Hello, everyone! A month ago I promised to write about various interesting things with the map in my opinion. Well the time has come.
Provinces quality
I just have to say thank you to @Pkeod
My first idea to apply this method was: make NE that blurs the image and set material with contrast shader. Then I realized that writing an algorithm for blurring is not the most pleasant thing. I also realized that processing for each image at runtime is terrible for performance.
I forgot about it for a while, but I remembered it during the development of a new game.
Now I pre-blur the images with a third party editor and then apply shader:
shader code
varying mediump vec2 var_texcoord0;
uniform lowp sampler2D texture_sampler;
uniform lowp vec4 tint;
const float threshold = .01;
void main()
{
// Pre-multiply alpha since all runtime textures already are
lowp vec4 tint_pm = vec4(tint.xyz * tint.w, tint.w);
vec4 c = texture2D(texture_sampler, var_texcoord0.xy);
vec3 res = smoothstep(.5-threshold, .5+threshold, c.xyz);
gl_FragColor = vec4(res,c.w) * tint_pm;
}
The screenshots were made at different times, so they have a different scale. In the case of distance fields, it is even larger.
I think the difference is obvious
ImageTransparencyCalculator
The name is really incomprehensible, I guess I should have thought a little longer.
How did I handle the click?
I used collisions in Civilization Path. Creating collisions that repeat the contours of the provinces is a pain. It takes a lot of time, and no matter how hard you try, the result will not be perfect.
At first I started using Convex hull shapes that were automatically generated from the image by a third-party editor. The result was better, and most importantly - faster.
Later in the Russian Telegram, there was a talk about click tracking and the idea of tracking a click using a color change and a render script came up. I tried to implement this but no success. In general, I doubt that this is possible in Defold.
I got inspired by this method and wrote a native extension. How it works:
The native extension takes the image data as input and rebuilds the buffer, leaving only the transparency data in it.
The resulting monochrome image can be reduced in NE. In this case, the accuracy drops, but the size of the generated file also decreases. This is just in case someone doesn’t want perfect accuracy.
Further, the file data is compressed by zlib.
And already on the basis of this data, the lua module calculates whether the transparent area was clicked.
I’m glad now I can perfectly track the click, and everything is automated.
I will also be glad if this is also useful to someone and saves time. I know that @russelkgd changed it for GUI. That’s cool.
Various dynamic objects
They are created using game object factories. From flags to nuclear explosions.
Performance is very important. Therefore, some objects are not always shown. Objects outside the camera are disabled.
I still use msg.post(“@render:”, “draw_line”) for lines, although I know it’s not good. I will definitely experiment with this. Maybe even the distance fields will try again 
It’s a sad that there is no slice 9 for sprites. I think this would be one of the solutions.