🐍 Tutorial · 6 min read

How to Open a CSV File in Python (csv Module & pandas)

To open a CSV in Python, use the built-in csv module for a plain read, or pandas.read_csv() when you plan to analyse the data. Both handle the parts that break a naive split — quoted fields, commas inside values, and line breaks inside a record. The two things that actually trip people up are the delimiter and the encoding, and each is one argument away.

The quick answer

Standard library, no installs:

import csv

with open("data.csv", newline="", encoding="utf-8") as f:
    for row in csv.reader(f):
        print(row)

Or with pandas, if the file is going to become a table you filter and summarise:

import pandas as pd

df = pd.read_csv("data.csv")
print(df.head())
print(df.shape)   # (rows, columns)

The newline="" in the first example is not decoration. The csv module needs to see the file's line breaks itself, so the standard library docs tell you to open the file with newline="" and let the parser split records.

Why not just split the text yourself?

Because a CSV is not a list of values separated by commas — it is a format described by RFC 4180, where any field may be wrapped in double quotes and may contain a comma or even a line break. Splitting on the comma gets those fields wrong. Here is the same three-record file read two ways.

A naive field split fragments the record wherever a comma sits inside quotes:

$ awk -F, '{print NF" fields: "$1" | "$2}' t.csv
3 fields: name | city
4 fields: "Doe |  Jane"
1 fields: and commas" | 
3 fields: Bob | Tokyo

The csv module returns the three records with the quoted comma kept intact:

import csv
for row in csv.reader(open("t.csv", newline="", encoding="utf-8")):
    print(row)

# ['name', 'city', 'note']
# ['Doe, Jane', 'Paris', 'likes "quotes"\nand commas']
# ['Bob', 'Tokyo', 'plain']

That output is from Python 3.13.5 on a file whose second field is "Doe, Jane" and whose third spans two physical lines. The naive split produced four fragments; the real parser produced three rows. This is the single best reason to use a CSV parser instead of split(",") in any language.

Open a CSV as dictionaries with csv.DictReader

If your file has a header row and you would rather address columns by name than by index, use csv.DictReader. It turns each record into a dictionary keyed by the header, which makes the rest of the code readable:

import csv

with open("data.csv", newline="", encoding="utf-8") as f:
    for row in csv.DictReader(f):
        print(row["name"], "->", row["city"])

Read only the columns you need with pandas

On a wide CSV, loading every column wastes time and memory. usecols restricts the read, and dtype stops pandas from guessing a type you did not want:

df = pd.read_csv("data.csv", usecols=["name", "city", "signup_date"])

Type guessing is worth watching: pandas will turn a column of ZIP codes or part numbers into integers and drop the leading zeros, and it will read a numeric-looking ID as a number unless you tell it otherwise. Pass dtype={"zip": str} for those columns. The same class of problem in Excel is covered in keep leading zeros in CSV.

Read a file that is too big to fit in memory

Do not hand a multi-gigabyte CSV to a single read_csv() call. Two habits keep memory flat:

• csv module: iterate csv.reader. It is a generator over the file, so only the current row is held at a time.
• pandas: pass chunksize and loop over the iterator:

for chunk in pd.read_csv("huge.csv", chunksize=100_000):
    process(chunk)   # each chunk is a DataFrame of 100k rows

If the job is really about size rather than code, the file may be beyond what a spreadsheet can even open — Excel stops at 1,048,576 rows and 16,384 columns per sheet, per Microsoft's documented limits. See a CSV too big for Excel and why a CSV is too large to analyse as-is.

The two gotchas: delimiter and encoding

Encoding. open() uses your platform's default encoding, which is why a UTF-8 file read on a Windows machine using cp1252 can raise UnicodeDecodeError or produce symbols. Be explicit: pass encoding="utf-8", and utf-8-sig if the file begins with a byte-order mark. If the file came from an older Windows export, cp1252 is the other common value. pandas takes the same argument in read_csv(), and mis-decoded text looks identical to the garbled-file symptom described in fix garbled CSV.

Delimiter. European exports often use semicolons, and TSV files use tabs. Pass delimiter=";" or delimiter="\t" to csv.reader, or sep=";" to read_csv(). You can also let csv.Sniffer guess, but a guess is still a guess — when the columns land wrong, set it by hand. The wider problem is covered in change a CSV delimiter.

If you would rather not write code

Reading a CSV in Python is the right tool when the file feeds a script. For a one-off look — confirm the columns, count the rows, check a few values — a browser-based viewer is faster and needs nothing installed. Drop the file into the free CSV analyzer and it renders as a table on your device, with no upload. To turn the same data into JSON instead, use the JSON ⇄ CSV converter. To run SQL against a CSV without a database, see query a CSV with SQL.

Frequently Asked Questions

How do I open a CSV file in Python?

Use the built-in csv module for a plain read: import csv, then open the file with newline="" and pass it to csv.reader or csv.DictReader. If you plan to analyse the data, pandas.read_csv("file.csv") loads it into a DataFrame in one line. Both parse quoted commas and embedded line breaks correctly.

Should I use the csv module or pandas?

Use csv when you only need to stream rows and want no dependencies — it ships with Python. Use pandas when you need filtering, grouping or statistics, because it gives you a DataFrame; it is a third-party library you install separately.

Why does my CSV look garbled or raise UnicodeDecodeError?

It is an encoding mismatch. open() defaults to the system encoding, so a UTF-8 file read as cp1252 fails, and a cp1252 file read as UTF-8 shows mangled characters. Pass encoding="utf-8" explicitly, and use "utf-8-sig" when the file starts with a byte-order mark.

How do I read a CSV that is too big for memory?

Do not load it whole. With csv, iterate csv.reader so only one row is held at a time. With pandas, pass chunksize to read_csv and loop over the chunks, or use usecols to read only the columns you need.