marinerg-schema

The MARINERG-i dataset metadata model, as a package.

marinerg_dataset.linkml.yaml (LinkML) defines the marine metadata fields and their controlled vocabularies. Everything else in the estate is a projection of it: the CKAN catalogue's form and validators, and the facility client's pre-publication checks.

It is a package because it has more than one consumer. It used to live inside the CKAN extension, and each consumer kept its own copy — a hand-written field list in marinerg-data, enum values restated in CKAN's form schema. Copies drift silently, and these had: the field list was three fields behind.

Install

pip install marinerg-schema

Use

from marinerg_schema import permitted_values, concept_uri, dataset_slots
from marinerg_schema.ckan import field_names

permitted_values("processing_level")   # frozenset({'L0', 'L1', 'L2', 'L3', 'L4'})
concept_uri("processing_level", "L1")  # 'http://vocab.nerc.ac.uk/collection/EPL/current/L1/'
dataset_slots()                        # the model's fields, in model order
field_names()                          # the keys a CKAN dataset takes top-level

registry_path() returns the generated registry YAML for consumers that want a file rather than the API.

Consumers

Predecessor

This model succeeds the MaRINET2 metadata schemas at github.com/Marinet2/metadata-schema (2020, registered at FAIRsharing bsg-s001497) — the dataset schema here, and MaRINET2's datastream schema by marinerg-data's run-manifest schema.

A successor, not a replacement: MaRINET2's schemas are the consortium-familiar backbone, and "Marinet2 NetCDF 0.1" stays a declared value in the conventions list. That repository is outside this group and still reads as current guidance, so a status banner there and a successor relation on its FAIRsharing record have to be requested from its owners.

Change the model

Edit the LinkML file, then regenerate the registry — never hand-edit it:

uv run python -m marinerg_schema
uv run python -m marinerg_schema --check   # what CI runs

A vocabulary change is a release: cut one, then move the consumers' pins. version: inside the LinkML file is the model's version and moves independently of the package's.

Develop

uv sync --group dev
uv run pytest
uv run ruff check src tests
uv run mypy src

The package floor is Python 3.10 — CKAN 2.11's runtime — even though the client that also consumes it needs 3.11. Ruff targets 3.10 for the same reason.

Licence

Copyright ICHEC. MIT — see LICENSE.