Python decimal localcontext: keep a rounding rule inside one calculation

Observed contract

A receipt amount with three decimal places is quantized to cents under ROUND_HALF_UP. The local context records that information was discarded. After the with block, the calling context still has its original rounding policy.

Boundary

Rounding mode and where rounding occurs are business rules. This fixture does not claim ROUND_HALF_UP is appropriate for every currency or jurisdiction. Construct Decimal from text when exact decimal input matters; binary floats can carry prior approximation.

Executable case

from decimal import Decimal, Inexact, ROUND_HALF_UP, getcontext, localcontext

prior_rounding = getcontext().rounding
with localcontext(rounding=ROUND_HALF_UP) as receipt_policy:
    rounded = Decimal("47.245").quantize(Decimal("0.01"))
    was_inexact = receipt_policy.flags[Inexact]

print("rounded", rounded)
print("inexact", was_inexact)
print("rounding_restored", getcontext().rounding == prior_rounding)

Output

rounded 47.25
inexact True
rounding_restored True

Cost

Decimal arithmetic has more overhead than integer-cent arithmetic. Context flags are sticky within a context, so clear or use a fresh local context when each operation needs an independent signal.

Common Mistakes

• Do not change a shared arithmetic policy for one receipt and forget to restore it.
• Inexact is a status flag unless its trap is enabled.
• A Decimal made from a float can preserve the float's approximation.

Connected lessons

• Python Decimal money: parse decimal text and choose rounding explicitly
• Python Decimal Inexact trap: reject fractional minor units
• Python amount parsing: define a wire format before Decimal

Test this contract.