marinerg-schema

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

marinerg_dataset.linkml.yaml, written in LinkML, defines the dataset fields and their controlled vocabularies. The CKAN catalogue's form and validators, and the facility client's checks, all take their fields and values from it. It's a package so that each of them uses the same version instead of a copy.

Install

pip install marinerg-schema

Use

from marinerg_schema import (
    check,
    concept_uri,
    dataset_slots,
    permitted_values,
    slot,
    value_titles,
)
from marinerg_schema.ckan import field_names

# frozenset({'L0', 'L1', 'L2', 'L3', 'L4'})
permitted_values("processing_level")
# 'http://vocab.nerc.ac.uk/collection/EPL/current/L1/'
concept_uri("processing_level", "L1")
# {'L0': 'L0 — Raw', ...}, for a form's choices
value_titles("ProcessingLevelEnum")
# The slot's range, constraints and RDF predicate
slot("device_trl")
# ValueError: Must be at most 9: '12'
check("device_trl", "12")
# The model's fields, in model order
dataset_slots()
# The keys a CKAN dataset takes at the top level
field_names()

The measured-parameter vocabulary says what a data channel measures. A facility template names each channel's parameter by its code:

from marinerg_schema import parameter, parameters

parameter("wind_speed").unit  # 'm s-1', as UDUNITS
parameter("wind_speed").cf_standard_name  # 'wind_speed'
parameter("roll").cf_standard_name  # None: no verified CF name

A CF or NERC mapping is filled only where the term was checked to exist. Lab and device quantities, such as mooring-line tension, often have none yet.

registry_path() returns the generated registry YAML for consumers that want a file rather than the API. vocab_path() returns our own vocabulary as SKOS Turtle: every term under https://vocab.marinerg-i.eu/, with its label, definition and close matches.

Consumers

Predecessor

This model follows on from MaRINET2's 2020 metadata schemas, at github.com/Marinet2/metadata-schema and FAIRsharing bsg-s001497. This package replaces its dataset schema. marinerg-data's run manifest replaces its datastream schema. "Marinet2 NetCDF 0.1" is still an allowed value in the conventions list.

The MaRINET2 repository still reads as current. Only its owners can add a notice there, or link this model from its FAIRsharing record.

Change the model

Edit the LinkML file, then regenerate the registry and the SKOS vocabulary. Never edit either by hand. The tests fail if they differ from the model.

uv run python -m marinerg_schema
uv run python -m marinerg_schema --check

Give every value a title, which forms show, and a meaning, its concept URI. Write NERC CURIEs with their trailing slash, such as nerc_epl:L1/, because NERC's concept URIs end in one. A value of our own also needs a description, which becomes its SKOS definition.

The package reads the model without LinkML, so it installs anywhere CKAN does. The tests run linkml-lint and check the package's reading against LinkML's own.

After a vocabulary change, cut a release, then move the pin in each consumer. The model's version is the package's.

Develop

uv sync --group dev
uv run tox

uv run tox runs the format check, lint, type check and tests, as CI does.

The package supports Python 3.10, because CKAN 2.11 runs on it. Ruff targets 3.10 for the same reason.

Release

Open the newest pipeline for main and press the play button on release patch, minor or major. The release pipeline publishes to PyPI, then tags the commit and creates the GitLab release.

Licence

Copyright ICHEC. MIT. See LICENSE.