#include "../graph/graph_internal.h"
#include <math.h>
#include <stdlib.h>
#include <string.h>

#define XTC_TAU 6.28318530717958647692

struct xtc_plan {
    size_t node_count, max_frames;
    xtc_node *nodes;
    size_t *order, *offsets, *sources;
    unsigned char *render;
    float *buffers, *peaks;
    double *phases;
    double sample_rate;
    float input_gain;
};

void xtc_plan_destroy(xtc_plan *plan)
{
    if (plan == NULL) return;
    free(plan->nodes);
    free(plan->order);
    free(plan->offsets);
    free(plan->sources);
    free(plan->render);
    free(plan->buffers);
    free(plan->peaks);
    free(plan->phases);
    free(plan);
}

xtc_result xtc_graph_compile(const xtc_graph *graph, double sample_rate,
                           size_t max_frames, xtc_plan **out_plan)
{
    xtc_plan *plan;
    xtc_result result;
    size_t node, edge, storage_nodes, cursor = 0U;
    int has_output = 0;
    if (out_plan == NULL) return XTC_ERR_INVALID_ARGUMENT;
    *out_plan = NULL;
    if (graph == NULL || !isfinite(sample_rate) || sample_rate <= 0.0 ||
        max_frames == 0U || max_frames > XTC_MAX_FRAMES) return XTC_ERR_INVALID_ARGUMENT;
    storage_nodes = graph->node_count == 0U ? 1U : graph->node_count;
    for (node = 0U; node < graph->node_count; ++node) {
        if (graph->nodes[node].type == XTC_NODE_OSCILLATOR &&
            (double)graph->nodes[node].value > sample_rate * 0.5)
            return XTC_ERR_INVALID_ARGUMENT;
        if (graph->nodes[node].type == XTC_NODE_OUTPUT) has_output = 1;
    }
    /* Fixed graph/frame limits bound all allocation products below. */
    plan = calloc(1U, sizeof(*plan));
    if (plan == NULL) return XTC_ERR_OUT_OF_MEMORY;
    plan->node_count = graph->node_count;
    plan->max_frames = max_frames;
    plan->sample_rate = sample_rate;
    plan->input_gain = 1.0f;
    plan->nodes = calloc(storage_nodes, sizeof(*plan->nodes));
    plan->order = calloc(storage_nodes, sizeof(*plan->order));
    plan->offsets = calloc(plan->node_count + 1U, sizeof(*plan->offsets));
    plan->sources = calloc(graph->edge_count + 1U, sizeof(*plan->sources));
    plan->render = calloc(storage_nodes, sizeof(*plan->render));
    plan->buffers = calloc(storage_nodes * max_frames, sizeof(*plan->buffers));
    plan->peaks = calloc(storage_nodes, sizeof(*plan->peaks));
    plan->phases = calloc(storage_nodes, sizeof(*plan->phases));
    if (plan->nodes == NULL || plan->order == NULL || plan->offsets == NULL ||
        plan->sources == NULL || plan->render == NULL || plan->buffers == NULL ||
        plan->peaks == NULL || plan->phases == NULL) {
        xtc_plan_destroy(plan);
        return XTC_ERR_OUT_OF_MEMORY;
    }
    result = xtc_graph_order(graph, plan->order);
    if (result != XTC_OK) { xtc_plan_destroy(plan); return result; }
    memcpy(plan->nodes, graph->nodes, plan->node_count * sizeof(*plan->nodes));
    for (node = 0U; node < plan->node_count; ++node) {
        int terminal = 1;
        plan->offsets[node] = cursor;
        for (edge = 0U; edge < graph->edge_count; ++edge) {
            if (graph->edges[edge].destination == node)
                plan->sources[cursor++] = graph->edges[edge].source;
            if (graph->edges[edge].source == node) terminal = 0;
        }
        plan->render[node] = (unsigned char)(has_output || graph->require_output_nodes ?
            plan->nodes[node].type == XTC_NODE_OUTPUT : terminal);
    }
    plan->offsets[plan->node_count] = cursor;
    *out_plan = plan;
    return XTC_OK;
}

xtc_result xtc_plan_reset(xtc_plan *plan)
{
    if (plan == NULL) return XTC_ERR_INVALID_ARGUMENT;
    memset(plan->phases, 0, plan->node_count * sizeof(*plan->phases));
    memset(plan->peaks, 0, plan->node_count * sizeof(*plan->peaks));
    return XTC_OK;
}

xtc_result xtc_plan_meter_peak(const xtc_plan *plan, size_t node_id, float *out_peak)
{
    if (plan == NULL || out_peak == NULL) return XTC_ERR_INVALID_ARGUMENT;
    size_t index;
    for (index = 0U; index < plan->node_count; ++index)
        if (plan->nodes[index].id == node_id) break;
    if (index == plan->node_count) return XTC_ERR_NOT_FOUND;
    node_id = index;
    if (plan->nodes[node_id].type != XTC_NODE_METER) return XTC_ERR_INVALID_ARGUMENT;
    *out_peak = plan->peaks[node_id];
    return XTC_OK;
}

xtc_result xtc_plan_process(xtc_plan *plan, const float *input, float *output,
                          size_t frame_count)
{
    size_t index, frame, edge;
    if (plan == NULL || output == NULL) return XTC_ERR_INVALID_ARGUMENT;
    if (frame_count > plan->max_frames) return XTC_ERR_CAPACITY;
    if (frame_count == 0U) return XTC_OK;
    /* All external input reads finish before output is written, allowing aliasing.
     * Nonfinite input/arithmetic becomes silence per sample, never persistent DSP state. */
    for (index = 0U; index < plan->node_count; ++index) {
        size_t node = plan->order[index];
        xtc_node spec = plan->nodes[node];
        float *buffer = plan->buffers + node * plan->max_frames;
        float peak = 0.0f;
        double phase = plan->phases[node];
        for (frame = 0U; frame < frame_count; ++frame) {
            float sample = 0.0f;
            if (spec.type == XTC_NODE_OSCILLATOR) {
                sample = (float)sin(XTC_TAU * phase);
                phase += (double)spec.value / plan->sample_rate;
                if (phase >= 1.0) phase -= 1.0;
            } else {
                if (plan->offsets[node] == plan->offsets[node + 1U])
                    sample = input == NULL ? 0.0f : input[frame] * plan->input_gain;
                for (edge = plan->offsets[node]; edge < plan->offsets[node + 1U]; ++edge)
                    sample += plan->buffers[plan->sources[edge] * plan->max_frames + frame];
                if (spec.type == XTC_NODE_GAIN) sample *= spec.value;
            }
            if (!isfinite(sample)) sample = 0.0f;
            buffer[frame] = sample;
            if (fabsf(sample) > peak) peak = fabsf(sample);
        }
        plan->phases[node] = phase;
        if (spec.type == XTC_NODE_METER) plan->peaks[node] = peak;
    }
    for (frame = 0U; frame < frame_count; ++frame) {
        float sample = 0.0f;
        for (index = 0U; index < plan->node_count; ++index)
            if (plan->render[index]) sample += plan->buffers[index * plan->max_frames + frame];
        output[frame] = isfinite(sample) ? sample : 0.0f;
    }
    return XTC_OK;
}

xtc_result xtc_graph_process_offline(xtc_graph *graph, const float *input,
                                   float *output, size_t frame_count, float input_gain)
{
    xtc_plan *plan = NULL;
    xtc_result result;
    if (input == NULL || output == NULL || !isfinite(input_gain))
        return XTC_ERR_INVALID_ARGUMENT;
    result = xtc_graph_compile(graph, 48000.0, frame_count == 0U ? 1U : frame_count, &plan);
    if (result != XTC_OK) return result;
    plan->input_gain = input_gain;
    result = xtc_plan_process(plan, input, output, frame_count);
    xtc_plan_destroy(plan);
    return result;
}