Specification: Overall Data Model

Profiles: basic v1

The MaRDI file format stores mathematical objects as annotated JSON trees. A file combines type information, serialized data, namespaces, and optional reference objects that provide the context required to interpret the payload.

Terminology

• A file is the top-level JSON object.
• A data type is the value of _type, either as a string or as an object with a name and optional params.
• A payload is the subtree stored under data.
• A profile is a namespace-specific encoding contract, for example the Oscar profile identified by an Oscar entry in _ns.
• A reference object is an entry in _refs, addressed by UUID and reused from types or payloads.

Core object members

• _type is required and names the data type that determines how data must be interpreted.
• data stores the serialized payload; it may be a string, array, object, or a foreign schema-defined subtree.
• _ns declares which namespace and software version define the semantics.
• _refs stores referenced objects so recursive and shared constructions can be serialized without duplication.

Design intent

• The format follows the paper's approach of separating syntax from semantics: JSON fixes the container syntax, while semantics are supplied by a concrete namespace and version.
• The format is intentionally extensible. New types, new namespaces, and namespace-specific payloads can be added without redesigning the whole file format.
• References are session-stable UUIDs rather than position-based indices, which makes reused mathematical context easier to track across files and workflows.