09_refactor_signals

Prompt file imported from acid0ikario/trade-bot (.codex/prompts/09_refactor_signals.md). Copyright stays with the author.

You are acting as a performance-minded Python developer. Refactor indicator and signal computation for speed and clarity without changing behaviour.

Context

• Current generate_signal recomputes indicators repeatedly and directly on demand.
• Project uses pandas-based indicators (EMA/RSI) and may add ADX/volume filters.

Goals

• Compute indicators once per DataFrame, expose reusable helpers, and avoid lookahead.
• Preserve signal semantics (use last CLOSED candle).

Tasks

1. src/bot/strategy.py

• Add calculate_indicators(df, cfg) -> DataFrame:
  • Returns a copy with columns: ema_fast, ema_slow, rsi.
  • If filters enabled by config: also compute adx, vol_sma.
  • Use vectorized operations only; no Python loops.
  • Include docstring explaining inputs/outputs and lookahead safety.
• Refactor generate_signal to:
  • Use precomputed columns when present, else call calculate_indicators.
  • Enforce candle-close policy (ignore the last incomplete bar by default).
  • Keep existing entry/exit rules intact.
• Add small helpers for readability:
  • _trend_up(row), _rsi_in_range(row, cfg), _is_pullback(df).

2. Tests

• Add fixtures with deterministic OHLCV data and precomputed indicators.
• Verify signals are identical to baseline for the same dataset (golden master style).
• Optional: simple timing check to show fewer indicator computations across multiple calls.

Acceptance criteria

• No change in generated signals vs baseline tests.
• Clear, documented code; helpers have docstrings and type hints.
• No lookahead bias introduced; last bar handling is explicit.