Geocoding is the process of converting place names or addresses into geographic coordinates – like latitude and longitude – or vice versa. With {opencage} you can geocode using the OpenCage API, either from place name to longitude and latitude (forward geocoding) or from longitude and latitude to the name and address of a location (reverse geocoding).

This vignette covers the setup process for working with {opencage} and basic workflows for both forward and reverse geocoding. Make sure to also check out “Customise your query” if you want a deeper dive into customizing the various parameters available through the OpenCage API. The “Output options” vignette shows additional workflows by modifying the form in which the geocoding results are returned.

Setup

Before you can use the {opencage} package and query the OpenCage API you need to first register with OpenCage. Additionally, you may want to set a rate limit (if you have a paid OpenCage plan), and you might want to prevent OpenCage from storing the content of your queries. In other words, you need to setup {opencage}, so let’s go through the process.

Authentication

To use the package and authenticate yourself with the OpenCage API, you will need to register at opencagedata.com/users/sign_up to get an API key. The “Free Trial” plan provides up to 2,500 API requests a day. There are paid plans available, if you need to run more API requests. After you have registered, you can generate an API key with the OpenCage dashboard.

Now we need to ensure that the functions in {opencage} can access your API key. {opencage} will conveniently retrieve your API key if it is saved in the environment variable "OPENCAGE_KEY". If it is not, oc_config() will help to set that environment variable.

Do not pass the key directly as a parameter to the function. Doing so risks exposing your API key via your script or your history. There are three safer ways to set your API key instead:

  1. Save your API key as an environment variable in .Renviron as described in What They Forgot to Teach You About R or Efficient R Programming. From there it will be fetched by all functions that call the OpenCage API. You do not even have to call oc_config() to set your key; you can start geocoding right away. If you have the {usethis} package installed, you can edit your .Renviron with usethis::edit_r_environ(). We strongly recommend storing your API key in the user-level .Renviron, as opposed to the project-level. This makes it less likely you will share sensitive information by mistake.

  2. If you use a package like {keyring} to store your credentials, you can safely pass your key in a script with a function call like oc_config(key = keyring::key_get("opencage")).

  3. If you call oc_config() in an interactive() session and the OPENCAGE_KEY environment variable is not set, it will prompt you to enter the key in the console.

Whatever method you choose, keep your API key secret. OpenCage also features best practices for keeping your API key safe.

Rate limit

A rate limit is used to control the rate of requests sent, so legitimate requests do not lead to an unintended Denial of Service attack. The rate limit allowed by the API depends on the OpenCage plan you have and ranges from 1 request/sec for the “Free Trial” plan to 15 requests/sec for the “Medium” or “Large” plans. See opencagedata.com/pricing for details and up-to-date information.

If you have a “Free Trial” account with OpenCage, you can skip to the next section, because the rate limit is already set correctly for you at 1 request/sec.

If you have a paid account, you can set the rate limit for the active R session with oc_config(rate_sec = n) where n is the appropriate rate limit. You can set the rate limit persistently across sessions by setting an oc_rate_sec option in your .Rprofile. If you have the {usethis} package installed, you can edit your .Rprofile with usethis::edit_r_profile().

Privacy

By default, OpenCage will store your queries on its server logs and will cache the forward geocoding requests on their side. They do this to speed up response times and to be able to debug errors and improve their service. Logs are automatically deleted after six months according to OpenCage’s page on data protection and GDPR.

If you have concerns about privacy and want OpenCage to have no record of your query, i.e. the place name or latitude and longitude coordinates you want to geocode, you can set a no_record parameter to TRUE, which tells the API to not log nor cache the queries. OpenCage still records that you made a request, but not the specific queries you made.

oc_config(no_record = TRUE) sets an oc_no_record option for the active R session, so it will be used for all subsequent OpenCage queries. You can set the oc_no_record option persistently across sessions in your .Rprofile.

For more information on OpenCage’s policies on privacy and data protection see the Legal section in their FAQs, their GDPR page, and, for the no_record parameter specifically, see the relevant blog post.

For increased privacy, {opencage} sets no_record to TRUE, by default. Please note, however, that {opencage} always caches the data it receives from the OpenCage API locally, but only for as long as your R session is alive (see below).

(Don’t) show API key

oc_config() has another argument, show_key. This is only used for debugging and we will explain it in more detail in vignette("output_options"). For now suffice it to say that your OpenCage API key will not be shown in any {opencage} output, unless you change this setting.

Altogether now

In sum, if you want to set your API key with {keyring}, set the rate limit to 10 (only do this if you have a paid account, please!), and do not want OpenCage to have records of your queries, you would configure {opencage} for the active session like this:

library("opencage")
oc_config(
  key = keyring::key_get("opencage"),
  rate_sec = 10,
  no_record = TRUE
)

Forward geocoding

Now you can start to geocode. Forward geocoding is from location name(s) to latitude and longitude tuple(s).

oc_forward_df(placename = "Sarzeau")
#> # A tibble: 1 x 4
#>   placename oc_lat oc_lng oc_formatted         
#>   <chr>      <dbl>  <dbl> <chr>                
#> 1 Sarzeau     47.5  -2.76 56370 Sarzeau, France

All geocoding functions are vectorised, i.e. you can geocode multiple locations with one function call. Note that behind the scenes the requests are still sent to the API one-by-one.

opera <- c("Palacio de Bellas Artes", "Scala", "Sydney Opera House")
oc_forward_df(placename = opera)
#> # A tibble: 3 x 4
#>   placename               oc_lat oc_lng oc_formatted                                                                     
#>   <chr>                    <dbl>  <dbl> <chr>                                                                            
#> 1 Palacio de Bellas Artes   19.4 -99.1  Palacio de Bellas Artes, Avenida Juárez, Centro Urbano, 06050 Mexico City, Mexico
#> 2 Scala                     45.5   9.19 Teatro alla Scala, Largo Antonio Ghiringhelli, Milan Milan, Italy                
#> 3 Sydney Opera House       -33.9 151.   Sydney Opera House, 2 Macquarie Street, Sydney NSW 2000, Australia

By default, oc_forward_df() only returns three results columns: oc_lat (for latitude), oc_lon (for longitude), and oc_formatted (the formatted address). As you can see, the results columns are all prefixed with oc_. If you specify oc_forward_df(output = all), you will receive all result columns, which are often quite extensive. Which columns you receive exactly depends on the information OpenCage returns for each specific request.

oc_forward_df(placename = opera, output = "all")
#> # A tibble: 3 x 30
#>   placename  oc_lat oc_lng oc_confidence oc_formatted  oc_northeast_lat oc_northeast_lng oc_southwest_lat oc_southwest_lng oc_iso_3166_1_a~
#>   <chr>       <dbl>  <dbl>         <int> <chr>                    <dbl>            <dbl>            <dbl>            <dbl> <chr>           
#> 1 Palacio d~   19.4 -99.1              9 Palacio de B~             19.4           -99.1              19.4           -99.1  MX              
#> 2 Scala        45.5   9.19             9 Teatro alla ~             45.5             9.19             45.5             9.19 IT              
#> 3 Sydney Op~  -33.9 151.               9 Sydney Opera~            -33.9           151.              -33.9           151.   AU              
#> # ... with 20 more variables: oc_iso_3166_1_alpha_3 <chr>, oc_category <chr>, oc_type <chr>, oc_city <chr>, oc_continent <chr>,
#> #   oc_country <chr>, oc_country_code <chr>, oc_museum <chr>, oc_neighbourhood <chr>, oc_postcode <chr>, oc_road <chr>,
#> #   oc_attraction <chr>, oc_county <chr>, oc_political_union <chr>, oc_state <chr>, oc_suburb <chr>, oc_arts_centre <chr>,
#> #   oc_house_number <chr>, oc_municipality <chr>, oc_state_code <chr>

You can also pass a data frame to oc_forward_df(). By default the results columns are added to the input data frame, which is useful for keeping information associated with the place names that are in separate columns. If you want a data frame with only the geocoding results, set bind_cols = FALSE.

concert_df <-
  data.frame(location = c("Elbphilharmonie", "Concertgebouw", "Suntory Hall"))
oc_forward_df(data = concert_df, placename = location)
#> # A tibble: 3 x 4
#>   location        oc_lat oc_lng oc_formatted                                                                 
#>   <chr>            <dbl>  <dbl> <chr>                                                                        
#> 1 Elbphilharmonie   53.5   9.98 Elbe Philharmonic Hall, Platz der Deutschen Einheit 1, 20457 Hamburg, Germany
#> 2 Concertgebouw     52.4   4.88 Concertgebouw, Concertgebouwplein 2, 1071 LN Amsterdam, Netherlands          
#> 3 Suntory Hall      35.7 140.   Suntory Hall, Karayan Plaza, Azabu, Minato, Japan

You can use it in a piped workflow as well.

library(dplyr, warn.conflicts = FALSE)
concert_df %>% oc_forward_df(location)
#> # A tibble: 3 x 4
#>   location        oc_lat oc_lng oc_formatted                                                                 
#>   <chr>            <dbl>  <dbl> <chr>                                                                        
#> 1 Elbphilharmonie   53.5   9.98 Elbe Philharmonic Hall, Platz der Deutschen Einheit 1, 20457 Hamburg, Germany
#> 2 Concertgebouw     52.4   4.88 Concertgebouw, Concertgebouwplein 2, 1071 LN Amsterdam, Netherlands          
#> 3 Suntory Hall      35.7 140.   Suntory Hall, Karayan Plaza, Azabu, Minato, Japan

Reverse geocoding

Reverse geocoding works in the opposite direction of forward geocoding: from a pair of coordinates to the name and address most appropriate for the coordinates.

oc_reverse_df(latitude = 51.5034070, longitude = -0.1275920)
#> # A tibble: 1 x 3
#>   latitude longitude oc_formatted                                                                            
#>      <dbl>     <dbl> <chr>                                                                                   
#> 1     51.5    -0.128 Prime Minister’s Office, Westminster, 10 Downing Street, London SW1A 2AA, United Kingdom

Note that all coordinates sent to the OpenCage API must adhere to the WGS 84 (also known as EPSG:4326) coordinate reference system in decimal format. This is the coordinate reference system used by the Global Positioning System. There is usually no reason to send more than six or seven digits past the decimal. Any further precision gets to the level of a centimeter.

Like oc_forward_df(), oc_reverse_df() is vectorised, can work with numeric vectors and data frames, supports the output = "all" argument and can be used with the {magrittr} pipe.

OpenCage only returns at most one result per reverse geocoding request.

Caching

OpenCage allows and supports caching. To minimize the number of requests sent to the API {opencage} uses {memoise} to cache results inside the active R session.

system.time(oc_reverse(latitude = 10, longitude = 10))
#>    user  system elapsed 
#>    0.00    0.00    0.86

system.time(oc_reverse(latitude = 10, longitude = 10))
#>    user  system elapsed 
#>    0.01    0.00    0.01

To clear the cache of all results either start a new R session or call oc_clear_cache().

oc_clear_cache()
#> [1] TRUE

system.time(oc_reverse(latitude = 10, longitude = 10))
#>    user  system elapsed 
#>    0.03    0.00    0.63

As you probably know, cache invalidation is one of the harder things to do in computer science. Therefore {opencage} only supports invalidating the whole cache and not individual records at the moment.

The underlying data at OpenCage is updated daily.

Further information

OpenCage supports a lot of parameters to either target your search area more specifically or to specify what additional information you need. See the “Customise your query” vignette for details.

Besides oc_forward_df() and oc_reverse_df(), which always return a single tibble, {opencage} has two sibling functions — oc_forward() and oc_reverse() — which can be used to return types of output. Depending on what you specify as the return parameter, oc_forward() and oc_reverse() will return either a list of tibbles (df_list, the default), JSON lists (json_list), GeoJSON lists (geojson_list), or the URL with which the API would be called (url_only). Learn more in the “Output options” vignette.

Please report any issues or bugs on our GitHub repository and post questions on discuss.ropensci.org.