Tcl Source Code

{
    /*
     * Remember to keep code here in some sync with the byte-compiled versions
     * in tclExecute.c (INST_STR_EQ, INST_STR_NEQ and INST_STR_CMP as well as
     * the expr string comparison in INST_EQ/INST_NEQ/INST_LT/...).
     */

    const char *string1, *string2;
    int length1, length2, i, match, length, nocase = 0, reqlength = -1;
    typedef int (*strCmpFn_t)(const char *, const char *, unsigned int);
    strCmpFn_t strCmpFn;

    if (objc < 3 || objc > 6) {
    str_cmp_args:
	Tcl_WrongNumArgs(interp, 1, objv,
		"?-nocase? ?-length int? string1 string2");
	return TCL_ERROR;
    }

    for (i = 1; i < objc-2; i++) {
	string2 = TclGetStringFromObj(objv[i], &length2);
	if ((length2 > 1) && !strncmp(string2, "-nocase", (size_t)length2)) {
	    nocase = 1;
	} else if ((length2 > 1)
		&& !strncmp(string2, "-length", (size_t)length2)) {
	    if (i+1 >= objc-2) {
		goto str_cmp_args;
	    }
	    i++;
	    if (TclGetIntFromObj(interp, objv[i], &reqlength) != TCL_OK) {
		return TCL_ERROR;
	    }

    /*
     * From now on, we only access the two objects at the end of the argument
     * array.
     */

    objv += objc-2;

    if ((reqlength == 0) || (objv[0] == objv[1])) {
	/*
	 * Always match at 0 chars of if it is the same obj.
	 */

	Tcl_SetObjResult(interp, Tcl_NewBooleanObj(1));
	return TCL_OK;
    }

    if (!nocase && TclIsPureByteArray(objv[0]) &&
	    TclIsPureByteArray(objv[1])) {
	/*
	 * Use binary versions of comparisons since that won't cause undue
	 * type conversions and it is much faster. Only do this if we're
	 * case-sensitive (which is all that really makes sense with byte
	 * arrays anyway, and we have no memcasecmp() for some reason... :^)
	 */

	string1 = (char *) Tcl_GetByteArrayFromObj(objv[0], &length1);
	string2 = (char *) Tcl_GetByteArrayFromObj(objv[1], &length2);
	strCmpFn = (strCmpFn_t) memcmp;
    } else if ((objv[0]->typePtr == &tclStringType)
	    && (objv[1]->typePtr == &tclStringType)) {
	/*
	 * Do a unicode-specific comparison if both of the args are of String
	 * type. In benchmark testing this proved the most efficient check
	 * between the unicode and string comparison operations.
	 */

	string1 = (char *) Tcl_GetUnicodeFromObj(objv[0], &length1);
	string2 = (char *) Tcl_GetUnicodeFromObj(objv[1], &length2);
	strCmpFn = (strCmpFn_t)
		(nocase ? Tcl_UniCharNcasecmp : Tcl_UniCharNcmp);
    } else {
	/*
	 * As a catch-all we will work with UTF-8. We cannot use memcmp() as
	 * that is unsafe with any string containing NUL (\xC0\x80 in Tcl's
	 * utf rep). We can use the more efficient TclpUtfNcmp2 if we are
	 * case-sensitive and no specific length was requested.
	 */

	string1 = (char *) TclGetStringFromObj(objv[0], &length1);
	string2 = (char *) TclGetStringFromObj(objv[1], &length2);
	if ((reqlength < 0) && !nocase) {
	    strCmpFn = (strCmpFn_t) TclpUtfNcmp2;
	} else {
	    length1 = Tcl_NumUtfChars(string1, length1);
	    length2 = Tcl_NumUtfChars(string2, length2);
	    strCmpFn = (strCmpFn_t) (nocase ? Tcl_UtfNcasecmp : Tcl_UtfNcmp);
	}
    }

    if ((reqlength < 0) && (length1 != length2)) {
	match = 1;		/* This will be reversed below. */
    } else {
	length = (length1 < length2) ? length1 : length2;
	if (reqlength > 0 && reqlength < length) {
	    length = reqlength;
	} else if (reqlength < 0) {
	    /*
	     * The requested length is negative, so we ignore it by setting it
	     * to length + 1 so we correct the match var.
	     */

	    reqlength = length + 1;
	}

	match = strCmpFn(string1, string2, (unsigned) length);
	if ((match == 0) && (reqlength > length)) {
	    match = length1 - length2;
	}
    }

    Tcl_SetObjResult(interp, Tcl_NewBooleanObj(match ? 0 : 1));
    return TCL_OK;
}

/*
 *----------------------------------------------------------------------

{
    /*
     * Remember to keep code here in some sync with the byte-compiled versions
     * in tclExecute.c (INST_STR_EQ, INST_STR_NEQ and INST_STR_CMP as well as
     * the expr string comparison in INST_EQ/INST_NEQ/INST_LT/...).
     */


    const char *string1, *string2;


    int length1, length2, i, match, length, nocase = 0, reqlength = -1;


    typedef int (*strCmpFn_t)(const char *, const char *, unsigned int);





    strCmpFn_t strCmpFn;













    if (objc < 3 || objc > 6) {
    str_cmp_args:
	Tcl_WrongNumArgs(interp, 1, objv,
		"?-nocase? ?-length int? string1 string2");
	return TCL_ERROR;
    }

    for (i = 1; i < objc-2; i++) {
	string2 = TclGetStringFromObj(objv[i], &length2);
	if ((length2 > 1) && !strncmp(string2, "-nocase", (size_t)length2)) {
	    nocase = 1;
	} else if ((length2 > 1)
		&& !strncmp(string2, "-length", (size_t)length2)) {
	    if (i+1 >= objc-2) {
		goto str_cmp_args;
	    }
	    i++;
	    if (TclGetIntFromObj(interp, objv[i], &reqlength) != TCL_OK) {
		return TCL_ERROR;
	    }
	} else {
	    Tcl_SetObjResult(interp, Tcl_ObjPrintf(
		    "bad option \"%s\": must be -nocase or -length",
		    string2));
	    Tcl_SetErrorCode(interp, "TCL", "LOOKUP", "INDEX", "option",
		    string2, NULL);
	    return TCL_ERROR;
	}
    }

    /*
     * From now on, we only access the two objects at the end of the argument
     * array.
     */

    objv += objc-2;

    if ((reqlength == 0) || (objv[0] == objv[1])) {
	/*
	 * Always match at 0 chars of if it is the same obj.
	 */

	Tcl_SetObjResult(interp, Tcl_NewBooleanObj(0));
	return TCL_OK;
    }

    if (!nocase && TclIsPureByteArray(objv[0]) &&
	    TclIsPureByteArray(objv[1])) {
	/*
	 * Use binary versions of comparisons since that won't cause undue
	 * type conversions and it is much faster. Only do this if we're
	 * case-sensitive (which is all that really makes sense with byte
	 * arrays anyway, and we have no memcasecmp() for some reason... :^)
	 */

	string1 = (char *) Tcl_GetByteArrayFromObj(objv[0], &length1);
	string2 = (char *) Tcl_GetByteArrayFromObj(objv[1], &length2);
	strCmpFn = (strCmpFn_t) memcmp;
    } else if ((objv[0]->typePtr == &tclStringType)
	    && (objv[1]->typePtr == &tclStringType)) {
	/*
	 * Do a unicode-specific comparison if both of the args are of String
	 * type. In benchmark testing this proved the most efficient check
	 * between the unicode and string comparison operations.
	 */

	string1 = (char *) Tcl_GetUnicodeFromObj(objv[0], &length1);
	string2 = (char *) Tcl_GetUnicodeFromObj(objv[1], &length2);
	strCmpFn = (strCmpFn_t)
		(nocase ? Tcl_UniCharNcasecmp : Tcl_UniCharNcmp);
    } else {
	/*
	 * As a catch-all we will work with UTF-8. We cannot use memcmp() as
	 * that is unsafe with any string containing NUL (\xC0\x80 in Tcl's
	 * utf rep). We can use the more efficient TclpUtfNcmp2 if we are
	 * case-sensitive and no specific length was requested.
	 */

	string1 = (char *) TclGetStringFromObj(objv[0], &length1);
	string2 = (char *) TclGetStringFromObj(objv[1], &length2);
	if ((reqlength < 0) && !nocase) {
	    strCmpFn = (strCmpFn_t) TclpUtfNcmp2;
	} else {
	    length1 = Tcl_NumUtfChars(string1, length1);
	    length2 = Tcl_NumUtfChars(string2, length2);
	    strCmpFn = (strCmpFn_t) (nocase ? Tcl_UtfNcasecmp : Tcl_UtfNcmp);
	}
    }

    length = (length1 < length2) ? length1 : length2;
    if (reqlength > 0 && reqlength < length) {
	length = reqlength;
    } else if (reqlength < 0) {
	/*
	 * The requested length is negative, so we ignore it by setting it to
	 * length + 1 so we correct the match var.
	 */

	reqlength = length + 1;
    }

    match = strCmpFn(string1, string2, (unsigned) length);
    if ((match == 0) && (reqlength > length)) {
	match = length1 - length2;
    }

    Tcl_SetObjResult(interp,
	    Tcl_NewIntObj((match > 0) ? 1 : (match < 0) ? -1 : 0));
    return TCL_OK;
}

/*
 *----------------------------------------------------------------------
 *
 * StringCatCmd --

    case INST_STR_EQ:
    case INST_STR_NEQ:		/* String (in)equality check */
    case INST_STR_CMP:		/* String compare. */
    stringCompare:
	value2Ptr = OBJ_AT_TOS;
	valuePtr = OBJ_UNDER_TOS;

	if (valuePtr == value2Ptr) {
	    match = 0;
	} else {
	    /*
	     * We only need to check (in)equality when we have equal length
	     * strings.  We can use memcmp in all (n)eq cases because we
	     * don't need to worry about lexical LE/BE variance.
	     */

	    typedef int (*memCmpFn_t)(const void*, const void*, size_t);
	    memCmpFn_t memCmpFn;
	    int checkEq = ((*pc == INST_EQ) || (*pc == INST_NEQ)
		    || (*pc == INST_STR_EQ) || (*pc == INST_STR_NEQ));

	    if (TclIsPureByteArray(valuePtr)
		    && TclIsPureByteArray(value2Ptr)) {
		s1 = (char *) Tcl_GetByteArrayFromObj(valuePtr, &s1len);
		s2 = (char *) Tcl_GetByteArrayFromObj(value2Ptr, &s2len);
		memCmpFn = memcmp;
	    } else if ((valuePtr->typePtr == &tclStringType)
		    && (value2Ptr->typePtr == &tclStringType)) {
		/*
		 * Do a unicode-specific comparison if both of the args are of
		 * String type. If the char length == byte length, we can do a
		 * memcmp. In benchmark testing this proved the most efficient
		 * check between the unicode and string comparison operations.
		 */

		s1len = Tcl_GetCharLength(valuePtr);
		s2len = Tcl_GetCharLength(value2Ptr);
		if ((s1len == valuePtr->length)
			&& (valuePtr->bytes != NULL)
			&& (s2len == value2Ptr->length)
			&& (value2Ptr->bytes != NULL)) {
		    s1 = valuePtr->bytes;
		    s2 = value2Ptr->bytes;
		    memCmpFn = memcmp;
		} else {
		    s1 = (char *) Tcl_GetUnicode(valuePtr);
		    s2 = (char *) Tcl_GetUnicode(value2Ptr);
		    if (
#ifdef WORDS_BIGENDIAN
			1
#else
			checkEq
#endif
			) {
			memCmpFn = memcmp;
			s1len *= sizeof(Tcl_UniChar);
			s2len *= sizeof(Tcl_UniChar);
		    } else {
			memCmpFn = (memCmpFn_t) Tcl_UniCharNcmp;
		    }
		}
	    } else {
		/*
		 * strcmp can't do a simple memcmp in order to handle the
		 * special Tcl \xC0\x80 null encoding for utf-8.
		 */

		s1 = TclGetStringFromObj(valuePtr, &s1len);
		s2 = TclGetStringFromObj(value2Ptr, &s2len);
		if (checkEq) {
		    memCmpFn = memcmp;
		} else {
		    memCmpFn = (memCmpFn_t) TclpUtfNcmp2;
		}
	    }

	    if (checkEq && (s1len != s2len)) {
		match = 1;
	    } else {
		/*
		 * The comparison function should compare up to the minimum
		 * byte length only.
		 */
		match = memCmpFn(s1, s2,
			(size_t) ((s1len < s2len) ? s1len : s2len));
		if (match == 0) {
		    match = s1len - s2len;
		}
	    }
	}

	/*
	 * Make sure only -1,0,1 is returned
	 * TODO: consider peephole opt.
	 */

    case INST_GT:
    case INST_LE:
    case INST_GE: {
	int iResult = 0, compare = 0;

	value2Ptr = OBJ_AT_TOS;
	valuePtr = OBJ_UNDER_TOS;









	if (GetNumberFromObj(NULL, valuePtr, &ptr1, &type1) != TCL_OK
		|| GetNumberFromObj(NULL, value2Ptr, &ptr2, &type2) != TCL_OK) {
	    /*
	     * At least one non-numeric argument - compare as strings.
	     */

 * Pointer to a heap-allocated string of length zero that the Tcl core uses as
 * the value of an empty string representation for an object. This value is
 * shared by all new objects allocated by Tcl_NewObj.
 */

MODULE_SCOPE char	tclEmptyString;





/*
 *----------------------------------------------------------------
 * Procedures shared among Tcl modules but not used by the outside world,
 * introduced by/for NRE.
 *----------------------------------------------------------------
 */

MODULE_SCOPE double	TclCeil(const mp_int *a);
MODULE_SCOPE void	TclChannelPreserve(Tcl_Channel chan);
MODULE_SCOPE void	TclChannelRelease(Tcl_Channel chan);
MODULE_SCOPE int	TclCheckArrayTraces(Tcl_Interp *interp, Var *varPtr,
			    Var *arrayPtr, Tcl_Obj *name, int index);
MODULE_SCOPE int	TclCheckBadOctal(Tcl_Interp *interp,
			    const char *value);

MODULE_SCOPE int	TclChanCaughtErrorBypass(Tcl_Interp *interp,
			    Tcl_Channel chan);
MODULE_SCOPE Tcl_ObjCmdProc TclChannelNamesCmd;
MODULE_SCOPE Tcl_NRPostProc TclClearRootEnsemble;
MODULE_SCOPE int	TclCompareTwoNumbers(Tcl_Obj *valuePtr,
			    Tcl_Obj *value2Ptr);
MODULE_SCOPE ContLineLoc *TclContinuationsEnter(Tcl_Obj *objPtr, int num,

			    Tcl_Obj *newValue, Tcl_Encoding encoding);
MODULE_SCOPE void	TclSignalExitThread(Tcl_ThreadId id, int result);
MODULE_SCOPE void	TclSpellFix(Tcl_Interp *interp,
			    Tcl_Obj *const *objv, int objc, int subIdx,
			    Tcl_Obj *bad, Tcl_Obj *fix);
MODULE_SCOPE void *	TclStackRealloc(Tcl_Interp *interp, void *ptr,
			    int numBytes);






MODULE_SCOPE int	TclStringMatch(const char *str, int strLen,
			    const char *pattern, int ptnLen, int flags);
MODULE_SCOPE int	TclStringMatchObj(Tcl_Obj *stringObj,
			    Tcl_Obj *patternObj, int flags);
MODULE_SCOPE void	TclSubstCompile(Tcl_Interp *interp, const char *bytes,
			    int numBytes, int flags, int line,
			    struct CompileEnv *envPtr);

 * The ANSI C "prototype" for this macro is:
 *
 * MODULE_SCOPE int	TclIsPureByteArray(Tcl_Obj *objPtr);
 *----------------------------------------------------------------
 */

MODULE_SCOPE int	TclIsPureByteArray(Tcl_Obj *objPtr);




/*
 *----------------------------------------------------------------
 * Macro used by the Tcl core to compare Unicode strings. On big-endian
 * systems we can use the more efficient memcmp, but this would not be
 * lexically correct on little-endian systems. The ANSI C "prototype" for
 * this macro is:

    if (TclIsPureByteArray(objPtr)) {
	int length;

	(void) Tcl_GetByteArrayFromObj(objPtr, &length);
	return length;
    }


    /*
     * OK, need to work with the object as a string.
     */

    SetStringFromAny(NULL, objPtr);
    stringPtr = GET_STRING(objPtr);
    numChars = stringPtr->numChars;

    /*
     * If numChars is unknown, compute it.
     */

    if (numChars == -1) {
	TclNumUtfChars(numChars, objPtr->bytes, objPtr->length);
	stringPtr->numChars = numChars;
    }
    return numChars;
}













































/*
 *----------------------------------------------------------------------
 *
 * Tcl_GetUniChar --
 *
 *	Get the index'th Unicode character from the String object. If index

    if (interp) {
	Tcl_SetObjResult(interp, Tcl_ObjPrintf(
		    "max size for a Tcl value (%d bytes) exceeded", INT_MAX));
	Tcl_SetErrorCode(interp, "TCL", "MEMORY", NULL);
    }
    return NULL;
}
















































































































































































/*
 *---------------------------------------------------------------------------
 *
 * TclStringFirst --
 *
 *	Implements the [string first] operation.