Entries by Rasmus

http://pipes.yahoo.com is a cool toy, and by toy I mean it in the useful and cant-stop-playing-with-it sense. My first impression when I saw an early version a couple of months ago was, "How the heck did they do that?" I was reading the Javascript source code for quite a while. Once you get beyond the fact that this is a browser-based app doing this without Flash, or Java or any similar cheats, you get down to what the app actually does.

Years ago I wrote this silly little Mashup example:

http://buzz.progphp.com/?q=4

It grabs an RSS feed, in this case the top daily search term % movers from http://buzz.yahoo.com/feeds/buzzoverm.xml which gives you an indication of what is on the minds of web searchers right now. I took these searches and did a Yahoo Image search and a News search and combined them in that oval interface you see. I had to do a bit of RSS and XML parsing to take these different data sources and combine them. This is what Pipes is all about. It provides a visual environment for manipulating data sources and then provides a number of different ways to get the results and integrate them into other things. Directly in your RSS reader is probably the simplest, but you could also feed it to PHP and do further data manipulation.

A simplified Pipes version of the above takes the same Buzz.yahoo.com RSS feed and does a Flickr search on each search term. The result looks like this:

http://pipes.yahoo.com/pipes/DnudMlO32xGDcIu7pRr_og

The point here is not the visual output. It is meant to be fed to something else. Hover over the "Subscribe" link on the right there. Then click on the "How this pipe was made" image on the left to see how it works.

This is a particularly lame and simple pipe. Some much cooler ones include:

Blog Buzz for Pipes combines a couple of different blog watching feeds, filters out duplicates and gives you a combined feed in reverse chronological order. When you look at how it was made it becomes immediately obvious what it does. You can save a copy and make your own version that watches for whatever terms you want.

Another interesting one takes the New York Times front page, runs a content analysis on it to get a set of representative keywords and then does a Flickr search on each of those. http://pipes.yahoo.com/pipes/vvW1cD212xGMiR9aqu5lkA/.

Here is a much more complex pipe that takes some user input. It finds apartments near things. In this case it looks for apartments within 2 miles of a Park in Palo Alto, California by searching Craigslist, then doing a location extraction and then doing a Yahoo! Local Search for that location.

Even if you have no use for processing data sources this way, open up one of these Pipes and drag the boxes around and watch the pipes react. Web apps don't get any cooler than this right now.

Want to work on some of the busiest and coolest web apps in the world?

Do you like Flickr, and want to work downtown San Francisco?

Or perhaps you are into music, movies or TV and want to work out of Santa Monica? Jumpcut? Or have you seen answers.yahoo.com? Address Book, Personals, Search, Premium Services, Hot Jobs? Want to do interesting things combining PHP and Flash?

Yes, I get a referral bonus, but I need more toys. You get a cool job though, so I think we are even.

Send me your resume and let me know what sort of stuff you are interested in or poke around on http://careers.yahoo.com/ and let me know which job interests you and I will forward your resume to the appropriate hiring manager.

[edited to remove RSS ad test I had forgotten about]

My previous entry way back on Aug.18 talked about the Nike+ipod widget I had picked up along with a first gen Nano. I figure it is time for an update. I still use it 5 or 6 times a week and I just logged my 100th run. There have been a few more actually, but I have had the odd Nano crash/corruption on me before I had a chance to upload a run.
Despite those occasional technical glitches, I still really like using it. I have been eyeing the Garmin Forerunner 305, but I actually like listening to music or a podcast as I run and it seems a bit much to strap a GPS to my wrist and also carry an mp3 player. Plus the thing costs $300+ vs. the $27 the Nike+ipod gadget costs, assuming you already have a Nano.

My 100th run looked like this:



620 miles later and some 80,000 calories burned I have gone from being a 205lb fat slow geek to a 175lb slow geek. Not entirely true, I lost about 10lb before I started running more seriously, so the Aug.18 to Jan.27 timeframe was about a 20lb drop. 80,000 calories and I read that one pound is about 3500 calories which seems to fit my numbers.

I ran a couple of hundred miles in a pair of Nike+ shoes. It is very convenient to have the sensor right in the sole of the shoe, but that is the only thing I liked about those shoes. Now I have a pair of New Balance 767's and have sewn the little sensor seated in a Switcheasy RunAway thing to the front of the shoe with the back looped through the bottom of the laces. That gets it as horizontal as possible and I find it is more accurate that way. Of course the RunAway plastic thing needs a bit of tape to make sure it doesn't swing or pop open. Here is a picture of the hack job:



I have also been playing with the run data a bit. Each run is stored in an XML file which is uploaded to the Nike site by itunes when you synch. You can easily read the files yourself by mounting the ipod and navigating to it. However, the red Nike interface you see above is a flash thing and since flash is obviously client-side the site needs some sort of API. With a little digging with Firebug it was trivial to figure out how it worked and I whipped up a little Serendipity plugin to show my stats. You can see it in the right column of this page. And I wrote a simple PHP 5 class that makes it rather easy to integrate this stuff into other apps. You can see the code here:

http://lerdorf.com/php/nikeplus.phps

Each run in the run list has an id and you can call the run() method in that class to get the raw data for that run. Graphing that gives us something like this for that same run I showed at the top:



The grey line is the raw data and the red line is a Lowess curve based on that. I find it interesting how noisy the raw data actually is. Especially for the shorter runs, but by applying a bit of local regression it cleans up nicely and data that seems accurate emerges. I can pick out the uphill stretches of my run from the graph above. If you click on the runs in the right sidebar section there you can see the graphs of my other runs.

If I could somehow find some free time I'd love to build a better tracking site than that Nike thing and then also have it support the Forerunner and whatever other devices are out there. But for now most of my free time is spent running. Probably better for me in the "long run" anyway.