Description:
------------
parse_url() returns the same result for these 2 distinct URLs:

http://example.com/foo
http://example.com/foo?

In the first URL, there is no query string, so the 'query' key is not set. This is OK. In the second URL, there is one, an empty one. As such, PHP should set the 'query' key in the result array to an empty string. At the moment, it's unset.

The issue with the current implementation is that libraries that parse and build URLs cannot differentiate these two, and will build a URL that's different from the parsed one.

Another issue is for libraries handling robots.txt files. According to Google (see https://developers.google.com/search/reference/robots_txt), the following rule:

Disallow: /*.php$

... will match '/filename.php', but not '/filename.php?'.

If I'm using parse_url() to extract the path and query string from the full URL, I will not be able to differentiate the two URLs above.

Test script:
---------------
var_export(parse_url('http://example.com/foo?'));

echo PHP_EOL, PHP_EOL;

var_export(parse_url('http://example.com/foo?', PHP_URL_QUERY));

Expected result:
----------------
array (
  'scheme' => 'http',
  'host' => 'example.com',
  'path' => '/foo',
  'query' => '',
)

''

Actual result:
--------------
array (
  'scheme' => 'http',
  'host' => 'example.com',
  'path' => '/foo',
)

NULL