How are intra-site notifications merged in segmentfault?

I feel that SF's on-site reminder design is very user-friendly. I have several questions after half a day: + merger of notification types (reply to notifications, reply notifications, and likes notifications) + merge the number of unread notifications (the same type of notifications, the number of unread notifications, 10 people with the same topic like, only one unnotified) + if... I feel that SF's site reminds me that the design is very user-friendly. I have several questions after half a day:
+ Merge notification types (reply to notifications, reply to notifications, and likes notifications)
+ Merge the number of unread notifications (the same type of notifications, the number of unread notifications, 10 people with the same topic like, only one not notified)
+ If only 20 pieces of data are retrieved on a page, but these 20 pieces are of the same type, how can we design them in scenarios where notifications need to be merged? There is only one merge, and there may be no problem in merging 20. But if it is a hot topic, Will 1000 replies be merged?

I want to know how SF is designed in these three cases.
If we consider data table splitting, how can we design such a merger?

Reply content:

I feel that SF's site reminds me that the design is very user-friendly. I have several questions after half a day:
+ Merge notification types (reply to notifications, reply to notifications, and likes notifications)
+ Merge the number of unread notifications (the same type of notifications, the number of unread notifications, 10 people with the same topic like, only one not notified)
+ If only 20 pieces of data are retrieved on a page, but these 20 pieces are of the same type, how can we design them in scenarios where notifications need to be merged? There is only one merge, and there may be no problem in merging 20. But if it is a hot topic, Will 1000 replies be merged?

I want to know how SF is designed in these three cases.
If we consider data table splitting, how can we design such a merger?

For OLTP transaction databases, they do not need to be merged during storage. statistics are collected based on various types of notifications when displayed.
Responses, replies, likes, and other events are different. If the user needs such information, they must be saved to the database.

Data Warehouses can collect statistics based on requirements, but determining the storage granularity is crucial. If the granularity is large, details will be lost. Small granularity and large data volume.