clarify input (take 2)

README.md

@@ -1,5 +1,5 @@
 # sparse-cubes
-Marching cubes for `(N, 3)` voxel coordinates - i.e. the equivalent of a 3D
+Marching cubes for `(N, 3)` voxel indices - i.e. the equivalent of a 3D
 sparse matrix in COOrdinate format.
 
 Running marching cubes directly on sparse voxels is faster and importantly much
@@ -27,8 +27,10 @@ pip3 install git+https://github.com/navis-org/sparse-cubes.git
 ```python
 >>> import sparsecubes as sc
 >>> import numpy as np
+>>> # Indices for two adjacent voxels
 >>> voxel_xyz = np.array([[0, 0, 0],
-...                       [0, 0, 1]])
+...                       [0, 0, 1]],
+...                      dtype='uint32')
 >>> m = sc.marching_cubes(voxel_xyz)
 >>> m
 <trimesh.Trimesh(vertices.shape=(12, 3), faces.shape=(20, 3))>

sparsecubes/core.py

@@ -27,11 +27,11 @@ def marching_cubes(voxels, spacing=None, step_size=1):
 
     Parameters
     ----------
-    voxels :    (N, 3) array
-                XYZ voxel coordinates to find isosurfaces for. The data type
-                should carry over to the mesh vertices - i.e. if `voxels` are
-                unsigned 32-bit integers, the mesh vertex coordinates should
-                also end up being unsigned 32-bit integers.
+    voxels :    (N, 3) array of integers
+                XYZ voxel coordinates (indices) to find isosurfaces for. The
+                data type should carry over to the mesh vertices - i.e. if
+                `voxels` are unsigned 32-bit integers, the mesh vertex
+                coordinates should also end up being unsigned 32-bit integers.
     spacing :   length-3 tuple of floats, optional
                 Voxel spacing in spatial dimensions corresponding to numpy array
                 indexing dimensions as in `voxels`.