Hi All,
We have Lance dataset that has custom pyarrow types, as well as the fixed_shape_tensor arrow extension type. The columns that have the custom types don't really need to be queried from the trino connector at the moment, but it appears that the presence of them causes the entire query to fail even when only querying columns with supported types
trino:default> select COUNT("metadata/mount") from dataset;
Query 20260410_015339_00006_mabtz failed: Cannot invoke "io.trino.spi.connector.ConnectorTableMetadata.getTableSchema()" because the return value of "io.trino.spi.connector.ConnectorMetadata.getTableMetadata(io.trino.spi.connector.ConnectorSession, io.trino.spi.connector.ConnectorTableHandle)" is null
The trino server logs have the following errors:
2026-04-10T01:53:39.669Z INFO Query-20260410_015339_00006_mabtz-224 org.apache.arrow.c.SchemaImporter Unrecognized extension type: arrow.fixed_shape_tensor
I confirmed that if I create a dataset without custom types, everything works fine. Other than making a shallow clone of the datasets with the problematic columns dropped (our current workaround), is it possible to support this?
Hi All,
We have Lance dataset that has custom pyarrow types, as well as the fixed_shape_tensor arrow extension type. The columns that have the custom types don't really need to be queried from the trino connector at the moment, but it appears that the presence of them causes the entire query to fail even when only querying columns with supported types
The trino server logs have the following errors:
I confirmed that if I create a dataset without custom types, everything works fine. Other than making a shallow clone of the datasets with the problematic columns dropped (our current workaround), is it possible to support this?