duckdb/duckdb_iceberg

binder error when directly scanning metadata file

Opened this issue · 2 comments

xiaom commented

Hey,

I am encountering an issue while trying to scan metadata files directly from an Iceberg table located in S3.

Here are the details:

  • Data Source: Iceberg table in S3, written by Flink with AWS Glue Catalog.
  • DuckDB Version: v0.9.1

Query

select * from  iceberg_scan('s3://<...>/metadata/00136-4330ee25-0db5-42a4-8230-035d739f76e7.metadata.json') limit 1;

Error:

Binder Error: Table "iceberg_scan_deletes" does not have a column named "file_path"

Would appreciate any insights or suggestions on what could be causing this issue. Thank you!

Fokko commented

Could you double-check the manifest file to see if the path is in there? This can be done easily using avro-tools tojson manifest-file.avro.

lamb-russell commented

i have a similar example in this notebook

key part is near the bottom
image