Optional Parameters

It's sometimes useful to define functions and methods with parameters that are optional. An optional parameter is one that a caller can include in a call to the function, but doesn't have to.

For a normal function or method, a caller must supply the same number of arguments that are specified in the definition of the routine. The system checks that the number of argument values supplied by the caller matches the number of parameter variables expected by the function definition, and throws an error if there's a mismatch.

When you define a function or method with optional parameters, though, you're telling the system that it's okay for a caller to omit one or more of the argument values. If the caller supplies a value for an optional parameter, the value is assigned to the parameter variable as usual. But if the caller omits a value for the corresponding argument position, the system doesn't throw an error; instead, it simply sets the parameter variable to a default value instead.

Optional parameters are useful in situations where you expect that the majority of calls to the function will use the same value for that argument. Rather than forcing those typical callers to type out that same value in every call, you can make the parameter optional, so that typical callers who just want the default can omit the value from the function call syntax. At the same time, the few cases where a different value is needed can still specify it just by adding the extra argument.

Declaring optional parameters

To make a parameter optional, simply put a question mark "?" after the parameter name in the function or method definition:

f(a, b?)
  "This is f: a=<<a>>, b=<<b>>\n";

This declares "a" as a regular parameter, and "b" as optional. Callers must supply either one or two arguments when calling f(): a value for "a" must always be supplied, but "b" can be omitted if desired.

When a caller omits an optional parameter, the system automatically assigns nil as the default value. Consider the following calls to f():

f(2, 3);

That'll produce this output:

This is f: a=1, b=
This is f: a=2, b=3

In the first call, we supplied only one argument value. This is assigned to "a", since it's the first parameter. Since there's no second argument value, the optional second parameter "b" gets the default value of nil (which is why we print out "b=" with no value - nil is displayed simply as empty when printed out). In the second call, we supplied two argument values, so "a" gets the first value and "b" gets the second.

The system matches up argument values to parameter names positionally, from left to right. (This doesn't apply to named arguments which are assigned explicitly by name; we'll see more about those below.) When a function has multiple optional parameters, they're assigned in left to right order.

f2(a?, b?, c?)
  "This is f2: a=<<a>>, b=<<b>>, c=<<c>>\n";

  f2(2, 3);
  f2(4, 5, 6);

Here's what the code above prints out:

This is f2: a=, b=, c=
This is f2: a=1, b=, c=
This is f2: a=2, b=3, c=
This is f2: a=4, b=5, c=6

As you'd expect, when the call has no argument values at all, the parameters are all set to nil, so they print out as empty. When one argument value is supplied, it's assigned to "a", since it's the first (leftmost) parameter in the list. When two are supplied, they go to "a" and "b".

Mixing normal and optional parameters

As we've seen, it's fine to use a mix of optional and normal parameters in the same function definition. But there's an important rule about the ordering: All parameters after the first optional parameter must also be optional.

The reason for this rule is that there'd be too much ambiguity in some cases without it. For example, let's suppose that you could define a function like this (you can't, because of the rule, but imagine for a moment we could):

g(a?, b, c?) { }  // illegal, because normal parameter b follows optional a

If you called this with g(1), it's fairly clear that "b" would be set to 1, since it's the only required variable. But how should we handle g(1, 2)? The most obvious handling would probably be to set a=1 and b=2, but this would mean that the first value isn't always assigned to "a" - in g(1) it's assigned to "b". It's also possible to think of other orderings that make their own kind of sense, such as setting b=1 and c=2 to preserve their relative order, or b=1 and a=2 on the theory that b should always get first dibs, being the one required parameter.

In any case, the compiler sidesteps the whole problem by disallowing this type of mixing. Optional parameters can only be followed by more optional parameters, so argument values are always assigned to parameter variables in a simple left-to-right order.

Declaring default parameter values

When you use the "?" suffix to mark a parameter as optional, you're also implicitly saying that the default value for the variable is nil when the caller doesn't supply an explicit value for it. But what if you want to use a different default?

The obvious approach might be to compare the value to nil, since that's the default that's assigned when the argument isn't supplied. If the value is nil, we'd set the variable to the default value we really wanted:

  if (a == nil)
    a = 'a default value';

  // ... 

But there's a problem: what if the caller wants to explicitly pass in nil as the value for "a"?


The approach we just took would make this impossible, because the function can't distinguish the case where the caller omits "a" entirely from the case where the caller explicit passes in nil as the value for "a".

A better approach would be to check argcount to test whether or not the caller included the argument:

  if (argcount < 1)
    a = 'a default value';

  // ...

That works, but it's what's known as brittle code - brittle in that it breaks if you bend it. The problem is that if you rearrange the argument list (by inserting another argument before "a", say), you have to remember to fix the argcount test to account for the change.

Fortunately, there's a better way, which doesn't involve a separate argcount test. The compiler provides some special syntax just for setting up custom default values. Rather than using the "?" suffix to define an optional parameters, you can instead use the special default value syntax, "= expression". Using this syntax, we'd redefine our example above like this:

h(a = 'a default value')
  // ...

This says that "a" should be set to the string 'a default value' only if the caller didn't supply a different value for it. If the caller does provide a value - even if the value is nil - "a" keeps the caller's value rather than the default.

Note that setting a default value with the "= expression" syntax also makes the parameter optional, as though it had the "?" suffix. You can't use both suffixes on the same variable, since there's no way for a parameter with a default not to be optional. This means that every parameter following a default-value parameter has to be optional, because of the ordering rule we saw earlier.

How defaults are assigned

A default value expression can be any valid expression, constant or non-constant. If it's not a constant, it's evaluated each time the function is called without an argument value for the parameter, immediately upon entry to the function. The expression has access to the other parameter variable names, but be aware that the optional parameters are initialized in left-to-right order. This means that if you have two default value expressions, the left one won't be able to access the value of the right one. It's not an error to do so, but the value will simply be nil. For example:

h2(a = 'b=<<b>>', b = 'a=<<a>>')
   "This is h2: a=<<a>>, b=<<b>>\n";

This looks confusingly circular: "a" uses the value of "b" and "b" uses the value of "a". But it's not an error, because of the simple left-to-right initialization rule. Here's how this is resolved. First, the system sets "a" and "b" to nil. Next, it initializes the optional parameters in left-to-right order. So we start with "a". If the caller supplied a value for "a", it's assigned to "a", otherwise we evaluate the default value expression 'b=<<b>>'. Since we haven't gotten around to "b" yet, it still has its initial value nil, so "a" is set to the string 'b='. We then move on to "b", assigning the value passed by the caller, or the default value expression 'a=<<a>>'. Since we've already finished setting up "a", this will expand to 'a=b=' if there was no caller value for "a", or 'a=x' if the caller supplied 'x' as the value for "a".

A default value expression is evaluated only when it's needed, which is to say, when the caller omits the corresponding argument. This is important when the expression has side effects, such as displaying a message. Any side effect will be triggered only on calls where the default value is actually needed.

Declaring named parameters as optional

Named arguments can be made optional in the same way as positional parameters. Simply put the ? suffix or the =default value assignment after the colon ":" suffix for the named argument:

f3(a, b:?, c: = 'c default')
  "This is f3: a=<<a>>, b=<<b>>, c=<<c>>\n";

This defines a function with one required positional argument "a", and two optional named arguments "b" and "c". Since "b" is marked as optional with the "?" suffix, it will be set to nil if the caller doesn't supply a value for it; "c", on the other hand, has an explicit default value expression that will be used if the caller omits a value. So the following series of calls:

f3(2, b:3);
f3(4, c:5);
f3(6, c:7, b:8);

...will produce this output:

This is f3: a=1, b=, c=c default
This is f3: a=2, b=3, c=c default
This is f3: a=4, b=, c=5
This is f3: a=6, b=8, c=7

Ordering for positional arguments

Remember the rule that an optional parameter can only be followed by more optional parameters? Well, there's an exception. It doesn't apply to named arguments.

This is because named arguments are effectively separate from the positional list, despite their syntactic commingling. The caller and callee both refer to these variables explicitly by name, so their positional order doesn't matter. So, it's fine to follow an optional positional parameter with a required named argument, and it's fine to follow an optional named argument with a required positional one (so long as there aren't any earlier optional positional parameters, of course).

Optional vs. varying parameters

Optional parameters are similar to the "..." syntax for varying argument lists, but not quite identical.

The "..." syntax says that a function can take any number of additional arguments beyond the ones explicitly named in the definition. This is extremely useful for situations where the number of additional arguments is truly unpredictable. For example, if you're setting up a proxy method that calls another method that's determined at run-time, you obviously can't know in advance how many arguments will be needed for the other method; "..." handles this by accepting whatever arguments are actually passed in. Similarly, a function that takes a message string and a series of substitution parameters would need "...", because the number of substitutions might be different in every message string.

Optional parameters, in contrast, are best for cases where you have a fixed number of parameters, but where one (or more) of the parameters will usually have a certain predictable value. That is, callers will almost always pass that same argument value for the parameter, except perhaps in a few special cases. For these situations, it's nice to let callers omit the argument value entirely when it's going to be that typical value. This saves a little typing on the calling side for the common invocation, while still letting callers supply a non-default value when needed.

By the way, it's legal to use the "..." syntax in a parameter list that has optional arguments. Everything works as normal: you find out how many parameters are present using argcount, and you access the unnamed parameters beyond the "..." using getArg().