How to remove empty query string from URL

Published:: 2015-05-28

Sometimes, there is an undesired question mark at the end of URL, e. g. /example/? instead of /example/. Question mark separates main part of the address from query string, but some browsers add it even if GET form which the URL is the result of submission of doesn’t contain fields and therefore query string is empty.

This happens, for example, when using link buttons in browsers based on Chromium / Blink (Opera 15+, Vivaldi) and WebKit (Safari) due to a bug (1, 2) they have. Trailing question mark when query string is empty can be automatically removed using server-side redirection.

PHP

A simplest solution is to use server-side scripting language PHP: full requested path is usually stored in server variable REQUEST_URI. If the first question mark in the variable value is its last character, we should redirect to URL with question mark removed.

$uri = $_SERVER['REQUEST_URI']; $qPos = strpos($uri, '?'); if ($qPos === strlen($uri) - 1) { header('HTTP/1.1 301 Moved Permanently'); header('Location: ' . substr($uri, 0, $qPos)); exit; }

This code should be inserted into index file of your website before any other PHP code. If altering original files of your CMS is undesired (e. g. to be able to update it), then you can define your own index file (e. g. custom-index.php), put the code above into it, and then include the true CMS’s index file (typically index.php) via a require_once statement.

Apache / .htaccess

With built-in features of the popular web server Apache, it is somewhat problematic to reliably distinguish cases when query string is empty (there is only question mark at the end of URL), and when there is no query string at all (URL does not even contain question mark).

Looks like the server variable REQUEST_URI in Apache (unlike PHP) does not contain question mark at all if query string is empty.

So we are forced to analize the server variable THE_REQUEST that contains not just the requested URL, but the entire HTTP request like GET uri HTTP/1.1:

RewriteEngine On RewriteCond %{THE_REQUEST} ^[^\s]+\s+[^?]*?\? RewriteCond %{QUERY_STRING} ="" # For any version of Apache: RewriteRule .? %{REQUEST_URI}? [R=301,L] # For Apache 2.4+: # RewriteRule .? %{REQUEST_URI} [R=301,L,QSD]

Request method (e. g. GET) and protocol (HTTP/1.1) do not typically contain question marks, so it is possible to use a simpler (though theoretically somewhat less reliable) way: checking value of the THE_REQUEST variable for question-mark existence in any part of request, not just in URL:

RewriteCond %{THE_REQUEST} \?

Comments

kaavain 2016-03-27

Works great, I use the PHP version because all other .htaccess solutions leads to broken links, missed images etc so I stopped to search for Apache way..
The only question is, I found
http://www.mysite/? links in a google and tried to rewrite them to http://www.mysite/
Using this method I've got

===
HTTP/1.1 301 Moved Permanently
Location: /
===

I guess it is not correct, isn't it? Although it works in browser, it looks strange and unlike Apache redirects I did for other situations… So I changed a little the PHP like this:

===
header('Location: http://'.$_SERVER['HTTP_HOST'].substr($uri, 0, $qPos));
===

And now

====
HTTP/1.1 301 Moved Permanently
Location: http://www.mysite/
====

May be it is better? Unfortunatelly I can't found how to detect http/https dynamically but actually I do not need this..

Anyway, thanks!
Marat Tanalin (author) 2016-03-27

kaavain:

all other .htaccess solutions leads to broken links, missed images etc so I stopped to search for Apache way

Did you try my code for Apache? I’ve tested all the snippets I’ve provided in this article before posting it. Still, I prefer PHP too since it’s much more flexible and doesn’t involve slow regular expressions for all URL-related operations, while using web server just for creating a single point of entry to the web application.

Although it works in browser, it looks strange and unlike Apache redirects I did for other situations

Both URL forms (a full absolute URL including protocol and host like http://example.com/foobar/, and a root-relative URL like /foobar/) are valid and work in modern browsers. You are free to use either. To be fair, in theory, some very old browsers could not support root-relative URLs, but I believe such old browsers don’t matter today.

I can't found how to detect http/https dynamically

When the page is opened via HTTPS, the HTTPS server variable ($_SERVER['HTTPS'] in PHP) has the on value.
kaavain 2016-03-29

=====
Did you try my code for Apache? I’ve tested all the snippets I’ve provided in this article before posting it. Still, I prefer PHP too since it’s much more flexible and doesn’t involve slow regular expressions for all URL-related operations, while using web server just for creating a single point of entry to the web application.
=====
Yes, I've tried a lot of ways, including yours. I do not clearly remember what happened in that approach, but it does not mater actually. May be the reason is that my CMS is old-fashioned and uses ?parameters instead of "normal" URLs…

=====
To be fair, in theory, some very old browsers could not support root-relative URLs, but I believe such old browsers don’t matter today.
=====
This is important as I'm trying to hold site functionality on all browsers and it works even on IE6

=====
When the page is opened via HTTPS, the HTTPS server variable ($_SERVER['HTTPS'] in PHP) has the on value.
=====
Got it, thanks. I'm not planning to use HTTPS at all, but I'm noticed.
Marat Tanalin (author) 2016-03-29

kaavain, IE6 does support root-relative URLs in the Location header. By saying very old browsers, I meant something absolutely irrelevant like IE3. (Though, actually, IE6 in particular does not matter today as well, moreso for such noncritical things as removing empty query string.)

For what it’s worth, there is a russian version of this blog post.
sS 2017-09-11

This is not working if the URL ends with two query strings (??)
combo 2017-11-05

OMG! This tutorial literally saved my life! Thanks a lot and all the best wishes from a webmaster in Germany that had to struggle with hundreds of urls with 'added whatever after the question mark'.

THANK YOU!