Fix random_projects route not returning the requested number of projects (#3758)

* Fix random_projects route not returning the requested number of projects

* fix(labrinth): further improve random project route SQL query

* chore: fix typo in comment

* tweak(labrinth): more apparent and fast randomness for `random_projects_get`

* tweak(labrinth): even better random projects query

* chore: address formatting review

---------

Co-authored-by: Alejandro González <me@alegon.dev>

apps/labrinth/.sqlx/query-1cefe4924d3c1f491739858ce844a22903d2dbe26f255219299f1833a10ce3d7.json

@@ -1,23 +0,0 @@
-{
-  "db_name": "PostgreSQL",
-  "query": "\n            SELECT id FROM mods TABLESAMPLE SYSTEM_ROWS($1) WHERE status = ANY($2)\n            ",
-  "describe": {
-    "columns": [
-      {
-        "ordinal": 0,
-        "name": "id",
-        "type_info": "Int8"
-      }
-    ],
-    "parameters": {
-      "Left": [
-        "Int8",
-        "TextArray"
-      ]
-    },
-    "nullable": [
-      false
-    ]
-  },
-  "hash": "1cefe4924d3c1f491739858ce844a22903d2dbe26f255219299f1833a10ce3d7"
-}

apps/labrinth/.sqlx/query-1d017ac5f5b1e76ec241533fd4d061c79f4e6d2f1701e727a7474fd5029a5492.json

@@ -0,0 +1,23 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "SELECT id FROM mods WHERE status = ANY($1)\n        ORDER BY id\n        LIMIT $2\n        OFFSET GREATEST(ROUND(RANDOM() * (SELECT COUNT(*) FROM mods WHERE status = ANY($1)))::int8 - $2, 0)",
+  "describe": {
+    "columns": [
+      {
+        "ordinal": 0,
+        "name": "id",
+        "type_info": "Int8"
+      }
+    ],
+    "parameters": {
+      "Left": [
+        "TextArray",
+        "Int8"
+      ]
+    },
+    "nullable": [
+      false
+    ]
+  },
+  "hash": "1d017ac5f5b1e76ec241533fd4d061c79f4e6d2f1701e727a7474fd5029a5492"
+}

apps/labrinth/migrations/20250608183828_random-project-index.sql

@@ -0,0 +1,12 @@
+-- This index substantially brings down the cost of the query plan for the
+-- hot query at `labrinth::routes::v3::projects::random_projects_get`, from
+-- 354.04..363.39 to 171.33..180.68 (~2x improvement).
+--
+-- The numbers above were calculated in a clean PostgreSQL 17.5.0 container
+-- with 10k mods created with the SQL below.
+--
+-- WITH seq AS (SELECT n FROM GENERATE_SERIES(1, 10000) AS n)
+-- INSERT INTO mods (id, team_id, name, summary, icon_url, license_url, slug, status)
+-- 	SELECT n, 1, n, '', '', '', n, (ARRAY['approved', 'pending'])[n % 2 + 1] from seq;
+
+CREATE INDEX mods_status ON mods(status);

apps/labrinth/src/routes/v3/projects.rs

@@ -94,14 +94,17 @@ pub async fn random_projects_get(
     })?;
 
     let project_ids = sqlx::query!(
-        "
-            SELECT id FROM mods TABLESAMPLE SYSTEM_ROWS($1) WHERE status = ANY($2)
-            ",
-        count.count as i32,
+        // IDs are randomly generated (see the `generate_ids` macro), so ID order is
+        // equivalent to a random order
+        "SELECT id FROM mods WHERE status = ANY($1)
+        ORDER BY id
+        LIMIT $2
+        OFFSET GREATEST(ROUND(RANDOM() * (SELECT COUNT(*) FROM mods WHERE status = ANY($1)))::int8 - $2, 0)",
         &*crate::models::projects::ProjectStatus::iterator()
             .filter(|x| x.is_searchable())
             .map(|x| x.to_string())
             .collect::<Vec<String>>(),
+        count.count as i32,
     )
     .fetch(&**pool)
     .map_ok(|m| db_ids::DBProjectId(m.id))