Re: RFC: Character set converters...

[Bryce McKinlay <mckinlay at redhat dot com>]

David Daney wrote:

> > Since then, Bryce has said that he thinks the nio API is too
> > heavyweight.  So, things are a bit undecided.
> Other that that I see no reason why gnu.java.nio.charset.* could not be
> made as efficient as gnu.gcj.convert.*
>  

How would you get data into the converter without either an extra layer 
of copying, or creating a new CharBuffer instance each time (via the 
wrap() method)? Seems to me you'd have to do one of these things for 
every write() call.

That additional overhead might be acceptable if either using the 
converters was simplified significantly, or it meant that a wider range 
of high quality converters became available to us, but I don't think 
either of those are the case. Implementing a new charset in the 
java.nio.charset API is, I think, be significantly more complicated than 
it would in the simpler gcj/classpath APIs.

This is just an argument against converting things like PrintStream to 
use java.nio.charset directly. I certainly don't see a problem with 
creating a gnu.gcj.convert wrapper for java.nio.charset in order to 
handle additional encodings that we don't currently have in the 
gnu.gcj.convert form.

Regards

Bryce