New subject: [xwiki-dev] Re: XWiki: au sujet d'un ancien patch i18n...

27 Apr 2007

Hi Gilles,

On Apr 27, 2007, at 10:41 AM, Gilles Serasset wrote:

...
  Hi Vincent, Hi all,

 mmm, sorry, I just tried to figfure out the problems from the doc.  
 There is a problem already here:

 First, the getXXX method do not return the same kind of info:

 getServletPath() - decoded
 getPathInfo() - decoded
 getRequestURI() - remain encoded
 getContextPath() - remain encoded
 getPathTranslated() - remain decoded

 You are using getServletPath which returns a decoded string when  
 getRequestURI gives you a encoded string. Hence, if a non ascii  
 char is used in the ServletPath, the length of the servletPath will  
 be shorter than its encoded conterpart. 
Yes this is a limitation and I assumed (probably wrongly) that  
servlet path wouldn't contain non ascii chars. If we have a better  
solution I'm all for it...

...
  Second, you don't have to fix the encoding anymore
as it is not  
 (incorrectly) decoded by the container. However, you have to decode  
 it correctly using the java.net.URLDecoder class (and passing it  
 you wiki encoding). 
I don't understand what you're saying here. All I can say is that the  
existing code wasn't working at all and it now works in all cases  
I've tried except when the user enters non ascii chars directly in  
the URL. Creating pages/links using XWiki works fine.

We discussed this with Sergiu and we'd like to modify xwiki so that:
- we assume that all URLs we get are URL encodings of UTF8 chars always
- when we generate URLs we UTF8 encode them (and then URL encode them)

...
  Third, in the case where you call getPathInfo, you
should put:

 path = fixDecodedURI(request, request.getPathInfo()); 
This is what you had before but it wasn't working... That's what I  
fixed :)

...
  Sorry, I'm teaching all day, so can't tell you
more now... 
Thanks
-Vincent

...
  Regards, Gilles,
 On 25 avr. 07, at 16:01, Vincent Massol wrote:

  Hi Gilles (and everyone interested),

 I have modified today the XWiki.getDocumentName() method. I've  
 seen that you had modified it in the past to add the call to  
 fixDecodedURI(). I think it's no longer required but I'd like to  
 have your opinion on that.

 Here's the modified version:

 public String getDocumentName(XWikiRequest request, XWikiContext  
 context)
 {
 [...]
         else {
             // TODO: Introduce a XWikiURL class in charge of  
 getting the information relevant
             // to XWiki from a request URL (action, space,  
 document name, file, etc)

             // Important: We cannot use getPathInfo() as the  
 container encodes it and different
             // containers encode it differently, depending on  
 their internal behavior and how
             // they are configured. Thus to make this container- 
 proof we use the
             // getRequestURI() which isn't modified by the  
 container and is thus only
             // URL-encoded.

             // Note: Ideally we should modify the  
 getDocumentNameFromPath method but in order
             // not to introduce any new bug right now we're  
 reconstructing a path info that we
             // pass to it using the following algorithm:
             //    path info = requestURI - (contextPath +  
 servletPath)
             String path;
             if (!request.getRequestURI().startsWith(
                 request.getContextPath() + request.getServletPath()))
             {
                 log.warn("Request URI [" + request.getRequestURI()  
 + "] should have matched "
                     + "context path [" + request.getContextPath()  
 + "] and servlet path ["
                     + request.getServletPath() + "]");
                 // Even though this branch shouldn't get executed  
 we never know what containers
                 // will return and thus in order to be safe we  
 fall back to the previous
                 // behavior which was to use getPathInfo() for  
 getting the path (with the
                 // potential issue with i18n encoding as stated  
 above).
                 path = request.getPathInfo();
             } else {
                 path = request.getRequestURI().substring(
                     request.getContextPath().length() +  
 request.getServletPath().length());
             }

             // TODO: I think it's safe to remove the fixDecodedURI  
 now that we're not using
             // getPathInfo() and thus the container doesn't encode  
 anything but I don't want
             // to break anything at this late stage (we're  
 releasing 1.0 RC2). Remove it when
             // we start working on 1.1.
             docname =
                 getDocumentNameFromPath(fixDecodedURI(request,  
 path), context);
         }
     }
     return docname;
 }

 Thanks a lot
 -Vincent 
 --
 Gilles Sérasset
 GETALP-LIG                         BP 53 - F-38041 Grenoble Cedex 9
 Phone: +33 4 76 51 43 80                   Fax:   +33 4 76 44 66 75