#pragma once

#include "screencast/capture/error.h"

#include <cstdint>
#include <memory>
#include <optional>
#include <string>
#include <variant>
#include <vector>

namespace sc {

enum class PixelFormat {
    Rgba,
    Bgrx, // 4 bytes/pixel, memory order B, G, R, unused
    Yuv420p,
};

// Opaque resource owned by the capture implementation.
struct CapturedFrame {
    int width = 0;
    int height = 0;
    uint64_t timestamp_ns = 0; // capture clock, monotonic
    PixelFormat pixel_format = PixelFormat::Rgba;
    int stride = 0; // bytes per row for the first plane; 0 means packed
    std::vector<std::byte> pixels;
};

// Capture target: whole monitor, specific window, or a region.
struct CaptureTargetWholeScreen {};
struct CaptureTargetWindow {
    std::string window_id;
};
struct CaptureTargetRegion {
    int x = 0;
    int y = 0;
    int width = 0;
    int height = 0;
};

using CaptureTarget = std::variant<CaptureTargetWholeScreen, CaptureTargetWindow, CaptureTargetRegion>;

class CaptureSession {
  public:
    virtual ~CaptureSession() = default;

    // Blocking call to acquire one frame. Returns std::nullopt on graceful stop.
    virtual std::optional<CapturedFrame> next_frame() = 0;

    // Request the session to stop. May be called from another thread.
    virtual void stop() = 0;
};

// Factory for the PipeWire / xdg-desktop-portal capture backend.
class CaptureFactory {
  public:
    static CaptureResult<std::unique_ptr<CaptureSession>> create(CaptureTarget target);
};

} // namespace sc