Package development process: Difference between revisions

Revision as of 00:23, 9 January 2013

Revision as of 21:29, 19 July 2016

Line 13:

For example, a [[Python (programming language)|Python]] package combines documentation, code and initial set up and possibly examples that could be used as [[unit testing|unit tests]] in a single file with a "py" extension.

By contrast, an [[R (programming language)|R]] package has documentation with examples in files separate from the code, possibly bundled with other material such as sample data sets and introductory ''vignettes''. The source code for an R package is contained in a directory with a master "description" file and separate subdirectories for documentation, code, optional data sets suits for unit or regression testing, and perhaps others.<ref>{{cite book|title=Writing R Extensions |authorlink=<!-- R Development Core Team --> |url =httphttps://cran.r-project.org/doc/manuals/R-exts.html }}</ref> A formal package compilation process

<ref>{{cite news

| last = Leisch | first = Friedrich

| title = Creating R Packages: A Tutorial

| url =httphttps://cran.r-project.org/doc/contrib/Leisch-CreatingPackages.pdf }}

</ref>

<ref>{{cite news

| last1 = Graves | first1 = Spencer B. | last2 = Dorai-Raj | first2=Sundar

| title = Creating R Packages, Using CRAN, R-Forge, And Local R Archive Networks And Subversion (SVN) Repositories

| url =httphttps://cran.r-project.org/doc/contrib/Graves+DoraiRaj-RPackageDevelopment.pdf}}

</ref> checks for errors of various types. This includes checking for syntax errors on both the documentation markup language and the code as well as comparing the arguments between documentation and code. Examples in the documentation are tested and produce error messages if they fail. This can be used as a primitive form of [[unit testing]]; more formal unit tests and [[regression testing]] can be included. This can improve software development productivity by making it easier to find bugs as the code is being developed. In addition, the documentation makes it easier to share code with others. It also makes it easier for a developer to use code written months or even years earlier. Routine checks are made of packages contributed to the R software repository CRAN (the Comprehensive R Archive Network) <ref>{{cite web

| title=The Comprehensive R Archive Network | url =http://cran.r-project.org}}</ref> and under development in companion open-source collaborative development web site, R-Forge.<ref>{{cite web | title = R-Forge | url=http://r-forge.r-project.org}}</ref> These checks compile the packages repeatedly on different platforms under different versions of the core R language. The results are made available to package maintainers. In this way, package contributors become aware of problems they might otherwise never encounter themselves, because they otherwise would not have easy access to those alternative test results.

An interesting research question would be to compare the quality of contributions to different [[software repository|software repositories]] and try to relate that to features of the language and accompanying package development process. This could include trying to compare the rate of growth of contributed software to the degree of formality and enforcement of standards for documentation, testing, and coding.