Serialization-json-pickle-csv

Serialization: json, csv, pickle

Serialization converts Python objects into a storable/transmittable format, and deserialization reverses it.

json: Human-Readable, Cross-Language

import json
 
data = {"name": "Bob", "age": 25, "active": True, "tags": ["a", "b"]}
 
# Object -> string
json_str = json.dumps(data)                       # compact string
json_str_pretty = json.dumps(data, indent=2)        # pretty-printed, human readable
json_str_sorted = json.dumps(data, sort_keys=True)     # keys alphabetically ordered
 
# string -> Object
parsed = json.loads(json_str)
 
# Object -> file
with open("data.json", "w") as f:
    json.dump(data, f, indent=2)
 
# file -> Object
with open("data.json") as f:
    loaded = json.load(f)

Type Mapping

Python	JSON
dict	object
list, tuple	array
str	string
int, float	number
True/False	true/false
None	null

JSON cannot represent all Python types natively

Tuples become JSON arrays and come back as lists (the tuple-ness is lost on round-trip). Custom objects, datetime, set, and Decimal all need custom handling via default= in json.dumps() or a custom decoder.

from datetime import datetime
 
def default_serializer(obj):
    if isinstance(obj, datetime):
        return obj.isoformat()
    raise TypeError(f"Type {type(obj)} not serializable")
 
json.dumps({"created": datetime.now()}, default=default_serializer)

csv: Tabular Data

import csv
 
# Writing
with open("output.csv", "w", newline="", encoding="utf-8") as f:
    writer = csv.writer(f)
    writer.writerow(["name", "age"])          # header row
    writer.writerow(["Bob", 25])
    writer.writerows([["Amy", 30], ["Tom", 22]])   # multiple rows at once
 
# Reading
with open("output.csv", encoding="utf-8") as f:
    reader = csv.reader(f)
    header = next(reader)          # consume the header row separately
    for row in reader:
        print(row)                    # each row is a list of strings

Always pass newline="" when opening a CSV file for writing

Without it, Windows systems can introduce extra blank lines between rows, because both Python’s text mode and the csv module try to handle newline translation.

DictReader / DictWriter (Preferred for Named Columns)

with open("output.csv", newline="", encoding="utf-8") as f:
    reader = csv.DictReader(f)
    for row in reader:
        print(row["name"], row["age"])     # access by column name, not position
 
with open("output.csv", "w", newline="", encoding="utf-8") as f:
    fieldnames = ["name", "age"]
    writer = csv.DictWriter(f, fieldnames=fieldnames)
    writer.writeheader()
    writer.writerow({"name": "Bob", "age": 25})

For anything beyond quick scripts, consider pandas

csv.DictReader is great for simple line-by-line processing. For actual data analysis (filtering, grouping, joining, statistics), pandas.read_csv() is far more capable, see the dedicated Pandas folder.

pickle: Python-Specific Binary Serialization

Can serialize almost any Python object (including custom classes), but is NOT human readable, NOT cross-language, and NOT safe to load from untrusted sources.

import pickle
 
data = {"name": "Bob", "scores": [1, 2, 3]}
 
# Object -> bytes
with open("data.pkl", "wb") as f:
    pickle.dump(data, f)
 
# bytes -> Object
with open("data.pkl", "rb") as f:
    loaded = pickle.load(f)

Never unpickle data from an untrusted source

Unlike json.loads(), pickle.load() can execute arbitrary code during deserialization. Only unpickle files you created yourself or that come from a fully trusted source. This is a genuine, well-documented security vulnerability, not a theoretical concern.

Choosing the Right Format

Need	Use
Human-readable, cross-language, web APIs	json
Simple tabular/spreadsheet-like data	csv (or pandas for analysis)
Full-fidelity Python objects, same-trust-boundary only	pickle
Structured tabular analytics on large data	Parquet (via pandas/pyarrow), not covered here