🔂 UnaryRetries (GoMT4)

Goal: apply a consistent retry pattern for unary RPC calls (quotes, account, orders) using helpers already present in this repo.

Real code refs:

• Backoff & helpers: examples/mt4/MT4Account.go (waitWithCtx, backoffDelay, maxRetries)
• Typical calls: Quote, AccountSummary, OrderSend, OrderModify, OrderClose

---

✅ Principles

• Retry only transient transport errors: codes.Unavailable (and network I/O errors if wrapped accordingly).
• Use per‑call timeout via context.WithTimeout to bound latency.
• Between attempts, sleep with backoff + jitter using waitWithCtx(ctx, backoffDelay(attempt)).
• Stop immediately on non‑transient (business) errors.

---

🧱 Skeleton helper (mirrors your account layer)

func (a *MT4Account) callWithRetry(ctx context.Context, fn func(context.Context) error) error {
    var last error
    for attempt := 0; attempt < maxRetries; attempt++ {
        if err := fn(ctx); err != nil {
            if st, ok := status.FromError(err); ok && st.Code() == codes.Unavailable {
                // transient: back off and retry
                if err := waitWithCtx(ctx, backoffDelay(attempt)); err != nil {
                    return err // context cancelled/deadline
                }
                last = err
                continue
            }
            return err // non‑transient → bubble up
        }
        return nil // success
    }
    return fmt.Errorf("max retries reached: %w", last)
}

In this project, a similar logic is called directly from the account methods so as not to duplicate backups throughout the project.

---

⏱️ Per‑call timeout wrapper

func withTimeout(parent context.Context, d time.Duration, fn func(context.Context) error) error {
    ctx, cancel := context.WithTimeout(parent, d)
    defer cancel()
    return fn(ctx)
}

---

💱 Example: robust Quote

err := withTimeout(ctx, 3*time.Second, func(c context.Context) error {
    return a.callWithRetry(c, func(cc context.Context) error {
        q, err := a.Quote(cc, symbol)
        if err != nil { return err }
        log.Printf("%s %.5f/%.5f @ %s", symbol, q.GetBid(), q.GetAsk(), q.GetTime().AsTime())
        return nil
    })
})
if err != nil { return fmt.Errorf("quote failed: %w", err) }

---

🧾 Example: AccountSummary health‑check

err := withTimeout(ctx, 3*time.Second, func(c context.Context) error {
    return a.callWithRetry(c, func(cc context.Context) error {
        _, err := a.AccountSummary(cc)
        return err
    })
})
if err != nil { return fmt.Errorf("health‑check failed: %w", err) }

---

🛒 Example: OrderSend (market)

err := withTimeout(ctx, 8*time.Second, func(c context.Context) error {
    return a.callWithRetry(c, func(cc context.Context) error {
        _, err := a.OrderSend(cc, symbol, side, volume, nil, &slip, sl, tp, &comment, &magic, nil)
        return err
    })
})
if err != nil { return fmt.Errorf("OrderSend failed: %w", err) }

---

🎛️ Tuning

Constants in examples/mt4/MT4Account.go:

const (
    backoffBase = 300 * time.Millisecond
    backoffMax  = 5 * time.Second
    jitterRange = 200 * time.Millisecond
    maxRetries  = 10
)

• VPS/LAN: backoffBase=150ms, backoffMax=3–5s, timeouts 2–3s for reads.
• Home/unstable: backoffMax=8–10s, timeouts 4–6s (reads) / 6–10s (trades).

---

⚠️ Pitfalls

• We will delete business errors (for example, invalid volume/price) — no need, return it immediately.
• There is no defer cancel() — goroutin leaks.
• Too aggressive backoff — "pounding" on the network; increase the "backoffBase" and jitter spread.
• One global context for all — it is better to have a separate timeout for each call.

---

🔗 See also

• HandleReconnect.md — for streaming and general strategy.
• `Reliability (en)' — summary recommendations on timeouts/retreats.
• GetQuote.md, PlaceMarketOrder.md — where it is applied live.