Money and other exact quantities: use Decimal, not float

What it is

A float is a binary fraction; 0.1 is stored as the nearest representable value, and 0.1 + 0.2 == 0.3 is false, as the Python tutorial shows. decimal.Decimal stores decimal digits exactly, supports user-defined precision and rounding modes, and quantize rounds to a fixed number of places.

Why it matters

Amounts of money, tax rates and quantities on invoices must add up to the cent. Accumulated float error shows up as one-cent differences that fail reconciliation and audits.

How to apply

• Construct decimals from strings or integers (Decimal("19.90")), never from floats, which carry the binary error in.
• Round explicitly at defined points (quantize(Decimal("0.01"), rounding=ROUND_HALF_EVEN) or the mode the domain prescribes).
• Alternatively store integers in minor units (cents) and convert only for display; divisions still need an explicit rounding rule.
• Serialise as strings in JSON to avoid float conversion in other systems.

Pitfalls

Mixing Decimal and float raises or silently converts depending on the operation. Database columns must be NUMERIC, not floating types. Percent calculations on already-rounded amounts differ from calculations on exact ones; define the order of rounding in the specification.