LNL (short for long number library) is a single-file open source C++ library for arithmetics with arbitrary length numbers. It is designed to be simple and easy to use, taking into account possible applications in cryptography and information security. Current version of LNL only supports arbitrary length integers.

• How to install
• How to use
  • General info
  • Creating an instance
  • Assigning a value
  • I/O operators
  • Useful getters
  • Relational operations
  • Mathematical operations
  • Random numbers
• Cryptography features
• Benchmarks
  • Multiplication
  • Division
  • Exponentiation
  • General benchmark info
• Testing the library
  • Run unit tests
  • Run benchmarks
• Contributing
• Licence

The whole library is located in the LNL.h header. You can find it in the include directory in this repository. Copy the this file to your project and include it just like any other header:

#include "<YourPath>/LNL.h"

⬆️ Back to contents

This library introduces a new namespace LNL (short for long number library). Inside this namespace there is a LNL::LongInt class (arbitrary length integer class) and some functions for mathematical operations.

It is possible to create a LNL::LongInt instance in different ways: with a short, int, long, long long or with unsigned equivalents of these types. It is also possible to create an instance with a std::string. If you do not pass a value to the constructor then an instance is initialized with 0 by default.

If you try to initialize with a std::string which contains non-digit characters (except front -) then an instance will be initialized with 0 and an error message will be printed.

Here are some examples:

LNL::LongInt x;                              // x = 0 (by default)
LNL::LongInt y(-123456789);                  // y = -123456789
LNL::LongInt z("1234567898765432123456789"); // z = 1234567898765432123456789

Assignment operator = is overloaded for short, int, long, long long and unsigned equivalents of these types, as well as for std::string and LNL::LongInt itself.

If you try to assign with a std::string which contains non-digit characters (except front "-") then a 0 value will be assigned and an error message will be printed.

Here are some examples:

LNL::LongInt x, y(16); // x = 0,   y = 16
y = 32;                // x = 0,   y = 32
y = "-64";             // x = 0,   y = -64
x = y;                 // x = -64, y = -64

Input and output operators >> and << are overloaded for the LNL::LongInt class and can be used to insert/extract LNL::LongInt instances as a stream of characters, just like with the standard version of these operators.

Here is an example code:

LNL::LongInt x;
std::cout << "Enter a number: ";
std::cin >> x;
std::cout << "The number is: " << x << "\n";

And if you run this code:

Enter a number: -123456789
The number is: -123456789

Below there is a list of some useful getters featured in LNL::LongInt class. For convenience and versatility of use, these functions come in two variations: as a class method and as an outside-a-class function.

List of useful getters:

• x.toString() and LNL::toString(x) return number value as a std::string.
• x.abs() and LNL::abs(x) return absolute number value (returns LNL::LongInt).
• x.size() and LNL::size(x) return number of digits in a number (returns long long).
• x.length() and LNL::length(x) are synonyms of x.size() and LNL::length(x).
• x.isEven() and LNL::isEven(x) check if a number is even (returns bool).
• x.isOdd() and LNL::isOdd(x) check if a number is odd (returns bool).
• x.isZero() and LNL::isZero(x) check if a number is equal to 0 (returns bool).
• x.isOne() and LNL::isOne(x) check if a number is equal to 1 (returns bool).
• x.isPositive() and LNL::isPositive(x) check if a number is positive (returns bool).
• x.isNegative() and LNL::isNegative(x) check if a number is negative (returns bool).

Here x is an instance of LNL::LongInt.

Relational operators ==, !=, >, >=, <, <= are overloaded for the LNL::LongInt class. There are also relational functions like:

• LNL::min(x,y) returns the least of two numbers (returns LNL::LongInt).
• LNL::max(x,y) returns the greatest of two numbers (returns LNL::LongInt).

Here x and y are instances of LNL::LongInt.

Operators +, -, *, /, %, <<, >> as well as operators +=, -=, *=, /=, %= are overloaded for the LNL::LongInt class and function just like you would expect them to. The same goes for unary operators +, - and ++, -- (both prefix and postfix).

Some other mathematical functions:

• LNL::pow(x,y) calculates x to the power of y (returns LNL::LongInt).
• LNL::div(x,y) performs a division of two numbers an returns a lidiv_t type structure (more about it below).
• LNL::mod(x,y) performs a modulo operation x mod y (returns LNL::LongInt). Consider the the difference between this modulo operation and the division remainder operator %!
• LNL::gcd(x,y) returns the greatest common divisor of two numbers (returns LNL::LongInt).
• LNL::lcm(x,y) returns the least common multiple of two numbers (returns LNL::LongInt).
• LNL::factorial(x) returns factorial of a number (returns LNL::LongInt).

Here x and y are instances of LNL::LongInt.

It was mentioned that LNL::div(x,y) function returns a lidiv_t type structure. This structure consists of a quotient and a remainder of a division, which looks like this:

struct lidiv_t {
	LongInt quot;
	LongInt rem;
};

Using LNL::div(x,y) is faster than operators / and % if you know you need both quotient and remainder.

There are multiple ways to generate random numbers in LNL:

• LNL::random() generates a random number of random size (returns LNL::LongInt).
• LNL::random(x,y) generates a random number in range from x to y (returns LNL::LongInt).
• LNL::random(s) generates a random number of size s (returns LNL::LongInt).

Here x and y are instances of LNL::LongInt and s is a long long.

LNL features some functions useful for cryptography and information security, which could be used for further implementation of more complex cryptography algorithms. Below is a list of currently implemented functions:

• x.isMillerRabinPrime() checks if a number is a prime using a probabilistic primality test (returns bool).
• LNL::randomPrime(s) generates a random prime number of size s (returns LNL::LongInt).
• LNL::modPowTwo(x,s,y) performs a modular exponentiation of x by 2 to the power of s with a modulus z (returns LNL::LongInt).
• LNL::modPow(x,y,z) performs a modular exponentiation of x by y with a modulus z (returns LNL::LongInt).
• LNL::shiftEncrypt(x, s) is a Caesar's cipher encryption which is performed on number x with a key s (returns void, encrypts the number x).
• LNL::shiftDecrypt(x, s) is a Caesar's cipher decryption which is performed on number x with a key s (returns void, decrypts the number x).

Here x, y, z are instances of LNL::LongInt and s is a long long.

⬆️ Back to contents

In this section you can find some benchmarks for the library.

LNL uses karatsuba divide-and-conquer algorithm for multiplication. It allows to reduce the number single-digit multiplications from $n^2$ (traditional multiplication) to at most $n^{log_23} \approx n^{1.585}$ when multiplying two $n$-digit numbers.

Benchmark plot for multiplication of two $n$-digit numbers:

Closeup of the benchmark plot:

Calculation time $time^{log_32}$ on the last plot is linear which indicates the correctness of the karatsuba algorithm. You can also see that this algorithm outperforms traditional multiplication because the $\sqrt{time}$ plot has logarithmic properties.

LNL uses traditional long division algorithm which has $n^2$ complexity at worst when dividing two $n$-digit numbers. The algorithm for division may be reworked in the future.

Benchmark plot for division of a $n$-digit number by some constant much smaller number:

Closeup of the benchmark plot:

Calculation time $\sqrt{time}$ is linear on this plot which points to a $n^2$ complexity as expected. However division is faster the closer the numbers are.

Here is a benchmark plot for division of two $n$-digit numbers:

LNL uses exponentiation by squaring which uses $log_2a$ multiplications when calculating $x^a$. However time complexity is also affected by the complexity of the multiplication algorithm itself (see the multiplication section above).

Benchmark plot for exponentiation:

You can also run benchmarks for other functions and operators of LNL with more specific benchmark parameters. For more information see the run benchmarks section.

⬆️ Back to contents

In the include folder in this repository you can find a second header called LNL-unit-tests.h. You can install this header just like the library itself. It is important to note that both headers should be located in the same directory. Inside this file there is a class LNL::Test which contains unit tests for the LNL::LongInt class.

To run all tests at once you can use code below:

LNL::Test::runAllTests();
LNL::Test::printTestReport();

Or if you forked the repository you can run:

g++ source/run-unit-tests.cpp -o run-unit-tests.out
./run-unit-tests.out

If you are using Linux OS, you can run benchmarks for some functions and operators to measure their execution time. You can find benchmark implementation in run-benchmarks.py and benchmark.cpp files in the source directory. Here is an example code that you can run in your terminal:

python3 source/run-benchmarks.py -path source/benchmark.cpp --pow

This will run benchmarks for function LNL::pow() and plot the results. For a more detailed understanding of how to run this script, use this command:

python3 source/run-benchmarks.py -h

⬆️ Back to contents

If you find any bugs or if you have improvement suggestions, please feel free to open an issue or a pull request. If you open a pull request, please add some comments to your changes and make sure all tests from LNL-unit-tests.h header are running correctly.

⬆️ Back to contents

All code is licensed under the Apache License, Version 2.0 (the "License"). You may modify, distribute and use this product in private or commercial purposes as long as you preserve the copyright and license notices. See the LICENSE file for more detail.

See the NOTICE file for copyright information.

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

⬆️ Back to contents