TLDR Tech / Practical
Evaluating PostgreSQL's ability to read Iceberg and Parquet data
The author analyzes how four different suppliers' implementations of PostgreSQL read Apache Iceberg and Parquet data directly through DuckDB. The comparison highlights differences in query planning, execution pushdown, heap-to-lake joins, and transactional write capabilities among Aurora PostgreSQL, pg_lake, ColdFront, and pg_clickhouse. It is noted that when these advanced features are not available, PostgreSQL often resorts to planning queries without sufficient statistical data.
Read the full story