/* * BigInt Base * (C) 1999-2011 Jack Lloyd * * Distributed under the terms of the Botan license */ #include #include #include #include #include namespace Botan { /* * Construct a BigInt from a regular number */ BigInt::BigInt(u64bit n) { set_sign(Positive); if(n == 0) return; const size_t limbs_needed = sizeof(u64bit) / sizeof(word); reg.resize(4*limbs_needed); for(size_t i = 0; i != limbs_needed; ++i) reg[i] = ((n >> (i*MP_WORD_BITS)) & MP_WORD_MASK); } /* * Construct a BigInt of the specified size */ BigInt::BigInt(Sign s, size_t size) { reg.resize(round_up(size, 8)); signedness = s; } /* * Construct a BigInt from a "raw" BigInt */ BigInt::BigInt(const BigInt& b) { const size_t b_words = b.sig_words(); if(b_words) { reg.resize(round_up(b_words, 8)); reg.copy(b.data(), b_words); set_sign(b.sign()); } else { reg.resize(2); set_sign(Positive); } } /* * Construct a BigInt from a string */ BigInt::BigInt(const std::string& str) { Base base = Decimal; size_t markers = 0; bool negative = false; if(str.length() > 0 && str[0] == '-') { markers += 1; negative = true; } if(str.length() > markers + 2 && str[markers ] == '0' && str[markers + 1] == 'x') { markers += 2; base = Hexadecimal; } else if(str.length() > markers + 1 && str[markers] == '0') { markers += 1; base = Octal; } *this = decode(reinterpret_cast(str.data()) + markers, str.length() - markers, base); if(negative) set_sign(Negative); else set_sign(Positive); } /* * Construct a BigInt from an encoded BigInt */ BigInt::BigInt(const byte input[], size_t length, Base base) { set_sign(Positive); *this = decode(input, length, base); } /* * Construct a BigInt from an encoded BigInt */ BigInt::BigInt(RandomNumberGenerator& rng, size_t bits) { set_sign(Positive); randomize(rng, bits); } /* * Swap this BigInt with another */ void BigInt::swap(BigInt& other) { reg.swap(other.reg); std::swap(signedness, other.signedness); } /* * Grow the internal storage */ void BigInt::grow_reg(size_t n) { reg.resize(round_up(size() + n, 8)); } /* * Grow the internal storage */ void BigInt::grow_to(size_t n) { if(n > size()) reg.resize(round_up(n, 8)); } /* * Comparison Function */ s32bit BigInt::cmp(const BigInt& n, bool check_signs) const { if(check_signs) { if(n.is_positive() && this->is_negative()) return -1; if(n.is_negative() && this->is_positive()) return 1; if(n.is_negative() && this->is_negative()) return (-bigint_cmp(data(), sig_words(), n.data(), n.sig_words())); } return bigint_cmp(data(), sig_words(), n.data(), n.sig_words()); } /* * Return byte n of this number */ byte BigInt::byte_at(size_t n) const { const size_t WORD_BYTES = sizeof(word); size_t word_num = n / WORD_BYTES, byte_num = n % WORD_BYTES; if(word_num >= size()) return 0; else return get_byte(WORD_BYTES - byte_num - 1, reg[word_num]); } /* * Return bit n of this number */ bool BigInt::get_bit(size_t n) const { return ((word_at(n / MP_WORD_BITS) >> (n % MP_WORD_BITS)) & 1); } /* * Return bits {offset...offset+length} */ u32bit BigInt::get_substring(size_t offset, size_t length) const { if(length > 32) throw Invalid_Argument("BigInt::get_substring: Substring size too big"); u64bit piece = 0; for(size_t i = 0; i != 8; ++i) { const byte part = byte_at((offset / 8) + (7-i)); piece = (piece << 8) | part; } const u64bit mask = (static_cast(1) << length) - 1; const size_t shift = (offset % 8); return static_cast((piece >> shift) & mask); } /* * Convert this number to a u32bit, if possible */ u32bit BigInt::to_u32bit() const { if(is_negative()) throw Encoding_Error("BigInt::to_u32bit: Number is negative"); if(bits() >= 32) throw Encoding_Error("BigInt::to_u32bit: Number is too big to convert"); u32bit out = 0; for(u32bit j = 0; j != 4; ++j) out = (out << 8) | byte_at(3-j); return out; } /* * Set bit number n */ void BigInt::set_bit(size_t n) { const size_t which = n / MP_WORD_BITS; const word mask = static_cast(1) << (n % MP_WORD_BITS); if(which >= size()) grow_to(which + 1); reg[which] |= mask; } /* * Clear bit number n */ void BigInt::clear_bit(size_t n) { const size_t which = n / MP_WORD_BITS; const word mask = static_cast(1) << (n % MP_WORD_BITS); if(which < size()) reg[which] &= ~mask; } /* * Clear all but the lowest n bits */ void BigInt::mask_bits(size_t n) { if(n == 0) { clear(); return; } if(n >= bits()) return; const size_t top_word = n / MP_WORD_BITS; const word mask = (static_cast(1) << (n % MP_WORD_BITS)) - 1; if(top_word < size()) for(size_t i = top_word + 1; i != size(); ++i) reg[i] = 0; reg[top_word] &= mask; } /* * Count how many bytes are being used */ size_t BigInt::bytes() const { return (bits() + 7) / 8; } /* * Count how many bits are being used */ size_t BigInt::bits() const { const size_t words = sig_words(); if(words == 0) return 0; size_t full_words = words - 1, top_bits = MP_WORD_BITS; word top_word = word_at(full_words), mask = MP_WORD_TOP_BIT; while(top_bits && ((top_word & mask) == 0)) { mask >>= 1; top_bits--; } return (full_words * MP_WORD_BITS + top_bits); } /* * Calcluate the size in a certain base */ size_t BigInt::encoded_size(Base base) const { static const double LOG_2_BASE_10 = 0.30102999566; if(base == Binary) return bytes(); else if(base == Hexadecimal) return 2*bytes(); else if(base == Octal) return ((bits() + 2) / 3); else if(base == Decimal) return static_cast((bits() * LOG_2_BASE_10) + 1); else throw Invalid_Argument("Unknown base for BigInt encoding"); } /* * Set the sign */ void BigInt::set_sign(Sign s) { if(is_zero()) signedness = Positive; else signedness = s; } /* * Reverse the value of the sign flag */ void BigInt::flip_sign() { set_sign(reverse_sign()); } /* * Return the opposite value of the current sign */ BigInt::Sign BigInt::reverse_sign() const { if(sign() == Positive) return Negative; return Positive; } /* * Return the negation of this number */ BigInt BigInt::operator-() const { BigInt x = (*this); x.flip_sign(); return x; } /* * Return the absolute value of this number */ BigInt BigInt::abs() const { BigInt x = (*this); x.set_sign(Positive); return x; } /* * Encode this number into bytes */ void BigInt::binary_encode(byte output[]) const { const size_t sig_bytes = bytes(); for(size_t i = 0; i != sig_bytes; ++i) output[sig_bytes-i-1] = byte_at(i); } /* * Set this number to the value in buf */ void BigInt::binary_decode(const byte buf[], size_t length) { const size_t WORD_BYTES = sizeof(word); clear(); reg.resize(round_up((length / WORD_BYTES) + 1, 8)); for(size_t i = 0; i != length / WORD_BYTES; ++i) { const size_t top = length - WORD_BYTES*i; for(size_t j = WORD_BYTES; j > 0; --j) reg[i] = (reg[i] << 8) | buf[top - j]; } for(size_t i = 0; i != length % WORD_BYTES; ++i) reg[length / WORD_BYTES] = (reg[length / WORD_BYTES] << 8) | buf[i]; } /* * Set this number to the value in buf */ void BigInt::binary_decode(const MemoryRegion& buf) { binary_decode(buf, buf.size()); } }