fdbgeo

go-fdbgeo

go get -u github.com/umpc/go-fdbgeo

Package fdbgeo contains tools for building geospatial layers using FoundationDB with Geohash-encoded keys.

This package uses its complimentary zrange package, for performing geospatial range queries using FoundationDB with Geohash-encoded keys and a search radius.

The RadialRange method appears to be sufficient for range queries of around 5,000km or less. Changes that efficiently add support for larger query ranges are welcome here.

Note: As of 5/26/2018, changes from apple/foundationdb PR #408 must be applied to use this package, or a runtime panic will occur during subspace encoding.

Example usage

...

rangeParams := fdbgeo.RadialRangeParams{
  Subspace:  dir,
  Radius:    32.18688,
  Latitude:  37.334722,
  Longitude: -122.008889,
}

keyRanges := rangeParams.RadialRange()

fdbRangeOptions := fdb.RangeOptions{
  Mode: fdb.StreamingModeWantAll,
}

ret, err := db.ReadTransact(func(tr fdb.ReadTransaction) (ret interface{}, e error) {
  rangeResults := make([]fdb.RangeResult, len(keyRanges))

  for i, keyRange := range keyRanges {
    rangeResults[i] = tr.GetRange(keyRange, fdbRangeOptions)
  }

  var kvList []fdb.KeyValue
  for _, rangeResult := range rangeResults {
    ri := rangeResult.Iterator()

    for ri.Advance() {
      kv := ri.MustGet()

      geohashID, err := fdbgeo.UnpackUint(kv.Key, -2)
      if err != nil {
        panic(err)
      }

      // Running WithinRadius before value parsing may benefit performance,
      // though it depends on the data model in use.
      if !rangeParams.WithinRadius(geohashID) {
        continue
      }

      kvList = append(kvList, kv)
    }
  }

  return kvList, e
})

...

Note: Geohash range searches are imprecise. Results should be filtered using the input radius where precision is desired.

References

• The RadialRange method was inspired by the algorithm in the "Search" section of this page.