| 1 | import Foundation |
| 2 | |
| 3 | // --------------------------------------------------------------- |
| 4 | // Screen share, receiving (spec §5.4) |
| 5 | // --------------------------------------------------------------- |
| 6 | // |
| 7 | // The web client is the contract here, and this file is its receiving |
| 8 | // half ported rule for rule: a sharer adds its display as one `sendonly` |
| 9 | // transceiver of its own, in a stream of its own, announces it with a |
| 10 | // `share {on:true}` broadcast, and stops it by stopping the transceiver |
| 11 | // and broadcasting `share {on:false}`. Nothing on the wire says which |
| 12 | // m-line is the screen. Each receiver works it out, and a native app and |
| 13 | // a browser looking at the same sender have to reach the same answer, or |
| 14 | // one of them captions a face as a screen. |
| 15 | |
| 16 | /// What one arriving track is to the person looking at it. |
| 17 | enum TrackRole: Equatable { |
| 18 | case audio, camera, screen |
| 19 | } |
| 20 | |
| 21 | /// What a connection has recorded about the peer on the other end: the |
| 22 | /// mid and stream its camera came in on, the stream everything else from |
| 23 | /// it has arrived in, and whether it has said it is sharing. room.js's |
| 24 | /// `entry` fields of the same names. |
| 25 | struct ScreenClues: Equatable { |
| 26 | var cameraMid = "" |
| 27 | var cameraStream = "" |
| 28 | var peerStream = "" |
| 29 | var sharing = false |
| 30 | } |
| 31 | |
| 32 | /// One arriving track, as the classifier needs it. `direction` is the |
| 33 | /// transceiver's negotiated direction when there is one and the direction |
| 34 | /// it was built with when there is not, spelled the way SDP spells it. |
| 35 | struct ArrivingTrack: Equatable { |
| 36 | var kind: String |
| 37 | var streamId = "" |
| 38 | var mid = "" |
| 39 | var direction = "" |
| 40 | } |
| 41 | |
| 42 | /// Tells a peer's screen from its camera. room.js's classifyTrack, in |
| 43 | /// the same order and for the same reasons, which are set out there at |
| 44 | /// length; the short form: |
| 45 | /// |
| 46 | /// 1. A video track arriving `recvonly` on a connection that sends a |
| 47 | /// camera is on an m-line the far side added, and the only one it adds |
| 48 | /// is the screen. |
| 49 | /// 2. The camera's own mid, once one is known. |
| 50 | /// 3. The camera's stream id, for a sender that reports no mid. |
| 51 | /// 4. The stream the peer's microphone came in: a sender puts its camera |
| 52 | /// there and its screen somewhere else. |
| 53 | /// |
| 54 | /// And when none of those can speak, the `share` message breaks the tie. |
| 55 | /// It is last on purpose: it says the peer is showing a screen somewhere, |
| 56 | /// not that this track is it. |
| 57 | /// |
| 58 | /// `hasVideo` is whether this connection carries a camera of ours. The |
| 59 | /// apps always add one, so rule 1 is the rule that decides in practice; |
| 60 | /// the rest are here because a browser on the far side may not have one, |
| 61 | /// and a port that dropped them would disagree with the browser the day |
| 62 | /// that mattered. |
| 63 | func classifyTrack(_ known: ScreenClues, _ track: ArrivingTrack, hasVideo: Bool) -> TrackRole { |
| 64 | if track.kind != "video" { return .audio } |
| 65 | if track.direction == "recvonly" && hasVideo { return .screen } |
| 66 | if !known.cameraMid.isEmpty && !track.mid.isEmpty { |
| 67 | return track.mid == known.cameraMid ? .camera : .screen |
| 68 | } |
| 69 | if !known.cameraStream.isEmpty && !track.streamId.isEmpty { |
| 70 | return track.streamId == known.cameraStream ? .camera : .screen |
| 71 | } |
| 72 | if !known.peerStream.isEmpty && !track.streamId.isEmpty { |
| 73 | return track.streamId == known.peerStream ? .camera : .screen |
| 74 | } |
| 75 | if known.cameraMid.isEmpty && known.cameraStream.isEmpty { |
| 76 | return known.sharing ? .screen : .camera |
| 77 | } |
| 78 | // A camera is known but there is nothing on this track to compare with |
| 79 | // it: a second video track is the screen. |
| 80 | return .screen |
| 81 | } |
| 82 | |
| 83 | /// The words under a screen. The browser's, character for character — |
| 84 | /// including the typographic apostrophe — so a person moving between the |
| 85 | /// app and a browser tab reads the same caption on the same screen. |
| 86 | public func screenCaption(_ name: String) -> String { name + "’s screen" } |
| 87 | |
| 88 | /// One thing the main stage can show: a person's camera or their screen. |
| 89 | public enum StageItem: Hashable, Sendable { |
| 90 | case camera(String) |
| 91 | case screen(String) |
| 92 | |
| 93 | public var peer: String { |
| 94 | switch self { |
| 95 | case .camera(let peer), .screen(let peer): return peer |
| 96 | } |
| 97 | } |
| 98 | } |
| 99 | |
| 100 | /// What the main stage shows; Speaker view's rules are in SpeakerView.swift. |
| 101 | public enum Stage {} |
| 102 | |
amadan