My friend Elizabeth needed to get a website up for her photography, so I said I'd see what I could do. After trying to customize Gallery (PHP), which is a really good way to go in general, I figured I should look into a quick, customized solution. I'm already familiar with the CakePHP framework (and since my host has crappy support for Ruby/Rails, I didn't want to use a Rails solution), I searched around a bit and found a good tutorial on how to set up a photo gallery with CakePHP using the Flickr API.

After about an hour I had a fully-functioning site up. The next few hours were spent tweaking the layout/design (and figuring out Internet Explorer bugs). I added some more JavaScript to the final solution that updates a caption in addition to the main photo. The results can be found at esoule.com. She's also got an online store at Etsy as well if you want to buy anything.

One thing that annoys me a lot on the web is the extremely inefficient use of space on many (if not most) web pages. Why does the header always take up a third of the page? Or take a look at this:

That's a screen shot of my fantasy baseball league's live scoring page on a 21" monitor (1280x1040 resolution) where I've set the browser window size to 1024x768. The three things I really want to see on that page are: (1) The score of this week's match-up (5-4), (2) How each player is doing today (the section below the score) and (3) how this week's stats break down by category (which isn't even close to being on the page.. it's way below).

So the problem with this page is that we have about 9.5 inches of screen height to work with, and the score takes up almost 2 inches (it could easily take up 1/2 an inch), and the rest of the junk above the score takes up about 5 inches. The ad banner at the top is huge, but hey, it's supposed to generate money for the site, so I can accept that (and by accept I mean I block it with adblock of course, but it still takes up the space). The biggest problem is that the section between the top banner ad and the scoreboard takes up about 3.5 inches, which is more than a third of the page, yet conveys almost nothing. I'm quite aware that I'm on the Big Sky Baseball 2007 (listed twice for some reason) page, so maybe they could make that text a little smaller? There's also at least an inch of wasted white space in the "GameCenter" section. The links below "GameCenter" could be next to it instead of below for one thing.

Of course, this is only one of many web sites with similar problems, although this is about as bad as it gets.

I'm by no means web design expert, but I'm pretty sure I'm right about this. I even get annoyed by the space taken up by the header on my own website, UrbanDrinks.com, and it's only about an inch or so.

In my job at Sun as a quality engineer I have made several web-based applications to help analyze and report our manufacturing quality data. One common situation is where we have a number of units that tested during a time period and a number of them failed, and we want to know if the fail rate was worse than a threshold fail rate.

For example, we might have a threshold fail rate of 10% (in reality the thresholds we use are much, much lower, but this is easier for demonstration purposes) and tested 100 units, 12 of which failed. If we treat these 100 units as a sample out of a theoretical infinite population, are we at least 95% confident that population's fail rate is greater than or equal to 10%? To find out, we can use the binomial distribution class from the rubystats Ruby library (I ported this from the PHPMath project. It's available at RubyForge or you can install it with "gem install rubystats" on a system that already has RubyGems installed).

Once you have the rubystats gem installed, you can test our scenario rather easily. Below is a code sample that tests 10 scenarios against the 10% threshold (called bad_fail_rate in the code). This consists of finding the cumulative probability of observing f or more failures if the theoretical infinite population's true fail rate is 0.10. We use the cdf method (cumulative density function) to calculate the probability of observing (f-1) or fewer failures, then subtract that value from 1. If it's less than 0.05 (the alpha value for 95% confidence, i.e. 1 - 0.95 = 0.05), then the fail rate is significant.

require 'rubygems'
require 'rubystats'
require 'binomial_distribution'

tested = [100, 68, 67, 96, 46, 2, 13, 33, 88, 71]
failed = [12,  9,  12, 7,  7,  0, 6,  4,  5,  5]

bad_fail_rate = 0.10
alpha = 0.05

for i in 0..9
  t = tested[i]
  f = failed[i]
  bin = BinomialDistribution.new(t,bad_fail_rate)
  cdf = bin.cdf(f-1)
  pval = 1 - cdf
  pval = sprintf("%.3f",pval).to_f
  status = pval <= alpha ? "RED ALERT" : "OK"
  puts "Tested: #{t}\tFailed: #{f}\tpval: #{pval}\tStatus:#{status}"
end

Which outputs:

Tested: 100     Failed: 12      pval: 0.297     Status:OK
Tested: 68      Failed: 9       pval: 0.237     Status:OK
Tested: 67      Failed: 12      pval: 0.033     Status:RED ALERT
Tested: 96      Failed: 7       pval: 0.856     Status:OK
Tested: 46      Failed: 7       pval: 0.172     Status:OK
Tested: 2       Failed: 0       pval: 1.0       Status:OK
Tested: 13      Failed: 6       pval: 0.001     Status:RED ALERT
Tested: 33      Failed: 4       pval: 0.423     Status:OK
Tested: 88      Failed: 5       pval: 0.947     Status:OK
Tested: 71      Failed: 5       pval: 0.85      Status:OK

As you can see, the first scenario where 12 out of 100 failed is not statistically significant at 95% confidence.

We've recently added quite a few places to the UrbanDrinks.com (a Portland happy hour website and side project of mine) database, thanks in part to the excellent Portland Happy Hour Guidebook, and now have over 180 different places in Portland with happy hours! That's a lot considering it doesn't count any of the surrounding metropolitan area. Downtown Portland alone has 52 happy hours today. Wow. When I started this project I had no idea how many happy hours in Portland there really were.

So despite the gray, drizzly days, Portland has got to be the happiest place in the country!

I'm really glad to see that Sun is supporting the JRuby project now, and seems to be supporting scripting languages in general. I feel like now I don't have to give excuses at work. I've been developing web applications used Sun-wide internally for over a year and a half now. I work as a quality engineer in Operations, so I'm not really surrounded by Java developers, but I've still received the occasional question asking why I didn't develop in Java (or Perl, another favorite here). The truth is that I just don't know much Java, and some attempts to learn J2EE a couple of years ago frustrated me a bit. Also, I already knew some PHP, so I developed in PHP at Sun for a couple of years until I got good at it. Then I found Rails and tried it out. I immediately got it installed on my server and started developing in Rails from then on.

I like Rails because it provides that perfect mix of coding ease and structure. The Ruby language is extremely flexible, easy to understand, and truly object-oriented. PHP provides much of this, but the language is much more verbose. PHP seems to be a gateway language for many new developers (myself included, aside from some C/C++ in college), and not surprisingly, there are a lot of bad PHP applications out there (I am the guilty author of some). Ruby, on the other hand, is being learned for the most part by new Rails developers. The Rails framework definitely helps new developers create well-organized applications from the beginning. Basically, if PHP would have become popular because of a framework, we wouldn't have as much bad PHP code out there. I really like PHP still, but I like the Ruby/Rails stack better. When I do write PHP code, I try to structure it as well as possible, usually with a variety of PEAR libraries and using CakePHP as a framework (CakePHP follows a lot of Rails-like conventions, and is a very good framework). I wish the PHP books I read when I first started out used a good framework to teach with. However, if they had, we might not have needed this "Rails Revolution".

With JRuby, I hope to see Ruby get faster, and maybe I'll learn Java so I can write fast, compiled portions of my future Rails apps. Long live Java and Ruby!

This is one of those events in life that makes you say "wow, what are the odds of that happening?".

Shortly after I graduated high school, I moved to Phoenix, Arizona to start college at DeVry in July 1997. I was the first of four roommates to arrive, and since the apartment was in my name, I was the one going through the paperwork process. One of my new roommates, whom I hadn't met yet, was Clint Carlson from Alaska, who was also going to be attending DeVry and was scheduled to arrive that same day. I was in the apartment management office waiting for my turn to sign some papers when I overheard the guy in front of me say he was Clint Carlson. So I of course introduced myself and asked him if he was from Alaska (to make sure it wasn't some other random Clint Carlson), and he replied "yes".

"So I guess we're going to be roommates," I responded.
"Um, no, my roommate is so-and-so."
"Uh.. you're Clint Carlson from Alaska, right? Are you going to DeVry?"
"Yeah, start in a few days."
"Um.. I think you're my roommate then. I'm rooming with Clint Carlson from Alaska who's going to start DeVry next week."
"Well, my roommate is so-and-so, so maybe you got the wrong guy.."

Yeah right. So I asked the apartment manager if there was more than one Clint Carlson moving today.

"Yep, I guess there are."
"Two Clint Carlsons from Alaska, moving into the same apartment complex in Phoenix, Arizona on the same day and both attending DeVry? AND I happen to be in the apartment management office when the wrong one is here signing papers?"

"Yep, two Clints from Alaska moving in today."
"Wow, what are the odds of that?"

So there you have it. That's probably the strangest coincidence that's happened to me. I'd like to hear some other stories.

p.s. I don't believe in any voodoo or fate stuff, so my statistics-oriented mind chalks this up to random chance alone.

I'm excited to say that we've finally launched the beta version of UrbanDrinks.com, a happy-hour website focused on Portland, Oregon.

This has been a side-project I've been working on since early July using Ruby on Rails (in fact, it's what my RubyQuiz post was all about). I've learned a lot so far about the Google Maps API and designing a database to handle all of the time ranges and associated happy hours, as well as just polishing (er... maybe "sanding" is a better term) my Ruby skills. I've also had the pleasure of pulling my hair out dealing with Internet Explorer CSS bugs.

I have to thank Guilhem Vellut for his wonderful Google Maps Rails Plugin. This plugin made the Google Maps portion of the project a thousand times easier. Of course, I should also thank the people who responded to the DayRange RubyQuiz and the Rails and Ruby developers in general. They really do make web development fun.

The site is still under development, so there could be some bugs. Definitely let us know if you find any or if you have any comments or suggestions. We also have an UrbanDrinks blog where we'll keep you posted on site updates and our opinions on various things related to happy hours. Enjoy!

If you've ever created a web application that deals with scheduling recurring events, you may have found yourself creating a method to convert a list of days into a more human-readable string. For example, suppose a musician plays at a certain venue on Monday, Tuesday, Wednesday, and Saturday. You could pass a list of associated day numbers to your object or method, which might return "Mon-Wed, Sat". The purpose of this quiz is to find the best "Ruby way" to generate this sentence-like string.

Here are some example lists of days and their expected returned strings: 1,2,3,4,5,6,7: Mon-Sun 1,2,3,6,7: Mon-Wed, Sat, Sun 1,3,4,5,6: Mon, Wed-Sat 2,3,4,6,7: Tue-Thu, Sat, Sun 1,3,4,6,7: Mon, Wed, Thu, Sat, Sun 7: Sun 1,7: Mon, Sun 1,8: ArgumentError

Five different artists have produced a bunch of paintings, and it's the player's task to be both the buyer and the seller, hopefully making a profit in both roles. He does this by putting a painting from his hand up for auction each turn. He gets the money if some other player buys it, but must pay the bank if he buys it for himself. After each round, paintings are valued by the number of paintings of that type that were sold. The broker with the most cash after four rounds is the winner.

If you run a company with a separate computer for each employee, you should really stop doing that and switch over to running a company with a Sun Ray for every employee. Yeah, this is a "plug" for Sun, but I'm serious about this. Sun recently set me up with a Sun Ray at home (a Sun Ray 170 to be exact), and I'm very impressed. We've been using Sun Rays at work for a few years now, but seeing them work so well from home is whole other level of technological coolness. Here's a short list of benefits:

• I don't have to maintain anything. All of the applications sit on centralized servers, which are much easier and cheaper to maintain than individual (especially Windows-based) systems.
• I can leave my session at work with various applications open, even if I'm in the middle of typing an email, and when I get home, I simply insert my badge/smart card and I can pick up where I left off.
• It's fast. Right now I'm typing this from 13 miles away from the servers and I can see no delay as I type. Sometimes the screen jitters a bit when I switch windows, but it's not bad at all.
• It's secure. I don't have to worry about setting up my laptop to be hacker-proof and don't have to worry about it being stolen.
• It's completely quiet and power efficient. The first time I used a Sun Ray at home, I found it almost too quiet in my condo. No fans, no heat, no hard drive noise.
• This 17" LCD is way nicer than my laptop display!

Heck, if I had a family, I'd set up a small Sun Ray network at home. Of course, I'd also make my kids use Solaris or Linux too...

I just took a class from Edward Tufte on how to display and present information. Tufte has relatively recently introduced the idea of sparklines (careful, that page has large graphics), which are essentially word-sized graphics that can fit right into a paragraph of text. I'm currently working on adding customizable sparklines for various baseball statistics, including team wins and losses for the season (thanks to sparkline.org's free PHP library). For example, the Oakland A's have made a dramatic comeback since May 29, as can be seen by this sparkline: (through July 28th). The upward whiskers are wins and the downward whiskers are losses. Red indicates a shutout, and the horizontal line under some of the whiskers indicates a home game. From this sparkline you can clearly see how the A's have turned around their season.

If you're a Major League Baseball fan and haven't seen it already, there's a great website out there called The Hardball Times that includes great daily articles as well as a wide variety of baseball statistics, especially some of the more hard-to-find Sabermetric stats like Win Shares, Pitching Runs, and Runs Created. I admit this is partially a shameless plug since I volunteered to develop the code for the statistics pages, but I have to say it is pretty darn nice stuff. If you're looking for baseball stats that you can sort and limit any which way you want without having to wait 30 seconds for a banner-ad-polluted page to render, The Hardball Times stats pages are probably something you'll appreciate.

For the techy folks, given that this is a Sun Microsystems-hosted blog, I developed the code in PHP 4 (I wish it were 5, but sometimes you have to deal with what the hosting company uses), Smarty, and MySQL 4.0.x. I used the great Propel object persistence and query toolkit for PHP, which is essentially a PHP version of Apache Torque, a peristence layer for Java. So far it's been great, but I do run into the occassional memory limit since our web host kills scripts that use more than about 16MB, and a Propel object can take up quite a bit of memory. I only run into this problem when running database loading scripts though, which cycle through all the players and do a lot of simple calculations to generate the various statistics. I also use some home-grown DB query classes for common queries that aren't Propel-friendly or efficient (at least in how I use Propel). The problem with a persistence layer that maps an object to a database table is that each object is generally an entire row of the table. If you need something like the average batters faced per game for pitchers in the American League, a persistence layer isn't really the way to go. I still abstract my queries from the DB using PEAR::DB (I could certainly use Creole, which is what Propel uses), so at least I've preserved some level of portability.

The coolest thing about this Hardball Times code is that I made most of it completely generic--it's sort of a framework really. Other people can use a web form to make changes to the sortable stats tables, including which columns are displayed, what the title of the column is, whether it should be included in the sort and limit drop-down lists, and what order the columns are displayed. All that is needed to add a new sortable web table is a group of Propel classes for the table and you're ready to go. This has come in very handy for my real job at Sun, where I create web applications to display statistical information on manufacturing data (such as fail rates, control charts, etc.).

For future PHP projects, I think I'll use the new Cake web development framework based on Ruby on Rails. I've tinkered with it a bit and so far it's great. One big advantage is that you don't have to generate the ORM classes for your database tables (although this is pretty easy with Propel), which makes updates to your code, especially database changes, quicker to implement.

This is the high school I went to. We were very lucky to have such a great science program. Keep in mind this is a school with less than 100 students in a town of about 450:

In the small community of Sunburst near the Canadian border, North Toole County High School teacher Larry Fauque accompanied four of his students to ISEF. Fauque estimates this was his 20th trip to ISEF. Three of his four students won awards at this year's national level of competition. The National Science Teachers Association has recognized Fauque's science curriculum over the last 34 years as one of the top 10 science research programs in the nation. Link

We might not have had AP classes, but classes like this made up for it. I'm also confident our English and Math curricula were as strong as most schools' AP curricula.