perf: limit trend directional movement scan

app/market/market_analysis.py

@@ -187,32 +187,28 @@ def _compute_trend_metrics(
 
         # Simple ADX approximation (simplified version)
         if len(medium) >= 14:
-            # Directional movement
-            plus_dm = []
-            minus_dm = []
+            # Directional movement. The simplified ADX below only uses up to
+            # the final 14 values, so skip older bars instead of materializing
+            # full-window arrays for every feature computation.
+            plus_dm_sum = 0
+            minus_dm_sum = 0
 
-            for i in range(1, len(medium)):
+            for i in range(max(1, len(medium) - 14), len(medium)):
                 up_move = medium[i].high - medium[i-1].high
                 down_move = medium[i-1].low - medium[i].low
 
                 if up_move > down_move and up_move > 0:
-                    plus_dm.append(up_move)
-                else:
-                    plus_dm.append(0)
+                    plus_dm_sum += up_move
 
                 if down_move > up_move and down_move > 0:
-                    minus_dm.append(down_move)
-                else:
-                    minus_dm.append(0)
+                    minus_dm_sum += down_move
 
             # Smooth and compute ADX (simplified)
-            if plus_dm and minus_dm:
-                avg_plus = sum(plus_dm[-14:]) / 14
-                avg_minus = sum(minus_dm[-14:]) / 14
-
-                if avg_plus + avg_minus > 0:
-                    dx = 100 * abs(avg_plus - avg_minus) / (avg_plus + avg_minus)
-                    metrics["adx"] = dx
+            if plus_dm_sum + minus_dm_sum > 0:
+                avg_plus = plus_dm_sum / 14
+                avg_minus = minus_dm_sum / 14
+                dx = 100 * abs(avg_plus - avg_minus) / (avg_plus + avg_minus)
+                metrics["adx"] = dx
 
         # Moving average slope (20 period)
         if len(medium) >= 20:

tests/test_market_analysis.py

@@ -0,0 +1,93 @@
+from datetime import datetime, timedelta, timezone
+
+from app.market.market_analysis import MarketAnalyzer
+from app.schemas import Candle
+
+
+def _reference_trend_metrics(recent, medium, long):
+    metrics = {}
+
+    if len(medium) >= 14:
+        plus_dm = []
+        minus_dm = []
+
+        for i in range(1, len(medium)):
+            up_move = medium[i].high - medium[i - 1].high
+            down_move = medium[i - 1].low - medium[i].low
+
+            if up_move > down_move and up_move > 0:
+                plus_dm.append(up_move)
+            else:
+                plus_dm.append(0)
+
+            if down_move > up_move and down_move > 0:
+                minus_dm.append(down_move)
+            else:
+                minus_dm.append(0)
+
+        if plus_dm and minus_dm:
+            avg_plus = sum(plus_dm[-14:]) / 14
+            avg_minus = sum(minus_dm[-14:]) / 14
+
+            if avg_plus + avg_minus > 0:
+                dx = 100 * abs(avg_plus - avg_minus) / (avg_plus + avg_minus)
+                metrics["adx"] = dx
+
+    if len(medium) >= 20:
+        price_change = medium[-1].close - medium[-20].close
+
+        if medium[-20].close > 0:
+            metrics["ma_slope"] = price_change / medium[-20].close
+
+    if len(recent) >= 10 and len(medium) >= 20:
+        recent_range = max(c.high for c in recent[-10:]) - min(c.low for c in recent[-10:])
+        medium_range = max(c.high for c in medium[-20:]) - min(c.low for c in medium[-20:])
+
+        if medium_range > 0:
+            range_ratio = recent_range / medium_range
+            metrics["range_compression_flag"] = range_ratio < 0.7
+            metrics["range_expansion_flag"] = range_ratio > 1.3
+
+    return metrics
+
+
+def _candles(count):
+    start = datetime(2026, 1, 1, tzinfo=timezone.utc)
+    candles = []
+    close = 100.0
+    for i in range(count):
+        close += ((i % 9) - 4) * 0.23
+        candles.append(
+            Candle(
+                timestamp=start + timedelta(minutes=15 * i),
+                open=close - 0.15,
+                high=close + 1.0 + (i % 4) * 0.17,
+                low=close - 0.8 - (i % 6) * 0.11,
+                close=close,
+                volume=1000 + (i % 23) * 13,
+            )
+        )
+    return candles
+
+
+def test_compute_trend_metrics_matches_reference_full_dm_arrays():
+    candles = _candles(2016)
+    recent = candles[-288:]
+    medium = candles[-2016:]
+    long = candles
+
+    assert MarketAnalyzer._compute_trend_metrics(recent, medium, long) == _reference_trend_metrics(
+        recent,
+        medium,
+        long,
+    )
+
+
+def test_compute_trend_metrics_preserves_short_window_behavior():
+    candles = _candles(14)
+
+    assert MarketAnalyzer._compute_trend_metrics(candles, candles, candles) == _reference_trend_metrics(
+        candles,
+        candles,
+        candles,
+    )