Thanks for this! I quickly scanned the code, and it made sense. I want to take a closer look, hopefully soon, to see if stuff can be simplified a bit. In the meantime, could you add a before and after of, say, a 20x20 grid next to what you already provided (which looks great, btw)?

Thank you for your response! I’ve added grid images with a 10x10 layout (as the 20x20 grid didn’t look as clean in the screenshot) while keeping the same line-dot pattern for better clarity.

There is still one issue unresolved regarding the alignment of hexes. Make the change shown below:

loc[:, 0] = loc[:, 0] * x_spacing + ((loc[:, 1]-1) % 2) * (x_spacing / 2)

@quaquel I don't know if you were able to see the review I started, in which I clearly attached images of your suggestion not working, so just to make sure I am attaching one image again, your suggestion offsets the placement of agents. I don't know if this is the intended behavior, if it is I shall change the code.

I had missed that reply. Sorry. I'll try to take a closer look myself later this week.

Finally had time to look at this again. My suggestion only changed the agent location, but the same correction evidently also needs to be made to the hex centers: so x = col * x_spacing + (row % 2 == 0) * (x_spacing / 2).

However, while figuring this out, I discovered 2 other bugs (which are also present in the current code).

1. If there are no agents, the modifications to loc given an error because loc = arguments["loc"].astype(float) is empty.
2. See the figure below. I created a simple grid with 1 central agent in blue and its neighbors in orange.

This is most definitely not right and seems to be related to mixing up row and column indexing. I'll try to figure this out asap.

I have added the reviewed changes, including the new empty loc check. And correct me if I'm wrong, the issue with neighbors existed in the previous code and was not introduced by this PR, right?

the issue with neighbors existed in the previous code and was not introduced by this PR, right?

Yes this issue predates your PR. See #2632

have added the reviewed changes, including the new empty loc check. And correct me if I'm wrong, the issue with neighbors existed in the previous code and was not introduced by this PR, right?

I suggest we first merge #2632 so I can fully visually check this PR.

I have merged the bugfix and updated this PR by rebasing.

@quaquel If there are no issues with the current approach, I would really appreciate it if this PR could be merged.

I hope to have time for some final checks later this afternoon CET time.

Ok, I checked stuff visually, and it looks all ok now. Merging this.

Thanks again!