Comments on: The Iterator Pattern

By: giulio

giulio — Tue, 08 Aug 2006 17:49:29 +0000

Why don’t to add an implementation template using C# enerics? Anyway… very good!

Bye,

Giulio

By: Jose Luis

Jose Luis — Sat, 20 Aug 2005 15:55:57 +0000

If I add the "friendly" class to be able to declare the iterator class into another unit separated from TCustomerList class, I would need to add into the TCustomerIterator unit the uses of the TCustomerList.

Then, when I would like to add the GetIterator into the TCustomerList to create the iterator from withing the list ifself, am I going to have a circular reference?

By: Shawn Stamps

Shawn Stamps — Sun, 10 Jul 2005 19:05:11 +0000

Yes, it is an infinite loop. The correct code would be:

Repeat

Result := inherited Next;

Until Not Result Or Result And CurrentItem.CreditStop;

I’ve coded that kind of error so many times that I usually catch it right after I type it in.

By: Stephen Melnyk

Stephen Melnyk — Tue, 23 Nov 2004 21:30:54 +0000

Thanks for the article, Joanna. I appreciate you taking the time to write it.

Could you clarify one point for me? In the Skip Iterators section, you implement TBadCustomerIterator.Next with:

repeat
Result := inherited Next;
until Result and CurrentItem.CreditStop

Isn't this an infinite loop? The last item in the list is going to return False from the inherited Next.

Stephen

By: VitoJeng

VitoJeng — Mon, 18 Oct 2004 09:25:03 +0000

Good article, Joanna!

By: Marcos Barreto

Marcos Barreto — Sun, 19 Sep 2004 14:13:08 +0000

Could you post a sample deletion with iterators?

Thanks!

By: Joanna Carter

Joanna Carter — Wed, 28 Jul 2004 10:40:36 +0000

> A re-evaluation every time the list changes could
> be very costly, especially when the structure is
> not linear (a tree or something).

The Observer Pattern does not re-evaluate the list it merely notifies the Iterator when a Deletion has occurred.

> The GoF book says that it is too costly to
> duplicate the list of items, but I disagree, there
> would have to be a lot of items in that list to be
> of any severe consequence when you are simply
> holding pointers to objects.

Duplicating the list is not the same as creating a list of references to items in an existing list. It involves creating a list of objects that are copies of the objects in the original list.

> You could however write a ReverseOrderIterator.

That's the beauty of Iterators:-)

By: Wilbert

Wilbert — Wed, 28 Jul 2004 10:12:02 +0000

> You would never normally use an Iterator for deletions

I disagree with this point. I think that an iterator can be used to perform any operation on a list.

> My standard Iterator is attached as an Observer to the list

A re-evaluation every time the list changes could be very costly, especially when the structure is not linear (a tree or something).

The GoF book says that it is too costly to duplicate the list of items, but I disagree, there would have to be a lot of items in that list to be of any severe consequence when you are simply holding pointers to objects.

You could however write a ReverseOrderIterator.

By: Joanna Carter

Joanna Carter — Tue, 27 Jul 2004 23:33:20 +0000

You would never normally use an Iterator for deletions unless you first created a List that contains the items to be deleted and then use the Iterator from that list to get the items to delete from the main list.

My standard Iterator is attached as an Observer to the list and when the contents of the list changes the Iterator is automatically invalidated by setting it to the end of the list.

By: Wilbert

Wilbert — Tue, 27 Jul 2004 22:40:36 +0000

Just a small point. I thought that the iterator was supposed to hold its own list of items rather than depending on the original list?

The purpose of this would be

while it.Next do
CustomerList.Delete(it.CurrentItem);

Which in your example would cause an index out of bounds.