Loading data

read_csv and read_excel, and the arguments that save you an hour.

What you'll be able to do

Load CSV and Excel files
Handle separators, encodings and dates
Read a specific Excel sheet

Reading a CSV

import pandas as pd

df = pd.read_csv('sales.csv')

That covers the easy case. The arguments below cover most of the rest.

The arguments worth knowing

Argument	Use
`sep=';'`	European CSVs are often semicolon-separated
`encoding='latin-1'`	When UTF-8 raises a decode error
`parse_dates=['order_date']`	Read dates as dates, not text
`dtype={'zip': str}`	Stop leading zeros being stripped
`usecols=['a','b']`	Read only what you need
`nrows=1000`	Sample a large file first
`thousands=','`	Parse `1,250` as a number
`na_values=['N/A','-']`	Treat placeholders as missing

The zip code problem

Leading zeros disappear

A zip code column of 07030 is read as the integer 7030. So are product codes, account numbers and anything else that looks numeric but is not.

df = pd.read_csv('data.csv', dtype={'zip': str, 'account_id': str})

If a column is an identifier rather than a quantity, read it as a string. You will not be doing arithmetic on it anyway.

Excel

df = pd.read_excel('report.xlsx', sheet_name='Q1')
all_sheets = pd.read_excel('report.xlsx', sheet_name=None)   # dict of DataFrames
df = pd.read_excel('report.xlsx', skiprows=3)                # skip title rows

sheet_name=None returning a dictionary keyed by sheet name is genuinely useful for workbooks with one sheet per month.

Other sources

pd.read_json('data.json')
pd.read_parquet('data.parquet')
pd.read_sql('SELECT * FROM orders', con=engine)
pd.read_clipboard()      # quick paste from Excel

Key points

Read identifier columns as str to preserve leading zeros
parse_dates at load time avoids converting later
sheet_name=None reads every Excel sheet into a dictionary

Check yourself