My side project from the Christmas break is now available on the iOS App Store: the Harrisburg Crime Tracker.

Recently I’ve been digging into Julia for technical and statistical computing. It’s still in its early stages and things break frequently, but I find it a lot more amenable to my style of thinking than the more established technical languages like Matlab/Octave or R.

As I’ve ironed out the basic issues in hbg-crime.org’s data collection process, I’ve been starting to think about adding some slightly more sophisticated analytics. One feature I had in mind from the beginning is classifying crime reports based on their neighborhood, so I dug around and found this decent approximation of Harrisburg neighborhood boundaries. Google allows you to export these maps as KML files, so I just extracted the coordinates into two-dimensional vectors directly in source:

For whatever unknowable reason, the Harrisburg PD publishes its crime blotter in PDF format. It’s simple monospaced text with no images or extra formatting, but it comes as a PDF that has to be stripped in order to parse its contents. Here’s how I did that with Clojure and PDFBox.

I’ve been having a reasonably productive time getting up to speed with ClojureScript for my Harrisburg city crime tracker. One difficult bit of trial-and-error I came across is successfully compiling the Google Maps API into my app with advanced optimizations set. The Google Closure compiler munges identifiers so you need an “externs” file so it knows which names correlate to what from your CLJS code.