Jump to content

Alpvax

Members
  • Posts

    304
  • Joined

  • Last visited

  • Days Won

    3

Posts posted by Alpvax

  1. There are 3 methods which sound like they do something very similar, overridden by various different blocks:

     

    • updatePostPlacement is used to make connections with adjacent blocks (fences/walls).
    /**
    * Update the provided state given the provided neighbor facing and neighbor state, returning a new state.
    * For example, fences make their connections to the passed in state if possible, and wet concrete powder immediately
    * returns its solidified counterpart.
    * Note that this method should ideally consider only the specific face passed in.
    */
    public BlockState updatePostPlacement(BlockState stateIn, Direction facing, BlockState facingState, IWorld worldIn, BlockPos currentPos, BlockPos facingPos)
    • neighborChanged appears to be used for detecting redstone changes (or water in the case of sponge)
    // No Javadoc
    public void neighborChanged(BlockState state, World worldIn, BlockPos pos, Block blockIn, BlockPos fromPos, boolean isMoving)
    • And the forge-added onNeighborChange is called from World#updateComparatorOutputLevel:
    /**
    * Called when a tile entity on a side of this block changes is created or is destroyed.
    * @param world The world
    * @param pos Block position in world
    * @param neighbor Block position of neighbor
    */
    default void onNeighborChange(BlockState state, IWorldReader world, BlockPos pos, BlockPos neighbor)

    My question is which method should be overridden for what purpose, as many of the call paths are similar. I have spent quite a bit of time trying to trace the call paths and have become lost multiple times.

    I was using neighborChanged, but switched to using updatePostPlacement (as that already supplies the direction, and sounds more like what I needed (making connections between blocks)). The effect in-game did not change, so I am wondering when I should be using each of these methods.

     

    There is also the following forge-added method, which doesn't appear to be overridden anywhere, and is only called by the blockstate method of the same name, which is itself not called:

    /**
    * Called on an Observer block whenever an update for an Observer is received.
    *
    * @param observerState The Observer block's state.
    * @param world The current world.
    * @param observerPos The Observer block's position.
    * @param changedBlock The updated block.
    * @param changedBlockPos The updated block's position.
    */
    default void observedNeighborChange(BlockState observerState, World world, BlockPos observerPos, Block changedBlock, BlockPos changedBlockPos)

     

  2. That is the only way. All blockstates are calculated on registration, so there is no way to save that to a single blockstate.

    If there are that many, you may need to rethink your approach.

    You could maybe store a chunk capability with a map of blockpos -> blockstate, but I'm not sure that would save you much (after all, that is what the world saves). And you'd have a much harder time sorting out the rendering.

  3. It is also possible to use NBT tags to create swords which have custom attributemodifiers and models in order to have more than the usual sword types.

    It can be done with any item, but you cannot add new functionality and behind the scenes they are still the vanilla sword item.

    See https://minecraft.gamepedia.com/Model#Item_predicates for more info on custom models (uses an NBT tag to select which model override to use).

  4. Even closer now.

    I do not want to update neighbours, as the property is changed from Block#neighborChanged, and I do not want to cause an infinite loop.

    //Called after updating property
    BlockState blockState = getBlockState();
    //Do not update neighbours
    getWorld().notifyBlockUpdate(getPos(), blockState, blockState, Constants.BlockFlags.BLOCK_UPDATE | Constants.BlockFlags.RERENDER_MAIN_THREAD); 
    requestModelDataUpdate();

    I also call requestModelDataUpdate at the end of handleUpdateTag.

    The data is now correct on the client (displayed when right clicked), but the render is always 1 step behind (Only updates once another block is placed).

  5. Almost fully working.

    Only thing I'm now struggling with is forcing my TileEntity to sync.

    I'm calling markDirty and requestModelDataUpdate after changing the property, but the client side data doesn't refresh until I close and reopen the world.

    Do I need to send a custom packet, or is there an easier way of forcing a resync?

  6. I am in the process of converting my existing blockstate based model to a tileentity due to excessive use of blockstates (4-value EnumProperty for each direction = 4096 + waterlogged = 8192 permutations!)

    I have an existing (generated) multipart blockstate json, is there any way to easily convert it to the same model, just using the data from the tileentity instead of the blockstate?

     

    I'm assuming I need to create a custom BakedModel, but I would ideally like to keep the current data-driven approach (one model for each of the 4 enum values).

     

    It's the first time I've started looking into custom models, and I think a TileEntityRenderer is overkill for static models. Please correct me if I'm wrong.

  7. 6 hours ago, frakier said:

    "resources\data\minecraft\tags\items\carnivore.json"

     

    6 hours ago, frakier said:

    "resources\data\mymod\tags\items\carnivore.json"

    These are 2 different tags ("minecraft:carnivore" and "mymod:carnivore" respectively).

    You should make 1 tag "mymod:carnivore" and add all the defaults to it (as in all the vanilla items, and your custom items if applicable), then if you want to add items in another mod, you also define the "mymod:carnivore" tag (not the "mysecondmod:carnivore", again that would be a different tag).

     

    As an extension on what draco said, the only time you should put tags in the minecraft domain is if you are overriding/adding to an existing vanilla tag.

  8. 17 hours ago, BlockyPenguin said:

    [ "one": {

    That is not valid json. Currently you have an array with keys.

    Either it should be an array of objects "[ { ..." or it should be an object with keys of "one, two,...".

    I would suggest going for the first option (especially as that is what you support in your code):

    17 hours ago, BlockyPenguin said:

    JsonArray jsonSteps = jsonMethod.getAsJsonArray("steps");

     

    • Like 1
  9. Worked around the issue by double unchecked casting the type:

    
    public abstract class IOType<T> extends ForgeRegistryEntry<IOType<?>> {
        @SuppressWarnings("unchecked")
        public static final Class<IOType<?>> CLASS_GENERIC = (Class<IOType<?>>)((Class<?>)IOType.class);
    }
    
    //Can then be used as follows
    public static final DeferredRegister<IOType<?>> IO_TYPES = DeferredRegister.create(IOType.CLASS_GENERIC, MODID);

    By putting the constant in the API class, it can be used by others who are consuming the API.

    • Like 1
  10. DeferredRegister has been updated recently to make it work with custom registries, but it still doesn't work with custom generic RegistryObjects (similar to TileEntityType<T>).

     

    When trying to create the DeferredRegister, it errors with: 

    Quote

    Required type: DeferredRegister<IOType<?>>
    Provided: DeferredRegister<IOType>
    Incompatible equality constraint: IOType<?> and IOType

    //Registry
      public static final DeferredRegister<IOType<?>> IO_TYPES = DeferredRegister.create(IOType.class, MODID);
    
    //IOType.java (currently just a stub)
    public abstract class IOType<T> extends ForgeRegistryEntry<IOType<?>> {
    }

     

    Is there a workaround for this issue, or can I just not use DeferredRegister? This will be available in an API, so I would ideally be able to support the more popular DeferredRegister approach.

     

    Removing the parametised type (DeferredRegister<IOType> IO_TYPES = ...) then throws the error of IOType not implementing IForgeRegistryEntry<IOType>, so that isn't a fix.

    I am wondering if the only workaround is to make the registry in a static initialiser (and overriding it in the NewRegistryEvent), but that seems like a step backwards.

     

    Does anyone have any suggestions?

×
×
  • Create New...

Important Information

By using this site, you agree to our Terms of Use.