Brick\Math

A PHP library to work with arbitrary precision numbers.
Installation
This library is installable via Composer:
composer require brick/math
Requirements
This library requires PHP 8.2 or later.
For PHP 8.1 compatibility, you can use version 0.13. For PHP 8.0, you can use version 0.11. For PHP 7.4, you can use version 0.10. For PHP 7.1, 7.2 & 7.3, you can use version 0.9. Note that PHP versions < 8.1 are EOL and not supported anymore. If you're still using one of these PHP versions, you should consider upgrading as soon as possible.
Although the library can work seamlessly on any PHP installation, it is highly recommended that you install the GMP or BCMath extension to speed up calculations. The fastest available calculator implementation will be automatically selected at runtime.
Project status & release process
While this library is still under development, it is well tested and considered stable enough to use in production environments.
The current releases are numbered 0.x.y. When a non-breaking change is introduced (adding new methods, optimizing existing code, etc.), y is incremented.
When a breaking change is introduced, a new 0.x version cycle is always started.
It is therefore safe to lock your project to a given release cycle, such as ^0.20.
If you need to upgrade to a newer release cycle, check the release history for a list of changes introduced by each further 0.x.0 version.
Package contents
This library provides the following public classes in the Brick\Math namespace:
- BigNumber: base class for
BigInteger,BigDecimalandBigRational - BigInteger: represents an arbitrary-precision integer number.
- BigDecimal: represents an arbitrary-precision decimal number.
- BigRational: represents an arbitrary-precision rational number (fraction), always reduced to lowest terms.
- RoundingMode: enum representing all available rounding modes.
And exceptions in the Brick\Math\Exception namespace.
Overview
Instantiation
The constructors of the classes are not public, you must use a factory method to obtain an instance.
All classes provide an of() factory method that accepts any of the following types:
BigNumberinstancesintnumbersstringrepresentations of integer, decimal and rational numbers
Example:
BigInteger::of(123546);
BigInteger::of('9999999999999999999999999999999999999999999');
BigDecimal::of('9.99999999999999999999999999999999999999999999');
BigRational::of('2/3');
Note that all of() methods accept all the representations above, as long as it can be safely converted to the current type:
BigInteger::of('1.00'); // 1
BigInteger::of('1.01'); // RoundingNecessaryException
BigDecimal::of('1/8'); // 0.125
BigDecimal::of('1/3'); // RoundingNecessaryException
BigRational::of('1.1'); // 11/10
BigRational::of('1.15'); // 23/20 (reduced to lowest terms)
> [!NOTE] > The of() factory method does not accept float values, because casting a float to string can be lossy. > To convert a float to a BigDecimal, use one of the dedicated methods: > > php > // Exact IEEE-754 representation — the value the float actually holds: > BigDecimal::fromFloatExact(0.1); // 0.1000000000000000055511151231257827021181583404541015625 > > // Shortest decimal that round-trips back to the same float: > BigDecimal::fromFloatShortest(0.1); // 0.1 >
Parsing untrusted input
of() places no hard limits on its input: a string with millions of digits is accepted as is, and a number in exponential notation is expanded to its full length, so a string as short as 1e1000000000 yields a number with a billion digits.
If your input comes from an untrusted source, such as an HTTP request, use parse() instead, which requires you to specify the allowed syntax and a maximum number of digits:
use Brick\Math\NumberSyntax;
BigDecimal::parse($input, allowedSyntax: NumberSyntax::DECIMAL, maxDigits: 20);
The $allowedSyntax parameter restricts the accepted notations. Plain integers such as 123 are always accepted, and each NumberSyntax case (DecimalPoint, Exponent, Fraction) allows one additional feature. The enum also provides constants for the most common combinations:
NumberSyntax::INTEGER— integers only:123NumberSyntax::DECIMAL— integers and decimal numbers:123,123.45; typical for monetary inputNumberSyntax::SCIENTIFIC— integers, decimal numbers and exponents:123,123.45,1.5e-3; accepts every JSON numberNumberSyntax::RATIONAL— integers and fractions:123,22/7NumberSyntax::ALL— the full syntax accepted byof():123,123.45,1.5e-3,22/7
The $maxDigits parameter limits the number of digits, counted both as written in the input and in the resulting number, so that a value such as 1e1000000000 is rejected before it is ever expanded:
BigDecimal::parse('123.45', allowedSyntax: NumberSyntax::DECIMAL, maxDigits: 20); // 123.45
BigDecimal::parse('1.2e3', allowedSyntax: NumberSyntax::DECIMAL, maxDigits: 20); // NumberFormatException (exponent not allowed)
BigDecimal::parse('1e1000000000', allowedSyntax: NumberSyntax::SCIENTIFIC, maxDigits: 20); // NumberFormatException (too many digits)
Immutability & chaining
The BigInteger, BigDecimal and BigRational classes are immutable: their value never changes, so that they can be safely passed around. All methods that return a BigInteger, BigDecimal or BigRational return a new object, leaving the original object unaffected:
$ten = BigInteger::of(10);
echo $ten->plus(5); // 15
echo $ten->multipliedBy(3); // 30
The methods can be chained for better readability:
echo BigInteger::of(10)->plus(5)->multipliedBy(3); // 45
Parameter types
All methods that accept a number: plus(), minus(), multipliedBy(), etc. accept the same types as of(). For example, given the following number:
$integer = BigInteger::of(123);
The following lines are equivalent:
$integer->multipliedBy(123);
$integer->multipliedBy('123');
$integer->multipliedBy($integer);
Just like of(), other types of BigNumber are acceptable, as long as they can be safely converted to the current type:
echo BigInteger::of(2)->multipliedBy(BigDecimal::of('2.0')); // 4
echo BigInteger::of(2)->multipliedBy(BigDecimal::of('2.5')); // RoundingNecessaryException
echo BigDecimal::of(2.5)->multipliedBy(BigInteger::of(2)); // 5.0
These parameters are converted with of(), so the same rules apply: for untrusted strings, use parse() first, and pass the resulting number to the method.
Division & rounding
BigInteger
By default, dividing a BigInteger returns the exact result of the division, or throws an exception if the remainder of the division is not zero:
echo BigInteger::of(999)->dividedBy(3); // 333
echo BigInteger::of(1000)->dividedBy(3); // RoundingNecessaryException
You can pass an optional rounding mode to round the result, if necessary:
echo BigInteger::of(1000)->dividedBy(3, RoundingMode::Down); // 333
echo BigInteger::of(1000)->dividedBy(3, RoundingMode::Up); // 334
If you're into quotients and remainders, there are methods for this, too:
echo BigInteger::of(1000)->quotient(3); // 333
echo BigInteger::of(1000)->remainder(3); // 1
You can even get both at the same time:
[$quotient, $remainder] = BigInteger::of(1000)->quotientAndRemainder(3);
BigDecimal
Dividing a BigDecimal always requires a scale to be specified. If the exact result of the division does not fit in the given scale, a rounding mode must be provided.
echo BigDecimal::of(1)->dividedBy('8', 3); // 0.125
echo BigDecimal::of(1)->dividedBy('8', 2); // RoundingNecessaryException
echo BigDecimal::of(1)->dividedBy('8', 2, RoundingMode::HalfDown); // 0.12
echo BigDecimal::of(1)->dividedBy('8', 2, RoundingMode::HalfUp); // 0.13
If you know that the division yields a finite number of decimals places, you can use dividedByExact(), which will automatically compute the required scale to fit the result, or throw an exception if the division yields an infinite repeating decimal:
echo BigDecimal::of(1)->dividedByExact(256); // 0.00390625
echo BigDecimal::of(1)->dividedByExact(11); // RoundingNecessaryException
BigRational
The result of the division of a BigRational can always be represented exactly:
echo BigRational::of('13/99')->dividedBy('7'); // 13/693
echo BigRational::of('13/99')->dividedBy('9/8'); // 104/891
Bitwise operations
BigInteger supports bitwise operations:
and()or()xor()not()
and bit shifting:
shiftedLeft()shiftedRight()
Exceptions
All exceptions thrown by this library implement the MathException interface. This means that you can safely catch all exceptions thrown by this library using a single catch clause:
use Brick\Math\BigDecimal;
use Brick\Math\Exception\MathException;
try {
$number = BigInteger::of(1)->dividedBy(3);
} catch (MathException $e) {
// ...
}
If you need more granular control over the exceptions thrown, you can catch the specific exception classes:
DivisionByZeroExceptionIntegerOverflowExceptionInvalidArgumentExceptionNegativeNumberExceptionNoInverseExceptionNumberFormatExceptionPlatformExceptionRandomSourceExceptionRoundingNecessaryException
Serialization
BigInteger, BigDecimal and BigRational can be safely serialized on a machine and unserialized on another, even if these machines do not share the same set of PHP extensions.
For example, serializing on a machine with GMP support and unserializing on a machine that does not have this extension installed will still work as expected.
PHPStan extension
A third-party PHPStan extension is available for this library. It provides more specific throw type narrowing for brick/math methods, so that PHPStan can infer the exact exception classes thrown. Note that this extension is not maintained by the author of brick/math.
Source
| 1 | ## Brick\Math |
| 2 | |
| 3 | <img src="https://raw.githubusercontent.com/brick/brick/master/logo.png" alt="" align="left" height="64"> |
| 4 | |
| 5 | A PHP library to work with arbitrary precision numbers. |
| 6 | |
| 7 | [](https://github.com/brick/math/actions) |
| 8 | [](https://codecov.io/github/brick/math) |
| 9 | [](https://packagist.org/packages/brick/math) |
| 10 | [](https://packagist.org/packages/brick/math) |
| 11 | [](http://opensource.org/licenses/MIT) |
| 12 | |
| 13 | ### Installation |
| 14 | |
| 15 | This library is installable via [Composer](https://getcomposer.org/): |
| 16 | |
| 17 | ```bash |
| 18 | composer require brick/math |
| 19 | ``` |
| 20 | |
| 21 | ### Requirements |
| 22 | |
| 23 | This library requires PHP 8.2 or later. |
| 24 | |
| 25 | For PHP 8.1 compatibility, you can use version `0.13`. For PHP 8.0, you can use version `0.11`. For PHP 7.4, you can use version `0.10`. For PHP 7.1, 7.2 & 7.3, you can use version `0.9`. Note that [PHP versions < 8.1 are EOL](http://php.net/supported-versions.php) and not supported anymore. If you're still using one of these PHP versions, you should consider upgrading as soon as possible. |
| 26 | |
| 27 | Although the library can work seamlessly on any PHP installation, it is highly recommended that you install the |
| 28 | [GMP](http://php.net/manual/en/book.gmp.php) or [BCMath](http://php.net/manual/en/book.bc.php) extension |
| 29 | to speed up calculations. The fastest available calculator implementation will be automatically selected at runtime. |
| 30 | |
| 31 | ### Project status & release process |
| 32 | |
| 33 | While this library is still under development, it is well tested and considered stable enough to use in production |
| 34 | environments. |
| 35 | |
| 36 | The current releases are numbered `0.x.y`. When a non-breaking change is introduced (adding new methods, optimizing |
| 37 | existing code, etc.), `y` is incremented. |
| 38 | |
| 39 | **When a breaking change is introduced, a new `0.x` version cycle is always started.** |
| 40 | |
| 41 | It is therefore safe to lock your project to a given release cycle, such as `^0.20`. |
| 42 | |
| 43 | If you need to upgrade to a newer release cycle, check the [release history](https://github.com/brick/math/releases) |
| 44 | for a list of changes introduced by each further `0.x.0` version. |
| 45 | |
| 46 | ### Package contents |
| 47 | |
| 48 | This library provides the following public classes in the `Brick\Math` namespace: |
| 49 | |
| 50 | - [BigNumber](https://github.com/brick/math/blob/0.20.0/src/BigNumber.php): base class for `BigInteger`, `BigDecimal` and `BigRational` |
| 51 | - [BigInteger](https://github.com/brick/math/blob/0.20.0/src/BigInteger.php): represents an arbitrary-precision integer number. |
| 52 | - [BigDecimal](https://github.com/brick/math/blob/0.20.0/src/BigDecimal.php): represents an arbitrary-precision decimal number. |
| 53 | - [BigRational](https://github.com/brick/math/blob/0.20.0/src/BigRational.php): represents an arbitrary-precision rational number (fraction), always reduced to lowest terms. |
| 54 | - [RoundingMode](https://github.com/brick/math/blob/0.20.0/src/RoundingMode.php): enum representing all available rounding modes. |
| 55 | |
| 56 | And [exceptions](#exceptions) in the `Brick\Math\Exception` namespace. |
| 57 | |
| 58 | ### Overview |
| 59 | |
| 60 | #### Instantiation |
| 61 | |
| 62 | The constructors of the classes are not public, you must use a factory method to obtain an instance. |
| 63 | |
| 64 | All classes provide an `of()` factory method that accepts any of the following types: |
| 65 | |
| 66 | - `BigNumber` instances |
| 67 | - `int` numbers |
| 68 | - `string` representations of integer, decimal and rational numbers |
| 69 | |
| 70 | Example: |
| 71 | |
| 72 | ```php |
| 73 | BigInteger::of(123546); |
| 74 | BigInteger::of('9999999999999999999999999999999999999999999'); |
| 75 | |
| 76 | BigDecimal::of('9.99999999999999999999999999999999999999999999'); |
| 77 | |
| 78 | BigRational::of('2/3'); |
| 79 | ``` |
| 80 | |
| 81 | Note that all `of()` methods accept all the representations above, *as long as it can be safely converted to |
| 82 | the current type*: |
| 83 | |
| 84 | ```php |
| 85 | BigInteger::of('1.00'); // 1 |
| 86 | BigInteger::of('1.01'); // RoundingNecessaryException |
| 87 | |
| 88 | BigDecimal::of('1/8'); // 0.125 |
| 89 | BigDecimal::of('1/3'); // RoundingNecessaryException |
| 90 | |
| 91 | BigRational::of('1.1'); // 11/10 |
| 92 | BigRational::of('1.15'); // 23/20 (reduced to lowest terms) |
| 93 | ``` |
| 94 | |
| 95 | > [!NOTE] |
| 96 | > The `of()` factory method does not accept `float` values, because casting a float to string can be lossy. |
| 97 | > To convert a float to a `BigDecimal`, use one of the dedicated methods: |
| 98 | > |
| 99 | > ```php |
| 100 | > // Exact IEEE-754 representation — the value the float actually holds: |
| 101 | > BigDecimal::fromFloatExact(0.1); // 0.1000000000000000055511151231257827021181583404541015625 |
| 102 | > |
| 103 | > // Shortest decimal that round-trips back to the same float: |
| 104 | > BigDecimal::fromFloatShortest(0.1); // 0.1 |
| 105 | > ``` |
| 106 | |
| 107 | #### Parsing untrusted input |
| 108 | |
| 109 | `of()` places no hard limits on its input: a string with millions of digits is accepted as is, and a number in |
| 110 | exponential notation is expanded to its full length, so a string as short as `1e1000000000` yields a number with |
| 111 | a billion digits. |
| 112 | |
| 113 | If your input comes from an untrusted source, such as an HTTP request, use `parse()` instead, which requires you |
| 114 | to specify the allowed syntax and a maximum number of digits: |
| 115 | |
| 116 | ```php |
| 117 | use Brick\Math\NumberSyntax; |
| 118 | |
| 119 | BigDecimal::parse($input, allowedSyntax: NumberSyntax::DECIMAL, maxDigits: 20); |
| 120 | ``` |
| 121 | |
| 122 | The `$allowedSyntax` parameter restricts the accepted notations. Plain integers such as `123` are always accepted, |
| 123 | and each `NumberSyntax` case (`DecimalPoint`, `Exponent`, `Fraction`) allows one additional feature. The enum also |
| 124 | provides constants for the most common combinations: |
| 125 | |
| 126 | - `NumberSyntax::INTEGER` — integers only: `123` |
| 127 | - `NumberSyntax::DECIMAL` — integers and decimal numbers: `123`, `123.45`; typical for monetary input |
| 128 | - `NumberSyntax::SCIENTIFIC` — integers, decimal numbers and exponents: `123`, `123.45`, `1.5e-3`; accepts every JSON number |
| 129 | - `NumberSyntax::RATIONAL` — integers and fractions: `123`, `22/7` |
| 130 | - `NumberSyntax::ALL` — the full syntax accepted by `of()`: `123`, `123.45`, `1.5e-3`, `22/7` |
| 131 | |
| 132 | The `$maxDigits` parameter limits the number of digits, counted both as written in the input and in the resulting number, |
| 133 | so that a value such as `1e1000000000` is rejected before it is ever expanded: |
| 134 | |
| 135 | ```php |
| 136 | BigDecimal::parse('123.45', allowedSyntax: NumberSyntax::DECIMAL, maxDigits: 20); // 123.45 |
| 137 | BigDecimal::parse('1.2e3', allowedSyntax: NumberSyntax::DECIMAL, maxDigits: 20); // NumberFormatException (exponent not allowed) |
| 138 | BigDecimal::parse('1e1000000000', allowedSyntax: NumberSyntax::SCIENTIFIC, maxDigits: 20); // NumberFormatException (too many digits) |
| 139 | ``` |
| 140 | |
| 141 | #### Immutability & chaining |
| 142 | |
| 143 | The `BigInteger`, `BigDecimal` and `BigRational` classes are immutable: their value never changes, |
| 144 | so that they can be safely passed around. All methods that return a `BigInteger`, `BigDecimal` or `BigRational` |
| 145 | return a new object, leaving the original object unaffected: |
| 146 | |
| 147 | ```php |
| 148 | $ten = BigInteger::of(10); |
| 149 | |
| 150 | echo $ten->plus(5); // 15 |
| 151 | echo $ten->multipliedBy(3); // 30 |
| 152 | ``` |
| 153 | |
| 154 | The methods can be chained for better readability: |
| 155 | |
| 156 | ```php |
| 157 | echo BigInteger::of(10)->plus(5)->multipliedBy(3); // 45 |
| 158 | ``` |
| 159 | |
| 160 | #### Parameter types |
| 161 | |
| 162 | All methods that accept a number: `plus()`, `minus()`, `multipliedBy()`, etc. accept the same types as `of()`. |
| 163 | For example, given the following number: |
| 164 | |
| 165 | ```php |
| 166 | $integer = BigInteger::of(123); |
| 167 | ``` |
| 168 | |
| 169 | The following lines are equivalent: |
| 170 | |
| 171 | ```php |
| 172 | $integer->multipliedBy(123); |
| 173 | $integer->multipliedBy('123'); |
| 174 | $integer->multipliedBy($integer); |
| 175 | ``` |
| 176 | |
| 177 | Just like `of()`, other types of `BigNumber` are acceptable, as long as they can be safely converted to the current type: |
| 178 | |
| 179 | ```php |
| 180 | echo BigInteger::of(2)->multipliedBy(BigDecimal::of('2.0')); // 4 |
| 181 | echo BigInteger::of(2)->multipliedBy(BigDecimal::of('2.5')); // RoundingNecessaryException |
| 182 | echo BigDecimal::of(2.5)->multipliedBy(BigInteger::of(2)); // 5.0 |
| 183 | ``` |
| 184 | |
| 185 | These parameters are converted with `of()`, so the same rules apply: for untrusted strings, use |
| 186 | [`parse()`](#parsing-untrusted-input) first, and pass the resulting number to the method. |
| 187 | |
| 188 | #### Division & rounding |
| 189 | |
| 190 | ##### BigInteger |
| 191 | |
| 192 | By default, dividing a `BigInteger` returns the exact result of the division, or throws an exception if the remainder |
| 193 | of the division is not zero: |
| 194 | |
| 195 | ```php |
| 196 | echo BigInteger::of(999)->dividedBy(3); // 333 |
| 197 | echo BigInteger::of(1000)->dividedBy(3); // RoundingNecessaryException |
| 198 | ``` |
| 199 | |
| 200 | You can pass an optional [rounding mode](https://github.com/brick/math/blob/0.20.0/src/RoundingMode.php) to round the result, if necessary: |
| 201 | |
| 202 | ```php |
| 203 | echo BigInteger::of(1000)->dividedBy(3, RoundingMode::Down); // 333 |
| 204 | echo BigInteger::of(1000)->dividedBy(3, RoundingMode::Up); // 334 |
| 205 | ``` |
| 206 | |
| 207 | If you're into quotients and remainders, there are methods for this, too: |
| 208 | |
| 209 | ```php |
| 210 | echo BigInteger::of(1000)->quotient(3); // 333 |
| 211 | echo BigInteger::of(1000)->remainder(3); // 1 |
| 212 | ``` |
| 213 | |
| 214 | You can even get both at the same time: |
| 215 | |
| 216 | ```php |
| 217 | [$quotient, $remainder] = BigInteger::of(1000)->quotientAndRemainder(3); |
| 218 | ``` |
| 219 | |
| 220 | ##### BigDecimal |
| 221 | |
| 222 | Dividing a `BigDecimal` always requires a scale to be specified. If the exact result of the division does not fit in |
| 223 | the given scale, a [rounding mode](https://github.com/brick/math/blob/0.20.0/src/RoundingMode.php) must be provided. |
| 224 | |
| 225 | ```php |
| 226 | echo BigDecimal::of(1)->dividedBy('8', 3); // 0.125 |
| 227 | echo BigDecimal::of(1)->dividedBy('8', 2); // RoundingNecessaryException |
| 228 | echo BigDecimal::of(1)->dividedBy('8', 2, RoundingMode::HalfDown); // 0.12 |
| 229 | echo BigDecimal::of(1)->dividedBy('8', 2, RoundingMode::HalfUp); // 0.13 |
| 230 | ``` |
| 231 | |
| 232 | If you know that the division yields a finite number of decimals places, you can use `dividedByExact()`, which will |
| 233 | automatically compute the required scale to fit the result, or throw an exception if the division yields an infinite |
| 234 | repeating decimal: |
| 235 | |
| 236 | ```php |
| 237 | echo BigDecimal::of(1)->dividedByExact(256); // 0.00390625 |
| 238 | echo BigDecimal::of(1)->dividedByExact(11); // RoundingNecessaryException |
| 239 | ``` |
| 240 | |
| 241 | ##### BigRational |
| 242 | |
| 243 | The result of the division of a `BigRational` can always be represented exactly: |
| 244 | |
| 245 | ```php |
| 246 | echo BigRational::of('13/99')->dividedBy('7'); // 13/693 |
| 247 | echo BigRational::of('13/99')->dividedBy('9/8'); // 104/891 |
| 248 | ``` |
| 249 | |
| 250 | #### Bitwise operations |
| 251 | |
| 252 | `BigInteger` supports bitwise operations: |
| 253 | |
| 254 | - `and()` |
| 255 | - `or()` |
| 256 | - `xor()` |
| 257 | - `not()` |
| 258 | |
| 259 | and bit shifting: |
| 260 | |
| 261 | - `shiftedLeft()` |
| 262 | - `shiftedRight()` |
| 263 | |
| 264 | #### Exceptions |
| 265 | |
| 266 | All exceptions thrown by this library implement the `MathException` interface. |
| 267 | This means that you can safely catch all exceptions thrown by this library using a single catch clause: |
| 268 | |
| 269 | ```php |
| 270 | use Brick\Math\BigDecimal; |
| 271 | use Brick\Math\Exception\MathException; |
| 272 | |
| 273 | try { |
| 274 | $number = BigInteger::of(1)->dividedBy(3); |
| 275 | } catch (MathException $e) { |
| 276 | // ... |
| 277 | } |
| 278 | ``` |
| 279 | |
| 280 | If you need more granular control over the exceptions thrown, you can catch the specific exception classes: |
| 281 | |
| 282 | - `DivisionByZeroException` |
| 283 | - `IntegerOverflowException` |
| 284 | - `InvalidArgumentException` |
| 285 | - `NegativeNumberException` |
| 286 | - `NoInverseException` |
| 287 | - `NumberFormatException` |
| 288 | - `PlatformException` |
| 289 | - `RandomSourceException` |
| 290 | - `RoundingNecessaryException` |
| 291 | |
| 292 | #### Serialization |
| 293 | |
| 294 | `BigInteger`, `BigDecimal` and `BigRational` can be safely serialized on a machine and unserialized on another, |
| 295 | even if these machines do not share the same set of PHP extensions. |
| 296 | |
| 297 | For example, serializing on a machine with GMP support and unserializing on a machine that does not have this extension |
| 298 | installed will still work as expected. |
| 299 | |
| 300 | #### PHPStan extension |
| 301 | |
| 302 | A third-party [PHPStan extension](https://github.com/simPod/phpstan-brick-math) is available for this library. It provides more specific throw type narrowing for brick/math methods, so that PHPStan can infer the exact exception classes thrown. Note that this extension is not maintained by the author of brick/math. |