rOpenSci: The pkgcheck package

Check whether a package is ready for submission to rOpenSci’s peer review system. The primary function collates the output of goodpractice, including R CMD check results, a number of statistics via the pkgstats package, and checks for package structure expected for rOpenSci submissions. The output of this function immediately indicates whether or not a package is “Ready to Submit”.

Installation

The easiest way to install this package is via the associated r-universe. As shown there, simply enable the universe with

options (repos = c (
    ropenscireviewtools = "https://ropensci-review-tools.r-universe.dev",
    CRAN = "https://cloud.r-project.org"
))

And then install the usual way with,

install.packages ("pkgcheck")

Alternatively, the package can be installed by running one of the following lines:

remotes::install_github ("ropensci-review-tools/pkgcheck")
pak::pkg_install ("ropensci-review-tools/pkgcheck")

The package can then loaded for use with

library (pkgcheck)

Setup

The pkgstats package also requires the system libraries ctags and GNU global to be installed. Procedures to install these libraries on various operating systems are described in the pkgstats README. This package also uses the GitHub GraphQL API which requires a local GitHub token to be stored with an unambiguous name including GITHUB, such as GITHUB_TOKEN (recommended), or GITHUB_PAT (for Personal Authorization Token). This can be obtained from GitHub (via your user settings), and stored using

Sys.setenv ("GITHUB_TOKEN" = "<my_token>")

This can also be set permanently by putting this line in your ~/.Renviron file (or creating this if it does not yet exist). Once pkgstats has been successfully installed, the pkgcheck package can then be loaded via a library call:

library (pkgcheck)

Usage

The package primarily has one function, pkgcheck, which accepts the single argument, path, specifying the local location of a git repository to be analysed. The following code generates a reproducible report by first downloading a local clone of a repository called srr-demo, which contains the skeleton of an srr (Software Review Roclets) package, generated with the srr_stats_pkg_skeleton() function:

mydir <- file.path (tempdir (), "srr-demo")
gert::git_clone ("https://github.com/mpadge/srr-demo", path = mydir)
x <- pkgcheck (mydir)

That object has default print and summary methods. The latter can be used to simply check whether a package is ready for submission:

summary (x)
## 
## ── demo 0.0.0.9000 ───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
## 
## ✔ Package name is available
## ✖ does not have a 'CITATION' file.
## ✖ does not have a 'codemeta.json' file.
## ✖ does not have a 'contributing' file.
## ✔ uses 'roxygen2'.
## ✔ 'DESCRIPTION' has a URL field.
## ✖ 'DESCRIPTION' does not have a BugReports field.
## ✖ Package has at no HTML vignettes
## ✔ All functions have examples.
## ✖ Package has no continuous integration checks.
## ✖ Package coverage is 0% (should be at least 75%).
## ✔ R CMD check found no errors.
## ✔ R CMD check found no warnings.
## 
## ℹ Current status:
## ✖ This package is not ready to be submitted.
## 

A package may only be submitted when the summary contains all ticks and no cross symbols. (These symbols are colour-coded with green ticks and red crosses when generated in a terminal; GitHub markdown only renders them in black-and-white.) The object returned from the pkgcheck function is a complex nested list with around a dozen primary components. Full information can be obtained by simply calling the default print method by typing the object name (x).

What is checked?

Summary of Check Results

Calling summary() on the object returned by the pkgcheck() function will generate a checklist like that shown above. This checklist will also be automatically generated when a package is first submitted to rOpenSci, and is used by the editors to assess whether to process a submission. Authors must ensure prior to submission that there are no red crosses in the resultant list. (In the unlikely circumstances that a package is unable to pass particular checks, explanations should be given upon submission about why those checks fail, and why review may proceed in spite of such failures.)

The full list of checks which packages are expected to pass currently includes:

1. Package must use roxygen2 for documentation.
2. Package must have a contributing.md file.
3. Package must have a CITATION file in the inst directory.
4. Package must have a codemeta.json file.
5. All exported functions must include examples in their documentation.
6. Left-assign operators must be used consistently throughout all code (so either all = or all <-, but not a mixture of both).
7. Package DESCRIPTION file must have a “URL” field.
8. Package DESCRIPTION file must have a “BugReports” field.
9. Package name must be available (or package must already be) on CRAN.
10. Package must have continuous integration tests.
11. Package must have test coverage of at least 75%.
12. No function should have a cyclomatic complexity of 15 or greater.
13. R CMD check (implemented via the rcmdcheck package) must generate no warnings or errors.
14. All statistical standards must be documented, as confirmed by the srr::srr_pre_submit() function.

Several of these are by default only shown when they fail; absence from a resultant checklist may be interpreted to indicate successful checks.

Detailed Check Results

Full details of check results can be seen by print-ing the object returned by the pkgcheck() function (or just by typing the name of this object in the console.) This object is itself a list including the following items:

names (x)

## [1] "pkg"          "info"         "checks"       "meta"         "goodpractice"

The first four of these contain information on the package. The remainder include:

• summary containing summary statistics on package structure (such as lines of code, and numbers of files and functions).
• git containing information and statistics on the git repository associated with a package (if any), including date of first commit, total number of commits and of committers.
• srr summarising information on statistical packages generated by the srr package, including a link to a locally-generated HTML report on standards compliance.
• file_list as a list of binary flags denoting presence of absence of all files required for an R package to be submitted to rOpenSci
• fns_have_exs with a list of any functions which do not have documented examples
• left_assigns with a logical value indicating whether or not global assignment operators (<<-) are used; and a tally of the two types of left-assignment operators (<- and =). One one of these should be non-zero, reflecting consistent usage of the same type.
• pkgstats containing several statistics of package structure, along with associated percentiles assessed against the entire distribution of all current CRAN packages.
• network_file with the path to a local HTML file containing a .visjs representation of the network of relationships between objects (such as functions) within a package, within and between each computer language used in the package.
• gp containing the output of the goodpractice pacakge, itself including results from:
  • the rcmdcheck pacakge for running R CMD check
  • the covr package for assessing code coverage
  • the cyclocomp package for quantifying the cyclomatic complexity of package functions
  • the desc package for analysing the structure of DESCRIPTION files
  • the lintr package
  • additional checks applied by the goodpractice pacakge

Note that results from lintr package are not reported in the check summary, and that lintr results are reported only in the detailed results, and have no influence on whether a package passes the summary checks.

HTML-formatted check results

Printing pkgcheck results to screen is nice, but sometimes it’s even better to have a nicely formatted full-screen representation of check results. The package includes an additional function, checks_to_markdown(), with a parameter, render, which can be set to TRUE to automatically render a HTML-formatted representation of the check results, and open it in a browser. The formatting differs only slightly from the terminal output, mostly through the components of goodpractice being divided into distinct headings, with explicit statements in cases where components pass all checks (the default screen output inherits directly from that package, and only reports components which do not pass all checks).

This checks_to_markdown() function returns the report in markdown format, suitable for pasting directly into a GitHub issue, or other markdown-compatible place. (The clipr package can be used to copy this directly to your local clipboard with write_clip(md), where md is the output of checks_to_markdown().)

Caching and running pkgcheck in the background

Running the pgkcheck function can be time-consuming, primarily because the goodpractice component runs both a full R CMD check, and calculates code coverage of all tests. To avoid re-generating these results each time, the package saves previous reports to a local cache, in a pkgcheck subdirectory of the location determined by

rappdirs::user_cache_dir ()

As explained in the help file for that function, these locations are:

System	location
Mac OS X	~/Library/Caches/pkgcheck
Linux	~/.cache/pkgcheck
Win XP	C:\\Documents and Settings\\<username>\\Local Settings\\Application Data\\<AppAuthor>\\pkgcheck\\Cache
Vista	C:\\Users\\<username>\\AppData\\Local\\<AppAuthor>\\pkgcheck\\Cache

You may manually erase the contents of this pkgcheck subdirectory at any time at no risk beyond additional time required to re-generate contents. This default location may also be over-ridden by setting an environmental variable named PKGCHECK_CACHE_DIR. By default checks presume packages use git for version control, with checks updated only when code is updated via git commit. Checks for packages that do not use git are updated when any files are modified.

The first time pkgcheck() is applied to a package, the checks will be stored in the cache directory. Calling that function a second time will then load the cached results, and so enable checks to be returned much faster. For code which is frequently updated, such as for packages working on the final stages prior to submission, it may still be necessary to repeatedly call pkgcheck() after each modification, a step which may still be inconveniently time-consuming. To facilitate frequent re-checking, the package also has a pkgcheck_bg() function which is effectively identical to the main pkgcheck() function, except it runs in the background, enabling you to continue coding while checks are running.

The pkgcheck_bg() function returns a handle to the callr::r_bg() process in which the checks are running. Typing the name of the returned object will immediately indicate whether the checks are still running, or whether they have finished. That handle is itself an R6 object with a number of methods, notably including get_result() which can be used to access the checks once the process has finished. Alternatively, as soon as the background process, the normal (foreground) pkgcheck() function may be called to quickly re-load the cached results.

Prior Work

The checklist package for “checking packages and R code”.

Code of Conduct

Please note that this package is released with a Contributor Code of Conduct. By contributing to this project, you agree to abide by its terms.