No mechanism for Orphaned Editions to be removed from Solr once resolved

[hornc]

Orphaned editions (ones without Works) are currently indexed in Solr as /works/<edition_ID> e.g /works/OL20454633M

This is a workaround to allow them to appear in search results

openlibrary/openlibrary/solr/update_work.py

Line 1127 in d1fb792

# When an edition does not contain a works list, create a fake work and index it.

When a work is added by a move or other edit, there is nothing that notices the change and removes the old /works/<edition_ID> from the index, so the fake work will continue to appear in search results.

/works/<edition_ID> needs a Delete update sent to Solr.

Currently the only way to remove them is to manually add the /works/<edition_ID> keys to admin/solr interface.

Suggestion:
When the solr updater script sees an edition edit in the Infogami logs, in the changeset / changes fields, it should send a delete request to Solr for the /works/<edition_ID> This will produce some delete requests for non-existent keys, but this does not cause a problem for solr. The alternative would be to query solr to find whether a '/works/` key exists for that edition, and delete if so. My thinking is that sending a delete anyway will put less load on solr but have the same end result.

[hornc]

Added the "blocker" tag because bulk or manual resolving orphaned editions is not much use if search results and author work lists are going to show them indefinitely.

[LeadSongDog]

These fake works are particularly egregious in that an edition without all three of an author, a publisher, and a publication place seems to create a new fake work even in the attempt to link that edition to an existing known work.

[hornc]

Resolved in cd67d7b