feat(cli): comments:dedupe cleanup command

Collapse historical duplicate comments produced before the 60-second
dedupe guard landed. Keeps the earliest row in each cluster, reparents
nested replies onto it, moves comment_likes rows, deletes the rest.

Default --window=60 matches the server guard; --dry-run shows the plan
without touching the DB. Transactional per group so a partial failure
does not leave orphaned likes.

Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>

app/Console/Commands/CommentsDedupeCommand.php

@@ -0,0 +1,122 @@
+<?php
+
+namespace App\Console\Commands;
+
+use App\Models\Comment;
+use Illuminate\Console\Command;
+use Illuminate\Support\Facades\DB;
+
+/**
+ * Clean up duplicate comments that pre-date the server-side 60-second dedupe
+ * guard. Duplicate = identical (user_id, video_id, parent_id, body) within a
+ * configurable gap. Keeps the earliest row, moves any likes onto it, deletes
+ * the rest. --dry-run shows what would happen without touching the DB.
+ */
+class CommentsDedupeCommand extends Command
+{
+    protected $signature = 'comments:dedupe
+        {--window=60 : Treat duplicates as same submission if within N seconds (default 60)}
+        {--dry-run : Print what would be removed; do not delete}';
+
+    protected $description = 'Collapse duplicate comments (same user, video, parent, body) into the earliest one';
+
+    public function handle(): int
+    {
+        $window = max(1, (int) $this->option('window'));
+        $dryRun = (bool) $this->option('dry-run');
+
+        // Fetch (user_id, video_id, parent_id, body) groups that appear >1 time.
+        // We ignore the window at the SQL layer and apply it per-group in PHP:
+        // that avoids a correlated subquery and is fast enough for a one-shot
+        // cleanup. For very large tables, chunk by user_id.
+        $groups = DB::table('comments')
+            ->selectRaw('user_id, video_id, parent_id, body, COUNT(*) as cnt')
+            ->groupBy('user_id', 'video_id', 'parent_id', 'body')
+            ->havingRaw('COUNT(*) > 1')
+            ->get();
+
+        if ($groups->isEmpty()) {
+            $this->info('No duplicate (user, video, parent, body) groups found.');
+            return self::SUCCESS;
+        }
+
+        $totalRemoved = 0;
+        $groupsTouched = 0;
+
+        foreach ($groups as $g) {
+            $rows = Comment::where('user_id', $g->user_id)
+                ->where('video_id', $g->video_id)
+                ->where(function ($q) use ($g) {
+                    $g->parent_id === null
+                        ? $q->whereNull('parent_id')
+                        : $q->where('parent_id', $g->parent_id);
+                })
+                ->where('body', $g->body)
+                ->orderBy('created_at')
+                ->orderBy('id')
+                ->get();
+
+            if ($rows->count() < 2) {
+                continue;
+            }
+
+            $keeper = $rows->first();
+            $cluster = collect([$keeper]);
+            $previous = $keeper;
+
+            foreach ($rows->slice(1) as $row) {
+                // Same cluster if row is within window of the previous row —
+                // which is the behavior the server guard enforces today.
+                if ($row->created_at->diffInSeconds($previous->created_at) <= $window) {
+                    $cluster->push($row);
+                    $previous = $row;
+                }
+            }
+
+            if ($cluster->count() < 2) {
+                continue;
+            }
+
+            $toDelete = $cluster->skip(1);
+            $groupsTouched++;
+            $totalRemoved += $toDelete->count();
+
+            $this->line(sprintf(
+                '  user=%d video=%d parent=%s keep=%d remove=%s body="%s"',
+                $g->user_id,
+                $g->video_id,
+                $g->parent_id ?? 'null',
+                $keeper->id,
+                $toDelete->pluck('id')->join(','),
+                mb_strimwidth(preg_replace('/\s+/', ' ', (string) $g->body), 0, 40, '…'),
+            ));
+
+            if ($dryRun) {
+                continue;
+            }
+
+            DB::transaction(function () use ($keeper, $toDelete) {
+                $ids = $toDelete->pluck('id')->all();
+
+                // Reparent any replies that were nested under a doomed row.
+                Comment::whereIn('parent_id', $ids)->update(['parent_id' => $keeper->id]);
+
+                // Move likes to the keeper, dropping dupes.
+                DB::table('comment_likes')
+                    ->whereIn('comment_id', $ids)
+                    ->update(['comment_id' => $keeper->id]);
+
+                Comment::whereIn('id', $ids)->delete();
+            });
+        }
+
+        $this->info(sprintf(
+            '%s %d duplicate rows across %d groups.',
+            $dryRun ? 'Would remove' : 'Removed',
+            $totalRemoved,
+            $groupsTouched,
+        ));
+
+        return self::SUCCESS;
+    }
+}