File Handling in Python

Use a context manager (with) so a file is closed even if reading or writing raises an exception. Text files should normally declare an encoding; UTF-8 is a portable default.

Text files and paths

pathlib.Path provides path operations without manual string concatenation.

from pathlib import Path
from tempfile import TemporaryDirectory

with TemporaryDirectory() as directory:
    path = Path(directory) / "notes.txt"
    path.write_text("alpha\nbeta\n", encoding="utf-8")
    lines = path.read_text(encoding="utf-8").splitlines()

print(lines)  # ['alpha', 'beta']

Path.read_text() is convenient for small files. Stream large files line by line instead of loading all content into memory.

CSV

Open CSV files with newline="" as required by the csv module, and specify the encoding separately.

import csv
from io import StringIO

buffer = StringIO(newline="")
writer = csv.DictWriter(buffer, fieldnames=["name", "score"])
writer.writeheader()
writer.writerow({"name": "Ada", "score": 9})
writer.writerow({"name": "Lin", "score": 8})

buffer.seek(0)
rows = list(csv.DictReader(buffer))
print(rows)

CSV values are read as strings unless you convert them. The format does not carry a universal schema, so document delimiter, quoting, missing-value, and type assumptions.

JSON

json.dump() and json.load() work with file objects; json.dumps() and json.loads() work with strings.

import json

record = {"name": "Ada", "active": True, "scores": [9, 8]}
encoded = json.dumps(record, ensure_ascii=False, sort_keys=True)
decoded = json.loads(encoded)

print(encoded)
print(decoded["scores"])  # [9, 8]

JSON has a smaller type system than Python. Dictionary keys become strings, tuples become arrays, and arbitrary Python objects are not serializable without an explicit conversion.

Binary and spreadsheet files

Use binary mode ("rb" or "wb") for bytes such as images or compressed data. .xlsx spreadsheets are not handled by the Python standard library; use a maintained library such as openpyxl or pandas and validate its version-specific API in the relevant data-tools lesson.

Never load untrusted pickle data. Unpickling can execute arbitrary code; use a data format designed for interchange when the source is not fully trusted.