From 71e8a93a71d3b2283fbec1316ade9a9509a930db Mon Sep 17 00:00:00 2001
From: Your Name <you@example.com>
Date: Fri, 25 Sep 2026 13:56:55 +0930
Subject: [PATCH] Dedup discover pool with aggregate to satisfy strict group by

---
 app/Services/DiscoverService.php              | 13 ++-
 .../Feature/DiscoverServiceDailyPoolTest.php  | 85 +++++++++++++++++++
 2 files changed, 94 insertions(+), 4 deletions(-)
 create mode 100644 tests/Feature/DiscoverServiceDailyPoolTest.php

diff --git a/app/Services/DiscoverService.php b/app/Services/DiscoverService.php
index c86723335..72d9346be 100644
--- a/app/Services/DiscoverService.php
+++ b/app/Services/DiscoverService.php
@@ -10,17 +10,22 @@ class DiscoverService
     public static function getDailyIdPool()
     {
         $min_id = SnowflakeService::byDate(now()->subMonths(3));
-        $sqld = db_is_mysql_maria();
 
+        // One post per author (the newest, MAX(id)) so the pool is not
+        // dominated by prolific accounts. Selecting the aggregated MAX(id) with
+        // GROUP BY profile_id is valid under ONLY_FULL_GROUP_BY and portable
+        // across every driver — the previous `select id ... group by profile_id`
+        // threw 1055 on strict MySQL/MariaDB, and was gated off entirely on
+        // other drivers (no dedup at all). inRandomOrder() still shuffles the
+        // pool. This mirrors the MAX(id)+GROUP BY dedup used elsewhere.
         return DB::table('statuses')
+            ->selectRaw('MAX(id) as id')
             ->whereNull('uri')
             ->whereType('photo')
             ->whereIsNsfw(false)
             ->whereVisibility('public')
-            ->when($sqld, function ($q, $sqld) {
-                return $q->groupBy('profile_id');
-            })
             ->where('id', '>', $min_id)
+            ->groupBy('profile_id')
             ->inRandomOrder()
             ->take(300)
             ->pluck('id');
diff --git a/tests/Feature/DiscoverServiceDailyPoolTest.php b/tests/Feature/DiscoverServiceDailyPoolTest.php
new file mode 100644
index 000000000..d60e09e5c
--- /dev/null
+++ b/tests/Feature/DiscoverServiceDailyPoolTest.php
@@ -0,0 +1,85 @@
+<?php
+
+use App\Models\Status;
+use App\Models\User;
+use App\Services\DiscoverService;
+use Illuminate\Foundation\Testing\LazilyRefreshDatabase;
+use Illuminate\Support\Facades\Redis;
+
+uses(LazilyRefreshDatabase::class);
+
+/*
+|--------------------------------------------------------------------------
+| DiscoverService daily id pool
+|--------------------------------------------------------------------------
+|
+| getDailyIdPool() de-duplicates to one photo per author. The old query did
+| `select id ... group by profile_id`, which throws ERROR 1055 under
+| ONLY_FULL_GROUP_BY (every strict MySQL/MariaDB deployment) and, on drivers
+| where it did run, was gated off entirely so prolific authors flooded the
+| pool. It now uses MAX(id)+GROUP BY: valid on every driver, one newest post
+| per author.
+|
+*/
+
+beforeEach(function () {
+    Redis::spy();
+});
+
+function publicPhoto(int $profileId): Status
+{
+    return Status::factory()->create([
+        'profile_id' => $profileId,
+        'type' => 'photo',
+        'scope' => 'public',
+        'visibility' => 'public',
+        'is_nsfw' => false,
+        'uri' => null,
+        'local' => true,
+    ]);
+}
+
+it('returns exactly one id per author', function () {
+    $author = User::factory()->create();
+    $author->refresh();
+
+    // Several public photos by the same author.
+    publicPhoto($author->profile_id);
+    publicPhoto($author->profile_id);
+    publicPhoto($author->profile_id);
+
+    $pool = DiscoverService::getDailyIdPool();
+
+    expect($pool)->toHaveCount(1);
+});
+
+it('keeps the newest post per author', function () {
+    $author = User::factory()->create();
+    $author->refresh();
+
+    $ids = collect([
+        publicPhoto($author->profile_id)->id,
+        publicPhoto($author->profile_id)->id,
+        publicPhoto($author->profile_id)->id,
+    ]);
+    $newestId = $ids->max();
+
+    $pool = DiscoverService::getDailyIdPool();
+
+    expect($pool->map(fn ($id) => (int) $id)->all())->toBe([(int) $newestId]);
+});
+
+it('includes one id from each of several authors', function () {
+    $a = User::factory()->create();
+    $a->refresh();
+    $b = User::factory()->create();
+    $b->refresh();
+
+    publicPhoto($a->profile_id);
+    publicPhoto($a->profile_id);
+    publicPhoto($b->profile_id);
+
+    $pool = DiscoverService::getDailyIdPool()->map(fn ($id) => (int) $id);
+
+    expect($pool)->toHaveCount(2);
+});