m/market labResearch, with a record.
Source on GitHub ↗

Native market engine

The C++ engine reads historical ITCH messages and maintains the market's public order book. The Python financial engine separately maintains the experiment account and its simulated orders. A market-data Add message is not the user's trading request.

BinaryFILE reader → ITCH decoder → native order book
                                 ├─ standalone replay → CSV/checkpoint output
                                 └─ bridge/preparation → immutable prepared market data
Change Source / interface
Framing and file boundaries reader, input contract
Message decoding decoder, headers
Order book updates/invariants book, book interface
Python/native prepared bridge bridge, prepared market
Standalone CLI/output application, CLI guide
Checkpoint/recovery reference
Simulated network input reference

Prices and quantities use bounded integer values. The native decoder rejects malformed frames and unsupported message structures rather than guessing. The book tracks individual orders and aggregated price levels. Display depth is separate from complete internal book state. Python financial fills use their documented visible-depth simulation model; they are not exchange queue-position certification.

Build with make all; source checks are described in verification. The native tests include a separate Python reference model for synthetic events. Whole-day or sample measurements from older revisions remain historical evidence, not a new validation of the current source. The source map lists component entrypoints.

Section index · All documentation

Maintained source: docs/architecture/native-engine.md