I’m new to UDF functions and I have created a BigQuery UDF that takes a polygon geometry and creates points with it. I’m trying to draw dot density maps (converting polygon + population number to points). I’ve adapted the code from this blog post. Because bigQuery doesn’t have a way to log variables, I’ve been testing things out in this codepen.
I’m at a point where the function seems to work right. The output is a geometry collection of points. It says in the bigquery docs that st_geogfromgeojson
can accept a geometry collection.
My UDF returns a stringified geometry collection.
But I cannot figure out why st_geogfromgeojson
doesn’t work. I can’t tell if I’m simply not unnesting something or what.
CREATE TEMP FUNCTION myFunc(feature string, ethnicity_column FLOAT64, year INT64) RETURNS string LANGUAGE js OPTIONS ( library=["https://storage.googleapis.com.../d3.js","https://storage.googleapis.com/.../turf.min.js","https://storage.googleapis.com/.../wellknown.js"] ) AS """ if (feature === undefined || feature === null) return; var feature_parsed = wellknown.parse(feature) const bounds = turf.bbox(feature_parsed); const populationData = Math.round(ethnicity_column / 10); if (!populationData) return; const x_min = bounds[0]; const y_min = bounds[1]; const x_max = bounds[2]; const y_max = bounds[3]; let hits = 0; let count = 0; const limit = populationData * 10; // limit test to 10x the population. let points = []; while (hits < populationData - 1 && count < limit) { const lat = y_min + Math.random() * (y_max - y_min); const lng = x_min + Math.random() * (x_max - x_min); const randomPoint = turf.point([lng, lat]); if (turf.booleanPointInPolygon(randomPoint, feature_parsed)) { points.push(randomPoint); hits++; } count++; } return JSON.stringify((turf.geometryCollection(points))); // return JSON.stringify(points) """; SELECT ST_GEOGFROMGEOJSON(JSON_EXTRACT((myFunc(st_astext(geom), white_pop, 2018)),'$')) FROM `myteam.kyle_data.blockgroups_with_acs`
But I keep hitting random errors like I’m not using the function right
I’m open to all suggestions. I return a string for simplicity but perhaps I need to use a STRUCT. Perhaps I should cut out turf from creating the points? I must be missing something here.
Advertisement
Answer
Two things that fixed it:
- Returning
array<String>
instead of string - CROSS JOIN UNNEST in the
select
CREATE TEMP FUNCTION myFunc(feature string, ethnicity_column FLOAT64) RETURNS array<String> LANGUAGE js OPTIONS ( library=["https://storage.googleapis.com/../d3.js","https://storage.googleapis.com/.../turf.min.js","https://storage.googleapis.com/.../wellknown.js"] ) AS """ geopath = d3.geoPath() if (feature === undefined || feature === null) return; var feature_parsed = wellknown.parse(feature) const bounds = turf.bbox(feature_parsed); const populationData = Math.round(ethnicity_column / 10); if (!populationData) return; const x_min = bounds[0]; const y_min = bounds[1]; const x_max = bounds[2]; const y_max = bounds[3]; let hits = 0; let count = 0; const limit = populationData; // limit test to 10x the population. let points = []; while (hits < populationData - 1 && count < limit) { const lat = y_min + Math.random() * (y_max - y_min); const lng = x_min + Math.random() * (x_max - x_min); const randomPoint = turf.point([lng, lat]); if (turf.booleanPointInPolygon(randomPoint, feature_parsed)) { points.push('POINT ('+lng+' '+lat+')'); hits++; } count++; } return points; """; SELECT st_geogfromtext(points) as the_geom, 'white' as ethnicity from (SELECT (myFunc(st_astext(geom), white_pop)) as points FROM `tableonmybq`) CROSS JOIN UNNEST(points) as points