Database Technology

The PostgreSQL global development group has released REL_19_RC1, marking the first release candidate for version 19. This build serves as the final candidate before the stable release, indicating that core functionality is feature-complete and entering the stabilization phase. It is available for download via the official GitHub releases page.

• RC1 signals feature freeze; only critical bug fixes are expected before final release.
• Downloadable directly from the official PostgreSQL GitHub releases repository.
• Suitable for testing major upgrade paths and verifying compatibility with new features.
• Scheduled for publication on 2026-10-12.

Christophe Pettus discusses using the post_auth_delay and pre_auth_delay GUCs to pause backend execution during connection establishment. These settings allow engineers to inspect the process state before or after authentication completes, which is critical for debugging complex connection issues. The approach provides a controlled way to analyze backend behavior without relying on external tracing tools.

• Use post_auth_delay to pause execution after successful authentication for state inspection.
• Use pre_auth_delay to pause execution before authentication begins for early debugging.
• These GUCs enable targeted backend analysis without modifying application code.
• Useful for diagnosing hang-ups or resource contention during connection setup.
• Consultant-tested technique for deep-dive PostgreSQL connection troubleshooting.

A new analysis benchmarks Liquid Clustering against traditional ZOrdering and partitioning strategies across four distinct workload types in Delta Lake. The author identifies two undocumented failure modes that currently impact production systems using these layout techniques. The article provides a decision framework to help engineers select the most appropriate physical data layout strategy for their specific use cases.

• Liquid Clustering offers alternatives to static partitioning and ZOrdering for optimizing data layout.
• Two undocumented failure modes exist that may affect current production Delta Lake deployments.
• Performance varies significantly by workload type; no single layout strategy fits all scenarios.
• Use the provided decision framework to match layout strategies to specific query patterns.

Ming Ying details how ParadeDB accelerates COUNT queries on Wikipedia data by implementing ctid-aware segment partitioning and an inverted DocId map with bitmap postings. The approach extends to broader search workloads, where the system achieved 10 to 22 times higher throughput than TIN for filtered and time-sorted Hacker News queries. This optimization strategy highlights specific internal mechanisms for scaling search performance within PostgreSQL.

• ctid-aware segment partitioning reduces I/O for COUNT operations on large datasets
• Inverted DocId maps with bitmap postings significantly speed up search filtering
• ParadeDB delivers 10-22x throughput gains over TIN for time-sorted queries
• Optimizations target specific search patterns rather than general table scans

Chris van Eijk benchmarks custom session settings in PostgreSQL 17, analyzing how NULL or empty string values are handled by current_setting and set_config. The analysis covers value persistence, default behaviors, and permission models for who can modify these configurations. This measurement clarifies edge cases in session-level parameter management.

• Clarifies NULL vs empty string handling in custom session settings
• Measures persistence duration of set_config values within sessions
• Details permission requirements for modifying configuration parameters
• Provides empirical data for PostgreSQL 17 session management