Abstract
Many real-world applications based on deductive databases require incrementally updating output relations (tables) in response to changes to input relations. To make such applications easier to implement we have created Differential Datalog (DDlog), a dialect of Datalog that automates incremental computation. A DDlog programmer writes traditional, non-incremental Datalog programs. The execution model of DDlog is however fully incremental: at runtime DDlog programs receive streams of changes to the input relations (insertions or deletions) and produce streams of corresponding changes to derived relations. The DDlog compiler translates DDlog programs to Differential Dataflow (DD) programs; DD provides an incremental execution engine supporting all the relational operators, including fixed-point. The DDlog runtime automatically maintains the indexes required to efficiently compute output updates.
The DDlog language is targeted for system builders. In consequence, the language emphasizes usability, by providing a rich type system, a powerful expression language, a module system, including string manipulation, arithmetic, and integration with C, Rust, and Java. The code is open-source, available using an MIT permissive license.