💥 Breaking changes

• Deprecated method BigDecimal::hasNonZeroFractionalPart() has been removed, use ! $number->getFractionalPart()->isZero() instead
• Exception constructors and factory methods are now @internal

✨ Compatibility improvements

• BigDecimal::fromFloatExact() now supports 32-bit PHP

✨ New features

New methods to create a BigDecimal from a float value:

• BigDecimal::fromFloatExact()
• BigDecimal::fromFloatShortest()

👌 Improvements

• Add @return non-empty-string to toString(), jsonSerialize() and __toString() (#111 by @vudaltsov)

💥 Breaking changes

• BigInteger::getLowestSetBit() now returns null instead of -1 when the number is zero
• Deprecated method BigRational::simplified() has been removed, as it is now a no-op

✨ New features

• New method: BigDecimal::getIntegralPart() returns the integral part as BigInteger (this method existed with a different signature in version 0.14, and was removed in 0.15)
• New method: BigDecimal::getFractionalPart() returns the fractional part as BigDecimal (this method existed with a different signature and meaning in version 0.14, and was removed in 0.15)

🗑️ Deprecations

• Method BigDecimal::hasNonZeroFractionalPart() is deprecated, use ->getFractionalPart()->isZero() instead

💥 Breaking changes

• floating-point inputs are no longer accepted by of() and arithmetic methods, use of((string) $float) to get the same behaviour as before (#105)
• BigRational is now always simplified to lowest terms: all operations, including of() and ofFraction(), now return a fraction in its simplest form (e.g. 2/3 instead of 4/6)
• BigDecimal::dividedBy() now requires the $scale parameter
• BigInteger::sqrt() and BigDecimal::sqrt() now default to RoundingMode::Unnecessary, explicitly pass RoundingMode::Down to get the previous behaviour
• BigInteger::mod() now uses Euclidean modulo semantics: the modulus must be strictly positive, and the result is always non-negative; this change aligns with Java's BigInteger.mod() behaviour
• BigInteger::mod(), modInverse() and modPow() now throw InvalidArgumentException (instead of NegativeNumberException) for negative modulus/exponent arguments
• MathException is now an interface instead of a class
• BigDecimal::getPrecision() now returns 1 for zero values
• BigNumber::min(), max() and sum() now throw an ArgumentCountError when called with no arguments (previously threw InvalidArgumentException)
• BigInteger::randomBits() and randomRange() now throw RandomSourceException when random byte generation fails or returns invalid data

Deprecated API elements removed:

• deprecated method BigInteger::testBit() has been removed, use isBitSet() instead
• deprecated method BigInteger::gcdMultiple() has been removed, use gcdAll() instead
• deprecated method BigDecimal::exactlyDividedBy() has been removed, use dividedByExact() instead
• deprecated method BigDecimal::getIntegralPart() has been removed (will be re-introduced as returning BigInteger in 0.16)
• deprecated method BigDecimal::getFractionalPart() has been removed (will be re-introduced as returning BigDecimal with a different meaning in 0.16)
• deprecated method BigDecimal::stripTrailingZeros() has been removed, use strippedOfTrailingZeros() instead
• deprecated method BigRational::nd() has been removed, use ofFraction() instead
• deprecated method BigRational::quotient() has been removed, use getIntegralPart() instead
• deprecated method BigRational::remainder() has been removed, use $number->getNumerator()->remainder($number->getDenominator()) instead
• deprecated method BigRational::quotientAndRemainder() has been removed, use $number->getNumerator()->quotientAndRemainder($number->getDenominator()) instead
• deprecated RoundingMode upper snake case constants (e.g. HALF_UP) have been removed, use the pascal case version (e.g. HalfUp) instead

The following breaking changes only affect you if you're using named arguments:

• BigInteger::mod() now uses $modulus as the parameter name
• BigInteger::modInverse() now uses $modulus as the parameter name
• BigInteger::modPow() now uses $exponent and $modulus as parameter names
• BigInteger::shiftedLeft() now uses $bits as the parameter name
• BigInteger::shiftedRight() now uses $bits as the parameter name
• BigInteger::isBitSet() now uses $bitIndex as the parameter name
• BigInteger::randomBits() now uses $bitCount as the parameter name
• BigDecimal::withPointMovedLeft() now uses $places as the parameter name
• BigDecimal::withPointMovedRight() now uses $places as the parameter name

The following breaking changes are unlikely to affect you:

• DivisionByZeroException::modulusMustNotBeZero() has been renamed to zeroModulus()
• DivisionByZeroException::denominatorMustNotBeZero() has been renamed to zeroDenominator()
• IntegerOverflowException::toIntOverflow() has been renamed to integerOutOfRange()
• RoundingNecessaryException::roundingNecessary() has been removed

🗑️ Deprecations

• Method BigRational::simplified() is deprecated, as it is now a no-op

✨ New features

• BigInteger::power(), BigDecimal::power() and BigRational::power() no longer enforce an exponent limit
• BigInteger::shiftedLeft() and BigInteger::shiftedRight() no longer enforce a limit on the number of bits
• BigRational::power() now accepts negative exponents
• New exception: InvalidArgumentException for invalid argument errors
• New exception: NoInverseException for modular inverse errors
• New exception: RandomSourceException for random source errors

👌 Improvements

• Narrowed parameter and return types with static analysis annotations (#108 by @simPod)

🗑️ Deprecations

• Method BigInteger::testBit() is deprecated, use isBitSet() instead

✨ New features

• New method: BigInteger::isBitSet() (replaces testBit())
• New method: BigNumber::toString() (alias of magic method __toString())

👌 Improvements

• Performance optimization of BigRational comparison methods
• More exceptions have been documented with @throws annotations

✨ New features

• clamp() is now available on the base BigNumber class

👌 Improvements

• Improved @throws exception documentation

🗑️ Deprecations

• Not passing a $scale to BigDecimal::dividedBy() is deprecated; $scale will be required in 0.15

👌 Improvements

• BigRational::toFloat() never returns NAN anymore

🗑️ Deprecations

• Not passing a rounding mode to BigInteger::sqrt() and BigDecimal::sqrt() triggers a deprecation notice: the default rounding mode will change from Down to Unnecessary in 0.15

✨ New features

• BigInteger::sqrt() and BigDecimal::sqrt() now support rounding
• abs() and negated() methods are now available on the base BigNumber class

👌 Improvements

• Alphabet is now checked for duplicate characters in BigInteger::(from|to)ArbitraryBase()
• BigNumber::ofNullable() is now marked as @pure

🗑️ Deprecations

• Passing a negative modulus to BigInteger::mod() is deprecated to align with Euclidean modulo semantics; it will throw InvalidArgumentException in 0.15
• Method BigDecimal::stripTrailingZeros() is deprecated, use strippedOfTrailingZeros() instead

✨ New features

• BigInteger::modPow() now accepts negative bases
• New method: BigDecimal::strippedOfTrailingZeros() (replaces stripTrailingZeros())

👌 Improvements

• clamp() methods are now marked as @pure