aboutsummaryrefslogblamecommitdiffstats
path: root/lib/eunit/doc/overview.edoc
blob: b4af31ae6afebacf33c65ea8bd80a49ad6196fb2 (plain) (tree)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725





                                                              
                                                     











































































































































































































































































































































































































































































































































































































































































































































                                                                          

                                                                        

                                                                            

























































































































































































                                                                        
                                                                                                       














































































































                                                                        
			-*- html -*-

	EUnit overview page

@title EUnit - a Lightweight Unit Testing Framework for Erlang

@author Richard Carlsson <carlsson.richard@gmail.com>
@author Micka�l R�mond <mickael.remond@process-one.net> 
        [http://www.process-one.net/]
@copyright 2004-2007 Micka�l R�mond, Richard Carlsson
@version {@version}, {@date} {@time}

@doc EUnit is a unit testing framework for Erlang. It is very powerful
and flexible, is easy to use, and has small syntactical overhead.

<ul>
<li>{@section Unit testing}</li>
<li>{@section Terminology}</li>
<li>{@section Getting started}</li>
<li>{@section EUnit macros}</li>
<li>{@section EUnit test representation}</li>
</ul>

EUnit builds on ideas from the family of unit testing frameworks for
Object Oriented languages that originated with JUnit by Beck and Gamma
(and Beck's previous framework SUnit for Smalltalk). However, EUnit uses
techniques more adapted to functional and concurrent programming, and is
typically less verbose than its relatives.

Although EUnit uses many preprocessor macros, they have been designed to
be as nonintrusive as possible, and should not cause conflicts with
existing code. Adding EUnit tests to a module should thus not normally
require changing existing code. Furthermore, tests that only exercise
the exported functions of a module can always be placed in a completely
separate module, avoiding any conflicts entirely.

== Unit testing ==

Unit Testing is testing of individual program "units" in relative
isolation. There is no particular size requirement: a unit can be a
function, a module, a process, or even a whole application, but the most
typical testing units are individual functions or modules. In order to
test a unit, you specify a set of individual tests, set up the smallest
necessary environment for being able to run those tests (often, you
don't need to do any setup at all), you run the tests and collect the
results, and finally you do any necessary cleanup so that the test can
be run again later. A Unit Testing Framework tries to help you in each
stage of this process, so that it is easy to write tests, easy to run
them, and easy to see which tests failed (so you can fix the bugs).

=== Advantages of unit testing ===

<dl>
  <dt>Reduces the risks of changing the program</dt>
  <dd>Most programs will be modified during their lifetime: bugs will be
  fixed, features will be added, optimizations may become necessary, or
  the code will need to be refactored or cleaned up in other ways to
  make it easier to work with. But every change to a working program is
  a risk of introducing new bugs - or reintroducing bugs that had
  previously been fixed. Having a set of unit tests that you can run
  with very little effort makes it easy to know that the code still
  works as it should (this use is called <em>regression testing</em>;
  see {@section Terminology}). This goes a long way to reduce the
  resistance to changing and refactoring code.</dd>
  <dt>Helps guide and speed up the development process</dt>
  <dd>By focusing on getting the code to pass the tests, the programmer
  can become more productive, not overspecify or get lost in premature
  optimizations, and create code that is correct from the very beginning
  (so-called <em>test-driven development</em>; see {@section
  Terminology}).</dd>
  <dt>Helps separate interface from implementation</dt>
  <dd>When writing tests, the programmer may discover dependencies
  (in order to get the tests to run) that ought not to be there, and
  which need to be abstracted away to get a cleaner design. This helps
  eliminate bad dependencies before they spread throughout the
  code.</dd>
  <dt>Makes component integration easier</dt>
  <dd>By testing in a bottom-up fashion, beginning with the smallest
  program units and creating a confidence in that they work as they
  should, it becomes easier to test that a higher-level component,
  consisting of several such units, also behaves according to
  specification (known as <em>integration testing</em>; see {@section
  Terminology}).</dd>
  <dt>Is self-documenting</dt>
  <dd>The tests can be read as documentation, typically showing both
  examples of correct and incorrect usage, along with the expected
  consequences.</dd>
</dl>

== Terminology ==

<dl>
  <dt>Unit testing</dt>
  <dd>Testing that a program unit behaves as it is supposed to do (in
  itself), according to its specifications. Unit tests have an important
  function as regression tests, when the program later is modified for
  some reason, since they check that the program still behaves according
  to specification.</dd>
  <dt>Regression testing</dt>
  <dd>Running a set of tests after making changes to a program, to check
  that the program behaves as it did before the changes (except, of
  course, for any intentional changes in behaviour). Unit tests are
  important as regression tests, but regression testing can involve more
  than just unit testing, and may also test behaviour that might not be
  part of the normal specification (such as bug-for-bug-compatibility).
  </dd>
  <dt>Integration testing</dt>
  <dd>Testing that a number of individually developed program units
  (assumed to already have been separately unit tested) work together as
  expected. Depending on the system being developed, integration testing
  may be as simple as "just another level of unit testing", but might
  also involve other kinds of tests (compare <em>system testing</em>).
</dd>
  <dt>System testing</dt>
  <dd>Testing that a complete system behaves according to its
  specification. Specifically, system testing should not require knowing
  any details about the implementation. It typically involves testing
  many different aspects of the system behaviour apart from the basic
  functionality, such as performance, usability, and reliability.</dd>
  <dt>Test-driven development</dt>
  <dd>A program development technique where you continuously write tests
  <em>before</em> you implement the code that is supposed to pass those
  tests. This can help you focus on solving the right problems, and not
  make a more complicated implementation than necessary, by letting the
  unit tests determine when a program is "done": if it fulfils its
  specifications, there is no need to keep adding functionality.</dd>
  <dt>Mock object</dt>
  <dd>Sometimes, testing some unit `A' (e.g., a function) requires that
  it collaborates somehow with some other unit `B' (perhaps being passed
  as an argument, or by reference) - but `B' has not been implemented
  yet. A "mock object" - an object which, for the purposes of testing
  `A', looks and behaves like a real `B' - might then be used instead.
  (This is of course only useful if it would be significantly more work
  to implement a real `B' than to create a mock object.)</dd>
  <dt>Test case</dt>
  <dd>A single, well-defined test, that somehow can be uniquely
  identified. When executed, the test case either <em>passes</em> or
  <em>fails</em>; the test report should identify exactly which test
  cases failed.</dd>
  <dt>Test suite</dt>
  <dd>A collection of test cases, generally with a specific, common
  target for testing, such as a single function, module, or subsystem. A
  test suite may also be recursively composed by smaller test
  suites.</dd>
</dl>

== Getting started ==
<ul>
  <li>{@section Including the EUnit header file}</li>
  <li>{@section Writing simple test functions}</li>
  <li>{@section Running EUnit}</li>
  <li>{@section Writing test generating functions}</li>
  <li>{@section An example}</li>
  <li>{@section Disabling testing}</li>
  <li>{@section Avoiding compile-time dependency on EUnit}</li>
</ul>

=== Including the EUnit header file ===

The simplest way to use EUnit in an Erlang module is to add the
following line at the beginning of the module (after the `-module'
declaration, but before any function definitions):
```-include_lib("eunit/include/eunit.hrl").'''

This will have the following effect:
<ul>
  <li>Creates an exported function `test()' (unless testing is turned
  off, and the module does not already contain a test() function), that
  can be used to run all the unit tests defined in the module</li>
  <li>Causes all functions whose names match `..._test()' or `..._test_()'
  to be automatically exported from the module (unless testing is
  turned off, or the `EUNIT_NOAUTO' macro is defined)</li>
  <li>Makes all the preprocessor macros of EUnit available, to help
  writing tests</li>
</ul>

<strong>Note:</strong> For `-include_lib(...)' to work, the Erlang
module search path <em>must</em> contain a directory whose name ends in
`eunit/ebin' (pointing to the `ebin' subdirectory of the EUnit
installation directory). If EUnit is installed as `lib/eunit' under your
Erlang/OTP system directory, its `ebin' subdirectory will be
automatically added to the search path when Erlang starts. Otherwise,
you need to add the directory explicitly, by passing a `-pa' flag to the
`erl' or `erlc' command. For example, a Makefile could contain the
following action for compiling `.erl' files:
```erlc -pa "path/to/eunit/ebin" $(ERL_COMPILE_FLAGS) -o$(EBIN) $<'''
or if you want Eunit to always be available when you run Erlang
interactively, you can add a line like the following to your
`$HOME/.erlang' file:
```code:add_path("/path/to/eunit/ebin").'''

=== Writing simple test functions ===

The EUnit framework makes it extremely easy to write unit tests in
Erlang. There are a few different ways of writing them, though, so we
start with the simplest:

A function with a name ending in `..._test()' is recognized by EUnit as
a simple test function - it takes no arguments, and its execution either
succeeds (returning some arbitrary value that EUnit will throw away), or
fails by throwing an exception of some kind (or by not terminating, in
which case it will be aborted after a while).

An example of a simple test function could be the following:
```reverse_test() -> lists:reverse([1,2,3]).'''
This just tests that the function `lists:reverse(List)' does not crash
when `List' is `[1,2,3]'. It is not a great test, but many people write
simple functions like this one to test the basic functionality of their
code, and those tests can be used directly by EUnit, without changes,
as long as their function names match.

==== Use exceptions to signal failure ====

To write more interesting tests, we need to make them crash (throw an
exception) when they don't get the result they expect. A simple way of
doing this is to use pattern matching with `=', as in the following
examples:
```reverse_nil_test() -> [] = lists:reverse([]).
   reverse_one_test() -> [1] = lists:reverse([1]).
   reverse_two_test() -> [2,1] = lists:reverse([1,2]).
'''
If there was some bug in `lists:reverse/1' that made it return something
other than `[2,1]' when it got `[1,2]' as input, then the last test
above would throw a `badmatch' error. The first two (we assume they do
not get a `badmatch') would simply return `[]' and `[1]', respectively,
so both succeed. (Note that EUnit is not psychic: if you write a test
that returns a value, even if it is the wrong value, EUnit will consider
it a success. You must make sure that the test is written so that it
causes a crash if the result is not what it should be.)

==== Using assert macros ====

If you want to use Boolean operators for your tests, the `assert'
macro comes in handy (see {@section EUnit macros} for details):
```length_test() -> ?assert(length([1,2,3]) =:= 3).'''
The `?assert(Expression)' macro will evaluate `Expression', and if that
does not evaluate to `true', it will throw an exception; otherwise it
just returns `ok'. In the above example, the test will thus fail if the
call to `length' does not return 3.

=== Running EUnit ===

If you have added the declaration
`-include_lib("eunit/include/eunit.hrl")' to your module, as described
above, you only need to compile the module, and run the automatically
exported function `test()'. For example, if your module was named `m',
then calling `m:test()' will run EUnit on all the tests defined in the
module. You do not need to write `-export' declarations for the test
functions. This is all done by magic.

You can also use the function {@link eunit:test/1} to run arbitrary
tests, for example to try out some more advanced test descriptors (see
{@section EUnit test representation}). For example, running
``eunit:test(m)'' does the same thing as the auto-generated function
``m:test()'', while ``eunit:test({inparallel, m})'' runs the same test
cases but executes them all in parallel.

==== Putting tests in separate modules ====

If you want to separate your test code from your normal code (at least
for testing the exported functions), you can simply write the test
functions in a module named `m_tests' (note: not `m_test'), if your
module is named `m'. Then, whenever you ask EUnit to test the module
`m', it will also look for the module `m_tests' and run those tests as
well. See `ModuleName' in the section {@section Primitives} for details.

==== EUnit captures standard output ====

If your test code writes to the standard output, you may be surprised to
see that the text does not appear on the console when the tests are
running. This is because EUnit captures all standard output from test
functions (this also includes setup and cleanup functions, but not
generator functions), so that it can be included in the test report if
errors occur. To bypass EUnit and print text directly to the console
while testing, you can write to the `user' output stream, as in
`io:format(user, "~w", [Term])'. The recommended way of doing this is to
use the EUnit {@section Debugging macros}, which make it much simpler.

=== Writing test generating functions ===

A drawback of simple test functions is that you must write a separate
function (with a separate name) for each test case. A more compact way
of writing tests (and much more flexible, as we shall see), is to write
functions that <em>return</em> tests, instead of <em>being</em> tests.

A function with a name ending in `..._test_()' (note the final
underscore) is recognized by EUnit as a <em>test generator</em>
function. Test generators return a <em>representation</em> of a <em>set
of tests</em> to be executed by EUnit.

==== Representing a test as data ====

The most basic representation of a test is a single fun-expression that
takes no arguments. For example, the following test generator:
```basic_test_() ->
       fun () -> ?assert(1 + 1 =:= 2) end.'''
will have the same effect as the following simple test:
```simple_test() ->
       ?assert(1 + 1 =:= 2).'''
(in fact, EUnit will handle all simple tests just like it handles
fun-expressions: it will put them in a list, and run them one by one).

==== Using macros to write tests ====

To make tests more compact and readable, as well as automatically add
information about the line number in the source code where a test
occurred (and reduce the number of characters you have to type), you can
use the `_test' macro (note the initial underscore character), like
this:
```basic_test_() ->
       ?_test(?assert(1 + 1 =:= 2)).'''
The `_test' macro takes any expression (the "body") as argument, and
places it within a fun-expression (along with some extra information).
The body can be any kind of test expression, just like the body of a
simple test function.

==== Underscore-prefixed macros create test objects ====

But this example can be made even shorter! Most test macros, such as the
family of `assert' macros, have a corresponding form with an initial
underscore character, which automatically adds a `?_test(...)' wrapper.
The above example can then simply be written:
```basic_test_() ->
       ?_assert(1 + 1 =:= 2).'''
which has exactly the same meaning (note the `_assert' instead of
`assert'). You can think of the initial underscore as signalling
<em>test object</em>.

=== An example ===

Sometimes, an example says more than a thousand words. The following
small Erlang module shows how EUnit can be used in practice.
```-module(fib).
   -export([fib/1]).
   -include_lib("eunit/include/eunit.hrl").

   fib(0) -> 1;
   fib(1) -> 1;
   fib(N) when N > 1 -> fib(N-1) + fib(N-2).

   fib_test_() ->
       [?_assert(fib(0) =:= 1),
	?_assert(fib(1) =:= 1),
	?_assert(fib(2) =:= 2),
	?_assert(fib(3) =:= 3),
	?_assert(fib(4) =:= 5),
	?_assert(fib(5) =:= 8),
	?_assertException(error, function_clause, fib(-1)),
	?_assert(fib(31) =:= 2178309)
       ].'''

(Author's note: When I first wrote this example, I happened to write a
`*' instead of `+' in the `fib' function. Of course, this showed up
immediately when I ran the tests.)

See {@section EUnit test representation} for a full list of all the ways
you can specify test sets in EUnit.

=== Disabling testing ===

Testing can be turned off by defining the `NOTEST' macro when compiling,
for example as an option to `erlc', as in:
```erlc -DNOTEST my_module.erl'''
or by adding a macro definition to the code, <em>before the EUnit header
file is included</em>:
```-define(NOTEST, 1).'''
(the value is not important, but should typically be 1 or `true').
Note that unless the `EUNIT_NOAUTO' macro is defined, disabling testing
will also automatically strip all test functions from the code, except
for any that are explicitly declared as exported.

For instance, to use EUnit in your application, but with testing turned
off by default, put the following lines in a header file:
```-define(NOTEST, true).
   -include_lib("eunit/include/eunit.hrl").'''
and then make sure that every module of your application includes that
header file. This means that you have a only a single place to modify in
order to change the default setting for testing. To override the `NOTEST'
setting without modifying the code, you can define `TEST' in a compiler
option, like this:
```erlc -DTEST my_module.erl'''

See {@section Compilation control macros} for details about these
macros.

=== Avoiding compile-time dependency on EUnit ===

If you are distributing the source code for your application for other
people to compile and run, you probably want to ensure that the code
compiles even if EUnit is not available. Like the example in the
previous section, you can put the following lines in a common header
file:
```-ifdef(TEST).
   -include_lib("eunit/include/eunit.hrl").
   -endif.'''
and, of course, also make sure that you place all test code that uses
EUnit macros within `-ifdef(TEST)' or `-ifdef(EUNIT)' sections.


== EUnit macros ==

Although all the functionality of EUnit is available even without the
use of preprocessor macros, the EUnit header file defines a number of
such macros in order to make it as easy as possible to write unit tests
as compactly as possible and without getting too many details in the
way.

Except where explicitly stated, using EUnit macros will never introduce
run-time dependencies on the EUnit library code, regardless of whether
your code is compiled with testing enabled or disabled.

<ul>
<li>{@section Basic macros}</li>
<li>{@section Compilation control macros}</li>
<li>{@section Utility macros}</li>
<li>{@section Assert macros}</li>
<li>{@section Macros for running external commands}</li>
<li>{@section Debugging macros}</li>
</ul>

=== Basic macros ===

<dl>
<dt>`_test(Expr)'</dt>
<dd>Turns `Expr' into a "test object", by wrapping it in a
fun-expression and a source line number. Technically, this is the same
as `{?LINE, fun () -> (Expr) end}'.
</dd>
</dl>

=== Compilation control macros ===

<dl>
<dt>`EUNIT'</dt>
<dd>This macro is always defined to `true' whenever EUnit is enabled at
compile time. This is typically used to place testing code within
conditional compilation, as in:
```-ifdef(EUNIT).
       % test code here
       ...
   -endif.'''
e.g., to ensure that the code can be compiled without including the
EUnit header file, when testing is disabled. See also the macros `TEST'
and `NOTEST'.
</dd>

<dt>`EUNIT_NOAUTO'</dt>
<dd>If this macro is defined, the automatic exporting or stripping of
test functions will be disabled.
</dd>

<dt>`TEST'</dt>
<dd>This macro is always defined (to `true', unless previously defined
by the user to have another value) whenever EUnit is enabled at compile
time. This can be used to place testing code within conditional
compilation; see also the macros `NOTEST' and `EUNIT'.

For testing code that is strictly dependent on EUnit, it may be
preferable to use the `EUNIT' macro for this purpose, while for code
that uses more generic testing conventions, using the `TEST' macro may
be preferred.

The `TEST' macro can also be used to override the `NOTEST' macro. If
`TEST' is defined <em>before</em> the EUnit header file is
included (even if `NOTEST' is also defined), then the code will be
compiled with EUnit enabled.
</dd>

<dt>`NOTEST'</dt>
<dd>This macro is always defined (to `true', unless previously defined
by the user to have another value) whenever EUnit is <em>disabled</em>
at compile time. (Compare the `TEST' macro.)

This macro can also be used for conditional compilation, but is more
typically used to disable testing: If `NOTEST' is defined
<em>before</em> the EUnit header file is included, and `TEST'
is <em>not</em> defined, then the code will be compiled with EUnit
disabled. See also {@section Disabling testing}.
</dd>

<dt>`NOASSERT'</dt>
<dd>If this macro is defined, the assert macros will have no effect,
when testing is also disabled. See {@section Assert macros}. When
testing is enabled, the assert macros are always enabled automatically
and cannot be disabled.
</dd>

<dt>`ASSERT'</dt>
<dd>If this macro is defined, it overrides the NOASSERT macro, forcing
the assert macros to always be enabled regardless of other settings.
</dd>

<dt>`NODEBUG'</dt>
<dd>If this macro is defined, the debugging macros will have no effect.
See {@section Debugging macros}. `NODEBUG' also implies `NOASSERT',
unless testing is enabled.
</dd>

<dt>`DEBUG'</dt>
<dd>If this macro is defined, it overrides the NODEBUG macro, forcing
the debugging macros to be enabled.
</dd>
</dl>

=== Utility macros ===

The following macros can make tests more compact and readable:

<dl>
<dt>`LET(Var,Arg,Expr)'</dt>
<dd>Creates a local binding `Var = Arg' in `Expr'. (This is the same as
`(fun(Var)->(Expr)end)(Arg)'.) Note that the binding is not exported
outside of `Expr', and that within `Expr', this binding of `Var' will
shadow any binding of `Var' in the surrounding scope.
</dd>
<dt>`IF(Cond,TrueCase,FalseCase)'</dt>
<dd>Evaluates `TrueCase' if `Cond' evaluates to `true', or otherwise
evaluates `FalseCase' if `Cond' evaluates to `false'. (This is the same
as `(case (Cond) of true->(TrueCase); false->(FalseCase) end)'.) Note
that it is an error if `Cond' does not yield a boolean value.
</dd>
</dl>

=== Assert macros ===

(Note that these macros also have corresponding forms which start with
an "`_'" (underscore) character, as in `?_assert(BoolExpr)', that create
a "test object" instead of performing the test immediately. This is
equivalent to writing `?_test(assert(BoolExpr))', etc.)

If the macro `NOASSERT' is defined before the EUnit header file is
included, these macros have no effect when testing is also disabled; see
{@section Compilation control macros} for details.

<dl>
<dt>`assert(BoolExpr)'</dt>
<dd>Evaluates the expression `BoolExpr', if testing is enabled. Unless
the result is `true', an informative exception will be generated. If
there is no exception, the result of the macro expression is the atom
`ok', and the value of `BoolExpr' is discarded. If testing is disabled,
the macro will not generate any code except the atom `ok', and
`BoolExpr' will not be evaluated.

Typical usage:
```?assert(f(X, Y) =:= [])'''

The `assert' macro can be used anywhere in a program, not just in unit
tests, to check pre/postconditions and invariants. For example:
```some_recursive_function(X, Y, Z) ->
       ?assert(X + Y > Z),
       ...'''
</dd>
<dt>`assertNot(BoolExpr)'</dt>
<dd>Equivalent to `assert(not (BoolExpr))'.
</dd>
<dt>`assertMatch(GuardedPattern, Expr)'</dt>
<dd>Evaluates `Expr' and matches the result against `GuardedPattern', if
testing is enabled. If the match fails, an informative exception will be
generated; see the `assert' macro for further details. `GuardedPattern'
can be anything that you can write on the left hand side of the `->'
symbol in a case-clause, except that it cannot contain comma-separated
guard tests.

The main reason for using `assertMatch' also for simple matches, instead
of matching with `=', is that it produces more detailed error messages.

Examples:
```?assertMatch({found, {fred, _}}, lookup(bloggs, Table))'''
```?assertMatch([X|_] when X > 0, binary_to_list(B))'''
</dd>
<dt>`assertEqual(Expect, Expr)'</dt>
<dd>Evaluates the expressions `Expect' and `Expr' and compares the
results for equality, if testing is enabled. If the values are not
equal, an informative exception will be generated; see the `assert'
macro for further details.

`assertEqual' is more suitable than than `assertMatch' when the
left-hand side is a computed value rather than a simple pattern, and
gives more details than `?assert(Expect =:= Expr)'.

Examples:
```?assertEqual("b" ++ "a", lists:reverse("ab"))'''
```?assertEqual(foo(X), bar(Y))'''
</dd>
<dt>`assertException(ClassPattern, TermPattern, Expr)'</dt>
<dt>`assertError(TermPattern, Expr)'</dt>
<dt>`assertExit(TermPattern, Expr)'</dt>
<dt>`assertThrow(TermPattern, Expr)'</dt>
<dd>Evaluates `Expr', catching any exception and testing that it matches
the expected `ClassPattern:TermPattern'. If the match fails, or if no
exception is thrown by `Expr', an informative exception will be
generated; see the `assert' macro for further details. The
`assertError', `assertExit', and `assertThrow' macros, are equivalent to
using `assertException' with a `ClassPattern' of `error', `exit', or
`throw', respectively.

Examples:
```?assertError(badarith, X/0)'''
```?assertExit(normal, exit(normal))'''
```?assertException(throw, {not_found,_}, throw({not_found,42}))'''
</dd>
</dl>

=== Macros for running external commands ===

Keep in mind that external commands are highly dependent on the
operating system. You can use the standard library function `os:type()'
in test generator functions, to produce different sets of tests
depending on the current operating system.

Note: these macros introduce a run-time dependency on the EUnit library
code, if compiled with testing enabled.

<dl>
<dt>`assertCmd(CommandString)'</dt>
<dd>Runs `CommandString' as an external command, if testing is enabled.
Unless the returned status value is 0, an informative exception will be
generated. If there is no exception, the result of the macro expression
is the atom `ok'. If testing is disabled, the macro will not generate
any code except the atom `ok', and the command will not be executed.

Typical usage:
```?assertCmd("mkdir foo")'''
</dd>
<dt>`assertCmdStatus(N, CommandString)'</dt>
<dd>Like the `assertCmd(CommandString)' macro, but generates an
exception unless the returned status value is `N'.
</dd>
<dt>`assertCmdOutput(Text, CommandString)'</dt>
<dd>Runs `CommandString' as an external command, if testing is enabled.
Unless the output produced by the command exactly matches the specified
string `Text', an informative exception will be generated. (Note that
the output is normalized to use a single LF character as line break on
all platforms.) If there is no exception, the result of the macro
expression is the atom `ok'. If testing is disabled, the macro will not
generate any code except the atom `ok', and the command will not be
executed.
</dd>
<dt>`cmd(CommandString)'</dt>
<dd>Runs `CommandString' as an external command. Unless the returned
status value is 0 (indicating success), an informative exception will be
generated; otherwise, the result of the macro expression is the output
produced by the command, as a flat string. The output is normalized to
use a single LF character as line break on all platforms.

This macro is useful in the setup and cleanup sections of fixtures,
e.g., for creating and deleting files or perform similar operating
system specific tasks, to make sure that the test system is informed of
any failures.

A Unix-specific example:
```{setup,
    fun () -> ?cmd("mktemp") end,
    fun (FileName) -> ?cmd("rm " ++ FileName) end,
    ...}'''
</dd>
</dl>

=== Debugging macros ===

To help with debugging, EUnit defines several useful macros for printing
messages directly to the console (rather than to the standard output).
Furthermore, these macros all use the same basic format, which includes
the file and line number where they occur, making it possible in some
development environments (e.g., when running Erlang in an Emacs buffer)
to simply click on the message and jump directly to the corresponding
line in the code.

If the macro `NODEBUG' is defined before the EUnit header file is
included, these macros have no effect; see
{@section Compilation control macros} for details.

<dl>
<dt>`debugHere'</dt>
<dd>Just prints a marker showing the current file and line number. Note
that this is an argument-less macro. The result is always `ok'.</dd>
<dt>`debugMsg(Text)'</dt>
<dd>Outputs the message `Text' (which can be a plain string, an IO-list,
or just an atom). The result is always `ok'.</dd>
<dt>`debugFmt(FmtString, Args)'</dt>
<dd>This formats the text like `io:format(FmtString, Args)' and outputs
it like `debugMsg'. The result is always `ok'.</dd>
<dt>`debugVal(Expr)'</dt>
<dd>Prints both the source code for `Expr' and its current value. E.g.,
`?debugVal(f(X))' might be displayed as "`f(X) = 42'". (Large terms are
shown truncated.) The result is always the value of `Expr', so this
macro can be wrapped around any expression to display its value when
the code is compiled with debugging enabled.</dd>
<dt>`debugTime(Text,Expr)'</dt>
<dd>Prints `Text' and the wall clock time for evaluation of `Expr'. The
result is always the value of `Expr', so this macro can be wrapped
around any expression to show its run time when the code is compiled
with debugging enabled. For example, `List1 = ?debugTime("sorting",
lists:sort(List))' might show as "`sorting: 0.015 s'".</dd>

</dl>


== EUnit test representation ==

The way EUnit represents tests and test sets as data is flexible,
powerful, and concise. This section describes the representation in
detail.

<ul>
<li>{@section Simple test objects}</li>
<li>{@section Test sets and deep lists}</li>
<li>{@section Titles}</li>
<li>{@section Primitives}</li>
<li>{@section Control}</li>
<li>{@section Fixtures}</li>
<li>{@section Lazy generators}</li>
</ul>

=== Simple test objects ===

A <em>simple test object</em> is one of the following:
<ul>
  <li>A nullary functional value (i.e., a fun that takes zero
      arguments). Examples:
```fun () -> ... end'''
```fun some_function/0'''
```fun some_module:some_function/0'''
  </li>
  <li>A tuple `{test, ModuleName, FunctionName}', where `ModuleName' and
      `FunctionName' are atoms, referring to the function
      `ModuleName:FunctionName/0'</li>
  <li>(Obsolete) A pair of atoms `{ModuleName, FunctionName}', equivalent to
      `{test, ModuleName, FunctionName}' if nothing else matches first. This
      might be removed in a future version.</li>
  <li>A pair `{LineNumber, SimpleTest}', where `LineNumber' is a
      nonnegative integer and `SimpleTest' is another simple test
      object. `LineNumber' should indicate the source line of the test.
      Pairs like this are usually only created via `?_test(...)' macros;
      see {@section Basic macros}.</li>
</ul>
In brief, a simple test object consists of a single function that takes
no arguments (possibly annotated with some additional metadata, i.e., a
line number). Evaluation of the function either <em>succeeds</em>, by
returning some value (which is ignored), or <em>fails</em>, by throwing
an exception.

=== Test sets and deep lists ===

A test set can be easily created by placing a sequence of test objects
in a list. If `T_1', ..., `T_N' are individual test objects, then `[T_1,
..., T_N]' is a test set consisting of those objects (in that order).

Test sets can be joined in the same way: if `S_1', ..., `S_K' are test
sets, then `[S_1, ..., S_K]' is also a test set, where the tests of
`S_i' are ordered before those of `S_(i+1)', for each subset `S_i'.

Thus, the main representation of test sets is <em>deep lists</em>, and
a simple test object can be viewed as a test set containing only a
single test; there is no difference between `T' and `[T]'.

A module can also be used to represent a test set; see `ModuleName'
under {@section Primitives} below.

=== Titles ===

Any test or test set `T' can be annotated with a title, by wrapping it
in a pair `{Title, T}', where `Title' is a string. For convenience, any
test which is normally represented using a tuple can simply be given a
title string as the first element, i.e., writing `{"The Title", ...}'
instead of adding an extra tuple wrapper as in `{"The Title", {...}}'.


=== Primitives ===

The following are primitives, which do not contain other test sets as
arguments:
<dl>
<dt>`ModuleName::atom()'
</dt>
<dd>A single atom represents a module name, and is equivalent to
`{module, ModuleName}'. This is often used as in the call
`eunit:test(some_module)'.
</dd>
<dt>`{module, ModuleName::atom()}'
</dt>
<dd>This composes a test set from the exported test functions of the
named module, i.e., those functions with arity zero whose names end
with `_test' or `_test_'. Basically, the `..._test()' functions become
simple tests, while the `..._test_()' functions become generators.

In addition, EUnit will also look for another module whose name is
`ModuleName' plus the suffix `_tests', and if it exists, all the tests
from that module will also be added. (If `ModuleName' already contains
the suffix `_tests', this is not done.) E.g., the specification
`{module, mymodule}' will run all tests in the modules `mymodule' and
`mymodule_tests'. Typically, the `_tests' module should only contain
test cases that use the public interface of the main module (and no
other code).
</dd>
<dt>`{application, AppName::atom(), Info::list()}'
</dt>
<dd>This is a normal Erlang/OTP application descriptor, as found in an
 `.app' file. The resulting test set consists of the modules listed in
 the `modules' entry in `Info'.
</dd>
<dt>`{application, AppName::atom()}'
</dt>
<dd>This creates a test set from all the modules belonging to the
specified application, by consulting the application's `.app' file
(see `{file, FileName}'), or if no such file exists, by testing all
object files in the application's <tt>ebin</tt>-directory (see `{dir,
Path}'); if that does not exist, the `code:lib_dir(AppName)' directory
is used.
</dd>
<dt>`Path::string()'
</dt>
<dd>A single string represents the path of a file or directory, and is
equivalent to `{file, Path}', or `{dir, Path}', respectively, depending
on what `Path' refers to in the file system.
</dd>
<dt>`{file, FileName::string()}'
</dt>
<dd>If `FileName' has a suffix that indicates an object file (`.beam'),
EUnit will try to reload the module from the specified file and test it.
Otherwise, the file is assumed to be a text file containing test
specifications, which will be read using the standard library function
`file:path_consult/2'.

Unless the file name is absolute, the file is first searched for
relative to the current directory, and then using the normal search path
(`code:get_path()'). This means that the names of typical "app" files
can be used directly, without a path, e.g., `"mnesia.app"'.
</dd>
<dt>`{dir, Path::string()}'
</dt>
<dd>This tests all object files in the specified directory, as if they
had been individually specified using `{file, FileName}'.
</dd>
<dt>`{generator, GenFun::(() -> Tests)}'
</dt>
<dd>The generator function `GenFun' is called to produce a test
set.
</dd>
<dt>`{generator, ModuleName::atom(), FunctionName::atom()}'
</dt>
<dd>The function `ModuleName:FunctionName()' is called to produce a test
set.
</dd>
<dt>`{with, X::any(), [AbstractTestFun::((any()) -> any())]}'
</dt>
<dd>Distributes the value `X' over the unary functions in the list,
turning them into nullary test functions. An `AbstractTestFun' is like
an ordinary test fun, but takes one argument instead of zero - it's
basically missing some information before it can be a proper test. In
practice, `{with, X, [F_1, ..., F_N]}' is equivalent to `[fun () ->
F_1(X) end, ..., fun () -> F_N(X) end]'. This is particularly useful if
your abstract test functions are already implemented as proper
functions: `{with, FD, [fun filetest_a/1, fun filetest_b/1, fun
filetest_c/1]}' is equivalent to `[fun () -> filetest_a(FD) end, fun ()
-> filetest_b(FD) end, fun () -> filetest_c(FD) end]', but much more
compact. See also {@section Fixtures}, below.
</dd>
</dl>

=== Control ===

The following representations control how and where tests are executed:
<dl>
<dt>`{spawn, Tests}'</dt>
<dd>Runs the specified tests in a separate subprocess, while the current
test process waits for it to finish. This is useful for tests that need
a fresh, isolated process state. (Note that EUnit always starts at least
one such a subprocess automatically; tests are never executed by the
caller's own process.)</dd>
<dt>`{spawn, Node::atom(), Tests}'</dt>
<dd>Like `{spawn, Tests}', but runs the specified tests on the given
Erlang node.</dd>
<dt>`{timeout, Time::number(), Tests}'</dt>
<dd>Runs the specified tests under the given timeout. Time is in
seconds; e.g., 60 means one minute and 0.1 means 1/10th of a second. If
the timeout is exceeded, the unfinished tests will be forced to
terminate. Note that if a timeout is set around a fixture, it includes
the time for setup and cleanup, and if the timeout is triggered, the
entire fixture is abruptly terminated (without running the
cleanup).</dd>
<dt>`{inorder, Tests}'</dt>
<dd>Runs the specified tests in strict order. Also see `{inparallel,
Tests}'. By default, tests are neither marked as `inorder' or
`inparallel', but may be executed as the test framework chooses.</dd>
<dt>`{inparallel, Tests}'</dt>
<dd>Runs the specified tests in parallel (if possible). Also see
`{inorder, Tests}'.</dd>
<dt>`{inparallel, N::integer(), Tests}'</dt>
<dd>Like `{inparallel, Tests}', but running no more than `N' subtests
simultaneously.</dd>
</dl>

=== Fixtures ===

A "fixture" is some state that is necessary for a particular set of
tests to run. EUnit's support for fixtures makes it easy to set up such
state locally for a test set, and automatically tear it down again when
the test set is finished, regardless of the outcome (success, failures,
timeouts, etc.).

To make the descriptions simpler, we first list some definitions:
<center>
<table border="0" cellspacing="4">
<tr>
<td>`Setup'</td><td>`() -> (R::any())'</td>
</tr>
<tr>
<td>`SetupX'</td><td>`(X::any()) -> (R::any())'</td>
</tr>
<tr>
<td>`Cleanup'</td><td>`(R::any()) -> any()'</td>
</tr>
<tr>
<td>`CleanupX'</td><td>`(X::any(), R::any()) -> any()'</td>
</tr>
<tr>
<td>`Instantiator'</td><td>`((R::any()) -> Tests) | {with, [AbstractTestFun::((any()) -> any())]}'</td>
</tr>
<tr>
<td>`Where'</td><td>`local | spawn | {spawn, Node::atom()}'</td>
</tr>
</table>
</center>
(these are explained in more detail further below.)

The following representations specify fixture handling for test sets:
<dl>
<dt>`{setup, Setup, Tests | Instantiator}'</dt>
<dt>`{setup, Setup, Cleanup, Tests | Instantiator}'</dt>
<dt>`{setup, Where, Setup, Tests | Instantiator}'</dt>
<dt>`{setup, Where, Setup, Cleanup, Tests | Instantiator}'</dt>
<dd>`setup' sets up a single fixture for running all of the specified
tests, with optional teardown afterwards. The arguments are described in
detail below.
</dd>
<dt>`{node, Node::atom(), Tests | Instantiator}'</dt>
<dt>`{node, Node::atom(), Args::string(), Tests | Instantiator}'</dt>
<dd>`node' is like `setup', but with a built-in behaviour: it starts a
slave node for the duration of the tests. The atom `Node' should have
the format `nodename@full.machine.name', and `Args' are the optional
arguments to the new node; see `slave:start_link/3' for details.
</dd>
<dt>`{foreach, Where, Setup, Cleanup, [Tests | Instantiator]}'</dt>
<dt>`{foreach, Setup, Cleanup, [Tests | Instantiator]}'</dt>
<dt>`{foreach, Where, Setup, [Tests | Instantiator]}'</dt>
<dt>`{foreach, Setup, [Tests | Instantiator]}'</dt>
<dd>`foreach' is used to set up a fixture and optionally tear it down
afterwards, repeated for each single one of the specified test sets.
</dd>
<dt>`{foreachx, Where, SetupX, CleanupX,
      Pairs::[{X::any(), ((X::any(), R::any()) -> Tests)}]}'</dt>
<dt>`{foreachx, SetupX, CleanupX, Pairs}'</dt>
<dt>`{foreachx, Where, SetupX, Pairs}'</dt>
<dt>`{foreachx, SetupX, Pairs}'</dt>
<dd>`foreachx' is like `foreach', but uses a list of pairs, each
containing an extra argument `X' and an extended instantiator function.
</dd>
</dl>

A `Setup' function is executed just before any of the specified tests
are run, and a `Cleanup' function is executed when no more of the
specified tests will be run, regardless of the reason. A `Setup'
function takes no argument, and returns some value which will be passed
as it is to the `Cleanup' function. A `Cleanup' function should do
whatever necessary and return some arbitrary value, such as the atom
`ok'. (`SetupX' and `CleanupX' functions are similar, but receive one
additional argument: some value `X', which depends on the context.) When
no `Cleanup' function is specified, a dummy function is used which has
no effect.

An `Instantiator' function receives the same value as the `Cleanup'
function, i.e., the value returned by the `Setup' function. It should
then behave much like a generator (see {@section Primitives}), and
return a test set whose tests have been <em>instantiated</em> with the
given value. A special case is the syntax `{with, [AbstractTestFun]}'
which represents an instantiator function that distributes the value
over a list of unary functions; see {@section Primitives}: `{with, X,
[...]}' for more details.

A `Where' term controls how the specified tests are executed. The
default is `spawn', which means that the current process handles the
setup and teardown, while the tests are executed in a subprocess.
`{spawn, Node}' is like `spawn', but runs the subprocess on the
specified node. `local' means that the current process will handle both
setup/teardown and running the tests - the drawback is that if a test
times out so that the process is killed, the <em>cleanup will not be
performed</em>; hence, avoid this for persistent fixtures such as file
operations. In general, 'local' should only be used when:
<ul>
  <li>the setup/teardown needs to be executed by the process that will
  run the tests;</li>
  <li>no further teardown needs to be done if the process is killed
  (i.e., no state outside the process was affected by the setup)</li>
</ul>

=== Lazy generators ===

Sometimes, it can be convenient not to produce the whole set of test
descriptions before the testing begins; for example, if you want to
generate a huge amount of tests that would take up too much space to
keep in memory all at once.

It is fairly easy to write a generator which, each time it is called,
either produces an empty list if it is done, or otherwise produces a
list containing a single test case plus a new generator which will
produce the rest of the tests. This demonstrates the basic pattern:

```lazy_test_() ->
       lazy_gen(10000).

   lazy_gen(N) ->
       {generator,
        fun () ->
            if N > 0 ->
                   [?_test(...)
                    | lazy_gen(N-1)];
               true ->
                   []
            end
        end}.'''

When EUnit traverses the test representation in order to run the tests,
the new generator will not be called to produce the next test until the
previous test has been executed.

Note that it is easiest to write this kind of recursive generator using
a help function, like the `lazy_gen/1' function above. It can also be
written using a recursive fun, if you prefer to not clutter your
function namespace and are comfortable with writing that kind of code.