Completely change biome code again

Turns out my previous grasp of how Minecraft does this was wrong.

This seems to be the correct way. One side effect is that biome data
now has less resolution. One only really notices this when looking at
water, for which Minecraft does not even use the water colours for
in-game, otherwise I can't really tell a big difference.

Fixes #1698.

overviewer_core/biome.py

@@ -14,31 +14,17 @@
 #    You should have received a copy of the GNU General Public License along
 #    with the Overviewer.  If not, see <http://www.gnu.org/licenses/>.
 
-import numpy
 
-class BiomeDispensary:
-    """Turns biome arrays of either 256 or 1024 integer values into 16x16 2d arrays,
-    which can then be retrieved for any Y level with get_biome.
-    """
-    def __init__(self, biome_array):
-        self.biome_len = len(biome_array)
-        if self.biome_len == 256:
-            self.biomes = [biome_array.reshape((16, 16))]
-        elif self.biome_len == 1024:
-            self.biomes = [None] * 16
-            # Each value in the biome array actually stands for 4 blocks, so we take the
-            # Kronecker product of it to "scale" it into its full size of 4096 entries
-            krond = numpy.kron(biome_array, numpy.ones((4)))
-            for i in range(0, 16):
-                # Now we can divide it into 16x16 slices
-                self.biomes[i] = krond[i * 256:(i + 1) * 256].reshape((16, 16))
-
-
-    def get_biome(self, y_level):
-        if self.biome_len == 256 or y_level < 0:
-            return self.biomes[0]
-        else:
-            # We clamp the value to a max of 15 here because apparently Y=16
-            # also exists, and while I don't know what biome level Mojang uses for
-            # that, the highest one is probably a good bet.
-            return self.biomes[min(y_level, 15)]
+def reshape_biome_data(biome_array):
+    biome_len = len(biome_array)
+    if biome_len == 256:
+        return biome_array.reshape((16, 16))
+    elif biome_len == 1024:
+        # Ok here's the big brain explanation:
+        # Minecraft's new biomes have a resolution of 4x4x4 blocks.
+        # This means for a 16x256x16 chunk column we get 64 times for the vertical,
+        # and 4x4 values for the horizontals.
+        # Minecraft Wiki says some dumb thing about how "oh it's ordered by Z, then X, then Y",
+        # but they appear to either be wrong or have explained it with the eloquence of a
+        # caveman.
+        return biome_array.reshape((4, 64, 4))

overviewer_core/src/iterate.c

@@ -109,12 +109,10 @@ static inline void load_chunk_section(ChunkData* dest, int32_t i, PyObject* sect
     dest->sections[i].data = (PyArrayObject*)PyDict_GetItemString(section, "Data");
     dest->sections[i].skylight = (PyArrayObject*)PyDict_GetItemString(section, "SkyLight");
     dest->sections[i].blocklight = (PyArrayObject*)PyDict_GetItemString(section, "BlockLight");
-    dest->sections[i].biomes = (PyArrayObject*)PyDict_GetItemString(section, "Biomes");
     Py_INCREF(dest->sections[i].blocks);
     Py_INCREF(dest->sections[i].data);
     Py_INCREF(dest->sections[i].skylight);
     Py_INCREF(dest->sections[i].blocklight);
-    Py_INCREF(dest->sections[i].biomes);
 }
 
 /* loads the given chunk into the chunks[] array in the state
@@ -132,12 +130,13 @@ bool load_chunk(RenderState* state, int32_t x, int32_t z, uint8_t required) {
     if (dest->loaded)
         return false;
 
+    /* set reasonable defaults */
+    dest->biomes = NULL;
     for (i = 0; i < SECTIONS_PER_CHUNK; i++) {
         dest->sections[i].blocks = NULL;
         dest->sections[i].data = NULL;
         dest->sections[i].skylight = NULL;
         dest->sections[i].blocklight = NULL;
-        dest->sections[i].biomes = NULL;
     }
     dest->loaded = 1;
 
@@ -167,6 +166,9 @@ bool load_chunk(RenderState* state, int32_t x, int32_t z, uint8_t required) {
         return true;
     }
 
+    dest->biomes = (PyArrayObject*)PyDict_GetItemString(chunk, "Biomes");
+    Py_INCREF(dest->biomes);
+    dest->new_biomes = PyObject_IsTrue(PyDict_GetItemString(chunk, "NewBiomes"));
 
     for (i = 0; i < PySequence_Fast_GET_SIZE(sections); i++) {
         PyObject* ycoord = NULL;
@@ -193,12 +195,12 @@ unload_all_chunks(RenderState* state) {
     for (i = 0; i < 3; i++) {
         for (j = 0; j < 3; j++) {
             if (state->chunks[i][j].loaded) {
+                Py_XDECREF(state->chunks[i][j].biomes);
                 for (k = 0; k < SECTIONS_PER_CHUNK; k++) {
                     Py_XDECREF(state->chunks[i][j].sections[k].blocks);
                     Py_XDECREF(state->chunks[i][j].sections[k].data);
                     Py_XDECREF(state->chunks[i][j].sections[k].skylight);
                     Py_XDECREF(state->chunks[i][j].sections[k].blocklight);
-                    Py_XDECREF(state->chunks[i][j].sections[k].biomes);
                 }
                 state->chunks[i][j].loaded = 0;
             }

overviewer_core/src/overviewer.h

@@ -29,9 +29,9 @@
 
 #define NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION
 
-// increment this value if you've made a change to the c extesion
+// increment this value if you've made a change to the c extension
 // and want to force users to rebuild
-#define OVERVIEWER_EXTENSION_VERSION 78
+#define OVERVIEWER_EXTENSION_VERSION 79
 
 #include <stdbool.h>
 #include <stdint.h>
@@ -84,10 +84,12 @@ typedef struct {
     int32_t loaded;
     /* chunk biome array */
     PyArrayObject* biomes;
+    /* whether this is a 3d biome array */
+    bool new_biomes;
     /* all the sections in a given chunk */
     struct {
         /* all there is to know about each section */
-        PyArrayObject *blocks, *data, *skylight, *blocklight, *biomes;
+        PyArrayObject *blocks, *data, *skylight, *blocklight;
     } sections[SECTIONS_PER_CHUNK];
 } ChunkData;
 typedef struct {
@@ -210,16 +212,21 @@ static inline uint32_t get_data(RenderState* state, DataType type, int32_t x, in
         data_array = state->chunks[chunkx][chunkz].sections[chunky].skylight;
         break;
     case BIOMES:
-        data_array = state->chunks[chunkx][chunkz].sections[chunky].biomes;
+        data_array = state->chunks[chunkx][chunkz].biomes;
     };
 
     if (data_array == NULL)
         return def;
 
     if (type == BLOCKS)
         return getArrayShort3D(data_array, x, y, z);
-    if (type == BIOMES)
-        return getArrayByte2D(data_array, x, z);
+    if (type == BIOMES) {
+        if (state->chunks[chunkx][chunkz].new_biomes) {
+            return getArrayByte3D(data_array, x / 4, y / 4, z / 4);
+        } else {
+            return getArrayByte2D(data_array, x, z);
+        }
+    }
     return getArrayByte3D(data_array, x, y, z);
 }
 

overviewer_core/world.py

@@ -26,7 +26,7 @@
 
 from . import nbt
 from . import cache
-from .biome import BiomeDispensary
+from .biome import reshape_biome_data
 
 """
 This module has routines for extracting information about available worlds
@@ -1372,23 +1372,21 @@ def get_chunk(self, x, z):
                 biomes = numpy.frombuffer(biomes, dtype=numpy.uint8)
             else:
                 biomes = numpy.asarray(biomes)
-            #biomes = biomes.reshape((16,16))
-            biome_giver = BiomeDispensary(biomes)
+            biomes = reshape_biome_data(biomes)
         else:
             # Worlds converted by Jeb's program may be missing the Biomes key.
             # Additionally, 19w09a worlds have an empty array as biomes key
             # in some cases.
-            #biomes = numpy.zeros((16, 16), dtype=numpy.uint8)
-            biome_giver = BiomeDispensary(numpy.zeros(256, dtype=numpy.uint8))
-        #chunk_data['Biomes'] = biomes
+            biomes = numpy.zeros((16, 16), dtype=numpy.uint8)
+        chunk_data['Biomes'] = biomes
+        chunk_data['NewBiomes'] = (len(biomes.shape) == 3)
 
         unrecognized_block_types = {}
         for section in chunk_data['Sections']:
 
             # Turn the skylight array into a 16x16x16 matrix. The array comes
             # packed 2 elements per byte, so we need to expand it.
             try:
-                section['Biomes'] = biome_giver.get_biome(section["Y"])
                 if 'SkyLight' in section:
                     skylight = numpy.frombuffer(section['SkyLight'], dtype=numpy.uint8)
                     skylight = skylight.reshape((16,16,8))
@@ -1612,9 +1610,6 @@ def get_chunk(self, x, z):
         for section in chunk_data['Sections']:
             section = dict(section)
             newsections.append(section)
-            biomes = numpy.swapaxes(section['Biomes'], 0, 1)
-            biomes = numpy.rot90(biomes, self.north_dir)
-            section['Biomes'] = numpy.swapaxes(biomes, 0, 1)
             for arrayname in ['Blocks', 'Data', 'SkyLight', 'BlockLight']:
                 array = section[arrayname]
                 # Since the anvil change, arrays are arranged with axes Y,Z,X
@@ -1626,7 +1621,15 @@ def get_chunk(self, x, z):
                 section[arrayname] = array
         chunk_data['Sections'] = newsections
 
-        # same as above, for biomes (Z/X indexed)
+        if chunk_data['NewBiomes']:
+            array = numpy.swapaxes(chunk_data['Biomes'], 0, 2)
+            array = numpy.rot90(array, self.north_dir)
+            chunk_data['Biomes'] = numpy.swapaxes(array, 0, 2)
+        else:
+            # same as above, for biomes (Z/X indexed)
+            biomes = numpy.swapaxes(chunk_data['Biomes'], 0, 1)
+            biomes = numpy.rot90(biomes, self.north_dir)
+            chunk_data['Biomes'] = numpy.swapaxes(biomes, 0, 1)
         return chunk_data
 
     def get_chunk_mtime(self, x, z):