What the Plot?!

A short tour of data visualization in Python.
Something of a companion talk to @jakevdp's survey - in this talk I drill down and give details on three popular Python plotting ecosystems.

Transcript

1. What the plot?! Tim Staley Oxford Python Meetup September 2018
2. What's up with Python Data-vis? My world, until recently: • Matplotlib • A bit of seaborn • What the hell's a bokeh anyway?
3. Then this happened
4. Credit: @jakevdp (PyCon 2017 talk)
5. Making sense of it all In this talk, I'll drill down, rather than survey the landscape (there will be code) • Brief refresher on Matplotlib and friends • Highlight some exciting new tools (holoviews! plotly!) • You don't have to know javascript!
6. THEME: Imperative Vs Declarative
7. (Thanks, @jakevdp!)
8. What's what, in Brief Matplotlib: Imperative, static • Works everywhere, plots anything, very robust • Super verbose • No interaction outside of zoomable notebook plots* • Slow for large data (though do note the rasterization hack option). * This is a lie for simplicity's sake.
9. What's what, in Brief Plotly: Imperative, Web / Interactive • Strong multi-language / collaboration support • Good (but limited) ready-made options • 3D plots! • Complex plotting still quite verbose / imperative • IMHO: Custom data-selection / interactivity dev-experience is painful right now (I may be doing it wrong).
10. Holoviews: Declarative, Web / Interactive • Wraps Bokeh/MPL - much less verbose than either • Declarative / data-driven interface ◦ This is weird at first but it's growing on me • Some rough edges still • Strong custom interaction / dashboarding potential • Links to datashader for big data What's what, in Brief
11. FIN
12. Further Reading • Jake Vanderplas' talk and slides • pythonplot.com • Galleries: MPL, Seaborn, Plotly, Holoviews • Notebooks accompanying this talk: https://github.com/timstaley/what-the-plot • Stuff I didn't mention: ◦ Altair ◦ Hypertools