CSV longitude and latitude columns swapped, points land off Somalia

• Header name,lat,lng with first row Beijing,116.397428,39.909230; reading [row.lat, row.lng] makes every point disappear without an error.
• Coordinates display as 39.9, 116.4 and the marker lands in the Indian Ocean off Somalia.
• In other files the latitude 116 is clamped to 90 and all points pile up at the North Pole.

1. Print the raw first line and confirm the header order matches the data order.
2. Check whether the code reads by column name or by index; index-based access must become name-based.
3. Compute the maximum absolute value of the latitude column before loading; above 90 means swapped or bad data.
4. Plot the first point alone and see whether it lands in Beijing or the Indian Ocean.
5. Remember the order differs: GeoJSON is [lng, lat], L.latLng is (lat, lng).

Parse with header: true and read row.lng and row.lat by name, then validate ranges before constructing points; if the pair is already swapped, exchange it once and keep one convention through the whole pipeline.

const pts = rows.map(r => [r[1], r[2]]);   // 实际拿到的是 [116.397428, 39.909230]
// -> Leaflet 认为纬度 116.397428，超限，点被丢弃

const { data } = Papa.parse(text, { header: true, skipEmptyLines: true });

const points = data.map((r, i) => {
  const lat = Number(r.lat ?? r.latitude);
  const lng = Number(r.lng ?? r.lon ?? r.longitude);
  if (!Number.isFinite(lat) || Math.abs(lat) > 90)
    throw new Error(`第 ${i + 2} 行纬度非法：${r.lat}`);
  if (!Number.isFinite(lng) || Math.abs(lng) > 180)
    throw new Error(`第 ${i + 2} 行经度非法：${r.lng}`);
  return { name: r.name, lat, lng };
});

L.geoJSON(points.map(p => ({
  type: 'Feature',
  properties: { name: p.name },
  geometry: { type: 'Point', coordinates: [p.lng, p.lat] }   // GeoJSON 永远是经度在前
})), {
  pointToLayer: (f, ll) => L.circleMarker(ll, { radius: 4 })
}).addTo(map);

• Swapping the values inside the Leaflet constructor only, which then breaks the GeoJSON export path that expects [lng, lat].
• Filtering out rows with latitude above 90 as 'bad data' when they are simply swapped longitudes.