In WdeNooy/UsingRTutorials: Provides learnr Tutorials for a Using R Course

knitr::opts_chunk$set(echo = TRUE)

Styles to be included at the top of the document

Code hint

More than one hint possible.

Markdown (text) hint

Only one hint possible.

__Hint:__ hinttext.

Programming Tip

__Programming Tip__ -

Exercise with result checker

gradethis::grade_result(
  pass_if(~ identical(.result, correct_result), "message"),
  fail_if(~ identical(.result, wrong_result), "message")
)

Exercise with conditions checker

# every pass_if condition must be satisfied
gradethis::grade_result_strict(
  pass_if(~ identical(.result, correct_result), "message"),
  fail_if(~ identical(.result, wrong_result), "message")
)

Exercise with code checker

__Hint:__ hinttext.

Or add , exercise.reveal_solution = FALSE to the solution chunk options:

"replace this by the correct code (pipe as one line)"

gradethis::grade_code(
  correct = "message", 
  incorrect = "message"
  )

Quiz

quiz(
  caption = "",
  question("Which package contains functions for installing other R packages?",
    answer("base"),
    answer("tools"),
    answer("utils", correct = TRUE),
    answer("codetools")
  ),
  question("Which of the R packages listed below are used to create plots?",
    answer("lattice", correct = TRUE),
    answer("tools"),
    answer("stats"),
    answer("grid", correct = TRUE)
  )
)

Installation

• Student must install the learnr (install.packages("learnr")) and gradethis (remotes::install_github("rstudio-education/gradethis", which will also install learnr) packages before a tutorial can be taken.
• Tutorials must be installed with devtools::install_github("WdeNooy/UsingRTutorials"). First install devtools package, which may require to update all other packages first.
• Updating the files on GitHub automatically updates the UsingRTutorials package: re-install/update it and you have the updated package.

Starting a tutorial

• console: learnr::run_tutorial("introduction", "UsingRTutorials"); opens in browser.
• Tutorials tab in RStudio: Press button Start Tutorial and, when started in the tab, open in new window (cannot be opened in a browser). NOTE: once the UsingRTutorials package is installed, the tutorials show up in the Tutorials tab (may take a few seconds), regardless of whether the tutorials package is loaded or not.

Notes on learnr

• While a learnr tutorial is running, RStudio Help and Packages tabs can be used, packages can be installed in RStudio, commands can be run from the console, and it is possible to run a shiny app from RStudio. So a lot of functionality of the RStudio interface can be used.
• Packages cannot be installed from a tutorial (must be done in RStudio).
• Packages (libraries) loaded within a tutorial (RMarkdown file) are available in all exercises within learnr, including exercises positioned before loading the package, but not within the RStudio interface.
• A data set loaded in the Rmd document is available for all exercises, including exercises positioned before loading the data set.
• A data set imported/read in one exercise (code box) cannot be used in another code box and is not present in the RStudio environment.
• It is not possible to run a shiny app from within the tutorial.
• I can’t make an external CSS file to work, so the (programming) tip style is defined in the tutorial.
• Environment setups for exercises via exercise.setup referencing a solution to a preceding exercise, which references another preceding exercise, seem to work if the tutorial is run/knitted from RStudio but not if the tutorial is started from the installed package. Create a new code chunk that sets up the environment for the exercise and reference this chunk in the exercise.setup argument. More generally, exercise setup code must be in the same section (## )?

IMPORTANT:

• learnr versions higher than 0.10.1 (and gradethis versions over 0.2.0.9000) will probably not accept the exercise solutions as set up now. The old learnr and gradethis vesions are saved in zip files, which can be unzipped into the folder containing the installed packages (check out: .libPaths()) to restore these versions. I checked that this works on a mac.
• Find a way to fix the learnr and gradethis packages at the start of the course. Copy (NOT fork) the GitHub sites to my GitHub site and have students install the packages from there?

Notes on gradethis

• The provided solution is directly accessible to the student unless we add one (not more!) Markdown (text) hint. A code chunk hint is displayed as a separate answer box if there is a code chunk check.
• Sometimes displays message “Last value being used to check answer is invisible. See ?invisible for more information”. Tell students to ignore this message.

Exercise setup (preparing data etc.):

• Chaining of solution code as setup (specify a-solution as exercise.setup, which has b-solution as setup) does not seem to work.
• Creating (new) data objects in a setup code chunk that is referenced with exercise.setup does not seem to work. A setup chunk (name-setup) does work (see crosstabs in Session 3).

With grade_code():

• If the correct answer does not specify argument names, argument names are considered incorrect and vice versa. Note that the “mapping =” argument does not throw an error.
• Can handle pipes and it can spot the location in a pipe where code starts to diverge as long as the correct answer is a pipe without new line characters. (Note: some gradethis builds throw an error: “Error occured while checking the submission”).
• We cannot address the code itself in grade_result() (or any of the other gradings?).
• For a chain of setups, e.g., code-checking question 3 requires the solution to code-checking question 2, which requires the solution to code-checking question 1, both the exercise and exercise-solution chunks of code-checking question 2 must refer to the solution of code-checking question 1.

With grade_result():

• Use a tilde if you use .result to refer to the resulting object: fail_if(~ ncol(.result) != 4, "Too few variables!”)
• but do not use a tilde if you use a function: fail_if(function(x) {ncol(x) != 4}, "Too few variables!”)
• if you use more than one criterion, put hem between {}: pass_if( ~ {nrow(.result) == 13 && ncol(.result) == 4 && identical(names(.result), c("friendships", "Freq", "Perc", "CumPerc")) && round(.result[[1, 3]]) == 13 && round(.result[[2, 4]]) == 25 }, “some message”)

Comparing ggplots:

• compare the code: any difference in the order of layers yields a fail message(?)
• compare the str (structure) output of each object:
  • Within the –check code chunk, create the correct plot: a <- ggplot() + <…>
  • Compare the user’s plot as a structure (list of strings?) to the correct plot pass_if( ~ diffobj::diffStr(.result, a)@diff.count.full == 0, "ok")
  • Or: A visual solution would be to look at the final svg produced by ggplot2. vdiffr::write_svg(). If the svg values are the same, then the image is the same. https://github.com/rstudio-education/gradethis/issues/97

WdeNooy/UsingRTutorials documentation built on Jan. 25, 2023, 2:39 a.m.