[cmollis]

this part is confusing to me in the doc.. I assume that you're using the httpfs (S3) extensions and perhaps doing scanning of the parquet files (which I think is actually streamed.. e.g. querying for a specific column values in a series of parquet files). We have a huge data set of hive-partitioned parquet files in s3 (e.g. /customerid/year/month/<series of parquet files>). Can i just scan these files using the glob pattern to retrieve data like I can with Athena? The extension doc seems to indicate that I can (from the doc: SELECT * FROM read_parquet('s3://bucket/*/file.parquet', HIVE_PARTITIONING = 1) where year=2013;) Or do I need to know which parquet files I'm looking for in S3 and bring them down to work on locally? If it's the former, then it seems equivalent to Athena..

[wenc]

No you can definitely use globs in DuckDB.

And no you don’t have to know the exact parquet file. You would treat the Hive partitioned data as a single dataset and DuckDB will scan it automatically. (Partition elimination, predicate pushdown etc all done automatically)

https://duckdb.org/docs/data/partitioning/hive_partitioning

[cmollis]

ok.. thanks.. I'll try it out. I can think of few use-case that we have where this might be a good alternative to athena.