Home

Awesome

OpenMapTiles Tools Build and publish to Docker

The OpenMapTiles toolbox for generating map vector tiles. It includes tools to prepare Imposm mappings and SQL files based on layers defined in OpenMapTiles or similar projects. It also includes map data downloading, parsing, debugging, and performance evaluation tools. We encourage other people to use this for their vector tile projects as well since this approach works well for us.

Docker Images

The repository hosts the code for the following Docker images. They can be built using make build-all-dockers.

openmaptiles-tools

A collection of tools for downloading, parsing, and generating map tiles described below.

import-data

Multiple data sources packaged for import into PostgreSQL DB, includes data from Natural Earth, water polygons, and lake centerlines.

postgis

An image with PostgreSQL database, PostGIS, and several other extensions, custom built for OpenMapTiles project.

postgis-preloaded

The above postgis image pre-loaded with the import-data. This image is mostly used for testing, and may not be appropriate for production. The image has hardcoded user openmaptiles and password openmaptiles.

generate-vectortiles

Legacy Mapnik-based image that simplifies tilelive-copy tile generation. Mapnik is outdated and is not actively maintained. Instead, use generate-tiles script from the openmaptiles-tools docker image.

Additional Documentation

See /docs

Usage

You need either just Docker or Python 3 installed on your system. You also need Docker-compose for testing. If running without Docker, see below for the list of additional tools and libraries.

Usage with Docker

The easiest is to use docker directly to run this command. You do not need to clone openmaptiles-tools locally, just clone the openmaptiles repo and run from its root.

Note: container scripts can only access files from the given directory and below, e.g. in this example - ${PWD} - current dir.

docker run -it --rm -u $(id -u ${USER}):$(id -g ${USER}) \
           -v "${PWD}:/tileset" \
           openmaptiles/openmaptiles-tools \
           <script-name> <script-parameters>

Where the <script-name> could be any of the scripts in the bin/ directory, e.g. generate-imposm3 openmaptiles.yaml.

Docker Volumes

Using without Docker

# Some tool require these packages. On Debian/Ubuntu you can install them with
sudo apt install graphviz sqlite3 aria2 osmctools

# install the package directly from git
python3 -m pip install git+https://github.com/openmaptiles/openmaptiles-tools

# Run the script you want, e.g. from the openmaptiles dir:
generate-imposm3 openmaptiles.yaml

# If the script doesn't run, make sure your PATH includes default PIP bin directory.
# On Debian/Ubuntu that would be  ~/.local/bin/  (under your home dir).
# Otherwise just run it with  ~/.local/bin/generate-imposm3 ...

Running from source

Make sure you have all dependencies from the Usage section above. You should have the latest Python (3.6+)

# Get OpenMapTiles layer data
git clone https://github.com/openmaptiles/openmaptiles.git
# Get the tools repo
git clone https://github.com/openmaptiles/openmaptiles-tools.git
# Run scripts from the root of the tools repository
cd openmaptiles-tools
# Install required dependencies (might need to run with sudo
python3.6 -m pip install -r requirements.txt
# The PYTHONPATH=$PWD allows script to find its modules located in the current dir.
PYTHONPATH=$PWD python3 bin/generate-imposm3 ../openmaptiles/openmaptiles.yaml

Tools Development

Use make test to run all of the tests locally. The Makefile will build a docker image with all the code, run all tests, and compare the build result with the files in the testdata/expected dir.

Run make rebuild-expected after you modify the output produced by the generation scripts. This will re-create the expected test results to match the actual ones, and make sure the changes are what you want.

Data Concepts

You define a self contained Layer together with SQL files and layer and data source definitions (like an imposm3 mapping file) that you can then reference in a Tileset where you mix and match with other layers.

Define your own Layer

Take a look or copy a standard layer like building to get started with your own layer. A layer consists out of a Layer definition written in YAML format.

There you specify the layer properties like id, buffer_size and possible Markdown documentation (description and fields). You can also reference SQL files in schema for writing the necessary queries for your layer or create generalized tables. We encourage you to have a function per layer which takes the bounding box and zoom level. This makes it easy to test and reuse.

If your data is based of OSM you can also directly reference a imposm3 mapping file to choose the OSM data you need.

layer:
  id: "building"
  description: Buildings from OpenStreetMap
  buffer_size: 4
  datasource:
    query: (SELECT geometry, render_height, class FROM layer_building(!bbox!, z(!scale_denominator!))) AS t
  fields:
    render_height: An approximated height from levels and height of building.
    class:
      description: Defines a subclass of a building (one of the known values).
      # Values can be either a list of strings, or a dictionary
      # Dictionary defines mapping of OSM values to the OMT field value
      values:
        school:
          # subclass IN ('school','kindergarten') OR subclass LIKE 'uni%'
          subclass: ['school','kindergarten','uni%']
        railway:
          # (subclass='station' AND mapping_key='railway')
          # OR subclass in ('halt','tram_stop','subway')
          - __AND__:
              subclass: 'station'
              mapping_key: 'railway'
          - subclass: ['halt', 'tram_stop', 'subway']
schema:
  - ./building.sql
datasources:
  - type: imposm3
    mapping_file: ./mapping.yaml

For the well known values (enums), the fields section can also contain the mapping of the input (OSM) values.

If a layer SQL files contains %%FIELD_MAPPING: class%%, generate-sql script will replace it

SELECT CASE
    %%FIELD_MAPPING: class%%
END, ...

into

SELECT CASE
    WHEN "subclass" IN ('school', 'kindergarten')
        OR "subclass" LIKE 'uni%' THEN 'school'
    WHEN ("subclass" = 'station' AND "mapping_key" = 'railway')
        OR "subclass" in ('halt','tram_stop','subway') THEN 'railway'
END, ...

Define your own Tileset

A Tileset defines which layer will be in your vector tile set (layers) and metadata used for generating a TM2Source project to actually generate the vector tiles.

tileset:
  layers:
    - layers/building/building.yaml
    - layers/housenumber/housenumber.yaml
    - layers/poi/poi.yaml
  name: Street Level
  description: A tileset showing street level info like building, housenumbers and POIs.
  attribution: "OpenStreetMap contributors"
  maxzoom: 14
  minzoom: 13
  center: [-12.2168, 28.6135, 4]
  bounds: [-180.0,-85.0511,180.0,85.0511]
  pixel_scale: 256
  defaults:
    srs: +proj=merc +a=6378137 +b=6378137 +lat_ts=0.0 +lon_0=0.0 +x_0=0.0 +y_0=0.0 +k=1.0 +units=m +nadgrids=@null +wktext +no_defs +over
    datasource:
      srid: 900913

Testing tiles

Tile size and PostgreSQL querying speed

Use test-perf to evaluate tile generation performance against a PostgreSQL database. This utility can test individual layers, several layers at once, as well as the whole tile. It has several pre-defined testing areas to provide cross-test consistency. Results are printed as histogram graphs, showing tile size distribution. The utility can compare current run with a previously saved one, highlighting large changes. If multiple zoom levels are tested, another histogram shows per-zoom size distribution. Run with --help to see all options.

test-perf <tileset> ...

Just like postserve below, test-perf requires PostgreSQL connection.

Realtime tile server

Postserve is an OpenMapTiles map vector tile test server that dynamically generates metadata and tiles directly from PostgreSQL database based on the tileset file definition.

postserve <tileset> ...

Use postserve <tileset> to start serving. Use --help to get the list of Postgres connection parameters. If you have a full planet database, you may want to use MIN_ZOOM=6 postserve ... to avoid accidental slow low-zoom tile generation.

Postserve quickstart with docker

docker run -it --rm -u $(id -u ${USER}):$(id -g ${USER}) \
    -v "${PWD}:/tileset" --net=host \
    openmaptiles/openmaptiles-tools \
    postserve openmaptiles.yaml

Viewing dynamic tiles

You can view tiles with any MVT-supporting viewer, such as:

Examining realtime tile content

Use debug-mvt tool to examine tile content. The tool will query PostgreSQL server and show layers with each data row and geometry type/size. This tool can limit output to just a few layers, optionally show all localized names, and show geometries as text. This example queries shows entries in the "place" layer for the specified tile:

$ debug-mvt openmaptiles.yaml 4/7/6 -l place
======= Layer place (extra name columns are hidden by default) =======
  capital  class    iso_a2    name                        name_de                     name_en                       rank       osm_id  is_valid_mvt    mvtgeometry    is_valid_geom    geometry
---------  -------  --------  --------------------------  --------------------------  --------------------------  ------  -----------  --------------  -------------  ---------------  ----------
           country  ES        Espaa                       Spanien                     Spain                            1   1483323000  True            POINT(32)      True             POINT(32)
           country  PT        Portugal                    Portugal                    Portugal                         1  23770282470  True            POINT(32)      True             POINT(32)
        4  city               Casablanca                  Casablanca                  Casablanca                       2    257307240  True            POINT(32)      True             POINT(32)
           country  MA        Maroc                       Marokko                     Morocco                          2   4324250410  True            POINT(32)      True             POINT(32)
        2  city               Madrid                      Madrid                      Madrid                           2    210682950  True            POINT(32)      True             POINT(32)
        2  city               Lisboa                      Lissabon                    Lisbon                           3   2659584900  True            POINT(32)      True             POINT(32)
        4  city               Oran                        Oran                        Oran                             4    275651030  True            POINT(32)      True             POINT(32)
        4  city               Valncia                     Valencia                    Valencia                         4    341056070  True            POINT(32)      True             POINT(32)
        4  city               Sevilla                     Sevilla                     Seville                          4   2488206560  True            POINT(32)      True             POINT(32)
        4  city               Fs                          Fs                          Fez                              4   2890354320  True            POINT(32)      True             POINT(32)
        2  city               Rabat                       Rabat                       Rabat                            4   2991208620  True            POINT(32)      True             POINT(32)
        ...

Profile PostgreSQL functions

Use profile-pg-func to compare PostgreSQL function execution speed. Each function is called thousands of times in several runs. The fastest and slowest runs are discarded. profile-pg-func can import SQL files before running the test, e.g. to add the latest developer versions of the function(s).

Show layer statistics for a field

Use layer-stats show per zoom statistics for some column (field) in a single layer. Supports several metrics:

Tools

Environment variables

Various tools require these environment variables to be set

Multi-streamed OSM Data Downloader

download-osm tool can be used to download an area extract or the entire planet file, and validate file content. The entire planet file is downloaded from all available OSM mirrors at the same time to distribute the load across mirrors, and makes download faster. The tool will ensure you get the latest version, verifies that all mirrors contain the same version, and validates the download with md5 hash. By default the tool will not download from the primary OSM site to reduce its load. Downloader uses aria2c. Downloader can also get files from Geofabrik, BBBike, and openstreemap.fr, or an arbitrary URL.

# downloads the latest version to the current directory
download-osm planet

# download to the target dir by passing -d ... param to aria2c
download-osm planet -- -d ./downloads

# download New Zealand extract from Geofabrik, together with the state file
download-osm geofabrik new-zealand --state state.txt

# List all extracts available from Geofabrik
download-osm list geofabrik

Generate SQL to create MVT tiles with PostGIS ST_AsMVT function

Uses tileset definition to create a PostgreSQL prepared or create function SQL code to generate an entire vector tile in the Mapbox Vector Tile format with a single getTile(z,x,y) query using PostGIS MVT support.

Use --help to get all parameters.

NOTE: Known bug is PostgreSQL JIT could make tile generation horribly slow in PG11+, and may need to be disabled.

generate-sqltomvt <tileset>

Generate Imposm3 Mapping File

Generate Imposm3 mapping file to import OSM data using a tileset definition file.

generate-imposm3 <tileset>

Generate SQL scripts

Assembles all SQL referenced in the layer definitions into an SQL script that can be executed with psql. If --dir option is given, generates .sql files that can be executed in parallel.

generate-sql <tileset>
generate-sql <tileset> --dir <outputdir>

Generate Markdown Documentation

Takes a tileset definition and generates Markdown documentation.

generate-doc <tileset>

Generate MBTiles tile snapshot file

generate-tiles script generate tiles from PostgreSQL using tilelive-copy with tilelive-pgquery. If run with LIST_FILE environment variable, this script will only generate tiles given in the list file, and it will ignore min/mid/max zoom values. For example, set LIST_FILE=/export/tiles.txt to match what was previously done with export-list.sh script.

If the LIST_FILE and MID_ZOOM environment variables are not set, this script will generate all tiles from MIN_ZOOM to MAX_ZOOM (inclusive).

If MID_ZOOM env var is set, generate-tiles will first generate all tiles from MIN_ZOOM to MID_ZOOM. Afterwards, mbtiles-tools impute finds the most duplicated small tiles at MID_ZOOM, and treats them as "empty" -- it copies empty tiles to the next MID_ZOOM+1 level as is, without regenerating them. The non-empty tiles at MID_ZOOM+1 are then generated as before using a list of tiles generated by the impute command. Once done, the same steps are repeated for MID_ZOOM+2, etc., until the MAX_ZOOM (inclusive).

Lastly, generate-tiles runs mbtiles-tools meta-generate to update mbtiles metadata unless TILESET_FILE is not set.

generate-tiles script does not take any parameters, but accepts many environment variables -- see code.

Generate ETL (Extract-Transform-Load) graph

dependency: graphviz

Takes a source code from the imposm3 mapping file and the SQL postprocessing code, and parsing for the etldoc: graphviz based comments, and generate an svg file. The .dot and the .svg filename prefix is etl_

generate-etlgraph <tileset> <target-directory>
generate-etlgraph layers/landcover/landcover.yaml  ./build/devdoc
generate-etlgraph layers/railway/railway.yaml      ./build/etlgraph

example:

input command: generate-etlgraph layers/landcover/landcover.yaml output fies:

Generate SQL query for a given layer and zoom level

example:

generate-sqlquery layers/landcover/landcover.yaml 14

Import and Update OSM data

The import-osm, import-update, and import-diff tools will import and update PostgreSQL database by running imposm. The tools expect these env vars: PGHOST, PGDATABASE, PGUSER, PGPASSWORD, and optionally PGPORT to connect to the PostgreSQL server, and a number of other vars for imposm configuration. See scripts.

Import Wikidata localized names

The import-wikidata tool searches for all wikidata tags in the database, and uses Wikidata Query Service to get the labels in all languages. This tool looks at all tables defined in layers' mapping files if they contain tags hstore field, and if the table has no _resolve_wikidata: false flag. The localized names are written to the wd_names table, which will be created if missing.

import-wikidata openmaptiles.yaml

Mbtiles file tools

This command allows users to examine and manipulate mbtiles file:

mbtiles-tools --help
mbtiles-tools ./data/tiles.mbtiles meta-all

Generate TM2Source Projects for Mapbox Studio Classic

Takes a tileset definition and generates a TM2Source YAML project file. You need to provide PostgreSQL database connection settings before generating the project.

generate-tm2source <tileset> --host="localhost" --port=5432 --database="osm" --user="osm" --password="osm"

Import OSM Borders

The import-borders script will take the first PBF file from the /import dir (by default), extract borders with osmborder tool, and import resulting CSV file into the database as osm_border_linestring table (by default). Example usages:

import-borders                      # Parse and import first PBF file in PBF_DATA_DIR
import-borders [import] planet.pbf  # Parse and import planet.pbf
import-borders parse planet.pbf     # Parse planet.pbf into a CSV file, but do not import
import-borders load borders.csv     # Load borders.csv into a table

This utility requires PostgreSQL's PG* environment variables, and optionally uses PBF_DATA_DIR, BORDERS_PBF_FILE, BORDERS_CSV_FILE, BORDERS_TABLE_NAME.

Importing into Postgres

The import-sql script can execute a single SQL file in Postgres when the file is given as the first parameter.

If ran without any arguments, import-sql executes all of the following:

Generating and importing SQL could be done in a single step with &&, e.g.

generate-sqltomvt openmaptiles.yaml > "$SQL_DIR/mvt.sql" && import-sql

Optionally you may pass extra arguments to psql by using PSQL_OPTIONS environment variable. For example PSQL_OPTIONS=-a makes psql echo all commands read from a file into stdout. PSQL_OPTIONS allows multiple arguments as well, and understands quotes, e.g. you can pass a whole query as a single argument surrounded by quotes -- PSQL_OPTIONS="-a -c 'SELECT ...'"

Environment variables

Most PostgreSQL-related images support standard PostgreSQL environment variables like PGUSER, PGPASSWORD, PGHOST, PGDATABASE, and optionally PGPORT.

The only exception is the postgis image, which uses different variables (POSTGRES_USER, POSTGRES_PASSWORD, POSTGRES_HOST, POSTGRES_DB, POSTGRES_PORT) during database creation due to how the official Docker postgres image has been set up.

Performance optimizations

Materialized views can be refreshed in parallel using refresh-views command. This could be especially useful if the CREATE MATERIALIZED VIEW statements had WITH NO DATA clause.