use strict; use warnings; package MooseX::Role::REST::Consumer; use MooseX::Role::Parameterized; use MooseX::Role::REST::Consumer::Response; use File::Spec; use REST::Consumer; use Try::Tiny; use URI::Escape; use Module::Load; our $VERSION = '0.003'; my @HTTP_METHODS = qw(get put post delete); parameter service_host => ( lazy => 1, default => '', ); parameter service_port => (); parameter resource_path => ( lazy => 1, default => sub { undef }, ); parameter content_type => ( isa => 'Str', required => 1, default => 'application/json', ); parameter header_exclude => ( isa => 'HashRef', default => sub {{}}, ); parameter retry => ( required => 1, default => 0, ); parameter timeout => ( default => 1, ); parameter query_params_mapping => ( isa => 'HashRef', default => sub {{}}, ); parameter useragent_class => ( isa => 'Str', ); role { my $p = shift; my $service_host = $p->service_host; my $service_port = $p->service_port; my $resource_path = $p->resource_path; my $content_type = $p->content_type; my $retry = $p->retry; my $timeout = $p->timeout; my $query_params_mapping = $p->query_params_mapping; my $useragent_class = $p->useragent_class; #Note: since this is a parametrized role then only one class will be closing #over this variable and this instance only depends on role parameters my $consumer_instance; #Note: $class varible in methods of this role can be an instance and a class name #For GET requests this is usually a class name, for POST requests #this would usually be an instance. #Be carefull! method 'consumer' => sub { my ($class) = @_; return $consumer_instance if($consumer_instance); my $user_agent; if($useragent_class) { Module::Load::load($useragent_class); $user_agent = $useragent_class->new; }; my $client = REST::Consumer->new( host => $service_host, timeout => $timeout, ($user_agent ? (ua => $user_agent) : ()), ($service_port ? (port => $service_port) : ()) ); $client->user_agent->use_eval(0); $consumer_instance = $client; return $client; }; method 'call' => sub { my ($class, $method, %params) = @_; my %request_headers = $class->request_headers( header_exclude => $p->header_exclude, headers => $params{headers}, content_type => $p->content_type, method => $method, ); $content_type = delete $request_headers{'Content-Type'}; if ($params{access_token}) { $request_headers{Authorization} = 'Bearer ' . $params{access_token}; } my %query_params = $class->query_params( params => $params{params}, query_params_mapping => $query_params_mapping, ); my $path = $class->request_path( resource_path => $resource_path, route_params => delete $params{route_params}, path => delete $params{path}, ); my ($data, $error, $timeout_error) = (undef, '', 0); my %request = ( params => \%query_params, content => $params{content}, path => $path, content_type => $content_type, headers => ( %request_headers ? [ %request_headers ] : undef ), ); #my $logger = $class->get_logger #$logger->start_request( # type => $method, # path => $path, # response => $data, # request => \%request #); my $consumer = $class->consumer; #we want same instance throughout the whole process $consumer->host(delete $params{host}) if $params{host}; if (my $timeout_override = delete $params{timeout} ) { $consumer->timeout($timeout_override); } my $try = 0; while ($try <= $retry) { my $is_success; try { $try++; $data = $consumer->$method(%request); $is_success = 1; } catch { $error = $_; $timeout_error++ if $error =~ /read timeout/; }; last if $is_success; } $consumer->timeout($timeout); $consumer->host($service_host); #$logger->finish($log_entry, { # type => $method, # path => $path, # response => $data, # request => \%request #}); # TODO: It's confusing as to how we handle errors. # ie: message should be called "error_message" and # we should inspect the response content for possible error messages return MooseX::Role::REST::Consumer::Response->new( data => $data, is_success => !$error, error_message => "$error", request => $consumer->last_request, content_type => $content_type, response => $consumer->last_response, ); }; # Note: REST::Consumer doesn't support OPTIONS/PATCH for my $method ( @HTTP_METHODS ) { method $method => sub { my $class = shift; $class->call($method, @_); } } method 'request_headers' => sub { my ($class, %params) = @_; my %request_headers = ($params{headers} ? %{delete $params{headers}} : ()); # Strange mutation going on here: # 1. First we set the content_type in the headers if we have one set in parameter definition # 2. However, we won't override anything that is passed explicitly into a method call Class->post # 3. Next we delete anything that needs to be removed from the header # 4. Finally we explicltly pull out content-type from the request_headers # to make REST::Consumer happy $request_headers{'Content-Type'} = $params{content_type} unless $request_headers{'Content-Type'}; delete @request_headers{@{$params{header_exclude}->{$params{method}}}} if $params{header_exclude}->{$params{method}}; return %request_headers; }; method 'request_path' => sub { my ($class, %params) = @_; my $resource_path = $params{resource_path}; my $route_params = $params{route_params}; my $params_path = $params{path}; my @path = (defined $resource_path ? $resource_path : ()); if($params_path) { if(ref($params_path) eq 'ARRAY') { push(@path, @$params_path); } else { push(@path, $params_path); } } my $path = File::Spec->catfile(@path) . ( $resource_path && $resource_path =~ m{[^/]+/$} ? '/' : ''); #We support two ways of substituting params here: # /:param/ - name has to have '/' or end of string after it # /has_:{param}_value - surround param name with '{}' #Note: we go through complete incremental parsing/fetching url params to avoid #params substituted values being mached against other params names #We also verify that all parameters have been substituted my $result = ''; while($path =~ /\G(.*?)(:(?:\{([^}]+?)\}|(\w+)(?=\W|$)))/gc) { $result .= $1; if($2) { my $name = $3 || $4; # only one of them could match if(exists $route_params->{$name}) { $result .= URI::Escape::uri_escape_utf8($route_params->{$name} // ''); } else { die "Found parameter $name in path but it wasn't set in parameters hash"; } } } if($path =~ /\G(.+)$/g) { $result .= $1; } return $result; }; method 'query_params' => sub { my ($class, %params) = @_; my %query_params; if($params{params}) { # TODO: I don't think having this strict mapping in place # makes a lot of sense. Ideally we should be able to pass in # any query params that we want if ( $params{query_params_mapping} ) { while(my ($name, $url_name) = each(%{$params{query_params_mapping}})) { # Check for method canness here probably $query_params{$url_name} = delete($params{params}->{$name}); } } else { %query_params = %{$params{params}}; } } return %query_params; }; method 'parameters' => sub { $p }; }; __END__ =pod =head1 NAME MooseX::Role::REST::Consumer =head1 VERSION version 0.003 =head1 SYNOPSIS package Foo; use Moose; with 'MooseX::Role::REST::Consumer' => { service_host => 'somewhere.over.the.rainbow', resource_path => '/path/to/my/resource/:id', }; my $object = Foo->get(route_params => {id => 1}); if ($object->is_success) { print $object->data->{something_that_came_back}; } =head2 DESCRIPTION At Shutterstock we love REST and we take it so seriously that we think our code should be RESTfully lazy. Now one can have a Moose model without needing to deal with all the marshalling details. =head3 Schema Definitions/Configuration When setting up a class the following are the supported parameters that L will support. For example a typical configuration would looke like the following: with 'MooseX::Role::REST::Consumer' => { service_host => 'host.name', resource_path => '/path/to/my/resource/:id' timeout => 10, }; =over =item content_type By default the content type is set to "application/json" =item header_exclude Acts as a filter, will exclude any header information. header_exclude => { post => 'X-Foo', } =item resource_path This is the path of the resource. IE: /foo/bar/:id The :id is a route parameter which will be filled in as specified by the "route_params" hashref. =item retry This is an explicit retry. Even if the service times out, it will retry using this value. This retry is different than what L offers. =item service_host The hostname to the service =item service_port The port for the hostname. =item timeout Configuration level timeout. This is global on each request. =item useragent_class Experimental way of overriding L's useragent. Right now MooseX::REST::Consumer uses L =back =head3 METHODS =over =item get( route_params => {...}, params => {...} ) Will use REST::Consumer::get to lookup a resource by supplied resource_path and substitution of route_params =item post( route_params => {...}, content => '...' ) Will perform a POST request with REST::Consumer::post. The data will the Content-Type of application/json by default. =item Other supported HTTP methods DELETE and PUT: delete(%params) and put(%params) =back =head3 Supported Method Parameters =over =item route_params => {...} These will be substituted into the package route definition. =item params => {...} Passed into L as a set of key/value query parameters. =item headers => {...} Any extra HTTP request headers to send. =item content => '' Passed into L. This is the body content of a request. =item timeout => '' Timeout override per request. Note that the 'timeout' is subject to interpretation by your underlying UserAgent class. For example, LWP::UserAgent treats the timeout as being C. This means that if you specify a timeout of 5 seconds and issue a request using LWP::UserAgent, each request that the UserAgent makes to fulfill your request will have its own timeout of 5 seconds. This becomes important if the API that you are talking to starts giving you 3xx redirects: while you might expect a timeout to occur within 5 seconds, the API might instruct your UserAgent to make a few subsequent requests, and each one will have your initial timeout applied to it. Different UserAgent classes implement timeouts differently. L, for example, has a global timeout value, where all requests must be fulfilled within C clock seconds. =back =head3 Response Object =over The response object is created and passed back whenever any of the supported HTTP methods are called. See L. =back =head2 =head1 SEE ALSO L, L, L =head2 AUTHORS The Shutterstock Webstack Team and alumni (Logan Bell, Jon Hogue, Vishal Kajjam, Belden Lyman, Nikolay Martynov, and Kurt Starsinic). =head2 COPYRIGHT AND LICENSE This software is copyright (c) 2014 by Shutterstock Inc. This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself. =cut