saved_tile_directory.md

Generating a Directory of Tiles

Avecado can be used to pre-render a set of vector tiles with the demo HTTP server.

To do so, a Mapnik XML file is needed. Mapbox Studio will do this, or Kosmtik can be used to convert a YAML layer file to Mapnik XML. Assuming a TileMill 2 vector source located in $TM2SOURCE is being used, the latter can be done with nodejs index.js export $TM2SOURCE/data.yml --output $TM2SOURCE/data.xml.

Avecado needs to be downloaded and built. When this is done, the HTTP server can be run. Adjust the number of threads to the number of CPU cores.

./avecado_server $TM2SOURCE/data.xml 4002 --thread-hint 4

In another console window, use curl and GNU Parallel to scrape tiles into a suitable directory structure. The Slippy Map tile directory structure is very simple which allows this. GNU Parallel can be obtained with sudo apt-get install parallel on Debian based systems.

Use the same number of jobs as avecado_server threads.

(for z in $(seq 0 8); do
  for y in $(seq 0 $((2**z - 1))); do
  mkdir -p ${z}/${y}
    for x in $(seq 0 $((2**z - 1))); do
      echo "${z}/${y}/${x}"
    done;
  done;
done;) | parallel --progress –jobs 4 curl -s --retry 5 --retry-delay 25 -o {}.pbf http://localhost:4002/{}.pbf
# Save the tile JSON too
curl -o tile.json http://localhost:4002/tile.json

Going to where the files are being saved and counting the number of files gives better view of progress. This can be done with

for d in `ls -v`; do echo $d: $(find $d -name '*.pbf' | wc -l)/$((4**d)); done

Downloading zoom levels above 8 may take considerable time, as there are a quarter million tiles on zoom 9, one million on zoom 10, and it increases exponentially.