Post
Here are the details copied from that post:
Why the campground had become so slow, and how we fixed that (browsing feeds is now up to 95% faster!):
When you're browsing your feed on Bonfire, each page on the app displays 20 activities, but there are tens of thousands of activities in total. Fetching all the activities at once, including related data like their authors and threads, and checking permissions for each activity can be slow and inefficient.
1/...
To address this challenge, we implemented deferred joins and boundary checking. Instead of fetching all the data and performing boundary checks for every activity before selecting the ones for a specific page, we defer these operations until after an initial pagination has been applied.
2/...
Here's how it works: First, the app creates an optimized query that loads only the necessary information needed for pagination. This includes filtering activities based on a specific feed, ensuring that only relevant activities are considered. This optimized query is designed to load less data from disk and thus is faster. It returns just the IDs of one or two pages, which could be up to 40 activities.
3/...
Next, the database takes the results of the optimized query and fetches the complete details of those activities. At this stage, it also computes the boundaries for each of those 40 activities, checking permissions to determine if the user has access to view them. The database filters out any activities that the user is not allowed to see. Finally, another round of pagination is applied to the filtered activities, ensuring that 1 to 20 activities are returned as the final result.
4/...
By deferring the retrieval of complete details and boundary computation until after pagination, the database minimizes the amount of data it needs to process. This significantly improves performance, even when the instance contains a large number of activities. It reduces the time and resources required to handle pagination, resulting in a much faster and more responsive user experience.
5/...
For those familiar with SQL, this looks something like (though the real query is much more complex):
```
SELECT * from activities
INNER JOIN (SELECT id from activities WHERE [...] LIMIT 40)
LEFT OUTER JOIN [...]
WHERE EXISTS (SELECT [...] WHERE boundaries.id = activities.id)
LIMIT 20
```
6/end
A space for Bonfire maintainers and contributors to communicate
