Devel::Decouple - Decouple code from imported functions
This module is intended to facilitate the testing and refactoring of legacy Perl code.
To generate a simple report about a module's or script's use of imported functions you can simply use the Devel::Decouple subclass Devel::Decouple::DB via the debugger.
Devel::Decouple
Devel::Decouple::DB
perl -d:Decouple::DB myscript.pl
Then, perhaps in a test file, you can redefine all those functions easily to decouple the problematic dependencies.
# for the given module automatically redefine ALL # imported functions as no-ops... the default my $DD = Devel::Decouple->new(); $DD->decouple( 'Some::Module' );
This module also provides for a high degree of customization of how, or whether, functions will be redefined, and to do so there is a clean declarative syntax.
# only decouple the named module from those modules that are explicitly listed my @modules = qw{ Another::Module And::Another }; $DD->decouple( 'Some::Module', from @modules ); # you MUST use a literal array here, not a list! # use the default (no-op) code stub... except where an alternative is # specified or the original imported function is explicitly preserved my @stooges = qw{ larry moe curly }; $DD->decouple( 'Some::Module', function 'foo', as { return 2.167 }, function 'bar', as { return 'hello' }, function 'baz', preserved, functions @stooges, # again, you MUST use a literal array here! as { return "I'm a stooge!" }); # define a custom default code stub, as with this simple stack trace, # that re-dispatches to the original code $DD->decouple( 'Some::Module', default_sub, as { warn "calling '", (caller(0))[3], "' from ", (caller(1))[3], " at line ", (caller(0))[2]; shift->(@_) });
When testing software it's often useful to use dummy, or 'mock', objects to represent external dependencies. This practice presupposes that the code being tested is object-oriented and can usually simply accept these objects to its constructor, i.e. the dependencies are loosely-coupled and we can use simple dependency-injection tactics to test the code.
If only the world were always so simple.
Unfortunately, legacy Perl code-bases often tightly couple external dependencies by creating those dependencies either inside modules, or inside monolithic scripts.
When faced with maintaining and extending such code this brings to light several obvious and serious problems. First, pervasive hardcoding and tight coupling ties us to a particular (often production) implementation. This is a testing nightmare and it can be dangerous to even attempt to test some code in production environments without long and careful consideration of the consequences. Second, it seems to require us to make a lot of changes all over the code-base to try to isolate particular moving-parts so that we can specify controlled behavior. This controlled behavior could stem the side effects of the tightly coupled code in order to facilitate testing, but how can we change the code safely and confidently to do this without first having tests? It's a chicken and egg scenario.
Further, because legacy Perl code uses predominantly functional interfaces, rather than object orientation, it often pollutes the consumers name-space with imported functions (and other symbols) that can be difficult to identify. Our only recourse in the past has generally been to eyeball the code, perhaps several thousand lines of it, and make careful notes.
Devel::Decouple is designed to do static analysis on the parts of the code that you intend to create Characterization tests (http://en.wikipedia.org/wiki/Characterization_Test) for to facilitate refactoring. It can programmatically identify the imported functions that have actually been called for the given name-space. It can also automatically install default stub functions into the symbol table to replace these functions, or allow the specification of individual and specialized behavior (e.g. returning mocked data from a tabular data-file or here-doc in a standard Test::More test script).
A simple contructor. It takes no arguments and returns an empty object.
A method to specify what and how to decouple the code in question. Please refer to the SYNOPSIS and "SYNTACTIC SUGAR" sections for full details about the options taken and how to use them.
Return a simple formatted report about the use of imported functions.
Un-patch the functions given in LIST.
Returns the name of the module being operated upon.
Returns the canonical name of the module being operated upon.
Returns the names of the modules that are being decoupled from the module being operated upon. That is, either the constrained list of modules given to decouple in the from sub-clause (but only if their functions were actually called), or, if unconstrained, the list of all modules whose exported functions were actually called in the primary code.
decouple
from
Returns the list of all imported functions that were actually called.
Returns the list of all functions, but not methods, in the primary code's immediate namespace.
Devel::Decouple provides a decalrative syntax to specify how and what to decouple. Each statement can be thought of as a clause, and each clause (or sub-clause) needs to be seperated with a (plain) comma (i.e. not a 'fat-comma', aka 'quoting-comma').
The order of the clauses is significant but intuitive. The first clause is always the name of the module to operate upon followed by an optional array of modules you wish to decouple listed in a from sub-clause. A function, functions, or default_sub clause is always followed by either an as sub-clause, which specifies the behavior you wish to implement for that entity, or the preserved marker.
function
functions
default_sub
as
preserved
The special marker preserved is used in place of an as sub-clause to indicate that no custom behavior should be added to the symbol table for the entity in question.
Specify the explicit list of modules to decouple from the primary module. If not provided Devel::Decouple will default to all modules whose imported functions were actually called.
The name of a function in the primary module's name-space for which you wish to specify custom behavior. Use an as clause to define the behavior itself.
The names of one or more functions in the primary module's name-space for which you wish to specify custom shared behavior. Use an as clause to define the behavior itself.
Behavior to install in the symbol table. The as sub-clause must follow function, functions, or default_sub separated by a comma.
Specify the default override behavior for any called (imported) function that has no other explicitly defined custom behavior and which is not preserved. Use an as clause to define the behavior itself. If not specified the default default_sub is a no-op that always simply returns undef.
undef
It is of course possible for the default_sub to specify that the previously defined behavior should be preserved:
# with the default_sub "preserved" and without any other behavior specified # the following statement would not affect the normal operation of My::Module... my $Decoupler = Devel::Decouple->new; $Decoupler->decouple( 'My::Module', default_sub, preserved );
If combined with specialized behavior in the as sub-clause of a function or functions clause this permits surgical precision in what functions are redefined and how. See also, "Devel::Decouple ADVANCED TOPICS".
The preserved marker can be used in place of any as sub-clause to indicate that the previous behavior of the imported function should be preserved.
Internally the patching of the symbol table is carried out by the Monkey::Patch module. As such there is a 'stack' of code definitions preserved for all function definitions up to and including the original source definition. As with all stacks we can push new entities (code definitions) onto the stack and also pop them off. By using this feature you can modify the behavior of your functions on-the-fly.
Monkey::Patch
There is also the ability to redispatch to the next most recent previously defined code on the stack. By using this feature you can essentially create custom wrappers. These wrappers can alter the code by adding new features and behavior to previous code definitions.
These are very powerful features and can be used to great advantage individually or together.
Push code-definition entities onto the stack by calling decouple on a new Devel::Decouple object instance and defining some new behavior, pop by calling revert for specified functions on that instance... or by undefing the object entirely.
revert
In contrast, calling decouple on an already defined Devel::Decouple object will overwrite the definitions at that location on the stack. This behavior may not be what you intend, and care should be taken to preserve the behavior that's already been set-up, e.g. by using a default_sub, preserved clause when redefining a Devel::Decouple object instance.
default_sub, preserved
# Adapted from override.t in the Devel::Decouple test suite... ### THE ORIGINAL BEHAVIOR # GOT EXPECTED is( TestMod::Baz::inhibit(), "I'm inhibited" ); is( TestMod::Baz::prohibit(), "I'm prohibited" ); ### OVERRIDING: pushing new function definitions onto the stack... ### PUSH my $DD1 = Devel::Decouple->new; $DD1->decouple( 'TestMod::Baz', from @modules, function 'prohibit', as { return 2 }, function 'inhibit', as { return 3 } ); is( TestMod::Baz::prohibit(), 2 ); is( TestMod::Baz::inhibit(), 3 ); ### PUSH my @functions = qw{ inhibit prohibit }; my $DD2 = Devel::Decouple->new; $DD2->decouple( 'TestMod::Baz', functions @functions, as { return "defined by \$DD2" } ); is( TestMod::Baz::inhibit(), "defined by \$DD2" ); is( TestMod::Baz::prohibit(), "defined by \$DD2" ); ### PUSH my @modules = qw{ TestMod::Foo TestMod::Bar }; my $DD3 = Devel::Decouple->new; $DD3->decouple( 'TestMod::Baz', from @modules, function 'prohibit', preserved, function 'inhibit', as { return "defined by \$DD3" } ); ### NOTE: 'prohibit' is 'preserved'... thus, 'defined by $DD2' is( TestMod::Baz::inhibit(), "defined by \$DD3" ); is( TestMod::Baz::prohibit(), "defined by \$DD2" ); ### PUSH my $DD4 = Devel::Decouple->new; $DD4->decouple( 'TestMod::Baz', from @modules, function 'inhibit', as { return "defined by \$DD4" } ); ### NOTE: 'prohibit' is redefined by the no-op 'default_sub' is( TestMod::Baz::inhibit(), "defined by \$DD4" ); is( TestMod::Baz::prohibit(), undef ); ### REVERTING: popping function definitions off the stack... ### POP undef $DD4; is( TestMod::Baz::inhibit(), "defined by \$DD3" ); is( TestMod::Baz::prohibit(), "defined by \$DD2" ); ### POP $DD3->revert( 'inhibit' ); is( TestMod::Baz::inhibit(), "defined by \$DD2" ); is( TestMod::Baz::prohibit(), "defined by \$DD2" ); ### POP undef $DD2; is( TestMod::Baz::inhibit(), 3 ); is( TestMod::Baz::prohibit(), 2 ); ### And so on...
Caution should be exercised here as the Devel::Decouple object instances are actually collections of stacks maintained for each function. It can become confusing to keep track of what behavior was specified in which objects and how those objects relate to the ordering of the underlying stacks. Please see "Devel::Decouple Testing Kata" for some ideas about how to deal with this issue.
For any function definition you provide you have access to the immediately previous code definition on the stack. This behavior is passed to your new function definition as a sub-ref in the first element of @_. The remainder of @_ contains the original arguments to the function.
@_
# Define a 'default_sub' that reports the arguments a function was called with # and then delegates back to the previous behavior on the stack... my $Decoupler = Devel::Decouple->new; $Decoupler->decouple( 'My::Module', default_sub, as { my $orig = shift; warn ("'", caller(0))[3], "' called with args: ", join '=|=', @_; $orig->(@_)}, # redispatch! );
If you wish, you can redispatch to any arbitrary depth down the stack of function definitions in this way, continually passing the original arguments along until ultimately terminating at the original source function definition.
Combining the use of the stack and this re-dispatch mechanism is a powerful way to implement stack traces, data serialization, and other exploratory testing for all the functions in a complex or convoluted name-space. Devel::Decouple isn't just for redefining the imported functions, although these can be automated to a greater extent by the use of the default_sub clause. Devel::Decouple can redefine native functions when listed explicitly in a function or functions clause.
Using such a technique can help you to understand the behavior of legacy Perl code so that you can clear the seemingly insurmountable hurdle of writing the initial tests. By automating much of the exploratory analysis of dense code it can also aid you in gradually adding increasingly granular test cases as you begin to better understand the behavior of your code. Once you understand the behavior of your code well and have a handle on complex dependencies then safely refactoring toward a more modern and maintainable idiom is possible.
Red. Green. Refactor.
Helpful code snippets coming soon...
Devel::Decouple requires no configuration files or environment variables.
Devel::Decouple does not currently identify imported symbols other than subroutines.
Please report any bugs or feature requests to <dev@namimedia.com>.
Montgomery Conner <mconner@cpan.org>
Copyright (c) 2012, NamiMedia <dev@namimedia.com>. 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 Devel::Decouple, copy and paste the appropriate command in to your terminal.
cpanm
cpanm Devel::Decouple
CPAN shell
perl -MCPAN -e shell install Devel::Decouple
For more information on module installation, please visit the detailed CPAN module installation guide.