Test::Builder::Tester - test testsuites that have been built with Test::Builder

    use Test::Builder::Tester tests => 1;
    use Test::More;

    test_out("not ok 1 - foo");
    test_err("# Failed test ($0 at line ".line_num(+1).")");
    fail("foo");
    test_test("fail works");

A module that helps you test testing modules that are built with Test::Builder.

The testing system is designed to be used by performing a three step process for each test you wish to test. This process starts with using CWtest_out and CWtest_err in advance to declare what the testsuite you are testing will output with Test::Builder to stdout and stderr.

You then can run the test(s) from your test suite that call Test::Builder. At this point the output of Test::Builder is safely captured by Test::Builder::Tester rather than being interpreted as real test output.

The final stage is to call CWtest_test that will simply compare what you predeclared to what Test::Builder actually outputted, and report the results back with a ok or not ok (with debugging) to the normal output.

These are the six methods that are exported as default.

test_out

test_err

Procedures for predeclaring the output that your test suite is expected to produce until CWtest_test is called. These procedures automatically assume that each line terminates with \n. So

   test_out("ok 1","ok 2");

is the same as

   test_out("ok 1\nok 2");

which is even the same as

   test_out("ok 1");
   test_out("ok 2");

Once CWtest_out or CWtest_err (or CWtest_fail or CWtest_diag) have been called once all further output from Test::Builder will be captured by Test::Builder::Tester. This means that your will not be able perform further tests to the normal output in the normal way until you call CWtest_test (well, unless you manually meddle with the output filehandles)

test_fail

Because the standard failure message that Test::Builder produces whenever a test fails will be a common occurrence in your test error output, rather than forcing you to call CWtest_err with the string all the time like so

    test_err("# Failed test ($0 at line ".line_num(+1).")");

CWtest_fail exists as a convenience method that can be called instead. It takes one argument, the offset from the current line that the line that causes the fail is on.

    test_fail(+1);

This means that the example in the synopsis could be rewritten more simply as:

   test_out("not ok 1 - foo");
   test_fail(+1);
   fail("foo");
   test_test("fail works");

test_diag

As most of the remaining expected output to the error stream will be created by Test::Builder's CWdiag function, Test::Builder::Tester provides a convience function CWtest_diag that you can use instead of CWtest_err. The CWtest_diag function prepends comment hashes and spacing to the start and newlines to the end of the expected output passed to it and adds it to the list of expected error output. So, instead of writing

   test_err("# Couldn't open file");

you can write

   test_diag("Couldn't open file");

Remember that Test::Builder's diag function will not add newlines to the end of output and test_diag will. So to check

   Test::Builder->new->diag("foo\n","bar\n");

You would do

  test_diag("foo","bar")

without the newlines.

test_test

Actually performs the output check testing the tests, comparing the data (with CWeq) that we have captured from Test::Builder against that that was declared with CWtest_out and CWtest_err. This takes name/value pairs that effect how the test is run.

title (synonym 'name', 'label')

The name of the test that will be displayed after the CWok or CWnot ok.

skip_out

Setting this to a true value will cause the test to ignore if the output sent by the test to the output stream does not match that declared with CWtest_out.

skip_err

Setting this to a true value will cause the test to ignore if the output sent by the test to the error stream does not match that declared with CWtest_err. As a convience, if only one argument is passed then this argument is assumed to be the name of the test (as in the above examples.) Once CWtest_test has been run test output will be redirected back to the original filehandles that Test::Builder was connected to (probably STDOUT and STDERR,) meaning any further tests you run will function normally and cause success/errors for Test::Harness.

line_num

A utility function that returns the line number that the function was called on. You can pass it an offset which will be added to the result. This is very useful for working out the correct text of diagnostic methods that contain line numbers. Essentially this is the same as the CW__LINE__ macro, but the CWline_num(+3) idiom is arguably nicer.

In addition to the six exported functions there there exists one function that can only be accessed with a fully qualified function call.

color

When CWtest_test is called and the output that your tests generate does not match that which you declared, CWtest_test will print out debug information showing the two conflicting versions. As this output itself is debug information it can be confusing which part of the output is from CWtest_test and which was the original output from your original tests. Also, it may be hard to spot things like extraneous whitespace at the end of lines that may cause your test to fail even though the output looks similar. To assist you, if you have the Term::ANSIColor module installed (which you should do by default from perl 5.005 onwards), CWtest_test can colour the background of the debug information to disambiguate the different types of output. The debug output will have it's background coloured green and red. The green part represents the text which is the same between the executed and actual output, the red shows which part differs. The CWcolor function determines if colouring should occur or not. Passing it a true or false value will enable or disable colouring respectively, and the function called with no argument will return the current setting. To enable colouring from the command line, you can use the Text::Builder::Tester::Color module like so:

   perl -Mlib=Text::Builder::Tester::Color test.t

Or by including the Test::Builder::Tester::Color module directly in the PERL5LIB.

Calls Test::Builder's CWno_ending method turning off the ending tests. This is needed as otherwise it will trip out because we've run more tests than we strictly should have and it'll register any failures we had that we were testing for as real failures.

The color function doesn't work unless Term::ANSIColor is installed and is compatible with your terminal.

Bugs (and requests for new features) can be reported to the author though the CPAN RT system: <http://rt.cpan.org/NoAuth/ReportBug.html?Queue=Test-Builder-Tester>

Copyright Mark Fowler <mark@twoshortplanks.com> 2002, 2004.

Some code taken from Test::More and Test::Catch, written by by Michael G Schwern <schwern@pobox.com>. Hence, those parts Copyright Micheal G Schwern 2001. Used and distributed with permission.

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

This code has been tested explicitly on the following versions of perl: 5.7.3, 5.6.1, 5.6.0, 5.005_03, 5.004_05 and 5.004.

Thanks to Richard Clamp <richardc@unixbeard.net> for letting me use his testing system to try this module out on.

Test::Builder, Test::Builder::Tester::Color, Test::More.