nakas's picture
Normalize longitudes to -180..180 and remove strict bounds so velocity particles render globally
9ed3c47
Raw
History Blame Contribute Delete
16.4 kB
import os
import json
from datetime import datetime
import math
from functools import lru_cache
from typing import Dict, Any, List
import numpy as np
from fastapi import FastAPI, Query
from fastapi.responses import JSONResponse, HTMLResponse
import gradio as gr
# Local import: vendored from working project
from backend.grib_wave_puller import GRIBWavePuller
app = FastAPI(title="Wave Visualizer API")
def _compute_uv_from_wave(height: np.ndarray, direction_deg: np.ndarray, scale: float = 0.1):
"""Compute U/V components from wave height and meteorological 'from' direction.
- height: significant wave height array (m)
- direction_deg: wave direction (deg, meteorological, coming from)
- scale: visualization scaling factor
"""
dir_rad = np.deg2rad(direction_deg)
mag = np.clip(height, 0, np.nanmax(height)) * scale
# Eastward (u) and northward (v) components; negative on v because 'from'
u = mag * np.sin(dir_rad)
v = -mag * np.cos(dir_rad)
return u, v
def _build_velocity_grib_json(lats: np.ndarray, lons: np.ndarray, u: np.ndarray, v: np.ndarray, ref_time: str) -> List[Dict[str, Any]]:
"""Build leaflet-velocity compatible JSON (Wind/Earth GRIB-like format).
Data must be provided on a regular lat-lon grid. Arrays are 2D with shape (ny, nx)
where ny=len(lats), nx=len(lons). Latitude should be provided in descending order
(north to south) to match common GRIB conventions; reorder if needed.
"""
# Ensure 1D coordinate arrays
lats_1d = lats if lats.ndim == 1 else lats[:, 0]
lons_1d = lons if lons.ndim == 1 else lons[0, :]
ny = int(len(lats_1d))
nx = int(len(lons_1d))
# If latitude increases northward, reverse to north->south
if ny > 1 and lats_1d[0] < lats_1d[-1]:
lats_1d = lats_1d[::-1]
u = np.flipud(u)
v = np.flipud(v)
# Normalize longitudes to [-180, 180) to avoid 0..360 grids causing clipping
# Then ensure they ascend west->east and reorder u/v columns accordingly.
if nx > 1:
lons_wrapped = ((np.asarray(lons_1d, dtype=float) + 180.0) % 360.0) - 180.0
order = np.argsort(lons_wrapped)
lons_1d = lons_wrapped[order]
if u.ndim == 2 and v.ndim == 2 and u.shape[1] == nx and v.shape[1] == nx:
u = u[:, order]
v = v[:, order]
la1 = float(lats_1d[0])
la2 = float(lats_1d[-1])
lo1 = float(lons_1d[0])
lo2 = float(lons_1d[-1])
# Grid spacing (approx)
dy = float(abs(lats_1d[1] - lats_1d[0])) if ny > 1 else 0.0
dx = float(abs(lons_1d[1] - lons_1d[0])) if nx > 1 else 0.0
# Sanitize arrays: replace NaN/Inf with zeros for JSON compliance
u = np.nan_to_num(np.asarray(u, dtype=float), nan=0.0, posinf=0.0, neginf=0.0)
v = np.nan_to_num(np.asarray(v, dtype=float), nan=0.0, posinf=0.0, neginf=0.0)
# Optional clamp to reasonable range (avoid absurd values)
# Here, clamp to [-20, 20] m/s just for safety in visualization
u = np.clip(u, -20.0, 20.0)
v = np.clip(v, -20.0, 20.0)
# Flatten row-major (lat-major first, then lon) matching header
u_data = u.flatten().tolist()
v_data = v.flatten().tolist()
header_common = {
"lo1": lo1,
"la1": la1,
"lo2": lo2,
"la2": la2,
"nx": nx,
"ny": ny,
"dx": dx,
"dy": dy,
"refTime": ref_time,
}
u_record = {
"header": {
**header_common,
"parameterCategory": 2,
"parameterNumber": 2, # U component
"parameterUnit": "m/s",
},
"data": u_data,
}
v_record = {
"header": {
**header_common,
"parameterCategory": 2,
"parameterNumber": 3, # V component
"parameterUnit": "m/s",
},
"data": v_data,
}
return [u_record, v_record]
@lru_cache(maxsize=16)
def get_puller() -> GRIBWavePuller:
return GRIBWavePuller()
@app.get("/data/points")
def data_points(hour: int = Query(0, ge=0, le=240)):
puller = get_puller()
result = puller.fetch_global_wave_data(hour)
if not result:
return JSONResponse(status_code=503, content={"error": "No data available"})
def _sanitize(obj):
if isinstance(obj, dict):
return {k: _sanitize(v) for k, v in obj.items()}
if isinstance(obj, list):
return [_sanitize(v) for v in obj]
if isinstance(obj, (np.floating,)):
v = float(obj)
return None if not math.isfinite(v) else v
if isinstance(obj, (np.integer,)):
return int(obj)
if isinstance(obj, float):
return None if not math.isfinite(obj) else obj
return obj
payload = {
"type": "points",
"refTime": result.get("timestamp"),
"points": result.get("sample_points", []),
}
return JSONResponse(content=_sanitize(payload))
@app.get("/data/velocity")
def data_velocity(hour: int = Query(0, ge=0, le=240), scale: float = Query(0.1)):
puller = get_puller()
result = puller.fetch_global_wave_data(hour)
if not result:
return JSONResponse(status_code=503, content={"error": "No data available"})
# If we have a downsampled UV grid, return leaflet-velocity JSON
grid_uv = result.get("grid_uv")
if grid_uv:
lats = np.array(grid_uv['lats'])
lons = np.array(grid_uv['lons'])
u = np.array(grid_uv['u'])
v = np.array(grid_uv['v'])
# Validate grid content; if empty or trivial, fall back to points
if (
u.size < 16 or v.size < 16 or
not np.isfinite(u).any() or not np.isfinite(v).any() or
(np.nanmax(np.abs(u)) < 1e-6 and np.nanmax(np.abs(v)) < 1e-6)
):
sample_points = result.get("sample_points", [])
return JSONResponse(content={"type": "points", "refTime": result.get("timestamp"), "points": sample_points})
payload = _build_velocity_grib_json(lats, lons, u, v, ref_time=result.get("timestamp", datetime.utcnow().isoformat()))
return JSONResponse(content=payload)
# Fallback to points if no grid is present
sample_points = result.get("sample_points", [])
payload = {"type": "points", "refTime": result.get("timestamp"), "points": sample_points}
# Sanitize for JSON compliance
def _san(obj):
if isinstance(obj, dict):
return {k: _san(v) for k, v in obj.items()}
if isinstance(obj, list):
return [_san(v) for v in obj]
if isinstance(obj, (np.floating,)):
v = float(obj)
return None if not math.isfinite(v) else v
if isinstance(obj, (np.integer,)):
return int(obj)
if isinstance(obj, float):
return None if not math.isfinite(obj) else obj
return obj
return JSONResponse(content=_san(payload))
def leaflet_html() -> str:
return """
<!doctype html>
<html>
<head>
<meta charset=\"utf-8\" />
<meta name=\"viewport\" content=\"width=device-width, initial-scale=1.0\" />
<link rel=\"stylesheet\" href=\"https://unpkg.com/leaflet@1.9.4/dist/leaflet.css\" />
<style>
html, body, #map { height: 100%; margin: 0; }
.leaflet-control-container .leaflet-top.leaflet-left { z-index: 1000; }
.control { position:absolute; top:10px; left:10px; z-index:1000; background:#fff; padding:8px; border-radius:4px; box-shadow:0 1px 3px rgba(0,0,0,0.3); pointer-events:auto; }
/* Animated flow for arrow polylines */
@keyframes arrow-dash {
0% { stroke-dashoffset: 0; }
100% { stroke-dashoffset: -20; }
}
/* Leaflet renders polylines as SVG paths */
.leaflet-overlay-pane path.arrow-line {
vector-effect: non-scaling-stroke;
fill: none;
stroke-linecap: butt;
stroke-dasharray: 6 10;
animation: arrow-dash 1.2s linear infinite;
}
</style>
</head>
<body>
<div id=\"map\"></div>
<div class=\"control\">
<label>Forecast hour: <input type=\"number\" id=\"hour\" min=\"0\" max=\"240\" step=\"6\" value=\"0\" /></label>
<button id=\"load\">Load Waves</button>
<span id=\"status\" style=\"margin-left:8px; font-size:12px; color:#333\"></span>
</div>
<script src=\"https://unpkg.com/leaflet@1.9.4/dist/leaflet.js\"></script>
<!-- Use a known-good Leaflet-Velocity build -->
<script src=\"https://cdn.jsdelivr.net/npm/leaflet-velocity@1.8.0/dist/leaflet-velocity.min.js\"></script>
<script>
const map = L.map('map').setView([20, 0], 2);
L.tileLayer('https://{s}.tile.openstreetmap.org/{z}/{x}/{y}.png', {
maxZoom: 6,
attribution: '&copy; OpenStreetMap contributors'
}).addTo(map);
let velocityLayer = null;
let pointLayer = null;
// Create a short dash polyline with attached metadata (wind-like particle)
function mkArrow(p, opts = {}) {
const dir = p.wave_direction ?? 0;
const height = p.wave_height ?? 0.5;
const period = p.wave_period ?? null;
// Keep segments short to resemble particles rather than long arrows
const len = Math.max(15000, Math.min(60000, height * 35000)); // meters
const rad = dir * Math.PI/180.0;
const dx = Math.sin(rad) * len;
const dy = -Math.cos(rad) * len;
const poly = L.polyline(
[[p.lat, p.lon], [p.lat + dy/1e6, p.lon + dx/1e6]],
{ color: '#e5242a', weight: 1.0, opacity: 0.9, className: 'arrow-line', ...opts }
);
// Attach metadata so we can tune styles after added to map
poly._waveMeta = { height, period, direction: dir };
return poly;
}
// Tune per-feature animation speed/appearance using period and height
function tuneArrowStyles(group) {
if (!group) return;
const tune = (layer) => {
const el = (layer.getElement && layer.getElement()) || layer._path;
if (!el || !layer._waveMeta) return;
const { height, period } = layer._waveMeta;
// Stroke weight by height (clamped)
const w = Math.max(0.6, Math.min(1.6, 0.8 + (height || 0) * 0.25));
layer.setStyle && layer.setStyle({ weight: w });
// Particle-like short dashes
const dashLen = Math.max(2, Math.min(10, 3 + (height || 0) * 1.2));
const gapLen = Math.round(dashLen * 1.4);
el.style.strokeDasharray = `${dashLen} ${gapLen}`;
// Animation speed by period: longer period => faster flow (shorter duration)
let dur;
if (period && isFinite(period)) {
// Map 2s..20s -> 1.2s..0.5s duration for livelier particles
const p = Math.max(2, Math.min(20, period));
dur = 1.2 - (p - 2) * ((1.2 - 0.5) / (20 - 2));
} else {
dur = 0.9; // default
}
el.style.animationDuration = `${dur.toFixed(2)}s`;
};
// Defer slightly to ensure SVG paths exist
setTimeout(() => {
group.eachLayer(tune);
}, 0);
}
async function load(hour) {
if (velocityLayer) { map.removeLayer(velocityLayer); velocityLayer = null; }
if (pointLayer) { map.removeLayer(pointLayer); pointLayer = null; }
const status = document.getElementById('status');
status.textContent = 'Loading...';
const res = await fetch(`/data/velocity?hour=${hour}`);
if (!res.ok) { alert('Failed to fetch data'); return; }
const payload = await res.json();
console.log('velocity payload', payload);
if (payload && payload.type === 'points') {
// Fallback: draw particle-like markers with direction
const features = payload.points.map(p => mkArrow(p));
pointLayer = L.layerGroup(features).addTo(map);
tuneArrowStyles(pointLayer);
status.textContent = `Rendered ${features.length} wave arrows`;
} else {
try {
// Expected: array of two GRIB-like records (u and v)
if (Array.isArray(payload) && payload.length >= 2 && payload[0].data && payload[0].data.length) {
// Quick sanity check: some non-zero magnitudes
const sample = payload[0].data.slice(0, 200);
const nz = sample.reduce((acc, v) => acc + Math.abs(v), 0);
if (nz < 1e-3) {
throw new Error('Velocity grid near-zero; fallback to points');
}
velocityLayer = L.velocityLayer({
data: payload,
displayValues: true,
displayOptions: {
velocityType: 'Wave',
position: 'bottomleft',
emptyString: 'No wave data',
speedUnit: 'm/s',
angleConvention: 'bearingCW',
showCardinal: true
},
// Settings aligned with the working wind demo
velocityScale: 0.01,
opacity: 0.9,
maxVelocity: 20,
particleMultiplier: 0.002,
lineWidth: 1.2,
frameRate: 15,
particleAge: 40,
fadeOpacity: 0,
animationDuration: 0,
// Remove strict bounds/wrap to support 0..360 or -180..180 grids
// Red gradient color scale
colorScale: [
"#4c0000", "#660000", "#800000", "#990000", "#b30000",
"#cc0000", "#e60000", "#ff0000", "#ff3333", "#ff6666", "#ff9999"
],
});
velocityLayer.addTo(map);
status.textContent = 'Velocity layer active';
// Also overlay a sparse set of arrows for immediate visual feedback
try {
const resPts = await fetch(`/data/points?hour=${hour}`);
if (resPts.ok) {
const pld = await resPts.json();
const pts = (pld.points || []).slice(0, 300);
const arrs = pts.map(p => mkArrow(p, { color: '#e5242a', opacity: 0.85 }));
pointLayer = L.layerGroup(arrs).addTo(map);
tuneArrowStyles(pointLayer);
}
} catch (e2) { console.warn('arrow overlay failed', e2); }
} else {
// Final fallback: fetch points explicitly
const res2 = await fetch(`/data/points?hour=${hour}`);
if (res2.ok) {
const payload2 = await res2.json();
console.log('points payload', payload2);
const features = (payload2.points || []).map(p => mkArrow(p));
if (features.length) {
pointLayer = L.layerGroup(features).addTo(map);
tuneArrowStyles(pointLayer);
status.textContent = `Rendered ${features.length} wave arrows`;
} else {
status.textContent = 'No wave data available';
}
} else {
status.textContent = 'Failed to fetch data';
}
}
} catch (e) {
console.warn('Velocity layer failed, falling back to points:', e);
const res2 = await fetch(`/data/points?hour=${hour}`);
const payload2 = await res2.json();
const features = payload2.points.map(p => mkArrow(p));
pointLayer = L.layerGroup(features).addTo(map);
tuneArrowStyles(pointLayer);
status.textContent = `Rendered ${features.length} wave arrows`;
}
}
}
document.getElementById('load').onclick = () => {
const h = parseInt(document.getElementById('hour').value || '0', 10);
load(h);
};
</script>
</body>
</html>
"""
@app.get("/map", response_class=HTMLResponse)
def map_page():
return leaflet_html()
@app.get("/", response_class=HTMLResponse)
def root_page():
return leaflet_html()
# Optional Gradio UI under /ui
with gr.Blocks(title="Wave Visualizer UI") as demo:
gr.Markdown("# Wave Visualizer\nUse the link below to open the map page.")
gr.HTML('<p><a href="/map" target="_blank">Open Map</a></p>')
from gradio.routes import mount_gradio_app
app = mount_gradio_app(app, demo, path="/ui")