overview

Overview

The job of the ViewPort is to coordinate the flow of information between the scenes and the drivers. Scenes and Drivers should not know anything about each other. An app should work identically from its point of view no matter if there is one, multiple, or no drivers currently running.

Drivers are all about rendering output and collecting input from a single source. Usually hardware, but can also be the network or files. Drivers only care about graphs and should not need to know anything about the logic or state encapsulated in Scenes.

The goal is to isolate app data and logic from render data and logic. The ViewPort is the connection between them that makes sense of the flow of information.

output

OUTPUT

Practically speaking, the ViewPort is the owner of the ETS tables that carry the graphs (and any other necessary support info). If the ViewPort crashes, then all that information needs to be rebuilt. The ViewPort monitors all running scenes and does the appropriate cleanup when any of them goes DOWN.

The scene is responsible for generating graphs and writing them to the graph ETS table. (Note: We also tried casting the graph to the ViewPort so that the table could be non-public, but that had issues)

The drivers should only read from the graph tables.

input

INPUT

When user input happens, the drivers send it to the ViewPort. Input that does not depend on screen position (key presses, audio window events, etc.) is sent to the root scene unless some other scene has captured that type of input (see captured input below).

If the input event does depend on position (cursor position, cursor button presses, scrolling, etc.) then the ViewPort needs to scan the hierarchical graph of graphs, to find the correct scene, and the item in that scene that was "hit". The ViewPort then sends the event to that scene, with the position projected into the scene's local coordinate space (via the built-up stack of matrix transformations)

captured-input

CAPTURED INPUT

A scene can request to "capture" all input events of a certain type. This means that all events of that type are sent to a certain scene process regardless of position or root. In this way, a text input scene nested deep in the tree can capture key presses. Or a button can capture cursor_pos events after it has been pressed.

If a scene has "captured" a position dependent input type, that position is projected into the scene's coordinate space before sending the event. Note that instead of walking the graph of graphs, the transforms provided in the input "context" field are used. You could, in theory change that to something else before capturing, but I wouldn't really recommend it.

Any scene can cancel the current capture. This would probably leave the scene that thinks it has "captured" the input in an inconsistent state, so this is not recommended.

dynamically-creating-view-ports

Dynamically Creating View Ports

Pass in the same set of opts that you would use when starting Scenic in your supervision tree. For example:

# Assuming you use the default view port name from the generator
{:ok, view_port} = Scenic.ViewPort.info(:main_viewport)

opts = [
  module: Scenic.Driver.Local,
  window: [resizeable: false, title: "My Example Scenic App"],
  on_close: :stop_system
]

{:ok, [opts]} = Scenic.Driver.validate([opts])
Scenic.ViewPort.start_driver(view_port, opts)