FlowLog: Efficient and Extensible Datalog via Incrementality
H. Zhao, Z. Yu, S. Rao, S. Frisk, Z. Fan, P. Koutris
VLDB 2026
FlowLog
A Datalog engine powered by Differential Dataflow. You write Datalog queries in Soufflé and FlowLog efficiently maintains the query results incrementally.
News: FlowLog has been accepted to VLDB 2026. Read it now.
Key Features
Soufflé Language
Write Datalog in the established Soufflé syntax.
Robustness First
Configure worst-case optimal query plans against runtime data skew.
Dual Modes
Optimize under both batch and incremental execution modes.
Efficiency
Scale (up and out) efficiently compared to state-of-the-art Datalog engines.
Extensibility
Easy to extend to any relational programs in operational semantics.
Outlook
Many optimization opportunities left on the table.
Publications
Evaluating Datalog over Semirings: A Grounding-based Approach
H. Zhao, S. Deep, P. Koutris, S. Roy, V. Tannen
PODS 2024
Evaluating Datalog via Structure-Aware Rewriting
H. Zhao, S. Deep, P. Koutris
Datalog 2.0 2024
Predicate Transfer: Efficient Pre-Filtering on Multi-Join Queries
Y. Yang, H. Zhao, X. Yu, P. Koutris
CIDR 2024
Ready to experiment?
Reproduce our VLDB results, plug in your Soufflé programs, or chat with us about shaping the roadmap. FlowLog thrives on real-world workloads.
Getting Involved
FlowLog is open-source on GitHub flowlog-rs under active development. If you're interested in discussing FlowLog query optimizations, email at hangdong@cs.wisc.edu.