Looks fine to me. Either that or use zip_longest. I think a rebase is needed.

I think I prefer the explicit check, mainly because otherwise I think you end up having to check for the raise of AttributeError in order to return False, because you don't want the __eq__ operator raising rather than returning False, unless I've missed a trick (I only realised this when switching to try what you'd said).

Thanks for the rebase prompt, I actually had forgotten that! (though I hope Jenkins & Travis wouldn't have)

Also, and this is probably beyond the scope of what I wanted to get into here, but in principle I think defects should be equal regardless of the order of the bboxes internally, just so long as their sets of pixels are the same. I tried adding this but lists of bboxes are not sortable with sorted() as they don't have an < operator, and I couldn't be bothered to work around that and so didn't put that in, but I think it might be worth considering as it's only a matter of time until that bites someone in exactly the same way subsets of defects just bit me. (And I realise it's quite likely to be me...)

I agree. Given that BoxI is mutable and can't be placed into a set, I think there are two options:

• Retrieve the coordinates of the Box2Is and store them as tuples in a set for comparison.
• Convert the defects to a mask and compare masks.

OK, I've merged this for now, and we can think about this.

I think which of your two approaches is right depends on what is desired given the following:
If one had two boxes totally touching in one defect set, and in the second that was a single box enclosing those two, would we want that to be equal or not? If so, then the second; if not then the first.

Technically they wouldn't be equal, but for most practical purposes I would think they would be. Thoughts?

I don't think I have an opinion on the difference. You may have to ask for input from interested parties.

Have filed DM-19901 for the next bit.