v0.7.0

v0.7.0 (2025-12-14)

• breaking: replace OrderBookID with SequenceID in BookLog. All events now have a globally increasing SequenceID for ordering, deduplication, and rebuild synchronization. Use LogType to determine if the event affects order book state.
• feature: add QuoteSize field to Order for Market orders to specify amount in quote currency (e.g., USDT). Mutually exclusive with Size.
• feature: add RejectReason field to BookLog with constants for reject scenarios (no_liquidity, price_mismatch, insufficient_size, would_cross_spread)
• feature: add TradeID field to BookLog for sequential trade identification (only set on Match events)
• feature: add Amount field to BookLog with pre-calculated Price × Size (only set on Match events)
• perf: optimize atomic operations by using Add() return value instead of separate Load() call
• refactor: split handleOrder into separate handlers for each order type
• refactor: replace Trade with BookLog to capture Open, Match, and Cancel events for full order book reconstruction
• feature: add AmendOrder to support modifying order price and size
• perf: implement sync.Pool for BookLog to reduce GC pressure and achieve zero-allocation for log events
• feature: add LogTypeReject to distinguish orders that never entered the book (e.g. failed PostOnly/IOC/FOK) from cancellations
• feature: add CalculateDepthChange helper to simplify downstream depth updates
• docs: improve readme document and provide more detail info
• fix: FOK order validation incorrectly used single order size instead of price level total size
• fix: FOK order validation did not properly reject when price doesn't match
• fix: depth() function off-by-one error returning limit-1 items instead of limit
• feature: add OrderBook.Shutdown(ctx) for graceful shutdown with pending order drain
• feature: add MatchingEngine.Shutdown(ctx) for graceful shutdown of all markets in parallel
• perf: reduce channel buffer sizes from 1,000,000 to 10,000 for orders and 100 for depth queries to reduce memory usage
• perf: pre-allocate slice capacity in order handlers to reduce dynamic allocations
• perf: cache time.Now().UTC() per order to reduce syscall overhead in high-frequency scenarios
• perf: replace atomic operations with direct variable access in queue (single-goroutine access pattern)