Add KDTreeMetadata to Python API

python/DEVELOP.md

@@ -0,0 +1,7 @@
+To run the docs locally:
+
+```
+uv sync --no-install-package geoindex-rs
+uv run --no-project maturin develop
+uv run --no-project mkdocs serve
+```

python/python/geoindex_rs/kdtree.pyi

@@ -35,15 +35,16 @@ def range(
     Results are the insertion indexes of items that match the query.
 
     Args:
-        index: the KDTree to search
-        min_x: The `min_x` coordinate of the query bounding box
-        min_y: The `min_y` coordinate of the query bounding box
-        max_x: The `max_x` coordinate of the query bounding box
-        max_y: The `max_y` coordinate of the query bounding box
+        index: the KDTree to search.
+        min_x: The `min_x` coordinate of the query bounding box.
+        min_y: The `min_y` coordinate of the query bounding box.
+        max_x: The `max_x` coordinate of the query bounding box.
+        max_y: The `max_y` coordinate of the query bounding box.
 
     Returns:
-        An Arrow array with the insertion indexes of query results.
+        A uint32-typed Arrow array with the insertion indexes of query results.
     """
+
 def within(
     index: IndexLike,
     qx: int | float,
@@ -55,13 +56,13 @@ def within(
     Results are the insertion indexes of items that match the query.
 
     Args:
-        index: the KDTree to search
-        qx: The `x` coordinate of the query point
-        qy: The `y` coordinate of the query point
+        index: the KDTree to search.
+        qx: The `x` coordinate of the query point.
+        qy: The `y` coordinate of the query point.
         r: The radius from the query point to use for searching.
 
     Returns:
-        An Arrow array with the insertion indexes of query results.
+        A uint32-typed Arrow array with the insertion indexes of query results.
     """
 
 class KDTreeBuilder:
@@ -149,3 +150,49 @@ class KDTree(Buffer):
     object.
     """
     def __repr__(self) -> str: ...
+
+class KDTreeMetadata:
+    """Common metadata to describe a KDTree.
+
+    This can be used to know the number of items, node information, or total byte size
+    of a KDTree.
+
+    Additionally, this can be used to know how much memory a KDTree **would use** with
+    the given number of items and node size. A KDTree with 1 million items and a node
+    size of 64 (the default) would take up 20 MiB.
+
+    ```py
+    from geoindex_rs import kdtree as kd
+
+    metadata = kd.KDTreeMetadata(num_items=1_000_000, node_size=64)
+    assert metadata.num_bytes == 20_000_008
+    ```
+    """
+
+    def __init__(
+        self,
+        num_items: int,
+        node_size: int = 64,
+        coord_type: Literal["float32", "float64"] = "float64",
+    ) -> None:
+        """Create a new KDTreeMetadata given a number of items and node size.
+
+        Args:
+            num_items: The number of items in the tree
+            node_size: The node size of the tree. Defaults to 16.
+            coord_type: The coordinate type to use in the tree. Currently only float32
+                and float64 are permitted. Defaults to None.
+        """
+    @classmethod
+    def from_index(cls, index: IndexLike) -> KDTreeMetadata:
+        """Create from an existing KDTree buffer."""
+    def __repr__(self) -> str: ...
+    @property
+    def num_items(self) -> int:
+        """The number of items indexed in the tree."""
+    @property
+    def node_size(self) -> int:
+        """The maximum number of items per node."""
+    @property
+    def num_bytes(self) -> int:
+        """The number of bytes that a KDTree with this metadata would have."""

python/python/geoindex_rs/rtree.pyi

@@ -36,6 +36,7 @@ def boxes_at_level(index: IndexLike, level: int) -> Array:
             The returned array is a a zero-copy view from Rust. Note that it will keep
             the entire index memory alive until the returned array is garbage collected.
     """
+
 def tree_join(
     left: IndexLike,
     right: IndexLike,
@@ -150,6 +151,7 @@ def partitions(index: IndexLike) -> RecordBatch:
             index. Therefore, the `indices` array will have type `uint16` if the tree
             has fewer than 16,384 items; otherwise it will have type `uint32`.
     """
+
 def partition_boxes(index: IndexLike) -> RecordBatch:
     """Extract the geometries of the spatial partitions from an RTree.
 
@@ -174,6 +176,7 @@ def partition_boxes(index: IndexLike) -> RecordBatch:
             data. The `partition_id` column will be `uint16` type if there are less than
             65,536 partitions; otherwise it will be `uint32` type.
     """
+
 def search(
     index: IndexLike,
     min_x: int | float,
@@ -199,7 +202,7 @@ def search(
 class RTreeMetadata:
     """Common metadata to describe an RTree.
 
-    This can be used to know the number of items, node informatino, or total byte size
+    This can be used to know the number of items, node information, or total byte size
     of an RTree.
 
     Additionally, this can be used to know how much memory an RTree **would use** with
@@ -335,7 +338,9 @@ class RTreeBuilder:
         It's important to add _arrays_ at a time. This should usually not be called in a loop.
 
         Args:
-            min_x: array-like input
+            min_x: array-like input. If this is the only provided input, it should
+                represent the entire bounding box, as described above. Otherwise, pass
+                four separate parameters.
             min_y: array-like input. Defaults to None.
             max_x: array-like input. Defaults to None.
             max_y: array-like input. Defaults to None.
@@ -347,7 +352,10 @@ class RTreeBuilder:
         """Sort the internal index and convert this class to an RTree instance.
 
         Args:
-            method: The method used for sorting the RTree. `"hilbert"` will use a [Hilbert Curve](https://en.wikipedia.org/wiki/Hilbert_R-tree#Packed_Hilbert_R-trees) for sorting; `"str"` will use the [Sort-Tile-Recursive](https://ia600900.us.archive.org/27/items/nasa_techdoc_19970016975/19970016975.pdf) algorithm. Defaults to `"hilbert"`.
+            method: The method used for sorting the RTree. Defaults to `"hilbert"`.
+
+                - `"hilbert"` will use a [Hilbert Curve](https://en.wikipedia.org/wiki/Hilbert_R-tree#Packed_Hilbert_R-trees) for sorting.
+                - `"str"` will use the [Sort-Tile-Recursive](https://ia600900.us.archive.org/27/items/nasa_techdoc_19970016975/19970016975.pdf) algorithm.
 
         Returns:
             An immutable RTree instance, which can be used for spatial queries.

python/src/kdtree/metadata.rs

@@ -0,0 +1,101 @@
+use geo_index::kdtree::{KDTreeIndex, KDTreeMetadata, DEFAULT_KDTREE_NODE_SIZE};
+use pyo3::prelude::*;
+use pyo3::types::PyType;
+
+use crate::coord_type::CoordType;
+use crate::kdtree::input::PyKDTreeRef;
+
+pub(crate) enum PyKDTreeMetadataInner {
+    Float32(KDTreeMetadata<f32>),
+    Float64(KDTreeMetadata<f64>),
+}
+
+impl PyKDTreeMetadataInner {
+    fn node_size(&self) -> u16 {
+        match self {
+            Self::Float32(meta) => meta.node_size(),
+            Self::Float64(meta) => meta.node_size(),
+        }
+    }
+
+    fn num_items(&self) -> u32 {
+        match self {
+            Self::Float32(meta) => meta.num_items(),
+            Self::Float64(meta) => meta.num_items(),
+        }
+    }
+
+    fn num_bytes(&self) -> usize {
+        match self {
+            Self::Float32(meta) => meta.data_buffer_length(),
+            Self::Float64(meta) => meta.data_buffer_length(),
+        }
+    }
+}
+
+#[pyclass(name = "KDTreeMetadata")]
+pub struct PyKDTreeMetadata(PyKDTreeMetadataInner);
+
+#[pymethods]
+impl PyKDTreeMetadata {
+    #[new]
+    #[pyo3(signature = (num_items, node_size = DEFAULT_KDTREE_NODE_SIZE, coord_type = None))]
+    fn new(num_items: u32, node_size: u16, coord_type: Option<CoordType>) -> Self {
+        let coord_type = coord_type.unwrap_or(CoordType::Float64);
+        match coord_type {
+            CoordType::Float32 => Self(PyKDTreeMetadataInner::Float32(KDTreeMetadata::<f32>::new(
+                num_items, node_size,
+            ))),
+            CoordType::Float64 => Self(PyKDTreeMetadataInner::Float64(KDTreeMetadata::<f64>::new(
+                num_items, node_size,
+            ))),
+        }
+    }
+
+    #[classmethod]
+    fn from_index(_cls: &Bound<PyType>, index: PyKDTreeRef) -> PyResult<Self> {
+        match index {
+            PyKDTreeRef::Float32(tree) => {
+                Ok(Self(PyKDTreeMetadataInner::Float32(*tree.metadata())))
+            }
+            PyKDTreeRef::Float64(tree) => {
+                Ok(Self(PyKDTreeMetadataInner::Float64(*tree.metadata())))
+            }
+        }
+    }
+
+    fn __repr__(&self) -> String {
+        format!(
+            "KDTreeMetadata(num_items={}, node_size={})",
+            self.0.num_items(),
+            self.0.node_size()
+        )
+    }
+
+    #[getter]
+    fn node_size(&self) -> u16 {
+        self.0.node_size()
+    }
+
+    #[getter]
+    fn num_items(&self) -> u32 {
+        self.0.num_items()
+    }
+
+    #[getter]
+    fn num_bytes(&self) -> usize {
+        self.0.num_bytes()
+    }
+}
+
+impl From<KDTreeMetadata<f32>> for PyKDTreeMetadata {
+    fn from(value: KDTreeMetadata<f32>) -> Self {
+        Self(PyKDTreeMetadataInner::Float32(value))
+    }
+}
+
+impl From<KDTreeMetadata<f64>> for PyKDTreeMetadata {
+    fn from(value: KDTreeMetadata<f64>) -> Self {
+        Self(PyKDTreeMetadataInner::Float64(value))
+    }
+}

python/src/kdtree/mod.rs

@@ -1,5 +1,6 @@
 mod builder;
 mod input;
+mod metadata;
 mod range;
 mod within;
 
@@ -19,6 +20,7 @@ pub fn register_kdtree_module(
 
     child_module.add_class::<builder::PyKDTree>()?;
     child_module.add_class::<builder::PyKDTreeBuilder>()?;
+    child_module.add_class::<metadata::PyKDTreeMetadata>()?;
     child_module.add_wrapped(wrap_pyfunction!(range::range))?;
     child_module.add_wrapped(wrap_pyfunction!(within::within))?;
 

python/src/rtree/metadata.rs

@@ -39,7 +39,7 @@ impl PyRTreeMetadataInner {
         }
     }
 
-    fn data_buffer_length(&self) -> usize {
+    fn num_bytes(&self) -> usize {
         match self {
             Self::Float32(meta) => meta.data_buffer_length(),
             Self::Float64(meta) => meta.data_buffer_length(),
@@ -57,12 +57,12 @@ impl PyRTreeMetadata {
     fn new(num_items: u32, node_size: u16, coord_type: Option<CoordType>) -> Self {
         let coord_type = coord_type.unwrap_or(CoordType::Float64);
         match coord_type {
-            CoordType::Float32 => Self(PyRTreeMetadataInner::Float32(
-                geo_index::rtree::RTreeMetadata::<f32>::new(num_items, node_size),
-            )),
-            CoordType::Float64 => Self(PyRTreeMetadataInner::Float64(
-                geo_index::rtree::RTreeMetadata::<f64>::new(num_items, node_size),
-            )),
+            CoordType::Float32 => Self(PyRTreeMetadataInner::Float32(RTreeMetadata::<f32>::new(
+                num_items, node_size,
+            ))),
+            CoordType::Float64 => Self(PyRTreeMetadataInner::Float64(RTreeMetadata::<f64>::new(
+                num_items, node_size,
+            ))),
         }
     }
 
@@ -113,7 +113,7 @@ impl PyRTreeMetadata {
 
     #[getter]
     fn num_bytes(&self) -> usize {
-        self.0.data_buffer_length()
+        self.0.num_bytes()
     }
 }