.
	 */

	condTokenPtr = &parsePtr->tokenPtr[firstIndex];
	condTokenPtr->size = (infoPtr->prevEnd - srcStart);
	condTokenPtr->numComponents = parsePtr->numTokens - (firstIndex+1);
    }
    return TCL_OK;
}

/*
 *----------------------------------------------------------------------
 *
 * ParseLorExpr --
 *
 *	This procedure parses a Tcl logical or expression:
 *	lorExpr ::= landExpr {'||' landExpr}
 *
 * Results:
 *	The return value is TCL_OK on a successful parse and TCL_ERROR
 *	on failure. If TCL_ERROR is returned, then the interpreter's result
 *	contains an error message.
 *
 * Side effects:
 *	If there is insufficient space in parsePtr to hold all the
 *	information about the subexpression, then additional space is
 *	malloc-ed.
 *
 *----------------------------------------------------------------------
 */

static int
ParseLorExpr(infoPtr)
    ParseInfo *infoPtr;		/* Holds the parse state for the
				 * expression being parsed. */
{
    Tcl_Parse *parsePtr = infoPtr->parsePtr;
    int firstIndex, code;
    CONST char *srcStart, *operator;
    
    HERE("lorExpr", 2);
    srcStart = infoPtr->start;
    firstIndex = parsePtr->numTokens;
    
    code = ParseLandExpr(infoPtr);
    if (code != TCL_OK) {
	return code;
    }

    while (infoPtr->lexeme == OR) {
	operator = infoPtr->start;
	code = GetLexeme(infoPtr); /* skip over the '||' */
	if (code != TCL_OK) {
	    return code;
	}
	code = ParseLandExpr(infoPtr);
	if (code != TCL_OK) {
	    return code;
	}

	/*
	 * Generate tokens for the LOR subexpression and the '||' operator.
	 */

	PrependSubExprTokens(operator, 2, srcStart,
	        (infoPtr->prevEnd - srcStart), firstIndex, infoPtr);
    }
    return TCL_OK;
}

/*
 *----------------------------------------------------------------------
 *
 * ParseLandExpr --
 *
 *	This procedure parses a Tcl logical and expression:
 *	landExpr ::= bitOrExpr {'&&' bitOrExpr}
 *
 * Results:
 *	The return value is TCL_OK on a successful parse and TCL_ERROR
 *	on failure. If TCL_ERROR is returned, then the interpreter's result
 *	contains an error message.
 *
 * Side effects:
 *	If there is insufficient space in parsePtr to hold all the
 *	information about the subexpression, then additional space is
 *	malloc-ed.
 *
 *----------------------------------------------------------------------
 */

static int
ParseLandExpr(infoPtr)
    ParseInfo *infoPtr;		/* Holds the parse state for the
				 * expression being parsed. */
{
    Tcl_Parse *parsePtr = infoPtr->parsePtr;
    int firstIndex, code;
    CONST char *srcStart, *operator;

    HERE("landExpr", 3);
    srcStart = infoPtr->start;
    firstIndex = parsePtr->numTokens;
    
    code = ParseBitOrExpr(infoPtr);
    if (code != TCL_OK) {
	return code;
    }

    while (infoPtr->lexeme == AND) {
	operator = infoPtr->start;
	code = GetLexeme(infoPtr); /* skip over the '&&' */
	if (code != TCL_OK) {
	    return code;
	}
	code = ParseBitOrExpr(infoPtr);
	if (code != TCL_OK) {
	    return code;
	}

	/*
	 * Generate tokens for the LAND subexpression and the '&&' operator.
	 */