github.com/shopspring/decimal

BEHAVIOR CHANGES

• NewFromString and UnmarshalBinary, and so JSON, text, SQL, gob and Spanner decoding, reject exponents larger than the new MaxDecodeExponent setting (default 10,000). Such input could hang the process or use gigabytes of memory. Raise MaxDecodeExponent if you need larger exponents #440
• Pow, PowWithPrecision, PowInt32 and PowBigInt round correctly. Results for fractional exponents have PowPrecisionNegativeExponent (16) decimal places, and negative powers that used to return 0 keep 16 significant digits, so 10^-18 is 0.000000000000000001. Results too large to represent return an error, or 0 from Pow, instead of hanging or panicking #439
• RoundUp, RoundDown, RoundCeil and RoundFloor always return exponent -places, like Round. Values don't change, but Exponent(), Coefficient() and String() with TrimTrailingZeros = false can #426 #441
• Truncate supports negative precision: 5432 truncated to -2 is 5400. Before, it returned 5432 unchanged #426
• Shift panics when the exponent overflows an int32, instead of wrapping around #437
• NewFromString rejects a sign right after a leading decimal point, such as ".-5" #447

FEATURES

• Add TrimTrailingZeros and UseScientificNotation settings for string output #389 #416
• Add Spanner support with EncodeSpanner and DecodeSpanner for Decimal and NullDecimal #395

ENHANCEMENTS

• Most operations are faster and allocate less, about twice as fast on average across the benchmarks #444 #445 #446 #447 #448 #449 #451
• Faster UnmarshalJSON, Scan and NewFromString #403
• ExpHullAbrham raises its partial sum to a power by squaring. Large inputs that took minutes now take seconds #450
• Document that NewFromFloat returns the shortest decimal that converts back to the same float, and how to convert large whole numbers exactly #442
• Split decimal.go into decimal.go, math.go and encoding.go #452
• Lint with golangci-lint in CI, and test on Go 1.26 and 1.27 #453 #454

BUGFIXES

• Fix Float64 hanging on decimals with huge exponents #438
• Fix ExpHullAbrham returning 10^exp instead of 1 for tiny inputs #433
• Fix a data race in ExpTaylor #414
• Fix NumDigits for some exact powers of ten #425
• Fix the omitzero struct tag for Decimal fields in nested structs #415
• Fix RoundCash giving wrong results with a low DivisionPrecision #443
• NullDecimal is no longer Valid after a failed Scan, UnmarshalJSON or DecodeSpanner #443