Drupal on lighttpd with clean urls (updated)

Posted: February 6th, by darix

Tags: clean url, drupal, lighttpd, mod_magnet

Many people have asked for an example for mod_magnet that doesnt involve Ruby on Rails.

I have choosen drupal. Drupal is a nice little content managment platform written in php. I used it before switching to rails. :) Though lets start.

As already mentioned in the “Dr. Magneto vs Mr 404 handler” article, drupal uses a small nice rewrite rule set:

RewriteCond %{REQUEST_FILENAME} !-f
RewriteCond %{REQUEST_FILENAME} !-d
RewriteRule ^(.*)$ index.php?q=$1 [L,QSA]

Ok. Lighttpd doesnt support -d/-f. So let’s use mod_magnet. For the example we assume drupal is already installed and accessible under “http://www.example.org/drupal/”. The needed drupal.lua can be downloaded here. The script is nearly as simple as the rails example. So is the integration with lighttpd. The configsnippet is:

$HTTP["url"] =~ "^/drupal" {
    # we only need index.php here.
    index-file.names = ( "index.php" )
    # for clean urls
    magnet.attract-physical-path-to = ( "/etc/lighttpd/drupal.lua" )
}

Last but not least you need to configure drupal to use clean urls. You go to “Admin” => “Site configuration” => “Clean urls” (quick link http://www.example.org/drupal/index.php?q=admin/settings/clean-urls ) and click on the “Run the clean URL test” link. If it was successful you can enable clean urls and save the settings.

“But my drupal runs under ‘/someotherpath’, what do i need to change?”

Of course the url conditional in the config snippet above. You also need to change the prefix in the drupal.lua. Do not append a trailing slash there.

Alternatives?

404 handler?

Of course it would work as well, but the 404 handler breaks the 403 return code, which is used by drupal.

url.rewrite-once?

This is a possible solution aswell. But you need to rewrite all static files to itself.

url.rewrite-once = (
    "^(/drupal/files.*)"       => "$1",
    "^(/drupal/themes.*)"      => "$1",
    "^/drupal/([^.?]*)\?(.*)$" => "/drupal/index.php?q=$1&$2",
    "^/drupal/([^.?]*)$"       => "/drupal/index.php?q=$1",
)

Killes mentioned another rewrite solution on #drupal on freenode:

url.rewrite-once = (
    "^/drupal/system/test/(.*)$" => "/drupal/index.php?q=system/test/$1",
    "^/drupal/([^.?]*)\?(.*)$"   => "/drupal/index.php?q=$1&$2",
    "^/drupal/([^.?]*)$"         => "/drupal/index.php?q=$1",
    "^/drupal/search/node/(.*)$" => "/drupal/index.php?q=search/node/$1" 
)

Security

A small side note, in SA-2006-006 the drupal team warned about a security hole through carefully named arbitrary scripts in the files directory. The apache workaround of course doesnt work here either. The following block should secure your server with lighttpd. :)

# "Fix" for Drupal SA-2006-006, requires lighttpd 1.4.13
$HTTP["url"] =~ "^/drupal/files" {
    cgi.assign = ()
    fastcgi.server = ()
    scgi.server = ()
}

For versions before 1.4.13 you need to move your cgi/fastcgi/scgi defintions into a conditional that inverts the match as shown in the example above. But i have to admit it complicates the config quite a bit. ;)

Update: the script was broken for prefices containing lua patterns special chars. This has been fixed and a new version was uploaded.

Harold

What if Drupal runs under the root of a (sub) domain?

Like: http://example.com

darix

$HTTP["host"] == "drupal.example.org" {
    # we only need index.php here.
    index-file.names = ( "index.php" )
    # for clean urls
    magnet.attract-physical-path-to = ( "/etc/lighttpd/drupal.lua" )
}

and set the prefix in drupal.lua to ”/”

fago

really interesting solution. Is there a way to set the lua variable through a lighttpd config variable? This would be really useful for automatic generated configs where the the drupal url can change…

X-Istence

What if you have multiple apps that have their own re-writing scheme? For example, on a website that I run we have two scripts that need to be rewritten. Drupal is in /, and the other script is /drupal-devel/. However when I go to http://example.com/drupal-devel/node/1 it goes to http://example.com/node/1. Which off course is not what I want.

fago

hm, perhaps it would be possible to set an evironment variable, that the lua script can read? that would help a lot and would be great enhancement for the combo drupal + lighty!

MrWerewolf

What if you want to integrate Gallery with Drupal? The following clean url solution is for the Drupal Gallery module. Replace the main loop code in drupal.lua with the following:

-- the magic ;) if (not file_exists(lighty.env["physical.path"])) then -- file still missing. pass it to the fastcgi backend request_uri = removePrefix(lighty.env["uri.path"], prefix) -- rules for drupal if request_uri then local uriquery = lighty.env["uri.query"] or "" if string.find(request_uri, "v/") == 1 then lighty.env["uri.path"] = prefix .. "/index.php" if #request_uri > 2 then g2path = string.sub(requesturi, 3) else g2_path = "" end lighty.env["uri.query"] = uriquery .. (uriquery ~= "" and "&" or "") .. "q=gallery&g2path=" .. g2path else lighty.env["uri.path"] = prefix .. "/index.php" lighty.env["uri.query"] = uriquery .. (uriquery ~= "" and "&" or "") .. "q=" .. request_uri end lighty.env["physical.rel-path"] = lighty.env["uri.path"] lighty.env["request.orig-uri"] = lighty.env["request.uri"] lighty.env["physical.path"] = lighty.env["physical.doc-root"] .. lighty.env["physical.rel-path"] end end -- fallthrough will put it back into the lighty request loop -- that means we get the 304 handling for free. ;)

I use the default modrewrite rules for accessing the non-embedded Gallery. You can find them here: http://codex.gallery2.org/Gallery2:LighttpdURL_Rewrite

thanks

finally your solution worked, thank you!

andrew

darix’s comment totally put me on the wrong track. i banged my head against this for quite a while before realizing that i needed: local prefix = ’’ now it works like a charm.

MrWerewolf

Sorry for the above comment being ugly, this should be prettier.

-- the magic ;)
if (not file_exists(lighty.env["physical.path"])) then
    -- file still missing. pass it to the fastcgi backend
    request_uri = removePrefix(lighty.env["uri.path"], prefix)
    -- rules for drupal
    if request_uri then
      local uriquery = lighty.env["uri.query"] or "" 
      if string.find(request_uri, "v/") == 1 then
        lighty.env["uri.path"]          = prefix .. "/index.php" 
        if #request_uri &gt; 2 then
          g2&lt;em&gt;path = string.sub(request&lt;/em&gt;uri, 3)
        else
          g2_path = "" 
        end
        lighty.env["uri.query"] = uriquery .. (uriquery ~= "" and "&amp;amp;" or "") .. "q=gallery&amp;amp;g2&lt;em&gt;path=" .. g2&lt;/em&gt;path
      else
        lighty.env["uri.path"]          = prefix .. "/index.php" 
        lighty.env["uri.query"] = uriquery .. (uriquery ~= "" and "&amp;amp;" or "") .. "q=" .. request_uri
      end
      lighty.env["physical.rel-path"] = lighty.env["uri.path"]
      lighty.env["request.orig-uri"]  = lighty.env["request.uri"]
      lighty.env["physical.path"]     = lighty.env["physical.doc-root"] .. lighty.env["physical.rel-path"]
    end
end
-- fallthrough will put it back into the lighty request loop
-- that means we get the 304 handling for free. ;)

guidot

The imagecache_module did not work when I used the above mentioned securityfix (for Drupal SA-2006-006).I don’t know how the fix can be done in another way.

waddles

MrWerewolf: it looks nicer, but one would need to convert your html elements: if #request_uri > 2 then g2path = string.sub(requesturi, 3) etc

darix

guidot: does it work with apache using the latest security config?

pebosi

Thanks, now i can try using lighty.