Go package for working with Who's On First documents and the Tile38 datastore
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.



Go tools for working with Who's On First documents and Tile38


This is tilting towards ready on the experimental-to-ready scale, but isn't quite there yet.


You will need to have both Go and the make programs installed on your computer. Assuming you do just type:

make bin

All of this package's dependencies are bundled with the code in the vendor directory.


Indexing Who's On First data is done using the wof-tile38-index utility described below. There are a few important things to remember about indexing:

  1. This is very much Who's On First (Mapzen) specific
  2. The goal is to index as little extra information as possible which ensuring the ability to generate a "minimal viable WOF record" and to perform basic container-ish (belongs to, placetype, etc.) queries.
  3. Those details are still a moving target.
  4. To whit, see the way we're encoding the repository name in to the first key itself? That is perhaps unnecessary.
  5. Whatever else this package holds hands with, now or in the future, it currently holds hands (tightly) with the lib_whosonfirst_spatial.php library in the whosonfirst-www-api repo and the py-mapzen-whosonfirst-tile38 library.


This stores a record's geometry (which may be a centroid or the actual GeoJSON geometry property) as well as the following numeric IDs:

  • wof:id
  • wof:placetype_id
  • wof:parent_id
  • wof:is_deprecated (stored as 0 or 1)
  • wof:is_superseded (stored as 0 or 1)

This we probably should store but aren't yet:

  • mz:is_current
  • mz:scale_rank

WOFID + "#meta"

This stores a following fields as a JSON encoded dictionary:

  • wof:name
  • wof:country


Querying Who's On First data can either be done by talking to a Tile38 server directly or using one of the utilities described below (currently there is only one).

Query results

Unless otherwise noted all query results are return as a simple list of JSON dictionaries. Paginated is implemented using a cursor parameter which is returned at top level of any query response. If present you should include it with the following query to return the next set of results. For example:

$> curl -s 'localhost:8080?bbox=-33.893217,151.165524,-33.840479,151.281223&per_page=1' | python -mjson.tool
    "cursor": 1,
    "results": [
            "wof:id": 1108814711,
            "wof:is_deprecated": 0,
            "wof:is_superseded": 0,
            "wof:parent_id": -1,
            "wof:placetype_id": 102312319,
            "wof:repo": "dxlabs"


All of these utilities assume that there is a running copy of the tile38-server (included in Tile38 package) that these utilities can communicate with. This package defines tools and utilities for working with tile38-server not to replace it.


Find Who's On First records in a Tile38 database that intersect a given bounding box.

./bin/wof-tile38-bboxd -h
Usage of ./bin/wof-tile38-bboxd:
  -host string
    	The address your HTTP server should listen for requests on (default "localhost")
  -port int
    	The port number your HTTP server should listen for requests on (default 8080)
  -tile38-collection string
    	The name of the Tile38 collection to read data from.
  -tile38-host string
    	The address your Tile38 server is bound to. (default "localhost")
  -tile38-port int
    	The port number your Tile38 server is bound to. (default 9851)

Query parameters

  • bbox required – Any bounding box format that is supported by the go-whosonfirst-bbox package.
  • ** scheme** optional A valid go-whosonfirst-bbox scheme
  • ** scheme** optional A valid go-whosonfirst-bbox order
  • per_page optional The number of results to include with a query. Default is 100.
  • cursor optional A pointer to the next set of results for your query.


This assumes you've created an index called dxlabs. See below for details.

$> wof-tile38-bboxd -tile38-collection dxlabs

$> curl 'localhost:8080?bbox=-33.893217,151.165524,-33.840479,151.281223&per_page=1&cursor=1'


Index one or more Who's On First records in a Tile38 database.

./bin/wof-tile38-index -h
Usage of ./bin/wof-tile38-index:
    	Go through all the motions but don't actually index anything.
  -geometry string
    	Which geometry to index. Valid options are: centroid, bbox or whatever is in the default GeoJSON geometry (default).
  -mode string
    	The mode to use importing data. Valid options are: directory, filelist and files. (default "files")
    	Enable the (walk.go) NFS kludge to ignore 'readdirent: errno' 523 errors
  -procs int
    	The number of concurrent processes to use importing data. (default is number of CPUs * 2)
  -tile38-collection string
    	The name of the Tile38 collection for indexing data.
  -tile38-host string
    	The address your Tile38 server is bound to. (default "localhost")
  -tile38-port int
    	The port number your Tile38 server is bound to. (default 9851)
    	Be chatty about what's happening. This is automatically enabled if the -debug flag is set.


For example, if you wanted to index all the localities in Who's On First:

$> wget https://whosonfirst.mapzen.com/bundles/wof-locality-latest-bundle.tar.bz2
$> tar -xvjf wof-locality-latest-bundle.tar.bz2
$> wof-tile38-index -collection whosonfirst -mode directory wof-locality-latest-bundle/data/

If you wanted to index one or more Who's On First "meta" files (they're just CSV files with a path column) you might do something like:

$> wof-tile38-index -tile38-collection whosonfirst-geom -mode meta /usr/local/data/whosonfirst-data/meta/wof-county-latest.csv:/usr/local/data/whosonfirst-data/data

The syntax for listing meta files to index is a pair of filesystem paths separated by a :. The first path is the path to the meta file and the second is the path to the directory containing the actual GeoJSON files. As of this writing it is assumed that the paths listed in the meta files are relative.

See also