Package that explicitly depend on rlang in their description file are now opting into a new snapshot display for errors, warnings, and messages. Previously this only concerned packages that explicitly depended on rlang >= 1.0.0. This display will eventually become the default for all packages.
Changes include:
Condition classes are no longer included in the snapshot by default. This is to avoid snapshot noise when upstream code adds or changes a class. For instance, r-devel has added classes to base errors.
Warnings and errors are now printed with rlang, including the call field. This makes it easy to monitor the full appearance of warning and error messages as they are displayed to users.
This change is part of a push towards mentioning the useful context of an error as part of messages, see the release notes of rlang 1.0.0 for more about this.
Test results show hyperlinks to failed expectation when supported (#1544).
testthat now uses brio for all reading and writing (#1120). This ensures that snapshots always use “” to separate lines (#1516).
expect_snapshot() no longer inadvertently trims trailing new lines off of errors and messages (#1509).
If expect_snapshot() generates a snapshot with different value but still compares as equal (e.g. because you’ve set a numeric tolerance), the saved values no longer update if another snapshot in the same file changes.
expect_snapshot() now only adds a .new file for the variants that actually changed, not all variants, while expect_snapshot_file() with variant with no longer immediately deletes .new files (#1468).
expect_snapshot_file() gains a transform argument to match expect_snapshot() (#1474). compare now defaults to NULL, automatically guessing the comparison type based on the extension.
expect_snapshot_file() now errors if the file being snapshot does not exist; SnapshotReporter also now treats the file directory as an absolute path (#1476, @malcolmbarrett)
New expect_snapshot_warning() to match expect_snapshot_error() (#1532).
JUnitReporter now includes skip messages/reasons (@rfineman, #1507)
local_reproducible_output() gains a lang argument so that you can optionally override the language used to translate error messages (#1483). It also sets the global option cli.num_colors in addition to crayon.enabled.
test_that() no longer inappropriately skips when calling expect_equal() when you’ve temporarily set the locale to non-UTF-8 (#1285).
skip_if_offline() now automatically calls skip_on_cran() (#1479).
snapshot_accept() and snapshot_review() now work with exactly the same file specification which can be a snapshot name, a file name, or a directory (#1546). They both work better with variants (#1508). Snapshot cleanup also removes all empty directories (#1457).
When a snapshot changes the hint also mentions that you can use snapshot_review() (#1500, @DanChaltiel) and the message tells you what variant is active (#1540).
JUnit reporter now includes skip messages/reasons (@rfineman, #1507).
Condition expectations like expect_error() now match across the ancestry of chained errors (#1493). You can disable this by setting the new inherit argument to FALSE.
Added preliminary support for rlang 1.0 errors. It is disabled by default for the time being. To activate it, specify rlang (>= 1.0.0) in your DESCRIPTION file (or >= 0.99.0.9001 if you’re using the dev version).
Once activated, snapshots will now use rlang to print error and warning messages, including the Error: and Warning: prefixes. This means the call field of conditions is now displayed in snapshots if present. Parent error messages are also displayed. Following this change, all snapshots including error and warning messages need to be revalidated.
We will enable the new rlang 1.0 output unconditionally in a future release.
expect_snapshot() gains a new argument cnd_class to control whether to show the class of errors, warnings, and messages.
The default is currently unchanged so that condition classes keep being included in snapshots. However, we plan to change the default to FALSE in an upcoming release to prevent distracting snapshot diffing as upstream packages add error classes. For instance, the development version of R is currently adding classes to basic errors, which causes spurious snapshot changes when testing against R-devel on CI.
If you depend on rlang 1.0 (see above), the default is already set to FALSE.
expect_snapshot() no longer processes rlang injection operators like !!.
Fixed bug in expectations with long inputs that use :: (#1472).
expect_snapshot() is no longer experimental.
expect_snapshot() and friends gets an experimental new variant argument which causes the snapshot to be saved in _snaps/{variant}/{test}.md instead of _snaps/{test}.md. This allows you to generate (and compare) unique snapshots for different scenarios like operating system or R version (#1143).
expect_snapshot() gains a transform argument, which should be a function that takes a character vector of lines and returns a modified character vector of lines. This makes it easy to remove sensitive (e.g. API keys) or stochastic (e.g. random temporary directory names) from snapshot output (#1345).
expect_snapshot_file() now replaces previous .new snapshot if code fails again with a different value.
expect_snapshot_value() now has an explicit tolerance argument which uses the testthat default, thus making it more like expect_equal() rather than expect_identical(). Set it to NULL if you want precise comparisons (#1309). expect_snapshot_value(style = "deparse") now works with negative values (#1342).
If a test containing multiple snapshots fails (or skips) in between snapshots, the later snapshots are now silently restored. (Previously this warned and reset all snapshots, not just later snapshots).
If you have multiple tests with the same name that use snapshots (not a good idea), you will no longer get a warning. Instead the snapshots will be aggregated across the tests.
Condition expectations now consistently return the expected condition instead of the return value (#1371). Previously, they would only return the condition if the return value was NULL, leading to inconsistent behaviour.
This is a breaking change to the 3rd edition. Where you could previously do:
expect_equal(expect_warning(f(), "warning"), "value")You must now use condition expectations on the outside:
expect_warning(expect_equal(f(), "value"), "warning")
# Equivalently, save the value before inspection
expect_warning(value <- f(), "warning")
expect_equal(value, "value")This breaking change makes testthat more consistent. It also makes it possible to inspect both the value and the warning, which would otherwise require additional tools.
Errors in test blocks now display the call if stored in the condition object (#1418). Uncaught errors now show their class (#1426).
Multi-line skips only show the first line in the skip summary.
expr_label(), which is used to concisely describe expressions used in expectations, now does a better job of summarising infix function (#1442).
local_reproducible_output() now sets the max.print option to 99999 (the default), so your tests are unaffected by any changes you might’ve made in your .Rprofile (1367).
ProgressReporter (the default only) now stops at the end of a file; this ensures that you see the results of all related tests, and ensures that snapshots are handled consistently (#1402).
ProgressReporter now uses an env var to adjust the maximum number of failures. This makes it easier to adjust when the tests are run in a subprocess, as is common when using RStudio (#1450).
skip_on_os() gains an arch argument so you can also choose to skip selected architectures (#1421).
test_that() now correctly errors when an expectation fails when run interactively (#1430).
test_that() now automatically and correctly generate an “empty test” skip if it only generates warnings or messages (and doesn’t contain any expectations).
testthat_tolerance() no longer has an unused argument.
The vendored Catch code used for use_catch() now uses a constant value for the stack size rather than relying on SIGSTKSZ. This fixes compatibility for recent glibc versions where SIGSTKSZ is no longer a constant.
Fixed an issue that caused errors and early termination of tests on R <= 3.6 when a failing condition expectation was signalled inside a snapshot.
expect_snapshot_file() gains a compare argument (#1378, @nbenn). This is a customisation point for how to compare old and new snapshot files.
The functions compare_file_binary() and compare_file_text() are now exported from testthat to be supplied as compare argument. These implement the same behaviour as the old binary argument which is now deprecated.
expect_snapshot() no longer deletes snapshots when an unexpected error occurs.
New announce_snapshot_file() function for developers of testthat extensions. Announcing a snapshot file allows testthat to preserve files that were not generated because of an unexpected error or a skip() (#1393). Unannounced files are automatically deleted during cleanup if the generating code isn’t called.
New expectation: expect_no_match(). It complements expect_match() by checking if a string doesn’t match a regular expression (@michaelquinn32, #1381).
Support setting the testthat edition via an environment variable (TESTTHAT_EDITION) as well (@michaelquinn32, #1386).
Failing expectations now include a backtrace when they’re not called directly from within test_that() but are instead wrapped in some helper function (#1307).
CheckReporter now only records warnings when not on CRAN. Otherwise failed CRAN revdep checks tend to be cluttered up with warnings (#1300). It automatically cleans up testthat-problems.rds left over from previous runs if the latest run is succesful (#1314).
expect_s3_class() and expect_s4_class() can now check that an object isn’t an S3 or S4 object by supplying NA to the second argument (#1321).
expect_s3_class() and expect_s4_class() format class names in a less confusing way (#1322).
expect_snapshot() collapses multiple adjacent headings of the same, so that, e.g., if you have multiple lines of code in a row, you’ll only see one “Code:” heading (#1311).
New testthat.progress.verbose_skips option. Set to FALSE to stop reporting skips as they occur; they will still appear in the summary (#1209, @krlmlr).
CheckReporter results have been tweaked based on experiences from running R CMD check on many packages. Hopefully it should now be easier to see the biggest problems (i.e. failures and errors) while still having skips and warnings available to check if needed (#1274). And now the full test name is always shown, no matter how long (#1268).
Catch C++ tests are no longer reported multiple times (#1237) and are automatically skipped on Solaris since Catch is not supported (#1257). use_catch() makes it more clear that your package needs to suggest xml2 (#1235).
auto_test_package() works once again (@mbojan, #1211, #1214).
expect_snapshot() gains new error argument which controls whether or not an error is expected. If an unexpected error is thrown, or an expected error is not thrown, expect_snapshot() will fail (even on CRAN) (#1200).
expect_snapshot_value(style = "deparse") handles more common R data structures.
expect_snapshot_value() now passes ... on to waldo::compare() (#1222).
expect_snapshot_file() gives a hint as to next steps when a failure occurs in non-interactive environments (with help from @maelle, #1179). expect_snapshot_*() gives a more informative hint when you’re running tests interactively (#1226).
expect_snapshot_*() automatically removes the _snaps directory if it’s empty (#1180). It also warns if snapshots are discarded because tests have duplicated names (#1278, @krlmlr).
local_reproducible_output() now sets the LANGUAGE env var to “en”. This matches the behaviour of R CMD check in interactive settings (#1213). It also now unsets RSTUDIO envvar, instead of setting it to 0 (#1225).
RstudioReporter has been renamed to RStudioReporter.
skip_if_not() no longer appends “is not TRUE” to custom messages (@dpprdan, #1247).
test_that() now warns (3e only) if code doesn’t have braces, since that makes it hard to track the source of an error (#1280, @krlmlr).
testhat 3.0.0 brings with it a 3rd edition that makes a number of breaking changes in order to clean up the interface and help you use our latest recommendations. To opt-in to the 3rd edition for your package, set Config/testthat/edition: 3 in your DESCRIPTION or use local_edition(3) in individual tests. You can retrieve the active edition with edition_get(). Learn more in vignette("third-edition").
context() is deprecated.
expect_identical() and expect_equal() use waldo::compare() to compare actual and expected results. This mostly yields much more informative output when the actual and expected values are different, but while writing it uncovered some bugs in the existing comparison code.
expect_error(), expect_warning(), expect_message(), and expect_condition() now all use the same underlying logic: they capture the first condition that matches class/regexp and allow anything else to bubble up (#998/#1052). They also warn if there are unexpected arguments that are never used.
The all argument to expect_message() and expect_warning() is now deprecated. It was never a particularly good idea or well documented, and is now superseded by the new condition capturing behaviour.
expect_equivalent(), expect_reference(), expect_is() and expect_that() are deprecated.
Messages are no longer automatically silenced. Either use suppressMessages() to hide unimportant messages, or expect_messsage() to catch important messages (#1095).
setup() and teardown() are deprecated in favour of test fixtures. See vignette("test-fixtures") for more details.
expect_known_output(), expect_known_value(), expect_known_hash(), and expect_equal_to_reference() are all deprecated in favour of expect_snapshot_output() and expect_snapshot_value().
test_that() now sets a number of options and env vars to make output as reproducible as possible (#1044). Many of these options were previously set in various places (in devtools::test(), test_dir(), test_file(), or verify_output()) but they have now been centralised. You can use in your own code, or when debugging tests interactively with local_test_context().
with_mock() and local_mock() are deprecated; please use the mockr or mockery packages instead (#1099).
New family of snapshot expectations (expect_snapshot(), expect_snapshot_output(), expect_snapshot_error(), and expect_snapshot_value()) provide “snapshot” tests, where the expected results are stored in separate files in test/testthat/_snaps. They’re useful whenever it’s painful to store expected results directly in the test files.
expect_snapshot_file() along with snapshot_review() help snapshot more complex data, with initial support for text files, images, and data frames (#1050).
See vignette("snapshotting") for more details.
CheckReporter (used inside R CMD check) now prints out all problems (i.e. errors, failures, warnings and skips; and not just the first 10), lists skips types, and records problems in machine readable format in tests/testthat-problems.rds (#1075).
New CompactProgressReporter tweaks the output of ProgressReporter for use with a single file, as in devtools::test_file(). You can pick a different default by setting testthat.default_compact_reporter to the name of a reporter.
ProgressReporter (the default reporter) now keeps the stack traces of an errors that happen before the before test, making problems substantially easier to track down (#1004). It checks if you’ve exceeded the maximum number of failures (from option testthat.progress.max_fails) after each expectation, rather than at the end of each file (#967). It also gains new random praise options that use emoji, and lists skipped tests by type (#1028).
StopReporter adds random praise emoji when a single test passes (#1094). It has more refined display of failures, now using the same style as CompactProgressReporter and ProgressReporter.
SummaryReporter now records file start, not just context start. This makes it more compatible with modern style which does not use context() (#1089).
All reporters now use exactly the same format when reporting the location of an expectation.
Warnings now include a backtrace, making it easier to figure out where they came from.
Catch C++ tests now provide detailed results for each test. To upgrade existing code, re-run testthat::use_catch() (#1008).
Many reporters (e.g. the check reporter) no longer raise an error when any tests fail. Use the stop_on_failure argument to devtools::test() and testthat::test_dir() if your code relies on this. Alternatively, use reporter = c("check", "fail") to e.g. create a failing check reporter.
New vignette("test-fixtures") describes test fixtures; i.e. how to temporarily and cleanly change global state in order to test parts of your code that otherwise would be hard to run (#1042). setup() and teardown() are superseded in favour of test fixtures.
New teardown_env() for use with withr::defer(). This allows you to run code after all other tests have been run.
New vignette("skipping") gives more general information on skipping tests, include some basics on testing skipping helpers (#1060).
ProgressReporter() and CheckReporter() list the number of skipped tests by reason at the end of the reporter. This makes it easier to check that you’re not skipping the wrong tests, particularly on CI services (#1028).
test_that() no longer triggers an error when run outside of tests; instead it produces a more informative summary of all failures, errors, warnings, and skips that occurred inside the test.
test_that() now errors if desc is not a string (#1161).
test_file() now runs helper, setup, and teardown code, and has the same arguments as test_dir() (#968). Long deprecated encoding argument has been removed.
test_dir() now defaults stop_on_failure to TRUE for consistency with other test_ functions. The wrap argument has been deprecated; it’s not clear that it should ever have been exposed.
New test_local() tests a local source package directory. It’s equivalent to devtools::test() but doesn’t require devtools and all its dependencies to be installed (#1030).
testthat no longer supports tests stored in inst/tests. This has been deprecated since testthat 0.11.0 (released in 2015). test_package() (previously used for running tests in R CMD check) will fail silently if no tests are found to avoid breaking old packages on CRAN (#1149).
capture_output() and verify_output() use a new testthat_print() generic. This allows you to control the printed representation of your object specifically for tests (i.e. if your usual print method shows data that varies in a way that you don’t care about for tests) (#1056).
context_start_file() is now exported for external reporters (#983, #1082). It now only strips first instance of prefix/suffix (#1041, @stufield).
expect_error() no longer encourages you to use class. This advice removes one type of fragility at the expense of creating a different type (#1013).
expect_known_failure() has been removed. As far as I can tell it was only ever used by testthat, and is rather fragile.
expect_true(), expect_false(), and expect_null() now use waldo to produce more informative failures.
verify_output() no longer always fails if output contains a carriage return character (") (#1048). It uses the pdf() device instead of png() so it works on systems without X11 (#1011). And it uses waldo::compare() to give more informative failures.
The last version of testthat introduced a performance regression in error assertions (#963). To fix it, you need to install rlang 0.4.2.
Fixed error assertions with rJava errors (#964).
Fixed issue where error and warning messages were not retrieved with conditionMessage() under certain circumstances.
This release mostly focusses on an overhaul of how testthat works with conditions (i.e. errors, warnings and messages). There are relatively few user-facing changes, although you should now see more informative backtraces from errors and failures.
Unexpected errors are now printed with a simplified backtrace.
expect_error() and expect_condition() now display a backtrace when the error doesn’t conform to expectations (#729).
expect_error(), expect_warning() and expect_message() now call conditionMessage() to get the condition message. This generic makes it possible to generate messages at print-time rather than signal-time.
expect_error() gets a better warning message when you test for a custom error class with regexp.
New exp_signal() function is a condition signaller that implements the testthat protocol (signal with stop() if the expectation is broken, with a continue_test restart).
Existence of restarts is first checked before invokation. This makes it possible to signal warnings or messages with a different condition signaller (#874).
ListReporter now tracks expectations and errors, even when they occur outside of tests. This ensures that stop_on_failure matches the results displayed by the reporter (#936).
You can silence warnings about untested error classes by implementing a method for is_uninformative_warning(). This method should be lazily registered, e.g. with vctrs::s3_register(). This is useful for introducing an experimental error class without encouraging users to depend on the class in their tests.
Respect options(warn = -1) to ignore all warnings (@jeroen #958).
Expectations can now be explicitly subclassed with new_expectation(). This constructor follows our new conventions for S3 classes and takes an optional subclass and optional attributes.
Unquoted inputs no longer potentially generate multiple test messages (#929).
verify_output() no longer uses quasiquotation, which fixes issues when verifying the output of tidy eval functions (#945).
verify_output() gains a unicode parameter to turn on or off the use of Unicode characters by the cli package. It is disabled by default to prevent the tests from failing on platforms like Windows that don’t support UTF-8 (which could be your contributors’ or your CI machines).
verify_output() now correctly handles multi-line condition messages.
verify_output() now adds spacing after condition messages, consistent with the spacing added after normal output.
verify_output() has a new syntax for inserting headers in output files: insert a "# Header" string (starting with # as in Markdown) to add a header to a set of outputs.
compare.numeric() uses a more sophisticated default tolerance that will automatically skip tests that rely on numeric tolerance if long doubles are not available (#940).
JunitReporter now reports tests in ISO 8601 in the UTC timezone and uses the maximum precision of 3 decimal places (#923).
test_rd() and add a couple of tests to hopefully detect the problem earlier in the future.verify_output() is designed for testing output aimed at humans (most commonly print methods and error messages). It is a regression test that saves output in a way that makes it easy to review. It is automatically skipped on CRAN (#782, #834).as.data.frame.testthat_results() now always returns a data frame with 13 columns (@jozefhajnala, #887).
auto_test_package() now correctly handles helper files (tests/testthat/helper-*.R), automatically reloading all code and rerunning all tests (@CorradoLanera, #376, #896).
expect_match() now displays info even when match length is 0 (#867).
expect_s3_class() gains new exact argument that allows you to check for an exact class match, not just inheritance (#885).
fail() and succeed() gain info argument, which is passed along to expect().
test_examples() gets some minor fixes: it now returns the results invisibly, doesn’t assume that examples should contain tests, and documents that you shouldn’t be using it routinely (#841).
test_file() only calls Reporter$end_context() if a context was started, fixing an error in TeamcityReporter (@atheriel, #883).
skip() now reports reason for skipping as: Reason: {skip condition} (@patr1ckm, #868).
skip_if() and skip_if_not() now report Reason: {skip condition} is TRUE and Reason: {skip condition} is not TRUE respectively (@ patr1ckm, #868).
skip_if_translated() now tests for translation of a specific message. This is more robust than the previous approach because translation happens message-by-message, not necessarily for the entire session (#879) (and in general, it’s impossible to determine what language R is currently using).
skip_on_covr() allows you to skip tests when covr is running. (@ianmcook, #895)
expect_known_value() gains a new serialisation version argument, defaulting to 2. Prevents the .rds files created to hold reference objects from making a package appear to require R >= 3.5 (#888 @jennybc).
New expect_visible() and expect_invisible() make it easier to check if a function call returns its result visibly or invisibly (#719).
New expect_mapequal(x, y) checks that x and y have the same names, and the same value associated with each name (i.e. they compare the values of the vector standardising the order of the names) (#863).
New expect_vector() is a wrapper around vctrs::vec_assert() making it easy to test against the vctrs definitions of prototype and size (#846). (Currently requires development version of vctrs.)
All expectations give clearer error messages if you forget the object or expected arguments (#743).
expect_equal() now correctly compares infinite values (#789).
In expect_equal_to_reference(), the default value for update is now FALSE (@BrodieG, #683).
expect_error() now returns the error object as documentated (#724). It also now warns if you’re using a classed expectation and you’re not using the class argument. This is good practice as it decouples the error object (which tends to be stable) from its rendering to the user (which tends to be fragile) (#816).
expect_identical() gains a ... argument to pass additional arguments down to identical() (#714).
expect_lt(), expect_lte(), expect_gt() expect_gte() now handle Inf and NA arguments appropriately (#732), and no longer require the inputs to be numeric.
expect_output() gains a width argument, allowing you to control the output width. This does not inherit from getOption("width"), ensuring that tests return the same results regardless of environment (#805).
expect_setequal() now works with more vector types (including lists), because it uses %in%, rather than sort(). It also warns if the inputs are named, as this suggests that your mental model of how expect_setequal() works is wrong (#750).
is_true() and is_false() have been deprecated because they conflict with other functions in the tidyverse.
Reporter documentation has been considerably improved (#657).
CheckReporter, used by R CMD check, now includes a count of warnings.
JUnitReporter no longer replaces . in class names (#753), and creates ouput that should be more compatible with Jenkins (#806, @comicfans).
ListReporter now records number of passed tests and original results in new columns (#675).
ProgressReporter, the default reporter, now:
Automatically generates a context from the file name. We no longer recommend the use of context() and instead encourage you to delete it, allowing the context to be autogenerated from the file name.
This also eliminates the error that occured if tests can before the first context() (#700, #705).
Gains a update_interval parameter to control how often updates are printed (default 0.1 s). This prevents large printing overhead for very fast tests. (#701, @jimhester)
Uses a 3 character wide column to display test successes, so up to 999 successful tests can be displayed without changing the alignment (#712).
reporter$end_reporter() is now only called when testing completes successfully. This ensures that you don’t get unnecessary output when the test fails partway through (#727).
skip_if_offline() skips tests if an internet connection is not available (#685).
skip_on_ci() skips tests on continuous integration systems (@mbjoseph, #825) by looking for a CI env var..
New testthat_examples() and testthat_example() make it easy to access new test files bundled with the package. These are used in various examples to make it easier to understand how to use the package.
New local_mock() which allows you to mock a function without having to add an additional layer of indentation as with with_mock() (#856).
auto_test_package() works better with recent devtools and also watches src/ for changes (#809).
expect_s3_class() now works with unquoting (@jalsalam, #771).
expectation objects now contain the failure message, even when successful (#836)
devtools::test() no longer fails if run multiple times within the same R session for a package containing Catch tests. (devtools #1832)
New testing_package() retrieves the name of the package currently being tested (#699).
run_testthat_tests C entrypoint is registered more robustly.
skip() now always produces a message of length 1, as expected elsewhere in testthat (#791).
Warnings are passed through even when options(warn = 2) is set (@yutannihilation, #721).
“Can’t mock functions in base packages”: You can no longer use with_mock() to mock functions in base packages, because this no longer works in R-devel due to changes with the byte code compiler. I recommend using mockery or mockr instead.
The order of arguments to expect_equivalent() and expect_error() has changed slightly as both now pass ... on another function. This reveals itself with a number of different errors, like:
If you see one of these errors, check the number, order, and names of arguments to the expectation.
“Failure: (unknown)”. The last release mistakenly failed to test bare expectations not wrapped inside test_that(). If you see “(unknown)” in a failure message, this is a failing expectation that you previously weren’t seeing. As well as fixing the failure, please also wrap inside a test_that() with an informative name.
“Error: the argument has already been evaluated”: the way in which expectations now need create labels has changed, which caused a couple of failures with unusual usage when combined with Reduce, lapply(), and Map(). Avoid these functions in favour of for loops. I also recommend reading the section below on quasiquotation support in order to create more informative failure messages.
expect_condition() works like expect_error() but captures any condition, not just error conditions (#621).
expect_error() gains a class argument that allows you to make an assertion about the class of the error object (#530).
expect_reference() checks if two names point to the same object (#622).
expect_setequal() compares two sets (stored in vectors), ignoring duplicates and differences in order (#528).
skip_if() makes it easy to skip a test when a condition is true (#571). For example, use skip_if(getRversion() <= 3.1) to skip a test in older R versions.
skip_if_translated() skips tests if you’re running in an locale where translations are likely to occur (#565). Use this to avoid spurious failures when checking the text of error messages in non-English locales.
skip_if_not_installed() gains new minimum_version argument (#487, #499).
We have identified a useful family of expectations that compares the results of an expression to a known good value stored in a file. They are designed to be use in conjunction with git so that you can see what precisely has changed, and revert it if needed.
expect_known_output() replaces expect_output_file(), which has been soft-deprecated. It now defaults to update = TRUE and warn, rather than failing on the first run. It gains a print argument to automatically print the input (#627). It also sets the width option to 80 to ensure consistent output across environments (#514)
expect_known_value() replaces expect_equal_to_reference(), which has been soft-deprecated. It gains an update argument defaulting to TRUE. This changes behaviour from the previous version, and soft-deprecated expect_equal_to_reference() gets update = FALSE.
expect_known_failure() stored and compares the failure message from an expectation. It’s a useful regression test when developing informative failure messges for your own expectations.
All expectations can now use unquoting (#626). This makes it much easier to generate informative failure messages when running tests in a for loop.
For example take this test:
When it fails, you’ll see the message Error: `f(i)` not equal to `i * 10`. That’s hard to diagnose because you don’t know which iteration caused the problem!
If you unquote the values using !!, you get the failure message `f(4L)` not equal to 40.. This is much easier to diagnose! See ?quasi_label() for more details.
(Note that this is not tidy evaluation per se, but is closely related. At this time you can not unquote quosures.)
New setup() and teardown() functions allow you to run at the start and end of each test file. This is useful if you want to pair cleanup code with the code that messes up state (#536).
Two new prefixes are recognised in the test/ directory. Files starting with setup are run before tests (but unlike helpers are not run in devtools::load_all()). Files starting with teardown are run after all tests are completed (#589).
All files are now read and written as UTF-8 (#510, #605).
is_testing() allows you to tell if your code is being run inside a testing environment (#631). Rather than taking a run-time dependency on testthat you may want to inline the function into your own package:
It’s frequently useful to combine with interactive().
A new default reporter, ReporterProgress, produces more aesthetically pleasing output and makes the most important information available upfront (#529). You can return to the previous default by setting options(testthat.default_reporter = "summary").
Output colours have been tweaked to be consistent with clang: warnings are now in magenta, and skips in blue.
New default_reporter() and check_reporter() which returns the default reporters for interactive and check environments (#504).
New DebugReporter that calls a better version of recover() in case of failures, errors, or warnings (#360, #470).
New JunitReporter generates reports in JUnit compatible format.
(#481, @lbartnik; #640, @nealrichardson; #575)
New LocationReporter which just prints the location of every expectation. This is useful for locating segfaults and C/C++ breakpoints (#551).
SummaryReporter recieved a number of smaller tweaks
Aborts testing as soon the limit given by the option testthat.summary.max_reports (default 10) is reached (#520).
New option testthat.summary.omit_dots = TRUE hides the progress dots speeding up tests by a small amount (#502).
Bring back random praise and encouragement which I accidentally dropped (#478).
New option testthat.default_check_reporter, defaults to "check". Continuous Integration system can set this option before evaluating package test sources in order to direct test result details to known location.
All reporters now accept a file argument on initialization. If provided, reporters will write the test results to that path. This output destination can also be controlled with the option testthat.output_file (#635, @nealrichardson).
is_null() and matches() have been deprecated because they conflict with other functions in the tidyverse (#523).Updated Catch to 1.9.6. testthat now understands and makes use of the package routine registration mechanism required by CRAN with R >= 3.4.0.
Better reporting for deeply nested failures, limiting the stack trace to the first and last 10 entries (#474).
Bare expectations notify the reporter once again. This is achieved by running all tests inside test_code() by default (#427, #498). This behaviour can be overridden by setting wrap = FALSE in test_dir() and friends (#586).
auto_test() and auto_test_package() provide hash parameter to enable switching to faster, time-stamp-based modification detection (#598, @katrinleinweber). auto_test_package() works correctly on windows (#465).
capture_output_lines() is now exported (#504).
compare.character() works correctly for vectors of length > 5 (#513, @brodieG)
compare.default() gains a max_diffs argument and defaults to printing out only the first 9 differences (#538).
compare.numeric() respects check.attributes() so expect_equivalent() correctly ignores attributes of numeric vectors (#485).
Output expectations (expect_output(), expect_message(), expect_warning(), and expect_silent()) all invisibly return the first argument to be consistent with the other expectations (#615).
expect_length() works with any object that has a length method, not just vectors (#564, @nealrichardson)
expect_match() now accepts explicit perl and fixed arguments, and adapts the failure message to the value of fixed. This also affects other expectations that forward to expect_match(), like expect_output(), expect_message(), expect_warning(), and expect_error().
expect_match() escapes special regular expression characters when printing (#522, @jimhester).
expect_message(), expect_warning() and expect_error() produce clearer failure messages.
find_test_scripts() only looks for \.[rR] in the extension (#492, @brodieG)
test_dir(), test_package(), test_check() unset the R_TESTS env var (#603)
test_examples() now works with installed packages as well as source packages (@jimhester, #532).
test_dir(), test_package(), and test_check() gain stop_on_failure and stop_on_waring arguments that control whether or not an error is signalled if any tests fail or generate warnings (#609, #619).
test_file() now triggers a gc() after tests are run. This helps to ensure that finalisers are run earlier (#535).
test_path() now generates correct path when called from within tools::testInstalledPackage() (#542).
test_path() no longer assumes that the path exists (#448).
test_that() calls without any expectations generate a default skip() (#413).
test_dir() gains load_helpers argument (#505).
show_failures() simply prints a failure if it occurs. This makes it easier to show failures in examples.
with_mock() disallows mocking of functions in base packages, because this doesn’t work with the current development version of R (#553).
std::logic_error() constructed with std::string() argument, to avoid build errors on Solaris.New expect_output_file() to compare output of a function with a text file, and optionally update it (#443, @krlmlr).
Properly scoped use + compilation of C++ unit testing code using Catch to gcc and clang only, as Catch includes code that does not strictly conform to the C++98 standard. (2)
Fixed an out-of-bounds memory access when routing Catch output through Rprintf(). (2)
Ensure that unit tests run on R-oldrel (remove use of dir.exists()).
Improved overriding of calls to exit() within Catch, to ensure compatibility with GCC 6.0. (@krlmlr)
Hardened formatting of difference messages, previously the presence of % characters could affect the output (#446, @krlmlr).
Fixed errors in expect_equal() when comparing numeric vectors with and without attributes (#453, @krlmlr).
auto_test() and auto_test_package() show only the results of the current test run and not of previously failed runs (#456, @krlmlr).
The expectation() function now expects an expectation type (one of “success”, “failure”, “error”, “skip”, “warning”) as first argument. If you’re creating your own expectations, you’ll need to use expect() instead (#437).
The expectation system got a thorough overhaul (#217). This primarily makes it easier to add new expectations in the future, but also included a thorough review of the documentation, ensuring that related expectations are documented together, and have evocative names.
One useful change is that most expectations invisibly return the input object. This makes it possible to chain together expectations with magrittr:
(And to make this style even easier, testthat now re-exports the pipe, #412).
The exception to this rule are the expectations that evaluate (i.e. for messages, warnings, errors, output etc), which invisibly return NULL. These functions are now more consistent: using NA will cause a failure if there is a errors/warnings/mesages/output (i.e. they’re not missing), and will NULL fail if there aren’t any errors/warnings/mesages/output. This previously didn’t work for expect_output() (#323), and the error messages were confusing with expect_error(..., NA) (#342, @nealrichardson + @krlmlr, #317).
Another change is that expect_output() now requires you to explicitly print the output if you want to test a print method: expect_output("a", "a") will fail, expect_output(print("a"), "a") will succeed.
There are six new expectations:
expect_type() checks the type of the object (#316), expect_s3_class() tests that an object is S3 with given class, expect_s4_class() tests that an object is S4 with given class (#373). I recommend using these more specific expectations instead of the more general expect_is().
expect_length() checks that an object has expected length.
expect_success() and expect_failure() are new expectations designed specifically for testing other expectations (#368).
A number of older features have been deprecated:
expect_more_than() and expect_less_than() have been deprecated. Please use expect_gt() and expect_lt() instead.
takes_less_than() has been deprecated.
not() has been deprecated. Please use the explicit individual forms expect_error(..., NA) , expect_warning(.., NA) and so on.
Now all expectations are also conditions, and R’s condition system is used to signal failures and successes (#360, @krlmlr). All known conditions (currently, “error”, “warning”, “message”, “failure”, and “success”) are converted to expectations using the new as.expectation(). This allows third-party test packages (such as assertthat, testit, ensurer, checkmate, assertive) to seamlessly establish testthat compatibility by issuing custom error conditions (e.g., structure(list(message = "Error message"), class = c("customError", "error", "condition"))) and then implementing as.expectation.customError(). The assertthat package contains an example.
The reporters system class has been considerably refactored to make existing reporters simpler and to make it easier to write new reporters. There are two main changes:
Reporters classes are now R6 classes instead of Reference Classes.
Each callbacks receive the full context:
add_results() is passed context and test as well as the expectation.test_start() and test_end() both get the context and test.context_start() and context_end() get the context.Warnings are now captured and reported in most reporters.
The reporter output goes to the original standard output and is not affected by sink() and expect_output() (#420, @krlmlr).
The default summary reporter lists all warnings (#310), and all skipped tests (@krlmlr, #343). New option testthat.summary.max_reports limits the number of reports printed by the summary reporter. The default is 15 (@krlmlr, #354).
MinimalReporter correct labels errors with E and failures with F (#311).
New FailReporter to stop in case of failures or errors after all tests (#308, @krlmlr).
New functions capture_output(), capture_message(), and capture_warnings() selectively capture function output. These are used in expect_output(), expect_message() and expect_warning() to allow other types out output to percolate up (#410).
try_again() allows you to retry code multiple times until it succeeds (#240).
test_file(), test_check(), and test_package() now attach testthat so all testing functions are available.
source_test_helpers() gets a useful default path: the testthat tests directory. It defaults to the test_env() to be consistent with the other source functions (#415).
test_file() now loads helpers in the test directory before running the tests (#350).
test_path() makes it possible to create paths to files in tests/testthat that work interactively and when called from tests (#345).
Add skip_if_not() helper.
Add skip_on_bioc() helper (@thomasp85).
make_expectation() uses expect_equal().
setup_test_dir() has been removed. If you used it previously, instead use source_test_helpers() and find_test_scripts().
source_file() exports the function testthat uses to load files from disk.
test_that() returns a logical that indicates if all tests were successful (#360, @krlmlr).
find_reporter() (and also all high-level testing functions) support a vector of reporters. For more than one reporter, a MultiReporter is created (#307, @krlmlr).
with_reporter() is used internally and gains new argument start_end_reporter = TRUE (@krlmlr, 355).
set_reporter() returns old reporter invisibly (#358, @krlmlr).
Comparing integers to non-numbers doesn’t raise errors anymore, and falls back to string comparison if objects have different lengths. Complex numbers are compared using the same routine (#309, @krlmlr).
compare.numeric() and compare.character() received another overhaul. This should improve behaviour of edge cases, and provides a strong foundation for further work. Added compare.POSIXt() for better reporting of datetime differences.
expect_identical() and is_identical_to() now use compare() for more detailed output of differences (#319, @krlmlr).
Added Catch v1.2.1 for unit testing of C++ code. See ?use_catch() for more details. (2)
Handle skipped tests in the TAP reporter (#262).
New expect_silent() ensures that code produces no output, messages, or warnings (#261).
New expect_lt(), expect_lte(), expect_gt() and expect_gte() for comparison with or without equality (#305, @krlmlr).
expect_output(), expect_message(), expect_warning(), and expect_error() now accept NA as the second argument to indicate that output, messages, warnings, and errors should be absent (#219).
Praise gets more diverse thanks to the praise package, and you’ll now get random encouragment if your tests don’t pass.
testthat no longer muffles warning messages. If you don’t want to see them in your output, you need to explicitly quiet them, or use an expectation that captures them (e.g. expect_warning()). (#254)
Use tests in inst/tests is formally deprecated. Please move them into tests/testthat instead (#231).
expect_match() now encodes the match, as well as the output, in the expectation message (#232).
expect_is() gives better failure message when testing multiple inheritance, e.g. expect_is(1:10, c("glm", "lm")) (#293).
Corrected argument order in compare.numeric() (#294).
comparison() constructure now checks its arguments are the correct type and length. This bugs a bug where tests failed with an error like “values must be length 1, but FUN(X[[1]]) result is length 2” (#279).
Added skip_on_os(), to skip tests on specified operating systems (2).
Skip test that depends on devtools if it is not installed (#247, @krlmlr)
Added skip_on_appveyor() to skip tests on Appveyor (@lmullen).
compare() shows detailed output of differences for character vectors of different length (#274, @krlmlr).
Detailed output from expect_equal() doesn’t confuse expected and actual values anymore (#274, @krlmlr).
Failure locations are now formated as R error locations.
Add an ‘invert’ argument to find_tests_scripts(). This allows one to select only tests which do not match a pattern. (#239, @jimhester).
Deprecated library_if_available() has been removed.
test (test_dir(), test_file(), test_package(), test_check()) functions now return a testthat_results object that contains all results, and can be printed or converted to data frame.
test_dir(), test_package(), and test_check() have an added ... argument that allows filtering of test files using, e.g., Perl-style regular expressions,or fixed character filtering. Arguments in ... are passed to grepl() (@leeper).
test_check() uses a new reporter specifically designed for R CMD check. It displays a summary at the end of the tests, designed to be <13 lines long so test failures in R CMD check display something more useful. This will hopefully stop BDR from calling testthat a “test obfuscation suite” (#201).
compare() is now documented and exported. Added a numeric method so when long numeric vectors don’t match you’ll see some examples of where the problem is (#177). The line spacing in compare.character() was tweaked.
skip_if_not_installed() skips tests if a package isn’t installed (#192).
expect_that(a, equals(b)) style of testing has been soft-deprecated. It will keep working, but it’s no longer demonstrated any where, and new expectations will only be available in expect_equal(a, b) style. (#172)
Once again, testthat suppresses messages and warnings in tests (#189)
New test_examples() lets you run package examples as tests. Each example counts as one expectation and it succeeds if the code runs without errors (#204).
New succeed() expectation always succeeds.
skip_on_travis() allows you to skip tests when run on Travis CI. (Thanks to @mllg)
colourise() was removed. (Colour is still supported, via the crayon package.)
Mocks can now access values local to the call of with_mock (#193, @krlmlr).
All equality expectations are now documented together (#173); all matching expectations are also documented together.
BDD: testhat now comes with an initial behaviour driven development (BDD) interface. The language is similiar to RSpec for Ruby or Mocha for JavaScript. BDD tests read like sentences, so they should make it easier to understand the specification of a function. See ?describe() for further information and examples.
It’s now possible to skip() a test with an informative message - this is useful when tests are only available under certain conditions, as when not on CRAN, or when an internet connection is available (#141).
skip_on_cran() allows you to skip tests when run on CRAN. To take advantage of this code, you’ll need either to use devtools, or run Sys.setenv(NOT_CRAN = "true"))
Simple mocking: with_mock() makes it easy to temporarily replace functions defined in packages. This is useful for testing code that relies on functions that are slow, have unintended side effects or access resources that may not be available when testing (#159, @krlmlr).
A new expectation, expect_equal_to_reference() has been added. It tests for equality to a reference value stored in a file (#148, @jonclayden).
auto_test_package() works once more, and now uses devtools::load_all() for higher fidelity loading (#138, #151).
Bug in compare.character() fixed, as reported by Georgi Boshnakov.
colourise() now uses option testthat.use_colours (default: TRUE). If it is FALSE, output is not colourised (#153, @mbojan).
is_identical_to() only calls all.equal() to generate an informative error message if the two objects are not identical (#165).
safe_digest() uses a better strategy, and returns NA for directories (#138, #146).
Random praise is renabled by default (again!) (#164).
Teamcity reporter now correctly escapes output messages (#150, @windelinckx). It also uses nested suites to include test names.
library_if_available() has been deprecated.Better default environment for test_check() and test_package() which allows S4 class creation in tests
compare.character() no longer fails when one value is missing.
testthat 0.8 comes with a new recommended structure for storing your tests. To better meet CRAN recommended practices, testthat now recommend that you to put your tests in tests/testthat, instead of inst/tests (this makes it possible for users to choose whether or not to install tests). With this new structure, you’ll need to use test_check() instead of test_packages() in the test file (usually tests/testthat.R) that runs all testthat unit tests.
The other big improvement to usability comes from @kforner, who contributed code to allow the default results (i.e. those produced by SummaryReporter) to include source references so you can see exactly where failures occured.
MultiReporter, which combines several reporters into one. (Thanks to @kforner)
ListReporter, which captures all test results with their file, context, test and elapsed time. test_dir, test_file, test_package and test_check now use the ListReporter to invisibly return a summary of the tests as a data frame. (Thanks to @kforner)
TeamCityReporter to produce output compatible with the TeamCity continuous integration environment. (Thanks to @windelinckx)
SilentReporter so that testthat can test calls to test_that. (Thanks to @craigcitro, #83)
expect_null() and is_null to check if an object is NULL (#78)
expect_named() and has_names() to check the names of a vector (#79)
expect_more_than(), is_more_than(), expect_less_than(), is_less_than() to check values above or below a threshold. (#77, thanks to @jknowles)
expect_that() (and thus all expect_* functions) now invisibly return the expectation result, and stops if info or label arguments have length > 1 (thanks to @kforner)
fixed two bugs with source_dir(): it did not look for the source scripts at the right place, and it did not use its chdir argument.
When using expect_equal() to compare strings, the default output for failure provides a lot more information, which should hopefully help make finding string mismatches easier.
SummaryReporter has a max_reports option to limit the number of detailed failure reports to show. (Thanks to @crowding)
Tracebacks will now also contain information about where the functions came from (where that information is available).
matches and expect_match now pass additional arguments on to grepl so that you can use fixed = TRUE, perl = TRUE or ignore.case = TRUE to control details of the match. expect_match now correctly fails to match NULL. (#100)
expect_output, expect_message, expect_warning and expect_error also pass … on to grepl, so that you can use fixed = TRUE, perl = TRUE or ignore.case = TRUE
Removed stringr and evaluate dependencies.
The not() function makes it possible to negate tests. For example, expect_that(f(), not(throws_error())) asserts that f() does not throw an error.
Make dir_state less race-y. (Thanks to @craigcitro, #80)
auto_test now pays attention to its ‘reporter’ argument (Thanks to @crowding, #81)
get_reporter(), set_reporter() and with_reporter() are now exported (#102)
Ignore attributes in is_true and is_false (#49)
make_expectation works for more types of input (#52)
Now works better with evaluate 0.4.3.
new fail() function always forces a failure in a test. Suggested by Richie Cotton (#47)
Added TapReporter to produce output compatible with the “test anything protocol”. Contributed by Dan Keshet.
Fixed where auto_test would identify the wrong files as having changed. (Thanks to Peter Meilstrup)
SummaryReporter: still return informative messages even if no tests defined (just bare expectations). (Fixes #31)
Improvements to reference classes (Thanks to John Chambers)
Bug fixes for when nothing was generated in gives_warning / shows_message. (Thanks to Bernd Bischl)
New make_expectation function to programmatically generate an equality expectation. (Fixes #24)
SummaryReporter: You don’t get praise until you have some tests.
Depend on methods rather than requiring it so that testthat works when run from Rscript
auto_test now normalises paths to enable better identification of file changes, and fixes bug in instantiating new reporter object.
All mutatr classes have been replaced with ReferenceClasses.
Better documentation for short-hand expectations.
test_dir and test_package gain new filter argument which allows you to restrict which tests are run.
autotest correctly loads code and executes tests in same environment
contexts are never closed before they are opened, and always closed at the end of file
fixed small bug in test_dir where each test was not given its own environment
all expect_* short cut functions gain a label argument, thanks to Steve Lianoglou
all expectations now have a shortcut form, so instead of expect_that(a, is_identical_to(b)) you can do expect_identical(a, b)
new shows_message and gives_warning expectations to test warnings and messages
expect_that, equals, is_identical_to and is_equivalent to now have additional label argument which allows you to control the appearance of the text used for the expected object (for expect_that) and actual object (for all other functions) in failure messages. This is useful when you have loops that run tests as otherwise all the variable names are identical, and it’s difficult to tell which iteration caused the failure.
executing bare tests gives nicer output
all expectations now give more information on failure to make it easier to track down the problem.
test_file and test_dir now run in code in separate environment to avoid pollution of global environment. They also temporary change the working directory so tests can use relative paths.
test_package makes it easier to run all tests in an installed package. Code run in this manner has access to non-exported functions and objects. If any errors or failures occur, test_package will throw an error, making it suitable for use with R CMD check.
colourise also works in screen terminal
equals expectation provides more information about failure
expect_that has extra info argument to allow you to pass in any extra information you’d like included in the message - this is very helpful if you’re using a loop to run tests
is_equivalent_to: new expectation that tests for equality ignoring attributes
library_if_available now works! (thanks to report and fix from Felix Andrews)
specify larger width and join pieces back together whenever deparse used (thanks to report and fix from Felix Andrews)
test_dir now looks for any files starting with test (not test- as before)