Parquet stats_min_value/stats_max_value do not respect timestamp configuration

[greaka]

What happens?

When reading parquet files and/or their metadata, timestamp column metadata statistics stats_min_value and stats_max_value seem to not respect the timestamps configuration of ms, resulting in where filters ignoring entire chunks.

To Reproduce

Example file using ms precision in stats:
2022-11-27T17 42 35.480941703Z.parquet.txt
select path_in_schema, stats_min_value, stats_max_value from 'file.parquet';
select * from 'file.parquet';
select * from 'file.parquet' where timestamp < '2022-01-01 00:00:00.000';
select * from 'file.parquet' where timestamp > '2022-01-01 00:00:00.000';
I've verified the raw values in the parquet file before filing this issue.

OS:

Linux-x64 & shell.duckdb.org

DuckDB Version:

0.6.0

DuckDB Client:

rust & wasm

Full Name:

Michl Steglich

Affiliation:

drf.rs

Have you tried this on the latest master branch?

• I agree

Have you tried the steps to reproduce? Do they include all relevant data and configuration? Does the issue you report still appear there?

• I agree

[Mytherin]

Thanks for the report! I have pushed a fix in #5540