Context parameters can be declared by-name to avoid a divergent inferred expansion. Example:

trait Codec[T]:
  def write(x: T): Unit

given intCodec: Codec[Int] = ???

given optionCodec[T](using ev: => Codec[T]): Codec[Option[T]] with
  def write(xo: Option[T]) = xo match
    case Some(x) => ev.write(x)
    case None =>

val s = summon[Codec[Option[Int]]]

s.write(Some(33))
s.write(None)

As is the case for a normal by-name parameter, the argument for the context parameter ev is evaluated on demand. In the example above, if the option value x is None, it is not evaluated at all.

The synthesized argument for a context parameter is backed by a local val if this is necessary to prevent an otherwise diverging expansion.

The precise steps for synthesizing an argument for a by-name context parameter of type => T are as follows.

1. Create a new given of type T:

   given lv: T = ???
   

   where lv is an arbitrary fresh name.

2. This given is not immediately available as candidate for argument inference (making it immediately available could result in a loop in the synthesized computation). But it becomes available in all nested contexts that look again for an argument to a by-name context parameter.

3. If this search succeeds with expression E, and E contains references to lv, replace E by

   { given lv: T = E; lv }
   

   Otherwise, return E unchanged.

In the example above, the definition of s would be expanded as follows.

val s = summon[Test.Codec[Option[Int]]](
  using optionCodec[Int](using intCodec)
)

No local given instance was generated because the synthesized argument is not recursive.