I remember the horror when I first came across implicit in scala, it sounded like a recipe for incomprehensible code. After a while I started to realize that there are a number of use cases where it could be very useful. An important example is method injection as it is discussed in the learning scalaz tutorial.

Essentially, with the combination of type class (in scala’s case just traits) and implicit, we can “create” new methods for existing data types without touching its source code. In languages like Haskell, this can be achieved by creating type class instance for a given data type, but it has the limitation of global uniqueness of instances, meaning that in a fully compiled program there can only exist at most one instance of a type class for a given data type. With the implicit approach, scala can have local type class instances, therefore different instances of a type class for the same data type. It can be pretty convenient to change behavior of a data type by just importing different implicits definitions from different packages.