geocoder ·

Use the power of MongoDB to make spatial queries, find boundaries and more for data analaysis from R.

roadmap

• list of UK geocodes (most used)
• list of another country’s?
• enable generating MongoDB spatial queries

Dependencies

• mongolite (main dependency)
• geojsonsf (perfect package for link between sf and Mongo spatial)
• jsonify (helping with validation)
• sf

Import vancouver geojson data

Given a local docker instance of MongoDb

# on Linux
docker pull mongo
docker run -d -p 27017:27017 --name mongodb mongo
# docker should be running at 27017

library(mongolite)
library(sf)
#> Linking to GEOS 3.7.2, GDAL 2.4.2, PROJ 5.2.0
test = mongo(url = "mongodb://localhost:27017")
class(test)
#> [1] "mongo"       "jeroen"      "environment"

Lets add real data (a small slice of Uber’s Vancouver dataset.)

# geojson file from Uberr
v.path = file.path(tempdir(), "vancouver.json")
if(!exists(v.path)) {
  v.url = paste0("https://github.com/layik/geocoder/releases/",
                 "download/data/v10.geojson")
  download.file(v.url, v.path)
}
v = geojsonsf::geojson_sf(v.path)
nrow(v)
#> [1] 10
class(v)
#> [1] "sf"         "data.frame"

# create a collection in your MongoDB instancee
vancouver = mongo("vancouver")
# clear it if has been run before
vancouver$drop()
vancouver$count()
#> [1] 0

Quietly populate the collection

library(geocoder)
gc_import_sf(v, collection = "vancounver")
#> Writing '10' documents...
#> There are '60' documents in collection: vancounver

Now lets query the collection

vancouver$find('{}', limit = 1)
#> data frame with 0 columns and 0 rows
vancouver$index((add = '{"geometry" : "2dsphere"}'))
#>   v key._id key.geometry              name             ns
#> 1 2       1         <NA>              _id_ test.vancouver
#> 2 2      NA     2dsphere geometry_2dsphere test.vancouver
#>   2dsphereIndexVersion
#> 1                   NA
#> 2                    3

# define a querry as mongolite does not accept somee queries
# a little cheating with copy and paste of one lon lat from the polygons in the data
qry <- '[
  {
    "$geoNear" : { 
      "near" : { "type" : "Point", "coordinates" : [ -123.1107886, 49.2718859 ] },
      "distanceField" : "dist.calculated",
      "maxDistance" : 3000,
      "spherical" : true
    }
  }
] '
r = vancouver$aggregate(pipeline = qry)
nrow(r)
#> [1] 0
# changing the maxDistance will increase the number returned
qry = gsub(x = qry, pattern = "3000", replacement = "6000")
r = vancouver$aggregate(pipeline = qry)
nrow(r)
#> [1] 0

development

To run all the tests, you will require mongo to run. Once you are sure that you have an instance (for example a container running as outliend above), you will need a flag in your R environment GEOCODER_MONGODB_AVAILABLE=true. You can use R package usethis::edit_r_environ().

But the steps should be straightforward for an R developer:

• clone the repo

• from RStudio (cmd/ctrl + T)

• OR from R console:

  • just run testthat::test_dir("pathToRepo/tests/testthat")

library("geocoder")
testthat::test_dir("tests/testthat")
#> ✓ |  OK F W S | Context
#> 
⠏ |   0       | gc-find
⠋ |   1       | gc-find
✓ |   4       | gc-find [1.4 s]
#> 
⠏ |   0       | gc-import
⠼ |   5       | gc-import
✓ |   6       | gc-import [1.2 s]
#> 
⠏ |   0       | gc-import-sf
✓ |   1       | gc-import-sf
#> 
⠏ |   0       | gc-setup
⠼ |   5       | gc-setup
✓ |   6       | gc-setup [0.7 s]
#> 
⠏ |   0       | gc-query
✓ |  23       | gc-query
#> 
#> ══ Results ═══════════════════════════════════════════════════════════════════
#> Duration: 3.7 s
#> 
#> OK:       40
#> Failed:   0
#> Warnings: 0
#> Skipped:  0

Acknowledgement

This work is funded under the Turing GeoVisualization Engine project.

Related work

Considered Mapit project, it is great for people who would like to have a web interface, for the use of this package, there is the underlying work of an R interface to MongoDB in mongolite and for this use casee, it might be better to use Mongo over PostgreSQL.

The list of UK geocodes, looks like before the ONS open geography portal, was compiled here.