roam: Remote Objects with Active-Binding Magic
Abstract
In R, data are most naturally distributed as objects bundled inside packages, but they comes with practical limits. Package size restriction of 5 MB make it difficult to include larger datasets, and updating data typically requires updating the entire package. The ‘roam’ package simplifies the creation of R objects that behave like regular data but are stored remotely and loaded when needed. This allows developers to work with larger datasets without inflating package size, while also enabling data to be updated independently of package releases. It also supports versioning, making it easier to manage and reproduce different versions of a dataset. In this talk, I will introduce the main idea behind ‘roam’, show how it can be used in practice, and briefly explain how it works behind the scenes.