Google translate

Question

Google

Let's run this script:

 <?php
// берём чешское слово 'Koláče' (пироги), подготавливаем к передаче через URL 
$text = urlencode('Koláče');
// формируем запрос
$query = "http://translate.google.com/translate_a/t?client=x&text={$text}&sl=cs&tl=en";
// который в итоге выглядит так: 
// http://translate.google.com/translate_a/t?client=x&text=Kol%C3%A1%C4%8De&sl=cs&tl=en


// посылаем для перевода
// этот кусок взял отсюда: 
//http://stackoverflow.com/questions/542046/php-file-get-contentsloc-fails
$curl = curl_init();
curl_setopt($curl, CURLOPT_RETURNTRANSFER, true);
curl_setopt($curl, CURLOPT_URL, $query);
$response = curl_exec($curl);
curl_close($curl);
// но вместо него можно и проще, результат будет одинаков:
// $response = file_get_contents($query);


echo $response;
?>

in response, we get a JSON object with this garbage:

{"sentences":[{
  "trans":"Kol\u0102\u0104\u00C4 e",
  "orig":"Kol\u0102\u0104\u00C4 e",
  "translit":"",
  "src_translit":""
}],"src":"cs","server_time":2}

Pay attention to the orig field , in some unknown way the string 'Kol%C3%A1%C4%8De' was transformed into 'Kol\u0102\u0104\u00C4 e'.
And if the request translate.google.com/translate_a/t?client=x&text=Kol%C3%A1%C4%8De&sl=cs&tl=en is simply entered into the address bar of the browser, we will get a beautiful correct answer:

{"sentences":[{
  "trans":"Pies",
  "orig":"Koláče",
  "translit":"",
  "src_translit":""
}],"src":"cs","server_time":41}

I would be very grateful if someone could explain why.

Add. information: the original string is in utf-8, and it doesn’t matter, because after urlencode we get only ASCII characters. It translates from English normally, which is logical, because the characters are not additionally encoded.

I suspect that browsers send some “correct” headers, but is this true, and if so, how to find them out, and most importantly, how to send them manually?

Answer 1

Looks like it depends on google. User-Agent:
You can check it at web-sniffer.net/ For browsers, the response comes in UTF-8, and for 'Web-sniffer' or empty User-Agent:- \u-encoded