[pull] develop from baserow:develop

.github/workflows/README.md

@@ -10,7 +10,7 @@ This workflow automatically updates the Database Team's GitHub Project board bas
 
 The workflow triggers on PR events (opened, review requested, review submitted, merged, etc.) and automatically:
 
-1. **Checks domain labels** - Only processes PRs with labels starting with `domain::database` or `domain::core`
+1. **Checks domain labels** - Only processes PRs with labels starting with `database 🗄`
 2. **Adds PR to project** - Ensures the PR is added to the Database Team project board
 3. **Updates Status field** - Sets the PR status based on its state:
    - `In Progress` - PR is a draft

.github/workflows/database-projects-issues-workflow.yml

@@ -2,7 +2,7 @@ name: Update Database Team Project Fields
 
 env:
   PROJECT_NUMBER: '3'
-  DOMAIN_LABELS: '["domain::database"]'
+  DOMAIN_LABELS: '["database 🗄️"]'
   STATUS_FIELD_NAME: 'Status'
   STATUS_TODO: 'Todo'
 
@@ -91,7 +91,7 @@ jobs:
               return;
             }
 
-            console.log(`Issue has domain label: ${labels.filter(l => l.startsWith('domain::')).join(', ')}`);
+            console.log(`Issue has domain label: ${labels.filter(l => domainLabels.some(d => l.startsWith(d))).join(', ')}`);
 
             // ============================================================
             // CHECK IF ALREADY IN PROJECT

.github/workflows/database-projects-pr-workflow.yml

@@ -2,7 +2,7 @@ name: Database Team PR Automation
 
 env:
   PROJECT_NUMBER: '3'
-  DOMAIN_LABELS: '["domain::database"]'
+  DOMAIN_LABELS: '["database 🗄️"]'
   STATUS_FIELD_NAME: 'Status'
   REVIEW_STATUS_FIELD_NAME: 'Review Status'
   STATUS_IN_PROGRESS: 'In Progress'
@@ -26,6 +26,7 @@ on:
 
 jobs:
   update-status:
+    if: github.actor != 'dependabot[bot]'
     runs-on: ubuntu-latest
     steps:
       - name: Update PR Project Status
@@ -114,7 +115,7 @@ jobs:
               return;
             }
 
-            console.log(`PR has domain label: ${labels.filter(l => l.startsWith('domain::')).join(', ')}`);
+            console.log(`PR has domain label: ${labels.filter(l => domainLabels.some(d => l.startsWith(d))).join(', ')}`);
 
             const statusField = project.fields.nodes.find(f => f.name === '${{ env.STATUS_FIELD_NAME }}');
             const reviewStatusField = project.fields.nodes.find(f => f.name === '${{ env.REVIEW_STATUS_FIELD_NAME }}');

backend/src/baserow/contrib/database/fields/registries.py

@@ -2127,31 +2127,29 @@ def get_group_by_field_filters_and_annotations(
         reversed_field = through_model._meta.get_fields()[1].name
         related_field = through_model._meta.get_fields()[2].name
 
-        if field_name not in cte:
-            row_ids = [row.id for row in rows]
-            # Improve performance of the query by creating a CTE with all the
-            # relationships of the field to group by. This is significantly faster than
-            # doing this every row in a separate subquery.
-            aggregated_cte = (
-                through_model.objects.filter(**{f"{reversed_field}_id__in": row_ids})
-                .values(f"{reversed_field}_id")
-                .annotate(
-                    res=ArrayAgg(
-                        F(f"{related_field}_id"),
-                        filter=Q(**{f"{related_field}_id__isnull": False}),
-                        order_by=self.get_group_by_aggregated_order(related_field),
-                    )
-                )
-            )
-            cte[field_name] = With(aggregated_cte, name=f"{field_name}_cte")
-
         filters = {field_name: value}
+        # Use a correlated subquery instead of a CTE so that the database
+        # computes each row's linked-ID array via an index lookup on the
+        # through table, without materialising the entire relationship set
+        # up-front. This is correct for all rows in the base_queryset
+        # (including off-page ones) and avoids dragging enhance_by_fields()
+        # joins into a CTE filter subquery.
         annotations = {
             field_name: Coalesce(
                 Subquery(
-                    cte[field_name]
-                    .queryset()
-                    .filter(**{f"{reversed_field}_id": OuterRef("id")})
+                    through_model.objects.filter(
+                        **{
+                            f"{reversed_field}_id": OuterRef("id"),
+                            f"{related_field}_id__isnull": False,
+                        }
+                    )
+                    .values(f"{reversed_field}_id")
+                    .annotate(
+                        res=ArrayAgg(
+                            F(f"{related_field}_id"),
+                            order_by=self.get_group_by_aggregated_order(related_field),
+                        )
+                    )
                     .values("res")[:1]
                 ),
                 Value([], output_field=ArrayField(IntegerField())),

backend/tests/baserow/contrib/database/field/test_link_row_field_type.py

@@ -2722,6 +2722,58 @@ def test_get_group_by_metadata_in_rows_with_many_to_many_field(data_fixture):
     }
 
 
+@pytest.mark.django_db
+def test_list_rows_group_by_link_row_counts_across_pages(api_client, data_fixture):
+    user, token = data_fixture.create_user_and_token(
+        email="test@test.nl", password="password", first_name="Test1"
+    )
+    table_a, table_b, link_a_to_b = data_fixture.create_two_linked_tables(user=user)
+
+    row_b1, row_b2 = (
+        RowHandler()
+        .force_create_rows(
+            user=user,
+            table=table_b,
+            rows_values=[{}, {}],
+        )
+        .created_rows
+    )
+
+    RowHandler().force_create_rows(
+        user=user,
+        table=table_a,
+        rows_values=[
+            {f"field_{link_a_to_b.id}": []},
+            {f"field_{link_a_to_b.id}": []},
+            {f"field_{link_a_to_b.id}": [row_b1.id]},
+            {f"field_{link_a_to_b.id}": [row_b1.id]},
+            {f"field_{link_a_to_b.id}": [row_b2.id]},
+            {f"field_{link_a_to_b.id}": [row_b2.id]},
+        ],
+    )
+
+    grid = data_fixture.create_grid_view(table=table_a)
+    data_fixture.create_view_group_by(view=grid, field=link_a_to_b)
+
+    url = reverse("api:database:views:grid:list", kwargs={"view_id": grid.id})
+    response = api_client.get(f"{url}?size=3", **{"HTTP_AUTHORIZATION": f"JWT {token}"})
+    assert response.status_code == HTTP_200_OK
+    response_json = response.json()
+
+    # Only groups present on the current page are returned, but their counts
+    # must reflect ALL matching rows in the view.  Before the fix, off-page
+    # rows were missing from the CTE so they were counted as "empty",
+    # producing count=5 for [] and count=1 for [row_b1] instead of 2 each.
+    assert response_json["group_by_metadata"] == {
+        f"field_{link_a_to_b.id}": unordered(
+            [
+                {f"field_{link_a_to_b.id}": [], "count": 2},
+                {f"field_{link_a_to_b.id}": [row_b1.id], "count": 2},
+            ]
+        )
+    }
+
+
 @pytest.mark.django_db
 def test_list_rows_with_group_by_link_row_to_multiple_select_field(
     api_client, data_fixture

changelog/entries/unreleased/bug/5047_fix_group_by_counts_for_link_row_field.json

@@ -0,0 +1,9 @@
+{
+    "type": "bug",
+    "message": "Fix group by counts for link row field",
+    "issue_origin": "github",
+    "issue_number": 5047,
+    "domain": "database",
+    "bullet_points": [],
+    "created_at": "2026-03-25"
+}

web-frontend/package.json

@@ -31,9 +31,9 @@
   },
   "dependencies": {
     "@nuxtjs/i18n": "10.2.4",
-    "@sentry/node": "10.40.0",
-    "@sentry/nuxt": "10.40.0",
-    "@sentry/vue": "10.40.0",
+    "@sentry/node": "10.46.0",
+    "@sentry/nuxt": "10.46.0",
+    "@sentry/vue": "10.46.0",
     "@tiptap/core": "^3.13.0",
     "@tiptap/extension-blockquote": "^3.13.0",
     "@tiptap/extension-bold": "^3.13.0",
@@ -103,12 +103,12 @@
     "normalize.css": "^8.0.1",
     "nuxt": "^3.21.2",
     "papaparse": "5.4.1",
-    "path-to-regexp": "8.2.0",
+    "path-to-regexp": "8.4.0",
     "posthog-js": "^1.232.4",
     "sass": "1.98.0",
     "thenby": "^1.3.4",
     "tippy.js": "^6.3.7",
-    "tiptap-markdown": "^0.8.9",
+    "tiptap-markdown": "0.9.0",
     "tldjs": "^2.3.1",
     "ulid": "^3.0.1",
     "vite-plugin-node-polyfills": "^0.24.0",