<xs:complexType name="switchType">
		<xs:sequence>
			<xs:element name="variable" type="variableType"/>
			<xs:element name="case" type="caseType"/>
			<xs:element name="default" type="caseType"/>
		</xs:sequence>
	</xs:complexType>
	<xs:complexType name="caseType">
		<xs:sequence>
			<xs:element name="constant" type="constantType"/>
			<xs:element name="block" type="blockType"/>
		</xs:sequence>
	</xs:complexType>

It may be better to change this as described here.

It seems to me this requires two node types ‘switch’ and ‘case’ because ‘default’ can be encoded as a special type of ‘case’.

But what should be the structure? I get the impression that Java and C# are defined in a slightly different way? In Java cases are like labels and if a break is not encountered then code execution just carries on past the label, but in C# cases are more like nodes?

Similarly I think we could encode the structure in two different ways (although I think the syntax independent of the semantics)

To explain what I mean I will use the following pseudo code:

switch (variable) {
case 0:
StatementA;
StatementB;
case 1:
StatementC;
default:
StatementD;
StatementE;
}

The first way to encode this would be to put ‘case’ nodes and all statements directly under the ‘switch’ node. If this is done then ‘case’ nodes would not have any child subnodes, this seems to me to give the cases the same sort of structure as labels. This is the way that the program is parsed at the moment (version 0.9).

The second way to encode this would be to put only ‘case’ nodes under the ‘switch’ and to put StatementA and StatementB under case 0, StatementC under case 1, StatementD and StatementE under default.

I seems to me that the second is a better structure and although there is a sort of implication that the statements belong to the cases it does not prevent code from flowing between cases.