ChatApp/apps/desktop/src-tauri/src/screen_sources.rs

// Source enumeration for the Discord-style screen-share picker. Split
// into two commands on the slow-vs-fast axis:
//
//   - list_screen_sources      — metadata only (no thumbnails). Fast; the
//                                picker shows names + placeholders instantly.
//   - capture_screen_source_thumbnail — one thumbnail at a time, keyed by
//                                       the id returned from the list.
//
// The frontend fans out the thumbnail calls via Promise.all so Tauri's
// command thread pool captures them in parallel — wall-clock time ends up
// bounded by the *slowest* source rather than the sum of all captures.
// The `id` field is emitted in Chromium's internal desktopCapturer format
// ("screen:<id>:0" / "window:<hwnd>:0") so the JS side can try to pass it
// straight into getUserMedia's `chromeMediaSourceId` constraint, or use
// it as the source key for the native capture pipeline in screen_capture.
//
// xcap abstracts the platform-specific capture APIs (Windows GDI + DXGI,
// macOS CoreGraphics/ScreenCaptureKit, X11) so the code here stays flat.
// Thumbnails are captured at native resolution, then letterbox-downscaled
// to fit a 320×180 box to keep the base64 payload small.

use base64::Engine;
use image::{ImageBuffer, Rgba};
use serde::Serialize;

const THUMB_MAX_W: u32 = 320;
const THUMB_MAX_H: u32 = 180;

#[derive(Serialize, Clone)]
#[serde(rename_all = "camelCase")]
pub struct ScreenSource {
    /// Chromium-format source id — stable within one enumeration call.
    pub id: String,
    /// Human-readable label for the picker (monitor name or window title).
    pub name: String,
    /// Discriminator for the grid grouping.
    pub kind: &'static str,
    /// Base64-encoded PNG, no data-URL prefix. None when the capture
    /// fails (minimised window, permission-denied surface, transient
    /// race). The UI renders a name-only card in that case.
    pub thumbnail_png: Option<String>,
    /// Native width of the full-res source — mostly informational, used
    /// by the UI for aspect-ratio styling of the card.
    pub width: u32,
    pub height: u32,
}

// Fast metadata-only enumeration. No image capture happens here — that's
// why it returns in tens of milliseconds instead of the multi-second
// wait the single-shot enumerate_screen_sources command had.
#[tauri::command]
pub fn list_screen_sources() -> Result<Vec<ScreenSource>, String> {
    let mut out: Vec<ScreenSource> = Vec::new();
    append_monitor_metadata(&mut out);
    append_window_metadata(&mut out);
    Ok(out)
}

// One-shot thumbnail capture by source id. Called N times in parallel from
// the frontend after the list lands so the total wall-clock is bounded by
// the slowest capture rather than the sum.
#[tauri::command]
pub fn capture_screen_source_thumbnail(source_id: String) -> Result<Option<String>, String> {
    if let Some(raw) = source_id.strip_prefix("screen:").and_then(strip_zero_suffix) {
        return Ok(capture_monitor_by_id(raw));
    }
    if let Some(raw) = source_id.strip_prefix("window:").and_then(strip_zero_suffix) {
        return Ok(capture_window_by_id(raw));
    }
    Err(format!("unknown source id format: {source_id}"))
}

fn strip_zero_suffix(s: &str) -> Option<&str> {
    s.strip_suffix(":0")
}

// Legacy one-shot all-in-one enumeration. Kept around so the frontend can
// fall back during rollout if the split pair throws; marked dead_code so
// the linker doesn't grumble when only the split variant is wired up.
#[allow(dead_code)]
#[tauri::command]
pub fn enumerate_screen_sources() -> Result<Vec<ScreenSource>, String> {
    let mut out: Vec<ScreenSource> = Vec::new();
    append_monitors(&mut out);
    append_windows(&mut out);
    Ok(out)
}

// ---------------------------------------------------------------------------
// Monitors
// ---------------------------------------------------------------------------

fn append_monitor_metadata(out: &mut Vec<ScreenSource>) {
    let monitors = match xcap::Monitor::all() {
        Ok(m) => m,
        Err(err) => {
            eprintln!("xcap Monitor::all failed: {err}");
            return;
        }
    };
    for (idx, m) in monitors.iter().enumerate() {
        out.push(ScreenSource {
            id: format!("screen:{}:0", m.id()),
            name: monitor_label(m, idx),
            kind: "screen",
            thumbnail_png: None,
            width: m.width(),
            height: m.height(),
        });
    }
}

#[allow(dead_code)]
fn append_monitors(out: &mut Vec<ScreenSource>) {
    let monitors = match xcap::Monitor::all() {
        Ok(m) => m,
        Err(err) => {
            eprintln!("xcap Monitor::all failed: {err}");
            return;
        }
    };
    for (idx, m) in monitors.iter().enumerate() {
        out.push(ScreenSource {
            id: format!("screen:{}:0", m.id()),
            name: monitor_label(m, idx),
            kind: "screen",
            thumbnail_png: capture_monitor_thumbnail(m),
            width: m.width(),
            height: m.height(),
        });
    }
}

fn capture_monitor_by_id(raw: &str) -> Option<String> {
    let raw_id: u32 = raw.parse().ok()?;
    let monitors = xcap::Monitor::all().ok()?;
    let target = monitors.into_iter().find(|m| m.id() == raw_id)?;
    capture_monitor_thumbnail(&target)
}

fn monitor_label(m: &xcap::Monitor, idx: usize) -> String {
    let name = m.name();
    if name.is_empty() {
        format!("Bildschirm {}", idx + 1)
    } else {
        name.to_string()
    }
}

fn capture_monitor_thumbnail(m: &xcap::Monitor) -> Option<String> {
    let image = m.capture_image().ok()?;
    encode_scaled_png(image)
}

// ---------------------------------------------------------------------------
// Windows
// ---------------------------------------------------------------------------

fn append_window_metadata(out: &mut Vec<ScreenSource>) {
    let windows = match xcap::Window::all() {
        Ok(w) => w,
        Err(err) => {
            eprintln!("xcap Window::all failed: {err}");
            return;
        }
    };
    for w in windows.iter() {
        if !is_shareable_window(w) {
            continue;
        }
        let title = w.title();
        if title.trim().is_empty() {
            continue;
        }
        out.push(ScreenSource {
            id: format!("window:{}:0", w.id()),
            name: title.to_string(),
            kind: "window",
            thumbnail_png: None,
            width: w.width(),
            height: w.height(),
        });
    }
}

#[allow(dead_code)]
fn append_windows(out: &mut Vec<ScreenSource>) {
    let windows = match xcap::Window::all() {
        Ok(w) => w,
        Err(err) => {
            eprintln!("xcap Window::all failed: {err}");
            return;
        }
    };
    for w in windows.iter() {
        if !is_shareable_window(w) {
            continue;
        }
        let title = w.title();
        if title.trim().is_empty() {
            continue;
        }
        out.push(ScreenSource {
            id: format!("window:{}:0", w.id()),
            name: title.to_string(),
            kind: "window",
            thumbnail_png: capture_window_thumbnail(w),
            width: w.width(),
            height: w.height(),
        });
    }
}

fn capture_window_by_id(raw: &str) -> Option<String> {
    let raw_id: u32 = raw.parse().ok()?;
    let windows = xcap::Window::all().ok()?;
    let target = windows.into_iter().find(|w| w.id() == raw_id)?;
    capture_window_thumbnail(&target)
}

fn is_shareable_window(w: &xcap::Window) -> bool {
    if w.is_minimized() {
        return false;
    }
    let width = w.width();
    let height = w.height();
    // Tooltips, invisible tray-helpers, etc. sit at or near zero size —
    // they'd clutter the picker grid and usually can't be captured anyway.
    if width < 80 || height < 60 {
        return false;
    }
    true
}

fn capture_window_thumbnail(w: &xcap::Window) -> Option<String> {
    let image = w.capture_image().ok()?;
    encode_scaled_png(image)
}

// ---------------------------------------------------------------------------
// Scaling + encoding
// ---------------------------------------------------------------------------

// Letterbox-shrink the captured image so the long edge is at most
// THUMB_MAX_W / THUMB_MAX_H. Keeps aspect ratio, skips upscaling entirely
// (tiny windows stay their captured size). Returns a base64 PNG string or
// None if encoding fails.
fn encode_scaled_png(src: ImageBuffer<Rgba<u8>, Vec<u8>>) -> Option<String> {
    let (w, h) = src.dimensions();
    if w == 0 || h == 0 {
        return None;
    }
    let scale = (THUMB_MAX_W as f32 / w as f32)
        .min(THUMB_MAX_H as f32 / h as f32)
        .min(1.0);
    let scaled = if scale < 1.0 {
        let new_w = ((w as f32) * scale).round().max(1.0) as u32;
        let new_h = ((h as f32) * scale).round().max(1.0) as u32;
        image::imageops::resize(&src, new_w, new_h, image::imageops::FilterType::Triangle)
    } else {
        src
    };
    let mut buf: Vec<u8> = Vec::new();
    let encoder = image::codecs::png::PngEncoder::new_with_quality(
        &mut buf,
        image::codecs::png::CompressionType::Fast,
        image::codecs::png::FilterType::Adaptive,
    );
    use image::ImageEncoder;
    encoder
        .write_image(
            scaled.as_raw(),
            scaled.width(),
            scaled.height(),
            image::ExtendedColorType::Rgba8,
        )
        .ok()?;
    Some(base64::engine::general_purpose::STANDARD.encode(&buf))
}