Add L-system trees as decorations

[cx384]

Goal of the PR

Resolves #7346

How does it work

It uses the same L-system as minetest.spawn_tree, but it's faster since it uses the engine version of make_ltree, which directly writes it into the VoxelManipulator during map generation and does not trigger a light update.

Roadmap

It's listed at https://dev.minetest.net/TODO.

To do

This PR is Ready for Review.

How to test

Add some Decorations, e.g. in MTG.
(aaa is a good seed for v7)

minetest.register_decoration({
	name = "default:testdeco",
	deco_type = "lsystem",
	place_on = {"default:dirt_with_grass", "default:sand"},
	sidelen = 80,
	fill_ratio = 0.005,
	y_max = 31000,
	y_min = 1,
	treedef = {
		axiom="FFFFFA",
		rules_a="[&&&F][&&&++++F][&&&----F]",
		trunk="default:coral_skeleton",
		leaves="default:permafrost_with_stones",
		angle=30,
		iterations=2,
		random_level=0,
		trunk_type="single",
		thin_branches=true,
	},
})

Also test if alias work:

• replace default:coral_skeleton with alias_test1,
• replace default:permafrost_with_stones with alias_test2
• and add:

minetest.register_alias("alias_test1", "default:sand")
minetest.register_alias("alias_test2", "default:snow")

[Wuzzy2]

In the documentation, I would remove the treedef example as its redundant. Referring to the minetest.spawn_tree function is enough.

[cx384]

I removed the redundant example, but I'm unsure what should be there, since for all other fields there are proper examples in the table. However, a reference to the L-system trees section is probably enough.
Moreover, I added a features flag and rebased.

[Wuzzy2]

A question: Since L-system trees might have some randomness involved, how does this randomness get applied in the mapgen?

Does this guarantee that equal world seeds will generate an identical map with identical L-system trees? If not, that should be fixed. Because I think all mapgen stuff (terrain, biomes, decorations, etc.) should be equal if the world seed is the same.

[cx384]

A question: Since L-system trees might have some randomness involved, how does this randomness get applied in the mapgen?

The position is calculated like for any other decoration, depending on the world seed, and the L-system seed gets calculated using the position. (Or the seed will be static if defined in the tree definition.)

minetest/src/mapgen/treegen.cpp

Line 154 in 3cac17d

seed = p0.X * 2 + p0.Y * 4 + p0.Z; // use the tree position to seed PRNG

Does this guarantee that equal world seeds will generate an identical map with identical L-system trees?

Yes.

The only concern may be, that two L-system decoration, which get placed at the same position in two different worlds, will look exactly the same. However, this does not happen too often and I doubt anyone would notice.

To fix this, the L-system tree seed could generally be calculated differently by taking the world seed into account (and not just the position), but this is out of scope for this PR, I think.

[cx384]

Rebased.

[SmallJoker]

PR works what it promises. I noticed that some decorations might generate in the same place, which is ugly and needs fixing by someone in a separate PR later on.

The comments aside - looks good so far. Thank you.

[cx384]

Thanks for the reviews, SmallJoker and sfan5.
I handled every suggestion, and rebased. (since it didn't compile with the Irrlicht version)
Note that

• read_deco_lsystem still needs a ndef as parameter since it can't access the one from DecoLSystem,
• read_tree_def moved to c_content.{cpp,h} and not c_converter.{cpp,h},
• and tree_def is a shared_ptr now.

[SmallJoker]

Works on my machine.

[sfan5]

works