Reading and understanding other people's code

Programmers read code far more than they write it. It is a skill you can train.

The method, in order:

1. Read the README and the directory structure — what kind of program is this?
2. Find the entry point: main(), if __name__ == "__main__", app = FastAPI(), a CLI command.
3. Follow a single flow all the way — one call, one button press. Not the whole codebase.
4. Search for names instead of reading everything: grep -rn "save_result" .
5. Run it with a breakpoint or a few prints and see what actually happens.
6. Read the tests — they show how the code is meant to be used, often more clearly than the code does.

You do not need to understand everything. You need to understand the piece you are going to change and what touches it.

# 1. An overview: what is there and how big is it?
ls -R src | head -40
wc -l src/**/*.py | sort -n | tail

# 2. Find the entry point
grep -rn "__main__\|def main\|FastAPI(" src/

# 3. Where is something defined and where is it used?
grep -rn "def save_result" src/            # the definition
grep -rn "save_result(" src/ | grep -v def # every call

# 4. What has changed recently, and why?
git log --oneline -15 -- src/model.py
git log -p -1 --follow src/model.py | head -60

Write a flow diagram in words while you read:

cli.py:main → reads the arguments → data.read_csv(file) → model.train(X, y) → report.write(result, out)

Four lines that you can show someone and ask «is that right?». It is the fastest route to understanding — and it reveals immediately if you have missed something.