Add initial support for geodatafusion spatial filters. (#31)

* Convert bbox representation to wkb.

* Register geodatafusion to support spatial filtering.

* Update py03 bindings to datafusion 52.

* Align zarrs_icechunk compatible icechunk-python versions.

* Include geodatafusion spatial function registration tests.

* Update geodatafusion python to 0.2.0 for datafusion 52 compatability.

* Fix cargo fmt issues.

* Fix ruff error.

* Update test data with new VariableLengthBytes bbox array.

* Export full API from Python module.

* Update deprecated register_table_provider to register_table.

Author: sharkinsspatial

Cargo.lock

@@ -4,13 +4,14 @@ version = "0.1.0"
 edition = "2024"
 
 [dependencies]
-arrow = "56.2.0"
-arrow-array = "56.2.0"
-arrow-schema = "56.2.0"
+arrow = "57.0.0"
+arrow-array = "57.0.0"
+arrow-schema = "57.0.0"
 async-trait = "0.1.89"
-datafusion = "50.2"
+datafusion = "52.0"
 futures = "0.3.31"
-geoarrow-schema = "0.6.1"
+geodatafusion = "0.3"
+geoarrow-schema = "0.7.0"
 icechunk = "0.3.16"
 object_store = "0.12.4"
 thiserror = "2"
@@ -23,5 +24,5 @@ zarrs_object_store = "0.5.0"
 zarrs_storage = { version = "0.4.0", features = ["async"] }
 
 [dev-dependencies]
-geoarrow-array = "0.6.1"
+geoarrow-array = "0.7.0"
 tokio = { version = "1.48", features = ["macros", "rt-multi-thread"] }

Cargo.toml

@@ -11,9 +11,7 @@ Users can define arbitrary schemas where the 1-dimensional arrays each use a `dt
 - Inside a Zarr group named `"meta"`
     - A `datetime64[ms]` array named `"date"` with `n` timestamps named `"date"` with `n` timestamps.
     - A `VariableLengthUTF8` array named `"collection"` with `n` string values.
-    - A `VariableLengthUTF8` array named `"bbox"` with `n` string values, stored as a `VariableLengthUTF8` array, where each string is a WKT-encoded Polygon (or MultiPolygon) with the bounding box of that Zarr record.
-
-        In the future, we will likely use a binary encoding like WKB, but Zarr's binary dtype is [not currently well-specified](https://github.com/zarr-developers/zarr-python/issues/3517).
+    - A `VariableLengthBytes` array named `"bbox"` with `n` binary values, where each value is a WKB-encoded Polygon (or MultiPolygon) with the bounding box of that Zarr record.
 
 This data schema may change over time.
 

README.md

@@ -1 +1 @@
-{"snapshot":"CMNB1D4K0S56F8Z7KA00"}
+{"snapshot":"ZKWW9E6BW6YFNNDKESR0"}