Test::Subunits - Extract subunit tests from within complex source code
This document describes Test::Subunits version 0.000003
# Extract and compile testable "subunits"... use Test::Subunits 'Your::Module::Here'; # Then test them in your preferred manner, for example... use Test::More; my @normalized_data = normalize_subunit(@data); ok !grep({!defined}, @normalized_data) => 'All data normalized'; my @sorted = sort_subunit(@unsorted); my $prev = shift @sorted; while (my $next = shift @sorted) { cmp_ok $prev, '>=', $next => "$prev >= $next"; $next = $prev; } done_testing();
"Unit testing is right. Unit testing works. Unit testing clarifies. Unit testing, for lack of a better word, is good."
...but unit testing can also be expensive. In particular, unit testing requires that your code be composed from a large number of small independently testable units (i.e. short subroutines). And, while that's good software engineering, it can sometimes be bad performance engineering.
Because that kind of highly decomposed then highly recomposed code also requires a large number of internal subroutine calls, and a lot of argument passing, either of which can reduce the performance of your code to unacceptable levels.
When that happens, the usual solution is to "inline" the original units of code: to create a single, larger, more complex subroutine that does everything in one place and as fast as possible.
But that kind of subroutine is also much more difficult to test (if it can be tested at all).
This module allows you to write fast-but-monolithic subroutines when you need to, but still be able to test individual sections of that code (known as "subunits") as if they were separate small-but-composable subroutines.
To do this, you annotate parts of your monolithic code with special comments. This module then uses those annotations to extract individual chunks of your code which it compiles into separate subroutines, which your test suite can then test independently.
In a sense, it's the exact opposite of inlining small subroutines into your code. The module effectively "out-lines" predetermined fragments of your code to create small subroutines that you can then test.
To extract subunits from a source file, you load the Test::Subunits module, passing it a single argument that specifies that source file. This argument can be either the full name of the module:
use Test::Subunits 'Your::Module::Name';
or a filepath to the module's source file:
use Test::Subunits 'Your/Module/Name.pm';
The module then searches the usual @INC path, looking for a matching source file. When it finds the file, it reads the source and extracts any subunits that you have specified using the "Subunit mark-up notation" which is described in the next section.
@INC
Normally, you specify these subunits as small subroutines, which are then compiled into your namespace, whereupon you can call them within your preferred testing framework (i.e. Test::More, Test::Class, Test::Effects, etc.)
Within your source file, you indicate subunits to be extracted using special comments that start with two # characters (i.e. using ordinary Perl comments where the very first character of the comment is another #)
#
There are three forms of these special comments: block inclusions, paragraph inclusions, and single-line inclusions...
Block inclusions include all the code and any extra mark-up found between two balanced delimiters: ##{ and ##}. For example, given the following source code:
##{
##}
package My::Module; sub frobnicate_widgets { my ($threshold, $widgets_ref) = @_; # Normalize widget list... $widgets_ref = [grep {defined} @{$widgets_ref}]; # Select widgets to be processed... my @selections; for my $list_elem (@{$widgets_ref}) { if ($list_elem > $threshold) { push @selections, $list_elem; } else { warn "Rejecting: $list_elem"; } } # Process selected widgets... return map { frobnicate($_) } @selections; }
you could mark up two subunits to be extracted for testing like so:
package My::Module; sub frobnicate_widgets { my ($threshold, $widgets_ref) = @_; # Normalize widget list... ##{ ## sub normalize_list { ## my $widgets_ref = shift; $widgets_ref = [grep {defined} @{$widgets_ref}]; ## return $widgets_ref ## } ##} # Select widgets to be processed... ##{ ## sub divide_list { ## my ($widgets_ref, $threshold) = @_; my @selections; for my $list_elem (@{$widgets_ref}) { if ($list_elem > $threshold) { push @selections, $list_elem; } else { warn "Rejecting: $list_elem"; } } ## return @selections; ## } ##} # Process selected widgets... return map { frobnicate($_) } @selections; }
This defines two spearate subunits, which would be extracted as:
sub normalize_list { my $widgets_ref = shift; $widgets_ref = [grep {defined} @{$widgets_ref}]; return $widgets_ref } sub divide_list { my ($widgets_ref, $threshold) = @_; my (@selections, @rejections); for my $list_elem (@{$widgets_ref}) { if ($list_elem > $threshold) { push @selections, $list_elem; } else { push @rejections, $list_elem; } } return [\@selections, \@rejections]; }
It is an error to specify a ##{ without a matching ##}, or vice versa.
Note that the ##{ and ##} must be on lines by themselves (i.e. the block starts on the line after the ##{ and finishes on the line before the ##}).
Because subunits are supposed to be small, coherent chunks of a larger subroutine, it is often the case that they will consist of a single code "paragraph". That is: they will consist of a sequence of statements with no blank lines between them.
For example, in the frobnicate_widgets() subroutine, the normalizing step is a single paragraph of code:
frobnicate_widgets()
# Normalize widget list... $widgets_ref = [grep {defined} @{$widgets_ref}];
Because this happens frequently, Test::Subunits provides a shorthand for the ##{..##} notation: ##:
##:
This shortcut does not require a closing delimiter because it always terminates at the next blank line (i.e. the next line that is empty or contains only whitespace characters).
For example, we could rewrite the normalizing subunit from:
##{ ## sub normalize_list { ## my $widgets_ref = shift; $widgets_ref = [grep {defined} @{$widgets_ref}]; ## return $widgets_ref ## } ##}
to:
##: ## sub normalize_list { ## my $widgets_ref = shift; $widgets_ref = [grep {defined} @{$widgets_ref}]; ## return $widgets_ref ## }
This may seem like only a trivial improvement, but becomes much more significant, and much more useful, when you are also using wrappers (as described later).
As the above examples imply, within a block-inclusion or paragraph-inclusion, you can also insert simple ## markers to inject extra "virtual" lines of code around the actual lines of code being extracted.
##
You can also use these kinds of inclusions to inject entirely independent code fragments. For example, to enable inline testing (as described later).
You can even use them to prevent the Test::Subunits module from extracting code from a particular source file:
## BEGIN{ die 'No Test::Subunits for you!' }
or, less belligerently:
## __END__
The previous examples demonstrate how to set up testable subunits as independent subroutines, by injecting a sub declaration, some parameter unpacking, and a return statement around a selection of actual code.
sub
For example, given a subroutine:
sub sort_data { my ($data_ref, $limit) = @_; $limit //= 0; my @data = grep {defined && $_ >= $limit} @{$data_ref}; return sort { $b <=> $a } @data; }
You could define three subunits (check_limit(), check_data(), and check_sort(), by annotating the source like so:
check_limit()
check_data()
check_sort()
sub sort_data { my ($data_ref, $limit) = @_; ##{ ## sub check_limit { ## my $limit = shift; $limit //= 0; ## return $limit; ## } ##} ##{ ## sub check_data { ## my ($limit, $data_ref) = @_; my @data = grep {defined && $_ >= $limit} @{$data_ref}; ## return @data; ## } ##} ##{ ## sub check_sort { ## my @data = @_; return sort { $b <=> $a } @data; ## } ##} }
Test::Subunits would then extract the following code:
sub check_limit { my $limit = shift; $limit //= 0; return $limit; } sub check_data { my ($limit, $data_ref) = @_; my @data = grep {defined && $_ >= $limit} @{$data_ref}; return @data; } sub check_sort { my @data = @_; return sort { $b <=> $a } @data; }
which you could test like so:
use Test::Subunits 'My::Module'; use Test::More; plan tests => 3; is check_limit(undef), 0 => 'Default limit'; is_deeply [check_data(42, [40..44])], [42..44] => 'Data checked'; is_deeply [check_sort(40..44)], [reverse(40..44)] => 'Sorted'; done_testing();
Manually setting up each subunit subroutine gives you complete flexibility but is often tedious, as almost every subroutine requires more or less the same "boilerplate" code wrapped around it: define the subroutine, unpack the arguments, execute the extracted code, return the results.
So Test::Subunits provides a short-cut for specifying those kinds of wrappers.
In all the preceding examples, the ##{ and ##: markers had nothing else on their line. However, if there is any extra text on the same line after those opening delimiters, that text is used as the declaration of the "boilerplate" wrapper to be placed around the extracted code.
The format of such declarations is:
<subname> ( <parameters> ) --> <return expr>
which is converted to:
sub <subname> { my ( <parameters> ) = @_; ... return <return expr>; }
So we could rewrite the example from the previous section:
like so:
sub sort_data { my ($data_ref, $limit) = @_; ##{ check_limit ($limit) --> $limit $limit //= 0; ##} ##{ check_data ($limit, $data_ref) --> @data my @data = grep {defined && $_ >= $limit} @{$data_ref}; ##} ##{ check_sort (@data) --> () return sort { $b <=> $a } @data; ##} }
or, even more concisely using paragraph subunits, like so:
sub sort_data { my ($data_ref, $limit) = @_; ##: check_limit ($limit) --> $limit $limit //= 0; ##: check_data ($limit, $data_ref) --> @data my @data = grep {defined && $_ >= $limit} @{$data_ref}; ##: check_sort (@data) --> () return sort { $b <=> $a } @data; }
Note that, in both the above examples, the wrapper declaration for check_sort() explicitly specified no return value (i.e. an empty list: --> ()). However, that return behaviour will actually be pre-empted by the extracted return statement itself. This is a common technique for testing any subunit that includes a return statement.
--> ()
return
As a further optimization, if you leave off the return value specification entirely (i.e. you don't specify a --> at all), then the wrapper simply returns its own parameter list.
-->
That is, a wrapper declaration of the form:
<subname> ( <parameters> )
is converted to:
sub <subname> { my ( <parameters> ) = @_; ... return <parameters>; }
So we could rewrite the previous example even more compactly:
sub sort_data { my ($data_ref, $limit) = @_; ##: check_limit ($limit) $limit //= 0; ##: check_data ($limit, $data_ref) --> @data my @data = grep {defined && $_ >= $limit} @{$data_ref}; ##: check_sort (@data) return sort { $b <=> $a } @data; }
Note that this wrapper approach is the recommended way of using Test::Subunits, as it imposes as little extra mark-up as possible on your original source code.
Using Test::Subunits to extract parts of your code into separate subroutines is the preferred way of handling testing, as the extracted subunits can then be tested repeatedly and independently.
However, it is also possible to use Test::Subunits to place your tests directly within your source code. Note that this in not the recommended approach as it couples your tests too tightly with your source, "fattens" your source code with too much mark-up, and also makes it far more difficult to run multiple tests on the same code fragment.
Nevertheless, inlined tests are possible. For example, in your source file:
package My::Module; ## use Test::More; ## plan tests => 3; sub frobnicate_widgets { my ($threshold, $widgets_ref) = @_; # Normalize widget list... ##: ## my $widgets_ref = [1,undef,2,undef,3]; $widgets_ref = [grep {defined} @{$widgets_ref}]; ## is_deeply $widgets_ref, [1..3] => 'Normalized'; # Select widgets to be processed... ##{ ## my $widgets_ref = [1..5]; ## my $threshold = 3; my @selections; for my $list_elem (@{$widgets_ref}) { if ($list_elem > $threshold) { push @selections, $list_elem; } else { warn "Rejecting: $list_elem"; } } ## is_deeply \@selections, [4..5] => 'Selectioned'; ##} # Process selected widgets... return ##: ## subtest 'Selections frobnicated' => sub { ## like $_, qr/<frob>/ for map { frobnicate($_) } @selections; ## } } ## done_testing();
and then, in your test file:
use Test::Subunits 'My::Module'; # (No other code required)
Test::Subunits always compiles any code it extracts into the current namespace in which the module itself is loaded. So if you write:
use Test::Subunits 'Some::Module'; # ...which contains a check_input() subunit # ...and later... my @result = check_input(@test_data);
then any subunits extracted from Some/Module.pm will be compiled into the default main namespace, and so you will be able to call them directly, as above.
main
If you prefer to extract subunits into some other namespace, just load Test::Subunits within that package instead, like so:
{ package Extracted::Subunits; use Test::Subunits 'Some::Module'; # ...which contains a check_input() subunit } # ...and later... my @result = Extracted::Subunits::check_input(@test_data);
So now you would need to qualify any call to an extracted subunit that is made outside the namespace it was extracted into.
This extract-into-the-current-namespace behaviour is particulaly useful if you are using an OO testing framework such as Test::Class, as it means you can install your subunits directly into your test class's package, simply by loading Test::Subunits as part of the class declaration.
No argument supplied to 'use Test::Subunits'
Loading this module requires a single argument: a string containing the name of another module, or else containing a path to an appropriate source file. That argument tells the module where to locate and extract your subunits.
But you loaded the module without specifying that argument, so it has no idea where to look.
You just need to add an argument to your use Test::Subunits statement.
use Test::Subunits
Unless, you're attempting something tricky that involves loading the module but not using its normal functionality, in which case you probably want:
use Test::Subunits ();
Test::Subunits can't locate requested source file: %s
Loading this module requires a single argument: the name of another module or else a path to an appropriate source file.
You specified such a argument when loading the module, but the module couldn't find that source anywhere under your library path (i.e. it wasn't in any of the directories listed in @INC).
Did you mistype the source file's name, or do you need to add its directory to @INC using a use lib statement?
use lib
Unmatched ##{
Unmatched ##}
##{ and ##} subunit markers must appear in matched pairs in your source file. But the module found an unmatched marker.
Did you (or your editor's helpful autoformatting) accidentally write ## { or ## } instead?
## {
## }
Unrecognized subunit marker %s
The module currently accepts only three kinds of valid subunit marker: block markers (##{), paragraph markers (##:), and single line markers (##).
Any other (potential) marker consisting of ## followed by a printable character is reserved for future use. You used such a marker, but it isn't the future yet.
Did you mean one of the existing markers instead? Perhaps you wanted a simple ##, but forgot to leave at least one whitespace character after it?
Invalid wrapper specification
The module found trailing text on a line with a ##{ or a ##: opening delimiter. The only trailing text allowed on such lines must be a valid wrapper declaration of the form:
<subname> (<parameters>) --> <return expr>
or:
<subname> (<parameters>)
But the module encountered some other kind of text after the delimiter, became hopelessly confused, burst into tears, and gave up.
Did you make a mistake in the wrapper specification format? If so, see above.
Or did you try to inject some extra code on the same line as the opening delimiter...which, unfortunately, you simply can't (yes, that's a deliberate design decision).
In that second case, you'll need to rewrite something like:
##{ my $code = 'here';
as:
##{ ## my $code = 'here';
Test::Subunits requires no configuration files or environment variables.
None.
Works under Perl 5.10 or later.
None reported.
No bugs have been reported.
Please report any bugs or feature requests to bug-test-subunits@rt.cpan.org, or through the web interface at http://rt.cpan.org.
bug-test-subunits@rt.cpan.org
My sincere thanks to Mathias Fischer, who first brought this conflict between the needs of testing and the constraints of code performance to my attention.
Damian Conway <DCONWAY@CPAN.org>
<DCONWAY@CPAN.org>
Copyright (c) 2015, Damian Conway <DCONWAY@CPAN.org>. All rights reserved.
This module is free software; you can redistribute it and/or modify it under the same terms as Perl itself. See perlartistic.
BECAUSE THIS SOFTWARE IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY FOR THE SOFTWARE, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE SOFTWARE "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE SOFTWARE IS WITH YOU. SHOULD THE SOFTWARE PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING, REPAIR, OR CORRECTION.
IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR REDISTRIBUTE THE SOFTWARE AS PERMITTED BY THE ABOVE LICENCE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE THE SOFTWARE (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD PARTIES OR A FAILURE OF THE SOFTWARE TO OPERATE WITH ANY OTHER SOFTWARE), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES.
To install Test::Subunits, copy and paste the appropriate command in to your terminal.
cpanm
cpanm Test::Subunits
CPAN shell
perl -MCPAN -e shell install Test::Subunits
For more information on module installation, please visit the detailed CPAN module installation guide.