Promise.hpp

Go to the documentation of this file.

#pragma once
 
#include <coroutine>
 
#include "MayaFlux/Core/ProcessingTokens.hpp"
 
namespace MayaFlux::Vruta {
 
class SoundRoutine;
class GraphicsRoutine;
class ComplexRoutine;
class Event;
class NetworkSource;
 
/**
 * @struct routine_promise
 * @brief Base coroutine promise type for audio processing tasks
 *
 * This promise_type serves as the base class for all coroutine promises in the
 * MayaFlux engine. It defines the common behavior and interface for
 * coroutines, including lifecycle management, state storage, and execution flags.
 *
 * The promise_type is a crucial component of C++20 coroutines that defines the
 * behavior of SoundRoutine coroutines. It serves as the control interface between
 * the coroutine machinery and the audio engine, managing:
 *
 * In the coroutine model, the promise object is created first when a coroutine
 * function is called. It then creates and returns the SoundRoutine object that
 * the caller receives. The promise remains associated with the coroutine frame
 * throughout its lifetime, while the RoutineType provides the external interface
 * to manipulate the coroutine.
 */
template <typename RoutineType>
struct MAYAFLUX_API routine_promise {
 
    RoutineType get_return_object()
    {
        return RoutineType(std::coroutine_handle<routine_promise>::from_promise(*this));
    }
 
    /**
     * @brief Determines whether the coroutine suspends immediately upon creation
     * @return A suspend never awaitable, meaning the coroutine begins execution immediately
     *
     * By returning std::suspend_never, this method indicates that the coroutine
     * should start executing as soon as it's created, rather than waiting for
     * an explicit resume call.
     */
    std::suspend_never initial_suspend() { return {}; }
 
    /**
     * @brief Determines whether the coroutine suspends before destruction
     * @return A suspend always awaitable, meaning the coroutine suspends before completing
     *
     * By returning std::suspend_always, this method ensures that the coroutine
     * frame isn't destroyed immediately when the coroutine completes. This gives
     * the scheduler a chance to observe the completion and perform cleanup.
     */
    std::suspend_always final_suspend() noexcept { return {}; }
 
    /**
     * @brief Handles the coroutine's void return
     *
     * This method is called when the coroutine executes a co_return statement
     * without a value, or reaches the end of its function body.
     */
    void return_void() { }
 
    /**
     * @brief Handles exceptions thrown from within the coroutine
     *
     * This method is called if an unhandled exception escapes from the coroutine.
     * The current implementation terminates the program, as exceptions in audio
     * processing code are generally considered fatal errors.
     */
    void unhandled_exception() { std::terminate(); }
 
    /**
     * @brief Token indicating how this coroutine should be processed
     */
    const ProcessingToken processing_token { ProcessingToken::ON_DEMAND };
 
    /**
     * @brief Flag indicating whether the coroutine should be automatically resumed
     *
     * When true, the scheduler will automatically resume the coroutine when
     * the current sample position reaches next_sample. When false, the coroutine
     * must be manually resumed by calling ::try_resume.
     */
    bool auto_resume = true;
 
    /**
     * @brief Flag indicating whether the coroutine should be terminated
     *
     * When set to true, the scheduler will destroy the coroutine rather than
     * resuming it, even if it hasn't completed naturally. This allows for
     * early termination of long-running coroutines.
     */
    bool should_terminate = false;
 
    /**
     * @brief Dictionary for storing arbitrary state data
     *
     * This map allows the coroutine to store and retrieve named values of any type.
     * It serves multiple purposes:
     * 1. Persistent storage between suspensions
     * 2. Communication channel between the coroutine and external code
     * 3. Parameter storage for configurable behaviors
     *
     * The use of std::any allows for type-safe heterogeneous storage without
     * requiring the promise type to be templated.
     */
    std::unordered_map<std::string, std::any> state;
 
    /**
     * @brief Flag indicating whether the coroutine should synchronize with the audio clock
     *
     * When true, the coroutine will be scheduled to run in sync with the specificed clock,
     * via tokens ensuring sample-accurate timing. When false, it has to be "proceeded" manually
     */
    const bool sync_to_clock = false;
 
    /**
     * @brief Amount of delay requested by the coroutine
     *
     * This value is set when the coroutine co_awaits a delay awaiter (e.g., SampleDelay).
     * It indicates how many time units the coroutine wishes to wait before resuming.
     */
    uint64_t delay_amount = 0;
 
    /**
     * @brief Stores a value in the state dictionary
     * @param key Name of the state value
     * @param value Value to store
     *
     * This method provides a type-safe way to store values of any type in the
     * state dictionary. The value is wrapped in std::any for type erasure.
     */
    template <typename T>
    inline void set_state(const std::string& key, T value)
    {
        state[key] = std::make_any<T>(std::move(value));
    }
 
    /**
     * @brief Retrieves a value from the state dictionary
     * @param key Name of the state value to retrieve
     * @return Pointer to the stored value, or nullptr if not found or type mismatch
     *
     * This method provides a type-safe way to retrieve values from the state
     * dictionary. It returns a pointer to the stored value if it exists and
     * has the requested type, or nullptr otherwise.
     */
    template <typename T>
    inline T* get_state(const std::string& key)
    {
        auto it = state.find(key);
        if (it != state.end()) {
            try {
                return std::any_cast<T>(&it->second);
            } catch (const std::bad_any_cast&) {
                return nullptr;
            }
        }
        return nullptr;
    }
 
    void domain_mismatch_error(const std::string& awaiter_name, const std::string& suggestion)
    {
        set_state("domain_error", awaiter_name + ": " + suggestion);
        should_terminate = true;
    }
};
 
/**
 * @struct audio_promise
 * @brief Coroutine promise type for audio processing tasks with sample-accurate timing
 *
 * The promise_type is a crucial component of C++20 coroutines that defines the
 * behavior of SoundRoutine coroutines. It serves as the control interface between
 * the coroutine machinery and the audio engine, managing:
 *
 * 1. Coroutine lifecycle (creation, suspension, resumption, and destruction)
 * 2. Timing information for sample-accurate scheduling
 * 3. State storage for persistent data between suspensions
 * 4. Control flags for execution behavior
 *
 * In the coroutine model, the promise object is created first when a coroutine
 * function is called. It then creates and returns the SoundRoutine object that
 * the caller receives. The promise remains associated with the coroutine frame
 * throughout its lifetime, while the SoundRoutine provides the external interface
 * to manipulate the coroutine.
 *
 * This separation of concerns allows the audio engine to schedule and manage
 * coroutines efficiently while providing a clean API for audio processing code.
 */
struct audio_promise : public routine_promise<SoundRoutine> {
    /**
     * @brief Creates the SoundRoutine object returned to the caller
     * @return A new SoundRoutine that wraps this promise
     *
     * This method is called by the compiler-generated code when a coroutine
     * function is invoked. It creates the SoundRoutine object that will be
     * returned to the caller and associates it with this promise.
     */
    SoundRoutine get_return_object();
 
    ProcessingToken processing_token { ProcessingToken::SAMPLE_ACCURATE };
 
    bool sync_to_clock = true;
 
    /**
     * @brief The sample position when this coroutine should next execute
     *
     * This is the core timing mechanism for sample-accurate scheduling.
     * When a coroutine co_awaits a SampleDelay, this value is updated to
     * indicate when the coroutine should be resumed next.
     */
    uint64_t next_sample = 0;
 
    /**
     * @brief The buffer cycle when this coroutine should next execute
     * Managed by BufferDelay awaiter. Incremented on each co_await BufferDelay{}.
     * Starts at 0, incremented to 1 on first await.
     */
    uint64_t next_buffer_cycle = 0;
 
    /**
     * @brief The active delay context for this coroutine
     *
     * This value indicates which type of delay (sample, buffer, event)
     * is currently being awaited by the coroutine. It helps the scheduler
     * determine how to manage the coroutine's timing.
     */
    DelayContext active_delay_context = DelayContext::NONE;
};
 
/**
 * @struct graphics_promise
 * @brief Coroutine promise type for graphics processing tasks with frame-accurate timing
 *
 * graphics_promise is the frame-domain equivalent of audio_promise. It manages the
 * state and lifecycle of GraphicsRoutine coroutines, providing frame-accurate timing
 * and scheduling for visual processing tasks.
 *
 * Key Architectural Notes:
 * - Frame timing is managed by FrameClock (self-driven wall-clock based)
 * - Unlike audio_promise (driven by RTAudio callbacks), graphics_promise observes
 *   the FrameClock which ticks independently in the graphics thread loop
 * - The promise doesn't care HOW timing advances, only that it receives tick updates
 * - Mirrors audio_promise architecture but for the visual/frame domain
 *
 * Timing Flow:
 * 1. Graphics thread loop: m_frame_clock->tick() (self-driven)
 * 2. GraphicsSubsystem::process() called
 * 3. Scheduler::process_frame_coroutines_impl() checks routines
 * 4. If current_frame >= next_frame, routine->try_resume(current_frame)
 * 5. Routine updates next_frame based on FrameDelay amount
 */
struct graphics_promise : public routine_promise<GraphicsRoutine> {
    /**
     * @brief Creates the GraphicsRoutine object returned to the caller
     * @return A new GraphicsRoutine that wraps this promise
     *
     * This method is called by the compiler-generated code when a coroutine
     * function is invoked. It creates the GraphicsRoutine object that will be
     * returned to the caller and associates it with this promise.
     */
    GraphicsRoutine get_return_object();
 
    /**
     * @brief Processing token indicating frame-accurate scheduling
     */
    ProcessingToken processing_token { ProcessingToken::FRAME_ACCURATE };
 
    /**
     * @brief Whether this routine should synchronize with FrameClock
     */
    bool sync_to_clock = true;
 
    /**
     * @brief The frame position when this coroutine should next execute
     *
     * This is the core timing mechanism for frame-accurate scheduling.
     * When a coroutine co_awaits a FrameDelay, this value is updated to
     * indicate when the coroutine should be resumed next.
     *
     * Example:
     * - Current frame: 1000
     * - co_await FrameDelay{5}
     * - next_frame becomes: 1005
     * - Routine resumes when FrameClock reaches frame 1005
     */
    uint64_t next_frame = 0;
 
    /**
     * @brief The active delay context for this coroutine
     *
     * This value indicates which type of delay (frame, event, etc.)
     * is currently being awaited by the coroutine. It helps the scheduler
     * determine how to manage the coroutine's timing and prevents
     * cross-domain contamination (e.g., audio delays don't affect graphics routines).
     *
     * Valid states for graphics routines:
     * - NONE: No active delay, can resume immediately
     * - FRAME_BASED: Waiting for a specific frame (FrameDelay)
     * - EVENT_BASED: Waiting for a window/input event (EventAwaiter)
     * - AWAIT: Temporary state during GetPromise awaiter
     */
    DelayContext active_delay_context = DelayContext::NONE;
 
    /**
     * @brief The amount of delay units for incremental delays
     *
     * When using delays that accumulate (like BufferDelay for audio),
     * this stores the increment amount. For graphics, this would be
     * the frame increment for continuous animations.
     *
     * Example:
     * ```cpp
     * auto animation = []() -> GraphicsRoutine {
     *     while (true) {
     *         render_frame();
     *         co_await FrameDelay{1};  // delay_amount = 1
     *     }
     * };
     * ```
     */
    uint64_t delay_amount = 0;
};
 
// TODO: Graphics features are not yet implemented, needs GL/Vulkan integration first
/** * @struct complex_promise
 * @brief Coroutine promise type for complex processing tasks with multi-rate scheduling
 */
struct complex_promise : public routine_promise<ComplexRoutine> {
    ComplexRoutine get_return_object();
 
    ProcessingToken processing_token { ProcessingToken::MULTI_RATE };
 
    bool sync_to_clock = true;
 
    uint64_t next_sample = 0;
 
    uint64_t next_frame = 0;
};
 
/**
 * @struct EventPromise
 * @brief Promise type for event-driven coroutines
 *
 * Unlike time-based promises (SampleClockPromise, FrameClockPromise),
 * EventPromise has no clock. Coroutines suspend/resume based on
 * discrete event signals, not periodic ticks.
 */
struct event_promise : public routine_promise<Event> {
    Event get_return_object();
 
    ProcessingToken processing_token { ProcessingToken::EVENT_DRIVEN };
 
    DelayContext active_delay_context { DelayContext::EVENT_BASED };
 
    /**
     * @brief Transfer ownership of a NetworkSource to this coroutine frame.
     * @param source Source to keep alive for the coroutine's lifetime.
     */
    void own(std::shared_ptr<Vruta::NetworkSource> source)
    {
        owned_sources.push_back(std::move(source));
    }
 
    /**
     * @brief Coroutine-owned NetworkSource instances.
     *
     * Sources deposited here via GetEventPromise live for exactly the
     * lifetime of the coroutine frame. Populated through co_await
     * GetEventPromise{ source } at coroutine startup.
     */
    std::vector<std::shared_ptr<Vruta::NetworkSource>> owned_sources;
};
 
}