Comment on page
The tiler is a module of the Analytics Toolbox for BigQuery that allows to process and visualize very large spatial datasets stored in BigQuery.
If you have small datasets in BigQuery (few megabytes), there are solutions such as BigQuery Geo Viz to visualize them; but if you have millions or even billions of rows, you will need a system to load them progressively on a map. The Tiler allows you to do that without having to move your data out of BigQuery.
The Tiler will process your data and create a complete tileset out of it. Each individual tile is a row in this table, with the tile coordinates and the encoded MVT stored in different columns:
The integration of tilesets with custom web map applications is also possible with CARTO Maps API, which will connect to BigQuery using your connection’s Service Account credentials to fetch and serve the tiles in a standard format, so they can be used with any web-mapping library or desktop GIS application.
The tiler module enables the creation of three types of tilesets through stored procedures: simple, aggregation and spatial index tilesets. Simple tilesets encode all the input features as is, while aggregation tilesets encode aggregations over the input features. Therefore, you should use simple tilesets for visualizing a dataset of world rivers, but use an aggregation tileset to visualize a heatmap of the trees distribution in New York City. On the other hand, spatial index tilesets allow the creation of tilesets aggregating data from an input table that uses H3 or Quadbin spatial indexes as geographic support systems.
We provide the following set of procedures to create tilesets:
- This procedure creates a simple tileset. You should use it if you have a dataset with any geography type (point, line, or polygon) and you want to visualize it at an appropriate zoom level.
- The geographies will be represented exactly as stored in BigQuery, which means that if they are too small to be visible at a certain zoom level they won’t be included in the tiles at that zoom level.
- The values associated with each feature are the same as the ones available in the source dataset.
carto.CREATE_TILESETis capable of finding the right configuration for your input data, whereas this procedure requires you to set them yourself.
- Please use this procedure only if you need a really specific configuration for your tileset or need to tweak a particular option that it’s not available in
- Use this procedure if you have a point dataset (or anything that can be converted to points, such as polygon centroids) and you want to see it aggregated.
- The points will be aggregated into cells. Each feature or cell represents all the points that fall under it, so the associated properties available for visualization are generated by aggregating the values in the source dataset.
- Values of individual points are available using
single_point_propertieswhich will only be included when a cell includes only one point. Remember that you could also get similar values with the aggregated properties using functions like
- Use this procedure if you have a dataset based on spatial indexes, and you want to build a tileset aggregating data in the same spatial index (H3 and QUADBIN are currently supported).
- Aggregated data will be computed for all levels between
- For each resolution level, all tiles for the area covered by the source table are added, with data aggregated at level
resolution + aggregation resolution.
The tiler is:
- Convenient – It can be run directly as SQL commands in BigQuery. The data never leaves BigQuery so you won’t have to worry about security and additional ETLs.
- Fast – CARTO BigQuery Tiler benefits from the massive scalability capabilities of BigQuery and can process hundreds of millions of rows in a few minutes.
- Scalable – This solution works well for 1M points or 100B points.
- Cost-effective – Since BigQuery separates storage from computing, the actual cost of hosting these tilesets is very low. Additionally, since the tiling process runs on-demand, you’ll only pay for that processing and you won’t need to have a cluster available 24/7. Finally, we have optimized how we serve the tiles, thanks to our partitioning algorithms.