The Code is nicely placed, i really appreciate the way u thought but i have a doubt,by using reflection does it become a perfo

Why bother with differently named methods like visitString, visitFloat etc. when you can just use method overloading e.g. visit(Float) and visit(String) and let the JVM select the right one ?

Also, the recursion in :

public void accept(Visitor visitor) {
visitor.visitTreeNode(this);
visitor.visitTreeNode(leftsubtree);
visitor.visitTreeNode(rightsubtree);
}
is wrong. If visitor.visitTreeNode does no navigation of the structure then this should be more like:
public void accept(Visitor visitor) {
visitor.visitTreeNode(this);
leftsubtree.accept(visitor);
rightsubtree.accept(visitor);
}
If it does do navigation it's still wrong, the left and right subtreest will be navigated twice.

I expect it will hit performance pretty hard. Look at the performance numbers these folks get with a similar pattern... http://www.cs.ucla.edu/~palsberg/paper/compsac98.pdf

You can use visit(Float) and visit(String) but anyway you have to cast the parameter on the caller.

I'll inform it to Hul Hul Ra.

...is that with its single inheritance java simply couldn't exist without reflection.

I hate this pattern more than liking it

What's the difference in the two different uses?
Printer printer = ...
Report report = ...
printer.print(report); //1
report.accept(printer); //2

Nice article. You should mention that with the relection approach you give up compile-time type safety... this can be costly in terms of maintenance and testing overheads.

A commons library implementing these ideas has been open sourced at
http://commons-visitor.sourceforge.net