Bug + fix found in reference implementation

[MarkusFaatz]

Hey there!
I recently searched for the EB-AFIT algorithm and I found this bug in the original C-Code project that seems to have also found his way to this C# project.

As Robertson also suggests a fix you might have a look at these two lines in your code

3DContainerPacking/src/CromulentBisgetti.ContainerPacking/Algorithms/EB_AFIT.cs

Lines 940 to 941 in 7f24c7b

	smallestZ.CumX = smallestZ.CumX - cboxx;
	itemsToPack[cboxi].CoordX = smallestZ.CumX - cboxx;

where it might be a mistake to decrement cboxx twice from the itemsToPack[cboxi].CoordX

[davidmchapman]

Very interesting! I'm open to updating the code. I've taken time away from this project but now have some more time to hop back into it. Soon I will run the tests on the before and after code to see how this affects them. I'll report back.

Thank you for bringing this up!

[davidmchapman]

I was able to verify both the issue and the proposed fix.

Following the original comment from @garrettsickles, I ran the packing algorithm on set 4 of the randomly generated problem sets from Appendix D of the original master's thesis. This test attempts to pack 1000+ items into a single 104x96x84 container.

Packing with the unfixed code yielded the same issue @garrettsickles noted, shown here:

Notably, you can also see where the accidental shifting started in this layer. That's what pushed the final item in the layer outside the bounds of the container:

With the proposed fix, the algorithm correctly closes that gap:

Great work, @garrettsickles, @robertson-so, and @MarkusFaatz!!

I will roll the fix into master in this repo soon. Thanks, everyone!

[davidmchapman]

Fix committed.