feat: Add ST_LineMerge

.github/workflows/ci.yml

@@ -23,6 +23,10 @@ jobs:
         with:
           components: rustfmt, clippy
       - uses: Swatinem/rust-cache@v2
+        - name: Install dependencies
+          run: |
+            sudo apt-get update
+            sudo apt-get install -y libgeos-dev
       - name: Clippy
         run: cargo clippy --all-features --tests -- -D warnings
       - name: Check

Cargo.toml

@@ -39,6 +39,7 @@ geodatafusion = { path = "rust/geodatafusion", version = "0.4.0" }
 geohash = "0.13.1"
 geojson = "0.24"
 geoparquet = "0.8.0"
+geos = { version = "11.1", default-features = false }
 http-range-client = { version = "0.9", default-features = false }
 object_store = "0.13.2"
 serde_json = "1"

DEVELOP.md

@@ -10,6 +10,7 @@ This is a Rust workspace with multiple crates:
     - Internally, each _provider_ of functions is organized in submodules:
         - `native/` - Operations that are natively implemented, without the use of other dependencies like `geo`
         - `geo/` - Operations implemented using the `geo` crate
+        - `geos/` - Operations implemented using the `geos` crate (bindings to the native GEOS library), gated behind the optional `geos` feature
         - `geohash/` - GeoHash encoding/decoding, using the `geohash` crate
 - `rust/geodatafusion-flatgeobuf` - FlatGeobuf format support
 - `rust/geodatafusion-geoparquet` - GeoParquet format support
@@ -104,8 +105,31 @@ Functions are organized by category in `rust/geodatafusion/src/udf/`:
 - `geo/processing/` - Processing functions (ST_Buffer, ST_Simplify)
 - `geo/relationships/` - Spatial relationships (ST_Intersects, etc.)
 - `geo/validation/` - Validation functions (ST_IsValid)
+- `geos/processing/` - GEOS-backed processing functions (ST_LineMerge)
 - `geohash/` - GeoHash functions
 
+### GEOS-backed functions
+
+Functions in the `geos/` provider link the [GEOS](https://libgeos.org/) library, and can be controlled via Cargo features.
+
+The version of GEOS required varies by UDF,
+so we expose cargo features with explicit minimum version numbers,
+like `geos-3_11` (GEOS >=3.11).
+
+#### GEOS linking
+
+This crate will try to link the system-provided GEOS library by default.
+If you want to statically link GEOS from a source build,
+add a direct dependency on the `geos` crate to your project,
+and enable the `static` feature like so:
+
+```toml
+geos = { version = "11.1.1", features = ["static"] }
+```
+
+If you're using cargo workspaces, make sure that
+the crates using geodatafusion directly reference `geos`.
+
 ### Code Style
 
 - Use meaningful variable and function names

README.md

@@ -297,7 +297,7 @@ Functions are explicitly modeled after the [PostGIS API](https://postgis.net/doc
 | ST_FilterByM                |             | Removes vertices based on their M value                                                           |
 | ST_GeneratePoints           |             | Generates a multipoint of random points contained in a Polygon or MultiPolygon.                   |
 | ST_GeometricMedian          |             | Returns the geometric median of a MultiPoint.                                                     |
-| ST_LineMerge                |             | Return the lines formed by sewing together a MultiLineString.                                     |
+| ST_LineMerge                | ✅          | Return the lines formed by sewing together a MultiLineString.                                     |
 | ST_MaximumInscribedCircle   |             | Computes the largest circle contained within a geometry.                                          |
 | ST_LargestEmptyCircle       |             | Computes the largest circle not overlapping a geometry.                                           |
 | ST_MinimumBoundingCircle    |             | Returns the smallest circle polygon that contains a geometry.                                     |

rust/geodatafusion/Cargo.toml

@@ -10,6 +10,14 @@ categories = { workspace = true }
 rust-version = { workspace = true }
 
 
+[features]
+# Enables GEOS-backed UDFs.
+# We only use explicitly versioned flags,
+# and the structure should be recursive
+# (with higher version targets implying lower version ones
+# in a chain.)
+geos-3_11 = ["geos/v3_11_0"]
+
 [dependencies]
 arrow-arith = { workspace = true }
 arrow-array = { workspace = true }
@@ -21,6 +29,7 @@ geoarrow-array = { workspace = true }
 geoarrow-expr-geo = { workspace = true }
 geoarrow-schema = { workspace = true }
 geohash = { workspace = true }
+geos = { workspace = true, optional = true }
 thiserror = { workspace = true }
 wkt = { workspace = true }
 

rust/geodatafusion/src/error.rs

@@ -19,6 +19,10 @@ pub(crate) enum GeoDataFusionError {
     #[error(transparent)]
     GeoArrow(#[from] GeoArrowError),
 
+    #[cfg(feature = "geos")]
+    #[error(transparent)]
+    Geos(#[from] geos::Error),
+
     #[error(transparent)]
     GeoHash(#[from] geohash::GeohashError),
 }
@@ -32,6 +36,8 @@ impl From<GeoDataFusionError> for DataFusionError {
             GeoDataFusionError::Arrow(err) => DataFusionError::ArrowError(Box::new(err), None),
             GeoDataFusionError::DataFusion(err) => err,
             GeoDataFusionError::GeoArrow(err) => DataFusionError::External(Box::new(err)),
+            #[cfg(feature = "geos")]
+            GeoDataFusionError::Geos(err) => DataFusionError::External(Box::new(err)),
             GeoDataFusionError::GeoHash(err) => DataFusionError::External(Box::new(err)),
         }
     }

rust/geodatafusion/src/lib.rs

@@ -20,6 +20,9 @@ pub fn register(session_context: &datafusion::prelude::SessionContext) {
 
     crate::udf::geo::validation::register(session_context);
 
+    #[cfg(feature = "geos-3_11")]
+    crate::udf::geos::processing::register(session_context);
+
     crate::udf::geohash::register(session_context);
 
     crate::udf::native::accessors::register(session_context);

rust/geodatafusion/src/udf/geos/mod.rs

@@ -0,0 +1,3 @@
+//! UDFs implemented via GEOS bindings.
+
+pub mod processing;