#pragma once

#include "screencast/capture/capture.h"
#include "screencast/codec/error.h"

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

namespace sc {

struct EncodedFrame {
    uint64_t capture_timestamp_ns = 0;
    uint32_t rtp_timestamp = 0;
    bool is_keyframe = false;
    std::vector<std::byte> data; // Annex-B H.264 NAL units
};

struct EncoderConfig {
    std::string codec_name = "h264";
    int width = 0;
    int height = 0;
    int frame_rate_num = 30;
    int frame_rate_den = 1;
    // CRF (constant rate factor, 0-51): the target visual quality level.
    // Lower = better quality. 23 is x264's default; screen content looks
    // good at 20-24.
    int crf = 22;
    // Maximum bitrate in kbps (the VBV cap). The encoder uses fewer bits on
    // static content and more on motion, but never exceeds this.
    int bitrate_kbps = 4000;
    bool hardware_accel = false;
};

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

    // Encode one captured frame. Returns an empty vector if the encoder emits
    // no packet for this frame.
    virtual CodecResult<std::vector<EncodedFrame>> encode(const CapturedFrame& frame) = 0;

    // Flush the encoder and return any remaining packets.
    virtual CodecResult<std::vector<EncodedFrame>> flush() = 0;

    // Force the next output frame to be a keyframe.
    virtual void request_keyframe() = 0;

    // Codec-specific parameters required by a decoder (SPS/PPS for H.264).
    virtual std::vector<std::byte> get_extradata() const = 0;
};

class EncoderFactory {
  public:
    static CodecResult<std::unique_ptr<Encoder>> create(const EncoderConfig& config);
};

} // namespace sc