[kaffe] Characters and Bytes conversion:

Ito Kazumitsu kaz at maczuka.gcd.org
Sun Jan 4 02:12:02 PST 2004


>>>>> ":" == Kaffe CVS <cvs-commits at kaffe.org> writes:

:> 2004-01-03  Ito Kazumitsu  <kaz at maczuka.gcd.org>
:> 
:>         * libraries/javalib/kaffe/io/ConverterAlias.java:
:>         Changed "EUC_JP" to "EUC-JP" so that iconv is directly called.
:> 
:>         * libraries/javalib/kaffe/io/ByteToCharEUC_JP.java,
:>         libraries/javalib/kaffe/io/CharToByteEUC_JP.java:
:>         Removed.

This change depends on the change I made last year:

| 2003-12-30  Ito Kazumitsu  <kaz at maczuka.gcd.org>
| 
|         * libraries/javalib/kaffe/io/ByteToCharConverter.java
|         (getConverterInternal),
|         libraries/javalib/kaffe/io/CharToByteConverter.java
|         (getConverterInternal):
|         For an encoding name that cannot be a class name, directly
|         return an instance of the iconv based converter.

These makes me think two things:

(1) For an encoding name XXX that can be a Java class name,
    CharToByteXXX or ByteToCharXXX is tried first and if
    there is no such class, then iconv is tried.

    Isn't it better to have a way to directly call iconv
    without trying CharToByteXXX or ByteToCharXXX?

(2) We are merging GNU Classpath's classes into kaffe.  If
    kaffe's java.io.InputStreamReader and the like are to
    be replaced by GNU Classpath's, a good characters and
    bytes conversion mechanism must be implemented in GNU
    Classpath.




More information about the kaffe mailing list