Test::Inline - Embed your tests in your code, next to what is being
Embedding tests allows tests to be placed near the code being tested.
This is a nice supplement to the traditional .t files.
How does it work?
"Test::Inline" lets you write small fragments of general or
function-specific testing code, and insert it anywhere you want in your
modules, inside a specific tagged POD segment, like the following.
# This code assumes we have a cpuinfo file
ok( -f /proc/cpuinfo, 'Host has a standard /proc/cpuinfo file' );
=begin testing label
# Test generation of the <label> HTML tag
is( My::HTML->label('foo'), '<label>foo</label>', '->label(simple) works' );
is( My::HTML->label('bar', 'foo'), '<label for="bar">foo</label>', '->label(for) works' );
You can add as many, or as few, of these chunks of tests as you wish.
The key condition when writing them is that they should be logically
independant of each other. Each chunk of testing code should not die or
crash if it is run before or after another chunk.
Using inline2test or another test compiler, you can then transform these
chunks in a test script, or an entire tree of modules into a complete
set of standard Test::More-based test scripts.
These test scripts can then be executed as normal.
What is Test::Inline good for?
"Test::Inline" is incredibly useful for doing ad-hoc unit testing.
In any large groups of modules, you can add testing code here, there and
everywhere, anywhere you want. The next time the test compiler is run, a
new test script will just appear.
This also makes it great for testing assumptions you normally wouldn't
bother to write run-time code to test. It ensures that your assumptions
about the way Perl does some operation, or about the state of the host,
are confirmed at install-time.
If your assumption is ever wrong, it gets picked up at install-time and
based on the test failures, you can correct your assumption.
It's also extremely useful for systematically testing self-contained
That is, any code which can be independantly tested without the need for
external systems such as databases, and that has no side-effects on
All of this code, written by multiple people, can then have one single
set of test files generated. You can check all the bits and pieces of a
large API, or anything you like, in fine detail.
Test::Inline also introduces the concept of unit-tested documentation.
Not only can your code be tested, but if you have a FAQ or some other
pure documentation module, you can validate that the documentation is
correct for the version of the module installed.
If the module ever changes to break the documentation, you can catch it
and correct the documentation.
What is Test::Inline bad for?
"Test::Inline" is not a complete testing solution, and there are several
types of testing you probably DON'T want to use it for.
* Static testing across the entire codebase
* Functional testing
* Tests with side-effects such as those that might change a testing
Because Test::Inline creates test scripts with file names that don't
start with a number (for ordering purposes), the first step is to create
your normal test scripts using file names in the CPAN style of
01_compile.t, 02_main.t, 03_foobar.t, and so on.
You can then add your testing fragments wherever you like throughout
your code, and use the inline2test script to generate the test scripts
for the inline tests. By default the test scripts will be named after
the packages/classes that the test fragments are found in.
Tests for Class::Name will end up in the file "class_name.t".
These test files sit quite happily alongside your number test scripts.
When you run the test suite as you normally would, the inline scripts
will be run after the numbered tests.
my $Tests = Test::Inline->new(
verbose => 1,
readonly => 1,
output => 'auto',
manifest => 'auto/manifest',
The "new" constructor creates a new test generation framework. Once the
constructor has been used to create the generator, the "add_class"
method can be used to specify classes, or class heirachies, to generate
verbose - The "verbose" option causes the generator to write state and
debugging information to STDOUT as it runs.
manifest - The "manifest" option, if provided, will cause a manifest
file to be created and written to disk. The manifest file contains a
list of all the test files generated, but listed in the prefered order
they should be processed to best satisfy the class-level dependency of
check_count - The "check_count" value controls how strictly the test
script will watch the number of tests that have been executed.
When set to false, the script does no count checking other than the
standard total count for scripts (where all section counts are known)
When set to 1 (the default), "Test::Inline" does smart count checking,
doing section-by-section checking for known-count sections only when the
total for the entire script is not known.
When set to 2 or higher, "Test::Inline" does full count checking, doing
section-by-section checking for every section with a known number of
file_content - The "file_content" option should be provided as a CODE
reference, which will be passed as arguments the "Test::Inline" object,
and a single Test::Inline::Script object, and should return a string
containing the contents of the resulting test file. This will be written
to the "OutputHandler".
output - The "output" option provides the location of the directory
where the tests will be written to. It should both already exist, and be
writable. If using a custom "OutputHandler", the value of "output"
should refer to the location within the OutputHandler that the files
will be written to.
readonly - The "readonly" option, if provided, indicates that any
generated test files should be created (or set when updated) with
read-only permissions, to prevent accidentally adding to or editing the
test scripts directly (instead of via the classes).
This option is currently disabled by default, by may be enabled by
default in a future release, so if you do NOT want your tests being
created as read-only, you should explicitly set this option to false.
InputHandler - The "InputHandler" option, if provided, supplies an
alternative "FileHandler" from which source modules are retrieved.
OuputHandler - The "OutputHandler" option, if provided, supplies an
alternative "FileHandler" to which the resulting test scripts are
Returns a new "Test::Inline" object on success.
Returns "undef" if there is a problem with one of the options.
The "exception" method returns a flag which indicates whether error will
be returned via exceptions.
The "InputHandler" method returns the file handler object that will be
used to find and load the source code.
The "ExtractHandler" accessor returns the object that will be used to
extract the test sections from the source code.
The "ContentHandler" accessor return the script content generation
The "OutputHandler" accessor returns the file handler object that the
generated test scripts will be written to.
add $file, $directory, \$source, $Handle
The "add" method is a parameter-sensitive method for adding something to
the build schedule.
It takes as argument a file path, a directory path, a reference to a
SCALAR containing perl code, or an IO::Handle (or subclass) object. It
will retrieve code from the parameter as appropriate, parse it, and
create zero or more Test::Inline::Script objects representing the test
scripts that will be generated for that source code.
Returns the number of test scripts added, which could be zero, or
"undef" on error.
$Tests->add_class( 'Foo::Bar' );
$Tests->add_class( 'Foo::Bar', recursive => 1 );
The "add_class" method adds a class to the list of those to have their
tests generated. Optionally, the "recursive" option can be provided to
add not just the class you provide, but all classes below it as well.
Returns the number of classes found with inline tests, and added,
including 0 if no classes with tests are found. Returns "undef" if an
error occurs while adding the class or it's children.
The "add_all" method will search the "InputHandler" for all *.pm files,
and add them to the generation set.
Returns the total number of test scripts added, which may be zero, or
"undef" on error.
The "classes" method returns a list of the names of all the classes that
have been added to the "Test::Inline" object, or the null list "()" if
nothing has been added.
For a given class name, fetches the Test::Inline::Script object for that
class, if it has been added to the "Test::Inline" object. Returns
"undef" if the class has not been added to the "Test::Inline" object.
For all of the classes added, the "filenames" method generates a map of
the filenames that the test files for the various classes should be
Returns a reference to a hash with the classes as keys, and filenames as
Returns 0 if there are no files to write.
Returns "undef" on error.
While the "filenames" method generates a map of the files for the
various classes, the "schedule" returns the list of file names in the
order in which they should actually be executed.
Returns a reference to an array containing the file names as strings.
Returns 0 if there are no files to write.
Returns "undef" on error.
The "manifest" generates the contents of the manifest file, if it is
both wanted and needed.
Returns the contents of the manifest file as a normal string, false if
it is either not wanted or needed, or "undef" on error.
The "save" method generates the test files for all classes, and saves
them to the "output" directory.
Returns the number of test files generated. Returns "undef" on error.
The "Extended =begin" syntax used for non-trivial sections is not
formalised as part of the POD spec yet, although it is on the track to
While simple '=begin testing' sections are fine and will pass POD
testing, extended begin sections may cause POD errors.
- Add support for "example" sections
- Add support for "=for" sections
Bugs should always be submitted via the CPAN bug tracker
Professional support, assistance, or customisations for large scale uses
of "Test::Inline" are available from <http://phase-n.com/>.
For other issues, contact the maintainer.
Adam Kennedy <email@example.com>
Thank you to Phase N (<http://phase-n.com/>) for permitting the open
sourcing and release of this distribution.
Copyright 2004 - 2013 Adam Kennedy.
This program is free software; you can redistribute it and/or modify it
under the same terms as Perl itself.
The full text of the license can be found in the LICENSE file included
with this module.