Math.sumPrecise

concepts

1. Float Drift Lab

   Side-by-side comparison of naive loop summation vs Math.sumPrecise. Add large sets of IEEE 754 floats and watch the error accumulate in the naive column while the precise column stays exact.

2. Sum REPL

   Enter any comma-separated numbers and instantly see the naive sum, Math.sumPrecise result, and the exact difference. Great for verifying where floating-point math breaks down.

3. Kahan Summation Visualizer

   Step through Kahan compensated summation — the algorithm behind Math.sumPrecise() — side-by-side with naïve addition. See the compensation term accumulate at each step and compare error magnitude across five different test datasets.

4. Financial Ledger

   Add transactions to a ledger and compare the running balance computed by naïve reduce() vs Math.sumPrecise(). Load the stress test — 40 micro-transactions plus the classic 0.1+0.2 entry — to see where real financial drift appears.

5. Compatibility Lab

   Detects Math.sumPrecise availability and runs a battery of floating-point precision test cases — [0.1, 0.2], [1e15, 0.1, -1e15], [0.1 × 10] — comparing naive reduce() against Math.sumPrecise. Provides the Kahan compensated summation fallback for non-supporting environments.

why it shipped

IEEE 754 floating-point arithmetic means that adding many values in a loop accumulates rounding errors — 0.1 + 0.2 famously yields 0.30000000000000004, not 0.3. Financial calculations, statistical aggregations, and physics simulations all suffer from this. Math.sumPrecise implements the Neumaier compensated-summation algorithm, which tracks the accumulated error and corrects for it. The TC39 proposal standardises the method so every JavaScript engine can provide a reliable, cross-platform precise sum without the developer needing to write or import a library.

references

• ChromeStatus entry
• TC39 proposal — Math.sumPrecise
• Chrome 147 release notes