🚸(backend) improve users similarity search and sort results

In some edge cases, the domain part the email addresse is
longer than the name part. Users searches by email similarity
then return a lot of unsorted results.

We can improve this by being more demanding on similarity when
the query looks like an email. Sorting results by the similarity
score is also an obvious improvement.

At the moment, we still think it is good to propose results with
a weak similarity on the name part because we want to avoid
as much as possible creating duplicate users by inviting one of
is many emails, a user who is already in our database.

Fixes 399

CHANGELOG.md

@@ -9,6 +9,10 @@ and this project adheres to
 
 ## [Unreleased]
 
+## Changed
+
+- 🚸(backend) improve users similarity search and sort results #391
+
 ## [1.7.0] - 2024-10-24
 
 ## Added

src/backend/core/api/viewsets.py

@@ -6,6 +6,7 @@
 
 from django.conf import settings
 from django.contrib.postgres.aggregates import ArrayAgg
+from django.contrib.postgres.search import TrigramSimilarity
 from django.core.exceptions import ValidationError
 from django.core.files.storage import default_storage
 from django.db.models import (
@@ -156,8 +157,21 @@ def get_queryset(self):
 
             # Filter users by email similarity
             if query := self.request.GET.get("q", ""):
+                # For performance reasons we filter first by similarity, which relies on an index,
+                # then only calculate precise similarity scores for sorting purposes
                 queryset = queryset.filter(email__trigram_word_similar=query)
 
+                queryset = queryset.annotate(
+                    similarity=TrigramSimilarity("email", query)
+                )
+                # When the query only is on the name part, we should try to make many proposals
+                # But when the query looks like an email we should only propose serious matches
+                threshold = 0.6 if "@" in query else 0.1
+
+                queryset = queryset.filter(similarity__gt=threshold).order_by(
+                    "-similarity"
+                )
+
         return queryset
 
     @decorators.action(

src/backend/core/tests/test_api_users.py

@@ -69,6 +69,48 @@ def test_api_users_list_query_email():
     assert user_ids == [str(nicole.id), str(frank.id)]
 
 
+def test_api_users_list_query_email_matching():
+    """While filtering by email, results should be filtered and sorted by similarity"""
+    user = factories.UserFactory()
+
+    client = APIClient()
+    client.force_login(user)
+
+    alice = factories.UserFactory(email="[email protected]")
+    factories.UserFactory(email="[email protected]")
+    michael_wilson = factories.UserFactory(email="[email protected]")
+    factories.UserFactory(email="[email protected]")
+    michael_brown = factories.UserFactory(email="[email protected]")
+    factories.UserFactory(email="[email protected]")
+
+    response = client.get(
+        "/api/v1.0/users/[email protected]",
+    )
+    assert response.status_code == 200
+    user_ids = [user["id"] for user in response.json()["results"]]
+    assert user_ids == [str(michael_wilson.id)]
+
+    response = client.get("/api/v1.0/users/[email protected]")
+
+    assert response.status_code == 200
+    user_ids = [user["id"] for user in response.json()["results"]]
+    assert user_ids == [str(michael_wilson.id), str(alice.id), str(michael_brown.id)]
+
+    response = client.get(
+        "/api/v1.0/users/[email protected]",
+    )
+    assert response.status_code == 200
+    user_ids = [user["id"] for user in response.json()["results"]]
+    assert user_ids == [str(alice.id)]
+
+    response = client.get(
+        "/api/v1.0/users/[email protected]",
+    )
+    assert response.status_code == 200
+    user_ids = [user["id"] for user in response.json()["results"]]
+    assert user_ids == [str(michael_wilson.id)]
+
+
 def test_api_users_list_query_email_exclude_doc_user():
     """
     Authenticated users should be able to list users