Name: rmT116609 Date: 06/17/2003


A DESCRIPTION OF THE REQUEST :
I've seen that there have been several requests for unsigned integral types in Java, and you have generally blown these off in a fairly cavalier manner. Usually you report that these types are unnecessary and can be worked around. In some cases you are correct, and I can write code that chews up CPU time masking and shifting bits around, using the char type as an unsigned temporary bucket, etc. However, I'm dealing with a stream of bytes that contains numeric values. I have to grab a few bytes at a time and make a number out of them - an integer, a long, whatever. Obviously, some of those bytes may have the high-order bit set, not because they are negative numbers but because that's how they fit in that particular byte in the sequence. Java's insistence on treating each byte as a signed value, and therefore doing sign-extension for me, is a major pain.
What this all boils down to is this: as a language vendor you should not be continually telling your user community that they are wrong when they request support from the language. Support the unsigned types because it is a reasonable thing to do, and stop pontificating. This is a major oversight in the language, and quite frankly, I'm amazed that you think it's okay to waste CPU run time rather than CPU compile time!

EXPECTED VERSUS ACTUAL BEHAVIOR :
EXPECTED -
Support the unsigned integral data types.
ACTUAL -
See above. Sometimes requires significant work-arounds.

CUSTOMER SUBMITTED WORKAROUND :
Bit shifting, masking.
(Review ID: 187835)
======================================================================