imclab / madpug

Code for Lightning talk at Madison Python User Group (PyCall and call-julia-from-python)

##Calling Julia from Python (and Python from Julia)

Code for Lightning talk at Madison Python User Group on Sept 26, 2013.

Since this is a Python User Group, let's drive the main event from Python.

Dependencies

• Install Julia
• at the Julia REPL, Pkg.add("PyCall")
• git clone https://github.com/JuliaLang/IJulia.jl
  • in IJulia.jl/python/, run python setup.py install (may need sudo)

Running the Code

To start with, clone this repo. Then, in the cloned directory, open a Python REPL.

Run the following code:

import pyiseven
pyiseven.even(5)

You should get False. On the surface this seems pretty innocuous. We imported a python file, and it took a long time to load. Then, we called a function, which told us that 5 is not even. However, that ignores key implementation details (which are hinted at by all those dependencies).

Let's start with the python file, pyiseven.py.

import julia
j = julia.Julia()

def even(x):
  j.run("using IsOdd")
  return not j.run("IsOdd.odd(" + str(x) + ")")

First, we import the thing we installed from inside the IJulia.jl/python directory. Then, we start up the Julia process we'll be talking to.

In the even function, we import a Julia package called IsOdd. That string argument to run is just a snippet of Julia code. After that, we return the opposite of running IsOdd's function odd on x.

So far, we've discovered that we called Julia code from Python. To see that happen in the Python REPL:

import julia
j = julia.Julia() # this will take a longish time
j.run("2+2") #=> 4
j.run("sin(pi)") #=> 1.2246467991473532e-16
j.run("x = 5") #=> 5
j.run("x += 2") #=> 2

Now, let's look at the Julia side of the code, in the IsOdd module.

module IsOdd

using PyCall
@pyimport pyiseven

function odd(x)
  if x == 1
    true
  elseif x == 0
    false
  else
    pyiseven.even(x-1)
  end
end

end

The module keyword creates a Module (a namespace); the last end keyword closes it.

using PyCall imports that Julia package we added in the Dependencies section. @pyimport is a macro from PyCall, which takes a Python package name as it's argument. After @pyimporting pyiseven, we can use dot-notation to access its members, including the even function.

In the odd function, we first take care of a couple of base cases in our recursion, then we call right back into that Python function (decrementing x so that we'll eventually finish).

There are some obvious short-comings of this code, such as not handling negative numbers, but it is an example of a minimal mutual recursion between Python and Julia. It also breaks on my machine if you pass even an integer greater than 200. I don't yet know why.

More PyCall fun

If you install matplotlib, then you can have more fun calling Python from Julia.

using PyCall
@pyimport pylab
x = linspace(0,2*pi,1000); y = sin(3*x + 4*cos(2*x));
pylab.plot(x, y; color="red", linewidth=2.0, linestyle="--")
pylab.show() 

This should pop open a window with a nice graph of a dotted red squiggley line. :)