Design Principles

DortDB is a TypeScript framework for querying existing in-memory data, built to be as configurable and modular as possible. A handful of principles explain most of its design decisions.

Language neutrality

Different query languages have different strengths, but the operations they express are ultimately similar: graph pattern matching, for instance, can be understood as a sequence of selections and joins. DortDB separates languages from planning and execution, and lets you choose which languages to load as plug-ins. No language is privileged: any loaded language can be the outer language, and any can be nested inside another.

Schema agnosticism

DortDB targets the browser and the querying of ordinary JavaScript values, which are dynamically typed and permissive. It therefore takes a schema-free approach: there is no schema to declare and no import step. This keeps registration a constant-time operation and keeps languages decoupled from the concrete shape of the data (that shape is handled by data adapters). The trade-off is that a few schema-dependent language features are restricted; see Limitations.

An algebra as the compilation target

To decouple languages from the rest of the framework, each language parses its queries into one unified algebra. The engine then reasons about algebraic operators rather than language-specific constructs, which is what makes the optimizer and executor language-agnostic. The algebra is described in the Formalism section.

Multi-language queries

Because every language compiles to the same representation, a single query can mix languages through nested language subqueries, and the whole thing is optimized and executed as one plan. See Cross-language Queries.

Extensibility and modularity

Extensibility is a fundamental goal. A new language can lean on the framework for most of the work, but every stage can be customized: you can add new algebra operators for new concepts or data models (XQuery's TreeJoin and ProjectionSize operators are examples), add optimizer rules, or add functions, aggregates, and index types. Modularity serves the same goal from the packaging side: small, separate packages keep web bundles small. See Extending DortDB.