second pass. low

bincio/edit/server.py

@@ -554,10 +554,15 @@ async def upload_activity(file: UploadFile = File(...)) -> JSONResponse:
     if suffix not in _SUPPORTED_SUFFIXES:
         raise HTTPException(400, f"Unsupported file type '{Path(name).suffix}'. Expected FIT, GPX, or TCX.")
 
+    _MAX_UPLOAD_BYTES = 50 * 1024 * 1024  # 50 MB
+    contents = await file.read()
+    if len(contents) > _MAX_UPLOAD_BYTES:
+        raise HTTPException(413, f"File too large ({len(contents)} bytes). Maximum is 50 MB.")
+
     staging = dd / "_uploads"
     staging.mkdir(exist_ok=True)
     staged = staging / name
-    staged.write_bytes(await file.read())
+    staged.write_bytes(contents)
 
     try:
         from bincio.extract.metrics import compute
@@ -592,7 +597,7 @@ async def upload_activity(file: UploadFile = File(...)) -> JSONResponse:
     except HTTPException:
         raise
     except Exception as exc:
-        raise HTTPException(422, str(exc))
+        raise HTTPException(422, f"Failed to process activity file: {type(exc).__name__}")
     finally:
         staged.unlink(missing_ok=True)
 

bincio/extract/cli.py

@@ -46,6 +46,9 @@ def _process_file(path: Path) -> dict:
     """Runs inside a worker process. Only receives a Path (tiny pickle).
     All heavy shared data (_known_hashes, _strava_lookup, etc.) is already
     in the worker's memory from the initializer — zero per-task overhead.
+
+    Writes to pending files (not final paths) so the main process can
+    arbitrate collisions and pick the best version.
     """
     from bincio.extract.metrics import compute
     from bincio.extract.parsers.factory import parse_file
@@ -80,11 +83,17 @@ def _process_file(path: Path) -> dict:
             activity, metrics, _output_dir,
             privacy=_privacy,
             rdp_epsilon=_rdp_epsilon,
+            pending=True,
         )
         summary = build_summary(activity, metrics, activity_id, _privacy)
     except Exception as exc:
         return {"status": "error", "path": str(path), "error": str(exc)}
 
+    # Quality signals for the main process to compare competing results
+    sensor_channels = sum(1 for v in [
+        metrics.avg_hr_bpm, metrics.avg_power_w, metrics.avg_cadence_rpm,
+    ] if v is not None)
+
     return {
         "status": "ok",
         "summary": summary,
@@ -94,6 +103,8 @@ def _process_file(path: Path) -> dict:
         "distance_m": metrics.distance_m,
         "source": summary.get("source"),
         "mmp": metrics.mmp,
+        "point_count": len(activity.points),
+        "sensor_channels": sensor_channels,
     }
 
 
@@ -177,6 +188,8 @@ def extract(
     summaries: list[dict] = []
     errors: list[tuple[str, str]] = []
     skipped = 0
+    # Collect all pending results, grouped by activity_id for collision arbitration
+    pending_by_id: dict[str, list[dict]] = {}
 
     with Progress(
         TextColumn("[progress.description]{task.description}"),
@@ -202,30 +215,61 @@ def extract(
                 elif result["status"] == "error":
                     errors.append((result["path"], result["error"]))
                 else:
-                    # Near-duplicate check — must be sequential (stateful)
+                    pending_by_id.setdefault(result["id"], []).append(result)
+
+    # ── Arbitrate collisions and finalize pending files ───────────────────────
+    from bincio.extract.writer import (
+        activity_quality, cleanup_pending, finalize_pending, write_athlete_json, write_index,
+    )
+
+    for activity_id, candidates in pending_by_id.items():
+        # Pick the best candidate by quality score
+        candidates.sort(key=activity_quality, reverse=True)
+        winner = candidates[0]
+
+        # Clean up losing candidates' pending files
+        for loser in candidates[1:]:
+            cleanup_pending(cfg.output_dir, activity_id, loser["hash"])
+            skipped += 1
+
+        # Near-duplicate check against already-known activities
         from datetime import datetime
-                    started_at = datetime.fromisoformat(result["started_at"])
+        started_at = datetime.fromisoformat(winner["started_at"])
-                    near_id = dedup.find_near_duplicate(started_at, result["distance_m"])
+        near_id = dedup.find_near_duplicate(started_at, winner["distance_m"])
 
         if near_id:
-                        canonical = dedup.pick_canonical(near_id, result.get("source"))
+            canonical = dedup.pick_canonical(near_id, winner.get("source"))
             if canonical != "__new__":
-                            _patch_duplicate_of(cfg.output_dir, result["id"], near_id)
+                # Existing is better — finalize winner as duplicate, then patch it
+                final_id = finalize_pending(cfg.output_dir, activity_id, winner["hash"])
+                _patch_duplicate_of(cfg.output_dir, final_id, near_id)
                 skipped += 1
                 continue
-                        _patch_duplicate_of(cfg.output_dir, near_id, result["id"])
+            # New is better — patch the existing one as duplicate
-                        dedup._records[near_id].duplicate_of = result["id"]
+            final_id = finalize_pending(cfg.output_dir, activity_id, winner["hash"])
+            _patch_duplicate_of(cfg.output_dir, near_id, final_id)
+            dedup._records[near_id].duplicate_of = final_id
+        else:
+            final_id = finalize_pending(cfg.output_dir, activity_id, winner["hash"])
+
+        # Update summary with the finalized ID (may include hash suffix)
+        summary = winner["summary"]
+        if final_id != activity_id:
+            summary = dict(summary)
+            summary["id"] = final_id
+            summary["detail_url"] = f"activities/{final_id}.json"
+            if summary.get("track_url"):
+                summary["track_url"] = f"activities/{final_id}.geojson"
 
         dedup.register(ActivityRecord(
-                        id=result["id"],
+            id=final_id,
-                        source_hash=result["hash"],
+            source_hash=winner["hash"],
             started_at=started_at,
-                        distance_m=result["distance_m"],
+            distance_m=winner["distance_m"],
-                        source=result.get("source"),
+            source=winner.get("source"),
         ))
-                    summaries.append(result["summary"])
+        summaries.append(summary)
 
-    from bincio.extract.writer import write_athlete_json, write_index
     existing = _load_existing_summaries(cfg.output_dir)
     merged = {s["id"]: s for s in existing}
     for s in summaries:

bincio/extract/parsers/gpx.py

@@ -76,6 +76,8 @@ def _apply_extensions(pt: gpxpy.gpx.GPXTrackPoint, dp: DataPoint) -> None:
                     dp.temperature_c = float(val)
                 elif tag == "speed":
                     dp.speed_kmh = float(val) * 3.6  # m/s → km/h
+                elif tag in ("pwr", "power", "watts"):
+                    dp.power_w = int(float(val))
 
 
 def _strip_ns(tag: str) -> str:

bincio/extract/parsers/tcx.py

@@ -97,8 +97,8 @@ def _parse_ts(s: str) -> datetime:
                 return datetime.strptime(s, fmt).replace(tzinfo=timezone.utc)
             except ValueError:
                 continue
-    # Numeric offset like +02:00 or -05:30 — parse with %z then convert to UTC
+    # Numeric offset like +02:00, -05:30, or +0200 — parse with %z then convert to UTC
-    m = _re.match(r"^(.+)([+-]\d{2}:\d{2})$", s)
+    m = _re.match(r"^(.+)([+-]\d{2}:?\d{2})$", s)
     if m:
         body, off = m.group(1), m.group(2).replace(":", "")
         for fmt in ("%Y-%m-%dT%H:%M:%S.%f", "%Y-%m-%dT%H:%M:%S"):

bincio/extract/simplify.py

@@ -67,8 +67,12 @@ def build_geojson(
         if p.lon is not None and p.lat is not None
     ]
 
-    # Parallel speed array for gradient coloring
+    # Parallel speed array for gradient coloring — same filter as coordinates
-    speeds = [round(p.speed_kmh, 2) if p.speed_kmh is not None else None for p in simplified]
+    speeds = [
+        round(p.speed_kmh, 2) if p.speed_kmh is not None else None
+        for p in simplified
+        if p.lon is not None and p.lat is not None
+    ]
 
     return {
         "type": "Feature",

bincio/extract/writer.py

@@ -33,8 +33,16 @@ def write_activity(
     privacy: str = "public",
     duplicate_of: str | None = None,
     rdp_epsilon: float = 0.0001,
+    pending: bool = False,
 ) -> str:
-    """Write {id}.json and (if GPS available) {id}.geojson. Returns the ID."""
+    """Write {id}.json and (if GPS available) {id}.geojson. Returns the ID.
+
+    When pending=True, writes to a uniquely-named pending file
+    ({id}.{hash[:8]}.pending.json) instead of the final path. This avoids
+    race conditions when multiple workers process activities with the same ID.
+    The main process is responsible for promoting pending files to final paths
+    via finalize_pending().
+    """
     activity_id = make_activity_id(activity)
     acts_dir = output_dir / "activities"
     acts_dir.mkdir(parents=True, exist_ok=True)
@@ -82,26 +90,108 @@ def write_activity(
         "custom": {},
     }
 
+    if pending:
+        # Write to a unique pending file — no collision possible
+        tag = activity.source_hash[-8:] if activity.source_hash else "unknown"
+        json_path = acts_dir / f"{activity_id}.{tag}.pending.json"
+    else:
         json_path = acts_dir / f"{activity_id}.json"
-    # Collision guard: if a *different* activity already has this ID, append a
+        # Legacy non-pending path: collision guard for callers that don't use
-    # short hash suffix to disambiguate (same hash = idempotent re-extract).
+        # the pending workflow (e.g. edit server upload_activity)
         if json_path.exists():
             existing = json.loads(json_path.read_text(encoding="utf-8"))
             if existing.get("source_hash") != activity.source_hash:
                 activity_id = f"{activity_id}-{activity.source_hash[-6:]}"
                 json_path = acts_dir / f"{activity_id}.json"
                 detail["id"] = activity_id
+
     json_path.write_text(json.dumps(detail, indent=2, ensure_ascii=False))
 
     # ── GeoJSON track ────────────────────────────────────────────────────────
     if has_gps:
         geojson = build_geojson(activity.points, activity_id, epsilon=rdp_epsilon)
+        if pending:
+            geojson_path = acts_dir / f"{activity_id}.{tag}.pending.geojson"
+        else:
             geojson_path = acts_dir / f"{activity_id}.geojson"
         geojson_path.write_text(json.dumps(geojson, indent=2, ensure_ascii=False))
 
     return activity_id
 
 
+def activity_quality(result: dict) -> int:
+    """Compute a quality score for an activity result from a worker.
+
+    Higher is better. Used by the main process to pick the best version
+    when multiple workers produce results for the same activity ID.
+    """
+    from bincio.extract.dedup import _SOURCE_QUALITY
+
+    score = 0
+    # Source type quality (FIT > GPX > TCX)
+    score += _SOURCE_QUALITY.get(result.get("source") or "", 0) * 100
+    # Sensor channel count
+    score += result.get("sensor_channels", 0) * 10
+    # Point count (more data = better)
+    score += min(result.get("point_count", 0), 50000) // 100
+    return score
+
+
+def finalize_pending(output_dir: Path, activity_id: str, source_hash: str) -> str:
+    """Promote a pending file to its final path via atomic rename.
+
+    If another activity already occupies the ID (different source_hash),
+    the pending file is disambiguated with a hash suffix.
+
+    Returns the final activity_id (may include suffix).
+    """
+    acts_dir = output_dir / "activities"
+    tag = source_hash[-8:] if source_hash else "unknown"
+
+    pending_json = acts_dir / f"{activity_id}.{tag}.pending.json"
+    pending_geojson = acts_dir / f"{activity_id}.{tag}.pending.geojson"
+
+    final_id = activity_id
+    final_json = acts_dir / f"{final_id}.json"
+
+    # Check for ID collision with a different activity
+    if final_json.exists():
+        existing = json.loads(final_json.read_text(encoding="utf-8"))
+        if existing.get("source_hash") != source_hash:
+            final_id = f"{activity_id}-{source_hash[-6:]}"
+            final_json = acts_dir / f"{final_id}.json"
+
+    # Update the ID inside the JSON if it changed
+    if final_id != activity_id and pending_json.exists():
+        detail = json.loads(pending_json.read_text(encoding="utf-8"))
+        detail["id"] = final_id
+        pending_json.write_text(json.dumps(detail, indent=2, ensure_ascii=False))
+
+    # Atomic rename: pending → final
+    if pending_json.exists():
+        pending_json.rename(final_json)
+
+    final_geojson = acts_dir / f"{final_id}.geojson"
+    if pending_geojson.exists():
+        # Update the ID in GeoJSON properties too
+        if final_id != activity_id:
+            geo = json.loads(pending_geojson.read_text(encoding="utf-8"))
+            geo["properties"]["id"] = final_id
+            pending_geojson.write_text(json.dumps(geo, indent=2, ensure_ascii=False))
+        pending_geojson.rename(final_geojson)
+
+    return final_id
+
+
+def cleanup_pending(output_dir: Path, activity_id: str, source_hash: str) -> None:
+    """Remove pending files for a losing activity (the one not chosen as canonical)."""
+    acts_dir = output_dir / "activities"
+    tag = source_hash[-8:] if source_hash else "unknown"
+    for suffix in (".pending.json", ".pending.geojson"):
+        p = acts_dir / f"{activity_id}.{tag}{suffix}"
+        p.unlink(missing_ok=True)
+
+
 def build_summary(
     activity: ParsedActivity,
     metrics: ComputedMetrics,

site/src/components/ActivityCharts.svelte

@@ -1,6 +1,6 @@
 <script lang="ts">
   import * as Plot from '@observablehq/plot';
-  import { onMount } from 'svelte';
+  import { onMount, onDestroy } from 'svelte';
   import type { Timeseries, AthleteZones } from '../lib/types';
 
   export let timeseries: Timeseries;
@@ -82,8 +82,15 @@
   // Range handles — reset whenever the metric or chart type changes
   let trimMin = 0;
   let trimMax = 100;
-  $: if (dataMin !== undefined) resetTrim(dataMin, dataMax);
+  let lastResetTab: Tab | null = null;
-  function resetTrim(lo: number, hi: number) { trimMin = lo; trimMax = hi; }
+  $: {
+    // Reset trim on tab change OR when data range changes
+    if (activeTab !== lastResetTab || trimMin < dataMin || trimMax > dataMax) {
+      trimMin = dataMin;
+      trimMax = dataMax;
+      lastResetTab = activeTab;
+    }
+  }
 
   $: step = (dataMax - dataMin) / 200 || 1;
 
@@ -116,6 +123,7 @@
 
   // ── Rendering ────────────────────────────────────────────────────────────
   onMount(() => { renderChart(); });
+  onDestroy(() => { chart?.remove(); chart = null; });
 
   $: if (chartEl) {
     activeTab; xMode; chartType; histData; histThresholds; alignZones;

site/src/components/AthleteView.svelte

@@ -27,7 +27,9 @@
   }
 
   onMount(async () => {
-    activeTab = (new URLSearchParams(window.location.search).get('tab') as Tab) ?? 'power';
+    const TABS: Tab[] = ['power', 'records', 'profile'];
+    const rawTab = new URLSearchParams(window.location.search).get('tab');
+    activeTab = TABS.includes(rawTab as Tab) ? (rawTab as Tab) : 'power';
     mounted = true;
     try {
       const [athleteRes, indexRes] = await Promise.all([

tests/test_merge.py

@@ -55,7 +55,7 @@ def test_parse_sidecar_frontmatter_only(tmp_path):
 # ── apply_sidecar ─────────────────────────────────────────────────────────────
 
 BASE_DETAIL = {
-    "id": "2024-01-01T08:00:00Z_cycling",
+    "id": "2024-01-01T080000Z-morning-ride",
     "title": "Morning Ride",
     "sport": "cycling",
     "started_at": "2024-01-01T08:00:00Z",
@@ -118,21 +118,21 @@ def data_dir(tmp_path):
     acts = tmp_path / "activities"
     acts.mkdir()
     # Two activities
-    for act_id, title in [
+    for act_id, title, sport, started_at in [
-        ("2024-01-01T08:00:00Z_cycling", "Morning Ride"),
+        ("2024-01-01T080000Z-morning-ride", "Morning Ride", "cycling", "2024-01-01T08:00:00Z"),
-        ("2024-01-02T09:00:00Z_running", "Easy Run"),
+        ("2024-01-02T090000Z-easy-run", "Easy Run", "running", "2024-01-02T09:00:00Z"),
     ]:
         detail = {
-            "id": act_id, "title": title, "sport": act_id.split("_")[1],
+            "id": act_id, "title": title, "sport": sport,
-            "started_at": act_id.split("_")[0],
+            "started_at": started_at,
             "description": "", "privacy": "public", "custom": {},
         }
         (acts / f"{act_id}.json").write_text(json.dumps(detail))
     # Index
     index = {"activities": [
-        {"id": "2024-01-01T08:00:00Z_cycling", "title": "Morning Ride",
+        {"id": "2024-01-01T080000Z-morning-ride", "title": "Morning Ride",
          "sport": "cycling", "started_at": "2024-01-01T08:00:00Z", "privacy": "public", "custom": {}},
-        {"id": "2024-01-02T09:00:00Z_running", "title": "Easy Run",
+        {"id": "2024-01-02T090000Z-easy-run", "title": "Easy Run",
          "sport": "running", "started_at": "2024-01-02T09:00:00Z", "privacy": "public", "custom": {}},
     ]}
     (tmp_path / "index.json").write_text(json.dumps(index))
@@ -145,20 +145,20 @@ def test_merge_all_no_sidecars(data_dir):
     merged = data_dir / "_merged"
     assert merged.exists()
     # Unmodified files are symlinked
-    detail_link = merged / "activities" / "2024-01-01T08:00:00Z_cycling.json"
+    detail_link = merged / "activities" / "2024-01-01T080000Z-morning-ride.json"
     assert detail_link.is_symlink()
 
 
 def test_merge_all_applies_sidecar(data_dir):
     edits = data_dir / "edits"
     edits.mkdir()
-    (edits / "2024-01-01T08:00:00Z_cycling.md").write_text(
+    (edits / "2024-01-01T080000Z-morning-ride.md").write_text(
         "---\ntitle: Epic Ride\nhighlight: true\n---\n\nWhat a day!"
     )
     n = merge_all(data_dir)
     assert n == 1
 
-    merged_json = data_dir / "_merged" / "activities" / "2024-01-01T08:00:00Z_cycling.json"
+    merged_json = data_dir / "_merged" / "activities" / "2024-01-01T080000Z-morning-ride.json"
     assert not merged_json.is_symlink()
     data = json.loads(merged_json.read_text())
     assert data["title"] == "Epic Ride"
@@ -166,41 +166,41 @@ def test_merge_all_applies_sidecar(data_dir):
     assert data["description"] == "What a day!"
 
     # Untouched activity is still a symlink
-    run_link = data_dir / "_merged" / "activities" / "2024-01-02T09:00:00Z_running.json"
+    run_link = data_dir / "_merged" / "activities" / "2024-01-02T090000Z-easy-run.json"
     assert run_link.is_symlink()
 
 
 def test_merge_all_private_filtered_from_index(data_dir):
     edits = data_dir / "edits"
     edits.mkdir()
-    (edits / "2024-01-01T08:00:00Z_cycling.md").write_text("---\nprivate: true\n---\n")
+    (edits / "2024-01-01T080000Z-morning-ride.md").write_text("---\nprivate: true\n---\n")
     merge_all(data_dir)
 
     index = json.loads((data_dir / "_merged" / "index.json").read_text())
     ids = [a["id"] for a in index["activities"]]
-    assert "2024-01-01T08:00:00Z_cycling" not in ids
+    assert "2024-01-01T080000Z-morning-ride" not in ids
-    assert "2024-01-02T09:00:00Z_running" in ids
+    assert "2024-01-02T090000Z-easy-run" in ids
 
 
 def test_merge_all_highlight_sorts_first(data_dir):
     edits = data_dir / "edits"
     edits.mkdir()
     # Highlight the older activity — it should appear first
-    (edits / "2024-01-01T08:00:00Z_cycling.md").write_text("---\nhighlight: true\n---\n")
+    (edits / "2024-01-01T080000Z-morning-ride.md").write_text("---\nhighlight: true\n---\n")
     merge_all(data_dir)
 
     index = json.loads((data_dir / "_merged" / "index.json").read_text())
     ids = [a["id"] for a in index["activities"]]
-    assert ids[0] == "2024-01-01T08:00:00Z_cycling"
+    assert ids[0] == "2024-01-01T080000Z-morning-ride"
 
 
 def test_merge_all_idempotent(data_dir):
     edits = data_dir / "edits"
     edits.mkdir()
-    (edits / "2024-01-01T08:00:00Z_cycling.md").write_text("---\ntitle: Renamed\n---\n")
+    (edits / "2024-01-01T080000Z-morning-ride.md").write_text("---\ntitle: Renamed\n---\n")
     merge_all(data_dir)
     merge_all(data_dir)  # second run should not error or double-apply
     data = json.loads(
-        (data_dir / "_merged" / "activities" / "2024-01-01T08:00:00Z_cycling.json").read_text()
+        (data_dir / "_merged" / "activities" / "2024-01-01T080000Z-morning-ride.json").read_text()
     )
     assert data["title"] == "Renamed"

tests/test_sport.py

@@ -11,6 +11,16 @@ def test_running_variants():
         assert normalise_sport(raw) == "running", raw
 
 
+def test_skiing_variants():
+    for raw in ("skiing", "alpine_skiing", "nordic_skiing", "backcountry_ski"):
+        assert normalise_sport(raw) == "skiing", raw
+
+
+def test_swimming_variants():
+    for raw in ("swimming", "swim", "open_water_swimming", "lap_swimming"):
+        assert normalise_sport(raw) == "swimming", raw
+
+
 def test_unknown_falls_back_to_other():
     assert normalise_sport("yoga") == "other"
     assert normalise_sport(None) == "other"

tests/test_writer.py

@@ -1,4 +1,5 @@
-from bincio.extract.writer import make_activity_id, _slugify
+from bincio.extract.writer import make_activity_id, build_summary, _slugify
+from bincio.extract.metrics import ComputedMetrics
 from bincio.extract.models import ParsedActivity, DataPoint
 from datetime import datetime, timezone
 
@@ -31,3 +32,52 @@ def test_slugify():
     assert _slugify("Morning Ride!") == "morning-ride"
     assert _slugify("  Vélo  ") == "velo"   # é → e via NFKD + ASCII
     assert _slugify("") == ""
+
+
+def test_id_utc_conversion():
+    """Non-UTC timestamps should be converted to UTC in the ID."""
+    from datetime import timedelta
+    tz_plus2 = timezone(timedelta(hours=2))
+    ts = datetime(2024, 6, 1, 9, 30, 12, tzinfo=tz_plus2)  # 07:30:12 UTC
+    act = ParsedActivity(
+        points=[DataPoint(timestamp=ts)],
+        sport="cycling",
+        started_at=ts,
+        source_file="test.fit",
+        source_hash="sha256:abc",
+    )
+    assert make_activity_id(act) == "2024-06-01T073012Z"
+
+
+def test_build_summary_required_fields():
+    """build_summary should include all fields needed by the schema."""
+    act = _dummy_activity("Test Ride")
+    metrics = ComputedMetrics(
+        distance_m=10000.0,
+        duration_s=3600,
+        moving_time_s=3500,
+        elevation_gain_m=100.0,
+        elevation_loss_m=95.0,
+        avg_speed_kmh=10.0,
+        max_speed_kmh=20.0,
+        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,
+        best_efforts=None,
+        best_climb_m=None,
+    )
+    summary = build_summary(act, metrics, "2024-06-01T073012Z-test-ride")
+    # Required fields per schema
+    assert summary["id"] == "2024-06-01T073012Z-test-ride"
+    assert summary["title"] == "Test Ride"
+    assert summary["sport"] == "cycling"
+    assert "started_at" in summary
+    assert "privacy" in summary
+    assert "detail_url" in summary
+    assert "track_url" in summary