Skip to content

Refactoring your code

How many times have you found yourself copy/pasting code from one notebook to another? If it the answer is "many", then this essay probably has something for you. We're going to look at the practice of "refactoring" code, and how it applies in a data science context.

Why refactor

When writing code, we intend to have a block of code do one thing. As such, its multiple application should have a single source of truth. However, the practice of copying and pasting code gives us multiple sources of truth. Refactoring code, thus, gives us a way of establishing a single source of truth for our functions, which can be called on in multiple situations.

When to refactor

The short answer is "basically whenever you find yourself hitting copy+paste" on your keyboard.

How do we refactor

The steps involved are as follows.

  1. Wrap the semi-complex block of code in a function.
  2. Identify what you would consider to be an "input" and "output" for the function.
  3. Take specific variable names and give them more general names.

An example

Let's take the example of a chunk of code that takes a protein sequence, compares it to a reference sequence, and returns all of the mutations that it has. (We will only implemenet a naive version for the sake of pedagogy.)

sequence1 = ...
sequence2 = ...

mutations = []
for i, (letter1, letter2) in enumerate(zip(sequence1, sequence2)):
    mutations.append(f"{letter1}{i+1}{letter2}")
mutations = "; ".join(m for m in mutations)

This more or less should accomplish what we want. Let's now apply the ideas behind refactoring to this code block.

def mutation_string(reference, sequence, sep="; "):
    mutations = []
    for i, (letter1, letter2) in enumerate(zip(reference, sequence)):
        mutations.append(f"{letter1}{i+1}{letter2}")
    return f"{sep}".join(m for m in mutations)

You'll notice the three steps coming into play.

Firstly, we simply shifted the main logic of the code into a function definition.

Secondly, we then generalized the function a bit, by renaming sequence1 and sequence2 to what we usually intend for it to be, a sequence of interest and a reference sequence.

Finally, we defined those two as inputs, alongside a keyword argument called sep, which defines the separator between each mutation.

Bonus

On the basis of this function definition, we can do some additional neat things!

For example, in protein sequence analysis, our reference sequence is usually kept constant. Hence, we can actually create a custom mutation_string for our reference sequence using functools.partial by fixing reference to a particular value, thus eliminating the need to repetitively pass in the same reference string.

from functools import partial

protein1 = ...  # define the string here.

prot1_mut_string = partial(mutation_string, reference=protein1)

protein2 = ...  # define the string here.

mutstring = prot1_mut_string(sequence=protein2)

Where should this function be refactored to

You can choose to keep it in the notebook, and that would be fine if the function was used only in a single notebook.

If you find yourself needing to call on that same function from another notebook, do the right thing and create a utils.py (or analogous) Python module that lives in the same directory as the notebook. Then, import the refactored function from utils.py.

If you feel sophisticated, you can also create a custom Python library for your project. I will address this in a separate essay.

An anti-pattern, though, would be to attempt to treat the notebook as source code and import the function from one notebook into another. Notebooks are great for one thing: weaving functions together into an integrarted analysis. I'm of the opinion that we should use a tool the way it was intended, and bring in other tools to do what we need. In this respect, I think that DataBricks notebooks does the wrong thing by bowing to bad human first instincts rather than encouraging productive behaviours.

Where do we find time to do this

I hear this concern, as I went through the same concerns myself.

Isn't it faster to just copy/paste the code? What if I don't end up reusing the code elsewhere? Isn't the time then wasted?

In thinking back to my own habits, I realized early on that doing this was not a matter of technical ability but rather a matter of mindset.

Investing the time into doing simple refactoring alongside my analyses does take immediate time away from the analysis. However, the deliberate practice of refactoring early on earns back multiples of the time spent as the project progresses.

Moreover, if and when the project gets handed over "in production", or at least shared with others to use, our colleagues can spend less time is spent navigating a spaghetti-like codebase, and more time can be spent building a proper mental model of the codebase to build on top of.

On the possiblity of not reusing the code elsewhere, I would strongly disagree. Refactoring is not a common skill, while copy/pasting code is. Every chance we get to refactor code is practicing the skill, which only gets sharper and more refined as we do it more. Hence, even for the sake of getting more practice makes it worthwhile to do refactoring at every chance.

Concluding words

I hope this mini-essay demystifies the practice of code refactoring, and gives you some ideas on how to make it part of your workflow.

Thank you for reading!

If you enjoyed this essay and would like to receive early-bird access to more, please support me on Patreon! A coffee a month sent my way gets you early access to my essays on a private URL exclusively for my supporters as well as shoutouts on every single essay that I put out.

Also, I have a free monthly newsletter that I use as an outlet to share programming-oriented data science tips and tools. If you'd like to receive it, sign up on TinyLetter!