关于Discoure热门排序

开源 Discourse
1

我发现 Discoure 热门排序 不是按照
image

显示数量感觉是乱排序的

image
image208×164 1.46 KB

我有的154的都在121的下面

不知道是按照什么来排序热门的

最新的排序

有的时候是按照最新发布的时间

有的时候回复帖子也算新发布的时间排序

我不得不去官网找到了一个

按照时间排序的组件

image

2

Discourse 的热门排序不是按照查看数量排序的。

Discourse 的热门排序有算法。

根据官方的说法：Calculating "Top" topics in Discourse - Using Discourse - Discourse Meta

会根据几个因素去计算得出一个值，然后按照这个值去排序。

当然你可以调整权重，当权重调整后显示结果就不一样了。

2024-08-16_07-07-20
2024-08-16_07-07-201117×1152 128 KB

上面还给了一个比较复杂的 SQL。

-- [params]
-- date :start_date = 26 apr 2020
-- date :end_date = 2 may 2020
-- double :log_views_multiplier = 2.0
-- double :first_post_likes_multiplier = 0.5
-- double :least_likes_per_post_multiplier = 3.0

WITH likes AS (
SELECT topic_id, SUM(like_count) AS count
FROM posts
WHERE created_at::date >= :start_date::date 
  AND created_at::date < :end_date::date
  AND deleted_at IS NULL
  AND NOT hidden
  AND post_type = 1
GROUP BY topic_id
),
op_likes AS (
SELECT topic_id, like_count AS count
FROM posts
WHERE created_at::date >= :start_date::date 
  AND created_at::date < :end_date::date
  AND post_number = 1
  AND deleted_at IS NULL
  AND NOT hidden
  AND post_type = 1
),
posts AS (
SELECT topic_id, GREATEST(COUNT(*), 1) AS count
FROM posts
WHERE created_at::date >= :start_date::date 
  AND created_at::date < :end_date::date
  AND deleted_at IS NULL
  AND NOT hidden
  AND post_type = 1
  AND user_id <> 0
GROUP BY topic_id
),
views AS (
SELECT topic_id, COUNT(*) AS count
FROM topic_views
WHERE viewed_at::date >= :start_date::date 
  AND viewed_at::date < :end_date::date
GROUP BY topic_id
),
category_definition_topic_ids AS (
  SELECT COALESCE(topic_id, 0) AS id FROM categories
),
top_topics AS(
SELECT 
topics.id AS topic_id,
topics.title,
topics.user_id, 
posts.count AS date_range_posts,
views.count AS date_range_views,
topics.views AS all_time_views,
topics.bumped_at,
(CASE
 WHEN topics.created_at::date < :start_date::date 
  AND topics.created_at::date >= :end_date::date
 THEN 0
 ELSE log(GREATEST(views.count, 1)) * :log_views_multiplier +
    op_likes.count * :first_post_likes_multiplier +
    CASE WHEN likes.count > 0 AND posts.count > 0
       THEN
        LEAST(likes.count / posts.count, :least_likes_per_post_multiplier)
       ELSE 0
    END +
    CASE WHEN topics.posts_count < 10 THEN
       0 - ((10 - topics.posts_count) / 20) * op_likes.count
    ELSE
       10
    END +
    log(GREATEST(posts.count, 1))
 END) AS score
FROM posts
INNER JOIN views ON posts.topic_id = views.topic_id
INNER JOIN likes ON posts.topic_id = likes.topic_id
INNER JOIN op_likes ON posts.topic_id = op_likes.topic_id
LEFT JOIN topics ON topics.id = posts.topic_id AND topics.deleted_at IS NULL
WHERE topics.deleted_at IS NULL
  AND topics.visible
  AND topics.archetype <> 'private_message'
  AND NOT topics.archived
  AND topics.id NOT IN (SELECT id FROM category_definition_topic_ids)
ORDER BY
  score DESC,
  topics.bumped_at DESC
)

SELECT * FROM top_topics WHERE score > 0

没有怎么仔细研究这个算法，应该还有点复杂。

对站点管理员来说，大致知道几个权重的配置应该就行了。