Repository navigation
Pow(-1, n) throws OverflowException for any integer |n| ≥ 2^31, although the result is ±1 #128
Description
What's wrong
The integer-power path in Pow (PreciseNumber/PreciseNumber.cs:1818) loops over power.Abs().To<int>(), so an exponent that doesn't fit in an int throws. IsUnit already short-circuits a base of +1, but a base of −1 gets no such shortcut, even though (−1)^n is always ±1.
Repro
Parse("-1").Pow(Parse("2147483648")); // OverflowException, expected 1
Parse("-1").Pow(Parse("-2147483648")); // OverflowException, expected 1
Parse("-1").Pow(Parse("1e20")); // OverflowException, expected 1
Parse("-1").Pow(Parse("2147483647")); // -1 (correct)
Parse("1").Pow(Parse("1e20")); // 1 (correct, short-circuited)For comparison, double.Pow(-1, 1e20) returns 1.
Suggested fix / acceptance criteria
- Before the loop, when the base is −1, return
IsEvenInteger(power) ? One : NegativeOne. - For any other base, an exponent that large really does overflow the result's exponent, so the current exception is fine there.
- Add tests for the three failing calls above.
Activity
matt-edmondson commentedon Sep 27, 2026
Triage
OverflowExceptionrather than returning a wrong value.PreciseNumber.PowPow.IsUnitcheck. A fix for Pow with a fractional exponent misrounds the last digit: Pow(19, 1.5) ends in …417 instead of …416 (double rounding without guard digits) #127 (also inPow) could carry it, as long as both keep their own tests.Generated by Claude Code