Loading a CSV File
pd.read_csv loads a CSV file into a DataFrame, usecols and nrows load only the columns and rows you need, sep reads a file whose values are split by semicolons, and a FileNotFoundError means the path doesn't lead from the folder the notebook runs in to the file.
Apprentice · Python · Updated
After this lesson you can
- Load a CSV file into a DataFrame with pd.read_csv, using usecols, nrows and sep when the file needs them
- Fix a FileNotFoundError by finding the folder a notebook runs in and writing the path to the file from there
Key takeaways
- pd.read_csv("sales.csv") reads a CSV file into a DataFrame, the first line becoming the column names.
- usecols= loads only the columns you list, and nrows= only the first rows, which is a quick way to peek at a very large file.
- sep= names the character between values when it isn't a comma. A file with semicolons read without sep=";" comes in as one long column.
- A path such as "sales.csv" is looked for from the folder the notebook runs in. A FileNotFoundError means it isn't there, and Path.cwd() shows which folder that is.
Most data reaches you as a file. A teacher posts a dataset for a class project, a manager exports last month's sales from a spreadsheet, an app lets you download your history: very often the file is a CSV. Loading it into a pandas table is the first line of almost every analysis, and this lesson is that line, with the options and the error you'll meet most.
The example is Sunnyside Farmers Market's sales for the 2026 season, sales.csv: one row for each product a vendor sold on a Saturday, 1,900 rows from April to October. The market and its vendors are our own made-up data.
From a file to a DataFrame
A CSV file ("comma-separated values") is a table saved as plain text: one row per line, with commas between the values, and the column names on the first line. These are the first lines of sales.csv:
date,vendor_id,product,category,units,price
2026-04-04,V01,lettuce,produce,21,3.00
2026-04-04,V01,kale,produce,18,3.50
2026-04-04,V01,spinach,produce,21,4.00
pandas is the Python library for tables, imported as pd. pd.read_csv reads the whole file into a DataFrame, pandas' table:
import pandas as pd
sales = pd.read_csv("sales.csv")
sales
date vendor_id product category units price
0 2026-04-04 V01 lettuce produce 21.0 3.0
1 2026-04-04 V01 kale produce 18.0 3.5
2 2026-04-04 V01 spinach produce 21.0 4.0
3 2026-04-04 V02 honey jar honey 20.0 12.0
4 2026-04-04 V02 honeycomb honey 4.0 15.0
... ... ... ... ... ... ...
1895 2026-10-31 V15 dried herbs produce 13.0 6.0
1896 2026-10-31 V16 knit hat crafts 3.0 25.0
1897 2026-10-31 V16 scarf crafts 2.0 30.0
1898 2026-10-31 V16 tote bag crafts 6.0 15.0
1899 2026-10-31 V16 dish cloth set crafts 9.0 10.0
[1900 rows x 6 columns]
A big table shows its first five and last five rows, with ... for the rest, and its size at the bottom: 1,900 rows and 6 columns. VS Code draws the table as a grid; this lesson shows it as pandas prints it as text. pandas worked out each column's type from the values, so units and price came in as numbers. (units shows 21.0 because a few units are blank in the file, which A First Look: head, info and describe explains.)
Only what you need: usecols and nrows
A file can have far more columns or rows than one question needs. usecols loads only the columns you list, and nrows only the first rows, not counting the header line:
pd.read_csv("sales.csv", usecols=["date", "product", "units"], nrows=5)
date product units
0 2026-04-04 lettuce 21
1 2026-04-04 kale 18
2 2026-04-04 spinach 21
3 2026-04-04 honey jar 20
4 2026-04-04 honeycomb 4
nrows is handy for a peek at a file too big to open comfortably. The names in usecols must match the file exactly: usecols=["Product"] stops with ValueError: Usecols do not match columns, columns expected but not found: ['Product'].
Another separator: sep
Hollow Oak Woodworks, one of the vendors, sends its own record from a spreadsheet that puts a semicolon between values, as many spreadsheets do where the comma is the decimal point:
date;product;units;price
2026-04-04;cutting board;5;35.00
2026-04-04;wooden spoon;9;9.00
2026-04-11;cutting board;3;35.00
2026-04-11;wooden spoon;7;9.00
read_csv looks for commas, finds none, and makes each whole line one value in a single column, date;product;units;price. sep names the separator:
pd.read_csv("hollow_oak.csv", sep=";")
date product units price
0 2026-04-04 cutting board 5 35.0
1 2026-04-04 wooden spoon 9 9.0
2 2026-04-11 cutting board 3 35.0
3 2026-04-11 wooden spoon 7 9.0
Four columns, as it should be. A table with one column whose name is full of semicolons or tabs is the sign a file needs sep (sep="\t" for tabs).
When the file isn't found
Type the name slightly wrong and pandas stops:
pd.read_csv("sale.csv")
FileNotFoundError: [Errno 2] No such file or directory: 'sale.csv'
That's the last line of the error. "sale.csv" is a relative path: pandas looks for it starting from the folder the notebook runs in, its working folder. In VS Code, a notebook runs in the folder it's saved in. When you're sure the name is right, check that folder:
from pathlib import Path
print(Path.cwd())
C:\Users\Sam\python-course\market
Path.cwd() ("current working directory") shows the folder pandas looks in; on a Mac it starts /Users/. If the CSV file sits in a folder inside that one, say data, the path names the folder too, with a forward slash, which works on Windows and on a Mac:
pd.read_csv("data/sales.csv").shape
(1900, 6)
Check yourself
2 questions to try on the lesson page.
So what
pd.read_csv with the right path and options turns a file into a table in one line, and when it can't find the file, Path.cwd() tells you where it looked.