Rewrite of IBM doublebyte charsets

Ulf Zibis Ulf.Zibis at
Tue May 19 10:02:51 UTC 2009

Am 19.05.2009 09:19, Xueming Shen schrieb:
> I will try the b2c[0x100][] idea to see the
> performance...I actually tried b2cIndex[0x100] +  b2c[], which did not 
> bring me the performance
> I expected, but that was a long time ago, so this time might be 
> different:-)

Even if there is no performance gain, the code will become more 
readable, and initC2B becomes more simple.
Additionally, when doing research on the generated source files (e.g. 
visual diff between different charsets), interpretation of the data 
becomes much more simple, as there would be exactly 1 String element for 
1 mapping segment, always starting at it's b2Min value as first left 
value of each chunk.
With that help, if would be easy to discover similar mapping for later 
tuning regarding sharing maps partly. :-)

Don't forget to try
    char c = b2cStr[byte1].charAt(byte2);
, using
    String[] b2cStr = new String[0x100];
, to avoid copying String[] to char[][].


More information about the core-libs-dev mailing list