// Copyright (c) 2017, the Dart project authors. Please see the AUTHORS file // for details. All rights reserved. Use of this source code is governed by a // BSD-style license that can be found in the LICENSE file. // part of dart.core; // Copyright 2009 The Go Authors. All rights reserved. // Use of this source code is governed by a BSD-style // license that can be found in the LICENSE file. /* * Copyright (c) 2003-2005 Tom Wu * Copyright (c) 2012 Adam Singer (adam@solvr.io) * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining * a copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including * without limitation the rights to use, copy, modify, merge, publish, * distribute, sublicense, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: * * The above copyright notice and this permission notice shall be * included in all copies or substantial portions of the Software. * * THE SOFTWARE IS PROVIDED "AS-IS" AND WITHOUT WARRANTY OF ANY KIND, * EXPRESS, IMPLIED OR OTHERWISE, INCLUDING WITHOUT LIMITATION, ANY * WARRANTY OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. * * IN NO EVENT SHALL TOM WU BE LIABLE FOR ANY SPECIAL, INCIDENTAL, * INDIRECT OR CONSEQUENTIAL DAMAGES OF ANY KIND, OR ANY DAMAGES WHATSOEVER * RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER OR NOT ADVISED OF * THE POSSIBILITY OF DAMAGE, AND ON ANY THEORY OF LIABILITY, ARISING OUT * OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. * * In addition, the following condition applies: * * All redistributions must retain an intact copy of this copyright notice * and disclaimer. */ @patch class BigInt implements Comparable { @patch static BigInt get zero => _BigIntImpl.zero; @patch static BigInt get one => _BigIntImpl.one; @patch static BigInt get two => _BigIntImpl.two; @patch static BigInt parse(String source, {int radix}) => _BigIntImpl.parse(source, radix: radix); @patch static BigInt tryParse(String source, {int radix}) => _BigIntImpl._tryParse(source, radix: radix); @patch factory BigInt.from(num value) => new _BigIntImpl.from(value); } int _max(int a, int b) => a > b ? a : b; int _min(int a, int b) => a < b ? a : b; /// Allocate a new digits list of even length. Uint32List _newDigits(int length) => new Uint32List(length + (length & 1)); /** * An implementation for the arbitrarily large integer. * * The integer number is represented by a sign, an array of 32-bit unsigned * integers in little endian format, and a number of used digits in that array. */ class _BigIntImpl implements BigInt { // Bits per digit. static const int _digitBits = 32; static const int _digitBase = 1 << _digitBits; static const int _digitMask = (1 << _digitBits) - 1; // Bits per half digit. static const int _halfDigitBits = _digitBits >> 1; static const int _halfDigitMask = (1 << _halfDigitBits) - 1; static final _BigIntImpl zero = new _BigIntImpl._fromInt(0); static final _BigIntImpl one = new _BigIntImpl._fromInt(1); static final _BigIntImpl two = new _BigIntImpl._fromInt(2); static final _BigIntImpl _minusOne = -one; static final _BigIntImpl _oneDigitMask = new _BigIntImpl._fromInt(_digitMask); static final _BigIntImpl _twoDigitMask = (one << (2 * _digitBits)) - one; static final _BigIntImpl _oneBillion = new _BigIntImpl._fromInt(1000000000); static const int _minInt = -0x8000000000000000; static const int _maxInt = 0x7fffffffffffffff; // Result cache for last _divRem call. // Result cache for last _divRem call. static Uint32List _lastDividendDigits; static int _lastDividendUsed; static Uint32List _lastDivisorDigits; static int _lastDivisorUsed; static Uint32List _lastQuoRemDigits; static int _lastQuoRemUsed; static int _lastRemUsed; static int _lastRem_nsh; /// Whether this bigint is negative. final bool _isNegative; /// The unsigned digits of this bigint. /// /// The least significant digit is in slot 0. /// The list may have more digits than needed. That is, `_digits.length` may /// be strictly greater than `_used`. /// Also, `_digits.length` must always be even, because intrinsics on 64-bit /// platforms may process a digit pair as a 64-bit value. final Uint32List _digits; /// The number of used entries in [_digits]. /// /// To avoid reallocating [Uint32List]s, lists that are too big are not /// replaced, but `_used` reflects the smaller number of digits actually used. /// /// Note that functions shortening an existing list of digits to a smaller /// `_used` number of digits must ensure that the highermost pair of digits /// is correct when read as a 64-bit value by intrinsics. Therefore, if the /// smaller '_used' number is odd, the high digit of that pair must be /// explicitly cleared, i.e. _digits[_used] = 0, which cannot result in an /// out of bounds access, since the length of the list is guaranteed to be /// even. final int _used; /** * Parses [source] as a, possibly signed, integer literal and returns its * value. * * The [source] must be a non-empty sequence of base-[radix] digits, * optionally prefixed with a minus or plus sign ('-' or '+'). * * The [radix] must be in the range 2..36. The digits used are * first the decimal digits 0..9, and then the letters 'a'..'z' with * values 10 through 35. Also accepts upper-case letters with the same * values as the lower-case ones. * * If no [radix] is given then it defaults to 10. In this case, the [source] * digits may also start with `0x`, in which case the number is interpreted * as a hexadecimal literal, which effectively means that the `0x` is ignored * and the radix is instead set to 16. * * For any int `n` and radix `r`, it is guaranteed that * `n == int.parse(n.toRadixString(r), radix: r)`. * * Throws a [FormatException] if the [source] is not a valid integer literal, * optionally prefixed by a sign. */ static _BigIntImpl parse(String source, {int radix}) { var result = _tryParse(source, radix: radix); if (result == null) { throw new FormatException("Could not parse BigInt", source); } return result; } /// Parses a decimal bigint literal. /// /// The [source] must not contain leading or trailing whitespace. static _BigIntImpl _parseDecimal(String source, bool isNegative) { const _0 = 48; int part = 0; _BigIntImpl result = zero; // Read in the source 9 digits at a time. // The first part may have a few leading virtual '0's to make the remaining // parts all have exactly 9 digits. int digitInPartCount = 9 - source.length.remainder(9); if (digitInPartCount == 9) digitInPartCount = 0; for (int i = 0; i < source.length; i++) { part = part * 10 + source.codeUnitAt(i) - _0; if (++digitInPartCount == 9) { result = result * _oneBillion + new _BigIntImpl._fromInt(part); part = 0; digitInPartCount = 0; } } if (isNegative) return -result; return result; } /// Returns the value of a given source digit. /// /// Source digits between "0" and "9" (inclusive) return their decimal value. /// /// Source digits between "a" and "z", or "A" and "Z" (inclusive) return /// 10 + their position in the ASCII alphabet. /// /// The incoming [codeUnit] must be an ASCII code-unit. static int _codeUnitToRadixValue(int codeUnit) { // We know that the characters must be ASCII as otherwise the // regexp wouldn't have matched. Lowercasing by doing `| 0x20` is thus // guaranteed to be a safe operation, since it preserves digits // and lower-cases ASCII letters. const int _0 = 48; const int _9 = 57; const int _a = 97; if (_0 <= codeUnit && codeUnit <= _9) return codeUnit - _0; codeUnit |= 0x20; var result = codeUnit - _a + 10; return result; } /// Parses the given [source] string, starting at [startPos], as a hex /// literal. /// /// If [isNegative] is true, negates the result before returning it. /// /// The [source] (substring) must be a valid hex literal. static _BigIntImpl _parseHex(String source, int startPos, bool isNegative) { int hexCharsPerDigit = _digitBits ~/ 4; int sourceLength = source.length - startPos; int used = (sourceLength + hexCharsPerDigit - 1) ~/ hexCharsPerDigit; var digits = _newDigits(used); int lastDigitLength = sourceLength - (used - 1) * hexCharsPerDigit; int digitIndex = used - 1; int i = startPos; int digit = 0; for (int j = 0; j < lastDigitLength; j++) { var value = _codeUnitToRadixValue(source.codeUnitAt(i++)); if (value >= 16) return null; digit = digit * 16 + value; } digits[digitIndex--] = digit; while (i < source.length) { digit = 0; for (int j = 0; j < hexCharsPerDigit; j++) { var value = _codeUnitToRadixValue(source.codeUnitAt(i++)); if (value >= 16) return null; digit = digit * 16 + value; } digits[digitIndex--] = digit; } if (used == 1 && digits[0] == 0) return zero; return new _BigIntImpl._(isNegative, used, digits); } /// Parses the given [source] as a [radix] literal. /// /// The [source] will be checked for invalid characters. If it is invalid, /// this function returns `null`. static _BigIntImpl _parseRadix(String source, int radix, bool isNegative) { var result = zero; var base = new _BigIntImpl._fromInt(radix); for (int i = 0; i < source.length; i++) { var value = _codeUnitToRadixValue(source.codeUnitAt(i)); if (value >= radix) return null; result = result * base + new _BigIntImpl._fromInt(value); } if (isNegative) return -result; return result; } /// Tries to parse the given [source] as a [radix] literal. /// /// Returns the parsed big integer, or `null` if it failed. /// /// If the [radix] is `null` accepts decimal literals or `0x` hex literals. static _BigIntImpl _tryParse(String source, {int radix}) { if (source == "") return null; var re = new RegExp(r'^\s*([+-]?)((0x[a-f0-9]+)|(\d+)|([a-z0-9]+))\s*$', caseSensitive: false); var match = re.firstMatch(source); int signIndex = 1; int hexIndex = 3; int decimalIndex = 4; int nonDecimalHexIndex = 5; if (match == null) return null; bool isNegative = match[signIndex] == "-"; String decimalMatch = match[decimalIndex]; String hexMatch = match[hexIndex]; String nonDecimalMatch = match[nonDecimalHexIndex]; if (radix == null) { if (decimalMatch != null) { // Cannot fail because we know that the digits are all decimal. return _parseDecimal(decimalMatch, isNegative); } if (hexMatch != null) { // Cannot fail because we know that the digits are all hex. return _parseHex(hexMatch, 2, isNegative); } return null; } if (radix is! int) { throw new ArgumentError.value(radix, 'radix', 'is not an integer'); } if (radix < 2 || radix > 36) { throw new RangeError.range(radix, 2, 36, 'radix'); } if (radix == 10 && decimalMatch != null) { return _parseDecimal(decimalMatch, isNegative); } if (radix == 16 && (decimalMatch != null || nonDecimalMatch != null)) { return _parseHex(decimalMatch ?? nonDecimalMatch, 0, isNegative); } return _parseRadix( decimalMatch ?? nonDecimalMatch ?? hexMatch, radix, isNegative); } /// Finds the amount significant digits in the provided [digits] array. static int _normalize(int used, Uint32List digits) { while (used > 0 && digits[used - 1] == 0) used--; return used; } /// Factory returning an instance initialized with the given field values. /// If the [digits] array contains leading 0s, the [used] value is adjusted /// accordingly. The [digits] array is not modified. _BigIntImpl._(bool isNegative, int used, Uint32List digits) : this._normalized(isNegative, _normalize(used, digits), digits); _BigIntImpl._normalized(bool isNegative, this._used, this._digits) : _isNegative = _used == 0 ? false : isNegative { assert(_digits.length.isEven); assert(_used.isEven || _digits[_used] == 0); // Leading zero for 64-bit. } /// Whether this big integer is zero. bool get _isZero => _used == 0; /// Allocates an array of the given [length] and copies the [digits] in the /// range [from] to [to-1], starting at index 0, followed by leading zero /// digits. static Uint32List _cloneDigits( Uint32List digits, int from, int to, int length) { var resultDigits = _newDigits(length); var n = to - from; for (var i = 0; i < n; i++) { resultDigits[i] = digits[from + i]; } return resultDigits; } /// Allocates a big integer from the provided [value] number. factory _BigIntImpl.from(num value) { if (value == 0) return zero; if (value == 1) return one; if (value == 2) return two; if (value.abs() < 0x100000000) return new _BigIntImpl._fromInt(value.toInt()); if (value is double) return new _BigIntImpl._fromDouble(value); return new _BigIntImpl._fromInt(value); } factory _BigIntImpl._fromInt(int value) { bool isNegative = value < 0; assert(_digitBits == 32); var digits = _newDigits(2); if (isNegative) { // Handle the min 64-bit value differently, since its negation is not // positive. if (value == _minInt) { digits[1] = 0x80000000; return new _BigIntImpl._(true, 2, digits); } value = -value; } if (value < _digitBase) { digits[0] = value; return new _BigIntImpl._(isNegative, 1, digits); } digits[0] = value & _digitMask; digits[1] = value >> _digitBits; return new _BigIntImpl._(isNegative, 2, digits); } /// An 8-byte Uint8List we can reuse for [_fromDouble] to avoid generating /// garbage. static final Uint8List _bitsForFromDouble = new Uint8List(8); factory _BigIntImpl._fromDouble(double value) { const int exponentBias = 1075; if (value.isNaN || value.isInfinite) { throw new ArgumentError("Value must be finite: $value"); } bool isNegative = value < 0; if (isNegative) value = -value; value = value.floorToDouble(); if (value == 0) return zero; var bits = _bitsForFromDouble; for (int i = 0; i < 8; i++) { bits[i] = 0; } bits.buffer.asByteData().setFloat64(0, value, Endian.little); // The exponent is in bits 53..63. var biasedExponent = (bits[7] << 4) + (bits[6] >> 4); var exponent = biasedExponent - exponentBias; assert(_digitBits == 32); // The significant bits are in 0 .. 52. var unshiftedDigits = _newDigits(2); unshiftedDigits[0] = (bits[3] << 24) + (bits[2] << 16) + (bits[1] << 8) + bits[0]; // Don't forget to add the hidden bit. unshiftedDigits[1] = ((0x10 | (bits[6] & 0xF)) << 16) + (bits[5] << 8) + bits[4]; var unshiftedBig = new _BigIntImpl._normalized(false, 2, unshiftedDigits); _BigIntImpl absResult = unshiftedBig; if (exponent < 0) { absResult = unshiftedBig >> -exponent; } else if (exponent > 0) { absResult = unshiftedBig << exponent; } if (isNegative) return -absResult; return absResult; } /** * Return the negative value of this integer. * * The result of negating an integer always has the opposite sign, except * for zero, which is its own negation. */ _BigIntImpl operator -() { if (_used == 0) return this; return new _BigIntImpl._(!_isNegative, _used, _digits); } /** * Returns the absolute value of this integer. * * For any integer `x`, the result is the same as `x < 0 ? -x : x`. */ _BigIntImpl abs() => _isNegative ? -this : this; /// Returns this << n*_digitBits. _BigIntImpl _dlShift(int n) { final used = _used; if (used == 0) { return zero; } final resultUsed = used + n; final digits = _digits; final resultDigits = _newDigits(resultUsed); for (int i = used - 1; i >= 0; i--) { resultDigits[i + n] = digits[i]; } return new _BigIntImpl._(_isNegative, resultUsed, resultDigits); } /// Same as [_dlShift] but works on the decomposed big integers. /// /// Returns `resultUsed`. /// /// `resultDigits[0..resultUsed-1] = xDigits[0..xUsed-1] << n*_digitBits`. static int _dlShiftDigits( Uint32List xDigits, int xUsed, int n, Uint32List resultDigits) { if (xUsed == 0) { return 0; } if (n == 0 && identical(resultDigits, xDigits)) { return xUsed; } final resultUsed = xUsed + n; assert(resultDigits.length >= resultUsed + (resultUsed & 1)); for (int i = xUsed - 1; i >= 0; i--) { resultDigits[i + n] = xDigits[i]; } for (int i = n - 1; i >= 0; i--) { resultDigits[i] = 0; } if (resultUsed.isOdd) { resultDigits[resultUsed] = 0; } return resultUsed; } /// Returns `this >> n*_digitBits`. _BigIntImpl _drShift(int n) { final used = _used; if (used == 0) { return zero; } final resultUsed = used - n; if (resultUsed <= 0) { return _isNegative ? _minusOne : zero; } final digits = _digits; final resultDigits = _newDigits(resultUsed); for (var i = n; i < used; i++) { resultDigits[i - n] = digits[i]; } final result = new _BigIntImpl._(_isNegative, resultUsed, resultDigits); if (_isNegative) { // Round down if any bit was shifted out. for (var i = 0; i < n; i++) { if (digits[i] != 0) { return result - one; } } } return result; } /// Same as [_drShift] but works on the decomposed big integers. /// /// Returns `resultUsed`. /// /// `resultDigits[0..resultUsed-1] = xDigits[0..xUsed-1] >> n*_digitBits`. static int _drShiftDigits( Uint32List xDigits, int xUsed, int n, Uint32List resultDigits) { final resultUsed = xUsed - n; if (resultUsed <= 0) { return 0; } assert(resultDigits.length >= resultUsed + (resultUsed & 1)); for (var i = n; i < xUsed; i++) { resultDigits[i - n] = xDigits[i]; } if (resultUsed.isOdd) { resultDigits[resultUsed] = 0; } return resultUsed; } /// Shifts the digits of [xDigits] into the right place in [resultDigits]. /// /// `resultDigits[ds..xUsed+ds] = xDigits[0..xUsed-1] << (n % _digitBits)` /// where `ds = ceil(n / _digitBits)` /// /// Does *not* clear digits below ds. /// /// Note: This function may be intrinsified. static void _lsh( Uint32List xDigits, int xUsed, int n, Uint32List resultDigits) { assert(xUsed > 0); final digitShift = n ~/ _digitBits; final bitShift = n % _digitBits; final carryBitShift = _digitBits - bitShift; final bitMask = (1 << carryBitShift) - 1; var carry = 0; for (int i = xUsed - 1; i >= 0; i--) { final digit = xDigits[i]; resultDigits[i + digitShift + 1] = (digit >> carryBitShift) | carry; carry = (digit & bitMask) << bitShift; } resultDigits[digitShift] = carry; } /** * Shift the bits of this integer to the left by [shiftAmount]. * * Shifting to the left makes the number larger, effectively multiplying * the number by `pow(2, shiftIndex)`. * * There is no limit on the size of the result. It may be relevant to * limit intermediate values by using the "and" operator with a suitable * mask. * * It is an error if [shiftAmount] is negative. */ _BigIntImpl operator <<(int shiftAmount) { if (shiftAmount < 0) { throw new ArgumentError("shift-amount must be positive $shiftAmount"); } if (_isZero) return this; final digitShift = shiftAmount ~/ _digitBits; final bitShift = shiftAmount % _digitBits; if (bitShift == 0) { return _dlShift(digitShift); } // Need one extra digit to hold bits shifted by bitShift. var resultUsed = _used + digitShift + 1; var resultDigits = _newDigits(resultUsed); _lsh(_digits, _used, shiftAmount, resultDigits); return new _BigIntImpl._(_isNegative, resultUsed, resultDigits); } /// resultDigits[0..resultUsed-1] = xDigits[0..xUsed-1] << n. /// Returns resultUsed. static int _lShiftDigits( Uint32List xDigits, int xUsed, int n, Uint32List resultDigits) { final digitsShift = n ~/ _digitBits; final bitShift = n % _digitBits; if (bitShift == 0) { return _dlShiftDigits(xDigits, xUsed, digitsShift, resultDigits); } // Need one extra digit to hold bits shifted by bitShift. var resultUsed = xUsed + digitsShift + 1; assert(resultDigits.length >= resultUsed + (resultUsed & 1)); _lsh(xDigits, xUsed, n, resultDigits); var i = digitsShift; while (--i >= 0) { resultDigits[i] = 0; } if (resultDigits[resultUsed - 1] == 0) { resultUsed--; // Clamp result. } else if (resultUsed.isOdd) { resultDigits[resultUsed] = 0; } return resultUsed; } /// resultDigits[0..resultUsed-1] = xDigits[0..xUsed-1] >> n. /// /// Note: This function may be intrinsified. static void _rsh( Uint32List xDigits, int xUsed, int n, Uint32List resultDigits) { assert(xUsed > 0); final digitsShift = n ~/ _digitBits; final bitShift = n % _digitBits; final carryBitShift = _digitBits - bitShift; final bitMask = (1 << bitShift) - 1; var carry = xDigits[digitsShift] >> bitShift; final last = xUsed - digitsShift - 1; for (var i = 0; i < last; i++) { final digit = xDigits[i + digitsShift + 1]; resultDigits[i] = ((digit & bitMask) << carryBitShift) | carry; carry = digit >> bitShift; } resultDigits[last] = carry; } /** * Shift the bits of this integer to the right by [shiftAmount]. * * Shifting to the right makes the number smaller and drops the least * significant bits, effectively doing an integer division by *`pow(2, shiftIndex)`. * * It is an error if [shiftAmount] is negative. */ _BigIntImpl operator >>(int shiftAmount) { if (shiftAmount < 0) { throw new ArgumentError("shift-amount must be positive $shiftAmount"); } if (_isZero) return this; final digitShift = shiftAmount ~/ _digitBits; final bitShift = shiftAmount % _digitBits; if (bitShift == 0) { return _drShift(digitShift); } final used = _used; final resultUsed = used - digitShift; if (resultUsed <= 0) { return _isNegative ? _minusOne : zero; } final digits = _digits; final resultDigits = _newDigits(resultUsed); _rsh(digits, used, shiftAmount, resultDigits); final result = new _BigIntImpl._(_isNegative, resultUsed, resultDigits); if (_isNegative) { // Round down if any bit was shifted out. if ((digits[digitShift] & ((1 << bitShift) - 1)) != 0) { return result - one; } for (var i = 0; i < digitShift; i++) { if (digits[i] != 0) { return result - one; } } } return result; } /// resultDigits[0..resultUsed-1] = xDigits[0..xUsed-1] >> n. /// Returns resultUsed. static int _rShiftDigits( Uint32List xDigits, int xUsed, int n, Uint32List resultDigits) { final digitShift = n ~/ _digitBits; final bitShift = n % _digitBits; if (bitShift == 0) { return _drShiftDigits(xDigits, xUsed, digitShift, resultDigits); } var resultUsed = xUsed - digitShift; if (resultUsed <= 0) { return 0; } assert(resultDigits.length >= resultUsed + (resultUsed & 1)); _rsh(xDigits, xUsed, n, resultDigits); if (resultDigits[resultUsed - 1] == 0) { resultUsed--; // Clamp result. } else if (resultUsed.isOdd) { resultDigits[resultUsed] = 0; } return resultUsed; } /// Compares this to [other] taking the absolute value of both operands. /// /// Returns 0 if abs(this) == abs(other); a positive number if /// abs(this) > abs(other); and a negative number if abs(this) < abs(other). int _absCompare(BigInt bigInt) { _BigIntImpl other = bigInt; return _compareDigits(_digits, _used, other._digits, other._used); } /** * Compares this to `other`. * * Returns a negative number if `this` is less than `other`, zero if they are * equal, and a positive number if `this` is greater than `other`. */ int compareTo(BigInt bigInt) { _BigIntImpl other = bigInt; if (_isNegative == other._isNegative) { var result = _absCompare(other); // Use 0 - result to avoid negative zero in JavaScript. return _isNegative ? 0 - result : result; } return _isNegative ? -1 : 1; } /// Compares `digits[0..used-1]` with `otherDigits[0..otherUsed-1]`. /// /// Returns 0 if equal; a positive number if larger; /// and a negative number if smaller. static int _compareDigits( Uint32List digits, int used, Uint32List otherDigits, int otherUsed) { var result = used - otherUsed; if (result == 0) { for (int i = used - 1; i >= 0; i--) { result = digits[i] - otherDigits[i]; if (result != 0) return result; } } return result; } /// resultDigits[0..used] = digits[0..used-1] + otherDigits[0..otherUsed-1]. /// used >= otherUsed > 0. /// /// Note: This function may be intrinsified. static void _absAdd(Uint32List digits, int used, Uint32List otherDigits, int otherUsed, Uint32List resultDigits) { assert(used >= otherUsed && otherUsed > 0); var carry = 0; for (var i = 0; i < otherUsed; i++) { carry += digits[i] + otherDigits[i]; resultDigits[i] = carry & _digitMask; carry >>= _digitBits; } for (var i = otherUsed; i < used; i++) { carry += digits[i]; resultDigits[i] = carry & _digitMask; carry >>= _digitBits; } resultDigits[used] = carry; } /// resultDigits[0..used-1] = digits[0..used-1] - otherDigits[0..otherUsed-1]. /// used >= otherUsed > 0. /// /// Note: This function may be intrinsified. static void _absSub(Uint32List digits, int used, Uint32List otherDigits, int otherUsed, Uint32List resultDigits) { assert(used >= otherUsed && otherUsed > 0); var carry = 0; for (var i = 0; i < otherUsed; i++) { carry += digits[i] - otherDigits[i]; resultDigits[i] = carry & _digitMask; carry >>= _digitBits; } for (var i = otherUsed; i < used; i++) { carry += digits[i]; resultDigits[i] = carry & _digitMask; carry >>= _digitBits; } } /// Returns `abs(this) + abs(other)` with sign set according to [isNegative]. _BigIntImpl _absAddSetSign(_BigIntImpl other, bool isNegative) { var used = _used; var otherUsed = other._used; if (used < otherUsed) { return other._absAddSetSign(this, isNegative); } if (used == 0) { assert(!isNegative); return zero; } if (otherUsed == 0) { return _isNegative == isNegative ? this : -this; } var resultUsed = used + 1; var resultDigits = _newDigits(resultUsed); _absAdd(_digits, used, other._digits, otherUsed, resultDigits); return new _BigIntImpl._(isNegative, resultUsed, resultDigits); } /// Returns `abs(this) - abs(other)` with sign set according to [isNegative]. /// /// Requirement: `abs(this) >= abs(other)`. _BigIntImpl _absSubSetSign(_BigIntImpl other, bool isNegative) { assert(_absCompare(other) >= 0); var used = _used; if (used == 0) { assert(!isNegative); return zero; } var otherUsed = other._used; if (otherUsed == 0) { return _isNegative == isNegative ? this : -this; } var resultDigits = _newDigits(used); _absSub(_digits, used, other._digits, otherUsed, resultDigits); return new _BigIntImpl._(isNegative, used, resultDigits); } /// Returns `abs(this) & abs(other)` with sign set according to [isNegative]. _BigIntImpl _absAndSetSign(_BigIntImpl other, bool isNegative) { var resultUsed = _min(_used, other._used); var digits = _digits; var otherDigits = other._digits; var resultDigits = _newDigits(resultUsed); for (var i = 0; i < resultUsed; i++) { resultDigits[i] = digits[i] & otherDigits[i]; } return new _BigIntImpl._(isNegative, resultUsed, resultDigits); } /// Returns `abs(this) &~ abs(other)` with sign set according to [isNegative]. _BigIntImpl _absAndNotSetSign(_BigIntImpl other, bool isNegative) { var resultUsed = _used; var digits = _digits; var otherDigits = other._digits; var resultDigits = _newDigits(resultUsed); var m = _min(resultUsed, other._used); for (var i = 0; i < m; i++) { resultDigits[i] = digits[i] & ~otherDigits[i]; } for (var i = m; i < resultUsed; i++) { resultDigits[i] = digits[i]; } return new _BigIntImpl._(isNegative, resultUsed, resultDigits); } /// Returns `abs(this) | abs(other)` with sign set according to [isNegative]. _BigIntImpl _absOrSetSign(_BigIntImpl other, bool isNegative) { var used = _used; var otherUsed = other._used; var resultUsed = _max(used, otherUsed); var digits = _digits; var otherDigits = other._digits; var resultDigits = _newDigits(resultUsed); var l, m; if (used < otherUsed) { l = other; m = used; } else { l = this; m = otherUsed; } for (var i = 0; i < m; i++) { resultDigits[i] = digits[i] | otherDigits[i]; } var lDigits = l._digits; for (var i = m; i < resultUsed; i++) { resultDigits[i] = lDigits[i]; } return new _BigIntImpl._(isNegative, resultUsed, resultDigits); } /// Returns `abs(this) ^ abs(other)` with sign set according to [isNegative]. _BigIntImpl _absXorSetSign(_BigIntImpl other, bool isNegative) { var used = _used; var otherUsed = other._used; var resultUsed = _max(used, otherUsed); var digits = _digits; var otherDigits = other._digits; var resultDigits = _newDigits(resultUsed); var l, m; if (used < otherUsed) { l = other; m = used; } else { l = this; m = otherUsed; } for (var i = 0; i < m; i++) { resultDigits[i] = digits[i] ^ otherDigits[i]; } var lDigits = l._digits; for (var i = m; i < resultUsed; i++) { resultDigits[i] = lDigits[i]; } return new _BigIntImpl._(isNegative, resultUsed, resultDigits); } /** * Bit-wise and operator. * * Treating both `this` and [other] as sufficiently large two's component * integers, the result is a number with only the bits set that are set in * both `this` and [other] * * Of both operands are negative, the result is negative, otherwise * the result is non-negative. */ _BigIntImpl operator &(BigInt bigInt) { _BigIntImpl other = bigInt; if (_isZero || other._isZero) return zero; if (_isNegative == other._isNegative) { if (_isNegative) { // (-this) & (-other) == ~(this-1) & ~(other-1) // == ~((this-1) | (other-1)) // == -(((this-1) | (other-1)) + 1) _BigIntImpl this1 = _absSubSetSign(one, true); _BigIntImpl other1 = other._absSubSetSign(one, true); // Result cannot be zero if this and other are negative. return this1._absOrSetSign(other1, true)._absAddSetSign(one, true); } return _absAndSetSign(other, false); } // _isNegative != other._isNegative var p, n; if (_isNegative) { p = other; n = this; } else { // & is symmetric. p = this; n = other; } // p & (-n) == p & ~(n-1) == p &~ (n-1) var n1 = n._absSubSetSign(one, false); return p._absAndNotSetSign(n1, false); } /** * Bit-wise or operator. * * Treating both `this` and [other] as sufficiently large two's component * integers, the result is a number with the bits set that are set in either * of `this` and [other] * * If both operands are non-negative, the result is non-negative, * otherwise the result us negative. */ _BigIntImpl operator |(BigInt bigInt) { _BigIntImpl other = bigInt; if (_isZero) return other; if (other._isZero) return this; if (_isNegative == other._isNegative) { if (_isNegative) { // (-this) | (-other) == ~(this-1) | ~(other-1) // == ~((this-1) & (other-1)) // == -(((this-1) & (other-1)) + 1) var this1 = _absSubSetSign(one, true); var other1 = other._absSubSetSign(one, true); // Result cannot be zero if this and a are negative. return this1._absAndSetSign(other1, true)._absAddSetSign(one, true); } return _absOrSetSign(other, false); } // _neg != a._neg var p, n; if (_isNegative) { p = other; n = this; } else { // | is symmetric. p = this; n = other; } // p | (-n) == p | ~(n-1) == ~((n-1) &~ p) == -(~((n-1) &~ p) + 1) var n1 = n._absSubSetSign(one, true); // Result cannot be zero if only one of this or a is negative. return n1._absAndNotSetSign(p, true)._absAddSetSign(one, true); } /** * Bit-wise exclusive-or operator. * * Treating both `this` and [other] as sufficiently large two's component * integers, the result is a number with the bits set that are set in one, * but not both, of `this` and [other] * * If the operands have the same sign, the result is non-negative, * otherwise the result is negative. */ _BigIntImpl operator ^(BigInt bigInt) { _BigIntImpl other = bigInt; if (_isZero) return other; if (other._isZero) return this; if (_isNegative == other._isNegative) { if (_isNegative) { // (-this) ^ (-other) == ~(this-1) ^ ~(other-1) == (this-1) ^ (other-1) var this1 = _absSubSetSign(one, true); var other1 = other._absSubSetSign(one, true); return this1._absXorSetSign(other1, false); } return _absXorSetSign(other, false); } // _isNegative != a._isNegative var p, n; if (_isNegative) { p = other; n = this; } else { // ^ is symmetric. p = this; n = other; } // p ^ (-n) == p ^ ~(n-1) == ~(p ^ (n-1)) == -((p ^ (n-1)) + 1) var n1 = n._absSubSetSign(one, true); // Result cannot be zero if only one of this or a is negative. return p._absXorSetSign(n1, true)._absAddSetSign(one, true); } /** * The bit-wise negate operator. * * Treating `this` as a sufficiently large two's component integer, * the result is a number with the opposite bits set. * * This maps any integer `x` to `-x - 1`. */ _BigIntImpl operator ~() { if (_isZero) return _minusOne; if (_isNegative) { // ~(-this) == ~(~(this-1)) == this-1 return _absSubSetSign(one, false); } // ~this == -this-1 == -(this+1) // Result cannot be zero if this is positive. return _absAddSetSign(one, true); } /// Addition operator. _BigIntImpl operator +(BigInt bigInt) { _BigIntImpl other = bigInt; if (_isZero) return other; if (other._isZero) return this; var isNegative = _isNegative; if (isNegative == other._isNegative) { // this + other == this + other // (-this) + (-other) == -(this + other) return _absAddSetSign(other, isNegative); } // this + (-other) == this - other == -(this - other) // (-this) + other == other - this == -(this - other) if (_absCompare(other) >= 0) { return _absSubSetSign(other, isNegative); } return other._absSubSetSign(this, !isNegative); } /// Subtraction operator. _BigIntImpl operator -(BigInt bigInt) { _BigIntImpl other = bigInt; if (_isZero) return -other; if (other._isZero) return this; var isNegative = _isNegative; if (isNegative != other._isNegative) { // this - (-other) == this + other // (-this) - other == -(this + other) return _absAddSetSign(other, isNegative); } // this - other == this - a == -(this - other) // (-this) - (-other) == other - this == -(this - other) if (_absCompare(other) >= 0) { return _absSubSetSign(other, isNegative); } return other._absSubSetSign(this, !isNegative); } /// Multiplies `xDigits[xIndex]` with `multiplicandDigits` and adds the result /// to `accumulatorDigits`. /// /// The `multiplicandDigits` in the range `i` to `i`+`n`-1 are the /// multiplicand digits. /// /// The `acculumatorDigits` in the range `j` to `j`+`n`-1 are the accumulator /// digits. /// /// Concretely: /// `accumulatorDigits[j..j+n] += xDigits[xIndex] * m_digits[i..i+n-1]`. /// Returns 1. /// /// Note: This function may be intrinsified. Intrinsics on 64-bit platforms /// process digit pairs at even indices and returns 2. static int _mulAdd( Uint32List xDigits, int xIndex, Uint32List multiplicandDigits, int i, Uint32List accumulatorDigits, int j, int n) { int x = xDigits[xIndex]; if (x == 0) { // No-op if x is 0. return 1; } int carry = 0; int xl = x & _halfDigitMask; int xh = x >> _halfDigitBits; while (--n >= 0) { int ml = multiplicandDigits[i] & _halfDigitMask; int mh = multiplicandDigits[i++] >> _halfDigitBits; int ph = xh * ml + mh * xl; int pl = xl * ml + ((ph & _halfDigitMask) << _halfDigitBits) + accumulatorDigits[j] + carry; carry = (pl >> _digitBits) + (ph >> _halfDigitBits) + xh * mh; accumulatorDigits[j++] = pl & _digitMask; } while (carry != 0) { int l = accumulatorDigits[j] + carry; carry = l >> _digitBits; accumulatorDigits[j++] = l & _digitMask; } return 1; } /// Multiplies `xDigits[i]` with `xDigits` and adds the result to /// `accumulatorDigits`. /// /// The `xDigits` in the range `i` to `used`-1 are the multiplicand digits. /// /// The `acculumatorDigits` in the range 2*`i` to `i`+`used`-1 are the /// accumulator digits. /// /// Concretely: /// `accumulatorDigits[2*i..i+used-1] += xDigits[i]*xDigits[i] + /// 2*xDigits[i]*xDigits[i+1..used-1]`. /// Returns 1. /// /// Note: This function may be intrinsified. Intrinsics on 64-bit platforms /// process digit pairs at even indices and returns 2. static int _sqrAdd( Uint32List xDigits, int i, Uint32List acculumatorDigits, int used) { int x = xDigits[i]; if (x == 0) return 1; int j = 2 * i; int carry = 0; int xl = x & _halfDigitMask; int xh = x >> _halfDigitBits; int ph = 2 * xh * xl; int pl = xl * xl + ((ph & _halfDigitMask) << _halfDigitBits) + acculumatorDigits[j]; carry = (pl >> _digitBits) + (ph >> _halfDigitBits) + xh * xh; acculumatorDigits[j] = pl & _digitMask; x <<= 1; xl = x & _halfDigitMask; xh = x >> _halfDigitBits; int n = used - i - 1; int k = i + 1; j++; while (--n >= 0) { int l = xDigits[k] & _halfDigitMask; int h = xDigits[k++] >> _halfDigitBits; int ph = xh * l + h * xl; int pl = xl * l + ((ph & _halfDigitMask) << _halfDigitBits) + acculumatorDigits[j] + carry; carry = (pl >> _digitBits) + (ph >> _halfDigitBits) + xh * h; acculumatorDigits[j++] = pl & _digitMask; } carry += acculumatorDigits[i + used]; if (carry >= _digitBase) { acculumatorDigits[i + used] = carry - _digitBase; acculumatorDigits[i + used + 1] = 1; } else { acculumatorDigits[i + used] = carry; } return 1; } /// Multiplication operator. _BigIntImpl operator *(BigInt bigInt) { _BigIntImpl other = bigInt; var used = _used; var otherUsed = other._used; if (used == 0 || otherUsed == 0) { return zero; } var resultUsed = used + otherUsed; var digits = _digits; var otherDigits = other._digits; var resultDigits = _newDigits(resultUsed); var i = 0; while (i < otherUsed) { i += _mulAdd(otherDigits, i, digits, 0, resultDigits, i, used); } return new _BigIntImpl._( _isNegative != other._isNegative, resultUsed, resultDigits); } // resultDigits[0..resultUsed-1] = // xDigits[0..xUsed-1]*otherDigits[0..otherUsed-1]. // Returns resultUsed = xUsed + otherUsed. static int _mulDigits(Uint32List xDigits, int xUsed, Uint32List otherDigits, int otherUsed, Uint32List resultDigits) { var resultUsed = xUsed + otherUsed; var i = resultUsed + (resultUsed & 1); assert(resultDigits.length >= i); while (--i >= 0) { resultDigits[i] = 0; } i = 0; while (i < otherUsed) { i += _mulAdd(otherDigits, i, xDigits, 0, resultDigits, i, xUsed); } return resultUsed; } // resultDigits[0..resultUsed-1] = xDigits[0..xUsed-1]^2. // Returns resultUsed = 2*xUsed. static int _sqrDigits( Uint32List xDigits, int xUsed, Uint32List resultDigits) { var resultUsed = 2 * xUsed; assert(resultDigits.length >= resultUsed); // Since resultUsed is even, no need for a leading zero for // 64-bit processing. var i = resultUsed; while (--i >= 0) { resultDigits[i] = 0; } i = 0; while (i < xUsed - 1) { i += _sqrAdd(xDigits, i, resultDigits, xUsed); } // The last step is already done if digit pairs were processed above. if (i < xUsed) { _mulAdd(xDigits, i, xDigits, i, resultDigits, 2 * i, 1); } return resultUsed; } // Indices of the arguments of _estimateQuotientDigit. // For 64-bit processing by intrinsics on 64-bit platforms, the top digit pair // of the divisor is provided in the args array, and a 64-bit estimated // quotient is returned. However, on 32-bit platforms, the low 32-bit digit is // ignored and only one 32-bit digit is returned as the estimated quotient. static const int _divisorLowTopDigit = 0; // Low digit of top pair of divisor. static const int _divisorTopDigit = 1; // Top digit of divisor. static const int _quotientDigit = 2; // Estimated quotient. static const int _quotientHighDigit = 3; // High digit of estimated quotient. /// Estimate `args[_quotientDigit] = digits[i-1..i] ~/ args[_divisorTopDigit]` /// Returns 1. /// /// Note: This function may be intrinsified. Intrinsics on 64-bit platforms /// process a digit pair (i always odd): /// Estimate `args[_quotientDigit.._quotientHighDigit] = digits[i-3..i] ~/ /// args[_divisorLowTopDigit.._divisorTopDigit]`. /// Returns 2. static int _estimateQuotientDigit(Uint32List args, Uint32List digits, int i) { // Verify that digit pairs are accessible for 64-bit processing. assert(digits.length >= 4); if (digits[i] == args[_divisorTopDigit]) { args[_quotientDigit] = _digitMask; } else { // Chop off one bit, since a Mint cannot hold 2 digits. var quotientDigit = ((digits[i] << (_digitBits - 1)) | (digits[i - 1] >> 1)) ~/ (args[_divisorTopDigit] >> 1); if (quotientDigit > _digitMask) { args[_quotientDigit] = _digitMask; } else { args[_quotientDigit] = quotientDigit; } } return 1; } /// Returns `trunc(this / other)`, with `other != 0`. _BigIntImpl _div(BigInt bigInt) { _BigIntImpl other = bigInt; assert(other._used > 0); if (_used < other._used) { return zero; } _divRem(other); // Return quotient, i.e. // _lastQuoRem_digits[_lastRem_used.._lastQuoRem_used-1] with proper sign. var lastQuo_used = _lastQuoRemUsed - _lastRemUsed; var quo_digits = _cloneDigits( _lastQuoRemDigits, _lastRemUsed, _lastQuoRemUsed, lastQuo_used); var quo = new _BigIntImpl._(false, lastQuo_used, quo_digits); if ((_isNegative != other._isNegative) && (quo._used > 0)) { quo = -quo; } return quo; } /// Returns `this - other * trunc(this / other)`, with `other != 0`. _BigIntImpl _rem(BigInt bigInt) { _BigIntImpl other = bigInt; assert(other._used > 0); if (_used < other._used) { return this; } _divRem(other); // Return remainder, i.e. // denormalized _lastQuoRem_digits[0.._lastRem_used-1] with proper sign. var remDigits = _cloneDigits(_lastQuoRemDigits, 0, _lastRemUsed, _lastRemUsed); var rem = new _BigIntImpl._(false, _lastRemUsed, remDigits); if (_lastRem_nsh > 0) { rem = rem >> _lastRem_nsh; // Denormalize remainder. } if (_isNegative && (rem._used > 0)) { rem = -rem; } return rem; } /// Computes this ~/ other and this.remainder(other). /// /// Stores the result in [_lastQuoRemDigits], [_lastQuoRemUsed] and /// [_lastRemUsed]. The [_lastQuoRemDigits] contains the digits of *both*, the /// quotient and the remainder. /// /// Caches the input to avoid doing the work again when users write /// `a ~/ b` followed by a `a % b`. void _divRem(_BigIntImpl other) { // Check if result is already cached. if ((this._used == _lastDividendUsed) && (other._used == _lastDivisorUsed) && identical(this._digits, _lastDividendDigits) && identical(other._digits, _lastDivisorDigits)) { return; } assert(_used >= other._used); var nsh = _digitBits - other._digits[other._used - 1].bitLength; // For 64-bit processing, make sure other has an even number of digits. if (other._used.isOdd) { nsh += _digitBits; } // Concatenated positive quotient and normalized positive remainder. // The resultDigits can have at most one more digit than the dividend. Uint32List resultDigits; int resultUsed; // Normalized positive divisor (referred to as 'y'). // The normalized divisor has the most-significant bit of its most // significant digit set. // This makes estimating the quotient easier. Uint32List yDigits; int yUsed; if (nsh > 0) { // Extra digits for normalization, also used for possible _mulAdd carry. var numExtraDigits = (nsh + _digitBits - 1) ~/ _digitBits + 1; yDigits = _newDigits(other._used + numExtraDigits); yUsed = _lShiftDigits(other._digits, other._used, nsh, yDigits); resultDigits = _newDigits(_used + numExtraDigits); resultUsed = _lShiftDigits(_digits, _used, nsh, resultDigits); } else { yDigits = other._digits; yUsed = other._used; // Extra digit to hold possible _mulAdd carry. resultDigits = _cloneDigits(_digits, 0, _used, _used + 1); resultUsed = _used; } Uint32List args = _newDigits(4); args[_divisorLowTopDigit] = yDigits[yUsed - 2]; args[_divisorTopDigit] = yDigits[yUsed - 1]; // For 64-bit processing, make sure yUsed, i, and j are even. assert(yUsed.isEven); var i = resultUsed + (resultUsed & 1); var j = i - yUsed; // tmpDigits is a temporary array of i (even resultUsed) digits. var tmpDigits = _newDigits(i); var tmpUsed = _dlShiftDigits(yDigits, yUsed, j, tmpDigits); // Explicit first division step in case normalized dividend is larger or // equal to shifted normalized divisor. if (_compareDigits(resultDigits, resultUsed, tmpDigits, tmpUsed) >= 0) { assert(i == resultUsed); resultDigits[resultUsed++] = 1; // Quotient = 1. // Subtract divisor from remainder. _absSub(resultDigits, resultUsed, tmpDigits, tmpUsed, resultDigits); } else { // Account for possible carry in _mulAdd step. resultDigits[resultUsed++] = 0; } if (resultUsed.isOdd) { resultDigits[resultUsed] = 0; // Leading zero for 64-bit processing. } // Negate y so we can later use _mulAdd instead of non-existent _mulSub. var nyDigits = _newDigits(yUsed + 2); nyDigits[yUsed] = 1; _absSub(nyDigits, yUsed + 1, yDigits, yUsed, nyDigits); // nyDigits is read-only and has yUsed digits (possibly including several // leading zeros) plus a leading zero for 64-bit processing. // resultDigits is modified during iteration. // resultDigits[0..yUsed-1] is the current remainder. // resultDigits[yUsed..resultUsed-1] is the current quotient. --i; while (j > 0) { var d0 = _estimateQuotientDigit(args, resultDigits, i); j -= d0; var d1 = _mulAdd(args, _quotientDigit, nyDigits, 0, resultDigits, j, yUsed); // _estimateQuotientDigit and _mulAdd must agree on the number of digits // to process. assert(d0 == d1); if (d0 == 1) { if (resultDigits[i] < args[_quotientDigit]) { // Reusing the already existing tmpDigits array. var tmpUsed = _dlShiftDigits(nyDigits, yUsed, j, tmpDigits); _absSub(resultDigits, resultUsed, tmpDigits, tmpUsed, resultDigits); while (resultDigits[i] < --args[_quotientDigit]) { _absSub(resultDigits, resultUsed, tmpDigits, tmpUsed, resultDigits); } } } else { assert(d0 == 2); assert(resultDigits[i] <= args[_quotientHighDigit]); if (resultDigits[i] < args[_quotientHighDigit] || resultDigits[i - 1] < args[_quotientDigit]) { // Reusing the already existing tmpDigits array. var tmpUsed = _dlShiftDigits(nyDigits, yUsed, j, tmpDigits); _absSub(resultDigits, resultUsed, tmpDigits, tmpUsed, resultDigits); if (args[_quotientDigit] == 0) { --args[_quotientHighDigit]; } --args[_quotientDigit]; assert(resultDigits[i] <= args[_quotientHighDigit]); while (resultDigits[i] < args[_quotientHighDigit] || resultDigits[i - 1] < args[_quotientDigit]) { _absSub(resultDigits, resultUsed, tmpDigits, tmpUsed, resultDigits); if (args[_quotientDigit] == 0) { --args[_quotientHighDigit]; } --args[_quotientDigit]; assert(resultDigits[i] <= args[_quotientHighDigit]); } } } i -= d0; } // Cache result. _lastDividendDigits = _digits; _lastDividendUsed = _used; _lastDivisorDigits = other._digits; _lastDivisorUsed = other._used; _lastQuoRemDigits = resultDigits; _lastQuoRemUsed = resultUsed; _lastRemUsed = yUsed; _lastRem_nsh = nsh; } // Customized version of _rem() minimizing allocations for use in reduction. // Input: // xDigits[0..xUsed-1]: positive dividend. // yDigits[0..yUsed-1]: normalized positive divisor. // nyDigits[0..yUsed-1]: negated yDigits. // nsh: normalization shift amount. // args: top y digit(s) and place holder for estimated quotient digit(s). // tmpDigits: temp array of 2*yUsed digits. // resultDigits: result digits array large enough to temporarily hold // concatenated quotient and normalized remainder. // Output: // resultDigits[0..resultUsed-1]: positive remainder. // Returns resultUsed. static int _remDigits( Uint32List xDigits, int xUsed, Uint32List yDigits, int yUsed, Uint32List nyDigits, int nsh, Uint32List args, Uint32List tmpDigits, Uint32List resultDigits) { // Initialize resultDigits to normalized positive dividend. var resultUsed = _lShiftDigits(xDigits, xUsed, nsh, resultDigits); // For 64-bit processing, make sure yUsed, i, and j are even. assert(yUsed.isEven); var i = resultUsed + (resultUsed & 1); var j = i - yUsed; var tmpUsed = _dlShiftDigits(yDigits, yUsed, j, tmpDigits); // Explicit first division step in case normalized dividend is larger or // equal to shifted normalized divisor. if (_compareDigits(resultDigits, resultUsed, tmpDigits, tmpUsed) >= 0) { assert(i == resultUsed); resultDigits[resultUsed++] = 1; // Quotient = 1. // Subtract divisor from remainder. _absSub(resultDigits, resultUsed, tmpDigits, tmpUsed, resultDigits); } else { // Account for possible carry in _mulAdd step. resultDigits[resultUsed++] = 0; } if (resultUsed.isOdd) { resultDigits[resultUsed] = 0; // Leading zero for 64-bit processing. } // Negated yDigits passed in nyDigits allow the use of _mulAdd instead of // unimplemented _mulSub. // nyDigits is read-only and has yUsed digits (possibly including several // leading zeros) plus a leading zero for 64-bit processing. // resultDigits is modified during iteration. // resultDigits[0..yUsed-1] is the current remainder. // resultDigits[yUsed..resultUsed-1] is the current quotient. --i; while (j > 0) { var d0 = _estimateQuotientDigit(args, resultDigits, i); j -= d0; var d1 = _mulAdd(args, _quotientDigit, nyDigits, 0, resultDigits, j, yUsed); // _estimateQuotientDigit and _mulAdd must agree on the number of digits // to process. assert(d0 == d1); if (d0 == 1) { if (resultDigits[i] < args[_quotientDigit]) { var tmpUsed = _dlShiftDigits(nyDigits, yUsed, j, tmpDigits); _absSub(resultDigits, resultUsed, tmpDigits, tmpUsed, resultDigits); while (resultDigits[i] < --args[_quotientDigit]) { _absSub(resultDigits, resultUsed, tmpDigits, tmpUsed, resultDigits); } } } else { assert(d0 == 2); assert(resultDigits[i] <= args[_quotientHighDigit]); if ((resultDigits[i] < args[_quotientHighDigit]) || (resultDigits[i - 1] < args[_quotientDigit])) { var tmpUsed = _dlShiftDigits(nyDigits, yUsed, j, tmpDigits); _absSub(resultDigits, resultUsed, tmpDigits, tmpUsed, resultDigits); if (args[_quotientDigit] == 0) { --args[_quotientHighDigit]; } --args[_quotientDigit]; assert(resultDigits[i] <= args[_quotientHighDigit]); while ((resultDigits[i] < args[_quotientHighDigit]) || (resultDigits[i - 1] < args[_quotientDigit])) { _absSub(resultDigits, resultUsed, tmpDigits, tmpUsed, resultDigits); if (args[_quotientDigit] == 0) { --args[_quotientHighDigit]; } --args[_quotientDigit]; assert(resultDigits[i] <= args[_quotientHighDigit]); } } } i -= d0; } // Return remainder, i.e. denormalized resultDigits[0..yUsed-1]. resultUsed = yUsed; if (nsh > 0) { // Denormalize remainder. resultUsed = _rShiftDigits(resultDigits, resultUsed, nsh, resultDigits); } return resultUsed; } int get hashCode { // This is the [Jenkins hash function][1] but using masking to keep // values in SMI range. // // [1]: http://en.wikipedia.org/wiki/Jenkins_hash_function int combine(int hash, int value) { hash = 0x1fffffff & (hash + value); hash = 0x1fffffff & (hash + ((0x0007ffff & hash) << 10)); return hash ^ (hash >> 6); } int finish(int hash) { hash = 0x1fffffff & (hash + ((0x03ffffff & hash) << 3)); hash = hash ^ (hash >> 11); return 0x1fffffff & (hash + ((0x00003fff & hash) << 15)); } if (_isZero) return 6707; // Just a random number. var hash = _isNegative ? 83585 : 429689; // Also random. for (int i = 0; i < _used; i++) { hash = combine(hash, _digits[i]); } return finish(hash); } /** * Test whether this value is numerically equal to `other`. * * If [other] is a [_BigIntImpl] returns whether the two operands have the * same value. * * Returns false if `other` is not a [_BigIntImpl]. */ bool operator ==(Object other) => other is _BigIntImpl && compareTo(other) == 0; /** * Returns the minimum number of bits required to store this big integer. * * The number of bits excludes the sign bit, which gives the natural length * for non-negative (unsigned) values. Negative values are complemented to * return the bit position of the first bit that differs from the sign bit. * * To find the number of bits needed to store the value as a signed value, * add one, i.e. use `x.bitLength + 1`. * * ``` * x.bitLength == (-x-1).bitLength * * new BigInt.from(3).bitLength == 2; // 00000011 * new BigInt.from(2).bitLength == 2; // 00000010 * new BigInt.from(1).bitLength == 1; // 00000001 * new BigInt.from(0).bitLength == 0; // 00000000 * new BigInt.from(-1).bitLength == 0; // 11111111 * new BigInt.from(-2).bitLength == 1; // 11111110 * new BigInt.from(-3).bitLength == 2; // 11111101 * new BigInt.from(-4).bitLength == 2; // 11111100 * ``` */ int get bitLength { if (_used == 0) return 0; if (_isNegative) return (~this).bitLength; return _digitBits * (_used - 1) + _digits[_used - 1].bitLength; } /** * Truncating division operator. * * Performs a truncating integer division, where the remainder is discarded. * * The remainder can be computed using the [remainder] method. * * Examples: * ``` * var seven = new BigInt.from(7); * var three = new BigInt.from(3); * seven ~/ three; // => 2 * (-seven) ~/ three; // => -2 * seven ~/ -three; // => -2 * seven.remainder(three); // => 1 * (-seven).remainder(three); // => -1 * seven.remainder(-three); // => 1 * ``` */ _BigIntImpl operator ~/(BigInt bigInt) { _BigIntImpl other = bigInt; if (other._used == 0) { throw const IntegerDivisionByZeroException(); } return _div(other); } /** * Returns the remainder of the truncating division of `this` by [other]. * * The result `r` of this operation satisfies: * `this == (this ~/ other) * other + r`. * As a consequence the remainder `r` has the same sign as the divider `this`. */ _BigIntImpl remainder(BigInt bigInt) { _BigIntImpl other = bigInt; if (other._used == 0) { throw const IntegerDivisionByZeroException(); } return _rem(other); } /// Division operator. double operator /(BigInt other) => this.toDouble() / other.toDouble(); /** Relational less than operator. */ bool operator <(BigInt other) => compareTo(other) < 0; /** Relational less than or equal operator. */ bool operator <=(BigInt other) => compareTo(other) <= 0; /** Relational greater than operator. */ bool operator >(BigInt other) => compareTo(other) > 0; /** Relational greater than or equal operator. */ bool operator >=(BigInt other) => compareTo(other) >= 0; /** * Euclidean modulo operator. * * Returns the remainder of the Euclidean division. The Euclidean division of * two integers `a` and `b` yields two integers `q` and `r` such that * `a == b * q + r` and `0 <= r < b.abs()`. * * The sign of the returned value `r` is always positive. * * See [remainder] for the remainder of the truncating division. */ _BigIntImpl operator %(BigInt bigInt) { _BigIntImpl other = bigInt; if (other._used == 0) { throw const IntegerDivisionByZeroException(); } var result = _rem(other); if (result._isNegative) { if (other._isNegative) { result = result - other; } else { result = result + other; } } return result; } /** * Returns the sign of this big integer. * * Returns 0 for zero, -1 for values less than zero and * +1 for values greater than zero. */ int get sign { if (_used == 0) return 0; return _isNegative ? -1 : 1; } /// Whether this big integer is even. bool get isEven => _used == 0 || (_digits[0] & 1) == 0; /// Whether this big integer is odd. bool get isOdd => !isEven; /// Whether this number is negative. bool get isNegative => _isNegative; _BigIntImpl pow(int exponent) { if (exponent < 0) { throw new ArgumentError("Exponent must not be negative: $exponent"); } if (exponent == 0) return one; // Exponentiation by squaring. var result = one; var base = this; while (exponent != 0) { if ((exponent & 1) == 1) { result *= base; } exponent >>= 1; // Skip unnecessary operation. if (exponent != 0) { base *= base; } } return result; } /** * Returns this integer to the power of [exponent] modulo [modulus]. * * The [exponent] must be non-negative and [modulus] must be * positive. */ _BigIntImpl modPow(BigInt bigExponent, BigInt bigModulus) { _BigIntImpl exponent = bigExponent; _BigIntImpl modulus = bigModulus; if (exponent._isNegative) { throw new ArgumentError("exponent must be positive: $exponent"); } if (modulus <= zero) { throw new ArgumentError("modulus must be strictly positive: $modulus"); } if (exponent._isZero) return one; final exponentBitlen = exponent.bitLength; if (exponentBitlen <= 0) return one; final bool cannotUseMontgomery = modulus.isEven || abs() >= modulus; if (cannotUseMontgomery || exponentBitlen < 64) { _BigIntReduction z = (cannotUseMontgomery || exponentBitlen < 8) ? new _BigIntClassicReduction(modulus) : new _BigIntMontgomeryReduction(modulus); var resultDigits = _newDigits(2 * z._normModulusUsed + 2); var result2Digits = _newDigits(2 * z._normModulusUsed + 2); var gDigits = _newDigits(z._normModulusUsed); var gUsed = z._convert(this, gDigits); // Initialize result with g. // Copy leading zero if any. for (int j = gUsed + (gUsed & 1) - 1; j >= 0; j--) { resultDigits[j] = gDigits[j]; } var resultUsed = gUsed; var result2Used; for (int i = exponentBitlen - 2; i >= 0; i--) { result2Used = z._sqr(resultDigits, resultUsed, result2Digits); if (exponent._digits[i ~/ _digitBits] & (1 << (i % _digitBits)) != 0) { resultUsed = z._mul(result2Digits, result2Used, gDigits, gUsed, resultDigits); } else { // Swap result and result2. var tmpDigits = resultDigits; var tmpUsed = resultUsed; resultDigits = result2Digits; resultUsed = result2Used; result2Digits = tmpDigits; result2Used = tmpUsed; } } return z._revert(resultDigits, resultUsed); } var k; if (exponentBitlen < 18) k = 1; else if (exponentBitlen < 48) k = 3; else if (exponentBitlen < 144) k = 4; else if (exponentBitlen < 768) k = 5; else k = 6; _BigIntReduction z = new _BigIntMontgomeryReduction(modulus); var n = 3; final k1 = k - 1; final km = (1 << k) - 1; List gDigits = new List(km + 1); List gUsed = new List(km + 1); gDigits[1] = _newDigits(z._normModulusUsed); gUsed[1] = z._convert(this, gDigits[1]); if (k > 1) { var g2Digits = _newDigits(2 * z._normModulusUsed + 2); var g2Used = z._sqr(gDigits[1], gUsed[1], g2Digits); while (n <= km) { gDigits[n] = _newDigits(2 * z._normModulusUsed + 2); gUsed[n] = z._mul(g2Digits, g2Used, gDigits[n - 2], gUsed[n - 2], gDigits[n]); n += 2; } } var w; var isOne = true; var resultDigits = one._digits; var resultUsed = one._used; var result2Digits = _newDigits(2 * z._normModulusUsed + 2); var result2Used; var exponentDigits = exponent._digits; var j = exponent._used - 1; var i = exponentDigits[j].bitLength - 1; while (j >= 0) { if (i >= k1) { w = (exponentDigits[j] >> (i - k1)) & km; } else { w = (exponentDigits[j] & ((1 << (i + 1)) - 1)) << (k1 - i); if (j > 0) { w |= exponentDigits[j - 1] >> (_digitBits + i - k1); } } n = k; while ((w & 1) == 0) { w >>= 1; --n; } if ((i -= n) < 0) { i += _digitBits; --j; } if (isOne) { // r == 1, don't bother squaring or multiplying it. resultDigits = _newDigits(2 * z._normModulusUsed + 2); resultUsed = gUsed[w]; var gwDigits = gDigits[w]; var ri = resultUsed + (resultUsed & 1); // Copy leading zero if any. while (--ri >= 0) { resultDigits[ri] = gwDigits[ri]; } isOne = false; } else { while (n > 1) { result2Used = z._sqr(resultDigits, resultUsed, result2Digits); resultUsed = z._sqr(result2Digits, result2Used, resultDigits); n -= 2; } if (n > 0) { result2Used = z._sqr(resultDigits, resultUsed, result2Digits); } else { var swapDigits = resultDigits; var swapUsed = resultUsed; resultDigits = result2Digits; resultUsed = result2Used; result2Digits = swapDigits; result2Used = swapUsed; } resultUsed = z._mul( result2Digits, result2Used, gDigits[w], gUsed[w], resultDigits); } while (j >= 0 && (exponentDigits[j] & (1 << i)) == 0) { result2Used = z._sqr(resultDigits, resultUsed, result2Digits); var swapDigits = resultDigits; var swapUsed = resultUsed; resultDigits = result2Digits; resultUsed = result2Used; result2Digits = swapDigits; result2Used = swapUsed; if (--i < 0) { i = _digitBits - 1; --j; } } } assert(!isOne); return z._revert(resultDigits, resultUsed); } // If inv is false, returns gcd(x, y). // If inv is true and gcd(x, y) = 1, returns d, so that c*x + d*y = 1. // If inv is true and gcd(x, y) != 1, throws Exception("Not coprime"). static _BigIntImpl _binaryGcd(_BigIntImpl x, _BigIntImpl y, bool inv) { var xDigits = x._digits; var yDigits = y._digits; var xUsed = x._used; var yUsed = y._used; var maxUsed = _max(xUsed, yUsed); final maxLen = maxUsed + (maxUsed & 1); xDigits = _cloneDigits(xDigits, 0, xUsed, maxLen); yDigits = _cloneDigits(yDigits, 0, yUsed, maxLen); int shiftAmount = 0; if (inv) { if ((yUsed == 1) && (yDigits[0] == 1)) return one; if ((yUsed == 0) || (yDigits[0].isEven && xDigits[0].isEven)) { throw new Exception("Not coprime"); } } else { if (x._isZero) { throw new ArgumentError.value(0, "this", "must not be zero"); } if (y._isZero) { throw new ArgumentError.value(0, "other", "must not be zero"); } if (((xUsed == 1) && (xDigits[0] == 1)) || ((yUsed == 1) && (yDigits[0] == 1))) return one; while (((xDigits[0] & 1) == 0) && ((yDigits[0] & 1) == 0)) { _rsh(xDigits, xUsed, 1, xDigits); _rsh(yDigits, yUsed, 1, yDigits); shiftAmount++; } if (shiftAmount >= _digitBits) { var digitShiftAmount = shiftAmount ~/ _digitBits; xUsed -= digitShiftAmount; yUsed -= digitShiftAmount; maxUsed -= digitShiftAmount; } if ((yDigits[0] & 1) == 1) { // Swap x and y. var tmpDigits = xDigits; var tmpUsed = xUsed; xDigits = yDigits; xUsed = yUsed; yDigits = tmpDigits; yUsed = tmpUsed; } } var uDigits = _cloneDigits(xDigits, 0, xUsed, maxLen); var vDigits = _cloneDigits(yDigits, 0, yUsed, maxLen + 2); // +2 for lsh. final bool ac = (xDigits[0] & 1) == 0; // Variables a, b, c, and d require one more digit. final abcdUsed = maxUsed + 1; final abcdLen = abcdUsed + (abcdUsed & 1) + 2; // +2 to satisfy _absAdd. var aDigits, bDigits, cDigits, dDigits; bool aIsNegative, bIsNegative, cIsNegative, dIsNegative; if (ac) { aDigits = _newDigits(abcdLen); aIsNegative = false; aDigits[0] = 1; cDigits = _newDigits(abcdLen); cIsNegative = false; } bDigits = _newDigits(abcdLen); bIsNegative = false; dDigits = _newDigits(abcdLen); dIsNegative = false; dDigits[0] = 1; while (true) { while ((uDigits[0] & 1) == 0) { _rsh(uDigits, maxUsed, 1, uDigits); if (ac) { if (((aDigits[0] & 1) == 1) || ((bDigits[0] & 1) == 1)) { if (aIsNegative) { if ((aDigits[maxUsed] != 0) || (_compareDigits(aDigits, maxUsed, yDigits, maxUsed)) > 0) { _absSub(aDigits, abcdUsed, yDigits, maxUsed, aDigits); } else { _absSub(yDigits, maxUsed, aDigits, maxUsed, aDigits); aIsNegative = false; } } else { _absAdd(aDigits, abcdUsed, yDigits, maxUsed, aDigits); } if (bIsNegative) { _absAdd(bDigits, abcdUsed, xDigits, maxUsed, bDigits); } else if ((bDigits[maxUsed] != 0) || (_compareDigits(bDigits, maxUsed, xDigits, maxUsed) > 0)) { _absSub(bDigits, abcdUsed, xDigits, maxUsed, bDigits); } else { _absSub(xDigits, maxUsed, bDigits, maxUsed, bDigits); bIsNegative = true; } } _rsh(aDigits, abcdUsed, 1, aDigits); } else if ((bDigits[0] & 1) == 1) { if (bIsNegative) { _absAdd(bDigits, abcdUsed, xDigits, maxUsed, bDigits); } else if ((bDigits[maxUsed] != 0) || (_compareDigits(bDigits, maxUsed, xDigits, maxUsed) > 0)) { _absSub(bDigits, abcdUsed, xDigits, maxUsed, bDigits); } else { _absSub(xDigits, maxUsed, bDigits, maxUsed, bDigits); bIsNegative = true; } } _rsh(bDigits, abcdUsed, 1, bDigits); } while ((vDigits[0] & 1) == 0) { _rsh(vDigits, maxUsed, 1, vDigits); if (ac) { if (((cDigits[0] & 1) == 1) || ((dDigits[0] & 1) == 1)) { if (cIsNegative) { if ((cDigits[maxUsed] != 0) || (_compareDigits(cDigits, maxUsed, yDigits, maxUsed) > 0)) { _absSub(cDigits, abcdUsed, yDigits, maxUsed, cDigits); } else { _absSub(yDigits, maxUsed, cDigits, maxUsed, cDigits); cIsNegative = false; } } else { _absAdd(cDigits, abcdUsed, yDigits, maxUsed, cDigits); } if (dIsNegative) { _absAdd(dDigits, abcdUsed, xDigits, maxUsed, dDigits); } else if ((dDigits[maxUsed] != 0) || (_compareDigits(dDigits, maxUsed, xDigits, maxUsed) > 0)) { _absSub(dDigits, abcdUsed, xDigits, maxUsed, dDigits); } else { _absSub(xDigits, maxUsed, dDigits, maxUsed, dDigits); dIsNegative = true; } } _rsh(cDigits, abcdUsed, 1, cDigits); } else if ((dDigits[0] & 1) == 1) { if (dIsNegative) { _absAdd(dDigits, abcdUsed, xDigits, maxUsed, dDigits); } else if ((dDigits[maxUsed] != 0) || (_compareDigits(dDigits, maxUsed, xDigits, maxUsed) > 0)) { _absSub(dDigits, abcdUsed, xDigits, maxUsed, dDigits); } else { _absSub(xDigits, maxUsed, dDigits, maxUsed, dDigits); dIsNegative = true; } } _rsh(dDigits, abcdUsed, 1, dDigits); } if (_compareDigits(uDigits, maxUsed, vDigits, maxUsed) >= 0) { _absSub(uDigits, maxUsed, vDigits, maxUsed, uDigits); if (ac) { if (aIsNegative == cIsNegative) { var a_cmp_c = _compareDigits(aDigits, abcdUsed, cDigits, abcdUsed); if (a_cmp_c > 0) { _absSub(aDigits, abcdUsed, cDigits, abcdUsed, aDigits); } else { _absSub(cDigits, abcdUsed, aDigits, abcdUsed, aDigits); aIsNegative = !aIsNegative && (a_cmp_c != 0); } } else { _absAdd(aDigits, abcdUsed, cDigits, abcdUsed, aDigits); } } if (bIsNegative == dIsNegative) { var b_cmp_d = _compareDigits(bDigits, abcdUsed, dDigits, abcdUsed); if (b_cmp_d > 0) { _absSub(bDigits, abcdUsed, dDigits, abcdUsed, bDigits); } else { _absSub(dDigits, abcdUsed, bDigits, abcdUsed, bDigits); bIsNegative = !bIsNegative && (b_cmp_d != 0); } } else { _absAdd(bDigits, abcdUsed, dDigits, abcdUsed, bDigits); } } else { _absSub(vDigits, maxUsed, uDigits, maxUsed, vDigits); if (ac) { if (cIsNegative == aIsNegative) { var c_cmp_a = _compareDigits(cDigits, abcdUsed, aDigits, abcdUsed); if (c_cmp_a > 0) { _absSub(cDigits, abcdUsed, aDigits, abcdUsed, cDigits); } else { _absSub(aDigits, abcdUsed, cDigits, abcdUsed, cDigits); cIsNegative = !cIsNegative && (c_cmp_a != 0); } } else { _absAdd(cDigits, abcdUsed, aDigits, abcdUsed, cDigits); } } if (dIsNegative == bIsNegative) { var d_cmp_b = _compareDigits(dDigits, abcdUsed, bDigits, abcdUsed); if (d_cmp_b > 0) { _absSub(dDigits, abcdUsed, bDigits, abcdUsed, dDigits); } else { _absSub(bDigits, abcdUsed, dDigits, abcdUsed, dDigits); dIsNegative = !dIsNegative && (d_cmp_b != 0); } } else { _absAdd(dDigits, abcdUsed, bDigits, abcdUsed, dDigits); } } // Exit loop if u == 0. var i = maxUsed; while ((i > 0) && (uDigits[i - 1] == 0)) --i; if (i == 0) break; } if (!inv) { if (shiftAmount > 0) { maxUsed = _lShiftDigits(vDigits, maxUsed, shiftAmount, vDigits); } return new _BigIntImpl._(false, maxUsed, vDigits); } // No inverse if v != 1. var i = maxUsed - 1; while ((i > 0) && (vDigits[i] == 0)) --i; if ((i != 0) || (vDigits[0] != 1)) { throw new Exception("Not coprime"); } if (dIsNegative) { if ((dDigits[maxUsed] != 0) || (_compareDigits(dDigits, maxUsed, xDigits, maxUsed) > 0)) { _absSub(dDigits, abcdUsed, xDigits, maxUsed, dDigits); if ((dDigits[maxUsed] != 0) || (_compareDigits(dDigits, maxUsed, xDigits, maxUsed) > 0)) { _absSub(dDigits, abcdUsed, xDigits, maxUsed, dDigits); } else { _absSub(xDigits, maxUsed, dDigits, maxUsed, dDigits); dIsNegative = false; } } else { _absSub(xDigits, maxUsed, dDigits, maxUsed, dDigits); dIsNegative = false; } } else if ((dDigits[maxUsed] != 0) || (_compareDigits(dDigits, maxUsed, xDigits, maxUsed) > 0)) { _absSub(dDigits, abcdUsed, xDigits, maxUsed, dDigits); if ((dDigits[maxUsed] != 0) || (_compareDigits(dDigits, maxUsed, xDigits, maxUsed) > 0)) { _absSub(dDigits, abcdUsed, xDigits, maxUsed, dDigits); } } return new _BigIntImpl._(false, maxUsed, dDigits); } /** * Returns the modular multiplicative inverse of this big integer * modulo [modulus]. * * The [modulus] must be positive. * * It is an error if no modular inverse exists. */ // Returns 1/this % modulus, with modulus > 0. _BigIntImpl modInverse(BigInt bigInt) { _BigIntImpl modulus = bigInt; if (modulus <= zero) { throw new ArgumentError("Modulus must be strictly positive: $modulus"); } if (modulus == one) return zero; var tmp = this; if (tmp._isNegative || (tmp._absCompare(modulus) >= 0)) { tmp %= modulus; } return _binaryGcd(modulus, tmp, true); } /** * Returns the greatest common divisor of this big integer and [other]. * * If either number is non-zero, the result is the numerically greatest * integer dividing both `this` and `other`. * * The greatest common divisor is independent of the order, * so `x.gcd(y)` is always the same as `y.gcd(x)`. * * For any integer `x`, `x.gcd(x)` is `x.abs()`. * * If both `this` and `other` is zero, the result is also zero. */ _BigIntImpl gcd(BigInt bigInt) { _BigIntImpl other = bigInt; if (_isZero) return other.abs(); if (other._isZero) return this.abs(); return _binaryGcd(this, other, false); } /** * Returns the least significant [width] bits of this big integer as a * non-negative number (i.e. unsigned representation). The returned value has * zeros in all bit positions higher than [width]. * * ``` * new BigInt.from(-1).toUnsigned(5) == 31 // 11111111 -> 00011111 * ``` * * This operation can be used to simulate arithmetic from low level languages. * For example, to increment an 8 bit quantity: * * ``` * q = (q + 1).toUnsigned(8); * ``` * * `q` will count from `0` up to `255` and then wrap around to `0`. * * If the input fits in [width] bits without truncation, the result is the * same as the input. The minimum width needed to avoid truncation of `x` is * given by `x.bitLength`, i.e. * * ``` * x == x.toUnsigned(x.bitLength); * ``` */ _BigIntImpl toUnsigned(int width) { return this & ((one << width) - one); } /** * Returns the least significant [width] bits of this integer, extending the * highest retained bit to the sign. This is the same as truncating the value * to fit in [width] bits using an signed 2-s complement representation. The * returned value has the same bit value in all positions higher than [width]. * * ``` * var big15 = new BigInt.from(15); * var big16 = new BigInt.from(16); * var big239 = new BigInt.from(239); * V--sign bit-V * big16.toSigned(5) == -big16 // 00010000 -> 11110000 * big239.toSigned(5) == big15 // 11101111 -> 00001111 * ^ ^ * ``` * * This operation can be used to simulate arithmetic from low level languages. * For example, to increment an 8 bit signed quantity: * * ``` * q = (q + 1).toSigned(8); * ``` * * `q` will count from `0` up to `127`, wrap to `-128` and count back up to * `127`. * * If the input value fits in [width] bits without truncation, the result is * the same as the input. The minimum width needed to avoid truncation of `x` * is `x.bitLength + 1`, i.e. * * ``` * x == x.toSigned(x.bitLength + 1); * ``` */ _BigIntImpl toSigned(int width) { // The value of binary number weights each bit by a power of two. The // twos-complement value weights the sign bit negatively. We compute the // value of the negative weighting by isolating the sign bit with the // correct power of two weighting and subtracting it from the value of the // lower bits. var signMask = one << (width - 1); return (this & (signMask - one)) - (this & signMask); } bool get isValidInt { assert(_digitBits == 32); return _used < 2 || (_used == 2 && (_digits[1] < 0x80000000 || (_isNegative && _digits[1] == 0x80000000 && _digits[0] == 0))); } int toInt() { assert(_digitBits == 32); if (_used == 0) return 0; if (_used == 1) return _isNegative ? -_digits[0] : _digits[0]; if (_used == 2 && _digits[1] < 0x80000000) { var result = (_digits[1] << _digitBits) | _digits[0]; return _isNegative ? -result : result; } return _isNegative ? _minInt : _maxInt; } /** * Returns this [_BigIntImpl] as a [double]. * * If the number is not representable as a [double], an * approximation is returned. For numerically large integers, the * approximation may be infinite. */ double toDouble() { const int exponentBias = 1075; // There are 11 bits for the exponent. // 2047 (all bits set to 1) is reserved for infinity and NaN. // When storing the exponent in the 11 bits, it is biased by exponentBias // to support negative exponents. const int maxDoubleExponent = 2046 - exponentBias; if (_isZero) return 0.0; // We fill the 53 bits little-endian. var resultBits = new Uint8List(8); var length = _digitBits * (_used - 1) + _digits[_used - 1].bitLength; if (length - 53 > maxDoubleExponent) return double.infinity; // The most significant bit is for the sign. if (_isNegative) resultBits[7] = 0x80; // Write the exponent into bits 1..12: var biasedExponent = length - 53 + exponentBias; resultBits[6] = (biasedExponent & 0xF) << 4; resultBits[7] |= biasedExponent >> 4; int cachedBits = 0; int cachedBitsLength = 0; int digitIndex = _used - 1; int readBits(int n) { // Ensure that we have enough bits in [cachedBits]. while (cachedBitsLength < n) { int nextDigit; int nextDigitLength = _digitBits; // May get updated. if (digitIndex < 0) { nextDigit = 0; digitIndex--; } else { nextDigit = _digits[digitIndex]; if (digitIndex == _used - 1) nextDigitLength = nextDigit.bitLength; digitIndex--; } cachedBits = (cachedBits << nextDigitLength) + nextDigit; cachedBitsLength += nextDigitLength; } // Read the top [n] bits. var result = cachedBits >> (cachedBitsLength - n); // Remove the bits from the cache. cachedBits -= result << (cachedBitsLength - n); cachedBitsLength -= n; return result; } // The first leading 1 bit is implicit in the double-representation and can // be discarded. var leadingBits = readBits(5) & 0xF; resultBits[6] |= leadingBits; for (int i = 5; i >= 0; i--) { // Get the remaining 48 bits. resultBits[i] = readBits(8); } void roundUp() { // Simply consists of adding 1 to the whole 64 bit "number". // It will update the exponent, if necessary. // It might even round up to infinity (which is what we want). var carry = 1; for (int i = 0; i < 8; i++) { if (carry == 0) break; var sum = resultBits[i] + carry; resultBits[i] = sum & 0xFF; carry = sum >> 8; } } if (readBits(1) == 1) { if (resultBits[0].isOdd) { // Rounds to even all the time. roundUp(); } else { // Round up, if there is at least one other digit that is not 0. if (cachedBits != 0) { // There is already one in the cachedBits. roundUp(); } else { for (int i = digitIndex; digitIndex >= 0; i--) { if (_digits[i] != 0) { roundUp(); break; } } } } } return resultBits.buffer.asByteData().getFloat64(0, Endian.little); } /** * Returns a String-representation of this integer. * * The returned string is parsable by [parse]. * For any `_BigIntImpl` `i`, it is guaranteed that * `i == _BigIntImpl.parse(i.toString())`. */ String toString() { if (_used == 0) return "0"; if (_used == 1) { if (_isNegative) return (-_digits[0]).toString(); return _digits[0].toString(); } // Generate in chunks of 9 digits. // The chunks are in reversed order. var decimalDigitChunks = []; var rest = isNegative ? -this : this; while (rest._used > 1) { var digits9 = rest.remainder(_oneBillion).toString(); decimalDigitChunks.add(digits9); var zeros = 9 - digits9.length; if (zeros == 8) { decimalDigitChunks.add("00000000"); } else { if (zeros >= 4) { zeros -= 4; decimalDigitChunks.add("0000"); } if (zeros >= 2) { zeros -= 2; decimalDigitChunks.add("00"); } if (zeros >= 1) { decimalDigitChunks.add("0"); } } rest = rest ~/ _oneBillion; } decimalDigitChunks.add(rest._digits[0].toString()); if (_isNegative) decimalDigitChunks.add("-"); return decimalDigitChunks.reversed.join(); } int _toRadixCodeUnit(int digit) { const int _0 = 48; const int _a = 97; if (digit < 10) return _0 + digit; return _a + digit - 10; } /** * Converts [this] to a string representation in the given [radix]. * * In the string representation, lower-case letters are used for digits above * '9', with 'a' being 10 an 'z' being 35. * * The [radix] argument must be an integer in the range 2 to 36. */ String toRadixString(int radix) { if (radix > 36) throw new RangeError.range(radix, 2, 36); if (_used == 0) return "0"; if (_used == 1) { var digitString = _digits[0].toRadixString(radix); if (_isNegative) return "-" + digitString; return digitString; } if (radix == 16) return _toHexString(); var base = new _BigIntImpl._fromInt(radix); var reversedDigitCodeUnits = []; var rest = this.abs(); while (!rest._isZero) { var digit = rest.remainder(base).toInt(); rest = rest ~/ base; reversedDigitCodeUnits.add(_toRadixCodeUnit(digit)); } var digitString = new String.fromCharCodes(reversedDigitCodeUnits.reversed); if (_isNegative) return "-" + digitString; return digitString; } String _toHexString() { var chars = []; for (int i = 0; i < _used - 1; i++) { int chunk = _digits[i]; for (int j = 0; j < (_digitBits ~/ 4); j++) { chars.add(_toRadixCodeUnit(chunk & 0xF)); chunk >>= 4; } } var msbChunk = _digits[_used - 1]; while (msbChunk != 0) { chars.add(_toRadixCodeUnit(msbChunk & 0xF)); msbChunk >>= 4; } if (_isNegative) { const _dash = 45; chars.add(_dash); } return new String.fromCharCodes(chars.reversed); } } // Interface for modular reduction. abstract class _BigIntReduction { int get _normModulusUsed; // Return the number of digits used by resultDigits. int _convert(_BigIntImpl x, Uint32List resultDigits); int _mul(Uint32List xDigits, int xUsed, Uint32List yDigits, int yUsed, Uint32List resultDigits); int _sqr(Uint32List xDigits, int xUsed, Uint32List resultDigits); // Return x reverted to _BigIntImpl. _BigIntImpl _revert(Uint32List xDigits, int xUsed); } // Montgomery reduction on _BigIntImpl. class _BigIntMontgomeryReduction implements _BigIntReduction { final _BigIntImpl _modulus; int _normModulusUsed; // Even if processing 64-bit (digit pairs). Uint32List _modulusDigits; Uint32List _args; int _digitsPerStep; // Number of digits processed in one step. 1 or 2. static const int _xDigit = 0; // Index of digit of x. static const int _xHighDigit = 1; // Index of high digit of x (64-bit only). static const int _rhoDigit = 2; // Index of digit of rho. static const int _rhoHighDigit = 3; // Index of high digit of rho (64-bit). static const int _muDigit = 4; // Index of mu. static const int _muHighDigit = 5; // Index of high 32-bits of mu (64-bit). _BigIntMontgomeryReduction(this._modulus) { _modulusDigits = _modulus._digits; _args = _newDigits(6); // Determine if we can process digit pairs by calling an intrinsic. _digitsPerStep = _mulMod(_args, _args, 0); _args[_xDigit] = _modulusDigits[0]; if (_digitsPerStep == 1) { _normModulusUsed = _modulus._used; _invDigit(_args); } else { assert(_digitsPerStep == 2); _normModulusUsed = _modulus._used + (_modulus._used & 1); _args[_xHighDigit] = _modulusDigits[1]; _invDigitPair(_args); } } // Calculates -1/x % _digitBase, x is 32-bit digit. // xy == 1 (mod m) // xy = 1+km // xy(2-xy) = (1+km)(1-km) // x(y(2-xy)) = 1-k^2 m^2 // x(y(2-xy)) == 1 (mod m^2) // if y is 1/x mod m, then y(2-xy) is 1/x mod m^2 // Should reduce x and y(2-xy) by m^2 at each step to keep size bounded. // // Operation: // args[_rhoDigit] = 1/args[_xDigit] mod _digitBase. static void _invDigit(Uint32List args) { var x = args[_xDigit]; var y = x & 3; // y == 1/x mod 2^2 y = (y * (2 - (x & 0xf) * y)) & 0xf; // y == 1/x mod 2^4 y = (y * (2 - (x & 0xff) * y)) & 0xff; // y == 1/x mod 2^8 y = (y * (2 - (((x & 0xffff) * y) & 0xffff))) & 0xffff; // y == 1/x mod 2^16 y = (y * (2 - x * y % _BigIntImpl._digitBase)) % _BigIntImpl._digitBase; // y == 1/x mod _digitBase y = -y; // We really want the negative inverse. args[_rhoDigit] = y & _BigIntImpl._digitMask; assert(((x * y) & _BigIntImpl._digitMask) == _BigIntImpl._digitMask); } // Calculates -1/x % _digitBase^2, x is a pair of 32-bit digits. // Operation: // args[_rhoDigit.._rhoHighDigit] = // 1/args[_xDigit.._xHighDigit] mod _digitBase^2. static void _invDigitPair(Uint32List args) { var two = _BigIntImpl.two; var mask32 = _BigIntImpl._oneDigitMask; var mask64 = _BigIntImpl._twoDigitMask; var xl = args[_xDigit]; // Lower 32-bit digit of x. var y = xl & 3; // y == 1/x mod 2^2 y = (y * (2 - (xl & 0xf) * y)) & 0xf; // y == 1/x mod 2^4 y = (y * (2 - (xl & 0xff) * y)) & 0xff; // y == 1/x mod 2^8 y = (y * (2 - (((xl & 0xffff) * y) & 0xffff))) & 0xffff; // y == 1/x mod 2^16 y = (y * (2 - ((xl * y) & 0xffffffff))) & 0xffffffff; // y == 1/x mod 2^32 var x = (args[_xHighDigit] << _BigIntImpl._digitBits) | xl; y *= 2 - x * y; // Masking with 2^64-1 is implied by 64-bit arithmetic. // y == 1/x mod _digitBase^2 y = -y; // We really want the negative inverse. args[_rhoDigit] = y & _BigIntImpl._digitMask; args[_rhoHighDigit] = (y >> _BigIntImpl._digitBits) & _BigIntImpl._digitMask; assert(x * y == -1); } // Operation: // args[_muDigit] = args[_rhoDigit]*digits[i] mod _digitBase. // Returns 1. // Note: Intrinsics on 64-bit platforms process digit pairs at even indices: // args[_muDigit.._muHighDigit] = // args[_rhoDigit.._rhoHighDigit] * digits[i..i+1] mod _digitBase^2. // Returns 2. static int _mulMod(Uint32List args, Uint32List digits, int i) { var rhol = args[_rhoDigit] & _BigIntImpl._halfDigitMask; var rhoh = args[_rhoDigit] >> _BigIntImpl._halfDigitBits; var dh = digits[i] >> _BigIntImpl._halfDigitBits; var dl = digits[i] & _BigIntImpl._halfDigitMask; args[_muDigit] = (dl * rhol + (((dl * rhoh + dh * rhol) & _BigIntImpl._halfDigitMask) << _BigIntImpl._halfDigitBits)) & _BigIntImpl._digitMask; return 1; } // result = x*R mod _modulus. // Returns resultUsed. int _convert(_BigIntImpl x, Uint32List resultDigits) { // Montgomery reduction only works if abs(x) < _modulus. assert(x.abs() < _modulus); assert(_digitsPerStep == 1 || _normModulusUsed.isEven); var result = x.abs()._dlShift(_normModulusUsed)._rem(_modulus); if (x._isNegative && !result._isNegative && result._used > 0) { result = _modulus - result; } var used = result._used; var digits = result._digits; var i = used + (used & 1); while (--i >= 0) { resultDigits[i] = digits[i]; } return used; } _BigIntImpl _revert(Uint32List xDigits, int xUsed) { // Reserve enough digits for modulus squaring and accumulator carry. var resultDigits = _newDigits(2 * _normModulusUsed + 2); var i = xUsed + (xUsed & 1); while (--i >= 0) { resultDigits[i] = xDigits[i]; } var resultUsed = _reduce(resultDigits, xUsed); return new _BigIntImpl._(false, resultUsed, resultDigits); } // x = x/R mod _modulus. // Returns xUsed. int _reduce(Uint32List xDigits, int xUsed) { while (xUsed < 2 * _normModulusUsed + 2) { // Pad x so _mulAdd has enough room later for a possible carry. xDigits[xUsed++] = 0; } var i = 0; while (i < _normModulusUsed) { var d = _mulMod(_args, xDigits, i); assert(d == _digitsPerStep); d = _BigIntImpl._mulAdd( _args, _muDigit, _modulusDigits, 0, xDigits, i, _normModulusUsed); assert(d == _digitsPerStep); i += d; } // Clamp x. while (xUsed > 0 && xDigits[xUsed - 1] == 0) { --xUsed; } xUsed = _BigIntImpl._drShiftDigits(xDigits, xUsed, i, xDigits); if (_BigIntImpl._compareDigits( xDigits, xUsed, _modulusDigits, _normModulusUsed) >= 0) { _BigIntImpl._absSub( xDigits, xUsed, _modulusDigits, _normModulusUsed, xDigits); } // Clamp x. while (xUsed > 0 && xDigits[xUsed - 1] == 0) { --xUsed; } return xUsed; } int _sqr(Uint32List xDigits, int xUsed, Uint32List resultDigits) { var resultUsed = _BigIntImpl._sqrDigits(xDigits, xUsed, resultDigits); return _reduce(resultDigits, resultUsed); } int _mul(Uint32List xDigits, int xUsed, Uint32List yDigits, int yUsed, Uint32List resultDigits) { var resultUsed = _BigIntImpl._mulDigits(xDigits, xUsed, yDigits, yUsed, resultDigits); return _reduce(resultDigits, resultUsed); } } // Modular reduction using "classic" algorithm. class _BigIntClassicReduction implements _BigIntReduction { final _BigIntImpl _modulus; // Modulus. int _normModulusUsed; _BigIntImpl _normModulus; // Normalized _modulus. Uint32List _normModulusDigits; Uint32List _negNormModulusDigits; // Negated _normModulus digits. int _modulusNsh; // Normalization shift amount. Uint32List _args; // Top _normModulus digit(s) and place holder for estimated // quotient digit(s). Uint32List _tmpDigits; // Temporary digits used during reduction. _BigIntClassicReduction(this._modulus) { // Preprocess arguments to _remDigits. var nsh = _BigIntImpl._digitBits - _modulus._digits[_modulus._used - 1].bitLength; // For 64-bit processing, make sure _negNormModulusDigits has an even number // of digits. if (_modulus._used.isOdd) { nsh += _BigIntImpl._digitBits; } _modulusNsh = nsh; _normModulus = _modulus << nsh; _normModulusUsed = _normModulus._used; _normModulusDigits = _normModulus._digits; assert(_normModulusUsed.isEven); _args = _newDigits(4); _args[_BigIntImpl._divisorLowTopDigit] = _normModulusDigits[_normModulusUsed - 2]; _args[_BigIntImpl._divisorTopDigit] = _normModulusDigits[_normModulusUsed - 1]; // Negate _normModulus so we can use _mulAdd instead of // unimplemented _mulSub. var negNormModulus = _BigIntImpl.one._dlShift(_normModulusUsed) - _normModulus; if (negNormModulus._used < _normModulusUsed) { _negNormModulusDigits = _BigIntImpl._cloneDigits( negNormModulus._digits, 0, _normModulusUsed, _normModulusUsed); } else { _negNormModulusDigits = negNormModulus._digits; } // _negNormModulusDigits is read-only and has _normModulusUsed digits (possibly // including several leading zeros) plus a leading zero for 64-bit // processing. _tmpDigits = _newDigits(2 * _normModulusUsed); } int _convert(_BigIntImpl x, Uint32List resultDigits) { var digits; var used; if (x._isNegative || x.compareTo(_modulus) >= 0) { var remainder = x._rem(_modulus); if (x._isNegative && !remainder._isNegative && remainder._used > 0) { remainder = _modulus - remainder; } assert(!remainder._isNegative); used = remainder._used; digits = remainder._digits; } else { used = x._used; digits = x._digits; } var i = used + (used & 1); // Copy leading zero if any. while (--i >= 0) { resultDigits[i] = digits[i]; } return used; } _BigIntImpl _revert(Uint32List xDigits, int xUsed) { return new _BigIntImpl._(false, xUsed, xDigits); } int _reduce(Uint32List xDigits, int xUsed) { if (xUsed < _modulus._used) { return xUsed; } // The function _BigIntImpl._remDigits(...) is optimized for reduction and // equivalent to calling // 'convert(revert(xDigits, xUsed)._rem(_normModulus), xDigits);' return _BigIntImpl._remDigits( xDigits, xUsed, _normModulusDigits, _normModulusUsed, _negNormModulusDigits, _modulusNsh, _args, _tmpDigits, xDigits); } int _sqr(Uint32List xDigits, int xUsed, Uint32List resultDigits) { var resultUsed = _BigIntImpl._sqrDigits(xDigits, xUsed, resultDigits); return _reduce(resultDigits, resultUsed); } int _mul(Uint32List xDigits, int xUsed, Uint32List yDigits, int yUsed, Uint32List resultDigits) { var resultUsed = _BigIntImpl._mulDigits(xDigits, xUsed, yDigits, yUsed, resultDigits); return _reduce(resultDigits, resultUsed); } }