Page 1 of 1

Stability of each mapgen

PostPosted: Sun Dec 24, 2017 16:44
by paramat
'Official stability' means we will not alter the terrain shape (that includes the structure of caves) unless it is unavoidable as part of a bugfix.
For non-mgv6 mapgens the biome system is separate and defined by MTGame, and is not stable, however changes are likely to be subtle and the worst that can happen is a few straight-edged biomes.

Mgv6 has been officially stable since 2012. Since mgv6 biomes are hardcoded into the engine the biomes are also stable.

Mgv7 terrain has been officially stable since 0.4.16 when it was made the default mapgen, however the optional floatlands (disabled by default) are not stable yet, maybe they will be for 0.5.0.

Mgsinglenode is obviously officially stable.

Mgv5, mgflat, mgfractal, mgvalleys and mgcarpathian terrains are not officially stable.

Re: Stability of each mapgen

PostPosted: Tue Dec 26, 2017 21:40
by Wuzzy
Thank you, this is interesting information, but how is this news? ;-)

Information like this needs to be written down in README files, documentation, etc. instead of scattering it in the forums.

Re: Stability of each mapgen

PostPosted: Mon Apr 16, 2018 19:20
by paramat
Updated to clarify that only 3 mapgens are officially stable.

Re: Stability of each mapgen

PostPosted: Tue Apr 17, 2018 00:09
by v-rob
I find it funny how mgv5 was made in 2011 and it isn't considered stable. :)

Singlenode is stable?! *Gasp!* :P

Re: Stability of each mapgen

PostPosted: Tue Apr 17, 2018 00:10
by rubenwardy
v-rob wrote:I find it funny how mgv5 was made in 2011 and it isn't considered stable. :)

Singlenode is stable?! *Gasp!* :P


The v5 we have now was made much later, around 2015 or 16 I forget. The original v5 was removed.

Re: Stability of each mapgen

PostPosted: Tue Apr 17, 2018 00:21
by v-rob
Yes, but it still uses the same noises (with a few modifications, I believe)

Re: Stability of each mapgen

PostPosted: Tue Apr 17, 2018 14:14
by twoelk
v-rob wrote:Yes, but it still uses the same noises (with a few modifications, I believe)

rather, total rewrite to achieve the same effect

Re: Stability of each mapgen

PostPosted: Tue Apr 17, 2018 19:36
by Wuzzy
What exactly are the differences between the historical mapgen v5 from 0.3 and the current mapgen v5?

So far the only obvious difference I know of is usage of the biome API.
I believe the terrain shape is mostly the same. But is it identical?

Re: Stability of each mapgen

PostPosted: Fri May 04, 2018 00:28
by paramat
For mgv5:
Terrain surface shape is identical or as close as it can be. The eased 3D noise new mgv5 uses may create a slightly different shape to the original 3D noise implementation.
The 3D noise tunnels are identical in shape.
Old mgv5 only had these tunnels and had random water and lava sources added underground. New mgv5 doesn't have these sources, instead it adds the 'large caves' from mgv6, which often contain water or lava, these were considered a better way to include underground liquids.
Also there is an option (default enabled) for the giant caverns deep underground.

Of course, the biomes, ores and decorations are completely different.

Re: Stability of each mapgen

PostPosted: Sat Jun 30, 2018 02:04
by Brian Gaucher
My question is.

When will mgv7, mgflat, and mgvalleys become stable?

Mgfractal will probably never be, given it isn't meant to be used in servers or for actually playing (mostly testing, proof-of-concept). And mgcarpathian is quite new.

Also, I think Minetest could have a tab that describes each MapGen. I still have no idea what they each are (other than fractal, flat and singlenode).

Re: Stability of each mapgen

PostPosted: Mon Jul 02, 2018 03:01
by paramat
Mgv7 surface and underground is stable. The floatlands are not but they are disabled by default, they also seem quite well tuned so they might become stable at 5.0.0.

Mgflat in flat mode is stable. The only uncertainty about stability is the 'lakes' and 'hills' options but there seems no reason to alter those.

Mgvalleys and mgv5 seem near-stable. maybe for 5.0.0.

We're very hesitant about officially announcing a mapgen as stable because that is a big commitment that rules out certain types of improvement. However mapgens are often near-stable long before being 'officially' stable.
If you wait for official stability you'll miss out on good stuff for years. It's better to accept that very occasionally there may be a discontinuity in terrain.

Re: Stability of each mapgen

PostPosted: Mon Jul 02, 2018 07:23
by hajo
paramat wrote:Mgv7 surface and underground is stable. The floatlands are not
Mgflat in flat mode is stable. The only uncertainty about stability is the 'lakes' and 'hills' options

mapgens are often near-stable long before being 'officially' stable.
very occasionally there may be a discontinuity in terrain.

Would it be possible to detect such discontinuities ?

Eg. with a mod that gets called whenever new terrain is generated (and writes a logfile),
and another mod that keeps emerging more and more of the map.

Re: Stability of each mapgen

PostPosted: Tue Jul 03, 2018 20:19
by paramat
A mod might be able to detect large flat walls running along mapchunk borders yes, and that mod could possibly also smooth the transitions.

Re: Stability of each mapgen

PostPosted: Tue Jul 03, 2018 21:05
by Krock
paramat wrote:and that mod could possibly also smooth the transitions.

Yes, such a mod already exists. However, detecting borders is slightly more complicated. It would need a on_generated callback which then loads the surrounding mapblocks to check whether there's a large change in terrain height in order to flatten it there. In addition to the height change, the biome might also have changed - which has then to be spread a bit to make the transition look acceptable.
It's much work to implement this - and the mod would cause additional lag for each new chunk to ensure a continuous terrain.

Re: Stability of each mapgen

PostPosted: Tue Jul 03, 2018 22:02
by ExeterDad
Krock wrote:Yes, such a mod already exists.

Whoa! How did I not know of this tool? Impressive videos!

Re: Stability of each mapgen

PostPosted: Tue Jul 03, 2018 22:26
by v-rob
Krock wrote:Yes, such a mod already exists.


excessive flattening + trees =
+ Spoiler

Re: Stability of each mapgen

PostPosted: Sat Jul 14, 2018 19:01
by paramat
I intend to have mgv5 and mgflat announced as officially stable for 5.0.0.
Mgvalleys is still somewhat of a mess, that i am slowly sorting out, so won't be for a while.
Mgcarpathian i might add a feature to (that has support from the mapgen creator), and is quite new, so again not for a while.

Re: Stability of each mapgen

PostPosted: Fri May 24, 2019 18:38
by uwu
How stable are v7's floatlands by now?

Re: Stability of each mapgen

PostPosted: Thu Jun 06, 2019 00:39
by paramat
There's been a complete change in the stating of mapgen stability.

Now, we only state which mapgens (or parts of mapgens) are 'highly unstable'.
Everything else should be considered as increasingly stable with time.
Now we recognise that stability is a smoothly varying thing, instead of 2 states: 'unstable' and 'stable'.
It is always possible, if it is absolutely necessary, that something may change, even in v6.
Any changes are likely to be minor, at worst, terrain being discontinuous between previously generated world and newly generated world.
Any mapgen that makes it into a stable release will be supported from then onwards.

Currently, only v7 floatlands are highly unstable. I am not happy with their generation, i might tune the parameters or even rewrite their generation method.

Re: Stability of each mapgen

PostPosted: Thu Jun 06, 2019 09:47
by Linuxdirk
paramat wrote:Currently, only v7 floatlands are highly unstable. I am not happy with their generation, i might tune the parameters or even rewrite their generation method.

They look good, though. Maybe just prevent water leaking off the islands.