Named parameters, also known as keyword arguments, have several advantages over positional arguments:

• convenience: they can be supplied in arbitrary order
• readability: their names serve as documentation at call site
• safety: it is impossible to accidentally mix them up

Consider a function to replace a substring with another string:

Text.replace path "$HOME" "/home/username/"

We want to replace references to the $HOME environment variable with a concrete directory. There is but one problem – we have supplied the text arguments in the wrong order.

Compare that to a newtype-based solution:

Text.replace
  (Needle "$HOME")
  (Replacement "/home/username/")
  (Haystack path)

Now that the function requires each argument to be wrapped in a newtype, we cannot mix them up – the compiler will report an error, and newtype constructors serve as documentation.

The problem with newtypes is that it is bothersome to create them for each parameter, they pollute the global namespace, and we still cannot supply wrapped arguments in arbitrary order.

With keyword arguments, none of that is a problem:

Text.replace ! #haystack path
             ! #needle "$HOME"
             ! #replacement "/home/username/"

Functions can declare their parameter names in pattern bindings:

replace (arg #needle -> n) (arg #replacement -> r) (arg #haystack -> h) =
  ...

Types are inferred, but it is possible to specify them. When the parameter names are specified in the type signature, they can be omitted from the pattern bindings:

replace ::
  "needle"      :! Text ->
  "replacement" :! Text ->
  "haystack"    :! Text ->
  Text
replace (Arg n) (Arg r) (Arg h) =
  ...

Keyword arguments have seamless interoperability with positional arguments when the function takes them last. Consider this function:

foo :: A -> B -> "x" :! C -> IO ()

There are several ways to invoke it:

(foo a b) ! #x c     -- parentheses for clarity
(foo a ! #x c) b     -- parentheses required
(foo ! #x c) a b     -- parentheses required

We can also supply keyword arguments using the with combinator instead of the ! operator:

(with (#x c) foo) a b  -- parentheses for clarity
with (#x c) (foo a b)  -- has the same effect

Both ! and with work in a similar manner: they traverse the spine of the function and supply the first keyword argument with a matching name.

For example:

bar             :: "x" :! A -> "y" :! B -> IO ()
bar ! #y b      :: "x" :! A             -> IO ()
with (#y b) bar :: "x" :! A             -> IO ()

There is also support for optional parameters. A function can specify default values for some of its arguments:

log ::
  "message"  :! Text ->
  "severity" :? Severity ->
  "handle"   :? Handle ->
  IO ()
log (arg    #message          -> msg)
    (argDef #severity Error   -> sev)
    (argDef #handle   stderr  -> hndl)
  = ...

Optional parameters are denoted with (:?) instead of (:!). Instead of arg to match on them, we must use either argDef to provide a default value or argF to get a value wrapped in Maybe (Just when the parameter was specified, Nothing when omitted).

At call site, optional parameters are passed using the same (!) operator:

log ! #message "All your base are belong to us"
    ! #severity Info
    ! #handle stdout

To use the default values for all unspecified optional parameters, we can pass defaults to the function:

log ! #message "Could not match type Int with type Bool"
    ! defaults

log ! #message "The password must contain a letter, \
               \a digit, and a plot twist"
    ! #severity Warning
    ! defaults

We can also pass defaults using with, which has the same effect as the (!) operator:

with defaults $
  log ! #message "Connection interrupted"
      ! #handle logfile