Differences with Logica¶

Synalog is a fork of Logica with a fast Rust core. It keeps Logica's syntax and semantics, with deliberate changes aimed at AI agents. For how both languages relate to classical Datalog, see Differences with Datalog.

Named attributes only¶

Synalog doesn't support positional attributes like Logica or Datalog — it only uses named attributes, which reduce agent mistakes. This feature is optional in Logica; Synalog makes it mandatory.

# Synalog — always named
Employee(name:, salary:)

# Logica/Datalog positional style — not supported
Employee(x, y)

As a consequence, the compiled SQL uses actual column names rather than Logica's col{i} format, making the output compatible with existing database schemas.

Pagination¶

Pagination is critical for AI agents with limited context windows. It also avoids loading large amounts of data into memory, enabling use on memory-constrained cloud infrastructure.

Synalog applies pagination at compile time via the limit and offset arguments of compile(). The limit is combined with the @Limit directive: actual_limit = min(limit, @Limit).

Compile-time verification¶

Synalog embeds a formal verifier that catches structural errors before any SQL is generated — variable safety, safe negation and aggregation, stratification, arity consistency, and recursion safety. Logica defers these to the database at execution time.

Performance¶

The compiler is written in Rust and exposed via PyO3, dramatically reducing compilation time compared to the Python implementation — which matters when an agent compiles programs inside a reasoning loop.