Author

Per Bothner <[email protected]>

Status

• Received: 2005/01/07
• Draft: 2005/01/28 - 2005/03/28
• Revised: 2005/10/18
• Revised: 2006/02/24
• Final: 2006/06/18

Abstract

This defines an API for writing test suites, to make it easy to portably test Scheme APIs, libraries, applications, and implementations. A test suite is a collection of test cases that execute in the context of a test-runner. This specifications also supports writing new test-runners, to allow customization of reporting and processing the result of running test suites.

Rationale

The Scheme community needs a standard for writing test suites. Every SRFI or other library should come with a test suite. Such a test suite must be portable, without requiring any non-standard features, such as modules. The test suite implementation or "runner" need not be portable, but it is desirable that it be possible to write a portable basic implementation.

There are other testing frameworks written in Scheme, including SchemeUnit. However SchemeUnit is not portable. It is also a bit on the verbose side. It would be useful to have a bridge between this framework and SchemeUnit so SchemeUnit tests could run under this framework and vice versa. There exists also at least one Scheme wrapper providing a Scheme interface to the standard JUnit API for Java. It would be useful to have a bridge so that tests written using this framework can run under a JUnit runner. Neither of these features are part of this specification.

This API makes use of implicit dynamic state, including an implicit test runner. This makes the API convenient and terse to use, but it may be a little less elegant and compositional than using explicit test objects, such as JUnit-style frameworks. It is not claimed to follow either object-oriented or functional design principles, but I hope it is useful and convenient to use and extend.

This proposal allows converting a Scheme source file to a test suite by just adding a few macros. You don't have to write the entire file in a new form, thus you don't have to re-indent it.

All names defined by the API start with the prefix test-. All function-like forms are defined as syntax. They may be implemented as functions or macros or built-ins. The reason for specifying them as syntax is to allow specific tests to be skipped without evaluating sub-expressions, or for implementations to add features such as printing line numbers or catching exceptions.

Specification

While this is a moderately complex specification, you should be able to write simple test suites after just reading the first few sections below. More advanced functionality, such as writing a custom test-runner, is at the end of the specification.

Writing basic test suites

Let's start with a simple example. This is a complete self-contained test-suite.

;; Initialize and give a name to a simple testsuite.
(test-begin "vec-test")
(define v (make-vector 5 99))
;; Require that an expression evaluate to true.
(test-assert (vector? v))
;; Test that an expression is eqv? to some other expression.
(test-eqv 99 (vector-ref v 2))
(vector-set! v 2 7)
(test-eqv 7 (vector-ref v 2))
;; Finish the testsuite, and report results.
(test-end "vec-test")

This testsuite could be saved in its own source file. Nothing else is needed: We do not require any top-level forms, so it is easy to wrap an existing program or test to this form, without adding indentation. It is also easy to add new tests, without having to name individual tests (though that is optional).

Test cases are executed in the context of a test runner, which is a object that accumulates and reports test results. This specification defines how to create and use custom test runners, but implementations should also provide a default test runner. It is suggested (but not required) that loading the above file in a top-level environment will cause the tests to be executed using an implementation-specified default test runner, and test-end will cause a summary to be displayed in an implementation-specified manner.

Simple test-cases

Primitive test cases test that a given condition is true. They may have a name. The core test case form is test-assert:

(test-assert [test-name] expression)

This evaluates the expression. The test passes if the result is true; if the result is false, a test failure is reported. The test also fails if an exception is raised, assuming the implementation has a way to catch exceptions. How the failure is reported depends on the test runner environment. The test-name is a string that names the test case. (Though the test-name is a string literal in the examples, it is an expression. It is evaluated only once.) It is used when reporting errors, and also when skipping tests, as described below. It is an error to invoke test-assert if there is no current test runner.

The following forms may be more convenient than using test-assert directly:

(test-eqv [test-name] expected test-expr)

This is equivalent to:

(test-assert [test-name] (eqv? expected test-expr))

Similarly test-equal and test-eq are shorthand for test-assert combined with equal? or eq?, respectively:

(test-equal [test-name] expected test-expr)
(test-eq [test-name] expected test-expr)

Here is a simple example:

(define (mean x y) (/ (+ x y) 2.0))
(test-eqv 4 (mean 3 5))

For testing approximate equality of inexact reals we can use test-approximate:

(test-approximate [test-name] expected test-expr error)

This is equivalent to (except that each argument is only evaluated once):

(test-assert [test-name]
  (and (>= test-expr (- expected error))
       (<= test-expr (+ expected error))))

Tests for catching errors

We need a way to specify that evaluation should fail. This verifies that errors are detected when required.

(test-error [[test-name] error-type] test-expr)

Evaluating test-expr is expected to signal an error. The kind of error is indicated by error-type.

If the error-type is left out, or it is #t, it means "some kind of unspecified error should be signaled". For example:

(test-error #t (vector-ref '#(1 2) 9))

This specification leaves it implementation-defined (or for a future specification) what form test-error may take, though all implementations must allow #t. Some implementations may support SRFI-35's conditions, but these are only standardized for SRFI-36's I/O conditions, which are seldom useful in test suites. An implementation may also allow implementation-specific exception types. For example Java-based implementations may allow the names of Java exception classes:

;; Kawa-specific example
(test-error <java.lang.IndexOutOfBoundsException> (vector-ref '#(1 2) 9))

An implementation that cannot catch exceptions should skip test-error forms.

Testing syntax

Testing syntax is tricky, especially if we want to check that invalid syntax is causes an error. The following utility function can help:

(test-read-eval-string string)

This function parses string (using read) and evaluates the result. The result of evaluation is returned from test-read-eval-string. An error is signalled if there are unread characters after the read is done. For example:
(test-read-eval-string "(+ 3 4)") evaluates to 7.
(test-read-eval-string "(+ 3 4") signals an error.
(test-read-eval-string "(+ 3 4) ") signals an error, because there is extra junk (i.e. a space) after the list is read.

The test-read-eval-string used in tests:

(test-equal 7 (test-read-eval-string "(+ 3 4)"))
(test-error (test-read-eval-string "(+ 3"))
(test-equal #\newline (test-read-eval-string "#\\newline"))
(test-error (test-read-eval-string "#\\newlin"))

;; Skip the next 2 tests unless srfi-62 is available.
(test-skip (cond-expand (srfi-62 0) (else 2)))
(test-equal 5 (test-read-eval-string "(+ 1 #;(* 2 3) 4)"))
(test-equal '(x z) (test-read-string "(list 'x #;'y 'z)"))

Test groups and paths

A test group is a named sequence of forms containing testcases, expressions, and definitions. Entering a group sets the test group name; leaving a group restores the previous group name. These are dynamic (run-time) operations, and a group has no other effect or identity. Test groups are informal groupings: they are neither Scheme values, nor are they syntactic forms.

A test group may contain nested inner test groups. The test group path is a list of the currently-active (entered) test group names, oldest (outermost) first.

(test-begin suite-name [count])

A test-begin enters a new test group. The suite-name becomes the current test group name, and is added to the end of the test group path. Portable test suites should use a sting literal for suite-name; the effect of expressions or other kinds of literals is unspecified.

Rationale: In some ways using symbols would be preferable. However, we want human-readable names, and standard Scheme does not provide a way to include spaces or mixed-case text in literal symbols.

The optional count must match the number of test-cases executed by this group. (Nested test groups count as a single test case for this count.) This extra test may be useful to catch cases where a test doesn't get executed because of some unexpected error.

Additionally, if there is no currently executing test runner, one is installed in an implementation-defined manner.

(test-end [suite-name])

A test-end leaves the current test group. An error is reported if the suite-name does not match the current test group name.

Additionally, if the matching test-begin installed a new test-runner, then the test-end will de-install it, after reporting the accumulated test results in an implementation-defined manner.

(test-group suite-name decl-or-expr ...)

Equivalent to:

(if (not (test-to-skip% suite-name))
  (dynamic-wind
    (lambda () (test-begin suite-name))
    (lambda () decl-or-expr ...)
    (lambda () (test-end suite-name))))

This is usually equivalent to executing the decl-or-exprs within the named test group. However, the entire group is skipped if it matched an active test-skip (see later). Also, the test-end is executed in case of an exception.

Handling set-up and cleanup

(test-group-with-cleanup suite-name
  decl-or-expr ...
  cleanup-form)

Execute each of the decl-or-expr forms in order (as in a <body>), and then execute the cleanup-form. The latter should be executed even if one of a decl-or-expr forms raises an exception (assuming the implementation has a way to catch exceptions).

For example:

(test-group-with-cleanup "test-file"
  (define f (open-output-file "log"))
  (do-a-bunch-of-tests f)
  (close-output-port f))

Conditonal test-suites and other advanced features

The following describes features for controlling which tests to execute, or specifing that some tests are expected to fail.

Test specifiers

Sometimes we want to only run certain tests, or we know that certain tests are expected to fail. A test specifier is one-argument function that takes a test-runner and returns a boolean. The specifier may be run before a test is performed, and the result may control whether the test is executed. For convenience, a specifier may also be a non-procedure value, which is coerced to a specifier procedure, as described below for count and name.

A simple example is:

(if some-condition
  (test-skip 2)) ;; skip next 2 tests

(test-match-name name)
The resulting specifier matches if the current test name (as returned by test-runner-test-name) is equals? to name.

(test-match-nth n [count])
This evaluates to a stateful predicate: A counter keeps track of how many times it has been called. The predicate matches the n'th time it is called (where 1 is the first time), and the next (- count 1) times, where count defaults to 1.

(test-match-any specifier ...)
The resulting specifier matches if any specifier matches. Each specifier is applied, in order, so side-effects from a later specifier happen even if an earlier specifier is true.

(test-match-all specifier ...)
The resulting specifier matches if each specifier matches. Each specifier is applied, in order, so side-effects from a later specifier happen even if an earlier specifier is false.

count (i.e. an integer)
Convenience short-hand for: (test-match-nth 1 count).

name (i.e. a string)
Convenience short-hand for (test-match-name name).

Skipping selected tests

In some cases you may want to skip a test.

(test-skip specifier)

Evaluating test-skip adds the resulting specifier to the set of currently active skip-specifiers. Before each test (or test-group) the set of active skip-specifiers are applied to the active test-runner. If any specifier matches, then the test is skipped.

For convenience, if the specifier is a string that is syntactic sugar for (test-match-name specifier). For example:

(test-skip "test-b")
(test-assert "test-a")   ;; executed
(test-assert "test-b")   ;; skipped

Any skip specifiers introduced by a test-skip are removed by a following non-nested test-end.

(test-begin "group1")
(test-skip "test-a")
(test-assert "test-a")   ;; skipped
(test-end "group1")      ;; Undoes the prior test-skip
(test-assert "test-a")   ;; executed

Expected failures

Sometimes you know a test case will fail, but you don't have time to or can't fix it. Maybe a certain feature only works on certain platforms. However, you want the test-case to be there to remind you to fix it. You want to note that such tests are expected to fail.

(test-expect-fail specifier)

Matching tests (where matching is defined as in test-skip) are expected to fail. This only affects test reporting, not test execution. For example:

(test-expect-fail 2)
(test-eqv ...) ;; expected to fail
(test-eqv ...) ;; expected to fail
(test-eqv ...) ;; expected to pass

Test-runner

A test-runner is an object that runs a test-suite, and manages the state. The test group path, and the sets skip and expected-fail specifiers are part of the test-runner. A test-runner will also typically accumulate statistics about executed tests,

(test-runner? value)
True iff value is a test-runner object.

(test-runner-current)
(test-runner-current runner)
Get or set the current test-runner. If an implementation supports parameter objects (as in SRFI-39), then test-runner-current can be a parameter object. Alternatively, test-runner-current may be implemented as a macro or function that uses a fluid or thread-local variable, or a plain global variable.

(test-runner-get)
Same as (test-runner-current), buth trows an exception if there is no current test-runner.

(test-runner-simple)
Creates a new simple test-runner, that prints errors and a summary on the standard output port.

(test-runner-null)
Creates a new test-runner, that does nothing with the test results. This is mainly meant for extending when writing a custom runner.

Implementations may provide other test-runners, perhaps a (test-runner-gui).

(test-runner-create)
Create a new test-runner. Equivalent to ((test-runner-factory)).

(test-runner-factory)
(test-runner-factory factory)
Get or set the current test-runner factory. A factory is a zero-argument function that creates a new test-runner. The default value is test-runner-simple, but implementations may provide a way to override the default. As with test-runner-current, this may be a parameter object, or use a per-thread, fluid, or global variable.

Running specific tests with a specified runner

(test-apply [runner] specifier ... procedure)
Calls procedure with no arguments using the specified runner as the current test-runner. If runner is omitted, then (test-runner-current) is used. (If there is no current runner, one is created as in test-begin.) If one or more specifiers are listed then only tests matching the specifiers are executed. A specifier has the same form as one used for test-skip. A test is executed if it matches any of the specifiers in the test-apply and does not match any active test-skip specifiers.

(test-with-runner runner decl-or-expr ...)
Executes each decl-or-expr in order in a context where the current test-runner is runner.

Test results

Running a test sets various status properties in the current test-runner. This can be examined by a custom test-runner, or (more rarely) in a test-suite.

Result kind

Running a test may yield one of the following status symbols:

'pass

The passed, as expected.

'fail

The test failed (and was not expected to).

'xfail

The test failed and was expected to.

'xpass

The test passed, but was expected to fail.

'skip

The test was skipped.

(test-result-kind [runner])
Return one of the above result codes from the most recent tests. Returns #f if no tests have been run yet. If we've started on a new test, but don't have a result yet, then the result kind is 'xfail is the test is expected to fail, 'skip is the test is supposed to be skipped, or #f otherwise.

(test-passed? [runner])
True if the value of (test-result-kind [runner]) is one of 'pass or 'xpass. This is a convenient shorthand that might be useful in a test suite to only run certain tests if the previous test passed.

Test result properties

A test runner also maintains a set of more detailed result properties associated with the current or most recent test. (I.e. the properties of the most recent test are available as long as a new test hasn't started.) Each property has a name (a symbol) and a value (any value). Some properties are standard or set by the implementation; implementations can add more.

(test-result-ref runner 'pname [default])
Returns the property value associated with the pname property name. If there is no value associated with 'pname return default, or #f if default isn't specified.

(test-result-set! runner 'pname value)
Sets the property value associated with the pname property name to value. Usually implementation code should call this function, but it may be useful for a custom test-runner to add extra properties.

(test-result-remove runner 'pname)
Remove the property with the name 'pname.

(test-result-clear runner)
Remove all result properties. The implementation automatically calls test-result-clear at the start of a test-assert and similar procedures.

(test-result-alist runner)
Returns an association list of the current result properties. It is unspecified if the result shares state with the test-runner. The result should not be modified, on the other hand the result may be implicitly modified by future test-result-set! or test-result-remove calls. However, A test-result-clear does not modify the returned alist. Thus you can archive result objects from previous runs.

Standard result properties

The set of available result properties is implementation-specific. However, it is suggested that the following might be provided:

'result-kind

The result kind, as defined previously. This is the only mandatory result property.
(test-result-kind runner) is equivalent to:
(test-result-ref runner 'result-kind)

'source-file

'source-line

If known, the location of test statements (such as test-assert) in test suite source code..

'source-form

The source form, if meaningful and known.

'expected-value

The expected non-error result, if meaningful and known.

'expected-error

The error-type specified in a test-error, if it meaningful and known.

'actual-value

The actual non-error result value, if meaningful and known.

'actual-error

The error value, if an error was signalled and it is known. The actual error value is implementation-defined.

Writing a new test-runner

This section specifies how to write a test-runner. It can be ignored if you just want to write test-cases.

Call-back functions

These call-back functions are methods (in the object-oriented sense) of a test-runner. A method test-runner-on-event is called by the implementation when event happens.

To define (set) the callback function for event use the following expression. (This is normally done when initializing a test-runner.)
(test-runner-on-event! runner event-function)

An event-function takes a test-runner argument, and possibly other arguments, depending on the event.

To extract (get) the callback function for event do this:
(test-runner-on-event runner)

To extract call the callback function for event use the following expression. (This is normally done by the implementation core.)
((test-runner-on-event runner) runner other-args ...)

The following call-back hooks are available.

(test-runner-on-test-begin runner)
(test-runner-on-test-begin! runner on-test-begin-function)
(on-test-begin-function runner)
The on-test-begin-function is called at the start of an individual testcase, before the test expression (and expected value) are evaluated.

(test-runner-on-test-end runner)
(test-runner-on-test-end! runner on-test-end-function)
(on-test-end-function runner)
The on-test-end-function is called at the end of an individual testcase, when the result of the test is available.

(test-runner-on-group-begin runner)
(test-runner-on-group-begin! runner on-group-begin-function)
(on-group-begin-function runner suite-name count)
The on-group-begin-function is called by a test-begin, including at the start of a test-group. The suite-name is a Scheme string, and count is an integer or #f.

(test-runner-on-group-end runner)
(test-runner-on-group-end! runner on-group-end-function)
(on-group-end-function runner)
The on-group-end-function is called by a test-end, including at the end of a test-group.

(test-runner-on-bad-count runner)
(test-runner-on-bad-count! runner on-bad-count-function)
(on-bad-count-function runner actual-count expected-count)
Called from test-end (before the on-group-end-function is called) if an expected-count was specified by the matching test-begin and the expected-count does not match the actual-count of tests actually executed or skipped.

(test-runner-on-base-end-name runner)
(test-runner-on-bad-end-name! runner on-bad-end-name-function)
(on-bad-end-name-function runner begin-name end-name)
Called from test-end (before the on-group-end-function is called) if a suite-name was specified, and it did not that the name in the matching test-begin.

(test-runner-on-final runner)
(test-runner-on-final! runner on-final-function)
(on-final-function runner)
The on-final-function takes one parameter (a test-runner) and typically displays a summary (count) of the tests. The on-final-function is called after called the on-group-end-function correspondiong to the outermost test-end. The default value is test-on-final-simple which writes to the standard output port the number of tests of the various kinds.

The default test-runner returned by test-runner-simple uses the following call-back functions:
(test-on-test-begin-simple runner)
(test-on-test-end-simple runner)
(test-on-group-begin-simple runner suite-name count)
(test-on-group-end-simple runner)
(test-on-bad-count-simple runner actual-count expected-count)
(test-on-bad-end-name-simple runner begin-name end-name)
You can call those if you want to write a your own test-runner.

Test-runner components

The following functions are for accessing the other components of a test-runner. They would normally only be used to write a new test-runner or a match-predicate.

(test-runner-pass-count runner)
Returns the number of tests that passed, and were expected to pass.

(test-runner-fail-count runner)
Returns the number of tests that failed, but were expected to pass.

(test-runner-xpass-count runner)
Returns the number of tests that passed, but were expected to fail.

(test-runner-xfail-count runner)
Returns the number of tests that failed, and were expected to pass.

(test-runner-skip-count runner)
Returns the number of tests or test groups that were skipped.

(test-runner-test-name runner)
Returns the name of the current test or test group, as a string. During execution of test-begin this is the name of the test group; during the execution of an actual test, this is the name of the test-case. If no name was specified, the name is the empty string.

(test-runner-group-path runner)
A list of names of groups we're nested in, with the outermost group first.

(test-runner-group-stack runner)
A list of names of groups we're nested in, with the outermost group last. (This is more efficient than test-runner-group-path, since it doesn't require any copying.)

(test-runner-aux-value runner)
(test-runner-aux-value! runner on-test)
Get or set the aux-value field of a test-runner. This field is not used by this API or the test-runner-simple test-runner, but may be used by custom test-runners to store extra state.

(test-runner-reset runner)
Resets the state of the runner to its initial state.

Example

This is an example of a simple custom test-runner. Loading this program before running a test-suite will install it as the default test runner.

(define (my-simple-runner filename)
  (let ((runner (test-runner-null))
	(port (open-output-file filename))
        (num-passed 0)
        (num-failed 0))
    (test-runner-on-test! runner
      (lambda (runner result)
        (case (cdr (assq 'result-kind result))
          ((pass xpass) (set! num-passed (+ num-passed 1)))
          ((fail xfail) (set! num-failed (+ num-failed 1)))
          (else #t))))
    (test-runner-on-final! runner
       (lambda (runner)
          (format port "Passing tests: ~d.~%Failing tests: ~d.~%"
                  num-passed num-failed)
	  (close-output-port port)))
    runner))

(test-runner-factory
 (lambda () (my-simple-runner "/tmp/my-test.log")))

Implementation

The test implementation uses cond-expand (SRFI-0) to select different code depending on certain SRFI names (srfi-9, srfi-34, srfi-35, srfi-39), or implementations (kawa). It should otherwise be portable to any R5RS implementation.

testing.scm

Examples

Here is srfi-25-test.scm, based converted from Jussi Piitulainen's test.scm for SRFI-25.

Test suite

Of course we need a test suite for the testing framework itself. This suite srfi-64-test.scm was contributed by Donovan Kolbly <[email protected]>.

Copyright

Copyright (C) Per Bothner (2005, 2006)

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Last modified: Sun Jan 28 13:40:18 MET 2007