CSV and JSON Round-Trips

Two serialization formats cover most day-to-day work. CSV is rows of fields (spreadsheets, exports); JSON is nested structures (APIs, configs).

import csv, json
from pathlib import Path

# CSV: always newline="" on open, always DictReader/DictWriter
with Path("people.csv").open(newline="", encoding="utf-8") as f:
    rows = list(csv.DictReader(f))          # each row: dict[str, str]

with Path("people.csv").open("w", newline="", encoding="utf-8") as f:
    writer = csv.DictWriter(f, fieldnames=["name", "age"])
    writer.writeheader()
    writer.writerows(rows)

# JSON: dumps/loads for strings, dump/load for files
record = {"name": "An", "tags": ["admin", "staff"]}
text = json.dumps(record, ensure_ascii=False, indent=2)
back = json.loads(text)

path = Path("record.json")
path.write_text(json.dumps(record), encoding="utf-8")
loaded = json.loads(path.read_text(encoding="utf-8"))

Round-trip honesty

Notice what each format does NOT preserve: CSV cells are strings — "30" must be converted back to int yourself. JSON has no tuples, no dates, and no Python-specific types. A "round trip" means designing the schema (what types, what's required) and validating at the boundary (module 5's discipline). Never feed untrusted JSON straight into your logic: check keys and types first.

Security footnote

json is safe for untrusted data. pickle is NOT — loading a malicious pickle executes code. Rule: json for anything that crosses a trust boundary, pickle only for your own trusted, local caches.