[DRILL-7797] Decimals are wrongly read from parquet files

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.18.0
• Component/s: Storage - Parquet
• Assignee: Unassigned
• Reporter: Jan Soubusta
• Created: 27/Oct/20 18:52
• Updated: 24/Jun/21 05:28
• Resolved: 23/Jun/21 15:32

Description

My setup:
Docker embedded Drill 1.18 (latest).

Parquet decimals are wrongly read by Drill, wrong huge / negative values are displayed.

Example small public file: 
https://gdc-tiger-test-data-eu-central.s3.eu-central-1.amazonaws.com/other_files/tpch/supplier.parquet/a51ab8fd-v_verticadb_node0001-140644911675136-0.parquet

The file was exported from Vertica database using EXPORT TO PARQUET statement.
My colleague utilizes his parquet reader written in C++ and this is his comment:

6th column S_ACCTBAL has type FIXED_LEN_BYTE_ARRAY and convertedType DECIMAL with scale 2 and precision 15.
I would say, that it is correctly conforming specification.