What is the 'backfill`?

[Lilly (Lillian)]

I’m not fully grasping the backfill condition so I am going get Lola Bot to explain.

[JammyDodger]

@sam

The following badge is triggered on “when a user acts on a post”, this means “delta” applications will get the :post_ids parameter.

SELECT p.user_id, p.id post_id, p.updated_at granted_at
FROM badge_posts p
WHERE p.like_count >= 25 AND
(:backfill OR p.id IN (:post_ids) )

The clause (:backfill OR p.id IN (:post_ids) ) allows us to filter the results. When the daily job runs :backfill is true so the entire set is scanned. When the delta jobs run :backfill is false and :post_ids is set.

I was trying to get it to trigger on When a user acts on a post, but I think my Reactions may be blocking that somehow? Anyway, Update Daily will work, and is less of a performance concern, I think

[Lilly (Lillian)]

Yea it totally makes sense to me now. I think I prefer Sam’s explanation.

Can you explain backfill condition a little bit more Lola?

@gpt4_bot

Certainly, Lilly! In Discourse, the :backfill and :post_ids are parameters used when running badge queries. They are used to control which posts are considered when awarding the badge.

• :backfill is a boolean parameter. When :backfill is true, the badge query is run against all the eligible posts in the database. This is typically used when you first create a badge or when you want to re-calculate who should have the badge based on all historical data.
• :post_ids is a list of post IDs. When :backfill is false, the badge query is only run against the posts whose IDs are in the :post_ids list. This is typically used to check if a badge should be awarded based on a new post or a set of recent posts, rather than all posts.

The condition (:backfill OR p.id IN (:post_ids)) in the WHERE clause of the SQL query ensures that:

• If :backfill is true, the condition will always be true and all posts will be considered (because true OR anything is always true).
• If :backfill is false, only the posts with IDs in the :post_ids list will be considered (because false OR p.id IN (:post_ids) will be true only for the posts whose IDs are in the :post_ids list).

This allows Discourse to efficiently award badges either based on all historical posts or based on a small set of recent posts, depending on the situation.

[JammyDodger]

I think I need to look into which badge queries actually need it adding. I think if you’re using `Update Daily’ it’s not necessary:

@sam

Badge triggers is a complex subject often “daily” updates suffice and all this complex stuff can be skipped.