JSON

Saving a dict

JSONJSONJavaScript Object Notation: a text format for lists, dicts, strings, numbers, true, false, and null. Almost every language and website reads it. is a text format for nested data: dicts, lists, strings, numbers, true, false, and null. Almost every language and website can read it. Python's json module turns your data into JSON text, which is called serializingserializingTurn data, like a dict or a list, into text that can be saved or sent, and read back later. it, and turns JSON text back into data:

import json

contacts = {"Selam": "0900 000 003", "Abel": "0900 000 001"}
with open("contacts.json", "w", encoding="utf-8") as f:
    json.dump(contacts, f)

with open("contacts.json", encoding="utf-8") as f:
    loaded = json.load(f)
print(loaded)
print(loaded == contacts)

dumpdumpTurn data into JSON text and write it to an open file: json.dump(data, f). json.dumps(data) returns the text instead. writes the data to an open file as JSON; loadloadRead JSON text from an open file and turn it back into Python data: json.load(f). json.loads(text) reads from a string. reads it back as a new dict. Look at contacts.json in Files: it looks a lot like Python, but JSON always uses double quotes. json.dumps() and json.loads(), with an s for string, do the same with a string instead of a file.

Amharic in JSON

Save a contact whose name is in Amharic, and JSON does something you might not expect. Predict:

import json

print(json.dumps("ሰላም"))

"\u1230\u120b\u121d"

By default, json writes every character beyond the basic English letters, digits, and symbols as \u and its code point in hex, the escape from lesson 6.3: \u1230 is ሰ. It's still correct JSON and loads back as ሰላም, but a person can't read the file. ensure_ascii=False keeps the characters as they are, and indent=2 puts each item on its own line:

import json

contacts = {"ሰላም": "0900 000 003", "Abel": "0900 000 001"}
with open("contacts.json", "w", encoding="utf-8") as f:
    json.dump(contacts, f, ensure_ascii=False, indent=2)

print(json.loads(json.dumps("ሰላም")) == "ሰላም")

Open contacts.json in Files. This is the line to remember for saving: json.dump(data, f, ensure_ascii=False, indent=2), with encoding="utf-8" on the open().

What changes on the way back

JSON has fewer types than Python. Strings, numbers, and lists come back as they were, and True, False, and None are saved as true, false, and null and come back as True, False, and None. Two things change. Predict:

import json

data = {1: "a", "t": (1, 2)}
print(json.loads(json.dumps(data)))

{'1': 'a', 't': [1, 2]}

If you picked the first answer, you expected JSON to give back exactly what you saved. If you picked the second, you caught the tuple but not the key. In JSON, a dict's keys are always strings, so 1 comes back as '1'; and JSON has no tuples, so (1, 2) comes back as the list [1, 2]. That bites when you save a dict with number keys:

Some values can't be saved at all. Python's datetime module has a date type for calendar dates (lesson 8.6 shows more): date(2026, 10, 1) is 1 October 2026, and its .isoformat() method gives it as text, "2026-10-01". JSON has no dates, so saving a date raises a TypeError; save its text instead:

import json
from datetime import date

paid = date(2026, 10, 1)
try:
    print(json.dumps({"date": paid}))
except TypeError as e:
    print(e)
print(json.dumps({"date": paid.isoformat()}))

except ... as e (lesson 7.3) gives the error a name, and print(e) shows its message. Sets can't be saved either; turn one into a list first. That's why Birr Ledger (lesson 8.9) keeps each date as text like "2026-10-01".

So when do you use which? CSV is for flat tables that people may open in a spreadsheet, one row per record, every value text. JSON is for nested data, like a list of dicts whose values are lists, and it keeps numbers, true/false, and null as types. A ledger of expense dicts is JSON's job.