FEB
22
2004

Playing with the Switch and Foreach Statements

I've been thinking about C's control statements (if, while, switch, for etc) for a little while, and I think there's some room for improvements in the later two...


switch

The C# guys have made two interesting modifications to C's switch statement: fall-through
is only allowed when there's no statement after the case (thus if a statement
follows a case, there must be a break or similar jump statement).
And it added the 'goto case' statement that can be used for
fall-through effects. Here's a C# snippet:

	void test(int a) {
		switch (a) {
		case 0:
			Console.WriteLine("blabla");
			break;
		case 1:
			Console.Write("More ");
			goto case 0;
		case 2:
		case 3:
			Console.WriteLine("lalala");
			break;
		}
	}

I think forbidding implicit fall-through is a good idea, it avoids
a common source of error. But if you forbid it, why is it still necessary
to write the redundant break statements for each case? Especially long
switch statements look much better without them:

	void test(int a) {
		switch (a) {
		case 0:
			Console.WriteLine("blabla");
		case 1:
			Console.WriteLine("More");
			goto case 0;
		case 2:
		case 3:
			Console.WriteLine("lalala");
		}
	}

Something that I am missing all the time is a way to describe value
ranges in a switch statement. The
constraint syntax from my
last entry
offers a neat solution for it:

	void test(int a) {
		switch (a) {
		case <0:
			System.out.println("a is negative");
		case 0:
			System.out.println("a is 0");
		case >0, <=100:
			System.out.println("a is between 1 and 100");
		case <200:
		case >500, <600:
			System.out.println("a is >100 and <200, or >500 and <600");
		default:
			System.out.println("a has some other value");
		}
	}

Allowing comma-separated lists of operator plus constant makes switch much more
powerful for many purposes.




foreach

Over the last years most C-based languages got a foreach statement to
iterate over collections and arrays, but every language got its own
syntax. This is what it looks like in C#:

	int sum = 0;
	foreach (int i in someArray)
		sum += i;

... and ECMAScript (JavaScript):

	var sum = 0;
	for (var i in someArray)
		sum += i;

... and Java 1.5:

	int sum = 0;
	for (int i: someArray)
		sum += i;

I definitely prefer Java's syntax as 'for' is shorter than 'foreach'
(without reducing readability) and I don't like making in
a keyword.

Unfortunately neither Java nor C# exploit their foreach as a nicer
alternative for regular for loops. The common C pattern

	for (int i = 0; i < 100; i++)
		System.out.println("Bla");

can be expressed with a foreach loop and a special object that
provides a sequence of integers, similar to
Python's range() function:

	for (int i: new Range(100))
		System.out.println("Bla");

It's looking nicer without the new keyword, as shown here:

	for (int i: Range(100))
		System.out.println("Bla");

Using other constructors it would be possible to specify at start value, to
go backwards or use bigger steps:

	for (int i: Range(50, 10, -1))
		System.out.println("i=" + i);

As for loops are pretty common, it may make sense to have a special
operator for simple ranges, like
Ruby does and
Project Alpha proposes:

	for (int i: 0...99)
		System.out.println("i=" + i);

The expression 'a...b' is short for '(a<=b)?Range(a, b+1):Range(b, a-1, -1)'.
Ranges can also be used for other purposes, for instance as function
arguments. With Range and an overloaded indexer iterating through the
first ten members of a list would be as easy as:

	void print10Numbers(List numberList) {
		for (int i: numberList[0...9])
			System.out.println("i=" + i);
	}

Unlike regular methods for slicing collections such as Java's List.subList()
this syntax gives more flexibility. You could get the first ten numbers in reverse order (numberList[9...0]), skip every second number (numberList[Range(0, numberList.size(), 2)]) and so on.