JSON and CSV in Python: Read, Write and Convert Data with Worked Examples

JSON and CSV both move data between programs, but toy-file code often breaks on nested values, quoted commas, numeric strings or malformed input. JSON can preserve nested structures and basic value types, while CSV is a flat table whose parsed cells begin as strings. The same student records can remain nested in JSON or become flat CSV rows, but the conversion must define a schema and preserve intended types. The Coding & Skill Development path connects these format boundaries to broader file-processing and programming practice.

JSON and CSV in Python represent different data shapes

JSON contains objects, arrays, strings, numbers, booleans and null. Its Python type map is: object to dict, array to list, string to str, integer-looking number to int, decimal to float, true and false to True and False, and null to None. Object keys are strings.

CSV is delimited rows, normally with a header. In 101,Asha,88,79,"Needs Python, SQL", the quoted comma belongs to one field. line.split(",") gives the wrong field count, so let csv.reader or csv.DictReader parse it.

Neither is universally better; choose by shape.

Python's JSON and CSV APIs: choose by input boundary

json.loads(text) converts a JSON string to Python objects; json.load(file_object) reads an open text file. json.dumps(value) returns a string; json.dump(value, file_object) writes to a file. The trailing s means the boundary is a string.

import json

item = json.loads(
    '{"topic": "CSV", "rows": 3, "ready": true, "note": null}'
)
print(item)
print(type(item["rows"]).__name__)
print(type(item["ready"]).__name__)
print(item["note"] is None)

The outputs are {'topic': 'CSV', 'rows': 3, 'ready': True, 'note': None}, int, bool and True. json.dumps({"b": 2, "a": 1}, sort_keys=True) returns '{"a": 1, "b": 2}' without altering the dictionary.

For CSV, reader and writer handle positional rows; DictReader and DictWriter use header-named fields. Header values become keys, but "101", "88" and "79" remain strings. CSV Files Explained: Parsing Rules, Worked Records and Exam Traps traces embedded newlines, escaped quotes and parser states; Python's csv module applies those rules after receiving the text.

Read, update and write a JSON file in Python

Start with students.json:

[
  {"roll_no": 101, "name": "Asha", "scores": {"python": 88, "sql": 79}, "active": true},
  {"roll_no": 102, "name": "Ravi", "scores": {"python": 76, "sql": 91}, "active": false}
]

Read, change and write the data:

import json

with open("students.json", "r", encoding="utf-8") as file:
    students = json.load(file)

asha_total = students[0]["scores"]["python"] + students[0]["scores"]["sql"]
asha_average = asha_total / 2

students[1]["active"] = True
students.append({
    "roll_no": 103, "name": "Meera",
    "scores": {"python": 84, "sql": 90}, "active": True
})

with open("students-updated.json", "w", encoding="utf-8") as file:
    json.dump(students, file, indent=2, ensure_ascii=False)

Before updating, students is a list of length 2; Asha's name is Asha, her Python score is integer 88, and Ravi's active value is Boolean False. Asha's total is 88 + 79 = 167, and her average is 167 / 2 = 83.5. indent=2 changes presentation; with UTF-8, ensure_ascii=False keeps Unicode readable.

Read, calculate and write CSV data with DictReader and DictWriter

Use this scores.csv with its quoted comma:

roll_no,name,python,sql,note
101,Asha,88,79,"Needs Python, SQL"
102,Ravi,76,91,Strong SQL
103,Meera,84,90,Balanced

import csv

qualified = []
with open("scores.csv", "r", encoding="utf-8", newline="") as file:
    reader = csv.DictReader(file)
    for row in reader:
        average = (int(row["python"]) + int(row["sql"])) / 2
        if average >= 85:
            qualified.append({
                "roll_no": row["roll_no"], "name": row["name"],
                "average": average
            })

with open("qualified.csv", "w", encoding="utf-8", newline="") as file:
    writer = csv.DictWriter(file, fieldnames=["roll_no", "name", "average"])
    writer.writeheader()
    writer.writerows(qualified)

The calculations are Asha (88 + 79) / 2 = 83.5, Ravi (76 + 91) / 2 = 83.5, and Meera (84 + 90) / 2 = 87.0. Only Meera passes average >= 85, so qualified.csv contains:

roll_no,name,average
103,Meera,87.0

The module handles quoting. newline="" prevents text-mode translation from interfering with CSV newline handling.

Convert JSON to CSV only after defining a flat schema

For the updated list, define roll_no,name,python,sql,active. Flatten student["scores"]["python"] into python and its SQL counterpart into sql:

import csv

with open("students.csv", "w", encoding="utf-8", newline="") as file:
    fields = ["roll_no", "name", "python", "sql", "active"]
    writer = csv.DictWriter(file, fieldnames=fields)
    writer.writeheader()
    for student in students:
        writer.writerow({
            "roll_no": student["roll_no"], "name": student["name"],
            "python": student["scores"]["python"],
            "sql": student["scores"]["sql"], "active": student["active"]
        })

The rows are 101,Asha,88,79,True, 102,Ravi,76,91,True and 103,Meera,84,90,True. All are true: Ravi was updated; Asha and Meera were already active.

Conversion back is not automatic. Apply int() to roll_no, python and sql, rebuild scores, and map active only through {"True": True, "False": False}. Reject other spellings. Validation and aggregation are a separate step; Python File Handling and Data Analysis: A Complete Worked Guide follows that end-to-end workflow after ingestion.

Common JSON and CSV errors: cause, failure and repair

Never use eval() for JSON. Where malformed external input is expected, catch JSONDecodeError and report its line and column. Confirm roll_no, name, python and sql occur in reader.fieldnames. Report every skipped row's number and reason.

JSON and CSV exercises: trace types, quoting and one-pass work

1. After data = json.loads('{"scores": [72, 88, 95]}'), data["scores"] is a list. Its total is 72 + 88 + 95 = 255, average is 255 / 3 = 85.0, and maximum is 95.

2. next(csv.reader(['Asha,"Python, SQL",88'])) returns ['Asha', 'Python, SQL', '88']. Quoting protects the comma, but the score stays a string because CSV carries text.

3. Add 104,Neha,92,86,Strong in both to scores.csv. Neha's average is (92 + 86) / 2 = 89.0, so the filter output has Meera 87.0, then Neha 89.0.

DictReader processes a fixed-width record in one pass, so scanning n rows takes O(n) row visits; Time Complexity and Asymptotic Notation: Big-O develops the notation. DictReader can yield one row at a time, while json.load materialises the whole document.

JSON and CSV in Python: the short version and next step

• Choose the format by data shape.

• Use load and dump for files, loads and dumps for strings.

• Open text files with UTF-8.

• Let the csv module handle quoting.

• Convert CSV strings before arithmetic.

• Validate malformed or missing fields explicitly.

Asha's JSON average is 83.5, and the first CSV filter row is 103,Meera,87.0. Read students-updated.json, write roll_no,name,python,sql,active, then read it back and rebuild nested JSON. Meera's final CSV row must be 103,Meera,84,90,True.

For a structured route through Python syntax, file handling and practice, continue with Python Course: Concepts, MCQs & Coding.