跟踪crate

提供日志记录、指标、内存和性能分析

在提供高可见性的同时，对执行关键路径的影响最小，tracing专注于为高性能应用提供可预测的性能。它最初是为视频游戏引擎设计的。

与其他跟踪crate不同，tracing不提供单个事件的钩子，而是一个事件流，在内部利用transit将事件序列化为二进制格式。旨在在进程后期消费，但也可以有效地通过网络发送。

示例

use micromegas_tracing::{
   span_scope, info, warn, error, debug, imetric, fmetric, guards, event,
};

// Initialize tracing, here with a null event sink, see `lgn-telemetry-sink` crate for a proper implementation
// libraries don't need (and should not) setup any TracingSystemGuard
let _tracing_guard = guards::TracingSystemGuard::new(
    8 * 1024 * 1024,
    1024 * 1024,
    16 * 1024 * 1024,
    std::sync::Arc::new(event::NullEventSink {})
);
let _thread_guard = guards::TracingThreadGuard::new();

// Create a span scope, this will complete when the scope is dropped, and provide the time spent in the scope
// Behind the scene this uses a thread local storage
// on an i9-11950H this takes around 40ns
span_scope!("main");

// Logging
info!("Hello world");
warn!("Hello world");
error!("Hello world");
debug!("Hello world");

// Metrics
imetric!("name", "unit", 0);
fmetric!("name", "unit", 0.0);