Tutorials · Chapter D (4/4) · ~9 min

File handling

Try it → see it → read → next

Read, transform, and save a small text file without leaking resources or data.

Try yourself

Code Lab

Data: tables & simple stats

Run the average, then print the top student name.

Python

# Tiny data table (list of dicts)
rows = [
  {"name": "Ada", "score": 88},
  {"name": "Lin", "score": 92},
  {"name": "Sam", "score": 79},
]

# Average score
avg = sum(r["score"] for r in rows) / len(rows)
print("average", round(avg, 1))

# Challenge: print the student with the highest score

Recap

What you just did

You moved data through a complete file loop: open → read → transform → write. Using a with block also closed the file automatically, even if later code failed. That habit matters once files become datasets and model artifacts.

Teach

How it works

from pathlib import Path

source = Path("labels.txt")
lines = source.read_text(encoding="utf-8").splitlines()
clean = [line.strip().lower() for line in lines if line.strip()]

Path("clean-labels.txt").write_text(
    "\n".join(clean),
    encoding="utf-8",
)
print(f"saved {len(clean)} labels")

For larger or streamed files, use with source.open(...) as file: and process one line at a time.

Path identifies where data lives
Read turns bytes into text or records
Transform makes the data useful
Write persists an artifact you can inspect

Mental model: files are handoff points between program runs.

Use it

When you'd use this

Loading prompt examples or labels from disk
Saving cleaned data before model training
Writing predictions and logs for later review

Watch out

Opening with write mode can replace an existing file. Check the target path before saving. Specify text encoding, avoid loading enormous files all at once, and never print secrets from configuration files into logs.

Try next

Try this next

Add one duplicate label, then update the transform so the output contains each label once.