Generation#

COGS 2 generation follows the same high-level flow for every publisher:

  1. Read and require the model’s CogsVersion before interpreting its versioned conventions.

  2. Read the folders, Markdown files, and CSV files that make up the model.

  3. Check the model for naming, datatype, inheritance, and other modeling errors.

  4. Resolve the model into a connected in-memory form with inherited properties, relationships, and datatype references worked out.

  5. Hand that resolved model to the selected publisher so it can write the target format.

The pages in this section describe how COGS concepts map into each generated format.

No publisher may skip those stages, continue after an error, mutate the shared model, or reinterpret source CSV. Source and target paths must not overlap, including through canonical paths or links.

Important

JSON Schema, XSD, C#, Python, and TypeScript implement the authoritative instance contract. UML/XMI is the authoritative structural model output, with PROJ2601 as its sole semantic exception; it is not an instance schema. OWL/RDF is emitted as W3C Turtle and is the authoritative ontology and class-semantics output, with OWL2002 and OWL2003 as its authority exceptions; ordering and unsupported lexical facets remain outside its authority. LinkML, DCTAP, GraphQL, DOT, and Sphinx are projections. A projection’s successful generation does not prove JSON/XML conformance and unsupported semantics must be diagnosed rather than silently dropped.

This Page