This is the source code and styles for the CartoDB Basemaps, designed by Stamen.

The code and styles here are intended for serving the basemaps on your own local CartoDB instance, if you just want to use them for your own map, you should use our hosted version.

• All CartoCSS styles live in /tilemill. The Tilemill Project MML (used for local Tilemill 1) and the project.json in the root (used for CartoDB) need to be manually kept in sync. Each layer should have a name the same as its CartoCSS ID, and a call to a SQL function.

  • The oddness here has to do with Tilemill and CartoDB parsing SQL differently, and also because CartoDB does not support zoom-dependent layers yet; so we instead make layers zoom-dependent by writing PL/PGSQL functions that take a mapnik !scaleDenominator!.
  • The stylesheet global_variables may be substituted in variant styles e.g. Dark Map.

• The project.json is transformed into a Windshaft MapConfig via util.js. All util does is inline text from stylesheets, and possibly toggles layers depending on which "toggle set" (e.g. Labels Off) they belong to. Disabling a layer is done by appending LIMIT 0 to the SQL query.

• There's two places where database stuff (materialized views, PL/PGSQL functions) is defined.

  • global_functions.sql is where all the functions go. This needs to be loaded first.
  • generalizations.yml describes the materialized views, this is read by generalizations_sql.js to output either raw SQL or issue queries.

Create a file in the root of this directory called config.json with your CartoDB host and API key like this:

{"api_key":"API_KEY", "cdb_url": "https://myuser.cartodb.com"}

1. run node import_files.js to import Natural Earth and Coastline data to your instance
2. Import an OSM extract into your db, using imposm3_mapping.json. See the Makefile or https://github.com/cartodb/osm
3. import global_functions.sql into your DB.
4. run generalizations_sql.js into your DB.
5. Ensure the Azo Sans, DejaVu unicode, unifont fonts are available on your instance.
6. Open index.html to see your basemap being rendered from CartoDB. The styles will be regenerated on refresh.

1. run sh download_natural_earth_data.sh to get all the NE files.
2. Run make coastline to get the water/land polygons into your DB
3. Import an OSM extract into your db. see the Makefile.
4. run sh ne2pgsql.sh to import Natural Earth shapefiles locally such that they resemble CartoDB tables. This also imports the z4to10.json file from this repository, which is a hand-curated list of low zoom city points.
5. import global_functions.sql into your DB (or run make globals)
6. run generalizations_sql.js into your DB (or run make generalizations)
7. run make install to get a local TileMill project
8. You may need the Azo Sans, DejaVu unicode, unifont fonts installed.

That should be it for getting the project running in a local TileMill.

The generalizations go into a different schema that the style (project.json) refers to. Creating all the materialized view takes 6-7 hours for the full planet (and requires that this be run on a machine with psql access). You can speed up the generalization creation by using the script and specifying a bigger amount of threads.

To create these:

nodejs generalizations_sql.js [TARGET_SCHEMA_NAME] [DATABASE_URL] [THREADS]

(On Debian, nodejs is the node executable.)

This creates MATERIALIZED VIEWS with the subsets of data required for each zoom.

As materialized view creation / refreshing is blocking process (as of PostgreSQL 9.3), you can use TARGET_SCHEMA_NAME to create the generalizations into different schemas (planet_blue or planet_green, for example) and then, when done, switch the schema referred to in project.json.

If you want to develop against CartoDB on an extract you can derive an extract from an existing OSM dump on public.planet like follows:

CREATE MATERIALIZED VIEW sf_madrid AS
SELECT id, osm_id, tags, the_geom
FROM public.planet
WHERE the_geom && ST_Transform(ST_SetSRID('BOX(-122.737 37.449,-122.011 37.955)'::box2d, 4326), 3857) OR
the_geom && ST_Transform(ST_SetSRID('BOX(-4.293 39.8,-3.057 41.03)'::box2d, 4326), 3857) OR
ORDER BY ST_GeoHash(ST_Transform(the_geom, 4326));
CREATE INDEX planet_the_geom_gist ON sf_madrid USING GIST(the_geom);

and build your generalizations off this instead of public.planet.

To create a "Named Map", so users can access the basemap without API key, running make named_maps will create them using the local styles.