* Initial work on new search backend
* Clean up search backends
* Return only the most relevant result per object
* Clear any pre-existing cached entries on cache()
* #6003: Implement global search functionality for custom field values
* Tweak field weights & document guidance
* Extend search() to accept a lookup type
* Move get_registry() out of SearchBackend
* Enforce object permissions when returning search results
* Add indexers for remaining models
* Avoid calling remove() on non-cacheable objects
* Use new search backend by default
* Extend search backend to filter by object type
* Clean up search view form
* Enable specifying lookup logic
* Add indexes for value field
* Remove object type selector from search bar
* Introduce SearchTable and enable HTMX for results
* Enable pagination
* Remove legacy search backend
* Cleanup
* Use a UUID for CachedValue primary key
* Refactoring search methods
* Define max search results limit
* Extend reindex command to support specifying particular models
* Add clear() and size to SearchBackend
* Optimize bulk caching performance
* Highlight matched portion of field value
* Performance improvements for reindexing
* Started on search tests
* Cleanup & docs
* Documentation updates
* Clean up SearchIndex
* Flatten search registry to register by app_label.model_name
* Clean up search backend classes
* Clean up RestrictedGenericForeignKey and RestrictedPrefetch
* Resolve migrations conflict
When using permissions that use tags, a user may receive multiple permissions
of the same type if multiple tags are assigned to the device. This causes the
RestrictedQuerySet class to generate a query similar to this:
>>> dcim.models.Device.objects.filter(Q(tags__name='tag1')|Q(tags__name='tag2'))
<ConfigContextModelQuerySet [<Device: device1>, <Device: device1>]>
This query returns the same object twice if both tags are assigned to it. This
is due to the use of the django-taggit library. The library's documentation
describes this behavior as expected and suggests using an explicit distinct()
call in queries to avoid duplicates.
However, the use of DISTINCT in queries has a global side effect -
deduplication of responses, which may or may not be acceptable behavior
(depending on further use). Since it is not known how RestrictedQuerySet will
be used in the rest of the code, it was decided to dedupe using a subquery.
