Functions, files and JSON

Real scripts reuse logic, handle failures, and read and write files. They also exchange data in JSON, the format APIs and AI tools speak. This lesson covers all of it.

Watch

Python Tutorial: Working with JSON Data using the json moduleCorey Schafer · 21 min

Functions

A function packages steps under a name so you can reuse and test them.

def parse_log_line(line):
    """Split a log line into date, time, level and message."""
    date, time, level, message = line.split(" ", 3)
    return {"date": date, "time": time, "level": level, "message": message}

line = "2026-03-01 10:02:11 ERROR disk /dev/sda1 is 93% full"
print(parse_log_line(line))

• def defines it, and return hands back a result.
• The text in triple quotes is a docstring: a short description of what it does.
• Give functions clear verb names, and keep each one doing one thing.
• Arguments can have defaults: def alert(metric, limit=90):.

Handling errors

Things fail: files are missing, networks drop, data is not what you expected. try and except let your script respond instead of crashing.

def to_percent(text):
    try:
        return float(text.strip("%"))
    except ValueError:
        return None

print(to_percent("93%"))
print(to_percent("n/a"))

Catch the specific error you expect (ValueError here), not everything. Catching all errors silently hides real bugs. When a failure should stop the program, let it, or raise your own error with a clear message.

Reading and writing files

Use with so the file is always closed, and pathlib for file paths that work on every system.

from pathlib import Path

folder = Path("data")
folder.mkdir(exist_ok=True)

notes = folder / "notes.txt"
notes.write_text("first line\nsecond line\n", encoding="utf-8")

with open(notes, encoding="utf-8") as f:
    for line in f:
        print(line.strip())

print(notes.exists(), notes.stat().st_size, "bytes")

Always pass encoding="utf-8" for text, or odd characters will break on some systems. Use mode="a" to append and never overwrite by accident.

JSON: the language of APIs

JSON is text that describes data with objects (key and value pairs), arrays (lists), strings, numbers, booleans and null. It maps almost one-to-one onto Python dictionaries and lists.

import json

server = {"name": "web-01", "cpu": 73.5, "tags": ["prod", "eu"], "maintenance": False}

text = json.dumps(server, indent=2)    # Python object -> JSON text
print(text)

back = json.loads(text)                # JSON text -> Python object
print(back["tags"][0])

with open("data/server.json", "w", encoding="utf-8") as f:
    json.dump(server, f, indent=2)

with open("data/server.json", encoding="utf-8") as f:
    print(json.load(f)["name"])

Memorize the four names: dumps and loads work on strings, dump and load work on files. Python True, False and None become true, false and null in JSON.

Nested data

API responses are often nested. Walk down one level at a time, and use .get() when parts may be missing.

import json

response = json.loads("""
{
  "user": {"name": "Ada", "teams": [{"name": "platform"}, {"name": "security"}]},
  "status": "ok"
}
""")

print(response["user"]["name"])
team_names = [t["name"] for t in response["user"]["teams"]]
print(team_names)
print(response.get("error", "no error"))

When you are unsure of the shape, print the response with json.dumps(..., indent=2) and read it first.

CSV files

CSV is the other common data format: rows of values separated by commas.

import csv

rows = [
    {"server": "web-01", "cpu": 91, "region": "eu"},
    {"server": "web-02", "cpu": 45, "region": "us"},
]

with open("data/servers.csv", "w", newline="", encoding="utf-8") as f:
    writer = csv.DictWriter(f, fieldnames=["server", "cpu", "region"])
    writer.writeheader()
    writer.writerows(rows)

with open("data/servers.csv", newline="", encoding="utf-8") as f:
    for row in csv.DictReader(f):
        print(row["server"], int(row["cpu"]))

Values read from a CSV are always text, which is why the code converts cpu with int(...). For anything bigger than a few rows, pandas (lesson 5) is easier.

Your own modules

When a script grows, move helper functions into another file, for example logtools.py, and use from logtools import parse_log_line. Keeping code in small files is a habit that makes both your own work and AI-written code easier to check.

Resources

Check yourself

1. Write a function parse_log_line that also returns None for a blank line. Test it on three lines.
2. Save a list of three server dictionaries to a JSON file, load it, and print the names.
3. What goes wrong if you use except: with no error type? Give an example.