• Part of Tracking issue: Misc improvements to multi-viewports #3556

In short

You now almost never need to use eframe::Frame - instead use ui.input(|i| i.viewport()) for information about the current viewport (native window), and use ctx.send_viewport_cmd to modify it.

In detail

This PR removes most commands from eframe::Frame, and replaces them with ViewportCommand.
So frame.close() becomes ctx.send_viewport_cmd(ViewportCommand::Close), etc.

frame.info().window_info is now also gone, replaced with ui.input(|i| i.viewport()).

frame.info().native_pixels_per_point is replaced with ui.input(|i| i.raw.native_pixels_per_point).

RawInput now contains one ViewportInfo for each viewport.

Screenshots are taken with ctx.send_viewport_cmd(ViewportCommand::Screenshots) and are returned in egui::Event which you can check with:

ui.input(|i| {
    for event in &i.raw.events {
        if let egui::Event::Screenshot { viewport_id, image } = event {
            // handle it here
        }
    }
});

Motivation

You no longer need to pass around the &eframe::Frame everywhere.
This also opens the door for other integrations to use the same API of ViewportCommands.