Friday, 25 January 2013

AMS 2013

As mentioned by Andrew, I spent a few days in early January in Texas for the 93rd American Meteorological Society annual meeting, with the majority of my time spent in the Python symposium. Interestingly, out of the nearly forty symposia/conferences taking place, this was the only one which was focussed on a single core technology. I'm not sure if that says more about the evangelical nature of the Python community, or the genuine benefits brought by Python adoption ... perhaps we're just lucky to have someone like Johnny Wei-Bing Lin putting in the hard work to get it organised. Thanks Johnny!

As you'd expect, there were plenty of interesting talks on a wide range of subjects, but a couple of themes stood out for me.

Firstly, with Iris we are trying to free scientists in the AOS (Atmospheric and Oceanic Sciences) from low-level issues (such as dealing with the intricacies of file formats), and let them work at a higher logical level. And it was good to see that same desire reflected in topics ranging from representing radar data to standardised calculations for potential temperature, etc. Not only is there an appetite for such things, there are other people who are trying to achieve the same goals, and it was great to be able to meet them.

Secondly, there there are plenty of up-and-coming technologies for improving Python's performance, whether that be handling large datasets, speeding up numeric computation, or wrestling with parallelism. Some projects have been around for a while (e.g. numexpr), others have a "Work in progress" sign on the door (e.g. numba), and still others have "Coming soon" (e.g. ODIN). But it's clear from the competition and flux that the domain as a whole is still in its early days. Perhaps the key pieces of the next generation/level of high-performance tools already exist but have yet to establish dominance, or maybe they've yet to be created. Either way, it's an exciting time to be involved.

And speaking of involvement, scalable performance is obviously not a need that's specific to AOS, so it's especially important to reach out to the broader scientific Python community. I'd like to echo a sentiment expressed in the symposium that we should be encouraging AOS attendance at the various SciPy events. Having been to EuroSciPy 2012, I can attest for the rich seam of technology, ideas, and inspiration that is out there just waiting to be mined.

If you weren't able to attend AMS2013 then keep an eye out for the recorded presentations. The AMS2013 website currently has them slated for release by "late February".