LIMITBOOK_

Nasdaq TotalView‑ITCH 5.0 · 2019‑12‑30 · 12:00–12:20 ET

Real Nasdaq depth,
rebuilt live
in your browser.

limitbook is a no_std Rust ITCH 5.0 engine, compiled to WebAssembly — parsing the raw exchange feed zero‑copy and reconstructing the limit order book for AAPL, TSLA and SPY as you watch.

engine throughput · measured in this tab msg/s

wall‑clock rate through the wasm engine, updating as the replay below runs

limitbook replay — itch50_20191230_midday · wasm32 AAPL · TSLA · SPY — 157,824 messages, mid‑day continuous trading
feed clock
best bid
spread
best ask
queue · price‑time
time & sales
cumulative depth
messages
replay pace
live orders
invariant violations
book checks
deep verify /1k msgs
Loading engine…

space play/pause  ·  15 speed  ·  seek  ·  click a price to inspect its queue, an order to track it, esc to reset every number on screen is read back out of the engine

What you're watching

One engine. The CLI runs it natively;
your browser runs it in WebAssembly.

No simulation layer, no pre-rendered data — the tab you're in does the same work a feed handler does, in four steps.

01

Raw ITCH bytes in

The wire format, unmodified

Your browser downloads ~1.6 MB of gzipped ITCH 5.0 — the binary format Nasdaq actually disseminates — and decodes it zero‑copy, reading fields straight off the byte slice with no intermediate allocation.

157,824 length‑prefixed messages
adds · executions · cancels · deletes · replaces

02

Price‑time priority queue

The same Rust the CLI runs

Each message mutates a limit order book kept in strict price‑time priority by limitbook-core, a #![no_std] crate compiled unchanged to wasm32-unknown-unknown — the exact code path behind the native CLI.

03

Reconstructed live, then proven

Tick‑for‑tick with the CLI

Every level above is computed in your tab and deep‑checked every 1,000 messages. At the end, the final book is compared tick‑for‑tick against limitbook replay — the parity badge you see land.

Measured, not estimated

The numbers behind the demo.

Every figure here is reproducible from the repo — the same fixtures, the same benchmark harness, and the same CI that guards the parity you just watched.

268.7M
messages · full trading day, zero violations

The complete 2019‑12‑30 TotalView sample day — 268,744,780 messages — replayed end to end through parser and book with zero invariant violations and zero parse errors.

31.9M msg/s
zero‑copy parse · single core

≈1 GB/s of raw feed decoded on one Xeon core, verified by a per‑run timestamp checksum. Full book reconstruction included: 1.38M msg/s.

wasm ≡ CLI
parity · enforced by CI

The browser build reproduces the CLI's final book state tick‑for‑tick on both committed fixtures — asserted on every push, and again live at the end of the replay above.

Methodology, hardware, and flamegraphs: BENCHMARKS.md.