Geohashing in R

Oliver Keyes


geohashes are a way of representing latitude/longitude pairs as individual, short strings of numbers and letters. The geohash package provides tools for creating and manipulating these.

Encoding and decoding

To turn a set of latitude/longitude pairs into a geohash, you need the gh_encode function. This accepts latitude, longitude and precision arguments - where precision refers to the number of characters the hash should take up and (correspondingly) how accurate it is. Any precision between 1 and 10 is acceptable - something outside that range will produce a warning, and result in a default of 6 being used.

# Encode!
gh_encode(42.60498046875, -5.60302734375, 5)
# [1] "ezs42"

To decode, you use gh_decode, which works the other way around - it accepts hashes, and returns a data.frame of latitude, longitude, and the error rates for each (since some degree of inaccuracy is introduced by the hashing process):

# Decode!
# lat      lng      lat_error  lng_error
# 42.60498 42.60498 0.02197266 0.02197266

Both functions are fully vectorised, so if you’ve got an entire set of latitudes and longitudes (or an entire set of hashes) it’s not a problem! They’re based on compiled code and can handle a million lat/long pairs (or hashes) in a third of a second.


Because geohashes are calculated for geographic “boxes”, they have neighbours; a hash that represents the block to the north, or south, or northwest, or…so on and so forth. You can extract these in one of two ways. With gh_neighbours, it’ll extract every hash around the hash you provide:

# Get all neighbours!
#  north northeast  east southeast south southwest  west northwest
#  ezs48     ezs49 ezs43     ezs41 ezs40     ezefp ezefr     ezefx

If you only want a specific neighbour, no problem! You can use convenience functions named after the cardinal and intercardinal points (north(), northwest() etc) to grab individual neighbours:

# Get the southeast neighbours!
# [1] "ezs41"

These functions are slower than encoding and decoding, since they’re a lot more complex, but still pretty fast.

Feature ideas or issues

If you have ideas for other geohash-related features, or run into a bug, the best approach is to either request/report it or add it!