Milestone 1: due Wednesday, Sep 4th by 11 pm

Milestone 2: due Wednesday, Sep 11th by 11 pm

Assignment type: Pair, you may work with one partner

Overview

In this assignment, you will implement a BigInt C++ data type which allows arbitrary-precision integer arithmetic.

Milestones, grading criteria

Milestone 1 (15% of the assignment grade):

• Implementation of functions (15%)
  • constructors
  • destructor
  • assignment operator
  • git_bits(unsigned) member function
  • get_bit_vector() member function
  • is_negative() member function
  • operator-() member function (unary minus)
  • to_hex() member function (conversion to base-16)

Milestone 2 (85% of the assignment grade):

• Implementation of functions (65%)
  • is_bit_set(unsigned) member function
  • operator+(const BigInt &) member function (addition)
  • operator-(const BigInt &) member function (two-operand subtraction)
  • operator*(const BigInt &) member function (multiplication—challenging!)
  • compare(const BigInt &) member function
  • operator<<(unsigned) member function (left shift)
  • operator/(const BigInt &) member function (division—challenging!)
  • to_dec() member function (conversion to base-10—challenging!)
• Comprehensiveness and quality of your unit tests (10%)
• Design and coding style (10%)

Getting started

To get started, download csf_assign01.zip, which contains the skeleton code for the assignment, and then unzip it.

Note that you can download the zipfile from the command line using the curl program:

curl -O https://jhucsf.github.io/spring2024/assign/csf_assign01.zip

We strongly recommend using a Git repository to store your code. Please do not use a public repository, however.

To compile the test program, run the commands

make depend
make

Note that make depend automatically determines header file dependencies.

To run the unit tests:

./bigint_tests

Machine integers, arbitrary-precision integers

As you know, the “built-in” C integer data types (int, uint64_t, etc.) have a finite number of bits, and so can represent only a finite set of possible values. Therefore, they don’t model mathematical integers with complete fidelity. However, they can be building blocks for the creation of an arbitrary-precision integer data type, where there is no fixed limit on the range of values that can be represented (other than machine limitations such as the amount of memory that can be addressed by the program.)

In this assignment, you will implement the BigInt C++ data type. It should have two member variables:

• a std::vector containing uint64_t values representing the magnitude of the BigInt value
• a bool representing whether or not the BigInt value is negative

The vector of uint64_t elements is the “bit string”. If you think about the magnitude of a BigInt value as an arbitrary integer, the bit string specifies that value in base 2. The first element contains the least-significant 64 bits of the bit string, the second element contains the next 64 bits of the bit string, etc. For example, consider the integer value \(2^{64} = 18,446,744,073,709,551,616\). This value is represented in base 2 as

\[10000000000000000000000000000000000000000000000000000000000000000\]

which is one followed by 64 zeroes. In hexadecimal (base 16), this value is represented as

\[10000000000000000\]

which is 1 followed by 16 zeroes. Note that each hex “digit” represents 4 bits (base-2 “digits”.)

Since a uint64_t value contains exactly 64 bits, the value \(2^{64}\) could be represented as two uint64_t values: the less-significant would have the value 0 (representing the lowest 64 bits, all of which are 0), and the more-significant would have the value 1 (meaning that there is a 1 in the \(2^{64}\) place in the base-2 representation.

In general, because a BigInt object contains a vector of uint64_t values, with no arbitrary limit on the number of elements the vector can hold, it can contain as many uint64_t values as it needs in order to represent any integer magnitude.

Since integers can be negative or non-negative, a BigInt will also have a bool value which is set to true if the BigInt is negative, false if non-negative. This makes BigInt a sign/magnitude integer representation.

Your tasks

You have two general tasks:

1. implement the member functions of BigInt
2. implement additional unit tests so that all member functions are tested thoroughly

The member functions have detailed documentation comments in bigint.h which describe how they should behave. There are also provided unit tests in bigint_tests.cpp which further clarify the intended behavior of the member functions. Note that while the provided unit tests are useful, and are a good starting point for validating the implementation of the member functions, there are important cases that they don’t test. So, it will be critical for you to write additional tests.

As documented above, there are two milestones. Milestone 1 tests only a limited subset of member functions, and the unit tests you write aren’t part of the grading criteria for Milestone 1. Milestone 2 requires you to implement the rest of the member functions, and also requires you to have comprehensive unit tests for all functions (including the ones tested in Milestone 1.)

Restrictions

You must adhere to the following restrictions in completing the assignment.

Only standard library classes/functions can be used. You aren’t allowed to use any external libraries in your implementation. However, you are free (and encouraged) to use any functionality in the C++ (or C) standard library.

Only 64-bit and smaller data types can be used. You aren’t allowed to use any data type whose representation is larger than 64 bits. (However, you won’t need to, and it wouldn’t make the job easier in any case.)

Original code only. It should go without saying that all of the code you submit must be your original work. Copying code from an external source would be a violation of academic ethics.

Recommendations and hints

This section has further recommendations and hints, in no particular order.

Don’t use floating point operations

You will not need to use floating point (float or double) operations. If you have a problem that you think requires floating point, there is definitely a way to solve the problem without floating point. For example, if you need to compute an arbitrary power of 2, don’t use the pow function. Instead, left shift the value 1UL the appropriate number of places. E.g., 1UL << n will be equal to \(2^{n}\) as long as n is in the range \(0 \ldots 63\).

This assignment is about integers.

Do use auto

We encourage you to use auto to infer types, since it makes code cleaner and more readable. For example, to iterate through a vector v:

for (auto i = v.begin(); i != v.end(); ++i) {
  // do something with *i
}

Helper functions

We recommend adding private member functions as necessary to support the implementation of the required public member functions. For example, the reference implementation defines the following private member functions:

bool is_zero() const;
static BigInt add_magnitudes(const BigInt &lhs, const BigInt &rhs);
static BigInt subtract_magnitudes(const BigInt &lhs, const BigInt &rhs);
static int compare_magnitudes(const BigInt &lhs, const BigInt &rhs);
BigInt div_by_2() const;

You aren’t required to implement these specific private member functions (or any private member functions for that matter.)

Addition of magnitudes

You should implement addition of magnitudes using the “grade school” algorithm. Think about each element of the vector of uint64_t values in a BigInt object as a “digit”. The uint64_t values just happen to be digits in base \(2^{64}\).

Start by adding the “digits” in the rightmost column to compute the rightmost digit of the result. Note that the computed “digit” is correct modulo \(2^{64}\). If the addition of the column values overflows, you will need to carry a 1 into the next column (to the left.) This is exactly analogous to base-10 addition. For example:

  16
+  7
----
  ??

When adding the digits in the right column (\(6 + 7\)), the sum is \(13\), which modulo the base (10) is 3. So, the rightmost digit of the sum is 3. However, the addition overflows, since \(13\) can’t be represented as a single base-10 digit. So, we will need to carry a 1 into the next column.

To detect overflow when adding uint64_t “digits”, you can use the standard trick for detecting unsigned integer overflow:

sum = a + b;
if (sum < a) {
  // overflow occurred
}

One concern when implementing addition is that the two BigInt values being added might have different numbers of uint64_t values in their bit string vectors. The get_bits() member function can be helpful for retrieving part of a bit string without needing to worry about how many elements the vector actually has.

Addition with negative values

You can handle negative values as follows:

• If two nonnegative values are added, or if two negative values are added, then the result has a magnitude that is the sum of the magnitudes of the operands, and the sign will be the same as the sign of both operands.
• In a “mixed” addition (one operand is non-negative and one operand is negative), the magnitude of the result is the difference \(a-b\), where \(a\) is the magnitude of the value with the larger magnitude, and \(b\) is the magnitude of the value with the smaller magnitude. The sign of the result is the same as the sign of the operand with the larger magnitude.

You will probably find it useful to create helper functions to add magnitudes and subtract magnitudes.

Subtraction

The two-operand subtraction operator operator-(const BigInt &) can be implemented by adding the negation of the right-hand operand to the value of the left hand operand. In other words, you can compute

as

Subtraction of magnitudes

Subtraction of magnitudes should always involve subtracting a smaller magnitude from a larger magnitude. As with adding magnitudes, you can use the “grade school” algorithm: start with the “digits” in the rightmost column. Each time a column difference requires subtracting a larger value from a smaller value, you will need to borrow 1 from the next column.

Again, as a base-10 analogy, in the subtraction

  16
-  7
----
  ??

the rightmost column difference \(6-7\) requires borrowing 1 from the next column since \(7\) is greater than \(6\).

Left shift

The left shift operator (operator<<(unsigned)) will require some careful thought. Here are some ideas that could make it simpler to implement:

• If the number of bits shifted is a multiple of 64, that is an “easy” case, since each uint64_t value in the result’s bit string will be identical to a corresponding uint64_t value in the original value’s bit string.
• The number of bits shifted is really only “interesting” modulo 64. For example, shifting left by 1 bit, shifting left by 65 bits, shifting left by 129 bits, etc., are very similar cases. The only way in which they differ is how many uint64_t values worth of 0 bits are incorporated into the least-significant part of the result’s bit string.

Conversion to hexadecimal (to_hex())

Converting a BigInt value to a hexadecimal string is fairly straightforward if you use a std::stringstream object to help with the formatting of the uint64_t values as hexadecimal. The std::hex, std::setfill, and std::setw manipulators will likely be helpful.

Don’t forget that a negative value needs a leading - sign. Also, make sure there are no unnecessary leading 0 digits. (Although, the value equal to \(0\) should be coverted to the string “0”.)

Multiplication

One way to implement multiplication is to break down one of the operands into powers of 2, multiply the other operand by each of those powers of 2, and add those partial products together.

For example, \(37 = 32 + 4 + 1 = 2^{5} + 2^{2} + 2^{0}\). The product \(37 \times m\) could therefore be computed as

Note that left-shifting a value by \(n\) is the same as multiplying it by \(2^{n}\). So, if you have implemented the is_bit_set(), operator<<, and operator+ member functions, you should have everything you need to implement multiplication.

You will need to think about how the sign of the result relates to the signs of the operands.

Division

A simple way to implement division is using binary search. In computing \(q = n / m\), where \(n\) is the dividend and \(m\) is the divisor, we can note that the quotient \(q\) will be in the range \(0 \ldots n\), inclusive. So, initially, \(0\) is the lower bound of the range, and \(n\) is the upper bound of the range. Repeatedly, we can choose a possible value of \(q\) midway between the lower and upper bounds. Depending on whether or not \(q \times m\) is greater than \(n\), we can revise either the lower or upper search bound. Eventually, the range should collapse to a single value, which is the computed quotient \(q\).

Note that this is not a particularly fast way to implement division, but it is adequate for this assignment.

When choosing a value midway between the current lower and upper bounds, it’s useful to have a “divide by 2” operation. Note that a right shift by 1 bit is effectively a division by 2.

The intended semantics of division is that the computed quotient has the largest magnitude such that multiplying it by the divisor yields a product whose magnitude is not greater than the dividend’s magnitude.

You will need to think about how the sign of the result relates to the signs of the operands.

Conversion to decimal (to_dec())

One algorithm for converting an integer to decimal (base 10) is repeatedly dividing by 10: the remainder of each division will yield one digit, in order from least significant to most significant.

Note that you could make this process more efficient by generating more than one digit at a time. For example, repeatedly dividing by 100 would yield two base-10 digits at a time. You should think about how many base-10 digits can be produced by each iteration of this process.

As with to_hex(), std::stringstream will be helpful. Also, don’t forget to prepend a leading - sign if the value is negative.

Writing tests

Your unit tests should test each required member function thoroughly. We recommend that you implement your tests by adding additional test functions to bigint_tests.cpp, rather than adding new tests to the provided test functions.

Your tests should try to create “interesting” scenarios for each tested member function. This includes things like

• zero vs. non-zero values
• negative vs. non-negative values (and combinations thereof)
• smaller vs. larger values
• etc.

A good mindset for testing is that you are an adversary of your own code, i.e., you are trying to make it break.

One approach that can be helpful is to write a program to generate test cases. Both Python and Ruby support arbitrary-precision integers as part of the core language. So, you can write a script in either of those languages which computes arbitrarily large integer values, does an operation on them, and then prints code of a test case to check whether BigInt produces the same result when doing the same computation. Some of the test cases in the provided unit tests were generated by a Ruby script. For example, here is an automatically-generated test for subtraction:

{
  BigInt left(
      {0x94e439a254295b2fUL, 0xc02d6dc0be0efef4UL, 0xe5156c9d912b61f2UL,
       0xb82729123ce1051eUL, 0x1d2c69a0ed4011c3UL, 0xf13f35779fd54911UL,
       0x15056f71d40516eaUL, 0xdb571f43f9416bdeUL, 0x7e21086e7df7095UL,
       0x797275a8e7538b0aUL, 0x18a6284e20e7893aUL});
  BigInt right(
      {0x9161ea05eb48510dUL, 0xb7a476402ef52acaUL, 0xdf96be7a926695adUL,
       0x53e8bc19a9c14029UL, 0xf87ee595e422d5f0UL, 0x72dd209be1d990cbUL,
       0xb991581205507625UL, 0x77bbceb930f0c50eUL, 0x862b240a5ee05327UL,
       0x44af5ae70f9c63b6UL, 0x30UL}, true);
  BigInt result = left - right;
  check_contents(result, {0x264623a83f71ac3cUL, 0x77d1e400ed0429bfUL,
                          0xc4ac2b182391f7a0UL, 0xc0fe52be6a24548UL,
                          0x15ab4f36d162e7b4UL, 0x641c561381aed9ddUL,
                          0xce96c783d9558d10UL, 0x5312edfd2a3230ecUL,
                          0x8e0d349146bfc3bdUL, 0xbe21d08ff6efeec0UL,
                          0x18a6284e20e7896aUL});
  ASSERT(!result.is_negative());
}

Note that automatically-generated test cases are a complement to hand-written test cases, not a substitute for them.

Submitting

Before submitting each milestone, you should edit README.txt to include information about

• the contributions of each team member
• any interesting implementation details or things we should know about your submission

You can use the command make solution.zip to create a zipfile with the required submission files.

Upload solution.zip to either Assignment 1 MS1 or Assignment 1 MS2 on Gradescope as appropriate.