Version 2.3.1

• Fixed a sprintf format string for long integers.

Version 2.3.0

• NULL is not longer considered to be atomic in future versions of R (c.f. https://stat.ethz.ch/pipermail/r-devel/2023-September/082892.html). To avoid breaking reverse dependencies, checkmate will stick to the old behavior until further notice.
• Fixed a warning in checkAtomic() (#245).

Version 2.2.0

• Fixed C compiler warnings for windows
• Added checkPermutation (#230).
• Better handling of long vectors (#231).
• Fixed a bounds check for numeric vectors and infinite bounds.

Version 2.1.0

• New arguments n.chars and max.chars for checkCharacter() and checkString().
• Checks for integerish now compare the tolerance with the difference to the nearest integer with > instead of >= to allow specifying a tolerance of exactly 0 (#177).
• Checks for integerish now check for class Date an POSIXt.
• Coercion of double to integer in assertInt() and assertIntegerish() now round to the nearest integer instead of always rounding via trunc().
• Fixed an error message where the wrong variable name was reported by assert() (#182).
• Checks on POSIXct dates with storage mode integer should now work instead of raising an exception (#175).
• *Matrix() and *Array() now allow different storage types than the one specified if all values are missing (#184).
• Function assert() now supports collecting assertions via AssertCollection (#112).
• New exported C function qcheck() (#180).
• Fixed a bug in checkFunction(..., ordered = TRUE) (#204).
• Removed deprecated S macro DOUBLE_EPS from C source.

Version 2.0.0

• Expectations now optionally support the package tinytest. tinytest is used as backend if it is attached, otherwise checkmate defaults to testthat. There is now also a vignette on how to setup checkmate for tinytest.
• Coercion now only affects double vectors.
• Improved error message for type detection in *List.
• Removed *Bit, the bit package is orphaned.
• Fixed documentation.

Version 1.9.4

• Fixed factors being detected as integerish.
• Fixed error message for name checks of vectors.

Version 1.9.3

• New argument extension for checkPathForOutput() (#162).
• Fixed handling of different NA types in all set functions (#158).
• expect_vector removed due to a nameclash with package testthat.

Version 1.9.2

• assert*(..., coerce = TRUE) does not drop names during conversion (#157), thanks to @mb706.
• Fixed documentation in checkDataFrame (#159), thanks to @harvey131.
• Changed heuristic in vname() to improve lookup of variable names.

Version 1.9.1

• Fix segfault on Solaris
• Fix warnings reported by rchk
• Fix checking private slots in checkR6 (#156)

Version 1.9.0

• Error messages now provide more information about error locations, e.g., the position of the first missing element in a vector.
• If the object to check is missing, assert-functions now give a better error message, test-functions are always FALSE and expect-functions always raise an exception.
• Checks for missingness and sort order optimized for ALTREPs.
• The calling frame reported in assertions is now identical to the calling frame reported by R’s stop() function (#117).
• Added checkDouble to explicitly check for non-integer numerics.
• Added checkRaw to check raw vectors.
• Added checkFormula to check formula objects.
• Added checkMultiClass to check for inheritance from a set of candidates
• Added checkDisjunct to check sets for being disjunct.
• Added abbreviation "p" to qassert to check for POSIXct objects.
• Added argument coerce to assertCount/assert_count, assertInt/assert_int and assertIntegerish/assert_integerish which optionally coerces x to integer after an successful assertion. This supersedes the functions asCount, asInt and asInteger (#77).
• Added arguments max.rows and max.cols to check for maximum number of rows and columns for matrices, data.frames, tibbles and data.tables.
• Added argument disjunct.from to *Names.
• Fixed an error message in checkChoice.
• Fixed *Function to work properly with Primitives.
• Fixed *List where the check for missingness was broken.
• Workaround for *DataTable for the detection of the number of rows of null data.tables: https://github.com/Rdatatable/data.table/issues/3149

Version 1.8.5

• Added *POSIXct to check POSIXct data-time objects in POSIXct format.
• The set functions optionally support the package fastmatch now.
• Argument sorted = TRUE is not passed to ls() anymore to support R versions prior to v3.2.0.

Version 1.8.4

• New functions to test bit vectors implemented in package bit.
• New functions to test R6 classes implemented in package R6.
• Always load (not attach) the respective namespace if checking for objects of type data.table, tibble, R6 or bit. This ensures that all operations work as expected after the check.
• *Names with type="unnamed" now works with NULL.
• New argument must.include for *Names.
• Fixed possible protection stack imbalance as reported by rchk.

Version 1.8.3

• New argument sorted (defaults to FALSE) for *Integer, *Integerish and Numeric to check for ascending order of vector elements.
• New argument null.ok (defaults to FALSE) for *Choice and *Class.
• *Subset now allows to pass empty vectors to choices.
• Improved error message for *Choice.
• The set family of functions is now more restrict regarding the class, e.g. they differentiate between factors and characters.
• *Character and *String now ignores missing values in regular expressions and for string length checks (using argument min.chars). To disallow missing values, set any.missing or na.ok, respectively.
• *Date now ignores missing values in for lower/upper bound checks. To disallow missing values, set any.missing to FALSE. Thanks to Will Beasley (@wibeasley) for the PR.
• Package microbenchmark is no longer strictly required to build the vignette. If not installed, some output and figures will be missing though.

Version 1.8.2

• *Matrix and *Array now additionally allow to check for integerish storage type via argument “mode”.
• Functions *Count, *Int, *Number, *Integer, *Integerish and *Numeric do not accept logical values any more.
• checkAtomicVector is now more restrictive and prohibits a dimension symbol. Thus, a matrix is not considered an atomic vector any more.
• Dropped support for AssertCollections in convert functions (asInt, asInteger and asCount).
• Added checkTibble.

Version 1.8.1

• Function test_file is longer exported.
• *Function does not longer lookup functions with match.fun. As a result, passing functions via the string of the function name stopped working.
• In qassert using f as first char in a rule now specifies factor (before: function).

Version 1.8.0

• Most functions now support the handling of default arguments encoded as NULL via argument null.ok.
• Functions *File and *Directory are deprecated due to name clashes and will be removed in a future version. Please use *FileExists or *DirectoryExists instead.
• New helper function matchArg to provide a simple an easy way for partial argument matching in combination with an AssertCollection.
• Added alias functions for all check functions (check_*) to provide support for the underscore programming style in assert().

Version 1.7.4

• Compatibility with the upcoming testthat version.
• expect_ functions now return the checked object invisibly.
• Changed default of argument .var.name for assertions and label for expectations: They now default to the return value of the exported function vname (instead of missing which confuses some linters).
• Fixed error message in convert functions: Variable name was not properly looked up by the heuristic.
• Fixed a bug in qassertr and qtestr where the error message was not properly generated if multiple rules were provided.
• New argument depth for qtestr to control the recursion depth while checking nested lists.

Version 1.7.3

• Added checkDate.
• Argument .var.name of assert functions now has as default value (instead of missing).
• Fixed a bug in *OS functions.
• Fixed a bug in *Directory functions.
• New argument extension for the *File family of functions.

Version 1.7.2

• Added checkOS().
• Argument fixed for *Character functions now accepts a string instead of a boolean value and thus can directly be used for a substring search.
• New arguments min.chars, pattern, fixed and ignore.case for the *String family of functions.
• Exported helper functions wf (which.first) and wl (which.last).
• Now importing the new backports package for functions lengths() and dir.exists.

Version 1.7.1

• Fixed a segfault while checking an upper bound in qassert/qtest.
• Some minor speedups

Version 1.7.0

• Added alias functions for all functions to support the underscore style, e.g. assert_numeric is the new alias for assertNumeric and test_matrix is the alias for test_matrix.
• All assert functions now invisibly return the tested object instead of TRUE and thus can be used with magrittr pipes.
• Improved speed for most functions by reducing the .Call overhead (Thanks to Hadley Wickham).
• Added *DataTable functions to properly test primary and secondary keys of data tables.
• Removed *Percentage family of functions.
• Exported functions makeAssertion, makeTest and makeExpectation to assist expanding the package with user-generated checks.
• Added functions makeAssertionFunction, makeTestFunction and makeExpectationFunction to automatically create the respective functions based on a provided check function.

Version 1.6.3

• Assertions can now be collected (via makeAssertCollection()) and reported (via reportAssertions()).
• qassert() can now perform bound checks on strings.
• The default for the parameter “ordered” of the *SetEqual functions is now set to FALSE, as described in the documentation.

Version 1.6.2

• Fixed a compile-time warning.
• checkmate does not import testthat anymore in order to speed up package loading times and to keep the dependencies at a minimum. The expect_* family of functions can still be used, the namespace will be loaded on demand.

Version 1.6.1

• New family of functions: expect_* is intended to be used in combination with testthat. But note that functions expect_null() and expect_named() are not provided to avoid name clashes with testthat.
• Added qexpect() and qexpectr().
• Added argument all.missing for checks of matricies and data frames.
• Added anyNaN().
• Clarified documentation for assert() and allMissing().
• Fixed a bug where bound checks were performed on missing values.
• Fixed a bug where missingness was not correctly detected in data frames.

Version 1.6.0

• Started to support long vectors.
• Added a short vignette.
• Improved documentation.
• New argument “combine” for assert() to allow combining check functions with an AND instead of an OR.

Version 1.5.3

• Fixed a bug regarding the number of rows in zero-column data frames.
• Fixed a bug where the type of lists with dimension attribute where reported as “array” or “matrix”.
• Family *Array: new arguments “min.d” and “max.d”.
• Family Array and Matrix: Argument “mode” now additionally accepts strings “list” and “atomic”.

Version 1.5.2

• Fixed: (assert|check|test)Character(NA_character_, min.chars = 1) does not eval to TRUE anymore.
• New arguments for *Factor functions: (n|min|max).levels.
• Improved error messages for type and length checks.
• Improved error messages for missing arguments.

Version 1.5.1

• Included a workaround for R’s nrow and ncol to properly work with data frames.
• Fixed a bug handling complex number in checks for integerish values.
• Improved documentation.

Version 1.5.0

• Added checkNames().
• Added checkPercentage().
• Added anyInfinite().
• Fixed error messages for some dimension checks.
• Fixed an error while checking numerics for finiteness.

Version 1.4

• Fixed a bug where rownames and colnames of data.frames where not retrieved correctly.
• Fixed a bug in checkVector() (wrong order of arguments in call to C).
• Filesystem access: checks for write and executable rights are now disabled on windows.

Version 1.3

• Fixed a bug where logical values passed a check for numerics in qassert.
• Family *SetEqual: new argument “ordered”.
• checkPathForOutput: new argument “overwrite”.

Version 1.2

• Fixed bug in checkList.
• Fixed dimnames check on empty matrices and data frames.
• Added *SetEqual functions.

Version 1.1

• Improved error messages in assert* functions.
• New argument ‘empty.ok’ for *Subset functions.
• assert() now returns TRUE invisibly (as documented).
• Fixed handling of zero-length arguments in checkFunction().
• Fixed error message if duplicated values where found.
• Fixed a missing check for row names in checkMatrix and checkDataFrame.

Version 1.0

• Initial release on CRAN.