Custom Loot Objects

Due to the complexity of the loot table system, there are several registries at work, all of which can be used by a modder to add more behavior.

All loot table related registries follow a similar pattern. To add a new registry entry, you generally extend some class or implement some interface that holds your functionality. Then, you define a codec for serialization, and register that codec to the corresponding registry, using DeferredRegister like normal. This goes along with the "one base object, many instances" approach most registries (for example also blocks/blockstates and items/item stacks) use.

Custom Loot Entry Types

To create a custom loot entry type, extend LootPoolEntryContainer or one of its two direct subclasses, LootPoolSingletonContainer or CompositeEntryBase. For the sake of example, we want to create a loot entry type that returns the drops of a entity - this is purely for example purposes, in practice it would be more ideal to directly reference the other loot table. Let's start by creating our loot entry type class:

// We extend LootPoolSingletonContainer since we have a "finite" set of drops.
// Some of this code is adapted from NestedLootTable.
public class EntityLootEntry extends LootPoolSingletonContainer {
    // A Holder for the entity type we want to roll the other table for.
    private final Holder<EntityType<?>> entity;

    // It is common practice to have a private constructor and have a static factory method.
    // This is because weight, quality, conditions, and functions are supplied by a lambda below.
    private EntityLootEntry(Holder<EntityType<?>> entity, int weight, int quality, List<LootItemCondition> conditions, List<LootItemFunction> functions) {
        // Pass lambda-provided parameters to super.
        super(weight, quality, conditions, functions);
        // Set our values.
        this.entity = entity;
    }

    // Static builder method, accepting our custom parameters and combining them with a lambda that supplies the values common to all entry types.
    public static LootPoolSingletonContainer.Builder<?> entityLoot(Holder<EntityType<?>> entity) {
        // Use the static simpleBuilder() method defined in LootPoolSingletonContainer.
        return simpleBuilder((weight, quality, conditions, functions) -> new EntityLootEntry(entity, weight, quality, conditions, functions));
    }

    // This is where the magic happens. To add an item stack, we generally call #accept on the consumer.
    // However, in this case, we let #getRandomItems do that for us.
    @Override
    public void createItemStack(Consumer<ItemStack> consumer, LootContext context) {
        // Get the entity's loot table. If it doesn't exist, an empty loot table will be returned, so null-checking is not necessary.
        LootTable table = context.getLevel().reloadableRegistries().getLootTable(entity.value().getDefaultLootTable());
        // Use the raw version here, because vanilla does it too. :P
        // #getRandomItemsRaw calls consumer#accept for us on the results of the roll.
        table.getRandomItemsRaw(context, consumer);
    }
}

Next up, we create a MapCodec for our loot entry:

// This is placed as a constant in EntityLootEntry.
public static final MapCodec<EntityLootEntry> CODEC = RecordCodecBuilder.mapCodec(inst ->
        // Add our own fields.
        inst.group(
                        // A value referencing an entity type id.
                        BuiltInRegistries.ENTITY_TYPE.holderByNameCodec().fieldOf("entity").forGetter(e -> e.entity)
                )
                // Add common fields: weight, display, conditions, and functions.
                .and(singletonFields(inst))
                .apply(inst, EntityLootEntry::new)
);

We then use this codec in registration:

public static final DeferredRegister<LootPoolEntryType> LOOT_POOL_ENTRY_TYPES =
        DeferredRegister.create(Registries.LOOT_POOL_ENTRY_TYPE, ExampleMod.MOD_ID);

public static final Supplier<LootPoolEntryType> ENTITY_LOOT =
        LOOT_POOL_ENTRY_TYPES.register("entity_loot", () -> new LootPoolEntryType(EntityLootEntry.CODEC));

Finally, in our loot entry class, we must override getType():

public class EntityLootEntry extends LootPoolSingletonContainer {
    // other stuff here

    @Override
    public LootPoolEntryType getType() {
        return ENTITY_LOOT.get();
    }
}

Custom Number Providers

To create a custom number provider, implement the NumberProvider interface. For the sake of example, let's assume we want to create a number provider that changes the sign of the provided number:

// We accept another number provider as our base.
public record InvertedSignProvider(NumberProvider base) implements NumberProvider {
    public static final MapCodec<InvertedSignProvider> CODEC = RecordCodecBuilder.mapCodec(inst -> inst.group(
            NumberProviders.CODEC.fieldOf("base").forGetter(InvertedSignProvider::base)
    ).apply(inst, InvertedSignProvider::new));

    // Return a float value. Use the context and the record parameters as needed.
    @Override
    public float getFloat(LootContext context) {
        return -this.base.getFloat(context);
    }

    // Return an int value. Use the context and the record parameters as needed.
    // Overriding this is optional, the default implementation will round the result of #getFloat.
    @Override
    public int getInt(LootContext context) {
        return -this.base.getInt(context);
    }

    // Return a set of the loot context params used by this provider. See below for more information.
    // Since we have a base value, we just defer to the base.
    @Override
    public Set<LootContextParam<?>> getReferencedContextParams() {
        return this.base.getReferencedContextParams();
    }
}

Like with custom loot entry types, we then use this codec in registration:

public static final DeferredRegister<LootNumberProviderType> LOOT_NUMBER_PROVIDER_TYPES =
        DeferredRegister.create(Registries.LOOT_NUMBER_PROVIDER_TYPE, ExampleMod.MOD_ID);

public static final Supplier<LootNumberProviderType> INVERTED_SIGN =
        LOOT_NUMBER_PROVIDER_TYPES.register("inverted_sign", () -> new LootNumberProviderType(InvertedSignProvider.CODEC));

And similarly, in our number provider class, we must override getType():

public record InvertedSignProvider(NumberProvider base) implements NumberProvider {
    // other stuff here

    @Override
    public LootNumberProviderType getType() {
        return INVERTED_SIGN.get();
    }
}

Custom Level-Based Values

Custom LevelBasedValues can be created by implementing the LevelBasedValue interface in a record. Again, for the sake of example, let's assume that we want to invert the output of another LevelBasedValue:

public record InvertedSignLevelBasedValue(LevelBasedValue base) implements LevelBaseValue {
    public static final MapCodec<InvertedLevelBasedValue> CODEC = RecordCodecBuilder.mapCodec(inst -> inst.group(
            LevelBasedValue.CODEC.fieldOf("base").forGetter(InvertedLevelBasedValue::base)
    ).apply(inst, InvertedLevelBasedValue::new));

    // Perform our operation.
    @Override
    public float calculate(int level) {
        return -this.base.calculate(level);
    }

    // Unlike NumberProviders, we don't return the registered type, instead we return the codec directly.
    @Override
    public MapCodec<InvertedLevelBasedValue> codec() {
        return CODEC;
    }
}

And again, we then use the codec in registration, though this time directly:

public static final DeferredRegister<MapCodec<? extends LevelBasedValue>> LEVEL_BASED_VALUES =
        DeferredRegister.create(Registries.ENCHANTMENT_LEVEL_BASED_VALUE_TYPE, ExampleMod.MOD_ID);

public static final Supplier<MapCodec<? extends LevelBasedValue>> INVERTED_SIGN =
        LEVEL_BASED_VALUES.register("inverted_sign", () -> InvertedSignLevelBasedValue.CODEC);

Custom Loot Conditions

To get started, we create our loot item condition class that implements LootItemCondition. For the sake of example, let's assume we only want the condition to pass if the player killing the mob has a certain xp level:

public record HasXpLevelCondition(int level) implements LootItemCondition {
    // Add the context we need for this condition. In our case, this will be the xp level the player must have.
    public static final MapCodec<HasXpLevelCondition> CODEC = RecordCodecBuilder.mapCodec(inst -> inst.group(
            Codec.INT.fieldOf("level").forGetter(HasXpLevelCondition::level)
    ).apply(inst, HasXpLevelCondition::new));
    
    // Evaluates the condition here. Get the required loot context parameters from the provided LootContext.
    // In our case, we want the KILLER_ENTITY to have at least our required level.
    @Override
    public boolean test(LootContext context) {
        Entity entity = context.getParamOrNull(LootContextParams.KILLER_ENTITY);
        return entity instanceof Player player && player.experienceLevel >= level; 
    }
    
    // Tell the game what parameters we expect from the loot context. Used in validation.
    @Override
    public Set<LootContextParam<?>> getReferencedContextParams() {
        return ImmutableSet.of(LootContextParams.KILLER_ENTITY);
    }
}

We can register the condition type to the registry using the condition's codec:

public static final DeferredRegister<LootItemConditionType> LOOT_CONDITION_TYPES =
        DeferredRegister.create(Registries.LOOT_CONDITION_TYPE, ExampleMod.MOD_ID);

public static final Supplier<LootItemConditionType> MIN_XP_LEVEL =
        LOOT_CONDITION_TYPES.register("min_xp_level", () -> new LootItemConditionType(HasXpLevelCondition.CODEC));

After we have done that, we need to override #getType in our condition and return the registered type:

public record HasXpLevelCondition(int level) implements LootItemCondition {
    // other stuff here

    @Override
    public LootItemConditionType getType() {
        return MIN_XP_LEVEL.get();
    }
}

Custom Loot Functions

To get started, we create our own class extending LootItemFunction. LootItemFunction extends BiFunction<ItemStack, LootContext, ItemStack>, so what we want is to use the existing item stack and the loot context to return a new, modified item stack. However, almost all loot functions don't directly extend LootItemFunction, but extend LootItemConditionalFunction instead. This class has built-in functionality for applying loot conditions to the function - the function is only applied if the loot conditions apply. For the sake of example, let's apply a random enchantment with a specified level to the item:

// Code adapted from vanilla's EnchantRandomlyFunction class.
// LootItemConditionalFunction is an abstract class, not an interface, so we cannot use a record here.
public class RandomEnchantmentWithLevelFunction extends LootItemConditionalFunction {
    // Our context: an optional list of enchantments, and a level.
    private final Optional<HolderSet<Enchantment>> enchantments;
    private final int level;
    // Our codec.
    public static final MapCodec<RandomEnchantmentWithLevelFunction> CODEC =
            // #commonFields adds the conditions field.
            RecordCodecBuilder.mapCodec(inst -> commonFields(inst).and(inst.group(
                    RegistryCodecs.homogeneousList(Registries.ENCHANTMENT).optionalFieldOf("enchantments").forGetter(e -> e.enchantments),
                    Codec.INT.fieldOf("level").forGetter(e -> e.level)
            ).apply(inst, RandomEnchantmentWithLevelFunction::new));
    
    public RandomEnchantmentWithLevelFunction(List<LootItemCondition> conditions, Optional<HolderSet<Enchantment>> enchantments, int level) {
        super(conditions);
        this.enchantments = enchantments;
        this.level = level;
    }
    
    // Run our enchantment application logic. Most of this is copied from EnchantRandomlyFunction#run.
    @Override
    public ItemStack run(ItemStack stack, LootContext context) {
        RandomSource random = context.getRandom();
        List<Holder<Enchantment>> stream = this.enchantments
                .map(HolderSet::stream)
                .orElseGet(() -> context.getLevel().registryAccess().registryOrThrow(Registries.ENCHANTMENT).holders().map(Function.identity()))
                .filter(e -> e.value().canEnchant(stack))
                .toList();
        Optional<Holder<Enchantment>> optional = Util.getRandomSafe(list, random);
        if (optional.isEmpty()) {
            LOGGER.warn("Couldn't find a compatible enchantment for {}", stack);
        } else {
            if (stack.is(Items.BOOK)) {
                stack = new ItemStack(Items.ENCHANTED_BOOK);
            }
            stack.enchant(enchantment, Mth.nextInt(random, enchantment.value().getMinLevel(), enchantment.value().getMaxLevel()));
        }
        return stack;
    }
}

We can then register the function type to the registry using the function's codec:

public static final DeferredRegister<LootItemFunctionType<?>> LOOT_FUNCTION_TYPES =
        DeferredRegister.create(Registries.LOOT_FUNCTION_TYPE, ExampleMod.MOD_ID);

public static final Supplier<LootItemFunctionType<RandomEnchantmentWithLevelFunction>> RANDOM_ENCHANTMENT_WITH_LEVEL =
        LOOT_FUNCTION_TYPES.register("random_enchantment_with_level", () -> new LootItemFunctionType(RandomEnchantmentWithLevelFunction.CODEC));

After we have done that, we need to override #getType in our condition and return the registered type:

public class RandomEnchantmentWithLevelFunction extends LootItemConditionalFunction {
    // other stuff here

    @Override
    public LootItemFunctionType getType() {
        return RANDOM_ENCHANTMENT_WITH_LEVEL.get();
    }
}