The Quiet Terminal

The csv module is still the right tool for small table jobs

2026-03-20 · 3 min read

Earlier this month I wanted to know what my coffee habit actually costs, and the answer lived in a 60-row CSV I keep in my notes folder. I very nearly installed pandas for this. The csv module — which ships with Python — did the whole job in a dozen lines, on a laptop with no dependencies and no internet if it came to that.

The data

date,coffee,size,price
2026-03-02,latte,large,4.80
2026-03-03,filter,small,2.60
2026-03-05,latte,large,4.80
2026-03-06,flat white,small,3.40

DictReader: rows as dictionaries

import csv

with open("coffee_log.csv", newline="", encoding="utf-8") as f:
    for row in csv.DictReader(f):
        print(row["date"], row["coffee"], row["price"])

DictReader reads the header row and uses it as the keys, so every row is a dict — no counting columns, no row[2] that silently breaks when someone reorders the file. (There's a plain csv.reader that hands you lists instead; I stopped using it the day a spreadsheet gained a column in the middle.) Two things to know: every value is a string, so convert with float() before doing math; and that mysterious newline="" is doing something important, which we'll get to.

One more thing the module handles so I never have to: quoting. flat white needed no quotes above because it has no comma, but a field like latte, oat milk gets quoted by DictWriter on the way out and un-quoted by DictReader on the way back. The writer and reader agree with each other, which is most of what you need from a file format.

Filter and sum: the whole reason I opened the file

import csv

total = 0.0
count = 0
with open("coffee_log.csv", newline="", encoding="utf-8") as f:
    for row in csv.DictReader(f):
        if row["coffee"] == "latte" and row["size"] == "large":
            total += float(row["price"])
            count += 1

print(f"{count} large lattes, {total:.2f} in total")

2 large lattes, 9.60 in total

Twelve lines including the blank ones, no imports beyond the standard library, and the answer was worse than I expected.

DictWriter and the `newline=""` gotcha

Writing rows back is symmetric:

import csv

row = {"date": "2026-03-20", "coffee": "espresso", "size": "small", "price": "2.20"}

with open("coffee_log.csv", "a", newline="", encoding="utf-8") as f:
    writer = csv.DictWriter(f, fieldnames=["date", "coffee", "size", "price"])
    writer.writerow(row)

Now the gotcha. The csv module writes its own \r\n line endings. If you open the file in text mode without newline="", Python also translates \n to \r\n on the way out, so every row ends with \r\r\n. The symptom: Excel shows a blank row between every line of data, on Windows only, and the file still "works", so the bug can hide for months. newline="" tells Python to keep its hands off the line endings because the csv module knows what it's doing. One argument, once, forever. (For a brand-new file, call writer.writeheader() first so the header row lands before the data.)

The opinion