athlete page first draft

bincio/extract/metrics.py

@@ -6,10 +6,14 @@ Uses inline haversine rather than geopy.geodesic to keep the hot path fast.
 
 import math
 from dataclasses import dataclass
+from datetime import datetime
 from typing import Optional
 
 from bincio.extract.models import DataPoint, ParsedActivity
 
+# Standard MMP durations (seconds). Log-spaced so the curve looks good on a log-x axis.
+MMP_DURATIONS_S = [1, 2, 5, 10, 15, 20, 30, 60, 120, 180, 300, 600, 1200, 1800, 3600]
+
 # Speed below which we consider the athlete stopped (km/h)
 _STOPPED_THRESHOLD_KMH = 1.0
 _EARTH_R = 6_371_000.0  # metres
@@ -42,6 +46,7 @@ class ComputedMetrics:
     bbox: Optional[tuple[float, float, float, float]]   # min_lon, min_lat, max_lon, max_lat
     start_latlng: Optional[tuple[float, float]]
     end_latlng: Optional[tuple[float, float]]
+    mmp: Optional[list[list[int]]]  # [[duration_s, avg_watts], ...] — None if no power data
 
 
 def compute(activity: ParsedActivity) -> ComputedMetrics:
@@ -58,6 +63,7 @@ def compute(activity: ParsedActivity) -> ComputedMetrics:
     max_pow = _max_nonnull([p.power_w for p in pts])
     bbox = _bbox(pts)
     start_ll, end_ll = _endpoints(pts)
+    mmp = compute_mmp(pts, activity.started_at)
 
     return ComputedMetrics(
         distance_m=distance_m,
@@ -75,9 +81,57 @@ def compute(activity: ParsedActivity) -> ComputedMetrics:
         bbox=bbox,
         start_latlng=start_ll,
         end_latlng=end_ll,
+        mmp=mmp,
     )
 
 
+# ── mean maximal power ────────────────────────────────────────────────────────
+
+def compute_mmp(pts: list[DataPoint], started_at: datetime) -> Optional[list[list[int]]]:
+    """Compute Mean Maximal Power curve at the standard MMP_DURATIONS_S.
+
+    Builds a 1 Hz power series (same downsampling as timeseries.py), then uses
+    a O(n) sliding-window sum for each duration.  Returns a list of
+    [duration_s, avg_watts] pairs (integers), or None when the activity has no
+    power data.  Only durations shorter than the total activity are included.
+    """
+    # 1 Hz downsample: at most one sample per second, skip sub-second duplicates.
+    # Seconds without a recorded sample are omitted (not zero-filled) so that
+    # paused-recording gaps don't silently lower power averages.
+    power_1hz: list[int] = []
+    last_t = -1
+    for p in pts:
+        t = int((p.timestamp - started_at).total_seconds())
+        if t < 0 or t == last_t:
+            continue
+        last_t = t
+        if p.power_w is not None:
+            power_1hz.append(p.power_w)
+
+    if len(power_1hz) < 2:
+        return None
+
+    n = len(power_1hz)
+    results: list[list[int]] = []
+
+    for d in MMP_DURATIONS_S:
+        if d > n:
+            break  # activity shorter than this duration — stop (durations are sorted)
+
+        # Sliding window of exactly d samples = d seconds at 1 Hz.
+        window_sum = sum(power_1hz[:d])
+        best = window_sum
+
+        for i in range(1, n - d + 1):
+            window_sum += power_1hz[i + d - 1] - power_1hz[i - 1]
+            if window_sum > best:
+                best = window_sum
+
+        results.append([d, round(best / d)])
+
+    return results if results else None
+
+
 # ── single-pass GPS stats ──────────────────────────────────────────────────────
 # distance, moving time, avg speed, and max speed are all derived from the same
 # per-segment loop, so we compute them in one pass instead of four.
@@ -209,4 +263,5 @@ def _empty() -> ComputedMetrics:
         avg_hr_bpm=None, max_hr_bpm=None,
         avg_cadence_rpm=None, avg_power_w=None, max_power_w=None,
         bbox=None, start_latlng=None, end_latlng=None,
+        mmp=None,
     )