Issues

No Floor, Ceiling or Truncate, and Round can only round half away from zero (no MidpointRounding / banker's rounding)

ktsu-dev/PreciseNumber#164 · opened

on Oct 7, 2026

1

Exp(x, 10) takes 6.9 s when x has 100,000 significant digits (3 ms once x is rounded to 25): the exp series never rounds its argument, and ExpM1/Exp2/Exp10/Cosh/Tanh inherit it

ktsu-dev/PreciseNumber#163 · opened

on Oct 7, 2026

1

RootN(2, 10000, 10) takes 11 s and RootN(2, 30000, 10) takes over 3 minutes: IntegerRootN's power-of-two seed needs ~0.1n–0.7n Newton steps

ktsu-dev/PreciseNumber#162 · opened

on Oct 7, 2026

1

Atanh is not odd near -1: Atanh(-(1 - 1e-41), 25) throws while Atanh(1 - 1e-41, 25) works, and nearer -1 the digits are wrong

ktsu-dev/PreciseNumber#158 · opened

on Oct 6, 2026

1

Acos near 1 loses its digits to cancellation: Acos(1 - 1e-80, 25) returns 5e-35, correct is 1.414e-40

ktsu-dev/PreciseNumber#157 · opened

on Oct 6, 2026

1

Sin/Cos/Tan near a multiple of π/2 return wrong digits and the wrong sign: Sin(PiTo(60), 25) = -3.75e-46, correct is +4.59e-60

ktsu-dev/PreciseNumber#156 · opened

on Oct 6, 2026

1

Hypot(3e1500000000, 4e1500000000) throws OverflowException instead of returning 5e1500000000, and Hypot(1e100000, 1) takes almost a second

ktsu-dev/PreciseNumber#155 · opened

on Oct 6, 2026

1

Sin, Atan, Tanh, Asinh, LogP1, Log and ExpM1 round the wrong way on midpoint arguments: LogP1(1.5e-15, 1) returns 2e-15 (correct: 1e-15)

ktsu-dev/PreciseNumber#154 · opened

on Oct 6, 2026

1

Parse("(1.5)"), Parse("1.5-") and Parse("$1,234.50", NumberStyles.Currency, en-US) throw FormatException: AllowParentheses, AllowTrailingSign and AllowCurrencySymbol are never honoured

ktsu-dev/PreciseNumber#152 · opened

on Oct 6, 2026

1

5 % 1e2000000000 throws OutOfMemoryException after 6 s instead of returning 5: Mod aligns exponents even when the answer needs no alignment

ktsu-dev/PreciseNumber#151 · opened

on Oct 6, 2026

1

Exp2(7000000000, 3) throws OverflowException while Exp2(7000000000.5, 3) works, and integer Exp2 ignores significantDigits: Exp2(5000000, 10) builds 1.5 million digits in 3.3 s

ktsu-dev/PreciseNumber#150 · opened

on Oct 6, 2026

1

Sin(1e-1000000, 10) takes 9 s and Exp(1e-2000000000, 10) runs out of memory, though the answers are x and 1: tiny series arguments (and Atan of huge ones) cost time proportional to the exponent

ktsu-dev/PreciseNumber#149 · opened

on Oct 6, 2026

1

Log(x, base, digits), Log2, Log2P1 and Log10P1 return more digits than requested when the quotient terminates: Log(1e5, 1e4, 1) returns 1.25

ktsu-dev/PreciseNumber#147 · opened

on Oct 5, 2026

1

The README benchmark chart has been stuck at 2.0.5 since v2.1.0: benchmark-history.yml triggers on `release: published`, which a GITHUB_TOKEN-created release never fires

ktsu-dev/PreciseNumber#146 · opened

on Oct 5, 2026

1

Log, Log2, Log10, Exp, Exp10, Pow, the hyperbolics, Acos and SinPi return about 50 wrong trailing digits when asked for more than ~140 digits, while Sin and Cos refuse the same request

ktsu-dev/PreciseNumber#143 · opened

on Sep 30, 2026

1

To<NFloat>(), NFloat.CreateChecked(x) and PreciseNumber.CreateChecked(nfloat) throw NotSupportedException, although ToPreciseNumber() accepts NFloat

ktsu-dev/PreciseNumber#138 · opened

on Sep 29, 2026

1

x + 0 costs time and memory proportional to x's exponent: 1e300000 + Zero takes ~1 s, and 1e2000000000 + Zero throws OutOfMemoryException

ktsu-dev/PreciseNumber#137 · opened

on Sep 29, 2026

1

1e2000000000 + 1e-2000000000 throws ArgumentOutOfRangeException ("exponent must be non-negative"): exponent alignment wraps int

ktsu-dev/PreciseNumber#135 · opened

on Sep 28, 2026

1

ToString/TryFormat throw FormatException for every format except "G": $"{x:F2}", "N2" and "0.00" all crash

ktsu-dev/PreciseNumber#134 · opened

on Sep 28, 2026

1

ExpM1(-1e10) throws OverflowException instead of returning -1, and ExpM1/LogP1/Sinh/Cosh/Asinh/Acosh take seconds to minutes for large arguments

ktsu-dev/PreciseNumber#133 · opened

on Sep 28, 2026

1

Pow with a fractional exponent misrounds the last digit: Pow(19, 1.5) ends in …417 instead of …416 (double rounding without guard digits)

ktsu-dev/PreciseNumber#127 · opened

on Sep 27, 2026

1

Investigate benchmark regressions in Max, CompareTo, ReduceSignificance, Pow, and Divide since 2.0

ktsu-dev/PreciseNumber#76 · opened

on Sep 15, 2026

5

Navigation Menu

Repository navigation

Issues

No Floor, Ceiling or Truncate, and Round can only round half away from zero (no MidpointRounding / banker's rounding)

Exp(x, 10) takes 6.9 s when x has 100,000 significant digits (3 ms once x is rounded to 25): the exp series never rounds its argument, and ExpM1/Exp2/Exp10/Cosh/Tanh inherit it

RootN(2, 10000, 10) takes 11 s and RootN(2, 30000, 10) takes over 3 minutes: IntegerRootN's power-of-two seed needs ~0.1n–0.7n Newton steps

Atanh is not odd near -1: Atanh(-(1 - 1e-41), 25) throws while Atanh(1 - 1e-41, 25) works, and nearer -1 the digits are wrong

Acos near 1 loses its digits to cancellation: Acos(1 - 1e-80, 25) returns 5e-35, correct is 1.414e-40

Sin/Cos/Tan near a multiple of π/2 return wrong digits and the wrong sign: Sin(PiTo(60), 25) = -3.75e-46, correct is +4.59e-60

Hypot(3e1500000000, 4e1500000000) throws OverflowException instead of returning 5e1500000000, and Hypot(1e100000, 1) takes almost a second

Sin, Atan, Tanh, Asinh, LogP1, Log and ExpM1 round the wrong way on midpoint arguments: LogP1(1.5e-15, 1) returns 2e-15 (correct: 1e-15)

Parse("(1.5)"), Parse("1.5-") and Parse("$1,234.50", NumberStyles.Currency, en-US) throw FormatException: AllowParentheses, AllowTrailingSign and AllowCurrencySymbol are never honoured

5 % 1e2000000000 throws OutOfMemoryException after 6 s instead of returning 5: Mod aligns exponents even when the answer needs no alignment

Exp2(7000000000, 3) throws OverflowException while Exp2(7000000000.5, 3) works, and integer Exp2 ignores significantDigits: Exp2(5000000, 10) builds 1.5 million digits in 3.3 s

Sin(1e-1000000, 10) takes 9 s and Exp(1e-2000000000, 10) runs out of memory, though the answers are x and 1: tiny series arguments (and Atan of huge ones) cost time proportional to the exponent

Log(x, base, digits), Log2, Log2P1 and Log10P1 return more digits than requested when the quotient terminates: Log(1e5, 1e4, 1) returns 1.25

The README benchmark chart has been stuck at 2.0.5 since v2.1.0: benchmark-history.yml triggers on `release: published`, which a GITHUB_TOKEN-created release never fires

Log, Log2, Log10, Exp, Exp10, Pow, the hyperbolics, Acos and SinPi return about 50 wrong trailing digits when asked for more than ~140 digits, while Sin and Cos refuse the same request

To<NFloat>(), NFloat.CreateChecked(x) and PreciseNumber.CreateChecked(nfloat) throw NotSupportedException, although ToPreciseNumber() accepts NFloat

x + 0 costs time and memory proportional to x's exponent: 1e300000 + Zero takes ~1 s, and 1e2000000000 + Zero throws OutOfMemoryException

1e2000000000 + 1e-2000000000 throws ArgumentOutOfRangeException ("exponent must be non-negative"): exponent alignment wraps int

ToString/TryFormat throw FormatException for every format except "G": $"{x:F2}", "N2" and "0.00" all crash

ExpM1(-1e10) throws OverflowException instead of returning -1, and ExpM1/LogP1/Sinh/Cosh/Asinh/Acosh take seconds to minutes for large arguments

Pow with a fractional exponent misrounds the last digit: Pow(19, 1.5) ends in …417 instead of …416 (double rounding without guard digits)

Investigate benchmark regressions in Max, CompareTo, ReduceSignificance, Pow, and Divide since 2.0