Skip to content

🚀 Performance Notes (GoMT4)

Practical tips to keep GoMT4 fast, stable, and resource‑efficient. Focus is on your current codebase (examples/, pb API, MT4 terminal on Windows).

🎯 Goals

  • Low latency for quotes/updates.
  • Throughput for bulk reads (history, many symbols).
  • Stability under loss/reconnects.
  • Predictable CPU/RAM usage.

🔥 Hot paths to watch

Keep these tight: avoid heavy logging, allocations, and repeated RPCs.

📦 Batch, don’t loop RPCs

Why: fewer round‑trips → lower latency and CPU per message.

📡 Streaming tuning

  • Backpressure: use the server’s knobs when available.

  • Example: OpenedOrdersProfitStreamRequest{ buffer_size } — set buffer high enough to absorb short spikes, but not so high that consumer lags unnoticed.

  • Consumer loop: drain fast, do work in a goroutine pool if processing is heavy.
  • Tick throttling: if you don’t need every tick, aggregate with a time.Ticker (e.g., 100–200ms) before updating UI/logs.
  • Timeouts: long‑lived streams should use WithCancel, not short timeouts; add an idle watchdog to reconnect when no data for X seconds.
ctx, cancel := context.WithCancel(parent)
defer cancel()
updates, errCh := svc.StreamQuotes(ctx, symbols)
for {
    select {
    case u := <-updates: /* cheap handling */
    case err := <-errCh: /* reconnect + backoff */
    case <-idle.C:      /* no data → ping/refresh */
    }
}

🧵 Concurrency patterns (Go)

  • One gRPC connection per MT4 instance; reuse *grpc.ClientConn (creating many is expensive).
  • Worker pool for CPU‑bound post‑processing (e.g., statistics on history), not for issuing parallel trade RPCs to the same MT4 terminal.
  • Channel fan‑out: if several consumers need the same stream, broadcast from one reader goroutine; don’t open multiple identical streams.

🧰 Symbol metadata cache

  • Cache ShowSymbolParams and ShowTickValues in memory (map by symbol). These values change rarely; re‑fetch only on symbol list changes or once per session.
  • Use cached digits/point/lot_step for rounding before any trade call (see Cookbook → RoundVolumePrice.md).

🧮 History I/O

  • Request time windows in chunks (days/weeks) and merge client‑side.
  • Prefer H1/H4/D1 where possible; M1 over long ranges is expensive.
  • For analytics, stream and process as you go instead of storing everything first.

🗜️ Compression & payloads

  • gRPC compression helps on large history responses; avoid compressing ultra‑short ticks.
  • Avoid logging full payloads in hot paths; log only counts and latency metrics.

🧠 Allocations & GC

  • Reuse buffers/slices for quote aggregation; pre‑allocate with make(..., cap) when size is known.
  • Convert time ↔ Timestamp carefully in tight loops; avoid repeated parsing/formatting.
  • Use strings.Builder for building CSV/TSV outputs.

⏱️ Timeouts & retries

  • Unary calls: context.WithTimeout (100ms–3s typical for local terminal). Add idempotency where safe and exponential backoff.
  • Streams: no short deadline; reconnect on error/idle with jittered backoff (see Cookbook → HandleReconnect.md, UnaryRetries.md).

🖥️ Windows & MT4 specifics

  • Keep MT4 responsive: exclude MT4 data directory from real‑time antivirus scanning if safe in your environment.
  • Power plan: use High performance to avoid CPU sleep throttling for low‑latency streams.
  • Symbol visibility: ensure symbols are visible before quote/history calls (Cookbook → EnsureSymbolVisible.md). Pre‑subscribe to reduce first‑tick latency.

đź“Š Observability

  • Add lightweight metrics: per‑RPC latency, stream message rate, reconnect counters.
  • Sampled logs (1/N) for hot paths; full debug logs only on demand.

âś… Checklist (TL;DR)

  • Use batch requests and paged/streamed history.
  • Keep one gRPC conn; reuse clients.
  • Cache SymbolParams / TickValues.
  • Fast stream consumers; apply throttling/aggregation if UI‑facing.
  • Timeouts for unary; cancel for streams; exponential backoff with jitter.
  • Minimal logging in hot paths; metrics for rates/latency.
  • Pre‑make symbols visible or subscribe early.
  • Prefer higher‑TF history when possible; chunk long ranges.

🗺️ Code map → where each tip lives in your repo

1) Batch, don’t loop RPCs

  • Batch quotes → examples/mt4/MT4_service.go → func (s *MT4Service) ShowQuotesMany(ctx context.Context, symbols []string) Search: ShowQuotesMany(
  • Single quote (avoid loops) → same file → ShowQuote(ctx, symbol string) Search: ShowQuote(ctx
  • History (paged/stream) → examples/mt4/MT4_service.go → StreamOrdersHistoryExample, StreamQuoteHistoryExample

2) Streaming tuning & backpressure

Tip → buffer_size: if you expose buffer_size for profits, it is wired at account layer:

  • examples/mt4/MT4Account.go → look for OnOpenedOrdersProfit(ctx,

3) One gRPC connection, reuse clients

  • Connection bootstrap → examples/main.go → creation of service & clients Search: NewMT4Service( / grpc.Dial(
  • Account/session holder → examples/mt4/MT4Account.go Search: type MT4Account struct / connect / login

4) Symbol metadata cache (params/tick-values)

  • Fetch params → examples/mt4/MT4_service.go → ShowSymbolParams
  • Fetch tick values → same file → ShowTickValues
  • Used for rounding → see order helpers (e.g., ShowOrderSendExample) where Digits, Point, LotStep are applied before send.

5) Rounding & validation before trades

6) Ensure symbol visible / pre‑subscribe

  • Ensure visible → examples/mt4/MT4Account.go Search: EnsureSymbolVisible / SymbolSelect / SymbolsGet
  • Pre‑subscribe quotes (lower first‑tick latency) → examples/mt4/MT4_service.go → StreamQuotes

7) History I/O in chunks

8) Timeouts & retries

  • Unary with timeout → examples/mt4/MT4_service.go → places that use context.WithTimeout(...) around quote/symbol calls.
  • Reconnect with backoff → reliability helpers you documented (Cookbook): Cookbook/Reliability/HandleReconnect.md and Cookbook/Reliability/UnaryRetries.md correspond to the code in examples/mt4/MT4_service.go (stream loops) and in connection handling in MT4Account.go.

9) Observability (logs/metrics)

  • Light logs around hot paths → examples/mt4/MT4_service.go stream handlers print Tick/Profit/Trade lines — replace with counters/rate meters in production.