Tcl Source Code

static int
StringRplcCmd(
    ClientData dummy,		/* Not used. */
    Tcl_Interp *interp,		/* Current interpreter. */
    int objc,			/* Number of arguments. */
    Tcl_Obj *const objv[])	/* Argument objects. */
{
    Tcl_UniChar *ustring;
    int first, last, length, end;

    if (objc < 4 || objc > 5) {
	Tcl_WrongNumArgs(interp, 1, objv, "string first last ?string?");
	return TCL_ERROR;
    }

    ustring = Tcl_GetUnicodeFromObj(objv[1], &length);
    end = length - 1;

    if (TclGetIntForIndexM(interp, objv[2], end, &first) != TCL_OK ||
	    TclGetIntForIndexM(interp, objv[3], end, &last) != TCL_OK){
	return TCL_ERROR;
    }

    /*


     * The following test screens out most empty substrings as
     * candidates for replacement. When they are detected, no
     * replacement is done, and the result is the original string,
     */

    if ((last < 0) ||		/* Range ends before start of string */
	    (first > end) ||	/* Range begins after end of string */
	    (last < first)) {	/* Range begins after it starts */

	/*
	 * BUT!!! when (end < 0) -- an empty original string -- we can
	 * have (first <= end < 0 <= last) and an empty string is permitted
	 * to be replaced.
	 */
	Tcl_SetObjResult(interp, objv[1]);
    } else {
	Tcl_Obj *resultPtr;

	/*
	 * We are re-fetching in case the string argument is same value as 
	 * an index argument, and shimmering cost us our ustring.
	 */

	ustring = Tcl_GetUnicodeFromObj(objv[1], &length);

	end = length-1;

	if (first < 0) {
	    first = 0;
	}

	resultPtr = Tcl_NewUnicodeObj(ustring, first);
	if (objc == 5) {

    Tcl_Interp *interp,		/* Tcl interpreter for context. */
    Tcl_Parse *parsePtr,	/* Points to a parse structure for the
				 * command. */
    Command *cmdPtr,		/* Points to defintion of command being
				 * compiled. */
    CompileEnv *envPtr)		/* Holds the resulting instructions. */
{
    Tcl_Token *tokenPtr, *valueTokenPtr;
    DefineLineInformation;	/* TIP #280 */
    int first, last;

    if (parsePtr->numWords < 4 || parsePtr->numWords > 5) {
	return TCL_ERROR;
    }
 
    /* Bytecode to compute/push string argument being replaced */
    valueTokenPtr = TokenAfter(parsePtr->tokenPtr);

    CompileWord(envPtr, valueTokenPtr, interp, 1);



    /*
     * Check for first index known and useful at compile time. 
     */
    tokenPtr = TokenAfter(valueTokenPtr);
    if (TclGetIndexFromToken(tokenPtr, TCL_INDEX_BEFORE, TCL_INDEX_AFTER,
	    &first) != TCL_OK) {
	goto genericReplace;
    }

    /*

     * Check for last index known and useful at compile time. 
     */

    tokenPtr = TokenAfter(tokenPtr);
    if (TclGetIndexFromToken(tokenPtr, TCL_INDEX_BEFORE, TCL_INDEX_AFTER,
	    &last) != TCL_OK) {
	goto genericReplace;
    }

    /* 
     * [string replace] is an odd bird.  For many arguments it is
     * a conventional substring replacer.  However it also goes out
     * of its way to become a no-op for many cases where it would be
     * replacing an empty substring.  Precisely, it is a no-op when
     *
     *		(last < first)		OR
     *		(last < 0)		OR
     *		(end < first)
     *
     * For some compile-time values we can detect these cases, and
     * compile direct to bytecode implementing the no-op.
     */

    if ((last == TCL_INDEX_BEFORE)		/* Know (last < 0) */
	    || (first == TCL_INDEX_AFTER)	/* Know (first > end) */

	/*
	 * Tricky to determine when runtime (last < first) can be
	 * certainly known based on the encoded values. Consider the
	 * cases...
	 *
	 * (first <= TCL_INDEX_END) &&
	 *	(last == TCL_INDEX_AFTER) => cannot tell REJECT
	 *	(last <= TCL_INDEX END) && (last < first) => ACCEPT
	 *	else => cannot tell REJECT
	 */
	    || ((first <= TCL_INDEX_END) && (last <= TCL_INDEX_END)
		&& (last < first))		/* Know (last < first) */
	/*
	 * (first == TCL_INDEX_BEFORE) &&
	 *	(last == TCL_INDEX_AFTER) => (first < last) REJECT
	 *	(last <= TCL_INDEX_END) => cannot tell REJECT
	 *	else		=> (first < last) REJECT
	 *







	 * else [[first >= TCL_INDEX_START]] &&
	 *	(last == TCL_INDEX_AFTER) => cannot tell REJECT
	 *	(last <= TCL_INDEX_END) => cannot tell REJECT
	 *	else [[last >= TCL_INDEX START]] && (last < first) => ACCEPT
	 */
	    || ((first >= TCL_INDEX_START) && (last >= TCL_INDEX_START)


		&& (last < first))) {		/* Know (last < first) */
	if (parsePtr->numWords == 5) {
	    tokenPtr = TokenAfter(tokenPtr);
	    CompileWord(envPtr, tokenPtr, interp, 4);
	    OP(		POP);		/* Pop newString */


	}

	/* Original string argument now on TOS as result */


	return TCL_OK;
    }



    if (parsePtr->numWords == 5) {
    /*

     * When we have a string replacement, we have to take care about
     * not replacing empty substrings that [string replace] promises
     * not to replace

     *
     * The remaining index values might be suitable for conventional
     * string replacement, but only if they cannot possibly meet the
     * conditions described above at runtime. If there's a chance they
     * might, we would have to emit bytecode to check and at that point
     * we're paying more in bytecode execution time than would make
     * things worthwhile. Trouble is we are very limited in
     * how much we can detect that at compile time. After decoding,
     * we need, first:
     *





     *		(first <= end)



     *
     * The encoded indices (first <= TCL_INDEX END) and
     * (first == TCL_INDEX_BEFORE) always meets this condition, but

     * any other encoded first index has some list for which it fails.
     *
     * We also need, second:

     *
     *		(last >= 0)
     *
     * The encoded indices (last >= TCL_INDEX_START) and
     * (last == TCL_INDEX_AFTER) always meet this condition but any
     * other encoded last index has some list for which it fails.
     *
     * Finally we need, third:
     *

     *		(first <= last)

     * 
     * Considered in combination with the constraints we already have,
     * we see that we can proceed when (first == TCL_INDEX_BEFORE)
     * or (last == TCL_INDEX_AFTER). These also permit simplification
     * of the prefix|replace|suffix construction. The other constraints,
     * though, interfere with getting a guarantee that first <= last. 
     */

    if ((first == TCL_INDEX_BEFORE) && (last >= TCL_INDEX_START)) {
	/* empty prefix */
	tokenPtr = TokenAfter(tokenPtr);
	CompileWord(envPtr, tokenPtr, interp, 4);
	OP4(		REVERSE, 2);
	if (last == TCL_INDEX_AFTER) {
	    OP(		POP);		/* Pop  original */
	} else {
	    OP44(	STR_RANGE_IMM, last + 1, TCL_INDEX_END);
	    OP1(	STR_CONCAT1, 2);
	}
	return TCL_OK;
    }




    if ((last == TCL_INDEX_AFTER) && (first <= TCL_INDEX_END)) {









	OP44(		STR_RANGE_IMM, 0, first-1);
	tokenPtr = TokenAfter(tokenPtr);
	CompileWord(envPtr, tokenPtr, interp, 4);
	OP1(		STR_CONCAT1, 2);

	return TCL_OK;
    }

	/* FLOW THROUGH TO genericReplace */

    } else {
	/* 


	 * When we have no replacement string to worry about, we may

	 * have more luck, because the forbidden empty string replacements
	 * are harmless when they are replaced by another empty string.
	 */

	if ((first == TCL_INDEX_BEFORE) || (first == TCL_INDEX_START)) {
	    /* empty prefix - build suffix only */

	    if ((last == TCL_INDEX_END) || (last == TCL_INDEX_AFTER)) {
		/* empty suffix too => empty result */
		OP(	POP);		/* Pop  original */
		PUSH	(	"");
		return TCL_OK;
	    }
	    OP44(	STR_RANGE_IMM, last + 1, TCL_INDEX_END);
	    return TCL_OK;
	} else {
	    if ((last == TCL_INDEX_END) || (last == TCL_INDEX_AFTER)) {
		/* empty suffix - build prefix only */
		OP44(	STR_RANGE_IMM, 0, first-1);
		return TCL_OK;
	    }
	    OP(		DUP);
	    OP44(	STR_RANGE_IMM, 0, first-1);
	    OP4(	REVERSE, 2);
	    OP44(	STR_RANGE_IMM, last + 1, TCL_INDEX_END);
	    OP1(	STR_CONCAT1, 2);
	    return TCL_OK;
	}
    }

    genericReplace:

	tokenPtr = TokenAfter(valueTokenPtr);
	CompileWord(envPtr, tokenPtr, interp, 2);
	tokenPtr = TokenAfter(tokenPtr);
	CompileWord(envPtr, tokenPtr, interp, 3);
	if (parsePtr->numWords == 5) {
	    tokenPtr = TokenAfter(tokenPtr);
	    CompileWord(envPtr, tokenPtr, interp, 4);
	} else {
	    PUSH(	"");
	}
	OP(		STR_REPLACE);
	return TCL_OK;

}

int
TclCompileStringTrimLCmd(
    Tcl_Interp *interp,		/* Used for error reporting. */
    Tcl_Parse *parsePtr,	/* Points to a parse structure for the command
				 * created by Tcl_ParseCommand. */

	    TclNewObj(objResultPtr);
	}
	TRACE_APPEND(("%.30s\n", O2S(objResultPtr)));
	NEXT_INST_F(9, 1, 1);

    {
	Tcl_UniChar *ustring1, *ustring2, *ustring3, *end, *p;
	int length3, endIdx;
	Tcl_Obj *value3Ptr;

    case INST_STR_REPLACE:
	value3Ptr = POP_OBJECT();
	valuePtr = OBJ_AT_DEPTH(2);
	endIdx = Tcl_GetCharLength(valuePtr) - 1;
	TRACE(("\"%.20s\" %s %s \"%.20s\" => ", O2S(valuePtr),
		O2S(OBJ_UNDER_TOS), O2S(OBJ_AT_TOS), O2S(value3Ptr)));
	if (TclGetIntForIndexM(interp, OBJ_UNDER_TOS, endIdx,
		    &fromIdx) != TCL_OK
	    || TclGetIntForIndexM(interp, OBJ_AT_TOS, endIdx,
		    &toIdx) != TCL_OK) {
	    TclDecrRefCount(value3Ptr);
	    TRACE_ERROR(interp);
	    goto gotError;
	}
	TclDecrRefCount(OBJ_AT_TOS);
	(void) POP_OBJECT();
	TclDecrRefCount(OBJ_AT_TOS);
	(void) POP_OBJECT();



	if ((toIdx < 0) ||
		(fromIdx > endIdx) ||
		(toIdx < fromIdx)) {
	    TRACE_APPEND(("\"%.30s\"\n", O2S(valuePtr)));
	    TclDecrRefCount(value3Ptr);
	    NEXT_INST_F(1, 0, 0);
	}

	if (fromIdx < 0) {
	    fromIdx = 0;
	}

	if (toIdx > endIdx) {
	    toIdx = endIdx;
	}

	if (fromIdx == 0 && toIdx == endIdx) {
	    TclDecrRefCount(OBJ_AT_TOS);
	    OBJ_AT_TOS = value3Ptr;
	    TRACE_APPEND(("\"%.30s\"\n", O2S(value3Ptr)));
	    NEXT_INST_F(1, 0, 0);
	}

	length3 = Tcl_GetCharLength(value3Ptr);

} {foo}
test string-14.17 {string replace} {
    string replace abcdefghijklmnop end end-1
} {abcdefghijklmnop}
test string-14.18 {string replace} {
    string replace abcdefghijklmnop 10 9 XXX
} {abcdefghijklmnop}
test string-14.19 {string replace} {
    string replace {} -1 0 A
} A

test string-15.1 {string tolower too few args} {
    list [catch {string tolower} msg] $msg
} {1 {wrong # args: should be "string tolower string ?first? ?last?"}}
test string-15.2 {string tolower bad args} {
    list [catch {string tolower a b} msg] $msg
} {1 {bad index "b": must be integer?[+-]integer? or end?[+-]integer?}}