Re: A few tips

[Per Bothner <per at bothner dot com>]

Dominique Boucher wrote:

> I run Kawa servlets with Tomcat 4.1 on Linux, using Sun’s JDK 2
> v1.4.1_01.

"Linux" means a lot of different things.  My impression is that many 
distributions (including Red Hat for sure) are moving towards using 
UTF-8 as the standard/default encoding.  If your files are instead 
ISO-8859-1, you'll probably have problems.

> Some of the configuration files contain
> Scheme strings with diacritics (French accents, for instance).

To be pedantic, the files only "contain" diacritics if interpreted using 
the correct encoding.  Specifically, the files are encoded in 
ISO-Latin-1, but your software environment thinks they use some other 
encoding, probably UTF-8.

This page http://fedora.redhat.com/docs/release-notes/ has some notes on 
encoding.  Red Hat believes "In the long term, all systems are expected 
to migrate to UTF-8, eliminating this issue."

My guess is you might need to change your LANG environment variable, 
unless you're willing to migrate to UTF-8.

> 2. For the constant Scheme strings in the source code, you must make
> sure that Kawa reads them properly when compiling. So add the mutation
> to 'port-char-encoding' on the command-line:
>
> shell> kawa -e '(set! port-char-encoding "ISO-8859-1")' -C
> sourcefile.scm

The same presumably also applies for files loaded with the -f flag.
And presumably for expressions type at the Kawa console.  I don't think 
setting port-char-encoding will help with the latter, since the the 
standard input has already been opened.  For that you need to set LANG.

> 3. Make sure the strings are not put in 'unescaped-data'. 

I assume the "not" was unintended.

> This way, all the special characters (those with French diacritics) will
> be translated to their equivalent numerical entities (é for é)
> properly.

Whether é is written as é or as &#233 should depend on the encoding used 
for the underlying PrintWriter.  Unfortunately, I don't know of any 
reliable way to get that.  However, we can use port-char-encoding.  We 
can also use OutputStreamWriter's getEncoding.

Does anyone know how one finds out in Java what the default (system) 
encoding is?

> [Note: these problems may be due to a special configuration of the C
> locale, but I can't modify it easily for Tomcat.

It's not a problem with the C local per se.  However, it may be a 
problem that Tomcat is *using* the C locale - or a UTF-8 locale.  I bet 
with the correct environment flags (LANG? LOCALE?) you could fix that.
--
	--Per Bothner
per@bothner.com   http://per.bothner.com/