Sum Sume usd_micros as integers in Python, not floats

Sume reports money as integer usd_micros, millionths of a dollar. Add the integers and convert once for display, instead of converting each row to a float dollar amount and summing.

Integer addition is exact, so a total never differs from the sum of its rows.

Names to know

Fields from the Runs and results page (read 2026-10-03).

Why floats drift

Binary floats cannot hold most decimal cents exactly, so a long sum of dollar floats can show a total like 0.30000000000000004. That makes a reconciliation fail by a fraction of a cent. Integers do not.

from decimal import Decimal

rows = [10_000, 20_000, 10_000]
print(0.1 + 0.2)
total = sum(rows)
print(total, Decimal(total) / Decimal(1_000_000))
print(f"${total // 10_000 / 100:.2f}")

Rounding at the edge

Round only for display and invoices, and decide your rule once, for example round half up to the cent. Null usage is unknown, not zero, so keep it out of the sum as in the null helper, and use the per-row cost walk for a batch.