Kevin Marsh

I finally took the plunge and built a Hackintosh.

I was against the idea for the longest time, previously willing to trade a bit of cash for the extra stability and legality that an Apple-Mac-proper offers. However, when comparing a Mac Pro to a homegrown solution the extra bit of cash is now a serious few thousand dollars.

Design Goals

I really wanted a fast machine for web development with Ruby and Rails, with enough horsepower to run Photoshop on some big files. This meant: SSD and RAM. Even with 12G of RAM I think the biggest speed differentiator on this machine that blows it out out of the water is the SSD. I’ll post some further benchmarks, but a quick comparison of a Rails unit test shows a speedup of about 7x from my Mac Book Pro.

The Goods

The Work

Getting everything to work together was a bit of a hassle, mostly due to never having to worry about drivers on OS X in the past and learning how all the kernel subsystems work. Kexts? (kernel extensions) SLE? (/System/Library/Extensions).

This tutorial (using iBoot and Multibeast) was a big help, as was the InsanelyMac Forum.

I had a lot of problems trying to get an nVidia GT240 to work with two displays, but I finally gave in and ordered a 9800GT which was reported to work perfectly — it did.

I’m not claiming this build is perfect, I still have a few odd issues with USB, but its close. And way more economical than buying a Mac Pro from Apple.

I’ve taken the plunge to a static-HTML blog with Jekyll. I’m impressed so far. I love having the whole she-bang in git, but then again the whole idea of a datastore in git has fascinated me from the start.

And pushing to GitHub pages (willcodeforfoo.github.com) just worked. A few minutes after pushing, I hit the URL. Half expecting an error page and half expecting just nothing I found the site. Just the way I hoped it would work.

I thought I’d share some of the things I encountered along the way to help folks just getting started.

Source Code

First up, feel free to check out the full source code that Jekyll uses to build all this html: http://github.com/willcodeforfoo/willcodeforfoo.github.com/tree/master

Importing from Mephisto

Here’s a hacky, super-simple, one-liner script that I threw together at the Mephisto Rails console:

Content.find_by_sql("SELECT id, permalink, body, published_at, title FROM contents WHERE type = 'Article' AND published_at IS NOT NULL ORDER BY published_at").each {|c| File.open("#{c.published_at.strftime('%Y-%m-%d')}-#{c.permalink}.textile", "w+"){|f| f.puts "---\nlayout: post\ntitle: #{c.title}\n---\n\n"; f.puts c.body }}

This will not work 100%, though. It doesn’t do anything fancy like output real YAML, with escaped characters. So if you have funky characters in your titles, things could blow up.

uninitialized constant Classifier::LSI::Matrix (NameError) when running with --lsi option

Applying this patch to Classifier’s lsi.rb file seemed to do the trick.

As an aside: I’m truly impressed with the Classifier gem. I’ve seen its Bayesian stuff before, but Jekyll was my first exposure to LSI.

Categories

I was thinking about adding other “types” of posts to my site, something along the lines of books I’m reading, music I’m enjoying, etc.

I decided I could use Categories to pull this off, so my structure looks something like this:

|-- CNAME
|-- _layouts
|   |-- book.html
|   |-- master.html
|   |-- music.html
|   `-- post.html
|-- about.textile
|-- articles
|   `-- _posts
|       |-- 2009-02-10-reducing-the-noise.textile
|       `-- 2009-02-12-jekyll.textile
|-- atom.xml
|-- books
|   `-- _posts
|       `-- 2009-02-12-click.textile
|-- images
|   `-- feed.png
|-- index.html
|-- movies
|   `-- _posts
|-- music
|   `-- _posts
|       `-- 2009-02-12-ray-lamontagne-gossip-in-the-grain.textile
`-- stylesheets
    `-- styles.css

Then on my homepage, for example, I can pull all the music entries with album art from Amazon by doing:

{ % for music in site.categories.music % }
  <a href=""><img alt="" src="http://ec1.images-amazon.com/images/P/.01.MZZZZZZZ.jpg"></a>
{ % endfor % }    

(I added an asin key to the post’s metadata header.) Slick!

Excerpts

Thankfully thanks to Jekyll’s slick metadata feature for posts, all I had to do was add a excerpt key to the header of a post and throw my excerpt there:

---
layout: post
title: Jekyll
excerpt: |-
  I've taken the plunge to a static-HTML blog with Jekyll. I'm impressed so far. I love having the whole she-bang in git, but then again the whole idea of a "datastore in git":/2008/04/13/git-as-a-data-store.html has fascinated me from the start.

  And pushing to GitHub pages (willcodeforfoo.github.com) just _worked_. A few minutes after pushing, I hit the URL. Half expecting an error page and half expecting just nothing I found the site. Just the way I hoped it would work.

  Big ups to "Michael Bleigh":http://mbleigh.com for making his whole blog available on "GitHub":http://github.com/mbleigh/mbleigh.github.com/tree/master. I didn't take much design-wise, but seeing how things were structured was a huge help. I did nick his Atom feed template.  
---

h3. Importing from Mephisto

Here's a hacky, super-simple, one-liner script that I threw together at the Mephisto Rails console:

The next gotcha was that I couldn’t get Jekyll to Textilize anything other than the content, and since my excerpt needed to be rendered as Textile, this was a problem.

So I submitted a patch that creates a filter called textilize that I can pass my excerpt into for formatting. Hopefully it gets applied to the master branch that GitHub Pages uses.

Inspiration

Big ups to Tom Preston-Werner for writing Jekyll and Michael Bleigh for making his whole blog’s source available on GitHub. I didn’t take much design-wise, but seeing how things were structured was a huge help. (I did nick his Atom feed template.)

I’m spending some time tonight to try and reduce some of the noise in my life. From a practical standpoint, this means I’m:

• Unsubscribing from a few RSS feeds that I end up just reading past anyway
• Deleting some iTunes tracks that I usually end skipping, both at my desk and in my car (the Skip Count in iTunes is great for this)
• Unsubscribing from a few mailing lists I rarely read, are definitely not active in, and end up just Google searching when I need a topical answer
• Taking steps to reduce blog spam
• Creating server-side and client-side rules to move bulk e-mail from companies like Dell to their own folder or tag, and skipping the inbox altogether. Even better, unsubscribing from most of these completely.
• Deleting rarely used Apps

While these minor things seem really trivial piece-by-piece, I think the sum of them cause a lot of distraction and cloud meaningful activities. As such, I plan to do this de-lousing on a periodic basis.

OK, this one is almost too hacky and simple to share, but I wanted to get it out there. Want to show your site’s users what’s changed? Just show them your git changelog:

It’s a bit dorky, and definately not as pretty as a properly parsed and formatted changelog, but for showing stakeholders what’s changed it’s been useful.

Here’s how it’s done (again, kind of a hack):

Controller

@code = `cd /var/www/vintageaerial.com/htdocs/shared/cached-copy && git-log --max-count=20 --no-merges`

View

<pre class="changelog"><%= @changes %></pre>

Stylesheet

pre.changelog { overflow: scroll; }

I definitely wouldn’t use this in high-traffic production areas without some caching. It’s also enforcing me to write better, more communicative changeset descriptions, because I know someone can read them.