NEWS

CBTF 0.6.0 (2026-02-13)

Rewrite fuzz() to use a backend based on the mirai package. This allows to parallelize computations and support a timeout to stop non-failing functions (#4).
Add the timeout argument to fuzz() to control the number of seconds to wait before interrupting the execution of a function being fuzzed.
Add the args argument to allow fuzzing multiple function arguments instead of only the first, with support for named and fixed arguments (#6, #7, #8).
Add support for what = NULL, which allows running the set of arguments listed in args with no modifications. This can be used after a fuzz run to collect results on just one set of problematic inputs.
Replace the show_all logical argument of print() with a show argument that accepts the class of results to be printed, any of "fail", "warn", "skip", "ok" or "all" (#9).

Use unicode symbols only on terminals that support UTF8, and respect the cli.unicode option, so that options(cli.unicode = FALSE) can be used to disable all unicode output.
Fix bug that caused failures in the +() operator to be whitelisted just because the + was interpreted as a regular expression quantifier.
Ignore deprecated functions by default in get_exported_functions().
Add more inputs to test_inputs(), bringing the total number of inputs from 60 to 70.
Move the NA inputs in test_inputs() from the "scalar" to the new "na" class.
Add the [[ operator to extract the results for a specific test input.
Stop exporting the namify() helper.

Streamline the progress report from fuzz().
In case both an error and a warning are raised by the function being fuzzed, we now report the error; previously the warning was reported because of how tryCatch() is implemented.
The length() S3 method for objects of class cbtf now counts the total number of tests performed, instead of the number of inputs tested.
Add the tabulate argument to summary() to control whether the tabulation of results should be printed to the terminal or not.
Change the internal structure of the cbtf object to store the names of the functions fuzzed only once.
Store the values of the ignore_patterns and ignore_warnings options in the cbtf object.
Add the whitelist() function to remove false positives from a cbtf object.
Make test_input() recognise the "help" keyword, in which case it returns a vector of valid input classes.
Add the namify() function to help create a named list of custom input objects.

Don't crash in get_exported_functions() if getNameSpaceExports() fails.
Store the error message in the results object also in case of whitelisted errors, so it can be reported from print(..., show_all = TRUE).
Add the tabulation shown in summary() as attribute "summary_table" of the data frame returned.
Make get_exported_functions() return only fuzzable functions, so that functions with no arguments are no longer returned.
Expand the documentation to satisfy the CRAN request.

Warnings containing the name of the function being fuzzed are no longer reported, to make their behaviour consistent with that of errors.
Add the package argument to fuzz() as an alternative and simpler way to specify the namespace where function names are searched.
Don't consider primitives (such as list() or +) to be functions with no arguments (#5).
Make the what argument to fuzz() accept a list of inputs, so that the function can test multiple inputs in the same run.
Add function test_inputs() to return a configurable list of problematic inputs.
Add the listify_what argument to fuzz() so that each input in what is also tested in its listified version.
Provide the S3 method for length().
Make the output from summary() more informative and include a column with the inputs tested in the data frame returned.

Return an object with the raw results from fuzz() and provide S3 methods for summary() and print() (#1).
Skip functions that call readline(), as in an interactive session they stall waiting for user input (#4).
Skip non-existing function names and functions that accept no arguments.
Show a progress bar during fuzz().
Prettify the output from fuzz(), summary() and print().

First version of the package.
The package website is now available at https://mcol.github.io/caught-by-the-fuzz/