Agfdhyk

Is there a term for this technique? One prominent example is the WinAPI: SendMessage( hwnd, msg, info1, info2 ) where parameters #3 and #4 only make sense per msg (which also means there are cases when only one or none of those two parameters are needed). See MSDN.

Rephrased: having an all-purpose function that always accepts multiple arguments, but interpreting them depends on a previous argument. I don't want to talk about open arrays, open arguments, typeless arguments... I know all that. That's not what I'm asking - I want to have the term for this type of functions (any maybe also how unspecific parameters are called).

This is not about casts or passing by reference - the parameter types are always the same. Other example: calculate( char operation, int a, int b ) which is then used as

1. 
   calculate( '+', 2, 5 ) (parameters #2 and #3 are summands)


2. 
   calculate( '/', 4, 2 ) (parameter #2 is the divident and parameter #3 is the divisor)


3. 
   calculate( '!', 3, 0 ) (parameter #2 is the factorial and parameter #3 is unused)

In all these cases the data type is always the same and never casted. But the meaning of parameters #2 and #3 differ per parameter #1. And since this is the case it is difficult to give those parameters a meaningful name. Of course the function itself most likely uses a switch(), but that is not subject to my question. How are parameters #2 and #3 called, where a distinct name cannot be found, but data types are always the same?

The fact the msg argument "changes" the parameters is through a simple switch statement. Each "msg" in the switch knows the parameters(with type) needed and casts them appropriately.

This "technique" is called passing by reference, or passing by address. The latter is usually used for method pointers.

There is no Special name if that is what you are asking. It is a regular function, method or procedure.

The referenced Function is a Win32 API Function, which may be referred to as a "Windows function call."

This is an example of a static Parameter and multiple Dynamic parameters.
The static is the "msg" and the dynamic is described as the following:

These parameters are generic pointers. Passed by reference. They can point to any data type or no value, ie null pointer. It is up to the sender to lock the memory in place, and the receiving method to interpret the pointer correctly (through pointer casts).

This is an example of typeless argument passing. The only thing passed is a memory address. It is dangerous since the types passed must be agreed upon ahead of time(by convention and not contract as with a typed language construct) and must match on both sides of the call.

This was common before C++, in the C days, we only had C structs to pass around. Leading to many General Fault Protection errors. Since then, typed interfaces mostly have replaced the generic equivalents through libraries. But the underlying Win32 methods remain the same. The main substantial change since its' inception is the acceptance of 64-bit pointers.

Although not widely supported, what you are referring to would be a dependently typed function (or dependently typed parameters).

To quote wikipedia on dependent types

A "pair of integers" is a type. A "pair of integers where the second is greater than the first" is a dependent type because of the dependence on the value.

The parameters could have a type that depends on a value. The type of info1 depends on the value msg as does info2.

In order to make this approach work in a language without dependent types, the dependent parameters are given a very generic type that is only refined later on when more information is available. When the type of msg becomes known (at runtime) only then is are the types of info1 and info2 assumed. Even though the language doesn't allow you to express this dependency, I would still call the approach a dependently type one.