Test::Base - A Data Driven Testing Framework

A new test module:

# lib/MyProject/Test.pm
package MyProject::Test;
use Test::Base -Base;

use MyProject;

package MyProject::Test::Filter;
use Test::Base::Filter -base;

sub my_filter {
    return MyProject->do_something(shift);
}

A sample test:

# t/sample.t
use MyProject::Test;

plan tests => 1 * blocks;

run_is input => 'expected';

sub local_filter {
    s/my/your/;
}

__END__

=== Test one (the name of the test)
--- input my_filter local_filter
my
input
lines
--- expected
expected
output

=== Test two
This is an optional description
of this particular test.
--- input my_filter
other
input
lines
--- expected
other expected
output

Testing is usually the ugly part of Perl module authoring. Perl gives you a standard way to run tests with Test::Harness, and basic testing primitives with Test::More. After that you are pretty much on your own to develop a testing framework and philosophy. Test::More encourages you to make your own framework by subclassing Test::Builder, but that is not trivial.

Test::Base gives you a way to write your own test framework base class that is trivial. In fact it is as simple as two lines:

package MyTestFramework;
use Test::Base -Base;

A module called MyTestFramework.pm containing those two lines, will give all the power of Test::More and all the power of Test::Base to every test file that uses it. As you build up the capabilities of MyTestFramework, your tests will have all of that power as well.

MyTestFramework becomes a place for you to put all of your reusable testing bits. As you write tests, you will see patterns and duplication, and you can "upstream" them into MyTestFramework. Of course, you don't have to subclass Test::Base at all. You can use it directly in many applications, including everywhere you would use Test::More.

Test::Base concentrates on offering reusable data driven patterns, so that you can write tests with a minimum of code. At the heart of all testing you have inputs, processes and expected outputs. Test::Base provides some clean ways for you to express your input and expected output data, so you can spend your

time focusing on that rather than your code scaffolding.

Test::Base extends Test::More and exports all of its functions. So you can basically write your tests the same as Test::More. Test::Base also exports many functions of its own:

is(actual, expected, [test-name])

This is the equivalent of Test::More's is function with one interesting twist. If your actual and expected results differ and the output is multi- line, this function will show you a unified diff format of output. Consider the benefit when looking for the one character that is different in hundreds of lines of output!

Diff output requires the optional Text::Diff CPAN module. If you don't have this module, the is() function will simply give you normal Test::More output. To disable diffing altogether, set the TEST_SHOW_NO_DIFFS environment variable (or $ENV{TEST_SHOW_NO_DIFFS}) to a true value. You can also call the no_diff function as a shortcut.

blocks( [data-section-name] )

The most important function is blocks. In list context it returns a list of Test::Base::Block objects that are generated from the test specification in the DATA section of your test file. In scalar context it returns the number of objects. This is useful to calculate your Test::More plan.

Each Test::Base::Block object has methods that correspond to the names of that object's data sections. There is also a name and a description method for accessing those parts of the block if they were specified.

The blocks function can take an optional single argument, that indicates to only return the blocks that contain a particular named data section. Otherwise blocks returns all blocks.

my @all_of_my_blocks = blocks;

my @just_the_foo_blocks = blocks('foo');

next_block()

You can use the next_block function to iterate over all the blocks.

while (my $block = next_block) {
    ...
}

It returns undef after all blocks have been iterated over. It can then be called again to reiterate.

first_block()

Returns the first block or undef if there are none. It resets the iterator to the next_block function.

run(&subroutine)

There are many ways to write your tests. You can reference each block individually or you can loop over all the blocks and perform a common operation. The run function does the looping for you, so all you need to do is pass it a code block to execute for each block.

The run function takes a subroutine as an argument, and calls the sub one time for each block in the specification. It passes the current block object to the subroutine.

run {
    my $block = shift;
    is(process($block->foo), $block->bar, $block->name);
};

run_is([data_name1, data_name2])

Many times you simply want to see if two data sections are equivalent in every block, probably after having been run through one or more filters. With the run_is function, you can just pass the names of any two data sections that exist in every block, and it will loop over every block comparing the two sections.

run_is 'foo', 'bar';

If no data sections are given run_is will try to detect them automatically.

NOTE: Test::Base will silently ignore any blocks that don't contain both sections.

is_deep($data1, $data2, $test_name)

Like Test::More's is_deeply but uses the more correct Test::Deep module.

run_is_deeply([data_name1, data_name2])

Like run_is_deeply but uses is_deep which uses the more correct Test::Deep.

run_is_deeply([data_name1, data_name2])

Like run_is but uses is_deeply for complex data structure comparison.

run_is_deeply([data_name1, data_name2])

Like run_is_deeply but uses is_deep which uses the more correct Test::Deep.

run_like([data_name, regexp | data_name]);

The run_like function is similar to run_is except the second argument is a regular expression. The regexp can either be a qr{} object or a data section that has been filtered into a regular expression.

run_like 'foo', qr{<html.*};
run_like 'foo', 'match';

run_unlike([data_name, regexp | data_name]);

The run_unlike function is similar to run_like, except the opposite.

run_unlike 'foo', qr{<html.*};
run_unlike 'foo', 'no_match';

run_compare(data_name1, data_name2)

The run_compare function is like the run_is, run_is_deeply and the run_like functions all rolled into one. It loops over each relevant block and determines what type of comparison to do.

NOTE: If you do not specify either a plan, or run any tests, the run_compare function will automatically be run.

delimiters($block_delimiter, $data_delimiter)

Override the default delimiters of === and ---.

spec_file($file_name)

By default, Test::Base reads its input from the DATA section. This function tells it to get the spec from a file instead.

spec_string($test_data)

By default, Test::Base reads its input from the DATA section. This function tells it to get the spec from a string that has been prepared somehow.

filters( @filters_list or $filters_hashref )

Specify a list of additional filters to be applied to all blocks. See FILTERS below.

You can also specify a hash ref that maps data section names to an array ref of filters for that data type.

filters {
    xxx => [qw(chomp lines)],
    yyy => ['yaml'],
    zzz => 'eval',
};

If a filters list has only one element, the array ref is optional.

filters_delay( [1 | 0] );

By default Test::Base::Block objects are have all their filters run ahead of time. There are testing situations in which it is advantageous to delay the filtering. Calling this function with no arguments or a true value, causes the filtering to be delayed.

use Test::Base;
filters_delay;
plan tests => 1 * blocks;
for my $block (blocks) {
    ...
    $block->run_filters;
    ok($block->is_filtered);
    ...
}

In the code above, the filters are called manually, using the run_filters method of Test::Base::Block. In functions like run_is, where the tests are run automatically, filtering is delayed until right before the test.

filter_arguments()

Return the arguments after the equals sign on a filter.

sub my_filter {
    my $args = filter_arguments;
    # is($args, 'whazzup');
    ...
}

__DATA__
=== A test
--- data my_filter=whazzup

tie_output()

You can capture STDOUT and STDERR for operations with this function:

my $out = '';
tie_output(*STDOUT, $out);
print "Hey!\n";
print "Che!\n";
untie *STDOUT;
is($out, "Hey!\nChe!\n");

no_diff()

Turn off diff support for is() in a test file.

default_object()

Returns the default Test::Base object. This is useful if you feel the need to do an OO operation in otherwise functional test code. See OO below.

WWW() XXX() YYY() ZZZ()

These debugging functions are exported from the Spiffy.pm module. See Spiffy for more info.

croak() carp() cluck() confess()

You can use the functions from the Carp module without needing to import them. Test::Base does it for you by default.

Test::Base allows you to specify your test data in an external file, the DATA section of your program or from a scalar variable containing all the text input.

A test specification is a series of text lines. Each test (or block) is separated by a line containing the block delimiter and an optional test name. Each block is further subdivided into named sections with a line containing the data delimiter and the data section name. A description of the test can go on lines after the block delimiter but before the first data section.

Here is the basic layout of a specification:

=== <block name 1>
<optional block description lines>
--- <data section name 1> <filter-1> <filter-2> <filter-n>
<test data lines>
--- <data section name 2> <filter-1> <filter-2> <filter-n>
<test data lines>
--- <data section name n> <filter-1> <filter-2> <filter-n>
<test data lines>

=== <block name 2>
<optional block description lines>
--- <data section name 1> <filter-1> <filter-2> <filter-n>
<test data lines>
--- <data section name 2> <filter-1> <filter-2> <filter-n>
<test data lines>
--- <data section name n> <filter-1> <filter-2> <filter-n>
<test data lines>

Here is a code example:

use Test::Base;

delimiters qw(### :::);

# test code here

__END__

### Test One
We want to see if foo and bar
are really the same...
::: foo
a foo line
another foo line

::: bar
a bar line
another bar line

### Test Two

::: foo
some foo line
some other foo line

::: bar
some bar line
some other bar line

::: baz
some baz line
some other baz line

This example specifies two blocks. They both have foo and bar data sections. The second block has a baz component. The block delimiter is ### and the data delimiter is :::.

The default block delimiter is === and the default data delimiter is --- .

There are some special data section names used for control purposes:

--- SKIP
--- ONLY
--- LAST

A block with a SKIP section causes that test to be ignored. This is useful to disable a test temporarily.

A block with an ONLY section causes only that block to be used. This is useful when you are concentrating on getting a single test to pass. If there is more than one block with ONLY, the first one will be chosen.

Because ONLY is very useful for debugging and sometimes you forgot to remove the ONLY flag before committing to the VCS or uploading to CPAN, Test::Base by default gives you a diag message saying I found ONLY ... maybe you're debugging?. If you don't like it, use no_diag_on_only.

A block with a LAST section makes that block the last one in the specification. All following blocks will be ignored.

The real power in writing tests with Test::Base comes from its filtering capabilities. Test::Base comes with an ever growing set of useful generic filters than you can sequence and apply to various test blocks. That means you can specify the block serialization in the most readable format you can find, and let the filters translate it into what you really need for a test. It is easy to write your own filters as well.

Test::Base allows you to specify a list of filters to each data section of each block. The default filters are norm and trim. These filters will be applied (in order) to the data after it has been parsed from the specification and before it is set into its Test::Base::Block object.

You can add to the default filter list with the filters function. You can specify additional filters to a specific block by listing them after the section name on a data section delimiter line.

Example:

use Test::Base;

filters qw(foo bar);
filters { perl => 'strict' };

sub upper { uc(shift) }

__END__

=== Test one
--- foo trim chomp upper
...

--- bar -norm
...

--- perl eval dumper
my @foo = map {
    - $_;
} 1..10;
\ @foo;

Putting a - before a filter on a delimiter line, disables that filter.

Each filter can take either a scalar or a list as input, and will return either a scalar or a list. Since filters are chained together, it is important to learn which filters expect which kind of input and return which kind of output.

For example, consider the following filter list:

norm trim lines chomp array dumper eval

The data always starts out as a single scalar string. norm takes a scalar and returns a scalar. trim takes a list and returns a list, but a scalar is a valid list. lines takes a scalar and returns a list. chomp takes a list and returns a list. array takes a list and returns a scalar (an anonymous array reference containing the list elements). dumper takes a list and returns a scalar. eval takes a scalar and creates a list.

A list of exactly one element works fine as input to a filter requiring a scalar, but any other list will cause an exception. A scalar in list context is considered a list of one element.

Data accessor methods for blocks will return a list of values when used in list context, and the first element of the list in scalar context. This is usually "the right thing", but be aware.

Test::Base comes with large set of stock filters. They are in the Test::Base::Filter module. See Test::Base::Filter for a listing and description of these filters.

Creating filter extensions is very simple. You can either write a function in the main namespace, or a method in the Test::Base::Filter namespace or a subclass of it. In either case the text and any extra arguments are passed in and you return whatever you want the new value to be.

Here is a self explanatory example:

use Test::Base;

filters 'foo', 'bar=xyz';

sub foo {
    transform(shift);
}

sub Test::Base::Filter::bar {
    my $self = shift;       # The Test::Base::Filter object
    my $data = shift;
    my $args = $self->current_arguments;
    my $current_block_object = $self->block;
    # transform $data in a barish manner
    return $data;
}

If you use the method interface for a filter, you can access the block internals by calling the block method on the filter object.

Normally you'll probably just use the functional interface, although all the builtin filters are methods.

Note that filters defined in the main namespace can look like:

sub filter9 {
    s/foo/bar/;
}

since Test::Base automatically munges the input string into $_ variable and checks the return value of the function to see if it looks like a number. If you must define a filter that returns just a single number, do it in a different namespace as a method. These filters don't allow the simplistic $_ munging.

Test::Base has a nice functional interface for simple usage. Under the hood everything is object oriented. A default Test::Base object is created and all the functions are really just method calls on it.

This means if you need to get fancy, you can use all the object oriented stuff too. Just create new Test::Base objects and use the functions as methods.

use Test::Base;
my $blocks1 = Test::Base->new;
my $blocks2 = Test::Base->new;

$blocks1->delimiters(qw(!!! @@@))->spec_file('test1.txt');
$blocks2->delimiters(qw(### $$$))->spec_string($test_data);

plan tests => $blocks1->blocks + $blocks2->blocks;

# ... etc

In Test::Base, blocks are exposed as Test::Base::Block objects. This section lists the methods that can be called on a Test::Base::Block object. Of course, each data section name is also available as a method.

name()

This is the optional short description of a block, that is specified on the block separator line.

description()

This is an optional long description of the block. It is the text taken from between the block separator and the first data section.

seq_num()

Returns a sequence number for this block. Sequence numbers begin with 1.

blocks_object()

Returns the Test::Base object that owns this block.

run_filters()

Run the filters on the data sections of the blocks. You don't need to use this method unless you also used the filters_delay function.

is_filtered()

Returns true if filters have already been run for this block.

original_values()

Returns a hash of the original, unfiltered values of each data section.

One of the nicest things about Test::Base is that it is easy to subclass. This is very important, because in your personal project, you will likely want to extend Test::Base with your own filters and other reusable pieces of your test framework.

Here is an example of a subclass:

package MyTestStuff;
use Test::Base -Base;

our @EXPORT = qw(some_func);

sub some_func {
    (my ($self), @_) = find_my_self(@_);
    ...
}

package MyTestStuff::Block;
use base 'Test::Base::Block';

sub desc {
    $self->description(@_);
}

package MyTestStuff::Filter;
use base 'Test::Base::Filter';

sub upper {
    $self->assert_scalar(@_);
    uc(shift);
}

Note that you don't have to re-Export all the functions from Test::Base. That happens automatically, due to the powers of Spiffy.

The first line in some_func allows it to be called as either a function or a method in the test code.

You might be thinking that you do not want to use Test::Base in you modules, because it adds an installation dependency. Fear not. Module::Install::TestBase takes care of that.

Just write a Makefile.PL that looks something like this:

use inc::Module::Install;

name            'Foo';
all_from        'lib/Foo.pm';

use_test_base;

WriteAll;

The line with use_test_base will automatically bundle all the code the user needs to run Test::Base based tests.

Test::Base automatically adds:

use strict;
use warnings;

to all of your test scripts and Test::Base subclasses. A Spiffy feature indeed.

This module started its life with the horrible and ridicule inducing name Test::Chunks. It was renamed to Test::Base with the hope that it would be seen for the very useful module that it has become. If you are switching from Test::Chunks to Test::Base, simply substitute the concept and usage of chunks to blocks.

Ingy döt Net <[email protected]>

Copyright 2005-2018. Ingy döt Net.

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

See http://www.perl.com/perl/misc/Artistic.html