The position of needle/haystack parameters in string functions is the reverse of the parameters in array functions.

strstr($haystack, $needle)
array_search($needle, $haystack)

I just assumed the developers were completely insane and chose parameter orders at random.

I decided to try this out by ordering the “Rocket Surgery Made Easy“ by Steve Krug, the sequel to “Don’t Make Me Think“. I haven’t finished the sequel yet, but I thought it would benefit the team to do a 15 minute presentation on the original:

I’m quite pleased with my first ever presentation (not including primary school). I assure you, it makes more sense if I’m standing there talking you through it!

Oh, and I’d highly recommend “Don’t Make Me Think” if you’re thinking of purchasing a web usability book. It really is quite good, although a little out of date.

SELECT ResponseXml.value('(//Reference)[1]', 'varchar(50)')
FROM Transaction

A fairly simple XQuery request of an XML column but annoyingly it didn’t work, and it turned out to to be because the ResponseXml had an unnamed XML namespace, which SQL Server doesn’t handle too nicely.

<Message xmlns="http://www.example.com/">
<Reference>123</Reference>
</Message>

It’s the xmlns attribute (XML namespace) that causes the problem apparently, and the simplest solution I found was this snippet:

SELECT ResponseXml.value('declare default element namespace "http://www.example.com/"; (//Reference)[1]', 'varchar(50)')
FROM Transaction

I just had to tell XQuery to use a different default namespace – a lot simpler than other suggested solutions such as removing the namespace from the XML.

Typically, Google’s way of tracking site usage is to ask you to put a small snippet of HTML onto your site which is essentially a javascript call to their servers. Great – except a lot of mobile phones don’t actually support javascript. So the only good solution is to run some code on the server. I don’t know if it’s related to Google’s purchase of AdMob, but they’ve done exactly that and provided server side code in various programming languages (C#, PHP, etc) to achieve this. See ga.php as a real example.

Great, now I’ve got beautiful graphs showing me how people use my site. But what’s that? 25% of my users live in the country (not set)? I’ve not heard of the place! It turns out those users more or less all used one browser: Opera Mini.

Oh Opera Mini, what a strange beast you are. While the Opera browser is a full normal browser, Opera Mini is a service that installs a thin client on your phone and Norwegian proxy servers visit the site for you. Norway? Well that’s a little better than living in (not set).

Luckily, this whole problem is fixed by a teensy change in their server side code. Rather than always fetching the browser’s current IP address, I tweaked Google’s PHP code so that it detects Opera Mini is in use and sends the real user’s IP address. To be honest, I’m a little confused why this isn’t in Google’s code in the first place.

So here it is. A minor fix to Google’s ga.php file (line 162) to improve Opera Mini handling:

Original line: "&utmip=" . getIP($_SERVER["REMOTE_ADDR"]);

and my version: "&utmip=" . getIP((stristr($_SERVER['HTTP_USER_AGENT'], 'opera mini') && array_key_exists('HTTP_X_FORWARDED_FOR', $_SERVER)) ? $_SERVER['HTTP_X_FORWARDED_FOR']: $_SERVER["REMOTE_ADDR"]);

That’s all. Now I get real country stats :)

Storage

Dabr.co.uk stores nothing. Seriously. Well, ok, it logs visits like 99% of websites do on the net. It doesn’t store logins, users, tweets, settings, anything. That’s great in terms of cheaper hosting for me but it restricts what I can make the code do.

@newdabr is different. Very different. Every minute it asks Twitter: “Wazzup??” and stores the response in a database. While @dabr is slow and has to talk to Twitter with every click you make, @newdabr is super-snappy and just asks the local database.

Advantages

So we have this cool database, but how does that help? The list is pretty big and here’s just a few nice examples:

• Pages should load in milliseconds, not seconds.
• Never miss a tweet. @newdabr knows all and can maybe resume from where you were last reading.
• The site can happily auto-refresh.
• Notifications! Yes really.
• Contact grouping – think Tweetdeck columns.
• Expanded URLs. Want to know what a TinyURL really points to? I could cache the results and let you know.
• Persistent settings – e.g. contact groups, colours
• Filtering – e.g. hiding certain hashtags
• Interested in links or photos your friends posted today? @newdabr could list them all for easy access. e.g. sidebar of today’s Twitpics.

Disadvantages

Since @newdabr needs to keep chatting with Twitter all day to work, I imagine it would cost more to run and possibly wouldn’t work with tens or hundreds of accounts. For that reason, it’s going to be a slow-starting project with just me using it at first and a few alpha testers along the way.

Timeline

What’s one of those? Sorry, I’ve got no idea when I’ll work on it but I will keep updating the @newdabr Twitter account any time I make changes.

My first point of call is to find out if any social networking sites (e.g. Brightkite) are interested in using this to create a great mobile web interface for Windows Mobile devices.

Manual labour

About 6 times per day, I press a button on my phone and it sends a text message to my website with my current cell tower ID. My website then takes over and does all the smart stuff by itself. If I’ve traveled somewhere new, it automatically looks up the cell ID in a large UK database using their API to find the latitude and longitude of the cell tower I’m currently connected to. But big decimal numbers showing my current latitude and longitude aren’t all that interesting on their own.

Sidebar widget

The next step in my process was to add the widget on the side of my website to show everyone my current location. It’s not particularly fascinating – either I’m in my home town, at work, or traveling somewhere in between. But I’ve just added something a little more interesting.

Google Earth

Whereas before the link on my current location just took you to the coordinates where I’m currently sat, it will now load up a KML file showing a nice picture of a man standing on my current location, and hopefully a number of little cell towers which will show where I’ve been over the last 7 days.

The historic data isn’t all that interesting in Google Maps, but if I open up my KML file in Google Earth then I expose the timeline controls and I can choose to see where I was at different times and even animate my location over time.

What’s next?

I’ve got a habit of always wanting to do more.  I’ve cleaned up my location database so I could add in other locations (i.e. not just cell towers) and try plotting even more data. I’d like to try the following:

• Load up GPS data for any hikes I go on
• Add a few key locations onto the map such as the Scout hut, my brother’s place in Canada, anything relevant to me
• Add some kind of refresh so that it updates automatically when I’m on the move
• Create some kind of JavScript timeline interface for Google Maps

“More than half of the word’s 7000 languages are endangered, because they consist of an unsustainably small – and declining – speaker base.” – New Scientist

On a much less epic scale, I’m concerned about my own knowledge of programming languages fading to nothing from lack of use. The best example for me is Perl.

Perl is the first server-side programming language I tried to learn to improve my websites. I made very basic guestbook scripts and that was about it, but I did (for a short time) know the syntax and some basic principles in the language. I’ve not used Perl for 5 years!

I started looking into PHP, and never looked back. So what do I remember of Perl now? Hardly anything at all.

With a few years of self-taught PHP under my belt, I’m now pushing slowly into ASP.NET. This time, I’m going to keep my PHP very much alive.