Another example API (less complete) based on the beam model where everything is streams and tables

PrototypeAPIStream.scala

  // Following model based on the description of 
  // Apache Beam in the book Streaming Systems by Reuven Lax; Tyler Akidau; Slava Chernyak

  /**
    * Tables are data at rest, and act as a container for data to accumulate and be observed over time.
    * K = Key
    * V = Value
    * W = Window
    * P = Partition
    */
  type Table[K, V, W, P]

  /**
    * Streams are data in motion, and encode a discretized view of the evolution of a table over time.
    * K = Key
    * V = Value
    * W = Window
    * P = Partition
    */
  type Stream[K, V, W, P]

  /**
    * Applying nongrouping operations to a stream alters the data in the stream while leaving them in motion,
    * yielding a new stream with possibly different cardinality.
    *
    * Stream -> Stream
    */
  trait NonGroupingOperations {
    def map[K, V1, V2, W, P](stream: Stream[K, V1, W, P])(f: V1 => V2): Stream[K, V2, W, P]
    def filter[K, V, W, P](stream: Stream[K, V, W, P])(f: V => Boolean): Stream[K, V, W, P]
    def assignWindow[K, V, W1, W2, P](stream: Stream[K, V, W1, P])(window: W2): Stream[K, V, W2, P]

  }

  /**
    * Grouping data within a stream brings those data to rest, yielding a table that evolves over time.
    *
    * - Windowing incorporates the dimension of event time into such groupings.
    *
    * - Merging windows dynamically combine over time, allowing them to reshape themselves in response to the data
    *   observed and dictating that key remain the unit of atomicity/parallelization, with window being a child
    *   component of grouping within that key.
    *
    * Stream -> Table
    */
  trait GroupingOperations {

  }

  /**
    * Triggering data within a table ungroups them into motion, yielding a stream that captures a view of the table’s
    * evolution over time.
    *
    * - Watermarks provide a notion of input completeness relative to event time, which is a useful reference point
    *   when triggering event-timestamped data, particularly data grouped into event-time windows from unbounded
    *   streams.
    *
    * - The accumulation mode for the trigger determines the nature of the stream, dictating whether it contains
    *   deltas or values, and whether retractions for previous deltas/values are provided.
    *
    * Table -> Stream
    */
  trait UnGroupingOperations {

  }