changelog

移动对象空间数据库中索引r树源码,能在eclipse下直接运行.

2003-03-05  root  <[e-mail]>

	* RTree.java (RTree): Major and sweeping changes! The buffer managment policy has changed.
	Now I write a node to the disk only when asked to flush or when a node is forced out of the 
	cache. Not yet fully tested, if you want to revert get the code of about 3 days back. Next
	note would be on the fully tested code.

2003-02-28  root  <[e-mail]>

	* RTree.java (RTree): Changed all the static methods to get root indix to instanciated methods.
	Root index is not read with i/o but instead read from the internal variable of fileHdr attibute.
	(adjustTree): Added another parameter - long parentIndex. This is the parent index of the root
	node. In normal cases this is Node.NOT_DEFINED, but in case of seeded trees this is the index
	of the slot node under which the tree is being grown.

2003-02-26  root  <[e-mail]>

	* Node.java (insertElement Element[]): This overloaded method is newly added to the class. The 
	advantage of this method over the other method is that this method helps when we have bulk loading.
	(writeNodeHeader): Added this method that writes to the passed DataOutputStream and not the 
	hard dick file directly.

2003-02-19  root  <[e-mail]>

	* Node.java (splitNode): Changed from linear pick seed to quadratic pick seed.

2003-02-14  root  <[e-mail]>

	* Node.java (nodeIndex): Changed it to long from int
	(parent): Changed to long as well
	These are the latest changes on 14/2/1003.                                                      
	The most major change is that all the pointers have been converted to <code>long</code>         
	from <code>int</code>. This has resulted in changing od file header, node header, and element   
	sizes.	                                                                                   
	This is what I wrote everywhere before converting all the pointers(leaf and nonleaf) to         
	long from Object:   <pre><code>
	<i>[1]</i>
	if(elmtType == LEAF_NODE){                                                             
	Integer intVal = (Integer)elmts[i].getPtr();
	this.elements[i] = new LeafElement(new Rect(elmts[i].getRect()), intVal.intValue());
	}
	else{                                                  
	Integer intVal = (Integer)elmts[i].getPtr();
	this.elements[i] = new NonLeafElement(new Rect(elmts[i].getRect()),intVal.intValue());
	}
	<i>[2]</i>
	if(elementType == LEAF_NODE){
	Integer intVal = (Integer)elements[i].getPtr();
	ds.writeInt(intVal.intValue());
	}	
	else{
	Integer nodePtr = (Integer)elements[i].getPtr();
	ds.writeInt(nodePtr.intValue());
	}
	<code></pre>
	The method getPtr or setPtr no longer returns or set <code>Object</code> but they now deal in   
	<code>long</code> values. Therefore now I simply call <code>getPtr</code> and gets me 
	<code>long</code>
	<br>
	There may be places where the non leaf index is directly typecasted into int from long 
	(<code>CachedNodes</code> and skipBytes (here)). This is no problem as we are never going to have 
	so many nodes that the Integer.MAX_VALUE is not enough.
	

2003-02-13  root  <[e-mail]>

	* LeafElement.java (recordPtr): changing record pointer to long