slama.dev

The C# programming language

notes · 19. 2. 2021 · [edit]

• Strings
  • Interning
  • System.Text.StringBuilder
  • Interpolation
• Chars
• File I/O
• Classes
  • Constructor
  • Static/class constructor
  • Destructors/finalizers
  • Inheritance
    • Virtual/abstract/new methods
    • Superclass to subclass conversion
      • is
      • as
    • System.Object
    • System.ValueType
    • sealed
  • Generic classes
• Variable scope
  • Local variables
• Exceptions
  • System.Exception
  • Syntax
  • Common exception classes
• using
• Properties
  • Auto-implemented
  • Read-only
  • Instantiating objects with properties
  • Parametric properties (indexers)
• Type System
  • Value types
    • Simple types
    • Nullable types
  • Reference types
    • Arrays
  • (un)boxing
• Functional elements
  • Local functions
  • Delegates
  • Lambda funtions
• Structures
• Visibility
• readonly
• =>
• ??, ??=, ?.
  • Arithmetic overflows
• References
  • ref
  • out
• var
• Interfaces
• Implementation details
  • Heaps and garbage collection
    • Large Object Heap (LOH)
    • Small Object Heap (SOH)
• Terms to know
  • CIL
  • CLR
  • GAC
  • JIT
  • AOT
  • BCL
  • .NET
  • Tiered compilation
  • CLS
  • typeof(Class)
  • nameof(x) [Microsoft Docs]
  • NUnit tests
• Data structures
  • Dictionaries
  • Queues
• Acknowledgements

Preface

This website contains my lecture notes from a lecture by Pavel Ježek from the academic year 2020/2021 (MFF UK). If you find something incorrect/unclear, or would like to contribute, feel free to submit a pull request (or let me know via email).

Besides this, they have been slightly extended by some C#-specific questions for my Bachelor’s state exam that were not covered in the lecture, namely generics and functional elements.

Strings

• internally an array of chars
• immutable (neither length nor contents)
  • concatenation creates new strings
    • String.concat(s1, s2, s3, ...); is good for concatenating a lot of them
      • s = "a" + "bcd" + "ef"; internally uses this method, so it isn’t slow
• System.String is the same as string
  • can’t ever be a name of a variable!
  • removes the confusion – what if someone decides to implement a String class
• == compares contents, char by char (same as .Equals())
  • we can use object.ReferenceEquals(o1, o2) if we want reference equality

Interning

• hashmap for reusing already created strings
• only for constants, not variables (only exception being "")!
• we can do this ourselves using String.Intern(str); – if the table contains it, return it; else add it to the table
  • should be limited use: strings can’t be removed from the table

System.Text.StringBuilder

• essentially a mutable string
• internally behaves like a dynamic array
  • quick operations like concatenation
• .ToString() returns a proper string (since we can’t use StringBuilder anywhere we want a string)
  • no copying happens – internally, the new string points to the contents of the StringBuilder (so we don’t have to copy twice)
  • if we were to modify the string builder again, it gets copied

Interpolation

• Console.Write("cislo = {0} a {1}", i, j");
  • same as Console.Write("cislo = " + i.ToString() + " a " + j.ToString()),
  • calls String.Format("format string", i, j)
    • creates new string

Console.WriteLine("{0}: Hello, {1}!", s1, s2);

// careful, what if s2 == "{0}"
Console.WriteLine("{0}: Hello, " + s2 + "!", s1, s2);

• interpolated strings – $"My name is {name} and I'm {age}."
  • is translated into a String.Format call (at compile time)… most of the time
  • if assigned to a FormattableString, its instance is created instead
    • .GetArgument(i) returns the i-th argument
    • .Format() creates a string
  • the {} blocks can be additionaly formatted using :<format> and , <format>

Chars

• System.Char == char (keyword)
• 2-byte UTF-16 character
  • some characters must be at least a string, since some UTF-16 characters can be up to 4 bytes
• \xABCD – unicode code in hex, consumes one to four hex characters
• \uABCD – unicode code in hex, consumes exactly four hex characters
• \UABCDEFGH – unicode code in hex, consumes exactly eight hex characters
  • note that while it does mean a single character, this has to be a string

File I/O

• if we’re dealing with binary I/O, use FileStream instead; this is for text

Classes

Constructor

• same syntax as C++, but it’s a good idea to add a public before it
• if none is specified, a default one without parameters is created
  • note that we have to write it explicitly if we want it besides one with parameters!
• translated to a .ctor method (see disassembly)
  • useful to know, since debug messages can contain this

class A {
	int x = stuff;
	
	A () {}
}

// is equivalent to (for each constructor!)
// note that stuff can be arbitrary code

class A {
	A () {
		int x = stuff;
	}
}

• everything in C# inherits : object == System.Object (if not inheriting anything)

When inheriting, the constructor of the predecessor is called like this:

class B : A {
	int x = stuff;
	
	B () {
		stuff2
	}
}

// is equivalent to (for each constructor!)
// note that stuff can be arbitrary code

class B : A {
	B () {
		int x = stuff;
		
		// constructor of A (without parameters)
		
		stuff2
	}
}

If no constructor without parameters exist (for example when a class contains one with parameters, which makes the one without parameters not generate) and we inherit the class without calling it, it won’t compile:

// THIS WON'T COMPILE!

class A {
	public A(int x) { }
}

class B : A {
}

We have to do this instead:

class A : B {
	A () : base(/*parameters for constructor of B*/) {stuff}
}

// is equivalent to (for each constructor!)
// note that stuff can be arbitrary code

class A : B {
	A () {
		int x = something;
		
		// constructor of B (with the appropriate parameters)
		
		stuff
	}
}

• calling one constructor from another constructor:
  • A(): this(constructor parameters) {}
  • the stuff that would be called before A() isn’t called (so we don’t do it twice)

Static/class constructor

• same as a regular constructor but for static variables
• called before the first time an object of the class is instantiated
  • if no object is constructed, it is never called
  • if we do Class a;, then it’s also not called!
• if none is specified, a default blank one is instantiated
  • this means that it can never have parameters
• internally called .cctor

class A {
	static A() { }
}

Destructors/finalizers

• same name as the class, but prepended with a tilde (~)
• has no return type and no parameters
• cannot be defined in structures
• called when the object is destroyed (by the GC)

Inheritance

class A { }      // some stuff
class B : A { }  // some stuff + some more stuff

A a = new A(); // is fine
a = new B();   // is also fine

• when inheriting, everything is inherited, except constructors
  • it would be a mess – which constructor would get called when?
• each class has exactly one predecessor
  • if none is specified, System.Object == object is used automatically

Virtual/abstract/new methods

• it sometimes makes sense to implement a function differently in a child… member hiding
  • if we’re looking at B like it’s A, the A implementation would be called (if it’s not virtual)
  • a warning is issued – did we really want to do this?
    • what if we called the method from some other method in A?
    • we can use new to let it know that we really wanted to do this

class A {
	public virtual void f() { Console.WriteLine("A"); }
	public virtual void g() { Console.WriteLine("A"); }
}

class B : A {
	public override void f() { Console.WriteLine("B"); }
	
	// new is optinal, suppresses a warning
	public new void g() { Console.WriteLine("B"); }
}

(new B()).f();  // prints B
(new B()).g();  // prints B

((A)(new B())).f();  // prints B
((A)(new B())).g();  // prints A

class Animal {
	public virtual void Name() { Console.WriteLine("Animal"); }
}

class Mammal: Animal {
	public override void Name() { Console.WriteLine("Mammal"); }
}

class Dog: Mammal {
	// new is optinal, suppresses a warning
	public new virtual void Name() { Console.WriteLine("Dog"); }
}

class Beagle: Dog {
	public override void Name() { Console.WriteLine("Beagle"); }
}

Dog pet = new Beagle();
pet.Name();              // prints Beagle

Animal pet = new Beagle();
pet.Name();              // prints Mammal

• new creates a new record, override overrides the current one:
  • it doesn’t matter that a method with the same name already existed

• if abstract is used, an entry in VTM is created but no implementation is provided
  • the entire class has to be abstract so it can’t be instantiated, since the method has to be overriden
• virtual methods can be called from the constructor and act correctly, since Type (and its VMT) must have been initialized by then
• virtual is essentially a promise to the user that it’s fine to provide an alternate implementation in the child without breaking the entire class
• base can be used to call a method from the parent non-virtually (even if it were)

Superclass to subclass conversion

• take A and B from the code above; B b = (B)a; has to be checked at runtime
  • the conversion is explicit, because it’s something to think about
• if it’s wrong InvalidCastException is thrown, so test the code!

is

• checks, whether object is of the given Type (or type of any of its children)
• runtime calls a method, not too quick of an operation – has to go through the class tree!
• if object is null, false is always returned, though it seems that null can be anything

as

• returns the object if it is of the given Type (or type of any of its children), else null

B b = a as B;  // assigns <code class="language-cs highlighter-rouge highlight" data-lang="cs"><span class="n">a</span></code> to <code class="language-cs highlighter-rouge highlight" data-lang="cs"><span class="n">b</span></code> if it's of the valid type
               // this is the reason why <code class="language-cs highlighter-rouge highlight" data-lang="cs"><span class="k">null</span> <span class="k">is</span> <span class="n">A</span></code> returns false

if (b != null) {
	// do stuff with <code class="language-cs highlighter-rouge highlight" data-lang="cs"><span class="n">B</span> <span class="n">b</span></code>
}

// is almost equivalent to (since C# 7.0)
// only difference is that <code class="language-cs highlighter-rouge highlight" data-lang="cs"><span class="n">b</span></code> is not initialized after

if (a is B b) {
	// do stuff with <code class="language-cs highlighter-rouge highlight" data-lang="cs"><span class="n">B</span> <span class="n">b</span></code>
}

System.Object

• inherited by all classes
• has the following members:
  • protected object MemberwiseClone();
    • creates a shallow copy of the object (byte by byte)
    • doesn’t make much sense for reference types
    • note that it is protected!
      • implement public A Clone() {...} to make it public
      • there is an IClonable interface, but it was before <T>…
  • public Type GetType();
    • done for all types (new Type("A"),…)
    • contains .Name and other stuff for reflection
    • also contains a virtual method table (VMT)
  • public virtual bool Equals(object o);
    • for checking, whether two objects are equal
  • public virtual string ToString();
    • the string representation of the object
    • shouldn’t be something too complicated (like a giant recursive function)
  • public virtual int GetHashCode();
    • hash code of the object; for hashmap-like data structures
  • public static bool ReferenceEquals(object objA, object objB);
    • whether the two objects are equal by reference

System.ValueType

• is inherited by all types (int, Nullable, bool, user-defined structures…)
• overrides Equals to be byte-equal (since it is a value type)
  • if it has reference members, uses reflection to check their values too – string, for example
    • better to implement it ourselves in this case

sealed

• prevents inheritance and allows for some code optimizations
  • no virtual function calls
• sealed class A {} – is not inheritable
• sealed override void m() – is not overridable

Generic classes

• useful when we have a lot of similar classes that differ by types
• can be used for structs and interfaces too

class FixedStack<T> {
	T[] a;
	int num;
	
	public FixedStack(int maxSize) {
		a = new T[maxSize];
	}
	
	public void push(T val) {
		a[num] = val;
		num += 1;
	}
	
	public T pop() {
		num -= 1;
		T ret = a[num];
		return ret;
	}
}

Variable scope

Local variables

• created each time we enter {, deleted each time we leave }
  • but is still extremely fast and highly optimized by the compiler
• the same name can’t be reused within the same scope, but be careful:

if <something> {int b;} else {int b;}  // this is ok
int b;  // this is not (already declared in a nested block)

• also, the compiler must know that the variable is initialized before use
  • a hack is to assign something garbage to f (and add a comment)

int e = 1, f;
if (e == 1) {f = 2;}
e = f; // error, f is not initialized in all paths

Exceptions

• all exceptions must inherit the System.Exception class
  • even the ones from other languages are wrapped in a SEHException
• takes a long time (a lot of things have to be collected), though a try block is basically free
• watch out for exceptions that
  1. can’t be caught (like StackOverflowException)
  2. probably shouldn’t be caught (like OutOfMemoryException)

System.Exception

Syntax

try {
	// stuff
} catch (Exception e) {
	// stuff only executed if the type matches the exception raised
} finally {
	// stuff always executed, even if the exception is not handled
	// for example, for closing IO/network resources
}

Common exception classes

• it’s important to distinguish exceptions due to bugs (NullReferenceException, IndexOutOfRangeException) and due to incorrect program usage (possibly FileNotFoundException and many others)

• Exception
  • SystemException
    • ArithmeticException
      • DivideByZeroException
      • OverflowException
      • ….
    • NullReferenceException
    • IndexOutOfRangeException
    • InvalidCastException
    • …
  • ApplicationException
    • user-defined exceptions
    • …
  • IOException
    • FileNotFoundException
    • DirectoryNotFoundException
    • …
  • WebException
  • …

using

Type x;
try {
	x = new Type();  // could raise an exception!
	
	// some code
} finally {
	if (x != null) x.Dispose();
}

• is (sort of, because x has a slightly different scope) equivalent to

using (type x = new Type()) {
	// some code
}

• is also equivalent to (since C# 8.0, with Dispose being called when x goes out of scope)

using type x = new Type();

• only works with objects that inherit IDisposable $\implies$ have a Dispose method

Properties

T Property {
	get { /* stuff to do */ }
	set { /* stuff to do (with a variable <code class="language-cs highlighter-rouge highlight" data-lang="cs"><span class="k">value</span></code>) */ }
}

• syntactic sugar for defining a T get() and void set(T value) methods
  • must be used with =, although it does generate methods get_* and set_* (so don’t name other methods like that)
• is nice when we want to let the programmer know that we’re just setting/getting something (although it may do something more complex)
  • it’s generally a good idea to not make it too slow, since the syntax looks instant
• when doing assemblies, properties make it so we don’t need to rebuild code that uses it, since the API stays the same (which can be desirable)
• interfaces can also contain them (not the implementation, just that it has to have one)

Auto-implemented

• if we’re just doing get => value and set = value, we can just do get; and set; and the code will get automatically generated
  • what if we want to make an interface out of it in the future?
  • what if we want to add constraints to the values in the future?
• we can also set default values: int X { get; set; } = 12;
• we can also explicitly set the accessibility of the functions: public int Length { get; private set; }
  • public affects get, since it doesn’t have an explicit modifier, but not set, because it has one

Read-only

• set can be omitted, which makes the property read-only
• this means it can only be changed in the constructor and nowhere else

Instantiating objects with properties

A a = new A { x = 3; y = 6 };

// is literally the same as

A a = new A();
a.x = 3;
a.y = 6;

• this implies that the setter can’t be private, since it’s just syntactic sugar!

Parametric properties (indexers)

• for defining indexing on an object of the given class
• can be overloaded, since it’s just a method (but might not be the smartest thing to do)
• are compiled into get_Item and set_Item

class File {
	FileStream s;
	
	public int this [int index] {
		get {
			s.Seek(index, SeekOrigin.Begin);
			return s.ReadByte();
		}
		
		set {
			s.Seek(index, SeekOrigin.Begin);
			return s.WriteByte((byte) value);
		}
	}
	
}

Type System

• we can’t choose whether something is a value or a reference – it is determined by its type

Value types

• variable is always passed by value (assigning copies it over)
• usually allocated in-place on the stack (with exceptions, like when a part of a class)
• contains:
  • structures
  • simple types (int, char)
  • Nullables (it’s just a struct)
  • enumerations
  • is weird, since it’s completely different to references that are allocated
• can implement interfaces, but boxing happens when we assign to I var
  • watch out for passing stuctures to function and doing something to them, since they will be boxed and nothing will change…

Simple types

• things like byte, short, int, char, bool…
• CLR and JIT knows about it, so operators (think +) is not a function call
• sizes are (almost) always defined (not like C++), except:
  • nint and nuint – native, platform-dependent (u)int (since C# 9.0)
  • bool is only true/false, so it isn’t defined too (implementation detail)
• decimal – exponent is decimal, so numbers like 0.1 are precies
  • the downside is that it is much slower
• all of them are perpendicular to one another in the type hierarchy, but there are conversions
  • it is an actual conversion, not like for reference types (just checks in that case)
  • implicit conversions happen automatically, while explicit are manual
    • although a conversion is implicit, it can have data loss (long $\rightarrow$ float)
  • they are not free, it can sometimes take quite a while (int $\rightarrow$ float)
  • no conversions from/to bool

• by default, all numeric constants (without a period) are int
  • however, we can do long a = 10;, it’s optimized to not do a conversion
  • when a specific constant type is desired, we can use the following suffixes:
    • unsigned int/long: u
    • long, unsigned long: l
    • unsigned long: ul
• by default, all decimal number constants are double
  • we can’t do float f = 2.5; (but must do 2.5f)
  • other suffixes include
    • d for double
    • m for decimal
• when designing APIs, it might be a good idea to do Int32 instead of int, so programmers from other languages now exactly what we mean

Nullable types

• int? x is a nullable variable – can contain its value, or null
• internally imlemented as a struct Nullable<T> that has T Value and bool HasValue memobers
• bool?’s null value is “i don’t know” so we can use | and & and get results we expect

Reference types

• variable is always passed by reference (assigning creates a new reference to it)
• the variable data is allocated and stored on managed heaps
• can only ever be assigned a reference to the heap
• can be null
  • actually, since C# 8.0, we can toggle this behavior and forbid it to contain a null
  • only gives warnings, since the compiler can’t always determine that it’s not from static analysis (although we can specify to treat certain warnings as errors)
• contains:
  • classes
  • interfaces
  • arrays
  • delegates
• takes up more memory, since it has to contain:
  • the reference to the heap, where the value is
  • heap overhead
    1. which type it is (reference to an instance of the class MethodTable)
       • is either $4/8B$, depending on the architecture
    2. syncblock – related to threads and locking
       • always $4B$ - padded to nearest $8B$

Arrays

• each new type inherits System.Array and is a new .NET type
• is sort of like generic types – code is generated for each variant

• array of struct automatically calls constructors on them so we don’t have to
• array of class is null by default
• jagged multidimensional arrays:
  • int[][] a = new int[2][]
  • can have different row lengths
  • takes a lot of memory (each row is a pointer) and is not generally nice to it
• rectangular multidimensional arrays:
  • single memory object
  • has to have same row lengths
  • nicer to memory
  • int[,] a = new int[2, 3]

(un)boxing

• crossing the bounday between a reference type and a value type
• object o = 5; creates a new object on the heap where the reference points to
• is immutable for simple value types, because… how would we modify it?

Functional elements

Local functions

• a function defined locally
• can access parameters and local variables

static void Main(string[] args) {
	WriteLine("hello");
	
	double arg(int n) {
		return double.Parse(args[n]);
	}
	
	double d = arg(0);
	double e = arg(1);
	WriteLine(d + e);
}

Delegates

• value representing function or method
• defined using the delegate keyword

using System;

static class Program {
	public delegate bool Condition(int i);
	
	public static void TestNumbers(Condition condition) {
		for (int i = 0; i < 10; i++) {
			if (condition(i))
				Console.WriteLine(i);
		}
	}

	public bool IsEven(int i) => i % 2 == 0;
	public bool IsOdd(int i) => i % 2 == 1;

	static void Main(string[] args) {
		TestNumbers(IsEven);
	}
}

• they can also be generic: delegate bool Condition<T>(T t)

Lambda funtions

• functions defined using an expression: parameters => expression
• can be converted into a delegate – IsEven above would be i => i % 2 == 0

Structures

• only makes sense when we want the value semantics but more complex behavior
  • Vector, Color, Complex structures, for example
• no inheritance, since the variable is always the value – assigning to variables would be broken (and is one of the main reasons for inheritance), which means:
  • no abstract, no virtual methods – all user-defined structures are essentially sealed
• as opposed to a class, a structure always has an empty constructor without parameters
  • we can’t define one, since it’s always there
  • is preserved when we define another one
• if they’re a part of a class/struct, the constructor is called automatically when initialized

using System;

struct A {
    public int num;
    public int num2;
}

struct B {
    public static A a1;
    public A a2;
}

static class Program {
    static void Main(string[] args) {
        B b = new B();
        Console.WriteLine(B.a1.num);  // is ok
        Console.WriteLine(b.a2.num);  // is also ok

        A a;
        Console.WriteLine(a.num);  // is NOT OK
                                   // we haven't called the constructor

        a.num = 5;
        Console.WriteLine(a.num);  // is ok

        A a2 = a;  // is still NOT OK
                   // the entire struct must be initialized

        a.num2 = 5;

        A a3 = a;  // is ok
    }
}

• careful with putting them in List<S> (or some other structure) and making them mutable – if we attempt to do something like myList[i].IncreaseByOne(), it will just modify the returned copy and not do anything
  • this doesn’t happen with arrays, since they’re not using an indexer
  • it’s generally a good idea to make structures immutable, because users will use it like this

Visibility

• by default:
  • visibility of classes/interfaces/structs is internal
  • visibility in classes is private
  • visibility in interfaces is public
    • it doesn’t make much sense to be private, since it’s a public contract

class A {
	private int x;
	
	// is ok, x is private in A
	public int GetX() {
		return x;
	}
	
	// is ALSO OK, since the code is inside A (B : A)
	public static void SetXOnB(B b) {
		b.x = 30;
	}
}

readonly

• used to set fields not assignable, except in the constructor
• can also be used in autoimplemented properties (int X { get; } = 5;)

=>

• syntactic sugar for:
  • when a non-void method returns something
  • when a void method has only one command

??, ??=, ?.

• conditional code execution, depending on whether something is null or not
• a ?? b returns a, if it isn’t null, else b
• x ??= y will be assigned if y isn’t null
• x?.m() will call the method if x isn’t null
• x?.f will get the field value, if x isn’t null

Arithmetic overflows

checked {
	// kód
}

• is not controlled when functions are called from this block (how could it, when they’re probably already translated…)

References

ref

• a tracking reference
• address is passed and automatically de-referenced
• can point to heap/stack/static data
• parameter must be a variable, not an expression (like a number)
• makes sense if we really want to modify the passed variable

void Inc(ref int x) { x += 1; }

void f() {
	int val = 3;
	Inc(ref val);  // val will be 4
}

• calling a method on an object is just a shortcut for doing Class.function(ref this, <other parameters>)

out

• an alternative to ref
• useful for returning/setting multiple things in a function
• the CIL code is exactly the same, but the compiler checks, whether each out parameter has been assigned to (since the caller wouldn’t know about the state of the variable)
  • we must assign the parameters (at least dummy values)

void Read(out int first, out int next) {
	first = Console.Read();
	next = Console.Read();
}

void Main() {
	int first, next;
	Read(out first, out next);
}

• the variable can be declared as a part of the function call

int a;
if (tryParse(something, out a))
	Console.WriteLine("We failed: " + a);

Console.WriteLine("We didn't fail: " + a);

// is the same as

if (tryParse(something, out int a))
	Console.WriteLine("We failed: " + a);

Console.WriteLine("We didn't fail: " + a);

var

• derived at compile time, depending on what is on the right side
• if we can’t determine the type, the code won’t compile
  • var x;
  • var y = (1, 2, 3);
  • var z = null
• the declaration is a comment – it’s unwise to write var everywhere:
  • var name = GetName(); – what does it return?
  • var d = new List<int>(); – what if I want to change List to a HashSet later (and interface would thus be better to use)?

Interfaces

• a contract – we can assign any class that implements an interface to variables with an interface (when we only require the functionality of the given interface)
  • we can do the same for a struct, but it involves boxing

interface Shape {
	double Perimeter();
	double Area();
}

• can’t be instantiated (it has no code)
• not entirely an abstract class
  • classes/interface can implement multiple interfaces
  • classes can inherit only a single class
  • interfaces can’t have code, abstract classes can

Implementation details

Heaps and garbage collection

• two heaps; behavior of the garbage collector to the two heaps is different
• real limit is around $1.4\ GB$ (for 32-bit systems)
  • OutOfMemoryException could happen sooner, since we could have a lot of holes in the given heap and the new object wouldn’t fit in – heap fragmentation
  • happens easily when resizing (dynamic array, for example), since we’re creating a new array of twice the size, that has to co-exist with the old one for a bit
• segment – reservation (virtual memory); around $~16\ MB$
  • varies greatly on the architecture and GC configuration!
• kvantum – commit (physical memory); around $~8\ kB$
  • varies greatly on the architecture and GC configuration!
• GC is generational
  • gen 0 – allocated just now
  • gen $n$ – survived $n$ garbage collections
  • GC of a given generation checks the generation and all above
  • the next generations are checked only if the previous ones didn’t free too much memory
  • ephemeral segment – the current newest segment of the garbage collector
    • this is where GC happens
    • once it is full, a new one is added and the old one becomes generation 2 segment
• GC class – for interacting with the GC:
  • checking the generation of the object
  • checking the number of generation collections
  • forcing a collection of a given generation
  • should be the last resort, since it usually does things well
• GC $\implies$ no memory leaks is not true (to an extent…):
  • when a global cache used in classes doesn’t get collected
  • when objects subscribe to events and don’t unsubscribe
  • when an anonymous method references an attribute from an object

Large Object Heap (LOH)

• if it’s larger than $85\ 000\ B$
• usually reached when the object contains an array
• no heap compacting (don’t move objects when others die)

Small Object Heap (SOH)

• if it’s smaller than $85\ 000\ B$
• yes heap compacting (move after the end of garbage collection)

Terms to know

CIL

• common intermediate language
• formerly known Microsoft Intermediate Language (MSIL) or Intermediate Language (IL)
• intermediate language binary instruction set for the CLI specification
• executed by CLI-compatible runtime like CLR
• object-oriented, stack-based, typically JITted into native code

CLR

• run-time environment for the .NET framework
• responsible for running .NET programs, regardless of the language
• contains things like GC and JIT compiler

GAC

• global assembly cache – stores assemblies to be shared by computer applications

JIT

• just in time – when it’s needed
• the default unit is 1 method
• function calls are translated to calls to stubs
  • short calls that call JIT and fix references to the method in the code
• compilation is spaced into the running time of the program
  • can mess benchmarks up – make sure we’re testing when the method has already been JITted
• also takes care of optimizations, since optimizing CIL code doesn’t make much sense
  • means that they can’t be overly aggressive
• does method inlining (moves body of the method to where it’s called)
  • can’t always be done (recursive functions)
  • generally happens for smaller ($<32\ B$ of machine code) methods that aren’t too complicated (no try/catch, for example)

AOT

• ahead-of-time – before it’s needed
• ngen.exe – pre-JITting code
• more intensive optimizations
• JITting can still happen when it’s needed, though
• will be an important feature in .NET 5

BCL

• base class library
• types for built-in CLI data types, basic file access, collections, formatting,…

.NET

• BCL + CLR

Tiered compilation

• multiple compilations of the same method that can be swapped
  • initially quickly JITted but kinda slow
  • JITted properly after a number of calls to the method

CLS

• common language specification
• what a language must do in order to be .NET-compliant
• specifies things like minimum types:
  • sbyte, ushort are not compliant, so making them parameters of a function called from some other DLL might not be the best idea; on the other hand, implementation details are quite fine

typeof(Class)

• return the Type instance of the class
• useful when we’re comparing types of variables

nameof(x) [Microsoft Docs]

• useful when debugging, when showing exceptions to users…
• better to do than "x", since we can rename using any IDE easily

Console.WriteLine(nameof(System.Collections.Generic));  // output: Generic
Console.WriteLine(nameof(List<int>));                   // output: List
Console.WriteLine(nameof(List<int>.Count));             // output: Count
Console.WriteLine(nameof(List<int>.Add));               // output: Add

var numbers = new List<int> { 1, 2, 3 };
Console.WriteLine(nameof(numbers));        // output: numbers
Console.WriteLine(nameof(numbers.Count));  // output: Count
Console.WriteLine(nameof(numbers.Add));    // output: Add

NUnit tests

using System;
using NUnit.Framework;

public class Tests
{
	[SetUp]
	public void Setup() { /* do something for setup (optional) */ }
	
	[Test]
	public void NameOfTheTest()
	{
		Assert.AreEqual("a", "a");
		Assert.AreNotEqual("a", "b");
		Assert.IsTrue(true);
		Assert.IsFalse(false);
		
		Assert.Throws<ArgumentException>(() =>
		{
			This doesn't throw!
		});
	}
}

Data structures

Dictionaries

• using System.Collections.Generic;

Queues

Acknowledgements

• Adam Dingle, whose Programming 2 notes I’ve shamelessly copied (for certain examples)