package Selenium::Remote::Mock::RemoteConnection; # ABSTRACT: utility class to mock the responses from Selenium server use strict; use warnings; use Moo; use JSON; use Carp; use Try::Tiny; use HTTP::Response; use Data::Dumper; extends 'Selenium::Remote::RemoteConnection'; has 'spec' => ( is => 'ro', default => sub { {} }, ); has 'mock_cmds' => ( is => 'ro', ); has 'fake_session_id' => ( is => 'lazy', builder => sub { my $id = join '', map +( 0 .. 9, 'a' .. 'z', 'A' .. 'Z' )[ rand( 10 + 26 * 2 ) ], 1 .. 50; return $id; }, ); has 'record' => ( is => 'ro', default => sub { 0 } ); has 'replay' => ( is => 'ro', ); has 'replay_file' => ( is => 'ro', ); has 'session_store' => ( is => 'rw', default => sub { {} } ); has 'session_id' => ( is => 'rw', default => sub { undef }, ); has 'remote_server_addr' => ( is => 'lazy', default => sub { 'localhost' } ); =for Pod::Coverage *EVERYTHING* =cut sub BUILD { my $self = shift; croak 'Cannot define replay and record attributes at the same time' if ( ( $self->replay ) && ( $self->record ) ); croak 'replay_file attribute needs to be defined' if ( ( $self->replay ) && !( $self->replay_file ) ); croak 'replay attribute needs to be defined' if ( !( $self->replay ) && ( $self->replay_file ) ); $self->port('4444'); if ( $self->replay ) { $self->load_session_store( $self->replay_file ); } } sub check_status { return; } sub load_session_store { my $self = shift; my $file = shift; croak "'$file' is not a valid file" unless ( -f $file ); open( my $fh, '<', $file ) or croak "Opening '$file' failed"; # here we use a fake session id since we have no way of figuring out # which session is good or not local $/ = undef; my $json = JSON->new; $json->allow_blessed; my $decoded_json = $json->allow_nonref(1)->utf8(1)->decode(<$fh>); close($fh); $self->session_store($decoded_json); } sub dump_session_store { my $self = shift; my ($file) = @_; open( my $fh, '>', $file ) or croak "Opening '$file' failed"; my $session_store = $self->session_store; my $dump = {}; foreach my $path ( keys %{$session_store} ) { $dump->{$path} = $session_store->{$path}; } my $json = JSON->new; $json->allow_blessed; my $json_session = $json->allow_nonref->utf8->pretty->encode($dump); print $fh $json_session; close($fh); } sub request { my $self = shift; my ( $resource, $params ) = @_; my $method = $resource->{method}; my $url = $resource->{url}; my $no_content_success = $resource->{no_content_success} // 0; my $content = ''; my $json = JSON->new; $json->allow_blessed; if ($params) { $content = $json->allow_nonref->utf8->canonical(1)->encode($params); } my $url_params = $resource->{url_params}; print "REQ: $method, $url, $content\n" if $self->debug; if ( $self->record ) { my $response = $self->SUPER::request( $resource, $params, 1 ); push @{ $self->session_store->{"$method $url $content"} }, $response->as_string; return $self->_process_response( $response, $no_content_success ); } if ( $self->replay ) { my $resp; my $arr_of_resps = $self->session_store->{"$method $url $content"} // []; if ( scalar(@$arr_of_resps) ) { $resp = shift @$arr_of_resps; $resp = HTTP::Response->parse($resp); } else { $resp = HTTP::Response->new( '501', "Failed to find a response" ); } return $self->_process_response( $resp, $no_content_success ); } my $mock_cmds = $self->mock_cmds; my $spec = $self->spec; my $cmd = $mock_cmds->get_method_name_from_parameters( { method => $method, url => $url } ); my $ret = { cmd_status => 'OK', cmd_return => 1 }; if ( defined( $spec->{$cmd} ) ) { my $return_sub = $spec->{$cmd}; my $mock_return = $return_sub->( $url_params, $params ); if ( ref($mock_return) eq 'HASH' ) { $ret->{cmd_status} = $mock_return->{status}; $ret->{cmd_return} = $mock_return->{return}; $ret->{cmd_error} = $mock_return->{error} // ''; } else { $ret = $mock_return; } $ret->{session_id} = $self->fake_session_id if ( ref($ret) eq 'HASH' ); } else { $ret->{sessionId} = $self->fake_session_id; } return $ret; } 1; __END__ =pod =head1 DESCRIPTION Selenium::Remote::Mock::RemoteConnection is a class to act as a short-circuit or a pass through to the connection to a Selenium Server. Using this class in place of L allows to: =over =item * record interactions with the Selenium Server into a JSON file =item * replay recorded interactions from a JSON file to mock answers from the Selenium Server =item * mock responses to specific functions =back =head1 SYNOPSIS =head2 Record interactions use strict; use warnings; use Selenium::Remote::Driver; use Selenium::Remote::Mock::RemoteConnection; # create a new Mock object to record the interactions with Selenium # Server my $mock_connection = Selenium::Remote::Mock::RemoteConnection->new( record => 1 ); # the Mock object is passed to the driver in place of what would be # a regular Selenium::Remote::RemoteConnection object my $driver = Selenium::Remote::Driver->new( remote_conn => $mock_connection ); # always store the session id, as it will become undef once # $driver->quit is called my $session_id = $driver->session_id; # do all the selenium things and quit $driver->get('http://www.google.com'); $driver->get('http://www.wikipedia.com'); $driver->quit; # dump the session to a file $mock_connection->dump_session_store( 'my_record.json' ); This code, above doing some basic Selenium interactions, will end up generating a JSON file containing all the requests and their responses for your Selenium session. The JSON file looks like this : '{ "HTTP_REQUEST_URL {request_parameters}":[response1,response2,...], ... }' The reason why we store array of responses is that the exact same request can be made more than once during a session, so we have to store every response to the same requests. =head2 Replay interactions #!perl use strict; use warnings; use Test::More; use Test::Selenium::Remote::Driver; use Selenium::Remote::Mock::RemoteConnection; my $mock_connection_2 = Selenium::Remote::Mock::RemoteConnection->new( replay => 1, replay_file => 'my_record.json' ); # javascript + version parameters added or else it will not work my $driver = Test::Selenium::Remote::Driver->new( remote_conn => $mock_connection_2, javascript => 1, version => '' ); $driver->get_ok('http://www.google.com'); $driver->get_ok('http://www.wikipedia.com'); $driver->quit; done_testing; Using the file generated with the recording snippet from the section before, we are able to mock the responses. Note that there is one small limitation (that I hope to remove in future versions), is that a record generated with L is not directly useable with L. This is mainly because the way the two instances are created are a bit different, which leads to different requests made, for creating a session for instance. For now, what works for sure is recording and replaying from the same class. =head2 Mock responses #!perl use Test::More; use Test::Selenium::Remote::Driver; use Selenium::Remote::WebElement; use Selenium::Remote::Mock::Commands; use Selenium::Remote::Mock::RemoteConnection; my $spec = { findElement => sub { my (undef,$searched_item) = @_; return { status => 'OK', return => { ELEMENT => '123456' } } if ( $searched_item->{value} eq 'q' ); return { status => 'NOK', return => 0, error => 'element not found' }; }, getPageSource => sub { return 'this output matches regex'}, }; my $mock_commands = Selenium::Remote::Mock::Commands->new; my $successful_driver = Test::Selenium::Remote::Driver->new( remote_conn => Selenium::Remote::Mock::RemoteConnection->new( spec => $spec, mock_cmds => $mock_commands ), commands => $mock_commands, ); $successful_driver->find_element_ok('q','find_element_ok works'); dies_ok { $successful_driver->find_element_ok('notq') } 'find_element_ok dies if element not found'; $successful_driver->find_no_element_ok('notq','find_no_element_ok works'); $successful_driver->content_like( qr/matches/, 'content_like works'); $successful_driver->content_unlike( qr/nomatch/, 'content_unlike works'); done_testing(); Mocking responses by hand requires a more advanced knowledge of the underlying implementation of L. What we mock here is the processed response that will be returned by L to '_execute_command' call. To accomplish this we need : =over =item * a spec: a HASHREF which keys are the name of the methods we want to mock. Note that those keys should also be valid keys from the _cmds attribute in L. The value of each key is a sub which will be given two parameters: =over =item * $url_params : the values that should have been replaced in the URL For instance, on the example above, it would have been: { session_id => 'some_session_id'} =item * $params : the original parameters of the request. On the example above it would have been: { value => 'q', using => 'xpath'} =back The sub used as a value in the spec is not expected to return anything, so you have to craft very carefully what you return so that it will produce the expected result. =item * a mock_cmd: a L object. This is used mainly to hijack the normal commands so that placeholders do not get replaced in the URLs. =back =head1 BUGS This code is really early alpha, so its API might change. Use with caution ! =cut