Sparkour Day

It's now been almost two months since I released Sparkour, the open-source collection of programming recipes for Apache Spark. Unlike DDMSence, which remained a niche oddity for its whole six year lifespan, Sparkour began to get traffic from Google almost immediately. Publishing DDMSence was like cornering the market on ergonomic attachments for a pasta maker -- only 10 people actually own a pasta maker in the world, and only 2 of them care enough about repetitive pasta injuries to buy your product.

There have been 442 unique visits to Sparkour since it was released, from a mix of countries, not all of which you would expect to have a thriving data science scene:

More interesting than the location is the corporation. Besides the usual assortment of universities, I've gotten hits from sources as varied as DuPont, ESPN, NOAA, Capital One, Ancestry.com, and all manner of health insurance companies. All of these companies have too much customer data and need something like Apache Spark to crunch the numbers so they can figure out what you ate for breakfast.

Here are the most popular articles that people read:

As for intrinsic satisfaction, I'm still enjoying the Sparkour experience. Unlike DDMSence, where there was nothing left to implement after a while, I have a nice backlog of potential articles that interest me. It usually takes about 4 hours to experiment with the concepts and write the source code (I start in Python, then convert into R, Scala, and finally, Java) and then another 3 hours to write the recipe prose. All of this costs me $0.12 an hour for servers, which I then shut down afterwards. Any given recipe is a low time commitment and I can release early and often!