#|----------------------------------------------------------------------------| | Copyright 1990 by the Massachusetts Institute of Technology, Cambridge MA. | | | | Permission to use, copy, modify, and distribute this software and its | | documentation for any purpose and without fee is hereby granted, provided | | that this copyright and permission notice appear in all copies and | | supporting documentation, and that the name of M.I.T. not be used in | | advertising or publicity pertaining to distribution of the software | | without specific, written prior permission. M.I.T. makes no | | representations about the suitability of this software for any purpose. | | It is provided "as is" without express or implied warranty. | | | | M.I.T. DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING | | ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL | | M.I.T. BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR | | ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, | | WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, | | ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS | | SOFTWARE. | |----------------------------------------------------------------------------|# (This is the December 19, 1990 version of brief documentation for the RT regression tester. A more complete discussion can be found in the article in Lisp Pointers.) The functions, macros, and variables that make up the RT regression tester are in a package called "RT". The ten exported symbols are documented below. If you want to refer to these symbols without a package prefix, you have to `use' the package. The basic unit of concern of RT is the test. Each test has an identifying name and a body that specifies the action of the test. Functions are provided for defining, redefining, removing, and performing individual tests and the test suite as a whole. In addition, information is maintained about which tests have succeeded and which have failed. <> deftest NAME FORM &rest VALUES Individual tests are defined using the macro DEFTEST. The identifying NAME is typically a number or symbol, but can be any Lisp form. If the test suite already contains a test with the same (EQUAL) NAME, then this test is redefined and a warning message printed. (This warning is important to alert the user when a test suite definition file contains two tests with the same name.) When the test is a new one, it is added to the end of the suite. In either case, NAME is returned as the value of DEFTEST and stored in the variable *TEST*. (deftest t-1 (floor 15/7) 2 1/7) => t-1 (deftest (t 2) (list 1) (1)) => (t 2) (deftest bad (1+ 1) 1) => bad (deftest good (1+ 1) 2) => good The FORM can be any kind of Lisp form. The zero or more VALUES can be any kind of Lisp objects. The test is performed by evaluating FORM and comparing the results with the VALUES. The test succeeds if and only if FORM produces the correct number of results and each one is EQUAL to the corresponding VALUE. <> *test* NAME-OF-CURRENT-TEST The variable *TEST* contains the name of the test most recently defined or performed. It is set by DEFTEST and DO-TEST. <> do-test &optional (NAME *TEST*) The function DO-TEST performs the test identified by NAME, which defaults to *TEST*. Before running the test, DO-TEST stores NAME in the variable *TEST*. If the test succeeds, DO-TEST returns NAME as its value. If the test fails, DO-TEST returns NIL, after printing an error report on *STANDARD-OUTPUT*. The following examples show the results of performing two of the tests defined above. (do-test '(t 2)) => (t 2) (do-test 'bad) => nil ; after printing: Test BAD failed Form: (1+ 1) Expected value: 1 Actual value: 2. <> *do-tests-when-defined* default value NIL If the value of this variable is non-null, each test is performed at the moment that it is defined. This is helpful when interactively constructing a suite of tests. However, when loading a test suite for later use, performing tests as they are defined is not liable to be helpful. <> get-test &optional (NAME *TEST*) This function returns the NAME, FORM, and VALUES of the specified test. (get-test '(t 2)) => ((t 2) (list 1) (1)) <> rem-test &optional (NAME *TEST*) If the indicated test is in the test suite, this function removes it and returns NAME. Otherwise, NIL is returned. <> rem-all-tests This function reinitializes RT by removing every test from the test suite and returns NIL. Generally, it is advisable for the whole test suite to apply to some one system. When switching from testing one system to testing another, it is wise to remove all the old tests before beginning to define new ones. <> do-tests &optional (OUT *STANDARD-OUTPUT*) This function uses DO-TEST to run each of the tests in the test suite and prints a report of the results on OUT, which can either be an output stream or the name of a file. If OUT is omitted, it defaults to *STANDARD-OUTPUT*. DO-TESTS returns T if every test succeeded and NIL if any test failed. As illustrated below, the first line of the report produced by DO-TEST shows how many tests need to be performed. The last line shows how many tests failed and lists their names. While the tests are being performed, DO-TESTS prints the names of the successful tests and the error reports from the unsuccessful tests. (do-tests "report.txt") => nil ; the file "report.txt" contains: Doing 4 pending tests of 4 tests total. T-1 (T 2) Test BAD failed Form: (1+ 1) Expected value: 1 Actual value: 2. GOOD 1 out of 4 total tests failed: BAD. It is best if the individual tests in the suite are totally independent of each other. However, should the need arise for some interdependence, you can rely on the fact that DO-TESTS will run tests in the order they were originally defined. <> pending-tests When a test is defined or redefined, it is marked as pending. In addition, DO-TEST marks the test to be run as pending before running it and DO-TESTS marks every test as pending before running any of them. The only time a test is marked as not pending is when it completes successfully. The function PENDING-TESTS returns a list of the names of the currently pending tests. (pending-tests) => (bad) <> continue-testing This function is identical to DO-TESTS except that it only runs the tests that are pending and always writes its output on *STANDARD-OUTPUT*. (continue-testing) => nil ; after printing: Doing 1 pending test out of 4 total tests. Test BAD failed Form: (1+ 1) Expected value: 1 Actual value: 2. 1 out of 4 total tests failed: BAD. CONTINUE-TESTING has a special meaning if called at a breakpoint generated while a test is being performed. The failure of a test to return the correct value does not trigger an error break. However, there are many kinds of things that can go wrong while a test is being performed (e.g., dividing by zero) that will cause breaks. If CONTINUE-TESTING is evaluated in a break generated during testing, it aborts the current test (which remains pending) and forces the processing of tests to continue. Note that in such a breakpoint, *TEST* is bound to the name of the test being performed and (GET-TEST) can be used to look at the test. When building a system, it is advisable to start constructing a test suite for it as soon as possible. Since individual tests are rather weak, a comprehensive test suite requires large numbers of tests. However, these can be accumulated over time. In particular, whenever a bug is found by some means other than testing, it is wise to add a test that would have found the bug and therefore will ensure that the bug will not reappear. Every time the system is changed, the entire test suite should be run to make sure that no unintended changes have occurred. Typically, some tests will fail. Sometimes, this merely means that tests have to be changed to reflect changes in the system's specification. Other times, it indicates bugs that have to be tracked down and fixed. During this phase, CONTINUE-TESTING is useful for focusing on the tests that are failing. However, for safety sake, it is always wise to reinitialize RT, redefine the entire test suite, and run DO-TESTS one more time after you think all of the tests are working.