Migrating from Drupal to Jekyll - antonizoon/antonizoon.github.io GitHub Wiki

The Bibliotheca Anonoma has put together a brand new, Jekyll (and Minimal Mistakes)-based website for the Yotsuba Society, which beats the rotting Drupal 7.x CMS it used to use.

And don't worry, all the old pagelinks and downloads still work (we worked hard to make that possible), so no need to change your bookmarks.

Situation

The Yotsuba Society, recently discovered that their website was getting spam sites injected into their aging Drupal 7.x installation.

Yes, with these messy, ancient CMS codebases, if you fail to update, security is like swiss cheese. It's a matter of time before your site gets exploited. And even if you think your site has nothing of value, it has valuable server space that spam sites can plaster their junk into.

The admin's original plan was simply to shut down the site indefinitely, but that would have meant the total death of the Yotsuba Society. A great example that sometimes, even the archivists have to archived. So I instead recommended that they abandon the CMS system entirely, and replace it with a Static HTML engine (such as Jekyll).

The CMS is often a necessary evil in large, account-based websites which implement forums, rich content, multiple posters, and such. However, by now the Yotsuba Society had significantly reduced it's operations, and was only using the CMS as a blog. Because Drupal 7 was now legacy code, it was starting to get spam ads injected into it to trick some poor saps.

Not everything has to be a giant Wordpress/Drupal/Joomla site. If you're only pushing information in a blog and some pages, perhaps with a few Disqus comments on the side, all you need is a static HTML engine, such as Jekyll.

• No more complex interfaces to wrestle with. Just write a page and push (via FTP or Git).
• No more bloated WYSIYWG editors. Just write in Markdown syntax, it's easy to learn and has massive potential. Dump YouTube embed codes in, dump internal image links, and be done with it.
• No more security issues. Bloated CMS systems, with their accounts and SQL injection vulnerabilities, offer a massive attack surface, especially if it's not well updated. With Static HTML, there's nothing for attackers to attack, other than the web server admin console.
• Static HTML uses significantly fewer resources. You don't need a PHP engine or SQL Database to generate pages on the fly anymore. The HTML pages are pregenerated on the author's computer, and just thrown at the viewer. In fact, you can serve them for free via Github Pages (which allows you to use your same ol' custom domains).

And yes, Static HTML engines support comments too, via Disqus.

Migration Rules

First, we must lay down some ground rules for website migration.

• NEVER cause link rot. - Isn't it annoying when your bookmarked links no longer work? Even if the original content has been imported, you will break links if you change them to a different URL structure. Link rot causes a whole host of issues, from reduced accessibility to reduced Google listing presence.
  • We have to make sure that links such as http://yotsubasociety.org/node/1 still work, or at least redirect in the new site. Luckily, Jekyll has a permalink: option to set custom URLs, and a Jekyll Redirect-from plugin to create multiple redirects. Say goodbye to link rot.
• Make sure all downloadable files remain in the same location. - Related to link rot. Ensure that any downloadable files can be accessed from the exact same URLs as you used to use.
• Retain as much functionality as possible. - Don't drop features that the old site used to have: no matter how trivial, someone out there really does use them. Try to find some kind of alternative that the new site can provide.
  • RSS Feeds - Yes, there are people out there that still use these, despite that Google Reader is long gone. How else would you subscribe to of blogs? Thankfully, most Jekyll themes come with a templatable feeds.xml that generates an RSS XML feed without any effort.
  • Comments - While the Yotsuba Society doesn't have comments, I would have used Disqus if it did, and imported the old comments with all history.
  • Multiple Authorship - This is a tricky one, given that it would be unsafe to give account access to all authors of the Yotsuba Society website. Since Jekyll is static HTML and best used with Git anyway, I would migrate the site to Github Pages, where you can give multiple users commit access to a repository.
• Teach the Website Owners well. - Make sure that the owners of the site actually know how to use the engine, and provide plenty of documentation.
  • Jekyll is pretty simple to work with, but the actual command-line tools that help create posts can be a pain to install on Windows. Unfortunately, since most of the world uses Windows, you'll have to figure out a suitable alternative.
  • The Github Online Repo Editor is actually quite nice, and you can even create files using it.
  • We then created a README.md file that provides full documentation on how to create blog posts, edit articles, and modify the site structure.

Importing a Drupal Website into Jekyll

Jekyll provides a fully-functional blog import system that will grab all pages, tags, categories from Drupal, by communicating with the SQL server.

Note: This script can't grab image nodes or comments! Remember to grab those manually or obtain via some other method!
Also, it only grabs blog posts. Home Pages and other special nodes also need to be obtained manually.

1. Install Ruby: Mac OS X - Linux
2. Install Node.js: Mac OS X - Linux
3. Open a command line.
4. Install Jekyll-Import. Some manual installation may be necessary.

• NOTE: This converter requires Sequel and the MySQL gems.
• MySQL has been superseded by MariaDB in Linux. Make sure to install mariadb and libmariadbclient-dev (for Debian)
• The MySQL gem can be difficult to install on OS X. Make sure to install MySQL, and use this argument instead: --with-mysql-config=/usr/local/mysql/bin/mysql_config

$ sudo gem install sequel
$ sudo gem install execjs
$ sudo gem install mysql --with-mysql-config=/usr/bin/mysql_config

4. Use the import commands shown on Jekyll-Import documentation: Drupal 6 - Drupal 7

You should edit the fields in the command before using it.

• dbname - The name of your database, just as Drupal asked you during installation. (check PHPMyAdmin)
• user - The username of the database user. Check your hosting provider's admin console.
• password - The password of the database user.
• host - The SQL Server hosting all the website data.
• prefix - The table prefixes used in the database. Usually, it's drupal_, but you might have set something different.

There are two ways to work with the Drupal SQL Database:

1. You can communicate directly with the curren website's Drupal SQL Database. However, it might be necessary to add your current IP as a trusted host (see your hosting provider for more details).
2. Alternatively, you can make a SQL database dump using PHPMyAdmin, import it, and host your own database on your computer. The host entry should then be localhost. This can be a bit more challenging to set up, but it reduces security vulnerabilities, and is necessary when a backup is all you've got.

For example, here is the command we used to communicate with the Yotsuba Society MySQL Database:

$ ruby -rubygems -e 'require "jekyll-import";
    JekyllImport::Importers::Drupal7.run({
      "dbname"   => "yotsubasociety",
      "user"     => "yotsubasociety",
      "password" => "put_your_password_here",
      "host"     => "mysql.yotsubasociety.org",
      "prefix"   => "drupal_"
    })'

Set up a SQL Backup and Export from it

Sometimes you might only have access to a SQL database backup. To export this database, you will have to create a local MySQL/MariaDB server, and import the database backup.

1. Install MySQL/MariaDB. Some Linux distros may have already migrated to MariaDB, which is a drop-in replacement for MySQL.

2. Start the MySQLd Service. On Debian/Ubuntu:

   sudo service start mysqld

3. Run the MySQL/MariaDB Secure Installation Script as root, and follow the instructions. Make sure to drop the test database and reload the privilege tables.

   sudo mysql_secure_installation

4. Once you're finished, log in as the root MySQL user.

   mysql -u root -p

5. Create a user to manage the databases. For example, I used user: localuser with password: password.

MariaDB [(none)]> CREATE USER 'localuser'@'localhost' IDENTIFIED BY 'password';
Query OK, 0 rows affected (0.00 sec)

MariaDB [(none)]> GRANT ALL PRIVILEGES ON *.* TO 'localuser'@'localhost'
    ->            WITH GRANT OPTION;
Query OK, 0 rows affected (0.00 sec)

6. Now create a new database to import the backup into. In this example, we call it yotsubasociety. Then quit.

MariaDB [(none)]> create database yotsubasociety;
Query OK, 1 row affected (0.00 sec)

MariaDB [(none)]> quit

Now you can import the database. As the localuser user, we access the localhost SQL server and insert the backup.sql dump into the yotsubasociety database.

mysql -u localuser -p -h localhost yotsubasociety < yotsubasociety352015.sql

Finally, here's the command we used to export to Jekyll from the localhost SQL Server. Usually, the prefix is drupal_.

$ ruby -rubygems -e 'require "jekyll-import";
    JekyllImport::Importers::Drupal7.run({
      "dbname"   => "yotsubasociety",
      "user"     => "localuser",
      "password" => "password",
      "host"     => "localhost",
      "prefix"   => "drupal_"
    })'

Cleaning up the Posts

Unfortunately, due to bugs in the Drupal import script:

• The author names aren't imported.
• The correct slugs aren't imported. In order to maintain compatibility with the old site's links, we need to grab the url_alias table as shown in the Drupal 7 database schema.
• The title is corrupted, as shown below:

---
layout: post
title: !binary |-
  SW50cm9kdWN0aW9uIHRvIFlvdHN1YmEgU29jaWV0eQ==
created: 1300717398
---

On the other hand, the correct title (well, lost underscores and Capitals) is always posted in the filename:

2014-03-04-the-complete-history-of-4chan.md

There must be some way we can make the title work. (though maybe it might actually display when compiled?)

Patching Jekyll-Import to work with Drupal 7

Jekyll-Import's Drupal 7 import scripts just aren't perfect or complete, as this bug report shows. We had to patch it to give us the correct title.

I have fixed this bug entirely by forcing the strings into Unicode format (and stripping any preceding and ending whitespace with strip):

# Get the relevant fields as a hash, delete empty fields and convert
# to YAML for the header
data = {
	'layout' => 'post',
	'title' => title.strip.force_encoding("UTF-8"),
	'created' => created,
	'excerpt' => summary
}.delete_if { |k,v| v.nil? || v == ''}.to_yaml

That way, it works even for those pesky titles with stray \xE2 junk that screws everything up.

Hacking the Jekyll-Import Gem

Before the change went live at the Jekyll-Import gem, I had to push the change to my existing gem to make the title work. First, we will have to be able to edit the gem, which is non-trivial. Thankfully, this user on StackOverflow has a handy Fork and Source method.

1. Fork the Jekyll-Import script to your own Github account.

2. Make your edits to the Drupal import script on your own repository. For example, I edited the following to add a permalink entry with spaces as underscores, and a better Title section:

      # Get the relevant fields as a hash, delete empty fields and convert
      # to YAML for the header
      data = {
        'layout' => 'post',
        'title' => '' + title,
        'permalink' => '/' + title.strip.gsub!(' ', '_'),
        'created' => created,
        'excerpt' => summary
      }.delete_if { |k,v| v.nil? || v == ''}.to_yaml
   

3. Now, clone your forked Jekyll-Import repository and add this to the Gemfile (also remove the gemspec line):

   gem 'gem_name', :git => 'git://github.com/username/foo'
   

4. Run the command bundle install to install the modified Gem.

5. Run the Jekyll-Import command you planned to use as usual. If it works, submit a pull request to the authors.

(though for me, the changes didn't stick, so I just hacked the file directly at: /usr/lib/ruby/gems/1.9.1/gems/jekyll-import-0.5.3/lib/jekyll-import/importers/drupal7.rb)

Obtain url_alias

If your website used a url_alias, as it probably should have, you will definitely want to inject them into the blog pages as a Jekyll permalink.

Sadly, as you see from the Drupal 7 Schema, the url_alias table is in no way relationally related to a node, which is just plain stupid and makes life hard.

I guess the only way to insert this right is to do it by hand. That's what we were forced to do.

First, use the following MySQL command to open the table and grab everything from the url_alias table.

MariaDB [(none)]> USE yotsubasociety
Database changed
MariaDB [yotsubasociety]> SELECT * FROM drupal_url_alias;

You will obtain all the url_alias from the database:

pid	source	alias
1	node/4	articles
2	node/6	the_otaku_continuum
3	node/7	contact
4	node/8	imageboardguide
5	node/9	boardsfaq
6	node/10	Channerstages
7	node/5	staff_recruitment_and_content_drive
9	node/12	archiveslaunch
10	node/13	An_Ode_to_the_F40PH
11	node/14	news_update_5_23_2011
12	node/15	Critique_and_Commentary
13	node/16	What_is_Chanthropology
14	node/17	originsofanon_pti
15	node/18	originsofanon_ptii
16	node/19	originsofanon_ptiii
17	node/20	20084chanthreads
18	node/21	yotsuba_archiver_changeover
19	node/22	the2ndchannel_pti
20	node/23	Raids_and_invasions_briefing
21	node/24	eternal_summer
22	node/25	Classic_4chan_memes
23	node/26	YS_Staff
24	node/27	Briefings
25	node/28	Futaba_Channel_Briefing
26	node/29	2channel_briefing
27	node/30	material_submission
28	node/31	Join_us
29	node/32	How_to_Easily_Archive_a_Web_Site
30	node/33	Television_Mostly_Sucks
31	node/34	briefing_channers_age
32	node/35	YSArchives_Update10_10_2011
33	node/36	burial_of_Illusion_of_rules_one_and_two
34	node/37	TheSecondChannel
35	node/38	TheThirdChannel
36	node/39	news_update-10-30-2011
37	node/40	re-savetheinternet
38	node/41	like_fish
39	node/42	Report_of_Joint_Operations
40	node/43	2011_Diaspora
41	node/44	americawonet
42	node/45	how_to_lurk_moar
43	node/46	Alternatives_to_Scared_Straight
44	node/47	chanverse_racism_briefing
45	node/48	1-26-2012_pressrelease
46	node/49	dealing_with_depression
47	node/50	Can_you_make_a_emblem
48	node/51	If_anyone_cares
49	node/52	sp_superbowl2012
50	node/53	ruoccupy
51	node/54	Nurse-kun_archives
52	node/55	facesofanonpti
53	node/56	facesofanonptii
54	node/57	YSchan_first_year
55	node/59	wikichan_restoration_complete
56	node/60	YSArchives_Update_5_6_2012
57	node/61	Computer games keep me mentally active
58	node/62	tuesday is backup day
59	node/63	2012Hibernation
60	node/64	Mostly_Perceived_Fear_and_Loathing_of_Reddit
61	node/65	Some_Chinese_Rail_Pics
62	node/66	two_panels-two_powerpoints
63	node/67	The_Imageboard_World_2
64	node/68	imageboardworld2_present
65	node/69	imageboardworld2_video
66	node/70	Links and Resources
67	node/71	russian_chanverse_notes
68	node/72	lurkmore_to_briefing
69	node/73	The_da_1chan_ru_came
70	node/74	the_russian_chanverse_part_2
71	node/75	savetheinternet_public_advisory
72	node/76	chanverse_review
73	node/77	two_major_updates
74	node/78	Routine_archival_ceased
75	node/79	19393370_hallo_I'm_from_Futaba
76	node/80	RIP_Yotsuba_Society_Archives
77	node/81	YSArchives_4chan
78	node/82	4chan_YSArchives_inducted_to_Stanford_Digital_Repository
79	node/83	YS_panel_presentations
80	node/84	The_Complete_History_of_4chan
81	node/85	Basic_Glossary_of_the_Chanverse
82	node/86	The_Silent_WebM_revolution
83	node/87	2channel_Japanese_Wiki_translation

I created this nice markdown table by taking the MySQL Table SELECT query, and changing the middle bar to Markdown format (changed all the + signs to |).

To ease the process of matching a node to a shortlink, I used a regex in the KDE Kate editor to match all node/## instances, and convert them into markdown links to the site in question. This way, I can just click a link as shown above to find the original title.

• Find: node/(\d+)
• Replace: [node/\1](http://www.yotsubasociety.org/node/\1)

Inject url_alias

Then, I open every single post dumped, and add permalink: /page_name_to_use in the YAML section. This can take a while.

I also use the handy Jekyll Redirect-from gem to make the /node/## links work as well:

redirect_from:
  - /node/1/
  - /Page_Name_In_Underscores/

Note: Make sure that the links all end in a /. This way, a folder with /node/1/index.html will be created, instead of /node/1, which is just spit at the browser as a downloadable binary.

Choose a Jekyll Theme

Now that you have the data exported, choose a Jekyll theme that fits for you. Here's a nice listing of some free, amazing themes.

Since I needed something simple, and since the Bibliotheca Anonoma Blog and my own blog uses the same theme, I used the very nice, simple, and responsive Minimal Mistakes theme.

All I had to do was git clone the theme, paste all the pages into the theme, and do a bit of configuration, and the Jekyll site is ready.

Hosting the Jekyll Site

For the most part, you will want to host your Jekyll Site using Github Pages. Github has the best support for Jekyll (they created it), and offers free limitless Static HTML hosting.

Just push your website to the master branch on your username.github.io repository, or to the gh-pages branch on any random Github repo. Github will automatically generate the website at username.github.io or username.github.io/repository.

For the Yotsuba Society, we had them create a new Github Organization (which is free), whereby we pushed the Jekyll website to the master branch on the repo: yotsubasociety.github.io. It was then instantly hosted at http://yotsubasociety.github.io without a hitch.

Notice that you can also use custom domains with Github Pages! We had to direct http://yotsubasociety.org to the Github Pages server, following the instructions provided by Github.

And now, the new Yotsuba Society Website is complete. What an amazing, difficult, but eye-opening journey.