extensions/migrations/0033_extensions_fts_20240603_1918.py

@@ -0,0 +1,41 @@
+# Generated by Django 4.2.11 on 2024-06-03 17:18
+
+from django.db import migrations
+
+
+def create_indexes(apps, schema_editor):
+    if schema_editor.connection.vendor != 'postgresql':
+        return
+    with schema_editor.connection.cursor() as cursor:
+        cursor.execute(
+            """
+            CREATE INDEX extensions_fts ON extensions_extension USING
+            gin ((to_tsvector('english', name) || ' ' || to_tsvector('english', description)))
+            """
+        )
+        cursor.execute('create extension if not exists pg_trgm;')
+        cursor.execute(
+            """
+            CREATE INDEX extensions_trgm_gin ON extensions_extension USING
+            gin((((name)::text || ' '::text) || description) gin_trgm_ops);
+            """
+        )
+
+
+def delete_indexes(apps, schema_editor):
+    if schema_editor.connection.vendor != 'postgresql':
+        return
+    with schema_editor.connection.cursor() as cursor:
+        cursor.execute('drop index extensions_fts')
+        cursor.execute('drop index extensions_trgm_gin')
+
+
+class Migration(migrations.Migration):
+
+    dependencies = [
+        ('extensions', '0032_extension_extensions__is_list_765936_idx_and_more'),
+    ]
+
+    operations = [
+        migrations.RunPython(create_indexes, delete_indexes)
+    ]

extensions/views/public.py

@@ -2,6 +2,7 @@ from collections import OrderedDict
 import logging
 
 from django.contrib.auth import get_user_model
+from django.db import connection
 from django.db.models import Count, Q
 from django.shortcuts import get_object_or_404, redirect
 from django.utils.translation import gettext_lazy as _
@@ -24,7 +25,15 @@ log = logging.getLogger(__name__)
 
 class ListedExtensionsView(ListView):
     model = Extension
-    queryset = Extension.objects.listed
+    queryset = Extension.objects.listed.prefetch_related(
+        'authors',
+        'latest_version__file',
+        'latest_version__tags',
+        'preview_set',
+        'preview_set__file',
+        'ratings',
+        'team',
+    )
     context_object_name = 'extensions'
 
 
@@ -34,20 +43,7 @@ class HomeView(ListedExtensionsView):
 
     def get_context_data(self, **kwargs):
         context = super().get_context_data(**kwargs)
-        q = (
-            super()
-            .get_queryset()
-            .prefetch_related(
-                'authors',
-                'latest_version__file',
-                'latest_version__tags',
-                'preview_set',
-                'preview_set__file',
-                'ratings',
-                'team',
-            )
-            .order_by('-average_score')
-        )
+        q = super().get_queryset().order_by('-average_score')
         context['addons'] = q.filter(type=EXTENSION_TYPE_CHOICES.BPY)[:8]
         context['themes'] = q.filter(type=EXTENSION_TYPE_CHOICES.THEME)[:8]
         return context
@@ -94,7 +90,7 @@ class SearchView(ListedExtensionsView):
         return sort_by
 
     def get_queryset(self):
-        queryset = super().get_queryset()
+        queryset = super().get_queryset().order_by(self._get_sort_by())
         if self.kwargs.get('tag_slug'):
             queryset = queryset.filter(
                 latest_version__tags__slug=self.kwargs['tag_slug']
@@ -108,26 +104,72 @@ class SearchView(ListedExtensionsView):
         if self.kwargs.get('type_slug'):
             _type = self._get_type_id_by_slug()
             queryset = queryset.filter(type=_type)
-        if 'q' in self.request.GET:
-            qs = self.request.GET['q'].split()
-            search_query = Q()
-            for token in qs:
-                search_query &= (
+
+        search_query = self.request.GET.get('q')
+        if not search_query:
+            return queryset
+
+        # WARNING: full-text search support only on postgres
+        if connection.vendor == 'postgresql':
+            queryset = self.postgres_fts(queryset, search_query)
+        else:
+            filter = Q()
+            for token in search_query.split():
+                filter &= (
                     Q(slug__icontains=token)
                     | Q(name__icontains=token)
                     | Q(description__icontains=token)
                     | Q(latest_version__tags__name__icontains=token)
                 )
-            queryset = queryset.filter(search_query).distinct()
-        return queryset.prefetch_related(
-            'authors',
-            'latest_version__file',
-            'latest_version__tags',
-            'preview_set',
-            'preview_set__file',
-            'ratings',
-            'team',
-        ).order_by(self._get_sort_by())
+            queryset = queryset.filter(filter).distinct()
+        return queryset
+
+    def postgres_fts(self, queryset, search_query):
+        """Postgres full text search (fast) and a fuzzy trigram search (slow) as a fallback.
+
+        Searches Extension name and description only, ranking name matches higher.
+        If we need to extend the functionality, it's better to consider using a different approach,
+        e.g. introduce meilisearch.
+
+        Limits the results size to 32 items (2 pages), assuming that nobody will click through many
+        pages if we failed to present the vital results on the first page.
+
+        Relies on indexed expressions:
+            CREATE INDEX extensions_fts ON extensions_extension USING
+            gin ((to_tsvector('english', name) || ' ' || to_tsvector('english', description)));
+            CREATE INDEX extensions_trgm_gin ON extensions_extension USING
+            gin((((name)::text || ' '::text) || description) gin_trgm_ops);
+
+        """
+        with connection.cursor() as cursor:
+            sql = """
+                select id
+                from extensions_extension
+                where (
+                    (to_tsvector('english', name) || ' ' || to_tsvector('english', description))
+                    @@ websearch_to_tsquery('english', %(query)s)
+                ) and is_listed
+                order by ts_rank(
+                    to_tsvector('english', name),
+                    websearch_to_tsquery('english', %(query)s)
+                ) desc
+                limit 32"""
+            cursor.execute(sql, {'query': search_query})
+            pks = [row[0] for row in cursor.fetchall()]
+            if not pks:
+                # fallback to fuzzy trigram search
+                sql = """
+                    select id
+                    from extensions_extension
+                    where ((name || ' ' || description) %%> %(query)s)
+                    and is_listed
+                    order by %(query)s <<<-> (name || ' ' || description)
+                    limit 32"""
+                cursor.execute(sql, {'query': search_query})
+                pks = [row[0] for row in cursor.fetchall()]
+            # pks are ordered by ranking, keep that order
+            # this approach is fine under the assumption that the list is small
+            return sorted(queryset.filter(pk__in=pks).order_by(), key=lambda x: pks.index(x.pk))
 
     def get_context_data(self, **kwargs):
         context = super().get_context_data(**kwargs)