Have a question about programming? In 2025, you're likely asking an AI chatbot — ChatGPT, Claude, or Gemini, to name just three — to answer your questions. But before chatbots, many coders used Stack Overflow, a question-and-answer site that was written and edited by contributors from around the world. Stack Overflow had (and still has!) answers to millions of questions about programming, tagged by language, library, and topic.
Not all answers on the site were accurate, and many developers copied solutions a bit too quickly, without truly understanding what they were doing. But Stack Overflow was certainly a big improvement over previous repositories for programming questions and answers. (It also provided participants in my classes with solutions to the exercises I was giving. Some of those solutions even worked!) I have to imagine that nearly every programmer in the last decade has used Stack Overflow at least once, and probably much more often than that.
For a number of years, Stack Overflow has surveyed its users to learn more about trends in programming. This year's results were published about two months ago as a set of summaries (https://survey.stackoverflow.co/2025/work/). Now that the raw data is downloadable from Stack Overflow, I thought it would be interesting to examine some of their findings, seeing if we can replicate them using Pandas.
Data and five questions
This week's data comes from Stack Overflow's 2025 survey. I found data is a bit hard to find, even though it's on the main "Stack Overflow Developer Survey" page at https://survey.stackoverflow.co/. Perhaps that's because the annual survey summary page (https://survey.stackoverflow.co/2025/work/) is more attractive, and is linked to by more sites.
Learning goals this week include memory optimization, categorization, grouping, and working with text (including regular expressions).
Paid subscribers, including members of my LernerPython.com site, can download the data file from the end of this message, as well as participate in office hours, download my notebooks, and open the notebooks with a single click.
Here are this week's questions. I'll be back tomorrow with my solutions and explanations.
- Read the 2025 survey data into a Pandas data frame. How much memory can you save by using categories instead of regular string ("object") columns? The
JobSatcolumn indicates, on a scale from 0-10, how satisfied people are in their jobs. Divide the population into "dissatisfied" (0-3), "neutral" (4-7), and "satisfied" (8-10). What proportion of respondents are in each category? - What proportion of developers have at least a four-year university degree? Do we see a difference between the United States and other countries? What proportion of developers have a master's degree or PhD? Again, do we see a difference between the United States and other countries? What country has the greatest proportion of developers with masters and PhD degrees?