Skip to content
Permalink
Branch: master
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
386 lines (335 sloc) 10.2 KB

Reko coding style document

This document formalizes the programming style conventions in use in the Reko project. When many contributors make changes to a shared codebase it becomes necessary to state the coding conventions in use to avoid unneccesary merge conflicts and disputes. None of these rules are absolute, but in general please adhere to these style rules unless you have a very good reason not to.

Formatting

Indentation in the source code is written with spaces, not tabs. Indentation stops are 4 spaces. Note: the Reko codebase is old and was originally written in C++ using tabs for indentation before being translated to C#. Some older sections of the sources still have tabs. To avoid pointless whitespace diffs in the code, those old code files have not been de-tabbified in bulk. However, if you are modifying one of these older files, prefer using spaces in the changed code.

Modern monitors make line lengths longer than 80 feasible. Still, very long lines are harder to read than shorter lines. If you're working with a line that is longer than 120 characters, consider either breaking it up into smaller subexpressions on separate lines, or at least fold the line into multiple lines.

Formatting rules are the default settings in Visual Studio. Some of these settings are repeated here for convenience.

Blocks starting with '{' are introduced on a separate line. Vertical space is no longer an issue now that that we have monitors capable of displaying more than 24 lines.

if (foo)
{
    bar();
}

// Please don't:
if (foo) {
    bar();
}

Spaces used to separate terms in expressions go 'inside' the expression, not outside:

a = b * c + (a / c);

// Please don't:
a = b*c + ( a/c );

No spaces after function applications, but spaces after C# keywords that are followed by parentheses:

myclass.MyMethod(foo);
if (foo.bar)  // ...
while (bar.baz()) // ...
return (a + b) * c;
return x;

// Please don't:
myclass.MyMethod (foo);
if(foo.bar) // ...
while( bar.baz ()) // ...
return(a+b)*c;
return (x);

Statements inside the substatements of if, while, do-while, and for statements are always enclosed in blocks, even if there is only a single statement. The only exceptions are if the single statemet is a control transfer instruction like return, break, continue, goto or throw or the empty statement of a while loop:

if (foo == null) 
{
    bar();
}
while (bar.foo())
{
    foo.bar();
}

// But...
if (foo == null)
    throw new ArgumentException(nameof(foo));
if (bar != foo)
    return false;
while (!done())
    ;

// Please don't:
if (foo == null)
    bar();
while (foo.bar()) bar.foo();
while (!done()) {}

Organization

C# classes are written as follows:

public class MyClass
{
    // Event definitions
    public event EventHandler FooChanged;
    public event EventHandler BarChanged;

    // Static and constant fields
    public const int MaxValue = 0x1234;
    public static readonly Dictionary<int, string> Lookups;

    // Member fields
    private string foo;
    private string bar;

    // Constructors
    public MyClass(string value)
    {
        this.foo = value;
    }

    // Properties

    public string Foo { get { return this.foo; } }

    // Methods:
    // Public methods first, then internal, protected, and finally private.
    public IExpansion FindExpansion(string expansionName) // ...

    protected IExpansion DoFindExpansion(
        string expansionName, 
        IEnumerable<IExpansion> items) // ...

    private bool IsValidExpansion(IExpansion expansion) // ...
}

Naming

We use camel-case consistently, unless using symbols from external sources that don't comply with this rule. Private C# fields and local variables are written with an initial lower case letter:

private int xRange;

Avoid using prefixes or suffixes to distinguish member varaibles from local variables. Use "this." to disambiguate:

private int xRange;
public void SetXRange(int xRange) 
{
    this.xRange = xRange;
}

// Please don't:
private int _xRange;
private int m_xRange; 

public void SetXRange(int xRange)
{
    m_xRange = xRange;
}

Public members are written with initial upper case letters, as are class and namespaces:

public int XRange { get { return xRange; } }
public virtual void Reset() {}
public class Modulator // ...

Prefer making member variable names longer and self-documenting and avoid cryptic abbreviations. Method parameters and local variables can be shorter, especially index or enumeration variables, but not cryptic:

public class Accumulator
{
    private int totalSum;
    private IEnvironment environment;

    public TheClass(IEnvironment env)
    {
        this.environment = env;
        this.totalSum = 0;
        foreach (var item in env.Items)
        {
            this.totalSum += item;
        }
    }
}

// Please don't:
public class Acc
{
    private int n;
    private IEnvironment e;

    public Acc(IEnvironment e)
    {
        // ...
    }
}

Well-established acronyms are fine:

public void RenderHtml()
{
    // ...
}

// Please don't:
public void RenderHyperTextMarkupLanguage()
{
    // ...
}

When defining a generic type with a single type argument, use the name T:

public class ProgramCollection<T> // ...

If it has more than one type argument, prefix the argument types with T:

public class EfficientDictionary<TKey, TValue> // ...

Usage of the var keyword in assignments

When it is clear what the type of the expression being assigned is, prefer using var, especially when dealing with generic types:

var dict = new Dictionary<Procedure,List<Block>>();
var binExpr = expr as BinaryExpression;

// Please don't
Dictionary<Procedure,List<Block>> dict = new Dictionary<Procedure,List<Block>>();
BinaryExpression binExpr = expr as BinaryExpression;

Methods

Try to keep methods small enough that they can fit in a page. It is harder to understand larger methods. In particular, avoid deeply nested statements; refactor them into smaller methods instead.

Resist the temptation to add comments to a large method to clarify it. The need for those comments suggests that you might need to break the method apart into smaller smaller methods; use the comments as the names for the smaller methods:

void LargeMethod()
{
    // INitialize the state
    /* Large amounts of code elided for clarity */

    // Process input
    /* Large amounts of code elided for clarity */

    // Validate state
    /* Large amounts of code elided for clarity */

    // Clean up
    /* Large amounts of code elided for clarity */
}

can be refactored to become:

void LargeMethod()
{
    InitializeState();
    ProcessInput();
    ValidateState();
    CleanUp();
}

Prefer "quick" exits from methods using return over nested if statements:

int Method(int arg1, float arg2)
{
    if (arg1 < 0)
        return 0;
    if (arg2 < 0.0F)
        return 1;
    //...
}

// Please don't:
int Method(int arg1, float arg2)
{
    if (arg1 < 0)
    {
        return 0;
    }
    else
    {
        if (arg2 < 0.0F)
        {
            return 1;
        }
    }
    //...
}

Unit tests names are a little different. Due to limitations in the NUnit test runner extension in VisualStudio, individual test methods should be named after the following schema:

{abbreviated class name}_{method being tested}_{special circumstances being tested}

E.g.

[Test]
public void X86DRw_mov() { /* ... */ }
public void X86DRw_call_importedFunction() { /* ... */ }

Comments

Comments are tricky. We want them to explain the code without pointlessly duplicating the code itself. Comments are not checked by the compiler, and they can grow stale when the code they are commenting is changed but the comments themselves are not:

// The following two parameters must be no larger than 100
public int AddTogether(params int[] args) // confusion????

Making an effort to use clear variable and function names helps in this regard. Well chosen names alleviate the need for comments:

public int SumOf(params int[] summands)

Prefer writing comments that explain "why", not "what".

int x = 1;                // Lua arrays start at offset 1.
// or even better:
const int LuaStartIndex = 1;
int x = LuaStartIndex;

// Please don't:
int x = 1;                // Set x to 1

Clarifying subtle details in algorithms with a comment is encouraged:

// We use the Lengauer-Tarjan algorithm to compute the dominator tree
// because its performance has been measured and been found best for
// code graphs typically found in binaries.

However, if you wish to document that the code is making assumptions about the current state of the program, consider using Debug.Assert or throwing an InvalidOperationException instead:

Debug.Assert(x != null, "x should be initialized before starting the frobulation process");
// or:
if (x != null)
    throw new InvalidOperationException("x should have been initialized before starting the frobulation process.")

// Please don't:

// x should be initialized here.

See here for more examples: Putting comments in code: the good, the bad, and the ugly

Exceptions

We throw exceptions for a few things. One is to catch programmer errors, typically when checking input parameters to constructors, methods, and property setters:

public void ReplaceWidget(Widget w)
{
    if (w == null)
        throw new ArgumentNullException(nameof(w));
    if (state == null || !state.Ready())
        throw new InvalidOperationException("A call to ReplaceWidget is not correct at this point.");
    // ...
}

We do not use exceptions to signal the end of an enumeration (I'm looking at you, Python!)

We do not use exceptions to perform non-local control flow. If you want setjmp/longjmp, you know where to find them:

// Please don't:
public bool CleverSearchAlgorithm<T>(Tree<T> tree, T value)
{
    try 
    {
        CleverImpl(tree, value);
    }
    catch (FoundItemException e)
    {
        return true;
    }
    return false;
}

private void CleverImpl<T>(Tree<T> tree, T value)
{
    if (t.Value == value)
        throw new FoundItemException();
    CleverImpl(tree.Left, value);
    CleverImpl(tree.Right, value);
}
You can’t perform that action at this time.