Level 2 · How big is the table?

Rows and columns

Before any analysis, check the size of what you’ve got:

order_items.shape        # (rows, columns)
order_items.shape[0]     # number of rows
order_items.shape[1]     # number of columns
len(order_items)         # number of rows, the most common way to write it

shape is an attribute, not a method, so it has no brackets. It’s a tuple, which you index like a list.

Row counts are your first sanity check. If the business says “about 30,000 order lines” and you see 3,000 or 300,000, stop and find out why before going further. Maybe the file was truncated, maybe you loaded the wrong table, maybe the business number is out of date. Any of those is worth knowing.

Column names

order_items.columns
# Index(['order_item_id', 'order_id', 'product_id', 'quantity',
#        'unit_price', 'discount_pct'], dtype='object')

list(order_items.columns)   # as a plain list

Column names are case-sensitive and must match exactly, so it’s worth looking at them before typing them in code.

Types with dtypes and info

Every column has a dtype (data type). The ones you’ll see most:

dtype	holds
int64	whole numbers
float64	decimals, and whole numbers in a column with missing values
bool	True / False
object	text (and anything mixed)
datetime64[ns]	dates and times, once converted

order_items.dtypes

info() puts everything in one view: number of rows, each column’s non-null count and dtype, and memory use.

customers.info()

RangeIndex: 2018 entries, 0 to 2017
Data columns (total 9 columns):
 #   Column           Non-Null Count  Dtype
---  ------           --------------  -----
 0   customer_id      2018 non-null   int64
 ...
 4   city             2004 non-null   object
 ...
 8   referral_source  1352 non-null   object

Two discoveries in one glance: some customers have no city, and about a third have no referral source. info() prints its report rather than returning a value, so use it to look, not to store.

Types tell you about the data

Types often reveal problems before you see them in rows. In web_sessions, customer_id is float64 rather than int64. Why would an ID be a decimal? Because some sessions are anonymous, and a classic NumPy integer column can’t hold missing values, so pandas stores the column as floats with NaN. Similarly, order_date shows datetime64[ns] only because the tables here were loaded with pd.read_csv(..., parse_dates=[...]). CSV files don’t carry types, so a plain read_csv would leave dates as object text. Spotting these early saves you from confusing errors later.

Other quick checks

order_items["order_id"].nunique()   # distinct orders that have lines
order_items.isna().sum()            # missing values per column

Common mistakes

• Writing order_items.shape() with brackets: TypeError: 'tuple' object is not callable.
• Storing order_items.info() in a variable. It prints and returns None.
• Using order_items.size for the row count. size is rows times columns.