Fix water/lava diggable by detecting fluid blocks from code

[AnonymoDGH]

Fixes water/lava being marked diggable: true.

The bug

The diggable logic is hardness != -1.0f && !(block instanceof AirBlock). Fluids have a hardness of 100 (not -1), so water and lava were incorrectly marked diggable in every version that has fluid blocks.

The fix

Detect fluid blocks from the game code instead of hardcoding names — per @extremeheat's guidance on #79:

• 1.13+ (Yarn & Mojmap): defaultState.getFluidState().isEmpty() — a block is diggable only if its default state has no fluid.
• 1.7–1.12.2 (legacyYarn): Material.isFluid() — !block.getMaterial().isFluid() for 1.7/1.8.9, !defaultState.getMaterial().isFluid() for 1.9.4–1.12.2.

No block names are hardcoded; the check queries the actual Minecraft source.

Verification

Every API was confirmed against the real mapping files for its era (Yarn v2 tiny mappings for 1.14/1.15/1.16/1.20, legacyYarn for 1.7–1.13, Mojmap for 1.21.5+):

• BlockState.getFluidState() (no-arg) exists from 1.13 onward (on AbstractBlock.AbstractBlockState, inherited by BlockState).
• FluidState.isEmpty() exists in all those versions.
• Material.isFluid() exists in all legacyYarn versions.

CI runs runServer for all 28 versions, which compiles and executes each generator, so any wrong API will fail the build.

[AnonymoDGH]

Verified against the freshly generated artifacts — diffing the PR branch output against main for all 28 versions:

diggable flips (the fix):

• water, lava: true → false in every version
• flowing_water, flowing_lava: true → false in 1.7–1.12.2
• Exactly those blocks — nothing else flipped (no seagrass/kelp/corals/sea_pickle/conduit/bubble_column)

example diffs (all other versions identical in shape):

1.13.2: water true->false, lava true->false   (bubble_column stays true)
1.18:   water true->false, lava true->false
26.1:   water true->false, lava true->false

Only other difference across all versions: the drops field on 122 blocks in various versions — runtime non-determinism from loot-table evaluation at server boot (differs between independent CI runs on main too); unrelated to this change.

CI: all 28 Build jobs green. Also confirmed the 1.13.2 class net/minecraft/class_3710 (unmapped in legacyYarn) resolves in the intermediary fallback namespace and is exactly FlowingFluidBlock: it extends Block, implements FluidDrainable, holds a FlowableFluid, and has getFluidState/bucket methods — checked against the 1.13.2 server jar.

[extremeheat]

Can you open PR in minecraft-data with the updated data?

[rom1504]

Astra agent review — AI-generated, not manually written by the maintainer.

At 4e3659d, this looks ready for maintainer merge review in the fluid-diggability scope. The current implementation distinguishes actual fluid blocks using the version’s native class/material APIs, rather than excluding every water-containing solid block. The author’s exact-head generated comparison reports only the water/lava variants changing, with all 28 generation jobs passing.

Coordinate the requested data update #1237 so users receive the corrected facts. I read the producer variants and companion data; I did not rerun all native generators or claim the older description of getFluidState().isEmpty() describes the final code.

Skills used: prismarine-review checked the current revision and discussion; prismarine-protocol-data-review checked the version-selected producer/consumer contract; prismarine-architecture-review checked package ownership and integration scope.