Filtering Collections

[Mark Herschberg]

I need to filter a Java Collection. Before I start reinventing the wheel, is there any standard way of doing this? Better yet, any small open source utility packages out there?
--Mark

[Jim Yingst]

I'm not sure what you mean here. Will this do:

  public void filter(Collection c) {
    for (Iterator it = c.iterator(); it.hasNext(); ) 
      if (!shouldBeRetained(it.next()))
        it.remove();
  }

where shouldBeRemoved(Object) is where you define the criteria for filtering. You can make this configurable by creating some sort if filtering criteria object, similar to FilenameFilter as used by File's list() method.
Depending on the situation you may also wish to leave the original Collection intact, and create a filtered copy:

  public Collection filter(Collection c) {
    Collection filtered = new ArrayList();
    for (Iterator it = c.iterator(); it.hasNext(); ) {
      Object obj = obj
      if (shouldBeRetained(obj))
        filtered.add(obj);
    }
    return filtered;
  }

Is this what you're looking for?
[ February 24, 2003: Message edited by: Jim Yingst ]

[Mark Herschberg]

Yeah, that's what I did. Of course, it's a very focused filter. Right now I'm filtering on an ID tag. Later I might want to filter on a date (either instead of, or additionally). So then I started envisioning interfaces and generic filters, etc. I was just wondering if there's a standard pattern for this. Right now I've got things like
Interfaces:

SecurityFilterablepublic long getSecurityID()

DateFilterable

public long getTimestamp()

(The names are terrible, any better ideas?)
And then a filter class with methods like filterOnSecurityID(Iterator iter, long id) and filterOnDate(Iterator iter, long start, long end).
Of course, I can make this more generic (but probably at an performance cost), by passing in not just an iterator, by some type of comparator. In some sense, I'm basically sorting, and then trunkating the collection.
I was just wondering if there was some standard way of doing this, so I don't reinvent the wheel.
--Mark
[ February 24, 2003: Message edited by: Mark Herschberg ]

[Jim Yingst]

I don't know of standard named patterns for this sort of thing, but that may be just because I don't spend enough time in our Patterns forum.
Of course, I can make this more generic (but probably at an performance cost),
Ah, you worry too much about performance at this point. (Well, probably.)
Usual practice with collections is to pass a Collection (or List/Set/whatever) rather than an Iterator. You can easily get an Iterator from a Collection, but it's more work to go the opposite direction.
For more generic filtering I'd probably go with something like this:

public class CollectionsUtil {
 
  public static interface Acceptor {
    boolean accept(Object obj);
  }
 
  public static void filter(Collection coll, Acceptor acc) {
    for (Iterator it = c.iterator(); it.hasNext(); ) 
      if (!acc.accept(it.next()))
        it.remove();
  }
}

And then to use it:

class SecurityFilter implements CollectionUtil.Acceptor {
  private Set validIDs = new HashSet();
  
  public addValidID(long id) { validIDs.add(new Long(id)); }
  
  public boolean accept(Object obj) {
    SecurityFilterable sf = (SecurityFilterable) obj;
    long id = sf.getSecurityID();
    return validIDs.contains(new Long(id));
  }
}   
class Demo
{
  public void method(Collection c) {
    SecurityFilter sf = new SecurityFilter();
    sf.addID(24601);
    sf.addID(2112);
    sf.addID(6);
    CollectionsUtil.filter(c, sf);
  }
}

Just off the top of my head; may or may not be a good fit for your project.

[Mark Herschberg]

Maybe I am misunderstanding the relationship between iterators and collections. When you pass in a collection, and modify it's iterator (e.g. remove objects from it), you're not modifying the underlying collection object, and so unless you create a new iterator with the non-removed items, you'll lose those changes.
On the other hand, passing in an iterator, and then modifying it, will result in having something filtered when the method returns, right?

--Mark

[Jim Yingst]

When you pass in a collection, and modify it's iterator (e.g. remove objects from it), you're not modifying the underlying collection object,
No, Iterator's remove() does affect the underlying Collection. From the API: "Removes from the underlying collection the last element returned by the iterator" The Iterator is a view of the Collection, not just a copy.
[ February 24, 2003: Message edited by: Jim Yingst ]

[Peter den Haan]

Look into the Jakarta Commons Collections, in particular FilterIterator, Predicate and CollectionUtils.filter(Collection, Predicate). I'm quite disappointed that closures and predicates aren't part of the Java Collections framework, to be honest.
- Peter
[ February 25, 2003: Message edited by: Peter den Haan ]

[Mark Herschberg]

That's what I love about JavaRanch, I'm always learning new things. :-)
Thanks Jim and Peter.
--Mark