How can I read CSV rows by column name using DictReader?

Use csv.DictReader: by default, it uses the first row as field names and gives you each following row as a dictionary keyed by those names. [S1] For example, if a header is name,department, a row can be accessed as row["department"].

The example below assumes the file is named people.csv, has a header row, and that name and department are required headers. It reports other headers, stops if a required header is absent, and flags rows with missing or extra fields. Adjust the assumed names and filename to match your file.

import csv

expected = {"name", "department"}
missing_value = object()

with open("people.csv", newline="", encoding="utf-8") as file:
    reader = csv.DictReader(
        file,
        restkey="__extra_fields__",
        restval=missing_value,
    )
    headers = reader.fieldnames or []
    header_set = set(headers)

    missing_headers = expected - header_set
    unexpected_headers = header_set - expected

    if missing_headers:
        print("Missing required headers:", sorted(missing_headers))
    if unexpected_headers:
        print("Unexpected headers:", sorted(unexpected_headers))

    if missing_headers:
        raise SystemExit("Cannot process rows without required headers")

    for row_number, row in enumerate(reader, start=2):
        extras = row.get("__extra_fields__")
        missing_fields = [
            name for name in expected
            if row.get(name, missing_value) is missing_value
        ]

        if extras or missing_fields:
            print(
                f"Row {row_number}: extra values={extras}, "
                f"missing fields={sorted(missing_fields)}"
            )
            continue

        print(row["name"], row["department"])

Here, restkey gives extra row values a named place, and restval marks missing values with a distinct sentinel. The standard defaults instead place extra values under None and fill missing values with None. [S1]

A dictionary can be created even when the CSV header does not match what your code expects. Checking reader.fieldnames lets you detect absent required columns before trying to read them. The documentation describes inspecting field names and comparing them with required names; the caller decides what to do when a required header is absent. [S1]

In the example, a missing required header is reported and processing stops. Unexpected headers are reported, but the program can continue because the chosen policy only requires name and department. That is a policy choice, not proof that the data is valid. DictReader's mapping does not validate that rows fit your intended schema. [S1]

No. A header check compares column names, not the number of values in every data row. By default, extra row values are collected under the key None, while missing values are filled with None; restkey and restval let you choose different markers. [S1]

The example uses a named key for extras and a unique sentinel for missing values, then prints a row-specific report and skips that row. You could instead collect the row for review or stop processing. Choose one policy deliberately. The documentation does not establish that a row's values are meaningful or correctly typed; those checks depend on your application and data.

Assume the required headers are name and department, and a particular row is short one field. The output would be similar to:

Row 3: extra values=None, missing fields=['department']

If a row has values beyond its headers, the report includes those values in extra values. The number in the sample is a record number: it starts at 2 because the first record is the header. DictReader's parsing also depends on the selected CSV dialect, so adapt the sample if the file uses a different delimiter or quoting convention. [S1]

Before relying on the results, decide whether unexpected headers are allowed, whether irregular rows should be skipped or stop the task, and what validation your application needs after parsing.