Web Automation fun with Selenium RC
2008-Nov-22 19:50 So, many of you are probably familiar with Selenium for deploying automated web-application testing. I've recently begun a small side project that involves automated calls to web-sites that are AJAX and Javascript heavy in order to scrape content from the pages. Typically, I would turn to my trusty friend cURL for most screen-scraping, but in cases where the content on a page is dynamically generated by AJAX, its a lot easier to emulate a browser as closely as possible. Enter Selenium-RC, and in my case, the PHP Selenium driver. More on this topic later, but for now, if you are facing a strange intermittent problem with the Selenium server complaining about SocketException: Broken Pipe , here is your solution: The command handler in the Testing/Selenium.php uses (apparently) an incomplete socket connection routine. I have found that the same handler provided by the PHPUnit's Selenium Driver is much more robust.
protected function doCommand($verb, $args = array())
{
$url = sprintf('http://%s:%s/selenium-server/driver/?cmd=%s', $this->host, $this->port, urlencode($verb));
for ($i = 0; $i < argnum =" strval($i" s="%s',">sessionId)) {
$url .= sprintf('&%s=%s', 'sessionId', $this->sessionId);
}
if (!$handle = fopen($url, 'r')) {
throw new Testing_Selenium_Exception('Cannot connected to Selenium RC Server');
}
/*
stream_set_blocking($handle, false);
stream_set_timeout($handle, 0, 60000);
$response = stream_get_contents($handle);
fclose($handle);
*/
stream_set_blocking($handle, 1);
stream_set_timeout($handle, 0, 60000);
$info = stream_get_meta_data($handle);
$response = '';
while (!$info['eof'] && !$info['timed_out']) {
$response .= fgets($handle, 4096);
$info = stream_get_meta_data($handle);
}
fclose($handle);
return $response;
}
Replace the doCommand function with the above code, and you should find better results. More on this to come...
View Printer Friendly Version
Email Article to Friend
Reader Comments