Re: [Qexo-general] Why isn't this working? Is it a bug?

[Per Bothner]

Both Craig's and Froncois's bug are "fixed" with the attached patch. 
However, the patch is not correct: First all the 'new TreeList' calls in 
getPosNext has not be fixed in the same way.  That's easy.  Harder is:
(1) It is not ok for gnu.lists to reference classes in gnu.xml - I don't 
want gnu.lists to depend on gnu.xml.
(2) Creating a NodeTree is correct when the TreeList is viewed as a 
sequence of document fragments, which is presumably going to be teh case 
for XQuery, but may not be the correct interpretation for other uses of 
Values and TreeList.

So I need to figure ut a clean(er) solution to this problem.

(Note the basic problem is how XQuery optimizes element constructors to 
SAX-like "event streams" using a Consumer in a function, and then result 
of the function needs to be gathered to an Object.)
--
        --Per Bothner
address@hidden   http://per.bothner.com/

Index: TreeList.java
===================================================================
RCS file: /cvs/kawa/kawa/gnu/lists/TreeList.java,v
retrieving revision 1.31
diff -u -r1.31 TreeList.java
--- TreeList.java       10 Jul 2003 18:44:07 -0000      1.31
+++ TreeList.java       23 Jul 2003 06:15:47 -0000
@@ -1333,7 +1333,12 @@
       return objects[datum-OBJECT_REF_SHORT];
     if (datum >= BEGIN_GROUP_SHORT
            && datum <= BEGIN_GROUP_SHORT+BEGIN_GROUP_SHORT_INDEX_MAX)
-      return new TreeList(this, index, index + data[index+1] + 2);
+      {
+       gnu.xml.NodeTree nodes = new gnu.xml.NodeTree();
+       consumeIRange(index, index + data[index+1] + 2, nodes);
+       return nodes;
+       //return new TreeList(this, index, index + data[index+1] + 2);
+      }
     /*
     if ((datum & 0xFF00) == BYTE_PREFIX)
       return Sequence.TEXT_BYTE_VALUE;