#include <gtest/gtest.h>

#include <memory>
#include <string>

#include "frame/frame.h"
#include "node.h"
#include "pose/pose_result.h"
#include "utils/simple_json.h"
#include "../plugins/action_recog/action_recog_node.h"

namespace rk3588 {
namespace {

SimpleJson ParseActionConfig(const std::string& text) {
    SimpleJson config;
    std::string err;
    const bool ok = ParseSimpleJson(text, config, err);
    EXPECT_TRUE(ok);
    return config;
}

TEST(ActionRecogTest, EmitsFallAfterRapidDropAndLowPosePersistence) {
    ActionRecogNode node;

    const std::string config_text = R"({
        "id": "action_evt",
        "events": [
            {
                "type": "fall",
                "window_ms": 1500,
                "min_drop_pixels": 120,
                "min_aspect_ratio_delta": 0.35,
                "activate_duration_ms": 0
            }
        ]
    })";

    SimpleJson config = ParseActionConfig(config_text);
    NodeContext ctx;
    auto out = std::make_shared<SpscQueue<FramePtr>>(4, QueueDropStrategy::DropOldest);
    ctx.output_queues.push_back(out);

    ASSERT_TRUE(node.Init(config, ctx));
    ASSERT_TRUE(node.Start());

    auto frame1 = std::make_shared<Frame>();
    frame1->width = 1920;
    frame1->height = 1080;
    frame1->pts = 1000;
    frame1->det = std::make_shared<DetectionResult>();
    frame1->det->img_w = 1920;
    frame1->det->img_h = 1080;
    frame1->det->items.push_back(Detection{0, 0.92f, Rect{800.0f, 240.0f, 120.0f, 320.0f}, 17});

    auto frame2 = std::make_shared<Frame>();
    frame2->width = 1920;
    frame2->height = 1080;
    frame2->pts = 1600;
    frame2->det = std::make_shared<DetectionResult>();
    frame2->det->img_w = 1920;
    frame2->det->img_h = 1080;
    frame2->det->items.push_back(Detection{0, 0.94f, Rect{760.0f, 460.0f, 260.0f, 140.0f}, 17});

    EXPECT_EQ(static_cast<int>(node.Process(frame1)), static_cast<int>(NodeStatus::OK));
    EXPECT_EQ(static_cast<int>(node.Process(frame2)), static_cast<int>(NodeStatus::OK));
    ASSERT_NE(frame2->behavior_events, nullptr);
    ASSERT_EQ(frame2->behavior_events->items.size(), 1u);
    EXPECT_EQ(frame2->behavior_events->items[0].type, BehaviorEventType::Fall);
    ASSERT_EQ(frame2->behavior_events->items[0].track_ids.size(), 1u);
    EXPECT_EQ(frame2->behavior_events->items[0].track_ids[0], 17);
}

TEST(ActionRecogTest, EmitsFightWhenTwoTracksStayCloseWithRepeatedMotion) {
    ActionRecogNode node;

    const std::string config_text = R"({
        "id": "action_evt",
        "events": [
            {
                "type": "fight",
                "window_ms": 1200,
                "proximity_pixels": 220,
                "min_motion_pixels": 70,
                "activate_duration_ms": 0
            }
        ]
    })";

    SimpleJson config = ParseActionConfig(config_text);
    NodeContext ctx;
    auto out = std::make_shared<SpscQueue<FramePtr>>(4, QueueDropStrategy::DropOldest);
    ctx.output_queues.push_back(out);

    ASSERT_TRUE(node.Init(config, ctx));
    ASSERT_TRUE(node.Start());

    auto frame1 = std::make_shared<Frame>();
    frame1->width = 1920;
    frame1->height = 1080;
    frame1->pts = 1000;
    frame1->det = std::make_shared<DetectionResult>();
    frame1->det->img_w = 1920;
    frame1->det->img_h = 1080;
    frame1->det->items.push_back(Detection{0, 0.90f, Rect{700.0f, 320.0f, 120.0f, 300.0f}, 31});
    frame1->det->items.push_back(Detection{0, 0.89f, Rect{860.0f, 320.0f, 120.0f, 300.0f}, 32});

    auto frame2 = std::make_shared<Frame>();
    frame2->width = 1920;
    frame2->height = 1080;
    frame2->pts = 1300;
    frame2->det = std::make_shared<DetectionResult>();
    frame2->det->img_w = 1920;
    frame2->det->img_h = 1080;
    frame2->det->items.push_back(Detection{0, 0.91f, Rect{780.0f, 340.0f, 120.0f, 300.0f}, 31});
    frame2->det->items.push_back(Detection{0, 0.92f, Rect{820.0f, 300.0f, 120.0f, 300.0f}, 32});

    EXPECT_EQ(static_cast<int>(node.Process(frame1)), static_cast<int>(NodeStatus::OK));
    EXPECT_EQ(static_cast<int>(node.Process(frame2)), static_cast<int>(NodeStatus::OK));
    ASSERT_NE(frame2->behavior_events, nullptr);
    ASSERT_EQ(frame2->behavior_events->items.size(), 1u);
    EXPECT_EQ(frame2->behavior_events->items[0].type, BehaviorEventType::Fight);
    ASSERT_EQ(frame2->behavior_events->items[0].track_ids.size(), 2u);
    EXPECT_EQ(frame2->behavior_events->items[0].track_ids[0], 31);
    EXPECT_EQ(frame2->behavior_events->items[0].track_ids[1], 32);
}

TEST(ActionRecogTest, EmitsFallFromPoseSequenceWhenBboxShapeChangeIsSmall) {
    ActionRecogNode node;

    const std::string config_text = R"({
        "id": "action_evt",
        "events": [
            {
                "type": "fall",
                "window_ms": 1500,
                "min_drop_pixels": 120,
                "min_aspect_ratio_delta": 10.0,
                "activate_duration_ms": 0,
                "pose_min_torso_drop_pixels": 120,
                "pose_max_upright_ratio": 0.6
            }
        ]
    })";

    SimpleJson config = ParseActionConfig(config_text);
    NodeContext ctx;
    auto out = std::make_shared<SpscQueue<FramePtr>>(4, QueueDropStrategy::DropOldest);
    ctx.output_queues.push_back(out);

    ASSERT_TRUE(node.Init(config, ctx));
    ASSERT_TRUE(node.Start());

    auto frame1 = std::make_shared<Frame>();
    frame1->width = 1920;
    frame1->height = 1080;
    frame1->pts = 1000;
    frame1->det = std::make_shared<DetectionResult>();
    frame1->det->img_w = 1920;
    frame1->det->img_h = 1080;
    frame1->det->items.push_back(Detection{0, 0.92f, Rect{800.0f, 220.0f, 180.0f, 300.0f}, 17});
    frame1->pose = std::make_shared<PoseResult>();
    frame1->pose->img_w = 1920;
    frame1->pose->img_h = 1080;
    PoseItem pose1;
    pose1.bbox = Rect{800.0f, 220.0f, 180.0f, 300.0f};
    pose1.score = 0.9f;
    pose1.keypoints.resize(17);
    pose1.keypoints[5] = PoseKeypoint{PosePoint2f{840.0f, 280.0f}, 0.9f};
    pose1.keypoints[6] = PoseKeypoint{PosePoint2f{940.0f, 280.0f}, 0.9f};
    pose1.keypoints[11] = PoseKeypoint{PosePoint2f{850.0f, 420.0f}, 0.9f};
    pose1.keypoints[12] = PoseKeypoint{PosePoint2f{930.0f, 420.0f}, 0.9f};
    frame1->pose->items.push_back(pose1);

    auto frame2 = std::make_shared<Frame>();
    frame2->width = 1920;
    frame2->height = 1080;
    frame2->pts = 1600;
    frame2->det = std::make_shared<DetectionResult>();
    frame2->det->img_w = 1920;
    frame2->det->img_h = 1080;
    frame2->det->items.push_back(Detection{0, 0.93f, Rect{780.0f, 420.0f, 200.0f, 280.0f}, 17});
    frame2->pose = std::make_shared<PoseResult>();
    frame2->pose->img_w = 1920;
    frame2->pose->img_h = 1080;
    PoseItem pose2;
    pose2.bbox = Rect{780.0f, 420.0f, 200.0f, 280.0f};
    pose2.score = 0.91f;
    pose2.keypoints.resize(17);
    pose2.keypoints[5] = PoseKeypoint{PosePoint2f{790.0f, 520.0f}, 0.9f};
    pose2.keypoints[6] = PoseKeypoint{PosePoint2f{970.0f, 520.0f}, 0.9f};
    pose2.keypoints[11] = PoseKeypoint{PosePoint2f{820.0f, 590.0f}, 0.9f};
    pose2.keypoints[12] = PoseKeypoint{PosePoint2f{940.0f, 590.0f}, 0.9f};
    frame2->pose->items.push_back(pose2);

    EXPECT_EQ(static_cast<int>(node.Process(frame1)), static_cast<int>(NodeStatus::OK));
    EXPECT_EQ(static_cast<int>(node.Process(frame2)), static_cast<int>(NodeStatus::OK));
    ASSERT_NE(frame2->behavior_events, nullptr);
    ASSERT_EQ(frame2->behavior_events->items.size(), 1u);
    EXPECT_EQ(frame2->behavior_events->items[0].type, BehaviorEventType::Fall);
    ASSERT_EQ(frame2->behavior_events->items[0].track_ids.size(), 1u);
    EXPECT_EQ(frame2->behavior_events->items[0].track_ids[0], 17);
}

TEST(ActionRecogTest, EmitsFightFromPoseMotionWhenBboxMotionIsSmall) {
    ActionRecogNode node;

    const std::string config_text = R"({
        "id": "action_evt",
        "events": [
            {
                "type": "fight",
                "window_ms": 1200,
                "proximity_pixels": 220,
                "min_motion_pixels": 1000,
                "activate_duration_ms": 0,
                "pose_min_wrist_motion_pixels": 120,
                "pose_max_wrist_distance_pixels": 120
            }
        ]
    })";

    SimpleJson config = ParseActionConfig(config_text);
    NodeContext ctx;
    auto out = std::make_shared<SpscQueue<FramePtr>>(4, QueueDropStrategy::DropOldest);
    ctx.output_queues.push_back(out);

    ASSERT_TRUE(node.Init(config, ctx));
    ASSERT_TRUE(node.Start());

    auto frame1 = std::make_shared<Frame>();
    frame1->width = 1920;
    frame1->height = 1080;
    frame1->pts = 1000;
    frame1->det = std::make_shared<DetectionResult>();
    frame1->det->img_w = 1920;
    frame1->det->img_h = 1080;
    frame1->det->items.push_back(Detection{0, 0.90f, Rect{700.0f, 320.0f, 120.0f, 300.0f}, 31});
    frame1->det->items.push_back(Detection{0, 0.89f, Rect{860.0f, 320.0f, 120.0f, 300.0f}, 32});
    frame1->pose = std::make_shared<PoseResult>();
    frame1->pose->img_w = 1920;
    frame1->pose->img_h = 1080;
    PoseItem left1;
    left1.bbox = Rect{700.0f, 320.0f, 120.0f, 300.0f};
    left1.keypoints.resize(17);
    left1.keypoints[9] = PoseKeypoint{PosePoint2f{780.0f, 430.0f}, 0.9f};
    left1.keypoints[10] = PoseKeypoint{PosePoint2f{790.0f, 450.0f}, 0.9f};
    PoseItem right1;
    right1.bbox = Rect{860.0f, 320.0f, 120.0f, 300.0f};
    right1.keypoints.resize(17);
    right1.keypoints[9] = PoseKeypoint{PosePoint2f{900.0f, 430.0f}, 0.9f};
    right1.keypoints[10] = PoseKeypoint{PosePoint2f{910.0f, 450.0f}, 0.9f};
    frame1->pose->items.push_back(left1);
    frame1->pose->items.push_back(right1);

    auto frame2 = std::make_shared<Frame>();
    frame2->width = 1920;
    frame2->height = 1080;
    frame2->pts = 1300;
    frame2->det = std::make_shared<DetectionResult>();
    frame2->det->img_w = 1920;
    frame2->det->img_h = 1080;
    frame2->det->items.push_back(Detection{0, 0.91f, Rect{710.0f, 325.0f, 120.0f, 300.0f}, 31});
    frame2->det->items.push_back(Detection{0, 0.92f, Rect{850.0f, 315.0f, 120.0f, 300.0f}, 32});
    frame2->pose = std::make_shared<PoseResult>();
    frame2->pose->img_w = 1920;
    frame2->pose->img_h = 1080;
    PoseItem left2;
    left2.bbox = Rect{710.0f, 325.0f, 120.0f, 300.0f};
    left2.keypoints.resize(17);
    left2.keypoints[9] = PoseKeypoint{PosePoint2f{860.0f, 430.0f}, 0.9f};
    left2.keypoints[10] = PoseKeypoint{PosePoint2f{870.0f, 450.0f}, 0.9f};
    PoseItem right2;
    right2.bbox = Rect{850.0f, 315.0f, 120.0f, 300.0f};
    right2.keypoints.resize(17);
    right2.keypoints[9] = PoseKeypoint{PosePoint2f{840.0f, 430.0f}, 0.9f};
    right2.keypoints[10] = PoseKeypoint{PosePoint2f{830.0f, 450.0f}, 0.9f};
    frame2->pose->items.push_back(left2);
    frame2->pose->items.push_back(right2);

    EXPECT_EQ(static_cast<int>(node.Process(frame1)), static_cast<int>(NodeStatus::OK));
    EXPECT_EQ(static_cast<int>(node.Process(frame2)), static_cast<int>(NodeStatus::OK));
    ASSERT_NE(frame2->behavior_events, nullptr);
    ASSERT_EQ(frame2->behavior_events->items.size(), 1u);
    EXPECT_EQ(frame2->behavior_events->items[0].type, BehaviorEventType::Fight);
    ASSERT_EQ(frame2->behavior_events->items[0].track_ids.size(), 2u);
    EXPECT_EQ(frame2->behavior_events->items[0].track_ids[0], 31);
    EXPECT_EQ(frame2->behavior_events->items[0].track_ids[1], 32);
}

TEST(ActionRecogTest, SupportsStructuredFusionConfigForFall) {
    ActionRecogNode node;

    const std::string config_text = R"({
        "id": "action_evt",
        "events": [
            {
                "type": "fall",
                "window_ms": 1500,
                "activate_duration_ms": 0,
                "bbox": {
                    "enabled": false
                },
                "pose": {
                    "enabled": true,
                    "min_torso_drop_pixels": 120,
                    "max_upright_ratio": 0.6
                },
                "fusion": {
                    "match_mode": "any"
                }
            }
        ]
    })";

    SimpleJson config = ParseActionConfig(config_text);
    NodeContext ctx;
    ASSERT_TRUE(node.Init(config, ctx));
    ASSERT_TRUE(node.Start());

    auto frame1 = std::make_shared<Frame>();
    frame1->width = 1920;
    frame1->height = 1080;
    frame1->pts = 1000;
    frame1->det = std::make_shared<DetectionResult>();
    frame1->det->items.push_back(Detection{0, 0.92f, Rect{800.0f, 220.0f, 180.0f, 300.0f}, 17});
    frame1->pose = std::make_shared<PoseResult>();
    PoseItem pose1;
    pose1.track_id = 17;
    pose1.bbox = Rect{800.0f, 220.0f, 180.0f, 300.0f};
    pose1.keypoints.resize(17);
    pose1.keypoints[5] = PoseKeypoint{PosePoint2f{840.0f, 280.0f}, 0.9f};
    pose1.keypoints[6] = PoseKeypoint{PosePoint2f{940.0f, 280.0f}, 0.9f};
    pose1.keypoints[11] = PoseKeypoint{PosePoint2f{850.0f, 420.0f}, 0.9f};
    pose1.keypoints[12] = PoseKeypoint{PosePoint2f{930.0f, 420.0f}, 0.9f};
    frame1->pose->items.push_back(pose1);

    auto frame2 = std::make_shared<Frame>();
    frame2->width = 1920;
    frame2->height = 1080;
    frame2->pts = 1600;
    frame2->det = std::make_shared<DetectionResult>();
    frame2->det->items.push_back(Detection{0, 0.93f, Rect{790.0f, 400.0f, 180.0f, 300.0f}, 17});
    frame2->pose = std::make_shared<PoseResult>();
    PoseItem pose2;
    pose2.track_id = 17;
    pose2.bbox = Rect{790.0f, 400.0f, 180.0f, 300.0f};
    pose2.keypoints.resize(17);
    pose2.keypoints[5] = PoseKeypoint{PosePoint2f{790.0f, 520.0f}, 0.9f};
    pose2.keypoints[6] = PoseKeypoint{PosePoint2f{970.0f, 520.0f}, 0.9f};
    pose2.keypoints[11] = PoseKeypoint{PosePoint2f{820.0f, 590.0f}, 0.9f};
    pose2.keypoints[12] = PoseKeypoint{PosePoint2f{940.0f, 590.0f}, 0.9f};
    frame2->pose->items.push_back(pose2);

    EXPECT_EQ(static_cast<int>(node.Process(frame1)), static_cast<int>(NodeStatus::OK));
    EXPECT_EQ(static_cast<int>(node.Process(frame2)), static_cast<int>(NodeStatus::OK));
    ASSERT_NE(frame2->behavior_events, nullptr);
    ASSERT_EQ(frame2->behavior_events->items.size(), 1u);
    EXPECT_EQ(frame2->behavior_events->items[0].type, BehaviorEventType::Fall);
}

TEST(ActionRecogTest, SuppressesFallFalsePositiveWhenPoseShapeConditionFails) {
    ActionRecogNode node;

    const std::string config_text = R"({
        "id": "action_evt",
        "events": [
            {
                "type": "fall",
                "window_ms": 1500,
                "activate_duration_ms": 0,
                "bbox": {
                    "enabled": false
                },
                "pose": {
                    "enabled": true,
                    "min_torso_drop_pixels": 120,
                    "max_upright_ratio": 0.8
                },
                "fusion": {
                    "match_mode": "any"
                }
            }
        ]
    })";

    SimpleJson config = ParseActionConfig(config_text);
    NodeContext ctx;
    ASSERT_TRUE(node.Init(config, ctx));
    ASSERT_TRUE(node.Start());

    auto frame1 = std::make_shared<Frame>();
    frame1->pts = 1000;
    frame1->det = std::make_shared<DetectionResult>();
    frame1->det->items.push_back(Detection{0, 0.92f, Rect{800.0f, 220.0f, 180.0f, 300.0f}, 17});
    frame1->pose = std::make_shared<PoseResult>();
    PoseItem pose1;
    pose1.track_id = 17;
    pose1.bbox = Rect{800.0f, 220.0f, 180.0f, 300.0f};
    pose1.keypoints.resize(17);
    pose1.keypoints[5] = PoseKeypoint{PosePoint2f{840.0f, 280.0f}, 0.9f};
    pose1.keypoints[6] = PoseKeypoint{PosePoint2f{940.0f, 280.0f}, 0.9f};
    pose1.keypoints[11] = PoseKeypoint{PosePoint2f{850.0f, 420.0f}, 0.9f};
    pose1.keypoints[12] = PoseKeypoint{PosePoint2f{930.0f, 420.0f}, 0.9f};
    frame1->pose->items.push_back(pose1);

    auto frame2 = std::make_shared<Frame>();
    frame2->pts = 1600;
    frame2->det = std::make_shared<DetectionResult>();
    frame2->det->items.push_back(Detection{0, 0.93f, Rect{790.0f, 400.0f, 180.0f, 300.0f}, 17});
    frame2->pose = std::make_shared<PoseResult>();
    PoseItem pose2;
    pose2.track_id = 17;
    pose2.bbox = Rect{790.0f, 400.0f, 180.0f, 300.0f};
    pose2.keypoints.resize(17);
    pose2.keypoints[5] = PoseKeypoint{PosePoint2f{860.0f, 520.0f}, 0.9f};
    pose2.keypoints[6] = PoseKeypoint{PosePoint2f{900.0f, 520.0f}, 0.9f};
    pose2.keypoints[11] = PoseKeypoint{PosePoint2f{865.0f, 700.0f}, 0.9f};
    pose2.keypoints[12] = PoseKeypoint{PosePoint2f{905.0f, 700.0f}, 0.9f};
    frame2->pose->items.push_back(pose2);

    EXPECT_EQ(static_cast<int>(node.Process(frame1)), static_cast<int>(NodeStatus::OK));
    EXPECT_EQ(static_cast<int>(node.Process(frame2)), static_cast<int>(NodeStatus::OK));
    ASSERT_NE(frame2->behavior_events, nullptr);
    EXPECT_TRUE(frame2->behavior_events->items.empty());
}

TEST(ActionRecogTest, SuppressesFightFalsePositiveWhenWristsStayFarApart) {
    ActionRecogNode node;

    const std::string config_text = R"({
        "id": "action_evt",
        "events": [
            {
                "type": "fight",
                "window_ms": 1200,
                "activate_duration_ms": 0,
                "bbox": {
                    "enabled": false,
                    "proximity_pixels": 220,
                    "min_motion_pixels": 0
                },
                "pose": {
                    "enabled": true,
                    "min_wrist_motion_pixels": 120,
                    "max_wrist_distance_pixels": 60
                },
                "fusion": {
                    "match_mode": "any"
                }
            }
        ]
    })";

    SimpleJson config = ParseActionConfig(config_text);
    NodeContext ctx;
    ASSERT_TRUE(node.Init(config, ctx));
    ASSERT_TRUE(node.Start());

    auto frame1 = std::make_shared<Frame>();
    frame1->pts = 1000;
    frame1->det = std::make_shared<DetectionResult>();
    frame1->det->items.push_back(Detection{0, 0.90f, Rect{700.0f, 320.0f, 120.0f, 300.0f}, 31});
    frame1->det->items.push_back(Detection{0, 0.89f, Rect{860.0f, 320.0f, 120.0f, 300.0f}, 32});
    frame1->pose = std::make_shared<PoseResult>();
    PoseItem left1;
    left1.track_id = 31;
    left1.bbox = Rect{700.0f, 320.0f, 120.0f, 300.0f};
    left1.keypoints.resize(17);
    left1.keypoints[9] = PoseKeypoint{PosePoint2f{760.0f, 430.0f}, 0.9f};
    left1.keypoints[10] = PoseKeypoint{PosePoint2f{770.0f, 450.0f}, 0.9f};
    PoseItem right1;
    right1.track_id = 32;
    right1.bbox = Rect{860.0f, 320.0f, 120.0f, 300.0f};
    right1.keypoints.resize(17);
    right1.keypoints[9] = PoseKeypoint{PosePoint2f{960.0f, 430.0f}, 0.9f};
    right1.keypoints[10] = PoseKeypoint{PosePoint2f{970.0f, 450.0f}, 0.9f};
    frame1->pose->items.push_back(left1);
    frame1->pose->items.push_back(right1);

    auto frame2 = std::make_shared<Frame>();
    frame2->pts = 1300;
    frame2->det = std::make_shared<DetectionResult>();
    frame2->det->items.push_back(Detection{0, 0.91f, Rect{710.0f, 325.0f, 120.0f, 300.0f}, 31});
    frame2->det->items.push_back(Detection{0, 0.92f, Rect{850.0f, 315.0f, 120.0f, 300.0f}, 32});
    frame2->pose = std::make_shared<PoseResult>();
    PoseItem left2;
    left2.track_id = 31;
    left2.bbox = Rect{710.0f, 325.0f, 120.0f, 300.0f};
    left2.keypoints.resize(17);
    left2.keypoints[9] = PoseKeypoint{PosePoint2f{820.0f, 430.0f}, 0.9f};
    left2.keypoints[10] = PoseKeypoint{PosePoint2f{830.0f, 450.0f}, 0.9f};
    PoseItem right2;
    right2.track_id = 32;
    right2.bbox = Rect{850.0f, 315.0f, 120.0f, 300.0f};
    right2.keypoints.resize(17);
    right2.keypoints[9] = PoseKeypoint{PosePoint2f{980.0f, 430.0f}, 0.9f};
    right2.keypoints[10] = PoseKeypoint{PosePoint2f{990.0f, 450.0f}, 0.9f};
    frame2->pose->items.push_back(left2);
    frame2->pose->items.push_back(right2);

    EXPECT_EQ(static_cast<int>(node.Process(frame1)), static_cast<int>(NodeStatus::OK));
    EXPECT_EQ(static_cast<int>(node.Process(frame2)), static_cast<int>(NodeStatus::OK));
    ASSERT_NE(frame2->behavior_events, nullptr);
    EXPECT_TRUE(frame2->behavior_events->items.empty());
}

}  // namespace
}  // namespace rk3588