COSC342: Lab 02

COSC342: Lab 02
Matrices and Vectors in C++
1 Introduction
In this lab you will practice using a simple Matrix library in C++. Online
documentation for this library is available at http://www.cs.otago.ac.nz/
cosc342/docs/matrices/index.html. The library provides two classes, Matrix
and Vector, and Vector is a subclass of Matrix.
Each class has a header file, Class.h, and an implementation, Class.cpp.
There is also a sample main file, matrixTest.cpp illustrating some basic operations, and a CMakeLists.txt file which we will use to build the project.
You can get the source code from Blackboard. Put it in a suitably named
place in your home directory, say .../cosc342/lab02-Matrices. Download
the code from Blackboard and unzip it in the directory you’ve made.
2 Building Projects with CMake in Visual Studio Code
There are a number of different toolchains for C++ on the lab machines. The option that we will use in the lab uses Visual Studio Code. Visual Studio Code
allows to load a CMake configuration file and automatically sets the project
structure up. Other options include the usage of a tool called CMake, which
creates projects for different platforms (we will not use this in the lab, but you
can use this for setting up makefiles). You can use this in other environments
as well, for example to make a project for Visual Studio under Windows.
In the lab code there is a file called CMakeLists.txt which describes the
project that needs to be compiled. Lets take a look inside that:
project ( Matrices )
add_executable ( Matrices
matrixMain . cpp
Matrix . h
Matrix . cpp
Vector . h
Vector . cpp
)
1
This is a very simple CMake file, which declares that we’re making a project
called Matrices. This is also the name of the executable program we’ll be
building, and that executable depends on the listed .cpp and .h files.
To set up the project, run the program called Visual Studio Code, which
is in the Applications directory. If you can’t find it press command-space and
type “Visual Studio Code” to search for it. Then go to File:Open and open
the directory with the code for this lab.
Now you need to configure the project. Go to the CMake tab and click
Configure as shown in the next figure.
You will be asked to install cmake tools. If cmake is not yet installed, type
Command P and type ext install cmake.
The next step will be to select a compiler. Visual Studio Code will ask you
what compiler kit to use. Select Clang.
2
2.1 Building with Visual Studio Code
After configuring the project, the next step is to build. Click on the Build icon
in the CMake tab.
After the application is build, you can run and debug it. For this you need
to configure a config file that describes where your applications is located and
also from which work folder you want to run it (in case you want to load any
additional files during runtime). In the Run and Debug tab click on Create a
launch.json file, select a debugger c++/dgb.
Specify your application executable. This means you need to tell Visual
Studio where your program is located. In our case this will be
${workspaceFolder}/build/Matrices as shown in the next figure. You also
might want to specify to use the terminal as output for better formatting. Set
externalConsole to true. All your output will now be shown in the terminal
while you can still use Visual Studio Code for debugging.
3
3 Debugging
Often if you are searching for a bug, debugging is a great option to step through
the code and watch the status of variables in your program In order to set a
debug point, select the Explorer tab, select on of the cpp files and just click next
to line numbers of code window. You can then use the watch option to inspect
variables (instead of printing them out manually).
4 Working with the Library
This may be your first time working with C++, but the sample file should give
you enough information to get started with the Matrix and Vector classes.
We’ll go through some basics, do some exercises, then you can go back to look
at the classes and their implementations in more detail.
Matrix and Vector objects are most commonly made with constructors that
take the size as parameters:
Matrix A (3 ,2); // A 3 x2 matrix
Vector v (3); // A 3 - vector
If no arguments are given then a 1 × 1 Matrix or Vector is made:
Matrix A ; // A 1 x1 matrix
Vector v ; // A 1 - vector
You can access the elements of a Matrix or Vector with a function-style
syntax. Following C/C++ convention, indices start from 0 rather than 1.
Matrix A (3 ,3); // A 3 x3 matrix
A (0 ,0) = 1; // Set top left element to 1
A (1 ,2) = 2; // Set 3 rd element of 2 nd row to 2
A (2 ,2) = 3; // Set bottom right element to 3
C++ uses streams for output, and the standard output stream, std::cout,
is defined in the iostream header. The << operator is used to send information
to the stream, so “Hello World” in C++ looks like this:
# include < iostream >
int main () {
std :: cout << " Hello , World " << std :: endl ;
return 0;
}
The special constant std::endl ends a line and flushes the stream. You can
also use << with Matrix and Vector objects:
std :: cout << A << std :: endl ;
Vectors are a single column, so it is often more convenient to output their
transpose:
4
std :: cout << v . transpose () << std :: endl ;
C++ supports operator overloading, which we’ll discuss in detail later. For
now all you need to know is that this means you can write fairly natural arithmetic expressions using Matrix and Vector objects:
Matrix A (3 ,3);
Vector v (3) , u (3);
u = A * v ;
5 Exercises
These exercises give you some more practice. For each exercise you should write
code to evaluate the required expressions, and display the results in the console
with the std::cout stream.
5.1 Vector Arithmetic
Let u = (1, −3, 2) and v = (3, 2, 0). Evaluate the following:
1. 2u
2. u + v
3. u − v
4. 2u − 3v
5.2 Dot and Cross Products
Vector objects have methods dot and cross which take another Vector as a
parameter and return the appropriate product.
Evaluate the following dot products:
1. (1, 2, −2) · (1, 2, −2)
2. (1, 2, −2) · (−2, 2, 1)
3. (1, 2, −2) · (2, 3, 1)
Evaluate the following cross products:
1. (1, 2, −2) × (−2, 2, 1)
2. (1, 2, −2) × (−2, −4, 4)
3. (1, 2, −2) · ((1, 2, −2) × (−2, 2, 1))
5
5.3 Matrix Arithmetic
Evaluate the following:
1. 2 
1 −2
3 4 
−

3 2
4 1
2. 
2 1
2 2  1 −2
−4 1 
3. 
1 −2
−4 1  2 1
2 2
5.4 Matrices and Vectors
These following exercises make use of the matrix A =

1 1 0
0 1 2
, and the vectors
u = (1, 2, −2) and v = (−2, 2, 1). Compute the following:
1. Au
2. uuT
3. u
Tv
6 Inside the Library
This may be your first time working with C++, but many of the concepts should
be familiar to you from your C and Java experience. For the most part you
should just be able to follow along for now, but here are some notes that might
help you make sense of things. I’d recommend skimming over these at first,
having a go at writing some code, and then coming back for a closer look. There
are also plenty of comments in the code which should help you get started, as
well as the online documentation.
Classes in C++ are defined using a similar syntax to Java (technically Java
borrowed a lot of syntax from C++), and each class typically has two files associated with it – a header file (e.g.: Matrix.h) and an implementation file (e.g.:
Matrix.cpp). If you open Matrix.h in a text editor, you will see the definitions
(or signatures) of many methods, but no implementation details. These details
are all in Matrix.cpp, and the header file just gives the interface.
Some of the syntax may be unfamiliar to you, in particular operator overloading. A good example is the multiplication method for Matrices:
friend Matrix operator *( const Matrix & lhs ,
const Matrix & rhs );
The short version of what this does, is allow us to write code like
6
Matrix A (3 ,3) , B (3 ,4) , C (3 ,4);
// Code to fill A and B with values
C = A * B ;
This code creates a 3×3 matrix called A, a 3×4 matrix called B, then multiplies
them together, storing the result in C.
In more detail:
• friend Tells us that even though this function is not exactly a method of
the Matrix class it can access the internal elements of Matrix objects as
if it was.
• Matrix Tells us that the result of multiplying two Matrix objects together
is another Matrix object.
• operator* Tells us that this function provides an implementation for the
multiplication operator. Basically, you can use A*B as a shorthand for
Matrix::operator*(A,B).
• Finally the parameters are Matrix objects. These are passed by reference
(&) to avoid copying complex objects, and the const keyword provides a
guarantee that if you call A*B, A and B will remain unchanged.
• The parameter names, lhs and rhs, are supposed to remind you that they
are on the left hand side and right hand side of the * operator respectively.
Operator overloading is a very powerful tool for providing concise and intuitive interfaces to mathematical objects like Matrices and Vectors. It also lets us
provide a few other useful things, such as overloading the assignment operator,
which is also used in the line C = A*B;, above. Another useful pair of overloads
are the stream operators, << and >>, which are used in C++ to provide a powerful syntax for input and output through streams. The classic Hello World
program in C++ is
# include < iostream >
int main () {
std :: cout << " Hello World " << std :: endl ;
return 0;
}
The iostream library provides input and output streams, one of which is
std::cout – this is an output stream which goes to the console (unless redirected by the operating system). The std:: part just says that the stream cout
exists in the standard namespace, which prevents cluttering the global namespace with lots of stuff. In this context the << operator provides the ability to
send information to the stream. In this case, we are sending a string, and this
gets printed in the console.
7
Overloading << means that we can just send Matrix operators to these
streams:
std :: cout << " Matrix A is : " << std :: endl ;
std :: cout << A << std :: endl ;
One final way that operator overloading is used is to access elements of
Matrix objects. This is done with the function call operator, operator().
There are two versions of this, and here are their signatures:
double & operator ()( size_t row , size_t col );
const double & operator ()( size_t row , size_t col ) const ;
These let us write code like
Matrix A (3 ,3);
double x = A (1 ,2);
On the first line, the brackets tell us that the parameters to send to the constructor are 3 and 3. On the second line, however, the brackets are calls to the
function call operator, which returns the corresponding element of the matrix
(indexed from 0, of course, since this is C++). So the second line gets the value
from the second row and third column of A, and stores it in x.
Note that the function call operators return a reference, this means that you
can also assign values to Matrix elements like this:
A (2 ,1) = 4.5;
This is also why there are two versions of the function call operator. C++ makes
a lot of use of the idea of const correctness, which is using the keyword const
to tell the compiler when things should not be altered. If A is const, then we
want to allow read access to its elements, but don’t want to be able to write
to them. The second version of the function call operator tells us that if we
call operator() on a const Matrix then we get back a const double&, and
so can’t assign a new value to it.
Have a look at the source code for the Matrix and Vector classes. Notice
that the output of the operators on Vector and Matrix objects might be a
Vector, a Matrix or a double (a scalar). This helps to catch errors at compile time, but other errors (such as multiplying Matrix objects whose interior
dimensions don’t agree) are caught at run time with asserts.
Have a look through the code and see how it works. Ask us if you have any
questions.
7 More Advanced Matrix Libraries
This Matrix code is designed to be easy to use and reasonably easy to understand. More advanced C++ programming methods allow for very efficient
implementations, but they are much harder to understand. One such library is
Eigen, which can be found at http://eigen.tuxfamily.org.
8
Eigen uses a feature C++ called templates, which are a bit like Java generics
(or again, Java generics are like C++ templates) but more powerful. Matrices in
Eigen are declared by giving their size and type as template parameters:
// A 3 x4 matrix of doubles
Eigen :: Matrix M <3 , 4 , double >
The use of templates allows the compiler to do a lot of reasoning about the size
and type of matrices. It means that the compiler can determine automatically
that the result of multiplying a 3 × 4 matrix by a 4 × 2 matrix is a 3 × 2 matrix.
If you try assigning this to an incorrectly sized matrix, then you get an error at
compile time rather than runtime.
Templates also allow for a somewhat arcane C++ technique called template
metaprogramming. It turns out that the template structure of C++ is Turing
complete, so you can do all sorts of things with it. This lets the compiler realise
when it is dealing with small matrices and unroll loops etc. to make faster code.
Other advanced features of Eigen like lazy evaluation, avoidance of temporary
objects, and vectorisation make the code very fast, while keeping a simple highlevel interface.
The Turing completeness of the template mechanism wasn’t planned, so the
syntax is rather ugly (and the compiler error messages worse). For our purposes,
however, code that you can understand is better than code which is very fast.
9