NAME Test::Inter - framework for more readable interactive test scripts

DESCRIPTION This is another framework for writing test scripts. Much of the syntax is loosely inspired by Test::More, and Test::Inter has most of it’s functionality, but it is not a drop-in replacement.

Test::More (and other existing test frameworks) suffer from two
weaknesses, both of which have prevented me from ever using them:

   None offer the ability to access specific tests in
   a reasonably interactive fashion, primarily for
   debugging purposes

   None offer the ability to write the tests in
   whatever format would make the tests the most
   readable

The way I write and use test scripts, existing Test::* modules are not
nearly as useful as they could be.

Test scripts written using Test::More work fine when running as part of
the test suite, but debugging an individual test requires extra steps,
and the tests themselves are not as readable as they should be.

INTERACTIVE EXECUTION One requirement that I have of a test framework is the ability to interact with it.

I do most of my debugging using test scripts. When I find a bug, I write
a test case for it (typically by adding it to an existing test script)
and then debug it using the test script. Then I leave the test there to
ensure that the bug won't come back (hopefully).

Since I use test scripts in a very interactive way (often in the
debugger), I want to be able to do the following trivially:

Easy access to a specific test or tests
    I'd like to be able to run only a single test, or a subset of tests.

Easily set breakpoints in the debugger
    Setting a breakpoint in the debugger to run up to the start of the
    Nth test is one of the most common tasks I want to do when I'm
    debugging a failed test.

To illustrate the first point, in Test::More, a series of tests might be
specified in a test script as shown in the following example (line
numbers added for convenience):

   ...

   100:  # test 1
   101:  $result = func("apples","bushels");
   102:  is($result, "enough");
   103:
   104:  # test 2
   105:  $result = func("grapefruit","tons");
   106:  is($result, "enough");
   107:
   108:  # test 3
   109:  $result = func("oranges","boatloads");
   110:  is($result, "insufficient");
   111:
   112:  # tests 4-6
   113:  foreach my $arg (qw(pears plums pineapple)) {
   114:    $result = func($arg,"boxes");
   115:    is($result, "enough");
   116:  }

   ...

Say you ran the test suite, and test 3 failed. To debug it you have to
open up the test script, find the 3rd test, and set the appropriate
breakpoint. In this case, you'll want to break at line 109.

None of these steps are impossible of course, but it will take some time
to get it right. It becomes harder when there are lots of tests (imagine
that you want to test the 117th test instead of the 3rd test) or when
tests are wrapped up in loops, embedded in subroutines, or other similar
situations.

As an example, what if it's the 5th test that fails in the example
above. Now the break point will be a conditional one, so you have to
figure out not only the line, but the condition the appropriate state
during that test. In this case, you need to stop at line 114 when $arg
is 'plums'.

Wouldn't it be far better to set a break point in func when the Nth test
is reached? With Test::Inter, you can.

So for the above script, the debugger commands that you would use to
debug the 3rd test are:

   Test::More :   b 109
   Test::Inter:   b func ($::TI_NUM==3)

and the 5th test are:

   Test::More :   b 114  ($arg eq 'plums')
   Test::Inter:   b func ($::TI_NUM==5)

It would also be nice to be able to skip the first two tests... perhaps
they take a long time to run, and I want to get right to work on test 3.
You can do this easily too by setting the $::TI_START variable.

There are some other variables that can be used to specify which test or
tests to run described in the "TEST::INTER VARIABLES" section below.

The other thing I want to do when I run the test scripts interactively
is to see more information which will assist in debugging a failed test.

This can be controlled with variables such as TI_QUIET, TI_MODE, and
TI_WIDTH described below in the "TEST::INTER VARIABLES" section.

READABLE TESTS The other feature that I wanted in a test suite is the ability to define the tests in a format that is natural and readable FOR THE TESTS. In almost every case, it is best to think of a test script as consisting of two separate parts: a script part, and a test part.

The script part of a test script is the least important part! It's
usually fairly trivial, rarely needs to be changed, and is not the focus
of the test script.

The tests part of the script IS the important part, and these should be
expressed in a form that is natural to them, easy to maintain, easy to
read, and easy to modify, and none of these should involve modifying the
script portion of the test script in general. Because the content of the
tests is the important part of the script, the emphasis should be in
making them more readable, even at the expense of the script portion. As
a general rule, if the script portion of the test script obscures the
tests in any way, it's not written correctly!

The solution to this is well understood, and is common to many other
systems where you are mixing two "languages". The task of correctly
specifying both the tests and the test script is virtually identical to
the task of creating a PHP script which consists of a mixture of PHP and
HTML, or the task of creating a template file using some templating
system where the file consists of a mixture of text to be displayed and
templating commands. It is well understood in each of these cases that
the more the two "languages" are interwoven, the less readable both are,
and the harder it is to maintain. The more you are able to separate the
two, the easier both are to read and maintain.

As often as possible, I want the tests to be written in some sort of
text format which can be easily viewed and modified (usually as a simple
table) with no perl commands interspersed. I want to the freedom to
define the tests in one section (a long string, the DATA section, or
even in a separate file) which is easily readable. This may introduce
the necessity of parsing it, but it makes it significantly easier to
maintain the tests.

This flexibility makes it much easier to read the tests (as opposed to
the script) which is the fundamental content of a test script.

Looking again at the example test script, you can see that there is far
too much perl interspersed with the tests.

It's difficult to read the tests individually in this script because
there is too much perl code among them, and virtually impossible to look
at them as a whole.

It is true that looking at this particular example, it is very simple...
but the script ISN'T the content you're interested in (and bear in mind
that many test scripts are nowhere near this simple). The REAL content
of this script are the tests, which consist of the function arguments
and the expected result. Although it's not impossible to see each of
these in the script above, it's not in a format that is conducive to
studying the tests, and especially not for examining the list of tests
as a whole.

Now, look at an alternate way of specifying the tests using this module:

   $tests = "

     apples     bushels   => enough

     grapefruit tons      => enough

     oranges    boatloads => insufficient

     pears      boxes     => enough

     plums      boxes     => enough

     pineapple  boxes     => enough

   ";

   $o->tests(tests => $tests,
             func  => \&func);

Here, it's easy to see the list of tests, and adding additional tests is
a breeze.

CREATING A TEST This module supports a number of methods for defining tests, so you can use whichever one is most convenient (including methods that are identical to Test::More if that really is the best method).

Every test may have several pieces of information:

A name
    Every test is automatically assigned a number, but it may be useful
    to specify a name of a test (which is actually a short description
    of the test). Whenever a test result is reported, the name will be
    given (if one was specified).

    The name may not have a '#' in it.

    The name is completely optional, but can serve to make the results
    more readable.

An expected result
    In order to test something, you need to know what result was
    expected (or in some cases, what result was NOT expected).

A function and arguments OR a result
    You also need to know the results that you're comparing to the
    expected results.

    This can be obtained by simply working with a set of results, or a
    function name and a set of arguments to pass to it.

Conditions
    It is useful to be able to specify state information at the start of
    the test suite (for example, to see if certain features are
    available), and some tests may only run if those conditions are met.

    If no conditions are set for a test, it will always run.

Todo tests
    Some tests may be marked as 'todo' tests. These are test which are
    allowed to fail (meaning that they have been put in place for an
    as-yet unimplemented feature). Since it is expected that the test
    will fail, the test suite will still pass, even if these tests fail.

    The tests will still run and if they pass, a message is issued
    saying that the feature is now implemented, and the tests should be
    graduated to non-todo state.

BASE METHODS new $o=newTest::Inter[o\; =\; new\; Test::Inter\; [$name] [%options];

    This creates a new test framework. There are several options which
    may be used to specify which tests are run, how they are run, and
    what output is given.

    The entire test script can be named by passing in $name.

    All options can be set in four different ways.

    First, you can pass in a hash of OPT = VAL> pairs in the new method.
    So, to set the start option, the %options) hash would contain:

       start => VALUE

    Second, you can set an environment variable. This will override any
    value passed in the first way. The environment variable is named
    TI_XXX where XXX is the fully capitalized option. So:

       $ENV{TI_START} = VALUE

    The third method, which overrides the previous two, is to set a
    global variable. It is also named TI_XXX in the main namespace, so
    to set it this way, set:

       $::TI_START = VALUE

    The final method is to call one of the methods below and these
    override all other methods.

    Each of the allowed options are described below in the following
    base methods:

       start
       end
       testnum
       plan
       abort
       quiet
       mode
       skip_all
       width
       use_lib

version
       $o->version();

    Returns the version of the module.

encoding
       $o->encoding($encoding);

    $encoding is any value that can be passed as an encoding to perl's
    Encode::decode function.

    Use this if your test strings contain characters in other encodings.

start
       $o = new Test::Inter 'start' => $N;
       $o->start($N)

    To define which test you want to start with, set the start option as
    described in the new method above.

    When the start test is defined, most tests numbered less than N are
    completely ignored. If the tests are being run quietly (see the
    quiet method below), nothing is printed out for these tests.
    Otherwise, a skip message is printed out.

    One class of tests IS still executed. Tests run using the require_ok
    or use_ok methods (to test the loading of modules) are still run.

    If no value (or a value of 0) is used, tests run from the first
    test.

end
       $o = new Test::Inter 'end' => $M;
       $o->end($M);

    To define which test you want to end with, set the end option as
    described in the new method above.

    When the end test is defined, all tests numbered more than M are
    completely ignored. If the tests are being run quietly (see the
    quiet method below), nothing is printed out for these tests.
    Otherwise, a skip message is printed out.

    If no value is given, it defaults to 0 (which means that all
    remaining tests are run).

testnum
       $o = new Test::Inter 'testnum' => $N;
       $o->testnum($N);

    To run only a single test, set the testnum option as described in
    the new method above.

    It is equivalent to setting both the start and end tests to $N.

plan
done_testing
       $o = new Test::Inter 'plan' => $N;
       $o->plan($n);

       $o->done_testing();
       $o->done_testing($n);

    The TAP API (the 'language' used to run a sequence of tests and see
    which ones failed and which ones passed) requires a statement of the
    number of tests that are expected to run.

    This statement can appear at the start of the test suite, or at the
    end.

    If you know in advance how many tests should run in the test script,
    you can set the plan option as described in the new method above to
    the number of tests.

    If you know how many tests should run at the end of the test script,
    you can pass in a non-zero integer to the done_testing method.

    Frequently, you don't really care how many tests are in the script
    (especially if new tests are added on a regular basis). In this
    case, you still need to include a statement that says that the
    number of tests expected is however many were run. To do this, call
    the done_testing method with no argument.

    NOTE: if the plan method is used, it MUST be used before any tests
    are run (including those that test the loading of modules). If the
    done_testing method is used, it MUST be called after all tests are
    run. You must specify a plan or use a done_testing statement, but
    you cannot do both.

    It is NOT strictly required to set a plan if the script is only run
    interactively, so if for some reason this module is used for test
    scripts which are not part of a standard perl test suite, the plan
    and done_testing statements are optional. As a matter of fact, the
    script will run just fine without them... but a perl installer will
    report a failure in the test suite.

abort
       $o = new Test::Inter 'abort' => 0/1/2;
       $o->abort(0/1/2);

    To define how you want a failure to be treated, set the abort option
    as described in the new method above. The abort option can take a
    value of 0, 1, or 2.

    If this is set to 1, the test script will run unmodified until a
    test fails. At that point, all remaining tests will be skipped. If
    it is set to 2, the test script will run until a test fails at which
    point it will exit with an error code of 1. With a value of 0,
    failed tests will be reported, but the script will continue.

    In both cases, todo tests will NOT trigger the abort behavior.

quiet
       $o = new Test::Inter 'quiet' => 0/1/2;
       $o->quiet(0/1/2);

    To define how you want failures to be reported, set the quiet option
    as described in the new method above. The quiet option can take a
    value of 0, 1, or 2.

    If this is set to 0 (the default), all information will be printed
    out. If it is set to 1, some optional information will not be
    printed. If it is set to 2, all optional information will not be
    printed.

mode
       $o = new Test::Inter 'mode' => MODE;
       $o->mode(MODE);

    Test::Inter scripts can be run in either an interactive mode, or as
    part of a test suite with different behaviors. To select the mode,
    set the mode option as described in the new method above. The mode
    option can take a value of 'inter' or 'test'.

    When run in test mode, it prints out the results using the TAP
    grammar (i.e. 'ok 1', 'not ok 3', etc.).

    When run in interactive mode, it prints out results in a more human
    readable format.

width
       $o = new Test::Inter 'width' => WIDTH;
       $o->width(WIDTH);

    The width option can be set as described in the new method above.

    WIDTH is the width of the terminal (for printing out failed test
    information). It defaults to 80, but it can be set to any width (and
    lines longer then this are truncated). If WIDTH is set to 0, no
    truncation is done.

use_lib
       $o = new Test::Inter 'use_lib' => VALUE;
       $o->use_lib(VALUE);
       $o->use_lib();

    By default, the library included in the module distribution will be
    added to the search path for modules, so a 'use MODULE' line should
    find the version stored in this module distribution.

    If VALUE is set to 'off', the search path will not be modified
    automatically.

    You may add the library path at a later time by calling:

       $o->use_lib('on');
       $o->use_lib();

    Note: both calls must be used. The first sets the option, the second
    actually modifies the search path.

skip_all
       $o = new Test::Inter 'skip_all' => REASON;
       $o->skip_all(REASON);

    The skip_all option can be set as described in the new method above.

    If this is set, the entire test script will be skipped for the
    reason given. This must be done before any test is run, and before
    any plan number is set.

    The skip_all can also be called at any point during the script (i.e.
    after tests have been run). In this case, all remaining scripts will
    be skipped.

       $o->skip_all(REASON,FEATURE,FEATURE,...);
       $o->skip_all('',FEATURE,FEATURE,...);

    This will skip all tests (or all remaining tests) unless all
    <FEATURE>s are available. REASON can be entered as an empty string
    and the reason the tests are skipped will be a message about the
    missing feature.

feature
       $o->feature($feature,$val);

    This defines a feature. If $val is non-zero, the feature is
    available. Otherwise it is not.

diag
note
       $o->diag($message);
       $o->note($message);

    Both of these print an optional message. Messages printed with the
    "note" method are always optional and will be omitted if the quiet
    option is set to 1 or 2. Messages printed with the "diag" method are
    optional and will not be printed if the quiet option is set to 2,
    but they will be printed if the quiet method is set to 1.

testdir
       $o->testdir();
       $o->testdir('mod');
       $o->testdir('lib');

    Occasionally, it may be necessary to know the directory where
    Test::Inter gets some of it's information. By default, the directory
    containing the tests will be returned, but if the optional argument
    'mod' is included, it will return the path to the module
    distribution (which should include both a lib and t subdirerctory).
    If the argument 'lib' is included, it will return the directory
    where the libraries are stored.

METHODS FOR LOADING MODULES Test scripts can load other modules (using either the perl “use” or “require” commands). There are three different modes for doing this which determine how this is done.

required
    By default, this is used to test for a module that is required for
    all tests in the test script.

    Loading the module is treated as an actual test in the test suite.
    The test is to determine whether the module is available and can be
    loaded. If it can be loaded, it is, and it is reported as a
    successful test. If it cannot be loaded, it is reported as a failed
    test.

    In the result of a failed test, all remaining tests will be skipped
    automatically (except for other tests which load modules).

feature
    In feature mode, loading the module is not treated as a test (i.e.
    it will not print out an 'ok' or 'not ok' line. Instead, it will set
    a feature (named the same as the module) which can be used to
    determine whether other tests should run or not.

forbid
    In a few very rare cases, we may want to test for a module but
    expect that it not be present. This is the exact opposite of the
    required mode.

    Successfully loading the module is treated as a test failure. In the
    event of a failure, all remaining tests will be skipped.

The methods available are:

require_ok
       $o->require_ok($module [,$mode]);

    This is used to load a module using the perl "require" function. If
    $mode is not passed in, the default mode (required) is used to test
    the existence of the module.

    If $mode is passed in, it must be either the string 'forbid' or
    'feature'.

    If $mode is 'feature', a feature named $module is set if the module
    was able to be loaded.

use_ok
       $o->use_ok(@args [,$mode]);

    This is used to load a module with "use", or check a perl version.

       BEGIN { $o->use_ok('5.010'); }
       BEGIN { $o->use_ok('Some::Module'); }
       BEGIN { $o->use_ok('Some::Module',2.05); }
       BEGIN { $o->use_ok('Some::Module','foo','bar'); }
       BEGIN { $o->use_ok('Some::Module',2.05,'foo','bar'); }

    are the same as:

       use 5.010;
       use Some::Module;
       use Some::Module 2.05;
       use Some::Module qw(foo bar);
       use Some::Module 2.05 qw(foo bar);

    Putting the use_ok call in a BEGIN block allows the functions to be
    imported at compile-time and prototypes are properly honored. You'll
    also need to load the Test::Inter module, and create the object in a
    BEGIN block.

    $mode acts the same as in the require_ok method.

METHODS FOR RUNNING TEST There are several methods for running tests. The ok, is, and isnt methods are included for those already comfortable with Test::More and wishing to stick with the same format of test script. The tests method is the suggested method though since it makes use of the full power of this module.

ok
       $o->ok(TESTS);

    A test run with ok looks at a result, and if it evaluates to 0 (or
    false), it fails. If it evaluates to non-zero (or true), it passes.

    These tests do not require you to specify the expected results. If
    expected results are given, they will be compared against the result
    received, and if they differ, a diagnostic message will be printed,
    but the test will still succeed or fail based only on the actual
    result produced.

    These tests require a single result and either zero or one expected
    results.

    To run a single test, use any of the following:

       $o->ok();          # always succeeds

       $o->ok($result);
       $o->ok($result,$name);
       $o->ok($result,$expected,$name);

       $o->ok(\&func);
       $o->ok(\&func,$name);
       $o->ok(\&func,$expected,$name);

       $o->ok(\&func,\@args);
       $o->ok(\&func,\@args,$name);
       $o->ok(\&func,\@args,$expected,$name);

    If $result is a scalar, the test passes if $result is true. If
    $result is a list reference, the test succeeds if the list contains
    any defined values. If $result is a hash reference, the test
    succeeds if the hash contains any key with a value that is not
    "undef".

    If "\&func" and "\@args" are passed in, then $result is generated by
    passing @args to &func and behaves identically to the calls where
    $result is passed in. If "\&func" is passed in but no arguments, the
    function takes no arguments, but still produces a result.

    If an expected value is passed in and the result does not match it,
    a diagnostic warning will be printed, even if the test passes.

is
isnt
       $o->is(TESTS);
       $o->isnt(TESTS);

    A test run with is looks at a result and tests to see if it is
    identical to an expected result. If it is, the test passes.
    Otherwise it fails. In the case of a failure, a diagnostic message
    will show what result was actually obtained and what was expected.

    A test run with isnt looks at a result and tests to see if the
    result obtained is different than an expected result. If it is
    different, the test passes. Otherwise it fails.

    The is method can be called in any of the following ways:

       $o->is($result,$expected);
       $o->is($result,$expected,$name);

       $o->is(\&func,$expected);
       $o->is(\&func,$expected,$name);

       $o->is(\&func,\@args,$expected);
       $o->is(\&func,\@args,$expected,$name);

    The isnt method can be called in exactly the same way.

    As with the ok method, the result can be a scalar, hashref, or
    listref. If it is a hashref or listref, the entire structure must
    match the expected value.

tests
       $o->tests($opt=>$val, $opt=>$val, ...)

    The options available are described in the following section.

file
       $o->file($func,$input,$outputdir,$expected,$name [,@args]);

    Sometimes it may be easiest to store the input, output, and expected
    output from a test in a text file. In this case, each line of output
    will be treated as a single test, so the output and expected output
    must match up exactly.

    $func is a reference to a function which will produce a temporary
    output file.

    If $input is specified, it is the name of the input file. If it is
    empty, no input file will be used. The input file can be fully
    specified, or it can be relative to the test directory.

    If $outputdir is passed in, it is the directory where the output
    file will be written. It can be fully specified, or relative to the
    test directory. If $outputdir is left blank, the temporary file will
    be written to the test directory.

    $expected is the name of a file which contains the expected output.
    It can be fully specified, or it will be checked for in the test
    directory.

    $name is the name of this series of tests.

    @args are extra arguments to pass to the test function.

    The function will be called with the arguments:

       &$func( [$input,] $output,@args);

    $input is only passed in if it was passed in to this method. If no
    input file is specified, nothing will be passed to the function.

    $output is the name of a temporary file where the output will be
    written to.

USING THE TESTS METHOD It is expected that most tests (except for those that load a module) will be run using the tests method called as:

   $o->tests(%options);

The following options are available:

name
       name => NAME

    This sets the name of this set of tests. All tests will be given the
    same name.

tests
func
expected
    In order to specify a series of tests, you have to specify either a
    function and a list of arguments, or a list of results.

    Specifying the function and list of arguments can be done using the
    pair:

       func  => \&FUNCTION
       tests => TESTS

    If the func option is not set, tests contains a list of results.

    A list of expected results may also be given. They can be included
    in the

       tests => TESTS

    option or included separately as:

       expected => RESULTS

    The way to specify these are covered in the next section SPECIFYING
    THE TESTS.

feature
disable
       feature => [FEATURE1, FEATURE2, ...]

       disable => [FEATURE1, FEATURE2, ...]

    The default set of tests to run is determined using the start, end,
    and skip_all methods discussed above. Using those methods, a list of
    tests is obtained, and it is expected that these will run.

    The feature and disable options modify the list.

    If the feature option is included, the tests given in this call will
    only run if ALL of the features listed are available.

    If the disable option is included, the tests will be run unless ANY
    of the features listed are available.

skip
       skip => REASON

    Skip these tests for the reason given.

todo
       todo => 0/1

    Setting this to 1 says that these tests are allowed to fail. They
    represent a feature that is not yet implemented.

    If the tests succeed, a message will be printed notifying the
    developer that the tests are now ready to promote to actual use.

SPECIFYING THE TESTS A series of tests can be specified in two different ways. The tests can be written in a very simple string format, or stored as a list.

Demonstrating how this can be done is best done by example, so let's say
that there is a function (func) which takes two arguments, and returns a
single value. Let's say that the expected output (and the actual output)
from 3 different sets of arguments is:

   Input   Expected Output  Actual Output
   -----   ---------------  -------------
   1,2     a                a
   3,4     b                x
   5,6     c                c

(so in this case, the first and third tests pass, but the 2nd one will
fail).

Specifying these tests as lists could be done as:

   $o->tests(
      func     => &func,
      tests    => [ [1,2], [3,4], [5,6] ],
      expected => [ [a],   [b],   [c] ],
   );

Here, the tests are stored as a list, and each element in the list is a
listref containing the set of arguments.

If the func option is not passed in, the tests option is set to a list
of results to compare with the expected results, so the following is
equivalent to the above:

   $o->tests(
      tests    => [ [a],   [x],   [c] ],
      expected => [ [a],   [b],   [c] ],
   );

If an argument (or actual result) or an expected result is only a single
value, it can be entered as a scalar instead of a list ref, so the
following is also equivalent:

   $o->tests(
      func     => &func,
      tests    => [ [1,2], [3,4], [5,6] ],
      expected => [ a,     b,     [c] ],
   );

The only exception to this is if the single value is itself a list
reference. In this case it MUST be included as a reference. In other
words, if you have a single test, and the expected value for this test
is a list reference, it must be passed in as:

   expected => [ [ \@r ] ]

NOT as:

   expected => [ \@r ]

Passing in a set of expected results is optional. If none are passed in,
the tests are treated as if they had been passed to the ok method (i.e.
if they return something true, they pass, otherwise they fail).

The second way to specify tests is as a string. The string is a
multi-line string with each tests being separate from the next test by a
blank line. Comments (lines which begin with '#') are allowed, and are
ignored. Whitespace at the start and end of the line is ignored.

The string may contain the results directly, or results may be passed in
separately. For example, the following all give the same sets of tests
as the example above:

   $o->tests(
      func     => &func,
      tests    => "
                   # Test 1
                   1 2 => a

                   # Test 2
                   3 4 => b

                   5 6 => c
                  ",
   );

   $o->tests(
      func     => &func,
      tests    => "
                   1 2

                   3 4

                   5 6
                  ",
       expected => [ [a], [b], [c] ]
   );

   $o->tests(
      func     => &func,
      tests    => [ [1,2], [3,4], [5,6] ],
      expected => "
                   a

                   b

                   c
                  ",
   );

   $o->tests(
      func     => &func,
      tests    => "
                   1 2

                   3 4

                   5 6
                  ",
      expected => "
                   a

                   b

                   c
                  ",
   );

The expected results may also consist of only a single set of results
(in this case, it must be passed in as a listref). In this case, all of
the tests are expected to have the same results.

So, the following are equivalent:

   $o->tests(
      func     => &func,
      tests    => "
                   1 2 => a b

                   3 4 => a b

                   5 6 => a b
                  ",
   );

   $o->tests(
      func     => &func,
      tests    => "
                   1 2

                   3 4

                   5 6
                  ",
      expected  => [ [a, b] ],
   );

   $o->tests(
      func     => &func,
      tests    => "
                   1 2

                   3 4

                   5 6
                  ",
      expected  => "a b",
   );

The number of expected values must either be 1 (i.e. all of the tests
are expected to produce the same value) or exactly the same number as
the number of tests.

The parser is actually quite powerful, and can handle multi-line tests,
quoted strings, and nested data structures.

The test may be split across any number of lines, provided there is not
a completely blank line (which signals the end of the test), so the
following are equivalent:

   tests => "a b c",
   tests => "a b
             c",

Arguments (or expected results) may include data structures. For
example, the following are equivalent:

   tests => "[ a b ] { a 1 b 2 }"
   tests => [ [ [a,b], { a=>1, b=>2 } ] ]

Whitespace is mostly optional, but there is one exception. An item must
end with some kind of delimiter, so the following will fail:

   tests => "[a b][c d]"

The first element (the list ref [a b]) must be separated from the second
element by the delimiter (which is whitespace in this case), so it must
be written as:

   tests => "[a b] [c d]"

As already demonstrated, hashrefs and listrefs may be included and
nested. Elements may also be included inside parens, but this is
optional since all arguments and expected results are already treated as
lists, so the following are equivalent:

   tests => "a b c"
   tests => "(a b) c"

Although parens are optional, they may make things more readable, and
allow you to use something other than whitespace as the delimiter. Since
parens are actually ignored, a string '()' is also ignored, so do not
use empty parentheses.

If the character immediately following the opening paren, brace, or
bracket is a punctuation mark, then it is used as the delimiter instead
of whitespace. For example, the following are all equivalent:

   [ a b c ]
   [a b c]
   [, a,b,c ]
   [, a, b, c ]

A delimiter is a single character, and the following may not be used as
a delimiter:

   any opening/closing characters () [] {}
   single or double quotes
   alphanumeric characters
   underscore

Whitespace (including newlines) around the delimiter is ignored, so the
following is valid:

   [, a,
      b,
      c ]

Two delimiters next to each other or a trailing delimiter produce an
empty string.

   "(,a,b,)" => (a, b, '')
   "(,a,,b)" => (a, '', b)

Hashrefs may be specified by braces and the following are equivalent:

   { a 1 b 2 }
   {, a,1,b,2 }
   {, a,1,b,2, }

Note that a trailing delimiter is ignored if there are already an even
number of elements, or an empty string otherwise.

Nested structures are allowed:

   "[ [1 2] [3 4] ]"

For example,

   $o->tests(
      func     => &func,
      tests    => "a [ b c ] { d 1 e 2 } => x y"
   );

is equivalent to:

   $o->tests(
      func     => &func,
      tests    => [ [a, [b,c], {d=>1,e=>2}] ],
      results  => [ [x,y] ],
   );

Any single value can be surrounded by single or double quotes in order
to include the delimiter. So:

   "(, a,'b,c',e )"

is equivalent to:

   "( a b,c e )"

Any single value can be the string '__undef__' which will be turned into
an actual undef. If the value is '__blank__' it is turned into an empty
string (''), though it can also be specified as '' directly. Any value
can have an embedded newline by including a __nl__ in the value, but the
value must be written on a single line.

Expected results are separated from arguments by ' => '.

TEST::INTER VARIABLES To summarize the information above, the following variables are used by Test::Inter. Each variable can be set in two different ways: as an environment variable and as a perl variable in the main namespace.

For example, the TI_END variable can be set as:

   $::TI_END
   $ENV{TI_END}

The following variables can be used to define which tests are run:

TI_START
    Set this to define the test you want to start with.

    Example: If you have a perl test script and you want to start
    running it at test 12, run the following shell commands:

       TI_START=12
       ./my_test_script.t

TI_END
    Set this to define the test you want to end with.

TI_TESTNUM
    Set this to run only a single test

There is also a variable TI_NUM (available only as $::TI_NUM) which is
set automatically by Test::Inter to be the test currently being run.

The following variables control what is output from the tests, and how
it is formatted:

TI_QUIET
    How verbose the test script is. Values are 0 (most verbose) to 2
    (least verbose).

TI_MODE
    How the output is formatted. Values are 'inter' (interactive mode)
    or 'test' (test suite mode). Interactive mode is easier to read.
    Test mode is for running as part of a test suite.

TI_WIDTH
    The width of the terminal.

The following variables control how some tests are run:

TI_NOCLEAN
    When running a file test, the temporary output file will not be
    removed if this is set.

HISTORY The history of this module dates back to 1996 when I needed to write a test suite for my Date::Manip module. At that time, none of the Test::* modules currently available in CPAN existed (the earliest ones didn’t come along until 1998), so I was left completely on my own writing my test scripts.

I wrote a very basic version of my test framework which allowed me to
write all of the tests as a string, it would parse the string, count the
tests, and then run them.

Over the years, the functionality I wanted grew, and periodically, I'd
go back and reexamine other Test frameworks (primarily Test::More) to
see if I could replace my framework with an existing module... and I've
always found them wanting, and chosen to extend my existing framework
instead.

As I've written other modules, I've wanted to use the framework in them
too, so I've always just copied it in, but this is obviously tedious and
error prone. I'm not sure why it took me so long... but in 2010, I
finally decided it was time to rework the framework in a module form.

I loosely based my module on Test::More. I like the functionality of
that module, and wanted most of it (and I plan on adding more in future
versions). So this module uses some similar syntax to Test::More (though
it allows a great deal more flexibility in how the tests are specified).

One thing to note is that I may have been able to write this module as
an extension to Test::More, but after looking into that possibility, I
decided that it would be faster to not do that. I did "borrow" a couple
of routines from it (though they've been modified quite heavily) as a
starting point for a few of the functions in this module, and I thank
the authors of Test::More for their work.

KNOWN BUGS AND LIMITATIONS None known.

SEE ALSO Test::More - the ‘industry standard’ of perl test frameworks

BUGS AND QUESTIONS If you find a bug in Test::Inter, there are three ways to send it to me. Any of them are fine, so use the method that is easiest for you.

Direct email
    You are welcome to send it directly to me by email. The email
    address to use is: sbeck@cpan.org.

CPAN Bug Tracking
    You can submit it using the CPAN tracking too. This can be done at
    the following URL:

    <http://rt.cpan.org/Public/Dist/Display.html?Name=Test-Inter>

GitHub
    You can submit it as an issue on GitHub. This can be done at the
    following URL:

    <https://github.com/SBECK-github/Test-Inter>

Please do not use other means to report bugs (such as forums for a
specific OS or Linux distribution) as it is impossible for me to keep up
with all of them.

When filing a bug report, please include the following information:

Test::Inter version
    Please include the version of Test::Inter you are using. You can get
    this by using the script:

       use Test::Inter;
       print $Test::Inter::VERSION,"\n";

If you want to report missing or incorrect codes, you must be running
the most recent version of Test::Inter.

If you find any problems with the documentation (errors, typos, or items
that are not clear), please send them to me. I welcome any suggestions
that will allow me to improve the documentation.

LICENSE This script is free software; you can redistribute it and/or modify it under the same terms as Perl itself.

AUTHOR Sullivan Beck (sbeck@cpan.org)