Using config with Posit Connect

The purpose of the config package is to respond to environment variables that could be set to different values on different machines. Specifically, config reacts to the value of R_CONFIG_ACTIVE, and this allows you to set configurations for different machines, e.g. in dev or production.

If you deploy your code to a Posit Connect instance, then you can use the fact that Connect sets the value of R_CONFIG_ACTIVE to rsconnect.

This means you can create a configuration file using the value rsconnect, for example:

config.yml

default:
  trials: 5
  dataset: "data-sampled.csv"
  
rsconnect:
  trials: 30
  dataset: "data.csv"

Note: this value can be configured by your system administrator, so check your own system configuration! See https://docs.posit.co/connect/admin/appendix/configuration/#R.ConfigActive for detail.

Using config for staging and prod on multiple servers

Many customers of Posit Connect will have separate Connect instances for the purposes of separating staging and production.

In this case, the system administrator needs to configure the R_CONFIG_ACTIVE environment variable to have different values on each machine. For example:

On the staging server: rsconnect-staging
On the production server: rsconnect-prod

Using these example values, you could configure the config.yml file to look like this:

config.yml

default:
  trials: 5
  dataset: "data-sampled.csv"

rsconnect:
  trials: 30
  
rsconnect-staging:
  inherits: rsconnect
  dataset: "data-staging.csv"
  
rsconnect-prod:
  inherits: rsconnect
  dataset: "data.csv"

Note that this configuration uses the inheritance of config yaml files, by having a common configuration for rsconnect and staging and prod configurations that inherit from the rsconnect configuration.

Using config for staging and prod on a single server

You may have the use case that you use the same Connect instance to host two instances of your app, for staging (test) and production.

The May 2023 release of Connect added support for allowing users to override the value of R_CONFIG_ACTIVE.

If using a version of Connect newer than 2023.05.0, you can change the value of R_CONFIG_ACTIVE using the envVars argument of rsconnect::deployApp() or the Connect Server API.

If using a version of Connect less than or equal to 2023.05.0, you can use config to set different values for staging and prod, by changing the environment variable that config::get() looks at.

Specifically, you must specify a different config argument to config::get(). For example, you can create a new environment variable for your Connect app, called R_CONFIG_ACTIVE_APP, and then use config::get() like this:

config::get("trials", config = Sys.getenv("R_CONFIG_ACTIVE_APP"))
#> [1] 5

Again, this example utilizes the inheritance of config yaml files, by having a common configuration for rsconnect and staging and prod configurations that inherit from the rsconnect configuration.

In such a case, you may have a config.yml file like this:

config.yml

default:
  trials: 5
  dataset: "data-sampled.csv"

rsconnect:
  trials: 30
  
staging:
  inherits: rsconnect
  dataset: "data-staging.csv"
  
prod:
  inherits: rsconnect
  dataset: "data.csv"

If you configure the environment variable correctly, you should get the appropriate dataset values for staging (data-staging.csv) and prod (data.csv):

config::get("dataset", config = "staging")
#> [1] "data-staging.csv"
config::get("dataset", config = "prod")
#> [1] "data.csv"

A utility function that might be helpful

If you are in the situation where you wish to use a different environment variable, you can potentially save a few keystrokes by defining a utility function that wraps around config::get(). Such a utility simply passes a different value for the config argument, like this:

cfg_get_app <- function (
  value = NULL, 
  config = Sys.getenv("R_CONFIG_ACTIVE_APP", "default"), 
  ...
) {
  config::get(value = value, config = config, ...)
}

You can then call this function from your app:

cfg_get_app("value")