Geospatial FAQ

Why Python, when there's spatial SQL?

Python is the go-to language for spatial data science. Although spatial SQL joins and transformations can be efficiently performed using PostGIS in an external database, you may eventually need to convert that data to Pandas and NumPy for further processing and analysis, especially for detailed operations on raster arrays. Additionally, you can run SQL directly on Fused using Python libraries like DuckDB, combining the strengths of both approaches.

What's the benefit of geo partitioning vector tables?

It enables efficient reading of large datasets by strategically partitioning GeoParquet files. Fused's GeoParquet format includes metadata that allows for spatial filtering of any dataset, loading only the chunks relevant to a specific area of interest. This approach reduces memory usage and allows you to work with any size dataset with just Python.

How can I create a set of tiles that cover a given geometry?

Use cases like creating chips may call for running a UDF across a set of tiles that fall within a given geometry. This can be done by creating a list of tiles with the mercantile library then calling the UDF in parallel.

import fused
import mercantile

bounds = [32.4203, -14.0933, 34.6186, -12.42826]

tile_list = list(mercantile.tiles(*bounds,zooms=[15]))

What's the best way to create a map from my UDF data and share it with external teams?

You can use the App Builder create an app that loads the UDF's data then create a shareable link.