Posts R Packages
Post
Cancel

R Packages

Feb 19, 2020 2020-02-19T00:00:00-06:00 on Programming

Create an R package

References:

Create from command line

  1. Create a folder named ‘greetings’: mkdir greetings and cd greetings.
  2. Write some R files in the folder, for example, hello.R and goodbye.R.(No C files.)
  3. Open R script: r --vanilla, run package.skeleton(name="greetings", code_files=c("hello.R", "goodbye.R")). Quit R, you will see a folder named greetings.
  4. Change directory into greetings: cd greetings. Edit DESCRIPTION file, add Encoding: UTF-8 to the end.
  5. Remove NAMESPACE file: rm NAMESPACE. Open R script: r --vanilla, run devtools::document() to update the NAMESPACE file.
  6. Edit .Rd files in the folder man, make sure title is edited. Remove the package rd file:rm greetings-package.Rd. If examples section is modified, then we need to add export(func) to NAMESPACE.
  7. Check the functions are working and the pacakges are loaded correctly. Change directory to the original greetings folder, the one with a subdirectory greetings and the R files. Open R script: r --vanilla, run
    1
2
3
4
5

     library(devtools)
 library(roxygen2)
 my.Rpackage <- as.package("greetings")
 load_all(my.Rpackage)
 document(my.Rpackage)
  8. Build, install, and check the package.
    1
2
3

     R CMD build greetings
 R CMD INSTALL greetings_1.0.tar.gz
 R CMD check --as-cran greetings_1.0.tar.gz
  9. Load the library in R script and try some functions.

Edit .Rd files

  1. Feel free to remove the .Rd files if you don’t plan to export the corresponding functions.
  2. Share one .Rd file. For example, use xkcd.Rd for dxkcd.Rd, pxkcd.Rd, qxkcd.Rd, rxkcd.Rd.
    • Rename dxkcd.Rd with xkcd.Rd, and remove the other Rd files.
    • Use alias to relate the other functions to this .Rd file.
  3. Check Writing R documentation files detaild Rd file rendering.
  4. View .Rd files conveniently. Go to the directory of the Rd file.
    1
2

     library(Rdpack)
 viewRd('./xkcd.Rd', type='html')

Do tests in the R package.

References:

  1. Get into the package directory, the one with NAMESPACE, with cd greetings. Open R script: r --vanilla.
    1

     devtools::use_testthat()

    or

    1
2

     library(devtools)
 use_testthat()

    This will:

    • Create a tests/testthat directory in the package directory.
    • Adds testthat to the Suggests field in the DESCRIPTION.
    • Creates a file tests/testthat.R that runs all your tests when R CMD check runs. Do not need to modify testthat.R.
  2. Move into the directory tests/testthat, and create test R files inside this directory. Each test R file serves for one single test.
    • Every test file’s name should start with test.
    • Need to include context() to describe the usage of this test file.
    • When testing the equivalence of numbers, control the output precision when running it in R by sprintf("%.20f", x), then paste the result to expect_equal().
    • For other test statements, check the reference for more detail.
    • For example, create a test file test_dxkcd.R to test for dxkcd function.
      1
2
3
4
5
6
7
8
9
10
11
12
13

        context("dxkcd")
  library(xkcd)

  test_that("the output is of the same length as max(len(y), len(sd.x))", {
  expect_equal(length(dxkcd(0.1, 1)), 1)
  expect_equal(length(dxkcd(0.1, c(1, 2, 3))), 3)
  expect_equal(length(dxkcd(c(0.1, 0.12), c(1, 1.5, 2, 2.5))), 4)
  expect_equal(length(dxkcd(c(0.1, 0.12, 0.13, 0.14), c(1, 1.5))), 4)
  })

  test_that("the output is correct within tolerance", {
      expect_equal(dxkcd(0.1, 1), 3.32703659106943439028)
  })
  3. Run the tests. Change directory to the package directory, the one containing NAMESPACE. Open R script: r --vanilla.
    1

     devtools::test()

    Correct mistakes refering to the prompt if there is any.

Commit the package to Github

  1. Remove the R files that are used to generate the package skeleton.
  2. Add .gitignore file to the repository you want to commit. Check my R.gitignore.
  3. Take the usual steps to commit.

R’s C Interface

References:

Rcpp

References:

Since there are so many subtle issue in Catalina, I turned to Ubuntu machine on Goolge Cloud. The installation of R follows the instruction How to Install R on Ubuntu 18.04.

sudo apt install libcurl4-openssl-dev sudo apt install libxml2-dev

Packages are installed as root sudo -i R, including:

  • Rcpp, devtools, bench, conformalInference,

Build R packages with Rstudio

About roxygen2 in Rstudio

  • There is an icon above the R file, called Code Tools. Put the cursor inside the function you want to write a document, click the icon, and choose Insert Roxygen Skeleton.
  • Add Roxygen: list(markdown = TRUE) to DESCRIPTION to enable markdown syntax in roxygen2.

About vignettes

  1. Build vignettes following the instructions in Writing vignettes.
  2. The default comment setting is "#>", and you can change it to ">" if you prefer. And also the collapse setting.
  3. html vignette saves memory, and this is why it is the default option. You can also choose to output to a pdf document by setting output: pdf_document.
  4. Ctrl + Shift + B will not generate vignettes. To build the package with vignettes, use command line R CMD build packagename and R CMD INSTALL tarballfile. I tried using devtools to sovle this problem, I ran devtools::build() and devtools::install(build_vignettes = TRUE). After doing this, we shall see the vignettes information by browseVignettes("packagename"). But everytime I do this, there would an error message when I try to see the help files. Command line is the best way for me so far.
  5. vignettes(package = "packagename"): show the vignette information.
  6. browseVignettes("packagename"): open the vignette files in a browser.

About datasets

  1. First, create a folder data/, and then save the dataset in .rda format: save(mydata, file="data/mydata.RData").
  2. Second, create a file mydata.R in the R/ folder, and write roxygen comments. Check Rd (documentation) tags and Including datasets for detailed instructions.
  3. Third, add LazyData: true to the DESCRIPTION file if it is not there.

Summary about steps after finishing editing files

1
2
3

library(devtools)
document()
test()

R CMD build …

R CMD INSTALL …

R CMD check …

1
2
3

library(package)
?function
browseVignettes("package")
Updated Feb 23, 2020 2020-02-23T15:14:05-06:00
R package
This post is licensed under CC BY 4.0

Cluster Notes

Python and R