Map generation algorithms

[galok]

We all know about minecraft decompiled sources, so called "Mod Coder Pack". So I want to ask, is MineTest map generation algorithms (v5, v6, v7) are based on decompiled Minecraft map generation algorithms or they are done from scratch?

[rubenwardy]

Done from scratch, I'm pretty sure

[lightonflux]

No MC code is in MT. No MT code is derived from MC code.

[galok]

Okay, I get it about the code. Tell me then, is there any point to read MC code after MT code (map generation), just how different they are? besides MC lack of comments and variable names. I mean, the result look quite similiar.

[qwertymine3]

Okay, I get it about the code. Tell me then, is there any point to read MC code after MT code (map generation), just how different they are? besides MC lack of comments and variable names. I mean, the result look quite similiar.

The biome placement code is very different to Minetest, however you don't need to read the code for this, this page covers it very well (look for the "grown biomes" section) http://mc-server.xoft.cz/docs/Generator.html.

[galok]

qwertymine3, thank you! This seems to be a very informative article. It's already answered a few questions for me.

[paramat]

I can confirm no core mapgens are derived from MC.

[galok]

have one more question on this topic. the point of generation algorithm is to generate exactly the same blocks in exact same chunk by demand, this also mean that order in which chunks are generated should not affect blocks that generated in specific chunks. Please, correct me if I'm wrong on this one, but else, if I get it right, then tell me: is use of a global random does kinda mess up this system? I mean this code in mapgen_v6.cpp:960

Your phone or window isn't wide enough to display the code box. If it's a phone, try rotating it to landscape mode.

Code: Select all

      // Put trees in random places on part of division
      for (u32 i = 0; i < tree_count; i++) {
         s16 x = myrand_range(p2d_min.X, p2d_max.X); // this one
         s16 z = myrand_range(p2d_min.Y, p2d_max.Y); // this one too


Position of trees would be different every time they will be generated for this chunk, right? (Implied that full chunk will be re-generated)

[paramat]

Most mapgens use pseudorandom randoms so will be deterministic, but for minor things like leaf randomness the random may not be deterministic.

[galok]

paramat, is there a particular reason for that?

[paramat]

I'm not sure which are deterministic and which are not, but it's possible some are not because they are so trivial, such as the randomness of leaves on a tree. Also mgv6 may be less deterministic because it's older code, newer mapgens have higher quality standards (thanks to hmmmm).

[galok]

paramat, okay, thank you, I think I got it. But which of the mapgens is a newest one? Judging from github it's a mapgen_v5.cpp, which was added at Nov 9, 2014. But a whole countdown thing from v6 and v7 is a bit tricky to understand. So which one is it?

[twoelk]

Legendary past

then

-v5 was made
-v6 was created and made default, v5 was dumped
-v7 was made and released before finished, v6 stayed default
-v5 was recreated from scratch and readded, v7 was made better, v6 optimised and stayed default
-valleys was created as lua mapgen and then rewritten and added to Mintest core as new mapgen, v6 was further optimized and stayed default
-more mapgens where added such as fractals (forgot when mathgen was dumped)
-mgflat was extended and is somewhat of a complete mapgen now

new-v5, default-v6, prereleased-v7, valleys and many others are all alive and kicking
most games run with v6, some recommend or even need v7, some subgames don't care, others use singlenode and apply their own on top

I may have missed some or mixed up some of the timeline but it should give a general idea

[paramat]

^ That.

Newest is mgflat.

[galok]

twoelk, thank you, you're most certainly brought order into this matter.