Indexicon for Python

Indexicon provides NumPy-native Python bindings for high-performance R-trees, Quad-trees, Oct-trees, and KD-trees backed by C++.

Installation

Indexicon requires Python 3.10 or newer.

pip install indexicon

Example

Every index supports bulk_load, insert, remove, query, knn, and statistics.

import numpy as np
import indexicon as idx

tree = idx.RTree(data="point", dims=2)
coords = np.array([[0.0, 0.0], [1.0, 1.0], [2.0, 2.0]])
tree.bulk_load(coords)

tree.insert(np.array([3.0, 3.0]))
hits = tree.query(np.array([0.5, 0.5]), np.array([2.5, 2.5]))
near = tree.knn(np.array([0.1, 0.1]), k=2)

print(hits)
print(near)
print(tree.statistics())

More examples are available in the python/examples directory.

Supported Indexes

Quad-tree splitting strategies are "pr", "pseudo_median", and "longest_axis". KD-tree splitting strategies are "longest_axis", "adaptive", and "round_robin".

Python bindings use float64 coordinates, int64 IDs, and a fixed node or bucket capacity of 128. Capacity and other template parameters can be customized through the C++ API.

Object IDs

Use custom IDs by passing an (N,) NumPy array of int64 values to bulk_load(..., ids=ids) or an integer to insert(..., id=100). When IDs are omitted, Indexicon assigns them automatically. Automatic IDs advance past nonnegative custom IDs and are not recycled after removal.

Citation

If you use Indexicon in a project, paper, benchmark, or product, please cite the Indexicon paper.

License

Indexicon is distributed under the MIT License.