Roundup of PyData London Meetup on 05-July-2016

My introduction to PyData London was through my boss Brendan, who recommended the conference to showcase a library that I'd built at the time. I got the wonderful opportunity to present at the conference last year, and was blown away by the community. Since then, I've tried to attend a few meetups, although I'm always on the waitlist. The community has grown massively from 1000+ members this time last year to over 3500 now... so much so that they have to host an overflow meetup at a pub nearby! The main meetup is hosted at AHL's office in Monument, and they generously provide beer and pizzas (and now gluten-free sandwiches!) to all attendees. Here's a roundup of last night's meetup #24.

It started off with community-related announcements by the organisers Emlyn Clay and Ian Ozsvald. Headlines are... NumFOCUS have received a generous grant from the Sloan Foundation to provide more awesomeness to the community. PyData EuroPython is taking place in Spain later this month and registration is still open. Ian, who has been advocating a switch to Python 3 for months, shared his survey results and it looks like one-third of the data scientists in London are now on Python 3.4+. We, at OpenSignal, have migrated most of our projects over to Python 3 as well. The package of the month is PyTables which provides a Pythonic layer for HDF5, a data model that works extremely well for time series data.

The first talk on writing clean code in Jupyter notebooks was co-presented by Vlad Kazantsev and Katya Nerush. I use notebooks a lot but mostly for prototyping / presenting results, and it was interesting to see that they were using it in production at Product Madness. The talk was mostly about coding best practices referencing "Uncle Bob" Martin's work on the clean coder. A cool package that was recommended was pytest-ipynb that can automatically detect and run tests within notebooks. You can find the slides of their talk here.

The second talk was by Dimitrios Kouzis-Loukas who took us through his journey of writing a book on Learning Scrapy. This was quite eye-opening for me and I was inspired to see that there were a hand full of people within the community who have written books. I've been dabbling with Scrapy and Scrapinghub lately and it became apparent to me during the talk that there's more for me to learn there... time to get the book. Effective product marketing I say! 

UPDATE [16-July-2016]: The slides of Dimitrios's presentation can be found here and here's the video recording on YouTube:

There were two lightning talks in the end - one by Steve Holden on Python Equivalences and the other by Lev Konstantinovskiy, who organises the London Data Science Journal Club meetup, on topic modelling in gensim. The evening ended with a hilarious take by Emelyn on Brexit from a data perspective. Here's an interesting stat... only 9% of the Leavers read the Guardian!

On the whole, it was a lovely evening with some great talks. Looking forward to the next.

Installing Scientific Python on Mac OS X using Homebrew and Pip

I encountered some difficulty while installing the SciPy stack on my Mac. I tried using Anaconda and Enthought Canopy as these contained all the essential packages in one installer. They both failed to install scipy though, possibly due to the issue with the Apple LLVM compiler that comes with the XCode 5.1.1 update. After some google-searching, I found this wonderful blog post that uses Homebrew and Pip and it almost worked for me. Here's how I fixed some of the additional issues, not detailed in that post.

Read More