R evolves every year. Each update
brings many bug fixes and performance improvements, but also sometimes
changes the default value of some parameters (such as switching from
stringsAsFactors = TRUE to FALSE in R
4.0.0).
While there is benefit in supporting older R versions, doing so adds maintenance burden, especially in a large ecosystem like easystats that contains 10 packages that are supposed to seamlessly work together. Therefore, easystats is supported (meaning each package is thoroughly tested) on the current R version, the future one (R-devel) and the previous five versions.
The current planned release schedule for large R updates is once a year in Spring. Accordingly, this table shows the corresponding latest R version and the oldest supported version:
| Year | Latest R version | Oldest supported R version |
|---|---|---|
| 2023 | 4.3 | 3.5 |
| 2024 | 4.4 | 3.6 |
| 2025 | 4.5 | 4.0 |
| 2026 | 4.6 | 4.1 |
| 2027 | 4.7 | 4.2 |
| 2028 | 4.8 | 4.3 |
(Of course, if the R release schedule changes or R releases a “5.0” version, the numbers in the table will change.)
The implications are simple:
If you are a user and your R version is below the oldest
supported one, you must update it to use the easystats
packages. Alternatively, you can also download a version of the package
that is compatible with your R version from the CRAN archive
and then install it locally. Finally, the package {remotes}
gives a way to install an old version of a package. For instance,
remotes::install_version("insight", version = "0.10.0")
installs a version of {insight} that is compatible with R
3.5.0.
If you are a developer and your package depends on one of the
easystats packages, the oldest version that your
package supports corresponds to the oldest version supported by
easystats. Note that this should be reflected in your
DESCRIPTION file with Depends: R (>= x.y.z).