These are the slides and IJulia notebooks from a talk by Steven G. Johnson at EuroSciPy 2014:
Julia (julialang.org) is a new language targeted at scientific computing, which combines the high-level abstractions and dynamic interactivity of languages like Python with the performance of low-level languages like C, thanks to being designed from the beginning for efficient just-in-time compilation by LLVM. But a major challenge for any young programming language is the availability of a large ecosystem of mature libraries and tools. To overcome this difficulty, Julia is "bootstrapping" off of the Python ecosystem, both by making it easy to call Python code and also by exploiting infrastructure such as IPython/Jupyter.
This talk will begin with an introduction to the Julia language, both explaining why it is able to attain C-like performance in many cases. At the same time, Julia supports a number of unusual programming features, such as multiple dispatch (a kind of generalization of object-oriented programming) and metaprogramming. We will also describe how Julia connects with Python via the PyCall package, which enables straightforward, low-overhead calls to Python libraries, copy-free sharing of NumPy arrays and other large data structures, and even sharing higher-order callback functions. This gives Julia direct access to SciPy and numerous other Python packages, such as SymPy and Matplotlib. Another key component of the Python universe is IPython, and we will explain how connecting to the IPython "Jupyter" front-end from an IJulia back-end allows Julia to benefit from IPython's rich multimedia notebook interface, and how Julia can even use IPython 2's interactive-widget infrastructure to provide truly interactive computations.
Although most Julia–Python interaction is from Julia users calling Python, there is potential for benefits to flow in both directions in the future. The same PyCall software that allows Julia code to call Python can also allow Python code to call Julia, with the same data-sharing and rich interactivity. And Julia code can also be compiled to C-compatible interfaces—currently, this is mainly used to pass Julia callback routines to C library functions (including libpython), but in the future the same facility should allow the generation of C-callable libraries written in Julia.
To use the IJulia notebooks, you will need to download Julia and install IJulia: see the IJulia installation instructions.
-
Slides in PDF (4MB) and PowerPoint (pptx) (3.5MB) formats.