Repository navigation
ReduceSignificance(0) returns 10 and ReduceSignificance(-1) returns 0 instead of rejecting the argument #126
Description
What's wrong
The public ReduceSignificance(int significantDigits) (PreciseNumber/PreciseNumber.cs:860-876) doesn't validate its argument. Every other precision parameter in the library throws ArgumentOutOfRangeException for values below 1: Divide, RootN, Sqrt, Sin, and ConstantTo for PiTo and friends. ReduceSignificance instead silently returns a meaningless value:
((PreciseNumber)5).ReduceSignificance(0) → 10
((PreciseNumber)5).ReduceSignificance(-1) → 0
The method also computes Exponent + significantDifference without checked, whereas the rest of the file guards exponent arithmetic with checked (see #120's exponent-overflow fix).
Why it matters
When a caller computes the precision, for example as digits - lost and it reaches 0 or goes negative, they get a wrong number (5 → 10) and no error. That is the worst outcome for a library whose purpose is precision.
Suggested fix
- Throw
ArgumentOutOfRangeExceptionwhensignificantDigits < 1, the same way the other precision parameters do. - Make the exponent addition
checked. - Add tests for 0 and for a negative value.
Activity
matt-edmondson commentedon Sep 27, 2026
Triage
ReduceSignificanceinPreciseNumber/PreciseNumber.cs. Owner: @matt-edmondsoncheckedexponent pattern from Throw OverflowException when Multiply or Divide leaves the int exponent range #120. No open PR covers it.ArgumentOutOfRangeExceptionforsignificantDigits < 1, make the exponent additionchecked, and add tests for 0 and for a negative value.Generated by Claude Code