Getting started

FerroReplay keeps time explicit. A strategy, matching engine, or analytics pipeline should depend on an injected Clock and let the replay producer advance VirtualClock from ordered event metadata instead of reading host wall-clock time.

Add the crate

[dependencies]
ferro-replay = { version = "0.1" }

Clock boundary

Production services usually receive Arc<dyn Clock>. Live paths use SystemClock; replay and tests use VirtualClock.

use std::sync::Arc;
 
use ferro_replay::{Clock, SystemClock, VirtualClock};
 
fn live_clock() -> Arc<dyn Clock> {
    Arc::new(SystemClock::new())
}
 
fn replay_clock(start_ns: i64) -> VirtualClock {
    VirtualClock::new(start_ns)
}

Contract

• Event order is explicit.
• Time comes from Clock, not direct process wall-clock reads.
• Replay advances VirtualClock to the next event timestamp before dispatch.
• Timers use interval so live and replay cadence behavior share one path.

Use the API reference for exact clock, virtual-clock, and interval types.