Question about how to fix `InvalidGeoArrow` error with Wkt Array #1410

C-Loftus · 2025-12-11T02:26:37Z

C-Loftus
Dec 11, 2025

Thank you for your excellent work on geoarrow.

I am trying to construct a minimum viable example for making geoparquet with wkt
- (For what its worth, I'd be happy to contribute this as an example once I clean it up if that would help others)
My understanding is that to construct geoparquet you need to use a builder struct
- I see builder structs for all geometry types and wkb but not wkt https://docs.rs/geoarrow-array/latest/geoarrow_array/builder/index.html
- As such I am just using the generic GeometryBuilder class after having converting the wkt (although perhaps there is a better way)

All my code compiles fine but I get an error saying the following which I am confused on how to fix.

called `Result::unwrap()` on an `Err` value: InvalidGeoArrow("Only FixedSizeList, Struct, Binary, LargeBinary, BinaryView, String, LargeString, and StringView arrays are unambigously typed for a GeoArrow type and can be used without extension metadata.\nEnsure your array input has GeoArrow metadata.")

My Code / Context

Below is a minimum viable example for constructing geoparquet. However, I am unclear why it is failing.

I have tried using both Int32 and Utf8 for the id field.
I make sure I am using arrays with the same types as my schema
I am calling to_array_ref and not doing any casting on the geometry array

use std::sync::Arc;

use arrow_array::{self, ArrayRef, Int32Array, RecordBatch};
use geo_types::Geometry;
use geoarrow_array::{GeoArrowArray, builder::GeometryBuilder};
use geoarrow_schema::GeometryType;
use wkt::TryFromWkt;

use arrow_schema::{DataType::Int32, Field, SchemaBuilder};
use geoarrow_schema::GeoArrowType;
use geoparquet::writer::{GeoParquetRecordBatchEncoder, GeoParquetWriterOptionsBuilder};
use parquet::arrow::ArrowWriter;

pub fn new_parquet_creator(
    file_name: &str,
) -> (GeoParquetRecordBatchEncoder, ArrowWriter<std::fs::File>) {
    let options = GeoParquetWriterOptionsBuilder::default()
        .set_primary_column("geometry".to_string())
        .build();

    let mut schema_builder = SchemaBuilder::new();

    let geoarrow_type = GeoArrowType::Geometry(GeometryType::default());
    let geometry_field = geoarrow_type.to_field("geometry", false);
    schema_builder.push(geometry_field);

    let geoconnex_pid = Field::new("id", Int32, false);
    schema_builder.push(geoconnex_pid);

    let schema = schema_builder.finish();

    let gpq_encoder = GeoParquetRecordBatchEncoder::try_new(&schema, &options).unwrap();

    let output_file = std::fs::File::create(file_name).unwrap();
    let parquet_writer =
        ArrowWriter::try_new(output_file, gpq_encoder.target_schema(), None).unwrap();

    (gpq_encoder, parquet_writer)
}

fn main() {
    let (mut gpq_encoder, mut parquet_writer) = new_parquet_creator("test.parquet");

    let col = Arc::new(Int32Array::from_iter_values([2, 2, 4])) as ArrayRef;

    let geometry1 = Geometry::try_from_wkt_str("POINT (1 2)").unwrap();
    let geometry2 = Geometry::try_from_wkt_str("POINT (1 3)").unwrap();
    let geometry3 = Geometry::try_from_wkt_str("POINT (1 3)").unwrap();

    let generic_geometry_type = GeometryType::new(Default::default());
    let mut builder = GeometryBuilder::new(generic_geometry_type);
    builder.push_geometry(Some(&geometry1)).unwrap();
    builder.push_geometry(Some(&geometry2)).unwrap();
    builder.push_geometry(Some(&geometry3)).unwrap();
    let geom_array = builder.finish();

    let batch =
        RecordBatch::try_from_iter([("geometry", geom_array.to_array_ref()), ("id", col)]).unwrap();

    for batch in [batch] {
        let encoded_batch = gpq_encoder.encode_record_batch(&batch).unwrap();
        parquet_writer.write(&encoded_batch).unwrap();
    }

    let kv_metadata = gpq_encoder.into_keyvalue().unwrap();
    parquet_writer.append_key_value_metadata(kv_metadata);
    parquet_writer.finish().unwrap();
}

Answered by kylebarron

Dec 11, 2025

I'm not sure if this is the exact issue you're hitting, but you'll also hit similar issues with this:

    let batch =
        RecordBatch::try_from_iter([("geometry", geom_array.to_array_ref()), ("id", col)]).unwrap();

When you're doing this, you're saying to RecordBatch to infer the schema for you. The problem is that when you call to_array_ref, the Arc<dyn Array> loses all information about the extension type.

Instead, you need to always explicitly pass the schema with RecordBatch::try_new, where you build the schema yourself. And any GeoArrow fields should be created with, e.g. GeoArrowType::to_schema.

This is mostly a documentation issue but I'm not sure where else to document it. I …

View full answer

kylebarron · 2025-12-11T02:55:06Z

kylebarron
Dec 11, 2025
Maintainer

called Result::unwrap() on an Err value: InvalidGeoArrow("Only FixedSizeList, Struct, Binary, LargeBinary, BinaryView, String, LargeString, and StringView arrays are unambigously typed for a GeoArrow type and can be used without extension metadata.\nEnsure your array input has GeoArrow metadata.")

Can you say what line of code that's coming from?

0 replies

kylebarron · 2025-12-11T02:59:58Z

kylebarron
Dec 11, 2025
Maintainer

I'm not sure if this is the exact issue you're hitting, but you'll also hit similar issues with this:

    let batch =
        RecordBatch::try_from_iter([("geometry", geom_array.to_array_ref()), ("id", col)]).unwrap();

When you're doing this, you're saying to RecordBatch to infer the schema for you. The problem is that when you call to_array_ref, the Arc<dyn Array> loses all information about the extension type.

Instead, you need to always explicitly pass the schema with RecordBatch::try_new, where you build the schema yourself. And any GeoArrow fields should be created with, e.g. GeoArrowType::to_schema.

This is mostly a documentation issue but I'm not sure where else to document it. I have one such note on GeoArrowType::to_data_type

3 replies

C-Loftus Dec 11, 2025
Author

Thank you very much for your reply.

Apologies previously the error on the unwrap was coming from this line

        let encoded_batch = gpq_encoder.encode_record_batch(&batch).unwrap();

I have since switched to this

    let batch =
        RecordBatch::try_new(Arc::new(schema), vec![geom_array.to_array_ref(), col]).unwrap();

and it appears to work, thank you!

C-Loftus Dec 11, 2025
Author

Full code snippet is here for others' context

use std::sync::Arc;

use arrow_array::{self, ArrayRef, Int32Array, RecordBatch};
use geo_types::Geometry;
use geoarrow_array::{GeoArrowArray, builder::GeometryBuilder};
use geoarrow_schema::GeometryType;
use wkt::TryFromWkt;

use arrow_schema::{DataType::Int32, Field, Schema, SchemaBuilder};
use geoarrow_schema::GeoArrowType;
use geoparquet::writer::{GeoParquetRecordBatchEncoder, GeoParquetWriterOptionsBuilder};
use parquet::arrow::ArrowWriter;

pub fn new_parquet_creator(
    schema: &Schema,
    file_name: &str,
) -> (GeoParquetRecordBatchEncoder, ArrowWriter<std::fs::File>) {
    let options = GeoParquetWriterOptionsBuilder::default()
        .set_primary_column("geometry".to_string())
        .build();

    let gpq_encoder = GeoParquetRecordBatchEncoder::try_new(schema, &options).unwrap();

    let output_file = std::fs::File::create(file_name).unwrap();
    let parquet_writer =
        ArrowWriter::try_new(output_file, gpq_encoder.target_schema(), None).unwrap();

    (gpq_encoder, parquet_writer)
}

fn main() {
    let mut schema_builder = SchemaBuilder::new();

    let geoarrow_type = GeoArrowType::Geometry(GeometryType::default());
    let geometry_field = geoarrow_type.to_field("geometry", false);
    schema_builder.push(geometry_field);

    let geoconnex_pid = Field::new("id", Int32, false);
    schema_builder.push(geoconnex_pid);

    let schema = schema_builder.finish();

    let (mut gpq_encoder, mut parquet_writer) = new_parquet_creator(&schema, "test.parquet");

    let col = Arc::new(Int32Array::from_iter_values([2, 2, 4])) as ArrayRef;

    let geometry1 = Geometry::try_from_wkt_str("POINT (1 2)").unwrap();
    let geometry2 = Geometry::try_from_wkt_str("POINT (1 3)").unwrap();
    let geometry3 = Geometry::try_from_wkt_str("POINT (1 3)").unwrap();

    let generic_geometry_type = GeometryType::new(Default::default());
    let mut builder = GeometryBuilder::new(generic_geometry_type);
    builder.push_geometry(Some(&geometry1)).unwrap();
    builder.push_geometry(Some(&geometry2)).unwrap();
    builder.push_geometry(Some(&geometry3)).unwrap();
    let geom_array = builder.finish();

    let batch =
        RecordBatch::try_new(Arc::new(schema), vec![geom_array.to_array_ref(), col]).unwrap();

    for batch in [batch] {
        let encoded_batch = gpq_encoder.encode_record_batch(&batch).unwrap();
        parquet_writer.write(&encoded_batch).unwrap();
    }

    let kv_metadata = gpq_encoder.into_keyvalue().unwrap();
    parquet_writer.append_key_value_metadata(kv_metadata);
    parquet_writer.finish().unwrap();
}

kylebarron Dec 11, 2025
Maintainer

If you have ideas for how to better document that, I'm all ears. Feel free to make an issue or PR

Navigation Menu

Question about how to fix `InvalidGeoArrow` error with Wkt Array #1410

Uh oh!

{{title}}

Uh oh!

Replies: 2 comments 3 replies

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{title}}

Uh oh!

Select a reply

Uh oh!

Question about how to fix InvalidGeoArrow error with Wkt Array #1410

Uh oh!

C-Loftus Dec 11, 2025

My Code / Context

Replies: 2 comments · 3 replies

Uh oh!

kylebarron Dec 11, 2025 Maintainer

Uh oh!

kylebarron Dec 11, 2025 Maintainer

Uh oh!

C-Loftus Dec 11, 2025 Author

Uh oh!

C-Loftus Dec 11, 2025 Author

Uh oh!

kylebarron Dec 11, 2025 Maintainer

Question about how to fix `InvalidGeoArrow` error with Wkt Array #1410

C-Loftus
Dec 11, 2025

Replies: 2 comments 3 replies

kylebarron
Dec 11, 2025
Maintainer

kylebarron
Dec 11, 2025
Maintainer

C-Loftus Dec 11, 2025
Author

C-Loftus Dec 11, 2025
Author

kylebarron Dec 11, 2025
Maintainer