bods-fixtures

Canonical BODS v0.4 fixtures and edge-case conformance suite, shared across the BODS adapter and query-engine ecosystem.

Why this exists

The BODS adapter suite (OpenCorporates, ICIJ, Kyckr, BrightQuery, FtM, AML AI, Lance, Neo4j, GQL…) currently ships per-repo fixtures. That means each repo decides independently what "anonymous person" or "circular ownership" looks like, and nothing validates that an entity observed in two sources produces equivalent BODS.

bods-fixtures is the single source of truth:

• a small, curated set of BODS v0.4 statement bundles, each targeting a specific semantic edge case
• a co-located .expected.md spec per fixture describing what a correct adapter / query engine should do with it
• a thin Python loader so adapters can parametrize their test suites against the shared pack

Install

pip install bods-v04-fixtures

The PyPI distribution is bods-v04-fixtures — the name makes the version scope explicit and leaves the generic bods-fixtures name available for future ecosystem-wide tooling. The Python import path is unchanged: from bods_fixtures import ....

Usage

from bods_fixtures import load, list_cases

fixture = load("core/01-direct-ownership")
fixture.statements                       # parsed BODS v0.4 statements
fixture.statement_ids                    # all statementIds
fixture.record_ids                       # all recordIds
fixture.by_record_type("person")         # filter by recordType: entity|person|relationship
fixture.find_record("rec-jane-smith-1975")  # look up a record by recordId
fixture.description                      # rendered .expected.md spec

list_cases()                             # all fixtures
list_cases("edge-cases")                 # one category

BODS v0.4 only. Fixtures use the canonical recordType / recordDetails envelope and string-based recordId references for relationships, per the official 0.4 schema. Adapters built against the older statementType / describedByEntityStatement shape (BODS v0.3) will need updating before they can consume this pack — which is one of the things this pack exists to surface.

Parametrised tests

import pytest
from bods_fixtures import list_cases, load

@pytest.mark.parametrize("name", list_cases())
def test_my_adapter_round_trips(name):
    fixture = load(name)
    result = my_adapter.convert(fixture.statements)
    assert ...

What's in the pack

Today (v0.1):

Planned (see CHANGELOG for target versions):

• core: person-with-control, entity-owns-entity, multiple-interest-types
• edge-cases: indirect chain, threshold boundaries, trust/nominee, joint ownership, historical interest, replaced statement
• cross-border: multi-jurisdiction chains, bearer shares
• cross-source: same entity from ≥2 source adapters, for consistency tests
• scale: synthetic 1k / 100k statement bundles for query-engine benchmarks
• invalid: deliberately malformed, for negative tests

Governance

• Versioning: semver. A fixture rename, removal, or semantic change bumps major. Field additions inside a fixture bump minor. Pure documentation edits bump patch.
• Adding a fixture: the .expected.md is reviewed like a spec change. It is the durable record of what the fixture is testing and what correct behaviour looks like.
• Removing a fixture: requires a deprecation cycle — mark in CHANGELOG one major version before removal.

Validation

The shipped tests are structural (references resolve, IDs unique, BODS version declared). For full schema validation against the BODS v0.4 JSON schema:

pip install 'bods-v04-fixtures[validate]'

(schema validation suite is a v0.2 deliverable.)

License

MIT. Fixture content is CC-BY-SA 4.0 to match BODS licensing conventions.