Refactor world generation by RecursivePineapple · Pull Request #39 · Cardinalstars/CubicChunks1710

RecursivePineapple · 2025-11-14T17:04:19Z

Work on worlgen framework some more
CoordinatePacker mixin
Prevent LightingManager NPE
Cleanup
Angelica interop
Lots of changes
spotless

…ther misc fixes

src/main/java/com/cardinalstar/cubicchunks/api/CCAPI.java

src/main/java/com/cardinalstar/cubicchunks/api/XZMap.java

...ava/com/cardinalstar/cubicchunks/mixin/early/common/worldgen/MixinChunkProviderGenerate.java

src/main/java/com/cardinalstar/cubicchunks/api/world/storage/ICubicStorage.java

src/main/java/com/cardinalstar/cubicchunks/util/DependencyGraph.java

src/main/java/com/cardinalstar/cubicchunks/api/worldgen/impl/StandardDependencyRegistry.java

Co-authored-by: Ryan Nasers <42074409+Cardinalstars@users.noreply.github.com>

src/main/java/com/cardinalstar/cubicchunks/async/TaskPool.java

Cardinalstars · 2026-01-11T19:20:19Z

src/main/java/com/cardinalstar/cubicchunks/async/TaskPool.java

+            TaskContainer<TTask, TResult> container = new TaskContainer<>(executor);
+            container.tasks.add(future);
+
+            TASK_QUEUE.addLast(container);


Suggested change

TASK_QUEUE.addLast(container);

TASK_QUEUE.add(container);

It's a nitpick, but this is a bit more clear imo

I prefer addLast because you can add to the front or to the end of the queue, and add isn't explicit enough IMO.

github-actions · 2026-01-11T19:32:42Z

#40

Co-authored-by: GitHub GTNH Actions <>

Cardinalstars · 2026-01-11T20:14:38Z

src/main/java/com/cardinalstar/cubicchunks/mixin/early/common/MixinCoordinatePacker.java

Is this something to eventually go into gtnhlib itself? Mixin on mods we control is sus obviously.

No, this is the best solution. It's ugly, but we can't add any indirection to these methods or it'll kill their performance. As it is, the JVM will inline and optimize everything as if the gtnhlib versions didn't exist.

Cardinalstars · 2026-01-11T20:22:07Z

src/main/java/com/cardinalstar/cubicchunks/mixin/early/common/MixinRegionFileCache.java

Just delete the method if we comment it out? Or is this something wip?

I commented it out because that method didn't exist in the newer version of regionlib, and I couldn't tell what it was supposed to do. I'll have to look at it again, I haven't looked into this much further.

Cardinalstars · 2026-01-11T21:32:44Z

src/main/java/com/cardinalstar/cubicchunks/mixin/early/common/MixinWorldServer.java

+    @Inject(method = "getChunkSaveLocation", at = @At("HEAD"), cancellable = true, remap = false)
+    private void redirectChunkSaveLocation(CallbackInfoReturnable<File> cir) {
+        if (this.theChunkProviderServer == null) {
+            WorldFormatSavedData format = WorldFormatSavedData.get((WorldServer) (Object) this);
+
+            // noinspection DataFlowIssue
+            cir.setReturnValue(
+                format.getFormat()
+                    .getWorldSaveDirectory(this.saveHandler, (WorldServer) (Object) this)
+                    .toFile());
+        }
+    }
+


What's the context on this change?

This decouples the world directory location from the anvil save format, since it wasn't being used at all. I forget why I had to do this, I think that world savedata wasn't being put in the right directories or something.

Cardinalstars · 2026-01-11T21:50:23Z

src/main/java/com/cardinalstar/cubicchunks/mixin/early/common/MixinWorld_HeightLimit.java

I just realized one of these is named _HeightLimit and the other _HeightLimits.
Maybe make them the same name.

Cardinalstars · 2026-01-11T21:58:21Z

src/main/java/com/cardinalstar/cubicchunks/modcompat/angelica/AngelicaInterop.java

+        return delegate;
+    }
+
+    public static void setDelegate(IAngelicaDelegate delegate) {


I currently see no uses of this. How is this supposed to be used?

It gets called in Angelica, then CC will call the various methods on the delegate to tell Angelica when cubes and columns are loaded. I haven't put up that PR yet because there are still some cube render culling issues.

Cardinalstars · 2026-01-12T19:54:29Z

src/main/java/com/cardinalstar/cubicchunks/network/PacketEncoderHeightMapUpdate.java

        }

-        return new PacketHeightMapUpdate(pos, updates, heights);
+        return new PacketHeightMapUpdate(pos, updates.clone(), heights);


I can sort of guess why you have to clone here, but I can't say I understand the full picture. Could you explain a bit? Is there some threading issue here that causes you needing to clone this?

I don't think there was an explicit bug, but I copied it just to be sure everything worked fine. The packet can stick around for any amount of time, but the provided updates list can be modified in the mean time.

src/main/java/com/cardinalstar/cubicchunks/server/CubicPlayerManager.java

Cardinalstars · 2026-01-13T22:04:53Z

src/main/java/com/cardinalstar/cubicchunks/server/chunkio/CubeIO.java

+
+        MinecraftForge.EVENT_BUS.post(event);
+
+        tag = event.tag;


Why reassign here?

Wait, the event can change the tag right?

Yeah, event handlers can do anything they want with the tag. It probably isn't used like that, but I did this just to make sure.

Cardinalstars · 2026-01-13T22:16:36Z

src/main/java/com/cardinalstar/cubicchunks/server/chunkio/CubeIO.java

+        TaskPool.submit(columnLoadExecutor, pos, tag -> {
+            if (tag == null) {
+                if (preloadFailures != null) preloadFailures.onColumnPreloadFailed(pos);
+            } else {
+                synchronized (columnCache) {
+                    columnCache.put(pos, new SaveData(tag.orElse(null), null, System.currentTimeMillis()));
+                }
+            }
+        });


Shouldn't you be checking tag.isEmpty, and otherwise you don't need to do the orElse(null)?

Inspection of the columnLoadExecutor makes me think that this can't really ever be a null value given from tag.

Yep you're right, I fixed it.

Cardinalstars · 2026-01-13T22:16:43Z

src/main/java/com/cardinalstar/cubicchunks/server/chunkio/CubeIO.java

+        synchronized (columnCache) {
+            columnCache.put(pos, new SaveData(tag, task, now));
+
+            if (columnCache.size() > 5000) {
+                int i = 0;
+
+                var iter = columnCache.object2ObjectEntrySet()
+                    .fastIterator();
+
+                while (columnCache.size() > 5000 && i++ < 100) {
+                    var e = iter.next();
+
+                    SaveData data = e.getValue();
+
+                    if (data.task != null && data.task.isDone()) {
+                        data.task = null;
+                    }
+
+                    if (data.task == null && data.lastAccess + EXPIRY < now) {
+                        iter.remove();
+                    }
+                }
+            }


I don't really understand this part of the saveColumn. Why do this in this method? Putting in the columnCache sure, but why do this iteration here if we want to make this as fast as possible. Can't some off-thread do this?

I'll make a task + executor for this. It's meant to be a garbage collector that scans the cache when it's full to remove old SaveData objects but it's definitely not ideal now that I'm looking at it again. The i++ portion basically only allows it to scan the first 100 elements in the map, and since I'm not using the getAndMoveToLast/putAndMoveToLast methods the map ordering is undefined.

Cardinalstars · 2026-01-13T22:17:41Z

src/main/java/com/cardinalstar/cubicchunks/server/chunkio/CubeIO.java

+    @Override
+    public void preloadCube(CubePos pos, CubeInitLevel wanted) {
+        TaskPool.submit(cubeLoadExecutor, pos, tag -> {
+            CubeInitLevel actual = !tag.isPresent() ? CubeInitLevel.None : IONbtReader.getCubeInitLevel(tag.get());


Suggested change

CubeInitLevel actual = !tag.isPresent() ? CubeInitLevel.None : IONbtReader.getCubeInitLevel(tag.get());

CubeInitLevel actual = tag.isEmpty() ? CubeInitLevel.None : IONbtReader.getCubeInitLevel(tag.get());

isEmpty was added in java 11, I can't use it. Intellij kept insisting on the same change so I had to disable that lint 🤣.

src/main/java/com/cardinalstar/cubicchunks/server/chunkio/CubeLoaderServer.java

src/main/java/com/cardinalstar/cubicchunks/util/biome3d/PalettizedBiomeArray.java

Cardinalstars · 2026-01-15T10:21:53Z

src/main/java/com/cardinalstar/cubicchunks/world/chunkloader/CubicChunkManager.java

What exactly is the problem here?

Forced cube loading wasn't important when I wrote this PR, so to keep things simple I just disabled most of its logic. It was highly coupled to the player manager. I wanted to revisit it later once the player manager matured (which it has at this point).

Cardinalstars

Once you look at these comments, I think it's ready. Shouldn't be much work either.

…anager.java Co-authored-by: Ryan Nasers <42074409+Cardinalstars@users.noreply.github.com>

* Work on worlgen framework some more * CoordinatePacker mixin * Prevent LightingManager NPE * Cleanup * Angelica interop * Lots of changes * spotless * Works somewhat better * Several improvements * MORE WORLDGEN FIXES!!! + chunk api/endless ids compat + 3d biomes + other misc fixes * Make sure it compiles * spotless * Fix EBS Y Level * Checkstyle fixes * fix dep versions * :clueless: * Update src/main/java/com/cardinalstar/cubicchunks/api/CCAPI.java Co-authored-by: Ryan Nasers <42074409+Cardinalstars@users.noreply.github.com> * Update dependency graphs * oops * spotlessApply (#40) Co-authored-by: GitHub GTNH Actions <> * Update src/main/java/com/cardinalstar/cubicchunks/server/CubicPlayerManager.java Co-authored-by: Ryan Nasers <42074409+Cardinalstars@users.noreply.github.com> * Review changes + biome fixes --------- Co-authored-by: Ryan Nasers <42074409+Cardinalstars@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

RecursivePineapple added 17 commits September 13, 2025 16:01

Work on worlgen framework some more

8fae17b

CoordinatePacker mixin

ab1db58

Prevent LightingManager NPE

2ce21a8

Cleanup

c2abab5

Angelica interop

0ff06b3

Merge branch 'master' into improve-worldgen

a78a771

Lots of changes

ac05363

spotless

6448476

Works somewhat better

b114a6b

Several improvements

2974efc

MORE WORLDGEN FIXES!!! + chunk api/endless ids compat + 3d biomes + o…

4012262

…ther misc fixes

Make sure it compiles

f574dbb

spotless

0c00345

Fix EBS Y Level

7f4a325

Checkstyle fixes

a902c30

fix dep versions

c11acb1

:clueless:

ba79156