discourse/script/import_scripts/question2answer.rb

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

576 lines
18 KiB
Ruby
Raw Normal View History

# frozen_string_literal: true
2018-06-28 09:55:38 +08:00
require 'mysql2'
require File.expand_path(File.dirname(__FILE__) + "/base.rb")
require 'htmlentities'
require 'php_serialize' # https://github.com/jqr/php-serialize
2018-06-28 09:55:38 +08:00
class ImportScripts::Question2Answer < ImportScripts::Base
BATCH_SIZE = 1000
# CHANGE THESE BEFORE RUNNING THE IMPORTER
DB_HOST ||= ENV['DB_HOST'] || "localhost"
DB_NAME ||= ENV['DB_NAME'] || "qa_db"
DB_PW ||= ENV['DB_PW'] || ""
DB_USER ||= ENV['DB_USER'] || "root"
2018-06-28 09:55:38 +08:00
TIMEZONE ||= ENV['TIMEZONE'] || "America/Los_Angeles"
TABLE_PREFIX ||= ENV['TABLE_PREFIX'] || "qa_"
def initialize
super
@old_username_to_new_usernames = {}
@tz = TZInfo::Timezone.get(TIMEZONE)
@htmlentities = HTMLEntities.new
@client = Mysql2::Client.new(
host: DB_HOST,
username: DB_USER,
password: DB_PW,
database: DB_NAME
)
end
def execute
import_users
import_categories
import_topics
import_posts
import_likes
import_bestanswer
2018-06-28 09:55:38 +08:00
post_process_posts
create_permalinks
end
def import_users
puts "", "importing users"
# only import users that have posted or voted on Q2A
# if you want to import all users, just leave out the WHERE and everything after it (and remove line 95 as well)
user_count = mysql_query("SELECT COUNT(userid) count FROM #{TABLE_PREFIX}users u WHERE EXISTS (SELECT 1 FROM #{TABLE_PREFIX}posts p WHERE p.userid=u.userid) or EXISTS (SELECT 1 FROM #{TABLE_PREFIX}uservotes uv WHERE u.userid=uv.userid)").first["count"]
2018-06-28 09:55:38 +08:00
last_user_id = -1
batches(BATCH_SIZE) do |offset|
users = mysql_query(<<-SQL
SELECT u.userid AS id, u.email, u.handle AS username, u.created AS created_at, u.loggedin AS last_sign_in_at, u.avatarblobid
FROM #{TABLE_PREFIX}users u
WHERE u.userid > #{last_user_id}
AND (EXISTS (SELECT 1 FROM #{TABLE_PREFIX}posts p WHERE p.userid=u.userid) or EXISTS (SELECT 1 FROM #{TABLE_PREFIX}uservotes uv WHERE u.userid=uv.userid))
ORDER BY u.userid
LIMIT #{BATCH_SIZE}
2018-06-28 09:55:38 +08:00
SQL
).to_a
break if users.empty?
last_user_id = users[-1]["id"]
users.reject! { |u| @lookup.user_already_imported?(u["id"].to_i) }
create_users(users, total: user_count, offset: offset) do |user|
email = user["email"].presence
username = @htmlentities.decode(user["email"]).strip.split("@").first
avatar_url = "https://your_image_bucket/#{user['cdn_slug']}" if user['cdn_slug']
{
id: user["id"],
name: "#{user['username']}",
username: "#{user['username']}",
password: user['password'],
2018-06-28 09:55:38 +08:00
email: email,
created_at: user["created_at"],
last_seen_at: user["last_sign_in_at"],
post_create_action: proc do |u|
@old_username_to_new_usernames[user["username"]] = u.username
end
}
end
end
end
def import_categories
puts "", "importing top level categories..."
categories = mysql_query("SELECT categoryid, parentid, title, position FROM #{TABLE_PREFIX}categories ORDER BY categoryid").to_a
top_level_categories = categories.select { |c| c["parentid"].nil? }
create_categories(top_level_categories) do |category|
{
id: category["categoryid"],
name: @htmlentities.decode(category["title"]).strip,
position: category["position"]
}
end
puts "", "importing children categories..."
children_categories = categories.select { |c| !c["parentid"].nil? }
top_level_category_ids = Set.new(top_level_categories.map { |c| c["categoryid"] })
# cut down the tree to only 2 levels of categories
children_categories.each do |cc|
while !top_level_category_ids.include?(cc["parentid"])
cc["parentid"] = categories.detect { |c| c["categoryid"] == cc["parentid"] }["parentid"]
end
end
create_categories(children_categories) do |category|
{
id: category["categoryid"],
name: @htmlentities.decode(category["title"]).strip,
position: category["position"],
parent_category_id: category_id_from_imported_category_id(category["parentid"])
}
end
end
def import_topics
puts "", "importing topics..."
topic_count = mysql_query("SELECT COUNT(postid) count FROM #{TABLE_PREFIX}posts WHERE type = 'Q'").first["count"]
2018-06-28 09:55:38 +08:00
last_topic_id = -1
batches(BATCH_SIZE) do |offset|
topics = mysql_query(<<-SQL
SELECT p.postid, p.type, p.categoryid, p.closedbyid, p.userid postuserid, p.views, p.created, p.title, p.content raw
FROM #{TABLE_PREFIX}posts p
WHERE type = 'Q'
AND p.postid > #{last_topic_id}
2018-06-28 09:55:38 +08:00
ORDER BY p.postid
LIMIT #{BATCH_SIZE}
SQL
).to_a
break if topics.empty?
last_topic_id = topics[-1]["postid"]
topics.reject! { |t| @lookup.post_already_imported?("thread-#{t["postid"]}") }
topics.reject! { |t| t["type"] == "Q_HIDDEN" }
2018-06-28 09:55:38 +08:00
create_posts(topics, total: topic_count, offset: offset) do |topic|
begin
raw = preprocess_post_raw(topic["raw"])
rescue => e
puts e.message
end
topic_id = "thread-#{topic["postid"]}"
t = {
id: topic_id,
user_id: user_id_from_imported_user_id(topic["postuserid"]) || Discourse::SYSTEM_USER_ID,
title: @htmlentities.decode(topic["title"]).strip[0...255],
category: category_id_from_imported_category_id(topic["categoryid"]),
raw: raw,
created_at: topic["created"],
visible: topic["closedbyid"].to_i == 0,
2018-06-28 09:55:38 +08:00
views: topic["views"],
}
t
end
# uncomment below lines to create permalink
topics.each do |thread|
topic_id = "thread-#{thread["postid"]}"
topic = topic_lookup_from_imported_post_id(topic_id)
if topic.present?
title_slugified = slugify(thread["title"], false, 50) if thread["title"].present?
url_slug = "qa/#{thread["postid"]}/#{title_slugified}" if thread["title"].present?
2018-06-28 09:55:38 +08:00
Permalink.create(url: url_slug, topic_id: topic[:topic_id].to_i) if url_slug.present? && topic[:topic_id].present?
end
end
end
end
def slugify(title, ascii_only, max_length)
words = title.downcase.gsub(/[^a-zA-Z0-9\s]/, '').split(" ")
word_lengths = {}
words.each_with_index do |word, idx|
word_lengths[idx] = word.length
end
remaining = max_length
2018-06-29 14:04:38 +08:00
if word_lengths.inject(0) { |sum, (_, v)| sum + v } > remaining
word_lengths = Hash[word_lengths.sort { |x, y| y[1] <=> x[1] }]
2018-06-28 09:55:38 +08:00
word_lengths.each do |idx, word_length|
if remaining > 0
remaining -= word_length
else
words[idx] = nil
end
end
end
words = words.compact.join("-")
end
def import_posts
puts "", "importing posts..."
post_count = mysql_query(<<-SQL
SELECT COUNT(postid) count
FROM #{TABLE_PREFIX}posts p
WHERE p.parentid IS NOT NULL
SQL
).first["count"]
last_post_id = -1
batches(BATCH_SIZE) do |offset|
posts = mysql_query(<<-SQL
SELECT p.postid, p.type, p.parentid, p.categoryid, p.closedbyid, p.userid, p.views, p.created, p.title, p.content,
parent.type AS parenttype, parent.parentid AS qid
2018-06-28 09:55:38 +08:00
FROM #{TABLE_PREFIX}posts p
LEFT JOIN qa_posts parent ON parent.postid = p.parentid
2018-06-28 09:55:38 +08:00
WHERE p.parentid IS NOT NULL
AND p.postid > #{last_post_id}
AND p.type in ('A','C')
AND p.closedbyid IS NULL
2018-06-28 09:55:38 +08:00
ORDER BY p.postid
LIMIT #{BATCH_SIZE}
SQL
).to_a
break if posts.empty?
last_post_id = posts[-1]["postid"]
posts.reject! { |p| @lookup.post_already_imported?(p["postid"].to_i) }
create_posts(posts, total: post_count, offset: offset) do |post|
begin
2018-06-29 14:04:38 +08:00
raw = preprocess_post_raw(post["content"])
2018-06-28 09:55:38 +08:00
rescue => e
puts e.message
end
next if raw.blank?
# this works as long as comments can not have a comment as parent
# it's always Q-A Q-C or A-C
if post['type'] == 'A' # for answers the question/topic is always the parent
topic = topic_lookup_from_imported_post_id("thread-#{post["parentid"]}")
next if topic.nil?
else
if post['parenttype'] == 'Q' # for comments to questions, the question/topic is the parent as well
topic = topic_lookup_from_imported_post_id("thread-#{post["parentid"]}")
next if topic.nil?
else # for comments to answers, the question/topic is the parent of the parent
topic = topic_lookup_from_imported_post_id("thread-#{post["qid"]}")
next if topic.nil?
end
end
next if topic.nil?
2018-06-28 09:55:38 +08:00
p = {
id: post["postid"],
user_id: user_id_from_imported_user_id(post["userid"]) || Discourse::SYSTEM_USER_ID,
topic_id: topic[:topic_id],
raw: raw,
created_at: post["created"],
}
if parent = topic_lookup_from_imported_post_id(post["parentid"])
p[:reply_to_post_number] = parent[:post_number]
end
p
end
end
end
def import_bestanswer
puts "", "importing best answers..."
ans = mysql_query("select postid, selchildid from qa_posts where selchildid is not null").to_a
ans.each do |answer|
begin
post = Post.find_by(id: post_id_from_imported_post_id("#{answer['selchildid']}"))
post.custom_fields["is_accepted_answer"] = "true"
post.save
topic = Topic.find(post.topic_id)
topic.custom_fields["accepted_answer_post_id"] = post.id
topic.save
rescue => e
puts "error acting on post #{e}"
end
end
end
2018-06-28 09:55:38 +08:00
def import_likes
puts "", "importing likes..."
likes = mysql_query(<<-SQL
SELECT postid, userid
FROM #{TABLE_PREFIX}uservotes u
WHERE u.vote=1
2018-06-29 14:04:38 +08:00
SQL
).to_a
2018-06-28 09:55:38 +08:00
likes.each do |like|
post = Post.find_by(id: post_id_from_imported_post_id("thread-#{like['postid']}"))
user = User.find_by(id: user_id_from_imported_user_id(like["userid"]))
begin
PostActionCreator.like(user, post) if user && post
2018-06-29 14:04:38 +08:00
rescue => e
puts "error acting on post #{e}"
end
2018-06-28 09:55:38 +08:00
end
end
def post_process_posts
puts "", "Postprocessing posts..."
current = 0
max = Post.count
Post.find_each do |post|
begin
new_raw = postprocess_post_raw(post.raw)
if new_raw != post.raw
post.raw = new_raw
post.save
end
rescue PrettyText::JavaScriptError
nil
ensure
print_status(current += 1, max)
end
end
end
def preprocess_post_raw(raw)
return "" if raw.blank?
raw.gsub!(/<a(?:.+)href="(\S+)"(?:.*)>(.+)<\/a>/i, '[\2](\1)')
raw.gsub!(/<p>(.+?)<\/p>/im) { "#{$1}\n\n" }
raw.gsub!('<br />', "\n")
raw.gsub!(/<strong>(.*?)<\/strong>/im, '[b]\1[/b]')
2018-06-28 09:55:38 +08:00
# decode HTML entities
raw = @htmlentities.decode(raw)
raw = ActionView::Base.full_sanitizer.sanitize raw
# fix whitespaces
raw.gsub!(/(\\r)?\\n/, "\n")
raw.gsub!("\\t", "\t")
# [HTML]...[/HTML]
raw.gsub!(/\[html\]/i, "\n```html\n")
raw.gsub!(/\[\/html\]/i, "\n```\n")
# [PHP]...[/PHP]
raw.gsub!(/\[php\]/i, "\n```php\n")
raw.gsub!(/\[\/php\]/i, "\n```\n")
# [HIGHLIGHT="..."]
raw.gsub!(/\[highlight="?(\w+)"?\]/i) { "\n```#{$1.downcase}\n" }
# [CODE]...[/CODE]
# [HIGHLIGHT]...[/HIGHLIGHT]
raw.gsub!(/\[\/?code\]/i, "\n```\n")
raw.gsub!(/\[\/?highlight\]/i, "\n```\n")
# [SAMP]...[/SAMP]
raw.gsub!(/\[\/?samp\]/i, "`")
# replace all chevrons with HTML entities
# NOTE: must be done
# - AFTER all the "code" processing
# - BEFORE the "quote" processing
raw.gsub!(/`([^`]+)`/im) { "`" + $1.gsub("<", "\u2603") + "`" }
raw.gsub!("<", "&lt;")
raw.gsub!("\u2603", "<")
raw.gsub!(/`([^`]+)`/im) { "`" + $1.gsub(">", "\u2603") + "`" }
raw.gsub!(">", "&gt;")
raw.gsub!("\u2603", ">")
# [URL=...]...[/URL]
raw.gsub!(/\[url="?([^"]+?)"?\](.*?)\[\/url\]/im) { "[#{$2.strip}](#{$1})" }
raw.gsub!(/\[url="?(.+?)"?\](.+)\[\/url\]/im) { "[#{$2.strip}](#{$1})" }
# [URL]...[/URL]
# [MP3]...[/MP3]
raw.gsub!(/\[\/?url\]/i, "")
raw.gsub!(/\[\/?mp3\]/i, "")
# [MENTION]<username>[/MENTION]
raw.gsub!(/\[mention\](.+?)\[\/mention\]/i) do
old_username = $1
if @old_username_to_new_usernames.has_key?(old_username)
old_username = @old_username_to_new_usernames[old_username]
end
"@#{old_username}"
end
# [FONT=blah] and [COLOR=blah]
2018-06-29 14:04:38 +08:00
raw.gsub!(/\[FONT=.*?\](.*?)\[\/FONT\]/im, '\1')
raw.gsub!(/\[COLOR=.*?\](.*?)\[\/COLOR\]/im, '\1')
raw.gsub!(/\[COLOR=#.*?\](.*?)\[\/COLOR\]/im, '\1')
2018-06-28 09:55:38 +08:00
2018-06-29 14:04:38 +08:00
raw.gsub!(/\[SIZE=.*?\](.*?)\[\/SIZE\]/im, '\1')
raw.gsub!(/\[h=.*?\](.*?)\[\/h\]/im, '\1')
2018-06-28 09:55:38 +08:00
# [CENTER]...[/CENTER]
2018-06-29 14:04:38 +08:00
raw.gsub!(/\[CENTER\](.*?)\[\/CENTER\]/im, '\1')
2018-06-28 09:55:38 +08:00
# [INDENT]...[/INDENT]
2018-06-29 14:04:38 +08:00
raw.gsub!(/\[INDENT\](.*?)\[\/INDENT\]/im, '\1')
raw.gsub!(/\[TABLE\](.*?)\[\/TABLE\]/im, '\1')
raw.gsub!(/\[TR\](.*?)\[\/TR\]/im, '\1')
raw.gsub!(/\[TD\](.*?)\[\/TD\]/im, '\1')
raw.gsub!(/\[TD="?.*?"?\](.*?)\[\/TD\]/im, '\1')
2018-06-28 09:55:38 +08:00
# [QUOTE]...[/QUOTE]
raw.gsub!(/\[quote\](.+?)\[\/quote\]/im) { |quote|
quote.gsub!(/\[quote\](.+?)\[\/quote\]/im) { "\n#{$1}\n" }
quote.gsub!(/\n(.+?)/) { "\n> #{$1}" }
}
# [QUOTE=<username>]...[/QUOTE]
raw.gsub!(/\[quote=([^;\]]+)\](.+?)\[\/quote\]/im) do
old_username, quote = $1, $2
if @old_username_to_new_usernames.has_key?(old_username)
old_username = @old_username_to_new_usernames[old_username]
end
"\n[quote=\"#{old_username}\"]\n#{quote}\n[/quote]\n"
end
# [YOUTUBE]<id>[/YOUTUBE]
raw.gsub!(/\[youtube\](.+?)\[\/youtube\]/i) { "\n//youtu.be/#{$1}\n" }
# [VIDEO=youtube;<id>]...[/VIDEO]
raw.gsub!(/\[video=youtube;([^\]]+)\].*?\[\/video\]/i) { "\n//youtu.be/#{$1}\n" }
# More Additions ....
# [spoiler=Some hidden stuff]SPOILER HERE!![/spoiler]
raw.gsub!(/\[spoiler="?(.+?)"?\](.+?)\[\/spoiler\]/im) { "\n#{$1}\n[spoiler]#{$2}[/spoiler]\n" }
# [IMG][IMG]http://i63.tinypic.com/akga3r.jpg[/IMG][/IMG]
raw.gsub!(/\[IMG\]\[IMG\](.+?)\[\/IMG\]\[\/IMG\]/i) { "[IMG]#{$1}[/IMG]" }
# convert list tags to ul and list=1 tags to ol
# (basically, we're only missing list=a here...)
# (https://meta.discourse.org/t/phpbb-3-importer-old/17397)
raw.gsub!(/\[list\](.*?)\[\/list\]/im, '[ul]\1[/ul]')
raw.gsub!(/\[list=1\](.*?)\[\/list\]/im, '[ol]\1[/ol]')
raw.gsub!(/\[list\](.*?)\[\/list:u\]/im, '[ul]\1[/ul]')
raw.gsub!(/\[list=1\](.*?)\[\/list:o\]/im, '[ol]\1[/ol]')
# convert *-tags to li-tags so bbcode-to-md can do its magic on phpBB's lists:
raw.gsub!(/\[\*\]\n/, '')
raw.gsub!(/\[\*\](.*?)\[\/\*:m\]/, '[li]\1[/li]')
raw.gsub!(/\[\*\](.*?)\n/, '[li]\1[/li]')
raw.gsub!(/\[\*=1\]/, '')
raw.strip!
raw
end
def postprocess_post_raw(raw)
# [QUOTE=<username>;<post_id>]...[/QUOTE]
raw.gsub!(/\[quote=([^;]+);(\d+)\](.+?)\[\/quote\]/im) do
old_username, post_id, quote = $1, $2, $3
if @old_username_to_new_usernames.has_key?(old_username)
old_username = @old_username_to_new_usernames[old_username]
end
if topic_lookup = topic_lookup_from_imported_post_id(post_id)
post_number = topic_lookup[:post_number]
topic_id = topic_lookup[:topic_id]
"\n[quote=\"#{old_username},post:#{post_number},topic:#{topic_id}\"]\n#{quote}\n[/quote]\n"
else
"\n[quote=\"#{old_username}\"]\n#{quote}\n[/quote]\n"
end
end
# remove attachments
raw.gsub!(/\[attach[^\]]*\]\d+\[\/attach\]/i, "")
# [THREAD]<thread_id>[/THREAD]
# ==> http://my.discourse.org/t/slug/<topic_id>
raw.gsub!(/\[thread\](\d+)\[\/thread\]/i) do
thread_id = $1
if topic_lookup = topic_lookup_from_imported_post_id("thread-#{thread_id}")
topic_lookup[:url]
else
$&
end
end
# [THREAD=<thread_id>]...[/THREAD]
# ==> [...](http://my.discourse.org/t/slug/<topic_id>)
raw.gsub!(/\[thread=(\d+)\](.+?)\[\/thread\]/i) do
thread_id, link = $1, $2
if topic_lookup = topic_lookup_from_imported_post_id("thread-#{thread_id}")
url = topic_lookup[:url]
"[#{link}](#{url})"
else
$&
end
end
# [POST]<post_id>[/POST]
# ==> http://my.discourse.org/t/slug/<topic_id>/<post_number>
raw.gsub!(/\[post\](\d+)\[\/post\]/i) do
post_id = $1
if topic_lookup = topic_lookup_from_imported_post_id(post_id)
topic_lookup[:url]
else
$&
end
end
# [POST=<post_id>]...[/POST]
# ==> [...](http://my.discourse.org/t/<topic_slug>/<topic_id>/<post_number>)
raw.gsub!(/\[post=(\d+)\](.+?)\[\/post\]/i) do
post_id, link = $1, $2
if topic_lookup = topic_lookup_from_imported_post_id(post_id)
url = topic_lookup[:url]
"[#{link}](#{url})"
else
$&
end
end
raw
end
def create_permalinks
puts '', 'Creating permalinks...'
# topics
Topic.find_each do |topic|
tcf = topic.custom_fields
if tcf && tcf["import_id"]
question_id = tcf["import_id"][/thread-(\d)/, 0]
url = "#{question_id}"
Permalink.create(url: url, topic_id: topic.id) rescue nil
end
end
# categories
2018-06-28 09:55:38 +08:00
Category.find_each do |category|
ccf = category.custom_fields
if ccf && ccf["import_id"]
url = category.parent_category ? "#{category.parent_category.slug}/#{category.slug}" : category.slug
Permalink.create(url: url, category_id: category.id) rescue nil
end
end
2018-06-28 09:55:38 +08:00
end
def parse_timestamp(timestamp)
Time.zone.at(@tz.utc_to_local(timestamp))
end
def mysql_query(sql)
@client.query(sql, cache_rows: true)
end
end
ImportScripts::Question2Answer.new.perform