[R] Wiki/revision control to management of CRAN package repository
Etienne Low-Décarie
etienne.low-decarie at mail.mcgill.ca
Mon Aug 22 06:02:57 CEST 2011
I propose the following humbly, with little know how as to how to implement, and realize it may have been proposed many times. It is just something I had on my mind.
Would it be possible/desirable to have the whole CRAN package repository accessible through a public wiki, forge or version control interface (ideally a fusion of the wiki and forge approach)?
It appears it would be a first for a software repository.
CRAN package repository is becoming a jungle of R code and may do well with currating and editorial effort. This can not/should not be the task of a single person or small group of people. Using a crowd sourced method by implementing a wiki approach to the CRAN package repository would allow for the rapid editing, sorting and improvement of this impressive and precious resource, while also improving the accessibility, visibility and quality of individual packages. It would also bind the
For example, such an interface would allow the cleaning up of the repository through the use of tagging of packages, using similar approaches to the wikipedia project (http://en.wikipedia.org/wiki/Wikipedia:Template_messages#Cleanup).
Such a tagging approach could be used within existing vcs, if the repository was migrated/mirrored within one of these systems.
Packages could be marked using tags for all following actions prior to implementing the action. Actions could be undertaken directly by package users after a delay or discussion.
Packages management/editorial effort:
-Merging/
-Combining packages that have:
-Large overlap in functionality
-Are largely interdependant
-Are only minor extensions of another package
-…
-Split/fork
-Subdividing behemoth packages into smaller packages with more specific tasks
-Categorization
-Packages could be sorted by use, improvement of Task View
-Tags, keywords could be added to packages for searching
-Packages could be placed in a hierarchy, not only by true dependance and reverse dependance, but also by logical dependance/reverse dependance
-ie. which package should probably be used with which package, an improvement on the see also help section
-Deletion
-Marking/tagging
-a stub/prototype
-broken
Package improvements
-Improving help files
-Adding functions
-Adding examples
-Requiring, improving or adding references
-References to the theory or approach used...
-A section could include a list of articles making use of the package, with package users encourage to enter this information
-This would allow package author recognition and allow a package impact factor
-Adding key words for indexing and searching
-Function improvement
-Adding compatibility with other packages/formats (including when merging packages)
-Speed improvements
Discussion
-On package improvements, management steps directly attached to the package
-Help discussion
These actions would be reversible, possibly with veto power from the author of the package.
Links:
http://www.rforge.net/
http://sourceforge.net/
http://channel9.msdn.com/Forums/Coffeehouse/174561-Coding-Wiki
http://code.google.com/p/mcover/
http://www.tigris.org/
This is just an idea I had on my mind.
Thank you
More information about the R-help
mailing list