Convert ESRI's Vector Tile Package (VTPK) to mbtiles

Resulting mbtiles can be used with mapbox-gl-js (Web), mapbox-gl-native (iOS/Android) or offline with QGIS via Vector-Tiles-Reader Plugin.

Disclaimer 1

I don't have access to any ESRI software or data. This is tested with the Austrian basemap.at Verwaltungsgrundkarte Vektor Offline (Beta) Österreich (EPSG:3857) only.

A few parameters are hard coded towards basemap.atas well, as some metadata values in the vtpkg are not correct.

Download vtpk here: https://www.data.gv.at/katalog/dataset/b694010f-992a-4d8e-b4ab-b20d0f037ff0

Disclaimer 2

No magic sauce or reverse engineering in the code, the internals of ESRI cache format are documented here: https://github.com/Esri/raster-tiles-compactcache

Another piece of software that does the same thing: https://github.com/syncpoint/openvtpk

Disclaimer 3

This is a one off proof of concept written months ago, putting it online now for educational purposes, unblocking offline and open source tool use. And free 🍻, maybe. Implementation is quick and dirty. No plans for further development. Feel free to clone, fork, ...

Note

For convenience a docker image is provided. For fast conversion Linux is recommended.

On my laptop:

• native on Windows
  • extracting tiles to local file system (xy-scheme): ~30 seconds
  • converting to mbtiles: ~50 seconds
• docker on Windows (Linux container)
  • extracting: ~3 minutes
  • converting: ~20 minutes(!)
• docker on Linux (Linux container)
  • extracting: ~4 seconds(!)
  • converting: ~30 seconds

Usage

• download bmapv_vtpk_3857.vtpk https://www.data.gv.at/katalog/dataset/b694010f-992a-4d8e-b4ab-b20d0f037ff0
• extract vtpk to a folder (vtpk is just a zip after all), eg
  • Linux: mkdir -p ~/basemap && unzip bmapv_vtpk_3857.vtpk -d ~/basemap/bmapv_vtpk_3857
  • Windows: md C:\basemap && 7z x bmapv_vtpk_3857.vtpk -oC:\basemap\bmapv_vtpk_3857

vtpk2mbtiles </path/to/extracted/vtpkg/folder> <output> <decompress tiles>

• </path/to/extracted/vtpkg/folder>: folder where the vtpkg has been extracted to
• <output>:
  • if it ends with .mbtiles destination format will be a mbtiles file
  • otherwise it is assumed to be a folder: single tiles will be placed there. With "current" vtpk version (Basemap_20190617): +162,000 tiles
• <decompress tiles>: if true tiles will be decompressed, if false (default) not.

Abort processing with Ctrl+C.

Linux

• convert to mbtiles, not decompressing tiles:

docker run -it --rm --name vtpk2mbtiles \
 --mount src="${HOME}/basemap",dst=/data,type=bind bergwerkgis/vtpk2mbtiles \
 /data/bmapv_vtpk_3857 \
 /data/bmapv.mbtiles \
 false

• extract tiles to file system, unzip tiles:

docker run -it --rm --name vtpk2mbtiles \
 --mount src="${HOME}/basemap",dst=/data,type=bind bergwerkgis/vtpk2mbtiles \
 /data/bmapv_vtpk_3857 \
 /data/bmapv-tiles \
 true

Windows

• Temporarily switch off Real-time protection (Virus & threat protection -> Virus & threat protection settings -> Manage settings -> Real-time protection -> Off). This will increase conversion speed considerably, especially when exporting to seperate tile files.

• Docker

  • Right click Docker icon in the taskbar: Switch to Linux Containers
  • Docker -> Settings -> Resources -> Filesharing: check the drive where the downloaded vtpk resides
  • Docker -> Settings -> Resources -> Advanced:
    • bump CPUs to the max
    • at least 3 GB Memory, as much as you can afford (expecially for mbtiles conversion)
  • Apply & Restart

• convert to mbtiles, not decompressing tiles:

docker run -it --rm ^
 --name vtpk2mbtiles ^
 -m 4g ^
 --cpus=%NUMBER_OF_PROCESSORS% ^
 --mount src="C:\basemap",dst=/data,type=bind bergwerkgis/vtpk2mbtiles ^
 /data/bmapv_vtpk_3857 ^
 /data/bmapv.mbtiles ^
 false

• extract to file system, unzip tiles:

docker run -it --rm ^
 --name vtpk2mbtiles ^
 -m 4g ^
 --cpus=%NUMBER_OF_PROCESSORS% ^
 --mount src="C:\basemap",dst=/data,type=bind bergwerkgis/vtpk2mbtiles ^
 /data/bmapv_vtpk_3857 ^
 /data/bmapv-tiles ^
 false

Converted mbtiles in mbview: