rero / rero-ils / 30270388542

91%

staging: 91%

Build Type

push

github

Committed by PascalRepond

Commit Message

fix(documents): count documents instead of holdings in org facet

The organisation -> library -> location facet is built on the
`nested_holdings` nested field, so terms sub-aggregations count holdings,
not documents: counts were inflated whenever a document had several
holdings in the same bucket.

Add a `reverse_nested` sub-aggregation (`record_count`) at each level to
count distinct parent documents, and expose it as `doc_count` in the
serializer.

Order the terms buckets by `record_count` in the query so the `size` cut
keeps the buckets with the most documents: Elasticsearch orders terms by
holdings count by default, so a library or location with many documents
but few holdings could otherwise be dropped before the serializer sees it.
The query now returns the buckets already ordered by document count, so
the serializer no longer re-sorts them.

Drop the `sum_other_doc_count` and `doc_count_error_upper_bound` terms
metadata: they still describe the holdings-count aggregation and no longer
match the rewritten (document-count) buckets. They cannot be recomputed for
documents, as a document may belong to several buckets.

* Closes #4003.

Co-Authored-by: Johnny Mariéthoz <johnny.mariethoz@rero.ch>
Co-Authored-by: Ana Gomes <anagoncalves7@icloud.com>

Coverage Stats

17 of 17 new or added lines in 1 file covered. (100.0%)

229 existing lines in 19 files now uncovered.

24338 of 26640 relevant lines covered (91.36%)

0.91 hits per line