discourse/app/models/hot_topic.rb

148 lines
6.7 KiB
Ruby

class HotTopic < ActiveRecord::Base
belongs_to :topic
belongs_to :category
# Here's the current idea behind the implementaiton of hot: random can produce good results!
# Hot is currently made up of a random selection of high percentile topics. It includes mostly
# new topics, but also some old ones for variety.
def self.refresh!
transaction do
exec_sql "DELETE FROM hot_topics"
# TODO, move these to site settings once we're sure this is how we want to figure out hot
max_hot_topics = 200 # how many hot topics we want
hot_percentile = 0.2 # What percentile of topics we consider good
older_percentage = 0.2 # how many old topics we want as a percentage
new_days = 21 # how many days old we consider old
no_old_in_first_x_rows = 8 # don't show old results in the first x rows
# Include all sticky uncategorized on Hot
exec_sql("INSERT INTO hot_topics (topic_id,
random_bias,
random_multiplier,
days_ago_bias,
days_ago_multiplier,
score,
hot_topic_type)
SELECT t.id,
calc.random_bias,
1.0,
0,
1.0,
calc.random_bias,
1
FROM topics AS t
INNER JOIN (SELECT id, RANDOM() as random_bias
FROM topics) AS calc ON calc.id = t.id
WHERE t.deleted_at IS NULL
AND t.visible
AND (NOT t.archived)
AND t.pinned_at IS NOT NULL
AND t.category_id IS NULL")
# Include high percentile recent topics
inserted_count = exec_sql("INSERT INTO hot_topics (topic_id,
category_id,
random_bias,
random_multiplier,
days_ago_bias,
days_ago_multiplier,
score,
hot_topic_type)
SELECT t.id,
t.category_id,
calc.random_bias,
0.05,
calc.days_ago_bias,
0.95,
(calc.random_bias * 0.05) + (days_ago_bias * 0.95),
2
FROM topics AS t
INNER JOIN (SELECT id,
RANDOM() as random_bias,
((1.0 - (EXTRACT(EPOCH FROM CURRENT_TIMESTAMP-created_at)/86400) / :days_ago) * 0.95) AS days_ago_bias
FROM topics) AS calc ON calc.id = t.id
WHERE t.deleted_at IS NULL
AND t.visible
AND (NOT t.closed)
AND (NOT t.archived)
AND t.pinned_at IS NULL
AND t.archetype <> :private_message
AND created_at >= (CURRENT_TIMESTAMP - INTERVAL ':days_ago' DAY)
AND t.percent_rank < :hot_percentile
AND NOT EXISTS(SELECT * FROM hot_topics AS ht2 WHERE ht2.topic_id = t.id)
LIMIT :limit",
hot_percentile: hot_percentile,
limit: ((1.0 - older_percentage) * max_hot_topics).round,
private_message: Archetype::private_message,
days_ago: new_days)
max_old_score = 1.0
# Finding the highest score in the first x rows
if HotTopic.count > no_old_in_first_x_rows
max_old_score = HotTopic.order('score desc').limit(no_old_in_first_x_rows).last.score
end
# Add a sprinkling of random older topics
exec_sql("INSERT INTO hot_topics (topic_id,
category_id,
random_bias,
random_multiplier,
days_ago_bias,
days_ago_multiplier,
score,
hot_topic_type)
SELECT t.id,
t.category_id,
calc.random_bias,
:max_old_score,
0,
1.0,
calc.random_bias * :max_old_score,
3
FROM topics AS t
INNER JOIN (SELECT id, RANDOM() as random_bias
FROM topics) AS calc ON calc.id = t.id
WHERE t.deleted_at IS NULL
AND t.visible
AND (NOT t.closed)
AND (NOT t.archived)
AND t.pinned_at IS NULL
AND t.archetype <> :private_message
AND created_at < (CURRENT_TIMESTAMP - INTERVAL ':days_ago' DAY)
AND t.percent_rank < :hot_percentile
AND NOT EXISTS(SELECT * FROM hot_topics AS ht2 WHERE ht2.topic_id = t.id)
LIMIT :limit",
hot_percentile: hot_percentile,
limit: (older_percentage * max_hot_topics).round,
private_message: Archetype::private_message,
days_ago: new_days,
max_old_score: max_old_score)
end
end
end
# == Schema Information
#
# Table name: hot_topics
#
# id :integer not null, primary key
# topic_id :integer not null
# category_id :integer
# score :float not null
# random_bias :float
# random_multiplier :float
# days_ago_bias :float
# days_ago_multiplier :float
# hot_topic_type :integer
#
# Indexes
#
# index_hot_topics_on_score (score)
# index_hot_topics_on_topic_id (topic_id) UNIQUE
#