Rails - Aug 2005 - Routing recognition problem: filenames containing '+'

All,

I recently upgraded to Rails 0.13.1 (under Windows XP, running webrick ) and I
am experiencing a problem.  I have a series of files in a directory called
$RAILS/public/files.  I can normally get at the file (called file.txt) by
pointing my browser to http://localhost:3000/files/file.txt.  If the file has a
plus sign (''+'') in the name (e.g. fil+e.txt), I get the
following error pasted
below.  Any thoughts on what''s going on here?  I tried using the HTML
code for
''+'' to no avail.

-cal


ActionController::RoutingError (Recognition failed for
"/files/fil+e.txt"):
   
C:/ruby/lib/ruby/gems/1.8/gems/actionpack-1.9.1/lib/action_controller/ \
routing.rb:444:in
`recognition_failed''
   
C:/ruby/lib/ruby/gems/1.8/gems/actionpack-1.9.1/lib/action_controller/ \
routing.rb:434:in
`recognize!''
    C:/ruby/lib/ruby/gems/1.8/gems/rails-0.13.1/lib/dispatcher.rb:32:in \
 `dispatch''
    C:/ruby/lib/ruby/gems/1.8/gems/rails-0.13.1/lib/webrick_server.rb:105: \
in
`handle_dispatch''
    C:/ruby/lib/ruby/gems/1.8/gems/rails-0.13.1/lib/webrick_server.rb:71: \
in
`service''
    C:/ruby/lib/ruby/1.8/webrick/httpserver.rb:104:in `service''
    C:/ruby/lib/ruby/1.8/webrick/httpserver.rb:65:in `run''
    C:/ruby/lib/ruby/1.8/webrick/server.rb:155:in `start_thread''
    C:/ruby/lib/ruby/1.8/webrick/server.rb:144:in `start''
    C:/ruby/lib/ruby/1.8/webrick/server.rb:144:in `start_thread''
    C:/ruby/lib/ruby/1.8/webrick/server.rb:94:in `start''
    C:/ruby/lib/ruby/1.8/webrick/server.rb:89:in `each''
    C:/ruby/lib/ruby/1.8/webrick/server.rb:89:in `start''
    C:/ruby/lib/ruby/1.8/webrick/server.rb:79:in `start''
    C:/ruby/lib/ruby/1.8/webrick/server.rb:79:in `start''
    C:/ruby/lib/ruby/gems/1.8/gems/rails-0.13.1/lib/webrick_server.rb:57: \
in
`dispatch''
    $RAILS/app/script/server:49

cal wrote:
> I recently upgraded to Rails 0.13.1 (under Windows XP, running webrick )
and I
> am experiencing a problem.  I have a series of files in a directory called
> $RAILS/public/files.  I can normally get at the file (called file.txt) by
> pointing my browser to http://localhost:3000/files/file.txt.  If the file
has a
> plus sign (''+'') in the name (e.g. fil+e.txt), I get the
following error pasted
> below.  Any thoughts on what''s going on here?  I tried using the
HTML code for
> ''+'' to no avail.
The HTML code?  Do you mean the HTTP escaped version of ''+''
which is, I
think, ''%2b''?  So, ''fil%2be.txt''.

Matthew Denner <matt@...> writes:
> 
> cal wrote:
> > I recently upgraded to Rails 0.13.1 (under Windows XP, running webrick
)
and I
> > am experiencing a problem.  I have a series of files in a directory
called
> > $RAILS/public/files.  I can normally get at the file (called file.txt)
by
> > pointing my browser to http://localhost:3000/files/file.txt.  If the
file
has a
> > plus sign (''+'') in the name (e.g. fil+e.txt), I get
the following error
pasted
> > below.  Any thoughts on what''s going on here?  I tried using
the HTML code
for
> > ''+'' to no avail.
> 
> The HTML code?  Do you mean the HTTP escaped version of
''+'' which is, I
> think, ''%2b''?  So, ''fil%2be.txt''.
> 
Yes.  I tried, as you indicate, ''fil%2be.txt''.  I received the
same
routing recognition message.  Perhaps there''s a way in my
$RAILS/config/routes.rb to let requests for files in
$RAILS/public/files to get passed as-is?

-cal

A followup:

I added an entry to my $RAILS/config/route.rb file which looks like this:

map.connect ''/files/:file'', :controller =>
''file'', :action => ''get''

I also added the appropriate controller and action, which simply performs
a render_text(@params).  Interestingly, the plus sign has been removed from
the parameter list.  This is the output:

When the URL is http://localhost:3000/files/test.txt, I see this:
{"action"=>"get",
"controller"=>"file",
"file"=>"test.txt"}

When the URL is http://localhost:3000/files/tes+t.txt, I see this:
{"action"=>"get",
"controller"=>"file", "file"=>"tes
t.txt"}

When I use the HTTP escaped character like this:
http://localhost:3000/files/tes%2Bt.txt

I get this promising result:
{"action"=>"get",
"controller"=>"file",
"file"=>"tes+t.txt"}

But removal of the line from routes.rb results in the same recognition error.

Any ideas from the RoR-ers out there?

-cal

> When the URL is http://localhost:3000/files/test.txt, I see this:
> {"action"=>"get",
"controller"=>"file",
"file"=>"test.txt"}
> 
> When the URL is http://localhost:3000/files/tes+t.txt, I see this:
> {"action"=>"get",
"controller"=>"file", "file"=>"tes
t.txt"}
> 
> When I use the HTTP escaped character like this:
> http://localhost:3000/files/tes%2Bt.txt
> 
> I get this promising result:
> {"action"=>"get",
"controller"=>"file",
"file"=>"tes+t.txt"}
> 
> But removal of the line from routes.rb results in the same recognition
error.
> 
> Any ideas from the RoR-ers out there?
> 
> -cal
Cal, these are just semi-educated guesses, but here''s some things to
ponder-- 
-- As you may know, in an http GET querystring the "+" is the
<space>
character (eg-- /show?1234+Main+Street)

-- mod_rewrite (used with fcgi) makes extensive use of regular
expressions, for which the "+" sign also has meaning (match one or
more
times).

Between these two points, I''d say your ''best'' bet is
to stear clear of
the "+" symbol in file names that are to be resources requested via a
URL, but you''ve probably already considered that and have good reason
for naming your files as such?

That being the case, I _suspect_ (remember that semi-educated disclaimer
above?) that the solution is to escape the + sign in the RegEx for your
route. But then, I stink at writing regular expressions :)

HTH,

Howard

The issue you are seeing is due to the differences in handling the + 
character. AFAIK, the standard defines that ''+'' =>
'' '' only in the query
part of the URL. However, Routing will produce strings with
''+'' inside
the path part of the URL. According to standards, this is incorrect. 
However, most webservers support(ed?) this usage.

Recently however, lighttpd changed to the standard use of +. Due to this 
change, building links to files inside public/  can be somewhat 
difficult. It is necessary for you to find out if your webserver will 
escape a ''+'' to a '' '' when it looks for a
file.

A while ago, I noticed that webrick did not behave as lighttpd did, so I 
patched webrick. Since then lighttpd changed their policy... :S I have 
not yet reverted the patch to webrick, and in reality I would like to 
see a way to have lighttpd behave as it did before.

HTH