Managing state

The notebook is an interactive computing environment. It lets you execute code in a step-by step fashion. This is powerful for thinking through a problem and taking your next steps based on what you see or do in the cell above. But it can lead to some challenging, non-intuitive behaviours if you run cells out of order or multiple times. In this notebook, we capture some “gotchas” and tips for avoiding tying your notebook in a knot. If you have ideas or experiences to contribute here, contributions to this repository are welcome!

import numpy as np
import matplotlib.pyplot as plt

Variable assignment

Sometimes as you are working, you might do simple computations, updates to a variable. One way to do this is to just over-write the variable name. The danger here is that if you re-run the second or third cell below, the value will update again, which may not be intended.

If you have a large notebook and update a parameter above, re-run the cells below, this can be easy to miss and generate non-intuitive results.

a = 2

a *= 2
print(a)

4

a = a*2
print(a)

8

A safer way to do this operation is to create a new variable with a different name. No matter how many times we execute the second cell below, the value will always be 4

a = 2

a_times_2 = a*2
print(a_times_2)

4

Scope of functions

The scope of variables assigned at the top level in the notebook (e.g. not inside a function or class) is global. This means you can still use them inside of a function. In the second cell below, we create a sine wave with a given frequency.

frequency = 1

def sketchy_sine_wave(x):
    return np.sin(2*np.pi*frequency*x)

x = np.linspace(0, 1, 100)
plt.plot(x, sketchy_sine_wave(x));

If we change the value of frequency, then the the behaviour of sine_wave changes. So even though the second cell has code that is identical to the cell above, the output is different. This can be confusing!

frequency = 2

plt.plot(x, sketchy_sine_wave(x));

A safer approach is to keep the scope of variables inside a function local to that function and assign default-values to parameters as appropriate. This will also allow you to much more easily pull functions / classes that you write in the notebook into a python file to create a module if you want to later re-use that code.

def safer_sine_wave(x, frequency=1):
    return np.sin(2*np.pi*frequency*x)