Dynamic type conversion?

Question

Hi everyone,
I'm working in a XML method invocation engine but at this moment I'm stucked trying to solve a little problem. I've a XML file looking this way:
<method class="java.lang.Math" name="min">
<arg>
<method class="java.lang.Math" name="max">
<arg type="float">
5
</arg>
<arg type="float">
10
</arg>
</method>
</arg>
<arg type="int">
100
</arg>
</method>
It's a simple example to explain myself, that equals a java expression like "Math.min(Math.max(5.0f,10.0f),100)". As you see min call corresponds to "int Math.min(int,int)" and max to "float Math.max(float,float). When I declare max as "int Math.max(int,int)" the return type equals the paramenter type for min and everything works. But in this example max returns float and min requires int. Using reflection API when I pass an array of arguments to Method.invoke which doesn't correspont to the expected input parameters for min it fails launching an exception because it needs an array like Object[]{int,int} and receives an array looking like this Object[]{float,int}.
My question is that if we know the return type of max using java.lang.reflect.Method.getReturnTypes() and the parameter type of min using java.reflect.Method.getParameterTypes(). How can i perform a dynamic conversion form a primitive type to another if i only know the types at runtime?
And what about objects? How can I convert from, lets say, Integer to Float? Is it posible to do it dynamically using reflection API or do I have to implement long conversion tables and methods?
Thanks for your help
Pep Ribas

Answer 1

You've got two problems here.
The first has to due with the fact that your interpreting of the XML source using Reflection is equivalent to compiling Java source. Your sample represents the following Java source fragment:    java.lang.Math.min(java.lang.Math.max(5.0f, 10.0f), 100);If you provided that line to the Java compiler, you would get a "possible loss of precision" error at the first argument to the min() method, where the compiler is seeing a float when it wants an int. This is a "widening" cast and is not allowed without an explicit cast. Your XML code needs a token for explicit casting. Your sample could look like this:    <method class="java.lang.Math" name="min">
        <arg>
            <cast type="int">
                <method class="java.lang.Math" name="max">
                    <arg type="float">
                        5
                    </arg>
                    <arg type="float">
                        10
                    </arg>
                </method>
            </cast>
        </arg>
        <arg type="int">
            100
        </arg>
    </method>This would represent a Java source fragment that correctly compiles:    java.lang.Math.min((int)java.lang.Math.max(5.0f, 10.0f), 100);The second problem is due to the ambiguous definition of the Reflection package. Even though Reflection provides ways to invoke methods, it was designed to browse the capabilities of a class and not to act as an interpreter. There is really no way to dynamically perform an explicit-cast without implementing the patterns of the Java implicit casting mechanisms and method look-up.
You might do better to compile your XML code into Java source text, and then programmatically pass that text to the Java compiler class sun.tools.javac.Main. The compiler will then perform the needed implicit casts to best-fit the available methods. If the compiler returns errors, you can map them back to the appropriate XML tokens and report it to the user.
I hope you find this of some help.