如何设计并建模用户索引以支持「谁查看过你」与「我查看过谁」查询

Great question! Let's walk through how to extend your existing user index to support both "Who viewed me?" and "Who have I viewed?" queries smoothly.

Core Requirements Recap #

We need to track two directional relationships from the view logs:

• For any user, know which other users have viewed them (viewed_by relationship)
• For any user, know which other users they've viewed (viewed_users relationship)

Index Modeling Solution #

Extend your user index with two array-type fields to store these relationships. Here's a concrete example (using a document-database style mapping, applicable to most search/index systems):

{
  "mappings": {
    "properties": {
      "first_name": { "type": "text" },
      "last_name": { "type": "text" },
      "gender": { "type": "keyword" },
      // Stores IDs of users who have viewed this user
      "viewed_by": { "type": "keyword" },
      // Stores IDs of users this user has viewed
      "viewed_users": { "type": "keyword" }
    }
  }
}

Why This Works #

• Direct Query Support: Each user's document directly holds the data needed for both queries, making lookups fast and simple.
• Scalable Updates: Every view action triggers a targeted update to two documents (no complex joins needed for basic queries).

Update Logic for View Actions #

Whenever a view occurs (e.g., UserA views UserB), you need to update both involved users' documents (with deduplication to avoid duplicate entries from repeated views):

1. For the viewer (UserA): Append UserB's ID to their viewed_users array only if it's not already present.
2. For the viewed user (UserB): Append UserA's ID to their viewed_by array only if it's not already present.

Example update script (pseudocode for clarity):

# Update UserA's viewed_users to add UserB (deduplicated)
if userA.viewed_users does not contain UserB.id:
    userA.viewed_users.append(UserB.id)
    save(userA)

# Update UserB's viewed_by to add UserA (deduplicated)
if userB.viewed_by does not contain UserA.id:
    userB.viewed_by.append(UserA.id)
    save(userB)

Query Examples #

1. "Who viewed me?" (For UserA) #

Fetch UserA's document and extract the viewed_by array. In your example, this would return ["UserD"]—all users who have viewed UserA. If you need full user details (not just IDs), run a multi-get query to pull in profiles for all IDs in the array.

2. "Who have I viewed?" (For UserD) #

Fetch UserD's document and extract the viewed_users array. In your example, this would return ["UserA"]—all users UserD has viewed.

Edge Cases & Optimizations #

• Deduplication: Always avoid adding duplicate user IDs to the arrays—repeated views shouldn't clutter the data.
• Large Arrays: For high-traffic platforms, these arrays could grow large. You might cap the array size (e.g., keep only the last 100 viewers) or use a separate view history table for full archival, while keeping the index arrays for fast recent queries.
• Consistency: Since updates involve two separate documents, handle failures gracefully (e.g., use retries or transactional updates if your system supports them).

内容的提问来源于stack exchange，提问作者Rpj