目录

本站技术栈(V3)

Rails / PostgreSQL / Sidekiq

最后还是用回了我最熟悉的Rails框架,能从容应对我对自己网站日益增长的各种前后端需求。

代码就不开源了,没啥意思。你写得比我好。

器件选型

这套工具链是我用得最顺手的:

实现功能

既然上了RoR,咱就豪放点,不用像 Sinatra实现 那样缩手缩脚的了,直接上Model。

/archives

Model

  • migration
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
# db/schema.rb
create_table "articles", force: :cascade do |t|
  t.string   "title"
  t.string   "subtitle"
  t.string   "type"
  t.text     "markdown_body"
  t.text     "body"
  t.boolean  "hidden",        default: false
  t.datetime "created_at",                    null: false
  t.datetime "updated_at",                    null: false
  t.text     "brief"
  t.boolean  "index_page",    default: false
  t.string   "file_path"
  t.string   "link"
  t.string   "category"
end

kramdown编译好了的markdown body直接扔进数据库(body字段)。其它能初始化好的字段全部存起来,供以后用。

建立搜索索引时现场建立一个plain_text,不另设字段。

  • Article

先来点验证:

1
2
3
4
5
class Article < ActiveRecord::Base
  validates :title, :markdown_body, :category, :file_path, presence: true
  validates :link, uniqueness: true
  validates :brief, length: { maximum: 60 }
end

想到 category 应当是个枚举型:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
class Article < ActiveRecord::Base
  extend Enumerize

  enumerize :category, in: {
    Default: "default",
    ACG: "acg",
    Software: "software",
    Hardware: "hardware",
    Game: "game",
    Network: "network",
    SoC: "soc",
  }, default: :Default
end

从源文件获取meta、编译到HTML均应当自动触发:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
class Article < ActiveRecord::Base
  before_update :init_article
  before_create :init_article

  private
  def init_article
    return if markdown_body.blank?
    self.body = Kramdown::Document.new(
      markdown_body,
      input: 'GFM',
      syntax_highlighter: :coderay,
      syntax_highlighter_opts: {
        span: { disable: true },
        block: { line_numbers: false }
      }
    ).to_html
    self.brief = self.body.gsub(/<.*?>/, "")[0,50] + "..." if self.brief.blank?
    self.link = "/#{self.category.value}/#{File.basename(self.file_path)[0..-4]}"
  end
end
  • Article#list

生成目录供右上角的 ToC 使用。

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
class Article < ActiveRecord::Base
  def self.list
    # 这玩意儿需要遍历数据库且不经常变动,做个缓存
    Rails.cache.fetch(self.cache_key, expires_in: 7.days) do
      result = {  }
      Article.find_each do |article|
        result[article.category.to_sym] = [] if result[article.category.to_sym].blank?
        result[article.category.to_sym] << {
          id: article.id,
          title: article.title,
          link: article.link,
        }
      end
      result
      # { :network => [{ id: 39232, title: "本站技术栈(V3)", link: '/network/tech-stack-ror' }] }
    end
  end

  # 其中,
  def self.cache_key
    "{article.list}" + Article.order("updated_at DESC").first.updated_at.to_i.to_s
  end
end

Controller、View

erb 自带了 content_for 这个注入信息的功能,不用像上次那样纠结了。

1
2
3
Rails.application.routes.draw do
  get 'archives/:category/:title' => "archives#article"
end
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
class ArchivesController < ApplicationController
  def article
    @page = "archives"
    @article = Article.find_by!(link: "/#{params[:category]}/#{params[:title]}")
    @list = Article.list
    @footer = {
      created_at: @article.created_at.to_s,
      updated_at: @article.updated_at.to_s,
    }.to_s
  end
end
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
<!-- app/views/archives/article.erb -->

<div class="row">
  <div class="columns large-12 small-12">
    <article>
      <h1><%= @article.title %>
      <% if @article.subtitle %>
        <small><%= @article.subtitle %></small>
      <% end %>
      <hr>
      </h1>

      <content>
        <%=raw @article.body %>
      </content>
    </article>
  </div>
</div>

自动更新

whenevercron 反人类的语法封装得非常好:

1
2
3
4
5
6
# config/schedule.rb

every 1.day, :at => '3:00 am' do
  runner "Article.find_each(&:update_from_file)"
  # Article.find_each { |article| article.update_from_file }
end

没做监视文件改动的 hot-reload,没想到什么好办法。找了一圈 Guard 的资料感觉它太重……

Twitter archive

我写一个使用API自动清理我的老推的机器人。流程是先更新新推到数据库,再删除早于一个时间窗口的推。

用到了一个强大的gem: <code>twitter</code>

它还有一个CLI工具,就叫 <code>t</code>

Migration

主要是参考Twitter官方的archive导出格式来的,增加了fav和RT的count:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
create_table "tweets", force: :cascade do |t|
  t.string   "tweet_id"
  t.datetime "timestamp"
  t.string   "source"
  t.string   "text"
  t.string   "in_reply_to_status_id"
  t.string   "in_reply_to_user_id"
  t.string   "retweeted_status_id"
  t.string   "retweeted_status_user_id"
  t.datetime "retweeted_status_timestamp"
  t.string   "expanded_urls"
  t.boolean  "deleted",                    default: false
  t.datetime "created_at",                                 null: false
  t.datetime "updated_at",                                 null: false
  t.integer  "favorite_count",             default: 0
  t.integer  "retweet_count",              default: 0
end