NAME HTTP::Response::Switch - handle many HTTP response possibilities VERSION This module is part of distribution HTTP-Response-Switch v1.1.1. This distribution's version numbering follows the conventions defined at semver.org . SYNOPSIS Define a "dispatcher" for the application code to use directly: package MyProject::WebResponses; use Moose; with 'HTTP::Response::Switch'; # All of the handlers are defined under this namespace. sub handler_namespace { 'MyProject::WebResponse' } # Always resort to these handlers before giving up on a response. sub default_handlers { qw( ConfirmAction LoginForm ) } # Throw an exception of this class if the response can't be handled # by any of the specified handlers. sub default_exception { 'MyProject::Error::BadWebResponse' } # Load all of the handlers and the exception class at compile time # (recommended). __PACKAGE__->load_classes; Then, in code that actually talks to the web server: my $http_response = WWW::Mechanize->new->post( URL, USER_DATA ); # Either the right data will be returned, or there was a problem # with some user data, or the session got logged out, or the system # is down for maintenance, or something entirely unanticipated will # happen. Whatever the case, do the right thing. use TryCatch; try { my @expected_data = MyProject::WebResponses->handle( $http_response, qw{ RawDataPage RawDataForm }, ); return SOMETHING_BUILT_FROM( @expected_data ); } catch (MyProject::Error::NeedConfirmation $e) { CLICK_YES; TRY_AGAIN; } catch (MyProject::Error::NotLoggedIn $e) { LOG_BACK_IN; TRY_AGAIN; } # Don't catch all possible exceptions here; let calling code decide # what to do in some cases. Also define each handler: package MyProject::WebResponse::ConfirmAction; ... package MyProject::WebResponse::LoginForm; ... package MyProject::WebResponse::RawDataPage; ... package MyProject::WebResponse::RawDataForm; ... See "SYNOPSIS" in HTTP::Response::Switch::Handler for example handler definitions. DESCRIPTION Sometimes the only possible way to communicate with an online service is through a web application intended for human consumption--dealing with cookies, forms and parsing of HTML, perhaps with the help of WWW::Mechanize and Web::Scraper. When automating such a web application, it may be unsafe to assume that a specific request will always trigger a specific response. For example, requesting bank transactions from an Internet Banking server could result in a response containing: 1 a CSV file of bank transactions (the "expected" response); 2 an HTML page indicating an input error and presenting a form; 3 an HTML page indicating that the session has been terminated; 4 an HTML page indicating an Internal Server Error; or 5 something else entirely unanticipated. Some of those "unexpected" or "undesired" responses may require special behaviour, in which case it isn't appropriate to simply "die" on such responses. Moreover, some of those responses could potentially occur on any communication with the web application, creating the need to test for them at every single point that the web server is contacted. This distribution aims to abstract away some of the code verbosity that would be required for this by providing Moose roles for: 1. "Handler" classes, which look at a specific HTTP::Response object with a single concern in mind and either return structured data, throw an exception, or indicate that they don't know how to handle that specific response. 2. A "dispatcher" class which takes an HTTP::Response object and passes it through a chain of the aforementioned "handler" classes until one of them returns structured data or throws an exception; or, if none of them do, indicates that that specific response truly is "unexpected." USAGE Refer to the "SYNOPSIS" above. Further information on configuring a "dispatcher" class follows in subsequent sections. See HTTP::Response::Switch::Handler for further information on writing "handler" classes. In order to better understand how this distribution's code is intended to be used, inspecting the source code of the following known dependent distributions may also be helpful: * Finance::Bank::Bankwest v1.2.2 and later METHODS TO DEFINE IN CONSUMING CLASSES These methods can be defined in a consuming class in order to customise functionality provided by the role. Some of these methods must be defined in order for the consuming class to work. handler_namespace sub handler_namespace { 'MyProject::WebResponse' } The namespace under which handler classes are to be found. This method must be defined in every consuming class. default_handlers sub default_handlers { qw( LoginForm ) } A list of handler classes (minus the "handler_namespace") that should be asked to process an HTTP::Response if no other handler accepts the response first. If not defined in the consuming class, an empty list is assumed. default_exception sub default_exception { 'MyProject::Error::BadWebResponse' } sub default_exception { die 'unexpected HTTP response' } # default The exception class to throw if, during a call to "handle" by external code, no handlers accept the HTTP::Response in question. The instance of this class will be passed the HTTP response object when created, via parameter "response". Alternatively, this method can be written to throw an exception directly (e.g. using "die"). If not defined in the consuming class, the default behaviour in this situation is to just "die" with the message "unexpected HTTP response". The Throwable distribution is suggested for building object-oriented exception classes. METHODS AVAILABLE FROM WITHIN CONSUMING CLASSES These methods are intended only to be called on a consuming class by that class itself, and not by code external to the class. load_classes __PACKAGE__->load_classes; Use Module::Find to locate all of the modules under the "handler_namespace" and load them into memory. Also load the "default_exception" class if one is specified. The handler modules must be loaded into memory before the first call to "handle" occurs. Loading them at compile time is recommended; to do this, call "load_classes" within the consuming class in the manner shown above, outside all "sub" definitions. load_handlers A deprecated alias for "load_classes". Historically, "load_handlers" did not load the "default_exception" class. METHODS AVAILABLE TO CALLERS OF CONSUMING CLASSES These methods are intended to be called directly on a consuming class by code external to the class. handle my @ret = MyDispatcher->handle( $http_response, @handlers ); Pass the HTTP::Response object in $http_response to each handler class defined in @handlers (if any, and minus the "handler_namespace"), then to each of the "default_handlers", until one of the handlers accepts the response for handling and either returns appropriate data or throws an appropriate exception. If none of the handlers accept this $http_response, throw the "default_exception". All of the handler classes (and the "default_exception" class) are assumed to have been loaded into memory *before* this method is first called. See "load_classes". SEE ALSO * HTTP::Response::Switch::Handler SUPPORT Bugs / Feature Requests Please report any bugs or feature requests by email to "bug-http-response-switch at rt.cpan.org", or through the web interface at . You will be automatically notified of any progress on the request by the system. Source Code The source code for this distribution is available online in a Git repository. Please feel welcome to contribute patches. git clone git://github.com/lx/perl5-HTTP-Response-Switch AUTHOR Alex Peters COPYRIGHT AND LICENSE This software is copyright (c) 2013 by Alex Peters. This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself. The full text of the license can be found in the 'LICENSE' file included with this distribution.