Skip to content

relationships.md

Latest commit

 

History

History
73 lines (55 loc) · 3.72 KB

relationships.md

File metadata and controls

73 lines (55 loc) · 3.72 KB
Raw

Resource relationships

Relationships drive the "Has part" / "Is part of" links on resource detail pages and the "Browse all X records..." search filter.

Elasticsearch mapping

The field dct_isPartOf_sm is defined in the index mapping (text + keyword subfield) so the "Browse all" filter works. If the index was created before this mapping was added, recreate the index and run make reindex, or run a full reindex so that documents include dct_isPartOf_sm and (if dynamic mapping was used) the keyword subfield is available for filtering.

Building relationships

  • DB table: resource_relationships (subject_id, predicate, object_id). Populated from the resources table columns (e.g. dct_isPartOf_sm, pcdm_memberOf_sm).
  • Import behavior: OpenGeoMetadata harvests and Kithe Bridge syncs sync resource_relationships automatically for imported records and unchanged resources that point at them. This includes inverse replacement links such as dct:isReplacedBy from a new record's dct_replaces_sm value.
  • Source direction: A child stores its parent ID in dct_source_sm. The relationship table exposes that as dct:source on the child and the inverse dct:isSourceOf on the parent. Resource pages label these "Source record" and "Derived records," respectively.
  • Make task (from project root): 
    make populate-relationships
    Runs scripts/populate_relationships.py inside the API container. Use this after bulk imports that do not already sync relationships incrementally, or when relationship-sync code changes and you want a full rebuild.
  • Search filter: "Browse all X records..." uses:
    • Has part: include_filters[dct_isPartOf_sm][]=<parent_id> — children must have the parent ID in resources.dct_isPartOf_sm.
    • Collection records: include_filters[pcdm_memberOf_sm][]=<collection_id> — member resources must have the collection ID in resources.pcdm_memberOf_sm. Filters are applied in Elasticsearch using the indexed fields (with .keyword for exact match). So:
    1. Run make populate-relationships after manual/bulk DB edits that bypass the normal importers.
    2. Run make reindex after manual/bulk DB edits so Elasticsearch has up-to-date dct_isPartOf_sm and pcdm_memberOf_sm for the search filters.

Querying the database

Use the ParadeDB (PostgreSQL) container. From the project root, with Docker running:

docker compose exec paradedb psql -U postgres -d btaa_geospatial_api -c "..."

Check "Has part" rows for a resource (e.g. Wabash parent)

Replace eee6150b-ce2f-4837-9d17-ce72a0c1c26f with your resource ID.

Rows in resource_relationships where this resource is the parent (dct:hasPart):

SELECT predicate, subject_id AS child_id, object_id AS parent_id, r.dct_title_s AS child_title
FROM resource_relationships rr
JOIN resources r ON r.id = rr.subject_id
WHERE rr.object_id = 'eee6150b-ce2f-4837-9d17-ce72a0c1c26f'
  AND rr.predicate = 'dct:hasPart'
ORDER BY r.dct_title_s;

Check which resources have a given parent in dct_isPartOf_sm

These are the documents that should match the "Browse all" search filter.

SELECT id, dct_title_s, "dct_isPartOf_sm"
FROM resources
WHERE "dct_isPartOf_sm" IS NOT NULL
  AND 'eee6150b-ce2f-4837-9d17-ce72a0c1c26f' = ANY("dct_isPartOf_sm")
ORDER BY dct_title_s;

If this returns 0 rows, the DB doesn’t have any children with that parent ID in dct_isPartOf_sm; the search filter will correctly return 0 results. If it returns 25 rows but search still returns 0, run make reindex so Elasticsearch gets the updated dct_isPartOf_sm values.

Count relationships by type

SELECT predicate, COUNT(*) FROM resource_relationships GROUP BY predicate ORDER BY predicate;