Cross-product calculator in Java
I am working my way through Norvig's book on AIP. There is an exercise in it on writing a cross-product function -
(defun cross-product (fn list-1 list-2)
(mappend #'(lambda (y)
(mapcar #'(lambda (x)
(funcall fn y x))
list-2))
list-1))
(defun mappend (fn the-list)
(if (null the-list)
nil
(append (funcall fn (first the-list))
(mappend fn (rest the-list)))))
I am trying to write an implementation in Java -
interface Function<T1, T2, T3> {
public T3 function(T1 t1, T2 t2);
}
public class CrossProduct<T1, T2> {
private List<T1> list1;
private List<T2> list2;
public CrossProduct(List<T1> t1, List<T2> t2) {
this.list1 = t1;
this.list2 = t2;
}
public <T3> List<T3> calculate(Function<T1, T2, T3> fn) {
List product = new ArrayList();
for (int i = 0; i < list1.size(); i++)
for (int j = 0; j < list2.size(); j++)
product.add(fn.function(list1.get(i), list2.get(j)));
return product;
}
}
Usage -
@Test
public void testWithStrings() {
List<String> list1 = new ArrayList<String>();
list1.add("6");
list1.add("8");
List<String> list2 = new ArrayList<String>();
list2.add("2");
list2.add("3");
List<String> product = new CrossProduct<String, String>(list1, list2)
开发者_StackOverflow社区 .<String> calculate(new Function<String, String, String>() {
public String function(String x, String y) {
return (String) x + (String) y;
}
});
Assert.assertEquals("62", product.get(0));
Assert.assertEquals("63", product.get(1));
Assert.assertEquals("82", product.get(2));
Assert.assertEquals("83", product.get(3));
}
Is there a better way of doing this?
It seems a little arbitrary to define your CrossProduct
class that way: why are the list args member variables, whereas the fn
is a method parameter? In fact, why is CrossProduct
a class at all? A cross product is a
list, but it's not a subtype of list, since a given list could both
- be expressed as a cross product in many different ways, and
- not have been constructed using the
crossproduct
function.
It's not natural to think of "cross product" as a type, IMO.
I would probably do something like
public class ListFunctions {
public static <T1, T2, T3> List<T3> crossProduct(List<T1> list1, List<T2> list2, Function<T1, T2, T3> fn) {
List<T3> product = new ArrayList<T3>();
for (int i = 0; i < list1.size(); i++)
for (int j = 0; j < list2.size(); j++)
product.add(fn.function(list1.get(i), list2.get(j)));
return product;
}
}
If you did want to define a class CrossProduct
for some reason (e.g. to implement lazy evaluation as salman suggested), I would say it's more OO to have all three args as member variables, and have the class implement List
, e.g.
public class CrossProduct<T1, T2, T3> implements List<T3> {
public CrossProduct(T1 list1, T2 list2, Function<T1, T2, T3> fn) {
// remember args...
}
// etc...
}
I don't know exactly which parameters you would like to improve. However, I would say I don't like N*M list size since it can be too big. If I knew that the result list can be immutable, then I would implement my own List which only calculates product(l1(i), l2(j))
when result.get(i*M+j-1)
is called. So I have not keep a long list (perhaps just a small cache if needed).
精彩评论