math

Improvements

• Safer conversion from float when using custom locales
• Much faster NativeCalculator implementation 🚀

You can expect at least a 3x performance improvement for common arithmetic operations when using the library on systems without GMP or BCMath; it gets exponentially faster on multiplications with a high number of digits. This is due to calculations now being performed on whole blocks of digits (the block size depending on the platform, 32-bit or 64-bit) instead of digit-by-digit as before.

New method

BigNumber::sum() returns the sum of one or more numbers.

Bug fix: of() factory methods could fail when passing a float in environments using a LC_NUMERIC locale with a decimal separator other than '.' (#20).

Thanks @manowark 👍

New method

BigDecimal::sqrt() calculates the square root of a decimal number, to a given scale.

New method

BigInteger::sqrt() calculates the square root of a number (thanks @peter279k).

New exception

NegativeNumberException is thrown when calling sqrt() on a negative number.

Performance update

• Further improvement of toInt() performance
• NativeCalculator can now perform some multiplications more efficiently

Performance optimization of toInt() methods.

Breaking changes

The following deprecated methods have been removed. Use the new method name instead:

New features

BigInteger has been augmented with 5 new methods for bitwise operations:

Thanks to @DASPRiD 👍

New method: BigDecimal::hasNonZeroFractionalPart()

Renamed/deprecated methods:

• BigDecimal::getIntegral() has been renamed to getIntegralPart() and is now deprecated
• BigDecimal::getFraction() has been renamed to getFractionalPart() and is now deprecated

Performance update

BigInteger::parse() and toBase() now use GMP's built-in base conversion features when available.

This is a maintenance release, no code has been changed.

• When installed with --no-dev, the autoloader does not autoload tests anymore
• Tests and other files unnecessary for production are excluded from the dist package

This will help make installations more compact.

Methods renamed:

• BigNumber:sign() has been renamed to getSign()
• BigDecimal::unscaledValue() has been renamed to getUnscaledValue()
• BigDecimal::scale() has been renamed to getScale()
• BigDecimal::integral() has been renamed to getIntegral()
• BigDecimal::fraction() has been renamed to getFraction()
• BigRational::numerator() has been renamed to getNumerator()
• BigRational::denominator() has been renamed to getDenominator()

Classes renamed:

• ArithmeticException has been renamed to MathException

The base class for all exceptions is now MathException.
ArithmeticException has been deprecated, and will be removed in 0.7.0.

A number of methods have been renamed:

• BigNumber:sign() is deprecated; use getSign() instead
• BigDecimal::unscaledValue() is deprecated; use getUnscaledValue() instead
• BigDecimal::scale() is deprecated; use getScale() instead
• BigDecimal::integral() is deprecated; use getIntegral() instead
• BigDecimal::fraction() is deprecated; use getFraction() instead
• BigRational::numerator() is deprecated; use getNumerator() instead
• BigRational::denominator() is deprecated; use getDenominator() instead

The old methods will be removed in version 0.7.0.

• Minimum PHP version is now 7.1; for PHP 5.6 and PHP 7.0 support, use version 0.5
• Deprecated method BigDecimal::withScale() has been removed; use toScale() instead
• Method BigNumber::toInteger() has been renamed to toInt()

BigNumber classes now implement JsonSerializable.
The JSON output is always a string.

Backport of two bug fixes from the 0.5 branch:

• BigInteger::parse() did not always throw InvalidArgumentException as expected
• Dividing by a negative power of 1 with the same scale as the dividend could trigger an incorrect optimization which resulted in a wrong result. See #6.

Backport of two bug fixes from the 0.5 branch:

• BigInteger::parse() did not always throw InvalidArgumentException as expected
• Dividing by a negative power of 1 with the same scale as the dividend could trigger an incorrect optimization which resulted in a wrong result. See #6.

This is a bugfix release. Dividing by a negative power of 1 with the same scale as the dividend could trigger an incorrect optimization which resulted in a wrong result. See #6.

The $scale parameter of BigDecimal::dividedBy() is now optional again.