在线用户数据存储方案咨询——协作故事写作应用场景

Hey David, great question—real-time, context-aware notifications make all the difference for collaborative storytelling apps, so let’s walk through the optimal end-to-end flow tailored to your use case.

When a user writes and submits a story segment, you want to make sure the data is safely stored before triggering any notifications (no ghost updates or broken alerts). Here’s the play:

• First, validate the submission (check for required fields, content length, basic formatting, etc.)
• Write the story content to your primary database—PostgreSQL with JSONB works great for flexible story data, or MongoDB if you prefer schema-less storage. Include critical fields like author_id, story_id, content, timestamp, and parent_story_id (if it’s a collaborative follow-up to someone else’s work).
• Log a structured event to an event store (either a dedicated table in your main DB or Redis Streams) with metadata needed for notifications: event_type (e.g., story_submitted or collaborative_submission), actor_id, target_story_id, target_author_id (for collaborative cases).

To keep things decoupled and scalable, use an event broker to handle notification triggers instead of tying them directly to the submission code:

• Set up a message queue (RabbitMQ, Kafka) or use Redis Pub/Sub to listen for the story_submitted/collaborative_submission events.
• Route each event to a dedicated notification service that handles audience logic:
  • Scenario 1: Original author submits new content: Fetch all users subscribed to this author from your subscription store (more on that below) and queue notifications for them.
  • Scenario 2: Collaborative subscriber submits content: Fetch the original story author and all users subscribed to the original author/story, then queue notifications for that group.

You need a fast way to get the list of users who need alerts—here’s how to structure it:

• Use a Redis Set for each author, where the key is subscribers:{author_id} and values are the user IDs of people following them. This lets you do O(1) lookups to get the full subscriber list instantly.
• If you also allow subscribing to individual stories (not just authors), add another set: story_subscribers:{story_id}.
• Keep this store in sync with your main database—whenever a user subscribes/unsubscribes, update both the DB record and the Redis Set to avoid stale data.

For that "almost instant" feel, combine WebSockets for online users and push notifications for offline users:

• Online users: Use a WebSocket server (like Socket.io for Node.js, or a managed service like Pusher) to push a notification directly to their connected client. The payload should include a link to the updated story, a content snippet, and the author’s name. Example code snippet for Socket.io:

  // In your notification service
  io.to(`user:${subscriberId}`).emit('story_update', {
    storyId: event.target_story_id,
    authorName: author.name,
    contentSnippet: `${event.content.slice(0, 100)}...`,
    timestamp: event.timestamp
  });
  

• Offline users: Store the notification in a user-specific notifications table in your DB, and send a push notification via APNs (iOS) or FCM (Android). When the user logs back in, fetch all unread notifications from the table.

Don’t forget the details that keep your app running smoothly:

• Idempotency: Add a unique event_id to each submission event, and track processed events to avoid sending duplicate notifications.
• Retry Logic: If a notification fails to send (e.g., WebSocket connection drops), add it to a retry queue with exponential backoff.
• Content Moderation: If your app has content checks, delay triggering notifications until the submission is approved—update the event status to approved and trigger alerts then.
• Unsubscribe Handling: Validate the subscriber list against the latest Redis data before sending to ensure notifications stop immediately when a user unsubscribes.

Hope this flow works for your collaborative storytelling app—if you need help picking specific tools or troubleshooting a part of the pipeline, just ask!

内容的提问来源于stack exchange，提问作者David Nankivell