python encoding tryout

[pvgenuchten]

hi i tried the python based encoder today on a CSV file

label,ID,Lat,Lon,pH ,Tot_C,Tot_N,P,K,Ca,Mg,Na,Mn,Cu,Fe,Zn,EC
example1,3,-0.78393,35.08891,5.7,1.84,0.18,10.9,0.44,12.13,2.1,0.45,1.3,0.28,120.4,8.4,
example2,4,-0.77556,35.10607,5.3,1.54,0.15,9.5,0.22,5.21,0.5,0.31,0.8,0.21,187,4.2,
example3,5,-0.79172,35.0989,5.4,1.82,0.18,6.3,0.68,8.06,1.6,0.21,0.8,0.35,133.7,4.7,
example4,6,-0.80082,35.1017,5.5,2.49,0.25,10,0.8,11.92,1.4,0.29,0.8,0.53,258.7,5.5,

using this script:

import csv
from pathlib import Path
import maplibre_tiles

features = []

with open("./points.csv", newline="") as f:
    reader = csv.DictReader(f)
 
    for row in reader:
        print(row)
        features.append({
            "type": "Feature",
            "geometry": {
                "type": "Point",
                "coordinates": [
                    float(row["Lon"]),
                    float(row["Lat"]),
                ],
            },
            "properties": {
                k: v for k, v in row.items() if k not in ("Lat", "Lon")
            },
        })

geojson = {
    "type": "FeatureCollection",
    "features": features,
}

mlt = maplibre_tiles.encode_geojson(
    geojson,
    name="points",
)

Path("points.mlt").write_bytes(mlt)

This throws an error

ValueError: input must be a GeoJSON FeatureCollection: invalid type: floating point `35.08891`, expected i32

From this error i get the impression that the use case is a bit different then expected (i haven't read docs in full detail)

Seems the mlt conversion happens at the tile level using the tiles internal projection? But in that scenario, I'm surpirsed that you use geojson as an input format in the examples, probably conversion between mvt <-> mlt is a more common case?

[CommanderStorm]

Yes, we currently expect the tile-local projection, that is so far not a bug..

Likely another method with the projection math would be helpfull though, unless that exists in the python ecosystem.. I am not an expert in python or projections.

To be able to accept lat/lon, we would need to know in addition which xyz this is, so that the bbox of the tile is clear.

Here is the respective section from our docs:

encode_geojson(geojson, name, extent=4096, *, tessellate=False, sort="auto", shared_dict=True, fsst=True, fastpfor=True) -> bytes encodes a single layer to an MLT blob.

geojson is a GeoJSON FeatureCollection. Geometry is in tile-local coordinate space (no projection), matching tilezen/mapbox-vector-tile's default. Coordinates must be integers and 2D. They must be JSON integers (2048), not floats: a float-typed value such as 2048.0 raises ValueError.

https://pypi.org/project/maplibre-tiles/#:~:text=geojson%20is%20a%20GeoJSON%20FeatureCollection.%20Geometry%20is%20in%20tile%2Dlocal%20coordinate%20space%20(no%20projection)%2C%20matching%20tilezen/mapbox%2Dvector%2Dtile%27s%20default

[pvgenuchten]

Thanx @CommanderStorm maybe interesting to mention where in the pipeline this library is of interest

• generate vector tiles from source data in arbitrary format (using pg-tiles, ogr, tipicanoe)
• then use the library to convert individual tiles to mlt?

Tipicanoe is actually in python, maybe you can use it as a library? Other option could be mapproxy. Reprojection is in pyproj, clipping and aggregation not sure