Use ChatGPT to Improve Code

Contributed by Mike Loukides

Difficulty level: Easy

You can ask ChatGPT for suggestions about how to improve programs that you’ve written. This can be the beginning of a larger strategy that culminates in ChatGPT writing new code. But it’s often best to start with a high-level discussion, just like you would with another engineer. Here’s an example:

Prompt: How could this program to read and analyze survey data in a spreadsheet be improved? Do not write code; just make suggestions.

from util import *
import pandas as pd

df = pd.read_excel("data.xlsx")
df.fillna("No answer", inplace=True)
cols = df.columns

totalrows = len(df)
print("total answers:", totalrows)

name = 0
for col in cols
    name += 1
    noans = len(df [df[col] == "No answer"])
    nonnull = totalrows - noans
    print(col)
    print("No answer:", noans, "Total answers:", nonnull)
    answers = set(df[col])
    for answer in answers :
        adf = df [ df[col] == answer ]
        nanswer = len(adf)
        p = pct(nanswer, nonnull)
        if answer != "No answer" :
            print("   ", answer, ":", nanswer, ":", p)
    print()

ChatGPT: Here are some suggestions to improve the program:

1. Function encapsulation: Consider breaking the code into smaller functions, each with a specific task, such as loading data, preprocessing, and analyzing the data. This will improve code readability and make it easier to maintain and debug.

2. Variable naming: Use more descriptive variable names to make the code more understandable. For instance, instead of using name, you ...