#include <doctest/doctest.h>

#include "arena.hpp"
#include "metric.hpp"

#include <atomic>
#include <cmath>
#include <latch>
#include <sstream>
#include <thread>
#include <vector>

TEST_CASE("metric validation functions") {
  SUBCASE("valid metric names") {
    CHECK(metric::is_valid_metric_name("valid_name"));
    CHECK(metric::is_valid_metric_name("ValidName"));
    CHECK(metric::is_valid_metric_name("valid:name"));
    CHECK(metric::is_valid_metric_name("_valid"));
    CHECK(metric::is_valid_metric_name("valid_123"));
    CHECK(metric::is_valid_metric_name("prometheus_metric_name"));
  }

  SUBCASE("invalid metric names") {
    CHECK_FALSE(metric::is_valid_metric_name(""));
    CHECK_FALSE(metric::is_valid_metric_name("123invalid"));
    CHECK_FALSE(metric::is_valid_metric_name("invalid-name"));
    CHECK_FALSE(metric::is_valid_metric_name("invalid.name"));
    CHECK_FALSE(metric::is_valid_metric_name("invalid name"));
  }

  SUBCASE("valid label keys") {
    CHECK(metric::is_valid_label_key("valid_key"));
    CHECK(metric::is_valid_label_key("ValidKey"));
    CHECK(metric::is_valid_label_key("valid123"));
    CHECK(metric::is_valid_label_key("_valid"));
  }

  SUBCASE("invalid label keys") {
    CHECK_FALSE(metric::is_valid_label_key(""));
    CHECK_FALSE(metric::is_valid_label_key("123invalid"));
    CHECK_FALSE(metric::is_valid_label_key("invalid:key"));
    CHECK_FALSE(metric::is_valid_label_key("invalid-key"));
    CHECK_FALSE(metric::is_valid_label_key("__reserved"));
    CHECK_FALSE(metric::is_valid_label_key("__internal"));
  }

  SUBCASE("valid label values") {
    CHECK(metric::is_valid_label_value("any_value"));
    CHECK(metric::is_valid_label_value("123"));
    CHECK(metric::is_valid_label_value("special-chars.allowed"));
    CHECK(metric::is_valid_label_value(""));
    CHECK(metric::is_valid_label_value("unicode测试"));
  }
}

TEST_CASE("counter basic functionality") {
  auto counter_family =
      metric::create_counter("test_counter", "Test counter help");

  SUBCASE("create counter with no labels") {
    auto counter = counter_family.create({});
    counter.inc(1.0);
    counter.inc(2.5);
    counter.inc(); // Default increment of 1.0
  }

  SUBCASE("create counter with labels") {
    auto counter =
        counter_family.create({{"method", "GET"}, {"status", "200"}});
    counter.inc(5.0);

    // Same labels should return same instance (idempotent)
    auto counter2 =
        counter_family.create({{"method", "GET"}, {"status", "200"}});
    counter2.inc(3.0);
  }

  SUBCASE("label sorting") {
    // Labels should be sorted by key
    auto counter1 =
        counter_family.create({{"z_key", "value"}, {"a_key", "value"}});
    auto counter2 =
        counter_family.create({{"a_key", "value"}, {"z_key", "value"}});

    // These should be the same instance due to label sorting
    counter1.inc(1.0);
    counter2.inc(2.0); // Should add to same counter
  }
}

TEST_CASE("gauge basic functionality") {
  auto gauge_family = metric::create_gauge("test_gauge", "Test gauge help");

  SUBCASE("gauge operations") {
    auto gauge = gauge_family.create({{"instance", "test"}});

    gauge.set(10.0);
    gauge.inc(5.0);
    gauge.dec(3.0);
    gauge.inc(); // Default increment
    gauge.dec(); // Default decrement
  }

  SUBCASE("gauge with multiple instances") {
    auto gauge1 = gauge_family.create({{"instance", "test1"}});
    auto gauge2 = gauge_family.create({{"instance", "test2"}});

    gauge1.set(100.0);
    gauge2.set(200.0);

    gauge1.inc(50.0);
    gauge2.dec(25.0);
  }
}

TEST_CASE("histogram basic functionality") {
  auto hist_family =
      metric::create_histogram("test_latency", "Test latency histogram",
                               metric::exponential_buckets(0.1, 2.0, 5));

  SUBCASE("histogram observations") {
    auto hist = hist_family.create({{"endpoint", "/api"}});

    hist.observe(0.05); // Below first bucket
    hist.observe(0.3);  // Between buckets
    hist.observe(1.5);  // Between buckets
    hist.observe(10.0); // Above all explicit buckets (goes in +Inf)
  }

  SUBCASE("histogram bucket validation") {
    // Buckets should be sorted and deduplicated, with +Inf added
    auto hist_family2 = metric::create_histogram(
        "test_hist2", "Test",
        std::initializer_list<double>{5.0, 1.0, 2.5, 1.0,
                                      0.5}); // Unsorted with duplicate

    auto hist = hist_family2.create({});
    hist.observe(0.1);
    hist.observe(1.5);
    hist.observe(100.0); // Should go in +Inf bucket
  }
}

TEST_CASE("histogram bucket generators") {
  SUBCASE("linear_buckets basic functionality") {
    // Linear buckets: start=0, width=10, count=5 -> {0, 10, 20, 30, 40}
    auto buckets = metric::linear_buckets(0.0, 10.0, 5);

    CHECK(buckets.size() == 5); // exactly count buckets
    CHECK(buckets[0] == 0.0);
    CHECK(buckets[1] == 10.0);
    CHECK(buckets[2] == 20.0);
    CHECK(buckets[3] == 30.0);
    CHECK(buckets[4] == 40.0);
  }

  SUBCASE("linear_buckets with non-zero start") {
    // Linear buckets: start=5, width=2.5, count=3 -> {5, 7.5, 10}
    auto buckets = metric::linear_buckets(5.0, 2.5, 3);

    CHECK(buckets.size() == 3);
    CHECK(buckets[0] == 5.0);
    CHECK(buckets[1] == 7.5);
    CHECK(buckets[2] == 10.0);
  }

  SUBCASE("linear_buckets edge cases") {
    // Zero count should give empty vector
    auto zero_buckets = metric::linear_buckets(100.0, 10.0, 0);
    CHECK(zero_buckets.size() == 0);

    // Negative start should work
    auto negative_buckets = metric::linear_buckets(-10.0, 5.0, 2);
    CHECK(negative_buckets.size() == 2);
    CHECK(negative_buckets[0] == -10.0);
    CHECK(negative_buckets[1] == -5.0);
  }

  SUBCASE("exponential_buckets basic functionality") {
    // Exponential buckets: start=1, factor=2, count=5 -> {1, 2, 4, 8, 16}
    auto buckets = metric::exponential_buckets(1.0, 2.0, 5);

    CHECK(buckets.size() == 5); // exactly count buckets
    CHECK(buckets[0] == 1.0);
    CHECK(buckets[1] == 2.0);
    CHECK(buckets[2] == 4.0);
    CHECK(buckets[3] == 8.0);
    CHECK(buckets[4] == 16.0);
  }

  SUBCASE("exponential_buckets different factor") {
    // Exponential buckets: start=0.1, factor=10, count=3 -> {0.1, 1, 10}
    auto buckets = metric::exponential_buckets(0.1, 10.0, 3);

    CHECK(buckets.size() == 3);
    CHECK(buckets[0] == doctest::Approx(0.1));
    CHECK(buckets[1] == doctest::Approx(1.0));
    CHECK(buckets[2] == doctest::Approx(10.0));
  }

  SUBCASE("exponential_buckets typical latency pattern") {
    // Typical web service latency buckets: 5ms, 10ms, 20ms, 40ms, 80ms, etc.
    auto buckets = metric::exponential_buckets(0.005, 2.0, 8);

    CHECK(buckets.size() == 8);
    CHECK(buckets[0] == doctest::Approx(0.005)); // 5ms
    CHECK(buckets[1] == doctest::Approx(0.010)); // 10ms
    CHECK(buckets[2] == doctest::Approx(0.020)); // 20ms
    CHECK(buckets[3] == doctest::Approx(0.040)); // 40ms
    CHECK(buckets[4] == doctest::Approx(0.080)); // 80ms
    CHECK(buckets[5] == doctest::Approx(0.160)); // 160ms
    CHECK(buckets[6] == doctest::Approx(0.320)); // 320ms
    CHECK(buckets[7] == doctest::Approx(0.640)); // 640ms
  }

  SUBCASE("exponential_buckets edge cases") {
    // Zero count should give empty vector
    auto zero_buckets = metric::exponential_buckets(5.0, 3.0, 0);
    CHECK(zero_buckets.size() == 0);
  }

  SUBCASE("bucket generators with histogram creation") {
    // Test that generated buckets work correctly with histogram creation
    auto linear_hist = metric::create_histogram(
        "linear_test", "Linear test", metric::linear_buckets(0, 100, 5));
    auto linear_instance = linear_hist.create({{"type", "linear"}});

    // Test observations fall into expected buckets
    linear_instance.observe(50);   // Should fall into 100 bucket
    linear_instance.observe(150);  // Should fall into 200 bucket
    linear_instance.observe(1000); // Should fall into +Inf bucket

    auto exp_hist =
        metric::create_histogram("exp_test", "Exponential test",
                                 metric::exponential_buckets(0.001, 10.0, 4));
    auto exp_instance = exp_hist.create({{"type", "exponential"}});

    // Test typical latency measurements
    exp_instance.observe(0.0005); // Should fall into 0.001 bucket (1ms)
    exp_instance.observe(0.005);  // Should fall into 0.01 bucket (10ms)
    exp_instance.observe(0.05);   // Should fall into 0.1 bucket (100ms)
    exp_instance.observe(5.0);    // Should fall into +Inf bucket
  }

  SUBCASE("prometheus compatibility verification") {
    // Verify our bucket generation matches Prometheus Go client behavior

    // Linear buckets equivalent to Prometheus LinearBuckets(0, 10, 5)
    auto our_linear = metric::linear_buckets(0, 10, 5);
    std::vector<double> expected_linear = {0, 10, 20, 30, 40};
    CHECK(our_linear == expected_linear);

    // Exponential buckets equivalent to Prometheus ExponentialBuckets(1, 2, 5)
    auto our_exp = metric::exponential_buckets(1, 2, 5);
    std::vector<double> expected_exp = {1, 2, 4, 8, 16};
    CHECK(our_exp == expected_exp);

    // Default Prometheus histogram buckets (exponential)
    auto default_buckets = metric::exponential_buckets(0.005, 2.5, 9);
    // Should be: .005, .0125, .03125, .078125, .1953125,
    // .48828125, 1.220703125, 3.0517578125, 7.62939453125
    CHECK(default_buckets.size() == 9);
    CHECK(default_buckets[0] == doctest::Approx(0.005));
    CHECK(default_buckets[1] == doctest::Approx(0.0125));
    CHECK(default_buckets[8] == doctest::Approx(7.62939453125));
  }
}

TEST_CASE("callback-based metrics") {
  auto counter_family =
      metric::create_counter("callback_counter", "Callback counter");
  auto gauge_family = metric::create_gauge("callback_gauge", "Callback gauge");

  SUBCASE("counter callback") {

    counter_family.register_callback({{"type", "callback"}},
                                     []() { return 42.0; });

    // Callback should be called during render
    Arena arena;
    auto output = metric::render(arena);
    CHECK(output.size() > 0);
  }

  SUBCASE("gauge callback") {
    gauge_family.register_callback({{"type", "callback"}},
                                   []() { return 123.5; });

    Arena arena;
    auto output = metric::render(arena);
    CHECK(output.size() > 0);
  }

  SUBCASE("callback conflict detection") {
    // First create a static instance
    auto counter = counter_family.create({{"conflict", "test"}});
    counter.inc(1.0);

    // Then try to register a callback with same labels - should abort
    // This is a validation test that would abort in debug builds
  }
}

TEST_CASE("prometheus text format rendering") {
  Arena arena;

  // Create some metrics
  auto counter_family =
      metric::create_counter("http_requests_total", "Total HTTP requests");
  auto counter = counter_family.create({{"method", "GET"}, {"status", "200"}});
  counter.inc(1000);

  auto gauge_family =
      metric::create_gauge("memory_usage_bytes", "Memory usage");
  auto gauge = gauge_family.create({{"type", "heap"}});
  gauge.set(1048576);

  auto hist_family = metric::create_histogram(
      "request_duration_seconds", "Request duration",
      metric::exponential_buckets(0.1, 2.0, 3)); // 0.1, 0.2, 0.4, 0.8
  auto hist = hist_family.create({{"handler", "api"}});
  hist.observe(0.25);
  hist.observe(0.75);
  hist.observe(1.5);

  SUBCASE("render format validation") {
    auto output = metric::render(arena);
    CHECK(output.size() > 0);

    // Basic format checks
    bool found_help = false;
    bool found_type = false;
    bool found_metric_line = false;

    for (const auto &line : output) {
      if (line.starts_with("# HELP"))
        found_help = true;
      if (line.find("# TYPE") != line.npos)
        found_type = true;
      if (line.find("http_requests_total") != std::string_view::npos)
        found_metric_line = true;
    }

    CHECK(found_help);
    CHECK(found_type);
    CHECK(found_metric_line);
  }

  SUBCASE("special value formatting") {
    auto special_gauge_family =
        metric::create_gauge("special_values", "Special value test");
    auto special_gauge = special_gauge_family.create({});

    special_gauge.set(std::numeric_limits<double>::infinity());
    auto output = metric::render(arena);

    // Should contain "+Inf" representation
    bool found_inf = false;
    for (const auto &line : output) {
      if (line.find("+Inf") != std::string_view::npos) {
        found_inf = true;
        break;
      }
    }
    CHECK(found_inf);
  }
}

TEST_CASE("thread safety") {
  constexpr int num_threads = 8;
  constexpr int ops_per_thread = 1000;

  SUBCASE("counter single-writer semantics") {
    auto counter_family =
        metric::create_counter("thread_test_counter", "Thread test");

    std::vector<std::thread> threads;
    std::latch start_latch{num_threads};

    // Each thread creates its own counter instance (safe)
    for (int i = 0; i < num_threads; ++i) {
      threads.emplace_back([&, i]() {
        auto counter =
            counter_family.create({{"thread_id", std::to_string(i)}});

        start_latch.arrive_and_wait();

        for (int j = 0; j < ops_per_thread; ++j) {
          counter.inc(1.0);
        }
      });
    }

    for (auto &t : threads) {
      t.join();
    }
  }

  SUBCASE("gauge multi-writer contention") {
    auto gauge_family =
        metric::create_gauge("thread_test_gauge", "Thread test gauge");

    std::vector<std::thread> threads;
    std::latch start_latch{num_threads};

    // Multiple threads create gauges with the same labels, writing to the same
    // underlying state, testing CAS contention.
    for (int i = 0; i < num_threads; ++i) {
      threads.emplace_back([&]() {
        auto gauge = gauge_family.create({{"shared", "true"}});
        start_latch.arrive_and_wait();

        for (int j = 0; j < ops_per_thread; ++j) {
          gauge.inc(1.0);
        }
      });
    }

    for (auto &t : threads) {
      t.join();
    }
  }

  SUBCASE("histogram single-writer per thread") {
    auto hist_family =
        metric::create_histogram("thread_test_hist", "Thread test histogram",
                                 std::initializer_list<double>{0.1, 0.5, 1.0});

    std::vector<std::thread> threads;
    std::latch start_latch{num_threads};

    for (int i = 0; i < num_threads; ++i) {
      threads.emplace_back([&, i]() {
        auto hist = hist_family.create({{"thread_id", std::to_string(i)}});

        start_latch.arrive_and_wait();

        for (int j = 0; j < ops_per_thread; ++j) {
          hist.observe(static_cast<double>(j) / ops_per_thread);
        }
      });
    }

    for (auto &t : threads) {
      t.join();
    }
  }

  SUBCASE("concurrent render calls") {
    // Multiple threads calling render concurrently should be safe (serialized
    // by mutex)
    auto counter_family = metric::create_counter("render_test", "Render test");
    auto counter = counter_family.create({});
    counter.inc(100);

    std::vector<std::thread> threads;
    std::latch start_latch{num_threads};
    std::atomic<int> success_count{0};

    for (int i = 0; i < num_threads; ++i) {
      threads.emplace_back([&]() {
        start_latch.arrive_and_wait();

        Arena arena;
        auto output = metric::render(arena);
        if (output.size() > 0) {
          success_count.fetch_add(1);
        }
      });
    }

    for (auto &t : threads) {
      t.join();
    }

    CHECK(success_count.load() == num_threads);
  }
}

TEST_CASE("thread counter cleanup bug") {
  return;
  SUBCASE(
      "counter and histogram values should persist after thread destruction") {
    auto counter_family = metric::create_counter(
        "thread_cleanup_counter", "Counter for thread cleanup test");
    auto histogram_family = metric::create_histogram(
        "thread_cleanup_histogram", "Histogram for thread cleanup test",
        metric::linear_buckets(0.0, 1.0, 5)); // buckets: 0, 1, 2, 3, 4

    // Variables to collect actual values from worker thread
    double counter_value_in_thread = 0;
    double histogram_sum_in_thread = 0;

    // Create thread that increments metrics and then exits
    std::thread worker([&]() {
      auto counter = counter_family.create({{"worker", "cleanup_test"}});
      auto histogram = histogram_family.create({{"worker", "cleanup_test"}});

      counter.inc(1.0);
      histogram.observe(1.5); // Should contribute to sum

      // Measure actual values from within the thread (before ThreadInit
      // destructor runs)
      Arena thread_arena;
      auto thread_output = metric::render(thread_arena);

      for (const auto &line : thread_output) {
        if (line.find("thread_cleanup_counter{worker=\"cleanup_test\"}") !=
            std::string_view::npos) {
          auto space_pos = line.rfind(' ');
          if (space_pos != std::string_view::npos) {
            auto value_str = line.substr(space_pos + 1);
            if (value_str.back() == '\n') {
              value_str.remove_suffix(1);
            }
            counter_value_in_thread = std::stod(std::string(value_str));
          }
        }
        if (line.find(
                "thread_cleanup_histogram_sum{worker=\"cleanup_test\"}") !=
            std::string_view::npos) {
          auto space_pos = line.rfind(' ');
          if (space_pos != std::string_view::npos) {
            auto value_str = line.substr(space_pos + 1);
            if (value_str.back() == '\n') {
              value_str.remove_suffix(1);
            }
            histogram_sum_in_thread = std::stod(std::string(value_str));
          }
        }
      }
    });

    // Wait for thread to complete and destroy (triggering ThreadInit
    // destructor)
    worker.join();

    // Measure values after thread cleanup
    Arena arena;
    auto output = metric::render(arena);

    double counter_value_after = 0;
    double histogram_sum_after = 0;

    for (const auto &line : output) {
      if (line.find("thread_cleanup_counter{worker=\"cleanup_test\"}") !=
          std::string_view::npos) {
        auto space_pos = line.rfind(' ');
        if (space_pos != std::string_view::npos) {
          auto value_str = line.substr(space_pos + 1);
          if (value_str.back() == '\n') {
            value_str.remove_suffix(1);
          }
          counter_value_after = std::stod(std::string(value_str));
        }
      }
      if (line.find("thread_cleanup_histogram_sum{worker=\"cleanup_test\"}") !=
          std::string_view::npos) {
        auto space_pos = line.rfind(' ');
        if (space_pos != std::string_view::npos) {
          auto value_str = line.substr(space_pos + 1);
          if (value_str.back() == '\n') {
            value_str.remove_suffix(1);
          }
          histogram_sum_after = std::stod(std::string(value_str));
        }
      }
    }

    // Values should have been captured correctly within the thread
    CHECK(counter_value_in_thread == 1.0);
    CHECK(histogram_sum_in_thread == 1.5);

    // The bug: These values should persist after thread cleanup but will be
    // lost because ThreadInit destructor erases per-thread state without
    // accumulating values
    CHECK(counter_value_after == 1.0);
    CHECK(histogram_sum_after == 1.5);

    // The bug: After thread destruction, the counter and histogram values are
    // lost because ThreadInit::~ThreadInit() calls
    // family->perThreadState.erase(thread_id) without accumulating the values
    // into global storage first. This causes counter values to "go backwards"
    // when threads are destroyed, violating the monotonic property of counters.
  }
}

TEST_CASE("memory management") {
  SUBCASE("arena allocation in render") {
    Arena arena;
    auto initial_used = arena.used_bytes();

    auto counter_family = metric::create_counter("memory_test", "Memory test");
    auto counter = counter_family.create(
        {{"large_label", "very_long_value_that_takes_space"}});
    counter.inc(42);

    auto output = metric::render(arena);
    auto final_used = arena.used_bytes();

    CHECK(output.size() > 0);
    CHECK(final_used > initial_used); // Arena was used for string allocation

    // All string_views should point to arena memory
    for (const auto &line : output) {
      CHECK(line.size() > 0);
    }
  }

  SUBCASE("arena reset behavior") {
    Arena arena;

    auto counter_family = metric::create_counter("reset_test", "Reset test");
    auto counter = counter_family.create({});
    counter.inc(1);

    // Render multiple times with arena resets
    for (int i = 0; i < 5; ++i) {
      auto output = metric::render(arena);
      CHECK(output.size() > 0);
      arena.reset(); // Should not affect metric values, only arena memory
    }

    // Final render should still work
    auto final_output = metric::render(arena);
    CHECK(final_output.size() > 0);
  }
}

TEST_CASE("histogram pending buffer thread cleanup bug") {
  for (int iterations = 0; iterations < 1000; ++iterations) {
    // This test demonstrates the bug where pending histogram observations
    // are lost when a thread dies because ThreadInit destructor doesn't
    // flush pending data into shared before accumulating into global state.

    metric::reset_metrics_for_testing();

    auto hist_family = metric::create_histogram(
        "pending_bug_test", "Test histogram for pending buffer bug",
        {1.0}); // Single bucket for simplicity

    std::atomic<bool> keep_rendering{true};
    constexpr int num_threads = 100;

    std::latch ready{2};

    // Background thread that calls render in a tight loop to hold global mutex
    std::thread render_thread([&]() {
      ready.arrive_and_wait();
      Arena arena;
      while (keep_rendering.load(std::memory_order_relaxed)) {
        metric::render(arena);
        arena.reset();
      }
    });
    // Don't spawn threads until render thread is running
    ready.arrive_and_wait();

    // Spawn threads that observe once and exit
    std::vector<std::thread> observer_threads;
    for (int i = 0; i < num_threads; ++i) {
      observer_threads.emplace_back([&hist_family]() {
        auto hist = hist_family.create({{"test", "observer"}});
        hist.observe(0.5); // Goes into first bucket (le="1.0")
        // Thread dies here - pending observations should be lost due to bug
      });
    }

    // Join all observer threads
    for (auto &t : observer_threads) {
      t.join();
    }

    // Stop render thread
    keep_rendering.store(false, std::memory_order_relaxed);
    render_thread.join();

    // Check if the worker's observations were preserved
    Arena arena;
    auto output = metric::render(arena);

    // First, let's debug what we actually got
    std::ostringstream debug_output;
    for (const auto &line : output) {
      debug_output << line;
    }
    std::string full_output = debug_output.str();

    // Parse the output to find the worker's bucket count for le="2.0"
    uint64_t worker_bucket_2_count = 0;
    bool found_worker_metric = false;

    // The render output alternates between metric name and value in separate
    // string_views
    for (size_t i = 0; i < output.size(); ++i) {
      const auto &line = output[i];
      // Look for: pending_bug_test_bucket{test="observer",le="1.0"}
      if (line.find("pending_bug_test_bucket{test=\"observer\",le=\"1.0\"}") !=
          std::string_view::npos) {
        found_worker_metric = true;
        // The value should be in the next element
        if (i + 1 < output.size()) {
          auto value_str = output[i + 1];
          // Remove trailing newline if present
          while (!value_str.empty() &&
                 (value_str.back() == '\n' || value_str.back() == '\r')) {
            value_str.remove_suffix(1);
          }
          try {
            worker_bucket_2_count = std::stoull(std::string(value_str));
          } catch (const std::exception &e) {
            MESSAGE("Failed to parse value: '"
                    << value_str << "' from metric line: '" << line << "'");
            MESSAGE("Full output:\n" << full_output);
            throw;
          }
        }
        break;
      }
    }

    REQUIRE(found_worker_metric); // The metric should exist

    // BUG: This will fail because pending observations are lost on thread death
    // Expected: num_threads observations (each thread made 1 observation)
    // Actual: less than num_threads (observations stuck in pending are lost
    // when threads die)
    CHECK_MESSAGE(
        worker_bucket_2_count == num_threads,
        "Expected "
            << num_threads << " observations but got " << worker_bucket_2_count
            << ". This indicates the pending buffer bug where observations "
            << "stuck in pending are lost when thread dies.");
  }
}

TEST_CASE("render output deterministic order golden test") {
  // Clean slate - reset all metrics before this test
  metric::reset_metrics_for_testing();

  Arena arena;

  // Create a comprehensive set of metrics with deliberate ordering
  // to test deterministic output

  // Create counters with different family names and labels
  auto z_counter_family =
      metric::create_counter("z_last_counter", "Last counter alphabetically");
  auto z_counter =
      z_counter_family.create({{"method", "POST"}, {"handler", "api"}});
  z_counter.inc(42.0);

  auto a_counter_family =
      metric::create_counter("a_first_counter", "First counter alphabetically");
  auto a_counter1 = a_counter_family.create({{"status", "200"}});
  auto a_counter2 = a_counter_family.create(
      {{"method", "GET"}}); // Should come before status lexicographically
  a_counter1.inc(100.0);
  a_counter2.inc(200.0);

  // Create gauges with different orderings
  auto m_gauge_family = metric::create_gauge("m_middle_gauge", "Middle gauge");
  auto m_gauge = m_gauge_family.create({{"type", "memory"}});
  m_gauge.set(1024.0);

  auto b_gauge_family = metric::create_gauge("b_second_gauge", "Second gauge");
  auto b_gauge = b_gauge_family.create({{"region", "us-west"}});
  b_gauge.set(256.0);

  // Create histograms
  auto x_hist_family = metric::create_histogram("x_histogram", "Test histogram",
                                                {0.1, 0.5, 1.0});
  auto x_hist = x_hist_family.create({{"endpoint", "/api/v1"}});
  x_hist.observe(0.25);
  x_hist.observe(0.75);

  // Add some callbacks to test callback ordering
  a_counter_family.register_callback({{"callback", "test"}},
                                     []() { return 123.0; });
  m_gauge_family.register_callback({{"callback", "dynamic"}},
                                   []() { return 456.0; });

  // Render the metrics
  auto output = metric::render(arena);

  // Concatenate all output into a single string
  std::ostringstream oss;
  for (const auto &line : output) {
    oss << line;
  }
  std::string actual_output = oss.str();

  // Define expected golden output - this represents the exact expected
  // deterministic order
  std::string expected_golden =
      "# HELP a_first_counter First counter alphabetically\n"
      "# TYPE a_first_counter counter\n"
      "a_first_counter{callback=\"test\"} 123.0\n"
      "a_first_counter{method=\"GET\"} 200.0\n"
      "a_first_counter{status=\"200\"} 100.0\n"
      "# HELP z_last_counter Last counter alphabetically\n"
      "# TYPE z_last_counter counter\n"
      "z_last_counter{handler=\"api\",method=\"POST\"} 42.0\n"
      "# HELP b_second_gauge Second gauge\n"
      "# TYPE b_second_gauge gauge\n"
      "b_second_gauge{region=\"us-west\"} 256.0\n"
      "# HELP m_middle_gauge Middle gauge\n"
      "# TYPE m_middle_gauge gauge\n"
      "m_middle_gauge{callback=\"dynamic\"} 456.0\n"
      "m_middle_gauge{type=\"memory\"} 1024.0\n"
      "# HELP x_histogram Test histogram\n"
      "# TYPE x_histogram histogram\n"
      "x_histogram_bucket{endpoint=\"/api/v1\",le=\"0.1\"} 0\n"
      "x_histogram_bucket{endpoint=\"/api/v1\",le=\"0.5\"} 1\n"
      "x_histogram_bucket{endpoint=\"/api/v1\",le=\"1.0\"} 2\n"
      "x_histogram_bucket{endpoint=\"/api/v1\",le=\"+Inf\"} 2\n"
      "x_histogram_sum{endpoint=\"/api/v1\"} 1.0\n"
      "x_histogram_count{endpoint=\"/api/v1\"} 2\n";

  // Check if output matches golden file
  if (actual_output != expected_golden) {
    MESSAGE("Render output does not match expected golden output.");
    MESSAGE("This indicates the deterministic ordering has changed.");
    MESSAGE("Expected output:\n" << expected_golden);
    MESSAGE("Actual output:\n" << actual_output);
    CHECK(false); // Force test failure
  } else {
    CHECK(true); // Test passes
  }
}