Data Formats: JSON, CSV & Serialization¶

import json
from pathlib import Path
from datetime import datetime
from decimal import Decimal

# Basic serialization
record = {
    "user": "ada",
    "scores": [90, 85, 92],
    "active": True,
    "metadata": None,
}
text = json.dumps(record)                     # compact: no whitespace
text = json.dumps(record, indent=2)           # pretty-printed
text = json.dumps(record, indent=2, sort_keys=True)  # sorted keys

# Deserialization
obj  = json.loads(text)                       # string → Python object
obj  = json.loads(Path("data.json").read_text(encoding="utf-8"))

# File I/O — preferred over read_text + loads
with open("data.json", "w", encoding="utf-8") as f:
    json.dump(record, f, indent=2, ensure_ascii=False)  # non-ASCII preserved

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

# Custom encoder for non-JSON-native types
class AppEncoder(json.JSONEncoder):
    def default(self, obj):
        if isinstance(obj, datetime):
            return obj.isoformat()          # "2026-06-27T14:32:01"
        if isinstance(obj, Decimal):
            return float(obj)               # or str for exact precision
        if isinstance(obj, set):
            return sorted(list(obj))        # sets → sorted lists
        return super().default(obj)

complex_obj = {
    "created_at": datetime(2026, 6, 27, 14, 32, 1),
    "price":      Decimal("19.99"),
    "tags":       {"python", "api"},
}
json.dumps(complex_obj, cls=AppEncoder, indent=2)
# {
#   "created_at": "2026-06-27T14:32:01",
#   "price": 19.99,
#   "tags": ["api", "python"]
# }

import csv
from pathlib import Path

# ── Reading with DictReader ───────────────────────────────────────────────────
# newline="" is required on all platforms to prevent double-newline on Windows
with open("customers.csv", newline="", encoding="utf-8") as f:
    reader = csv.DictReader(f)
    # reader.fieldnames is populated from the header row
    for row in reader:
        print(row["name"], row["email"], row["spend"])
        # All values are strings — convert types yourself

# Handling missing columns gracefully
with open("customers.csv", newline="", encoding="utf-8") as f:
    reader = csv.DictReader(f)
    records = []
    for row in reader:
        records.append({
            "name":  row["name"].strip(),
            "email": row["email"].lower().strip(),
            "spend": float(row.get("spend", "0") or "0"),
        })

# ── Writing with DictWriter ───────────────────────────────────────────────────
fieldnames = ["id", "name", "email", "spend"]
with open("output.csv", "w", newline="", encoding="utf-8") as f:
    writer = csv.DictWriter(f, fieldnames=fieldnames)
    writer.writeheader()
    writer.writerow({"id": 1, "name": "Ada",  "email": "ada@x.io", "spend": 299.0})
    writer.writerow({"id": 2, "name": "Bob",  "email": "bob@y.io", "spend": 85.0})
    writer.writerows([                               # batch write
        {"id": 3, "name": "Carol", "email": "c@z.io", "spend": 412.0},
        {"id": 4, "name": "Dave",  "email": "d@w.io", "spend": 55.0},
    ])

# ── Tab-separated and custom delimiters ──────────────────────────────────────
with open("data.tsv", newline="", encoding="utf-8") as f:
    reader = csv.DictReader(f, delimiter="\t")  # TSV
    for row in reader:
        ...

# ── Reading from in-memory string (useful for testing) ───────────────────────
import io
csv_text = "name,score\nAda,95\nBob,82\n"
reader = csv.DictReader(io.StringIO(csv_text))
rows = list(reader)
# [{'name': 'Ada', 'score': '95'}, {'name': 'Bob', 'score': '82'}]

import pickle, json
from pathlib import Path

# pickle: full Python objects — TRUSTED DATA ONLY
data = {"model": some_sklearn_model, "created": datetime.now()}
Path("model.pkl").write_bytes(pickle.dumps(data))
loaded = pickle.loads(Path("model.pkl").read_bytes())  # ONLY safe for your own files

# NEVER do this with external data:
# user_data = request.body
# obj = pickle.loads(user_data)   # REMOTE CODE EXECUTION VULNERABILITY

# For ML models specifically: use joblib (wrapper around pickle with compression)
# import joblib; joblib.dump(model, "model.joblib"); joblib.load("model.joblib")

# For configs and API data: JSON is always the right answer
# For dataclasses → dict for JSON: use dataclasses.asdict()
from dataclasses import dataclass, asdict

@dataclass
class Config:
    host: str
    port: int
    debug: bool

cfg = Config(host="localhost", port=8080, debug=True)
json_str = json.dumps(asdict(cfg), indent=2)
# {"host": "localhost", "port": 8080, "debug": true}

# Restore: pass dict fields back as kwargs
cfg2 = Config(**json.loads(json_str))