I'd like to suggest rOpenSci as a potential source for reliable packages we can depend on. We also have the option to add dependencies as SUGGESTS packages for optional functionality that might only be needed by a subset of users (e.g. If a package is only using one function from a large package it might make sense to implement single function and remove the dependency. I like the idea trying to limit dependencies and analyzing our dependencies. A significant benefit of taking on dependencies is that we save ourselves the work of implementing and maintaining some functionality. This conversation might be better framed as a tradeoff between "build yourself" vs "borrow and share", a decision that Stephen Wali outlined in the open source workshop earlier this year. Once criteria have been established, we could form a list of 'approved' packages we can depend on. For example, most of tidyVerse is probably reliable. Criteria could be defined for what are considered 'reliable' dependencies, based for example on age and development state of the package, and an assessment of the community behind the package. If possible, we'd like to restrict our set of dependencies to those that we believe are reliable. Therefore, one effort could focus on analyzing the list of dependencies per HADES package, and the overlap between dependencies. If possible, different HADES packages should have overlapping dependencies, to minimize the overall dependencies of HADES. Some HADES packages have a lot of dependencies, and one reason is that some direct dependencies have a large list of dependencies themselves. We therefore would like to minimize our number of dependencies. by renv), and tend to be the cause when things break.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |