Rust Order Book & Matching Engine

The problem

I wanted to understand order-book behavior through a core small enough to reason about, but strict enough that fills, ordering, and replay could be checked instead of hand-waved.

What I built

• Built typed market rules for prices, quantities, order IDs, validation, and structured errors.
• Implemented limit and market orders, FIFO price-time matching, partial fills, cancellation, snapshots, events, and deterministic replay.
• Added an Axum HTTP adapter and a Next.js learning terminal for submitting orders, changing market scenarios, and inspecting the live book.

Key decisions

• Use exact decimal arithmetic for engine values instead of floating-point amounts.
• Keep the matching engine single-threaded and in memory so ordering stays explicit and deterministic.
• Separate the core library from HTTP, serialization, and UI adapters.
• Use sequence numbers, invariant checks, property tests, benchmarks, and fuzz targets to make behavior measurable.

How I check the work

• Unit, integration, property, replay, serialization, and documentation tests pass with the locked dependency graph.
• Clippy passes across all targets and features with warnings denied.
• Replay tests confirm the same command sequence produces the same events and final book state.