grade_code() no longer fails if .envir_result or .envir_solution is missing (#355).
detect_mistakes() now keeps a version of standardized user and solution code with and without default arguments added. Missing arguments are detected by comparing the user code with defaults to the solution code without defaults. Surplus arguments are detected by comparing the user code without defaults to the solution code with defaults (#356).
This helps avoid spurious feedback when comparing code that involves S3 methods. If the user’s code differs from the solution code in a way that means a different S3 method is used, the standardized code may gain different default arguments. This could result in feedback about missing or surplus arguments that were added by code standardization rather than by the student, which is not actionable feedback. By no longer looking for default arguments that are missing or surplus in the user code, we ensure that students receive more actionable feedback, likely about the incorrect argument that resulted in the use of a different S3 method.
This fixes a bug where a list could be marked as equal to another object with the same contents but a different class, e.g. list(a = 1, b = 2) and c(a = 1, b = 2) or data.frame(a = 1, b = 2).
Fix bug where call_standardise_formals() would fail when given a purrr::map() function where .f is an index rather than a function (#359).
gradethis 0.2.13
code_feedback() now standardizes arguments to functions defined within student and solution code before comparing code. It also now successfully standardizes arguments passed through ... by mapping functions into functions defined by setup code (#349).
gradethis_equal() now has a method for objects of class list. If two lists are not identical() and their lengths are the same, gradethis_equal() is applied to each element pairwise. This allows special logic for specific classes to be used for list elements of that class (#351).
call_standardise_formals() now applies ggplot2::standardise_aes_names() to all arguments of ggplot2 functions. This means it no longer detects inconsequential differences between, e.g., color = and colour = (#353).
When an exercise without an -error-check chunk returns an error, the default feedback now includes both the error message and code feedback if fail.hint = TRUE. Previously the error message was only shown when fail.hint = FALSE and code feedback was only shown when fail.hint = TRUE (#352).
gradethis 0.2.12.9005
gradethis_equal() now has default arguments of x = .result and y = .solution (#347).
New function: with_exercise() allows you to evaluate an expression as if it were run inside grade_this() with an associated exercise. It can be used alongside mock_this_exercise() for testing grading code outside the context of a learnr tutorial (#333).
.envir_solution is now included in the grade_this() check environment alongside .envir_prep and .envir_result. .envir_solution contains the state of the environment used to execute solution code, just as .envir_result does for student code (#333).
gradethis 0.2.11.9000
Bug fixes
Code feedback is now disabled for non-R exercise engines (#321).
More operators were added to our list of infixes (#327).
gradethis 0.2.10.9000
Feedback for error messages are now slightly more generic and refer to your code instead of your R code (@Laura-Puckett#318).
The solution_code argument of code_feedback() is now the single entry-point for solution code in code_feedback(), fail_if_code_feedback() and other functions that work with both .solution_code and .solution_code_all. In these cases, the default argument values will use multiple solutions if they exist (#305).
pass_if_equal() supports multiple solutions when authors set y = .solution_all to compare the student’s result with all solutions. For additional details, please see the new section, Comparing with Multiple Solutions, in ?pass_if_equal (#306).
A new vignette("multiple_solutions") describes how gradethis can be used to provide feedback for exercises with more than one solution (#312).
gradethis_error_checker() gains the hint argument from fail() which follows the global gradethis.fail.hint option. When FALSE, the error feedback won’t include code feedback hints (thanks @cswclui, #315).
gradethis_exercise_checker() can now be configure to evaluate solution code for non-R exercise engines by providing a function of code and envir to the solution_eval_fn argument. In addition to evaluation R exercise solutions, gradethis will now also evaluate SQL exercise solutions (#316).
gradethis 0.2.8.9000
gradethis now includes support for multiple solutions with different results. The solutions are made available in the .solution_all object in grade_this() grading code. pass_if_equal() gains support for multiple solutions as its y argument. If .solution_all is used as the y argument of pass_if_equal(), it will return a passing grade if x matches any of the multiple solutions (#296).
gradethis 0.2.6.9000
You may now call return() in grade_this() grading code to exit grading early. This is allowed in code and error checking code, but will result in an “internal error” when used in the -check chunk grading code (#284).
gradethis now includes low-level support for multiple solutions. Authors can add multiple solutions in the -solution chunk, separated by code section headers, e.g. # ----. (note only trailing dashes, ----, are supported). These additional solutions are made available in the .solution_code_all object in grade_this() grading code and are named with the code section name if one is provided, e.g. # first ----. When multiple solutions are provided using the code section comments, .solution and .solution_code will default to the last solution. (#286)
gradethis can now be used to grade non-R code. If learnr can evaluate the non-R code and return the result of the submitted code as an R object, then gradethis can be used to grade the submission result. Grading code is still written in R and code feedback tool designed for R will not work as expected (#290).
grade_this_code() gains an action argument, allowing authors to choose if grade_this_code() should only "pass" or "fail" the user’s submission. By default, grade_this_code() uses action = "both" to maintain current behavior. (#276)
When combined with learnr version 0.10.1.9017 or later, gradethis’ exercise checking function will not require that grading code absolutely return feedback unless exercise checking is at the "check" stage. (#276)
Breaking changes
Errors in the grading code are now returned as neutral grades rather than failing grades. The feedback message and type can be changed with two new arguments to gradethis_setup(): grading_problem.message and grading_problem.type (#256).
gradethis 0.2.4.9000
Added gradethis_error_checker(), a more robust checking function for general use when the student’s submission should not throw an error. (#234)
Breaking changes
grade_this() no longer automatically converts errors to fail() grades, instead authors need to wrap unit-test-style code in fail_if_error() to convert them to grades. This helps to better differentiate between unexpected errors resulting from the author’s grading code and portions of the grading code where errors are expected and indicative of a problem with the user’s code. (#254)
gradethis 0.2.3.9001
All failinggraded() helper functions, like fail() etc, now take a hint argument that when TRUE adds a code feedback hint to the custom feedback message. The default value of this argument can be set via gradethis_setup(). (#216)
Passing and failing graded() helper functions gain a praise or encourage argument (respectively) to prepend a random praising phrase when passing or append a random encouraging phrase when failing. The default values of these arguments can be set via gradethis_setup(). (#227)
New function: fail_if_code_feedback(). Returns an incorrect grade when there are differences between the user code and solution code (if it exists). (#228)
We now use placeholder sentinel objects as function argument defaults to signal that a function will find the object in the grade_this() check environment. The help page ?grade_this-objects describes these objects and documents their purpose, which you can also find by calling one of the placeholders, e.g. .result. (#232)
Breaking changes
The x argument of pass_if() and fail_if() has been renamed cond and both functions now work inside grade_this(), although the function and formula versions are not supported there. (#216)
gradethis_setup() now uses a new argument order that favors the gradethis-specific options (#212).
gradethis.code.feedback is now gradethis.maybe_code_feedback. (#219)
The space_before and space_after arguments of maybe_code_feedback() have been deprecated in favor of more flexible arguments before and after that accept arbitrary strings to add before or after the message. (#219)
The “Insert Exercise …” RStudio Addins were removed from gradethis (#196).
The names of several global options were changed in #210. This will only affect users who were setting the options directly rather than using gradethis_setup(). The name changes are:
gradethis.code.partial_matching is now gradethis.allow_partial_matching
gradethis.code.feedback is now gradethis.fail_code_feedback
gradethis.code.correct is now gradethis.code_correct
gradethis.code.incorrect is now gradethis.code_incorrect
gradethis_glue_correct is now gradethis.glue_correct
gradethis_glue_incorrect is now gradethis.glue_incorrect
gradethis_glue_correct_test is now gradethis.glue_correct_test
gradethis_glue_incorrect_test is now gradethis.glue_incorrect_test
pass_if_equal() now compares the submitted .result to the exercise .solution by default (#203).
New function: debug_this() returns information about the various objects available for use by grading code. It is designed for tutorial authors to use during development and can be used in *-check chunks or inside grade_this() grading code and the feedback is shown in the tutorial when clicking on Submit Answer. See ?debug_this for more information. (#203)
graded() and its pass/fail helper functions now accept type and location parameters that specify how the feedback is displayed (see learnr Custom Checking for more details about these options).
Bug fixes
The gradethis.glue_pipe option is now called gradethis.pipe_warning as it sets the default value of the pipe_warning() function. pipe_warning() can be included in the glue strings of other messages, such as those set by gradethis.code.incorrect (#193).
The glue_pipe argument of glue_code() is now deprecated (#193).
gradethis 0.2.0.9001
Checking exercise code with blanks, e.g. ____, now returns clear feedback that students should replace the ____ with code. (#153)
The exercise.parse.error global option now accepts a function with one argument. The function is given the gradethis check environment with an additional .error object containing the parse error condition. (#153)
Improved code feedback for function definitions will detect mistakes in function arguments (#178)
gradethis now accepts markdown or an htmltools tags or a tagList() object for feedback messages. Markdown processing is handled via commonmark. Internally, all code feedback messages now use markdown. (#189)
Breaking changes
Improved code feedback for function definitions will detect mistakes in function arguments (#178)
New function: grade_this(expr). Evaluates the expression and returns the first grade that is called or error that is thrown.
New function: grade_this_code(correct, incorrect). Makes a grade from comparing the user code against the solution code. This is a simplified version of grade_code().
New function: code_feedback(). Compares the user code against the solution code.
