Interestingly, I didn't get any responses to my original mail on the topic. Since we need UTF-8 support and because part of our application is in Struts (which doesn't have a problem with UTF-8 encoded property files) we'd like to avoid encoding the property files in Latin-1. I made a quick hack for supporting UTF-8 encoded property files as outlined here (http://www.thoughtsabout.net/blog/archives/000044.html), added the class and made a few changes to take it into use in our build of MyFaces. Another benefit of the wrapper is that you could easily implement caching � la Struts MessageResources. I tested it and it seems to work fine. I could easily create a patch for it; Does anybody else in MyFaces community think it'd be useful to add support for UTF-8 encoded Property files?
Kalle > -----Original Message----- > From: Korhonen, Kalle > Sent: Wednesday, April 20, 2005 11:12 AM > To: 'MyFaces Discussion' > Subject: UTF-8 support > > Looks like the default MyFaces implementation uses the > standard PropertyResourceBundle, which uses the standard > Properties class that in turn assumes only Latin-1 as > encoding of the property file. That naturally breaks all the > UTF-8 encoded double-byte characters in the properties file. > I definitely don't want to use native2ascii, which is just > too much hassle, but just simply encode my Properties files as UTF-8. > > I'm surprised that I seem to be the first person hitting this > problem. I search the mail archives for "utf*" but got > nothing on this. I assume and recall reading from some place > that you could replace the default ResourceBundle > implementation, similarly to overriding other configuration > in JSF's pluggable architecture. Anybody done this and care > to share any samples? > > Struts implements its own resource classes that deal with > UTF-8 encoded files properly. In fact, those classes are not > even inherited or using any of the ResourceBundle/Properties > base classes. Anybody done an adaptation of that code as a > ResourceBundle? A few months ago, I looked around quickly but > couldn't find any easily available, does anyone know if there are any? > > Also, looks like HtmlResponseWriterImpl is encoding all > characters out of latin-1 as unicode character references. > For performance reasons, I don't think we should do that if > the response stream specifies UTF-8 as encoding. There might > be other complications that need to be considered, but I > might file a bug on it later and patch it once I get the > UTF-8 properties file working. > > (A little off-topic) Finally, I could also consider 1.5 Java > Properties.loadFromXml(), but I'm a little hesitant to use it > because of possible performance penalties and because I don't > really need xml format. Don't really understand why Sun > wouldn't just add support for UTF-8 for Properties.load() and > store() since it shouldn't break anything. Anyway, anybody > have any experience on loadFromXml and an opinion on pros and > cons of it? The technical writers have barely wrapped their > heads around the concept of parametrized resources, so I'm > afraid introducing xml into the soup is just more confusing. > > Kalle
