Question Re: Extending AbstractList

Question

Hi,
I'm trying to create a custom collection class (MyList) which functions similarly to a LinkedList and will contain elements of MyClass but where I can override the add(), remove() and get() methods. Rather than having those methods operate on the position of the element in the list I want them to function by examining a specific value of each MyClass element (using a for each loop and MyClass.getId() for example). I would also make add() check for duplicate ID's before adding a new MyClass to the list. Additionally, I'd like maintain the for each looping ability of a list externally so from MyOtherClass I can loop on MyList directly. Anyway, my research led me to conclude that extending AbstractList was the approach I needed to take, although I've listed other alternatives that occurred to me below.
There are a couple of things I'm confused about though and I'm having a hard time find answers to these specific questions.
1. Do I need a instance variable which will hold the items in my list? For example, do I need to have an ArrayList<MyClass> or LinkedList<MyClass> internally in MyList? If so, I'm confused on how to make the for (MyClass myClass : MyList) loop work.
2. If I don't have the above, what's the proper way to implement the size() method?
3. This is my class declaration...it works but it feels wrong...is it actually the correct way to do it?
public class MyList extends AbstractList<MyClass>4. I'm using this class as a singleton using what I read was a thread safe method (code follows). Are there additional considerations I need to make for the overridden methods or the class declaration? I'm using a public static final instance variable like this:
public final static MyList INSTANCE = new MyList();And then accessing it like this:
MyList myList = MyList.INSTANCE;Probably overly complicated but I need to make sure my app is only ever using one list. Anyway, I've tried very hard to find some examples but no joy. In the end I think I have 3 choices:
1. Copy LinkedList and change the methods I need. For some reason that feels like major overkill. I'm willing to do it if it's the best way though.
2. I could extend LinkedList but that presents a couple of problems also. First, it's not a good idea to extend concrete classes (or so I read) and second, I can't then override the add method which is the whole point.
3. The last alternative my limited brain power can come up with, but which makes me lose the for each looping I think, is to just make a non-extending class and manipulate an internal LinkedList. I also lose the ability to have my parameter be a List instead of MyList.
While I could probably force any of the above to work I'm trying to bend my brain, learn something new, and come up with an elegant solution. Any feedback or advice is appreciated.
Thanks,
Pablo

Answer 1

You can do that. Do you intend to modify the set from multiple threads, or create it in one thread and iterate over it in multiple others?Yes, items can be added and removed from multiple threads, as well as updated.
I don't recall you saying that. It sounds like you might actually want a SortedMap.True, I didn't explicitly state that, but I believe I did imply it by saying I wanted to override add(), get() and remove() to use the id.
I have no idea what you mean by that first bit. Do you mean you want a single iterator used by multiple threads, as opposed to multiple threads each having their own iterator?Sorry, I have a hard time explaining myself when trying to learn something new I guess. Yes, multiple threads will need their own iterator I believe. There is a display table showing various properties of each item, then there are currently two different processes which poll the list for specific statuses of each item (like ReadyToSend or Pending) and do independent tasks. These processes are based on a single abstract class since they are both web service clients but with different request/response implementations. My thought was that each of these processes (3 so far) would need their own iterator or possibly their own subsets from the master list based on status. I haven't gotten that far so I don't know. Another process is checking for newly submitted requests which come from an external system via files. It parses the file, stores the object values in the database, and then adds the item to the list. Rather than adding the item directly I was going to pass only the item ID to the add() method and use the factory class I've already created to create a new item from the database to add to the list. This is to pick up valid timestamp values from the record (they're inserted using CURRENT_TIMESTAMP rather than Timestamp objects). The alternative was to store the item and then read it back again. That seemed less elegant. This is the only process doing adds. The ID is actually an externally derived number which is supposed to always be unique but there's nothing to prevent a duplicate. I planned on using the list to control duplicates. If a duplicate is added it's status would be changed in the database to indicate that and it would not be added to the list.
[Clarification] While the list itself will be accessed from multiple locations, I don't expect any individual item to be accessed simultaneously. I have a feeling that's an important distinction. The processes reading and modifying each item will be working on specific their own specific item status.
A lot of this is still roughed out only and needs more refactoring. I started the refactoring by working on the list first. I was passing around an open ResultSet which I know is bad but it allowed me to flesh out the rest more quickly. Good idea or not it's what I did.
Anyway, there would seem to be multiple classes needing access to the list for multiple reasons and yet the list needs singularity to prevent duplicate tasking. I didn't want to get into having to remove the item from one list and add it to a different one when the status changed. I figured the easiest thing to do was have a master list class of some type and have all the various classes use that as a singleton.
Sorry, not a very succinct explanation I guess but we weren't getting far with me beating around the bush with abstract MyXXX examples. Does this change your thoughts on my implementation idea?
Read that chapter of Effective Java. If the chapter you're referring to is Chapter 3, I finally found it and I will read it. There is also a Chapter 5 which deals with Generics but doesn't have any mention of equals() which I'm reading also. And, I ordered the book. ;-)
The bottom line is I still don't understand how I can have a single instance of a TreeMap or SortedMap or MapOfNarnia without having a class to contain it and offer it to calling classes. I realize this is probably a stupid question, but exactly how would I do that without doing something like my last MyClasses example?
Unfortunately I don't have an infinite amount of time to finish this project so I may have to go with a less elegant solution than I hoped. I'm still learning generics and collections and it's obvious I've got quite a bit to learn before I can use them effectively. That being said, which of my bad ideas is the least bad?
By the way, I just want to say I really do appreciate the effort all you experts put into to answering questions for us newbs. I realize it's a significant investment of your time and I appreciate it.
Edited by: Pablo_Vadear on Dec 24, 2009 5:14 AM