One SQL to Rule Them All
summary
Paper proposes a concrete, implementable model (accepted at SIGMOD 2019) that unifies batch and streaming SQL using three parts: time-varying relations as the core abstraction, explicit event-time via schema timestamps plus watermarks, and small DDL/CL options to control materialization. Examples show table-valued window functions, watermark-based completeness, and change‑log outputs using EMIT STREAM and EMIT STREAM AFTER WATERMARK (or AFTER DELAY). Implementations and prototypes exist in Apache Calcite, Flink (partial), and Beam.
AI-written summary · about
chapters (6)
- 0:00Introduction and motivation
- 3:40Time varying relations concept
- 10:50Event time semantics
- 21:14Materializing results as streams
- 33:36Adoption and future work
- 38:54Summary and conclusion
related talks
21mStream Processing and SQL
41mStreaming SQL
41mData Engineers: Respect the Resiliency of SQL
48mUnlocking Real-time Insights: Enhancing Your Databases With Stream Processing
30mThe Future of Data is Words
23mBetter, Faster, Stronger Streaming: Your First Dive into Flink SQL
43mUnified stream and batch processing with Apache Flink’s relational APIs
33mStreaming Databases