A view from the boundary – JuliaCon 2015


Re-posted from: http://ljuug.tumblr.com/post/123026364378

by Malcolm Sherrington

After JuliaCon 2014, Julia was described in a blog posting as a language for Geeks. I had recently formed the London Julia User Group, specifically to spread the word that Julia is definitely NOT for Geeks.  True, it is being developed by some remarkable people, but it’s strength is that is possibly the easiest and most elegant of languages  to learn and use. The astonishing thing is just how much can be achieved in Julia alone, without having to dip a toe in other languages such as C/C++ or Java.

So I decided to do my bit and write a book aimed at the data scientist and the would be convert to Julia, even though the language has not yet reached version 1.0. With the book nearing its final stages I indulged myself with a trip to this year’s JuliaCon, especially since it was being held in its spiritual home at MIT, a place I’d not visited for over 25 years.

The first day was given over to an introductory workshop by fellow Brit, David Sanders.  I had already seen David’s excellent offerings at SciPy 2014 on Youtube – talk about putting your head in the lion’s den! This workshop was equally enjoyable and I recommend it to any beginners to Julia and but also to those who have progressed further.

The next two days were devoted a series of streamed talks, soon to be put up on the Internet. My personal interests were satiated by a talk on Econometrics from Spencer Lyon, and then, as a member of the Baker Street Irregulars Astros here in London, I attended a talk by Kyle Barbary on the work of the JuliaAstro group.  There was also a couple of talks: on visualisation from Zachary Yedida ( Simple Fast Multimedia Library), and one by Shashi Gowda (Escher: A new way to make and deploy GUIs), both of which sadly will not make it into the first edition of the book.  

There was also yet more seminal work from our own Mike Innes, he of Lazy.jl and the Juno IDE, this time on building web-based applications from Julia. How he has managed to combine this with his final year studies in Physics at Oxford I cannot imagine. Hopefully Mike can come and tell us all about his work soon at the LJuUG

On the next day there was a presentation by Keno Fischer about the amazing work he has been doing on combining Julia and C++. Apparently this has involved debugging and contributing to the next release of the LLVM compiler as well as work on the Julia side.  

However the highlight of the conference for me came on the last day in the form of the workshop by the Julia Parallel group.  Even now I am completely blown away by this.  Anyone who has struggled with current approaches to parallelism based on MPI and Hadoop will appreciate the importance of the work this group is doing. I believe this may well be the killer-app which wooes users from the darkside.  After listening to and meeting the Julia Parallel group I can see that it is in safe hands.

On a personal note,  I met with one of the book’s reviewers, Dan (Milktrader) Wlasiuk and promised to make real contributions to the JuliaQuant group. Together with my work on Econometrics at London University (Birkbeck) in the coming year I hope to make some material developments to Julia, rather than just writing about it.

Next year my view may be from the square and not from the boundary. Who knows?