Autotiling does not work in combination with half offsets.

[michalove]

Godot version:
3.0

Issue description:
Autotiling determines the tile to be drawn by looking at the neighbor cells. These are defined by having an x- or y-index that differs by at most 1. With an activated half-offset, this neighborhood-logic breaks.

What is the upper-right neighbor in a regular grid, can become two different things in a half-offset grid: Either a cell that does not touch the original cell anymore, or a cell that shares half an edge with the cell.
This are diagonal neighbors on the regular grid:

And here are two pairs of diagonal neighbors on the half offset grid

That means the concept of a neighbor depends on what cell you are looking at.
Therefore it does not make sense to look at the neighboring cells to determine the tile to draw.

I have a suggestion for fixing that. Maybe you find it helpful:
I suggest the following change in the half-offset functionality: Instead of making the cells go zig-zag, when turning on half offset, make them go skewed (Each row is shifted by one half cell more than the previous one, instead of going back and forth). With that arrangement, autotiling would work for half offset grid, without any other changes.

[SpechtMagnus]

I like your suggestion but I think it would mess with the quadrants used for tilemap optimisation. Another possibility would be to use the central bit of the autotile bitmask to determine whether the tile is on a odd or even row. What do you think about that?
For obvious reasons this is only a solution using the wildcard functionallity I just introduced. #27565

[SpechtMagnus]

@MarianoGnu What would you suggest to solve this problem. I could implement it but I would really like to hear a second opinion on this.

[michalove]

Here is an opinion from me, a random guy from the internet:
Yes, the suggestion with the central bit would work. It would be a solution that requires only a minimum of changes in the code.
From a user perspective, however, this solution does not seem optimal to me. The central-bit solution would again lead to redundancies in the bitmasks (for each tile-configuration, the user has to input two bitmasks, one for even and one for odd rows/columns). Furthermore, this solution is quiet unintuitive (why should autotiling lead to different results depending on the absolute position of the tile? For the user and for the tiling-logic, the odd and even columns are irrelevant), so it would required proper documentation.

A very intuitive solution would be to add two more types of bitmasks (besides the 2x2 and 3x3 one), namely for the two possible half offset grids. These would consist of 7 bits (center bit plus 6 neighbor bits), three in the middle, two on the left and two on the right (of top and bottom for the other half offset mode). Here is how I imagine this to look like:

To sum up, I suggest: Add two new types of bitmasks (vertical half offset and horizontal half offset), Then adapt the tilemap code, so the bitmask is assigned correctly (when the user turns on one of the half offsets, the the bitmask computation needs to adapt).

I believe, this is the way, the user thinks. Fiddling around with the center bit unnecessary work for the user and can easily be done automatically (once it is implemented). Of course, I know, this solution is more difficult to implement. I hope my ideas are helpful for your discussion.

[SpechtMagnus]

I actually like this idea quite a lot as it would probably be the most intuitive way to implement this without breaking backwards campatibility. I think I could implement this very quickly but I would really like to here the opinion of someone who knows a little bit more about the plans for this engine as I'm just a developer.
MarianoGnu, @akien-mga or @reduz possibly

[KoBeWi]

Can anyone still reproduce this bug in Godot 3.2.3 rc3 or any later release?

If yes, please ensure that an up-to-date Minimal Reproduction Project (MRP) is included in this report (a MRP is a zipped Godot project with the minimal elements necessary to reliably trigger the bug). You can upload ZIP files in an issue comment with a drag and drop.

[akien-mga]

No answer, so closing.

Please comment if you can still reproduce it, and this issue can be reopened. Or open a new issue with a MRP if this would be better suited.

[Merlin1846]

It's less of a bug and more of just the way the tile set auto tile works, so until one of the suggested fixes are implemented it will always be reproducible.