curved walls

[lmangall]

If the curved walls are appearing only (or mostly) on the edges of the image, can't we just hide it by making the actual image bigger than the window ? In this way the part where the problem is most visible will be out of view. Of course it's "dirt under the carpet" kind of solution...

[552020]

Yes we can. But is the problem otherwise hard to solve. Did you understand it? I would try to understand it first and then if it's impossible (or if it would take too much time) to solve or too hard to understand, I would be ok with the under the carpet solution.

[lmangall]

No I didn't understood the problem. I guess if we have to change the algo part of the code it might be doable, now it isn't a big part of it...

[552020]

I would say it's the algo part for sure, no?

Basically, it could be three things:

• the raycaster
• the fisheye correction
• maybe the render wall function, I mean the part where the height of the wall is calculated.

But I have no so much idea of what is going on in the renderer. Do you prefer to go for the carpet solution?

[lmangall]

This line is the fisheye correction:
data->ray->length *= cos(normalize_angle(data->ray->angle_rd - data->player->orientation_angle_rd));

Most likely it is the raycaster (the algorithm), because it was confirmed by two persons...

I think changing the algo would mean changing at least:
void update_ray(t_data *data) void update_length_and_collision_orientation(t_data *data) void render_wall_background(t_data *data)

[lmangall]

For documentation:

With fov 120°, without fisheye correction:

---

With fov 120°, with fisheye correction:

°

[lmangall]

data->ray->length *= cos(normalize_angle((data->ray->angle_rd - data->player->orientation_angle_rd) / 2));

---

	data->ray->length *= cos(normalize_angle(data->ray->angle_rd - data->player->orientation_angle_rd));

>

[lmangall]

data->ray->length *= cos(normalize_angle((data->ray->angle_rd - data->player->orientation_angle_rd) / 7));

_______

data->ray->length *= cos(normalize_angle((data->ray->angle_rd - data->player->orientation_angle_rd) / 6));

[lmangall]

Just posting to document and as a not to self
So, I was focused on the corner and thought I had it as good as it can be untill I realised the distorsion moved to the center of the screen...

I also started implementing a function that would calculate the fisheye compensation proportionaly to the FOV...

[lmangall]

I talked with a couple people, who already did cub3D, and they argued that it's not really something we should correct... In Cub3d the FOV is not a user defined parameter and ours is fine with 60, which is standard.

I will look a bit more into it nontheless

[lmangall]

` double angle;

angle = data->player->orientation_angle_rd - data->ray->angle_rd;
data->ray->length = data->ray->length * cos(angle);`

fov 120: