Replace TclCountSpaceRuns() with TclMaxListLength() which is the function we

actually want.

4 files changed, 45 insertions, 24 deletions

diff --git a/ChangeLog b/ChangeLog
index 856780e..f7e8a0d 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,9 @@
+2011-05-02  Don Porter  <dgp@users.sourceforge.net>
+
+	* generic/tclInt.h:	Replace TclCountSpaceRuns() with
+	* generic/tclListObj.c:	TclMaxListLength() which is the function we
+	* generic/tclUtil.c:	actually want.
+
 2011-04-28  Don Porter  <dgp@users.sourceforge.net>
 
 	* generic/tclInt.h:	New utility routines:
diff --git a/generic/tclInt.h b/generic/tclInt.h
index 7d9f066..e3aaccb 100644
--- a/generic/tclInt.h
+++ b/generic/tclInt.h
@@ -2900,8 +2900,6 @@ MODULE_SCOPE void	TclContinuationsEnterDerived(Tcl_Obj *objPtr,
 MODULE_SCOPE ContLineLoc *TclContinuationsGet(Tcl_Obj *objPtr);
 MODULE_SCOPE void	TclContinuationsCopy(Tcl_Obj *objPtr,
 			    Tcl_Obj *originObjPtr);
-MODULE_SCOPE int	TclCountSpaceRuns(CONST char *bytes, int numBytes,
-			    CONST char **endPtr);
 MODULE_SCOPE void	TclDeleteNamespaceVars(Namespace *nsPtr);
 /* TIP #280 - Modified token based evulation, with line information. */
 MODULE_SCOPE int	TclEvalEx(Tcl_Interp *interp, const char *script,
@@ -3011,6 +3009,8 @@ MODULE_SCOPE Tcl_Obj *	TclLsetFlat(Tcl_Interp *interp, Tcl_Obj *listPtr,
 			    Tcl_Obj *valuePtr);
 MODULE_SCOPE Tcl_Command TclMakeEnsemble(Tcl_Interp *interp, const char *name,
 			    const EnsembleImplMap map[]);
+MODULE_SCOPE int	TclMaxListLength(CONST char *bytes, int numBytes,
+			    CONST char **endPtr);
 MODULE_SCOPE int	TclMergeReturnOptions(Tcl_Interp *interp, int objc,
 			    Tcl_Obj *const objv[], Tcl_Obj **optionsPtrPtr,
 			    int *codePtr, int *levelPtr);
diff --git a/generic/tclListObj.c b/generic/tclListObj.c
index 9437f7e..03843b8 100644
--- a/generic/tclListObj.c
+++ b/generic/tclListObj.c
@@ -1783,27 +1783,21 @@ SetListFromAny(
 
     /*
      * Parse the string into separate string objects, and create a List
-     * structure that points to the element string objects. We use a modified
-     * version of Tcl_SplitList's implementation to avoid one malloc and a
-     * string copy for each list element. First, estimate the number of
-     * elements by counting the number of space characters in the list.
-     */
-
-    estCount = TclCountSpaceRuns(string, length, &limit) + 1;
-
-    /*
-     * Allocate a new List structure with enough room for "estCount" elements.
-     * Each element is a pointer to a Tcl_Obj with the appropriate string rep.
-     * The initial "estCount" elements are set using the corresponding "argv"
-     * strings.
+     * structure that points to the element string objects. 
+     *
+     * First, allocate enough space to hold a (Tcl_Obj *) for each
+     * (possible) list element.
      */
 
+    estCount = TclMaxListLength(string, length, &limit);
+    estCount += (estCount == 0); /* Smallest List struct holds 1 element. */
     listRepPtr = AttemptNewList(interp, estCount, NULL);
     if (listRepPtr == NULL) {
 	return TCL_ERROR;
     }
     elemPtrs = &listRepPtr->elements;
 
+    /* Each iteration, parse and store a list element */
     for (p=string, lenRemain=length, i=0;
 	    lenRemain > 0;
 	    p=nextElem, lenRemain=limit-nextElem, i++) {
diff --git a/generic/tclUtil.c b/generic/tclUtil.c
index fdfb190..d4630af 100644
--- a/generic/tclUtil.c
+++ b/generic/tclUtil.c
@@ -88,15 +88,19 @@ const Tcl_ObjType tclEndOffsetType = {
 /*
  *----------------------------------------------------------------------
  *
- * TclCountSpaceRuns --
+ * TclMaxListLength --
  *
  *	Given 'bytes' pointing to 'numBytes' bytes, scan through them and
  *	count the number of whitespace runs that could be list element
  *	separators.  If 'numBytes' is -1, scan to the terminating '\0'.
+ *	Not a full list parser.  Typically used to get a quick and dirty
+ *	overestimate of length size in order to allocate space for an
+ *	actual list parser to operate with.
  *
  * Results:
- *	Returns the count.  If 'endPtr' is not NULL, writes a pointer to
- *	the end of the string scanned there.
+ *	Returns the largest number of list elements that could possibly
+ *	be in this string, interpreted as a Tcl list.  If 'endPtr' is not
+ *	NULL, writes a pointer to the end of the string scanned there.
  *
  * Side effects:
  *	None.
@@ -105,13 +109,22 @@ const Tcl_ObjType tclEndOffsetType = {
  */
 
 int
-TclCountSpaceRuns(
+TclMaxListLength(
     CONST char *bytes,
     int numBytes,
     CONST char **endPtr)
 {
     int count = 0;
 
+    if ((numBytes == 0) || ((numBytes == -1) && (*bytes == '\0'))) {
+	/* Empty string case - quick exit */
+	goto done;
+    }
+
+    /* No list element before leading white space */
+    count += 1 - TclIsSpaceProc(*bytes); 
+
+    /* Count white space runs as potential element separators */
     while (numBytes) {
 	if ((numBytes == -1) && (*bytes == '\0')) {
 	    break;
@@ -131,6 +144,11 @@ TclCountSpaceRuns(
 	bytes++;
 	numBytes -= (numBytes != -1);
     }
+
+    /* No list element following trailing white space */
+    count -= TclIsSpaceProc(bytes[-1]); 
+
+    done:
     if (endPtr) {
 	*endPtr = bytes;
     }
@@ -469,15 +487,18 @@ Tcl_SplitList(
     int length, size, i, result, elSize, brace;
 
     /*
-     * Figure out how much space to allocate. There must be enough space for
-     * both the array of pointers and also for a copy of the list. To estimate
-     * the number of pointers needed, count the number of space characters in
-     * the list.
+     * Allocate enough space to work in. A (CONST char *) for each
+     * (possible) list element plus one more for terminating NULL,
+     * plus as many bytes as in the original string value, plus one
+     * more for a terminating '\0'.  Space used to hold element separating
+     * white space in the original string gets re-purposed to hold '\0'
+     * characters in the argv array.
      */
 
-    size = TclCountSpaceRuns(list, -1, &end) + 2;
+    size = TclMaxListLength(list, -1, &end) + 1;
     length = end - list;
     argv = ckalloc((size * sizeof(char *)) + length + 1);
+
     for (i = 0, p = ((char *) argv) + size*sizeof(char *);
 	    *list != 0;  i++) {
 	const char *prevList = list;