Knute Johnson wrote:
I've been trying to clean up some really old code and I've hit some
snags. I've got several modified JLists and the ListCellRenderers for
them and thought it would make sense to have generic classes that could
be extended for different data types that need to be displayed. The
example below displays InetAddresses in the JList. I've got another
implementation of JList that does a lot more than what this one does but
I wanted to keep this example simple and focus on the ListCellRenderer.
Without knowing what directions you intend/expect to extend the code it's a
little difficult to make sensible suggestions.
So, proceeding with due absence of sense...
I don't see what the clazz variable (and the associated instance checks) are
buying you. You could try removing them and see if that makes your code come
to life. It would certainly remove some of the "kludgy" smell.
If that doesn't help, then it may be that you're being bothered by the way that
ListCellRenderer has (so to speak) too many responsibilities -- it's
simultaneously about /how/ something is displayed and about /what/ is
displayed.
So (this is only my guess, of course) you're being pulled in two directions: on
the one hand your /how/ is pretty much fixed and you want to use nice generic
code for it whatever the types of the values, but the /what/ is not generic and
will change according to each application. You are being required to do
rampant subclassing of something that shouldn't /need/ subclassing.
If that sounds as if it might relate to your problem, you could introduce a new
kind of object who's only responsibility is to take a value, the element of the
List (of type Object -- unless you want to mess with generics), and return a
String which is to be the contents of the cell. It might also have a parallel
method which takes the (same) List element and returns an Icon (or null).
The new object would have appropriate subclasses for InetAddress etc. (which
might or might not appear formally as separate classes in your source -- you
could instead use anonymous classes to make the getTextFor() and getIconFor()
methods "pluggable").
So then you have just one subclass of ListCellRenderer which holds an instance
of [a subclass of] this new class (instead of the clazz variable), and which
always blindly uses that to "fetch" the appropriate String and/or Icon for use
for each cell.
It's difficult to think of a good name for the new kind of object (and it's
subclasses), but in this case I don't think that's the danger sign it normally
would be: the reason that it's difficult to find a name is that the obvious
names (some variant on "Renderer") are already taken. "TextSupplier" perhaps
(or ContentSupplier, or ContentTranslator) ?
It could well be that this kind of picture is over-engineered for your purposes
(I can't tell). But even if it is, then the real culprit is the Swing
architecture which is under-engineered (here), so something may be needed to
compensate.
-- chris