I got all the bitwise operators working using the new ShortcutExpr model which maps to methods. Plus I reworked JCodeAsm to generate the correct bytecode optimizations for those cases. Soon as fix some of the auto-cast stuff, I should be able to move most of the BinaryExpr operators to ShortcutExpr.

One thing that really bothers me is that C, Java, and C# all have this weird precedence for bitwise operators which is lower than equality or relational operators (I think it is a holdover from C days where non-zero was true)...

http://java.sun.com/docs/books/tutorial/java/nutsandbolts/expressions.html http://msdn.microsoft.com/library/default.asp?url=/library/en-us/csspec/html/vclrfcsharpspec_7_2_1.asp http://www.zenspider.com/Languages/Ruby/QuickRef.html#21

For example I hate having to write this in Java:

(flags & mask) != 0

I think of bitwise and, or, and xor like math operators:

flags & mask != 0

This is what I am thinking for Fan operator precedence (basically Ruby precedence):

++ -- ! ~ - (T)x
* / %
+ -
<< >>
&
| ^
>  >=  <  <= is as
<=> == === !=
&&
||
= x=

It is the same as C/Java/C# except that we move bitwise & | ^ up above relational operators and make | and ^ the same precedence.

Do you guys agree? Doesn't seem like a change that would really bother or trip up hard-code Java or C# programmers.

I understand what <=> does, but what is the point of having it? I don't think it's a necessary operator and it doesn't seem like something anyone would ever use. Especially since we have the full set of other comparison operators.

I guess it really isn't needed. But it is trivial to implement, and seems like it should be present since it the "master operator" for all comparisions.

I vote to keep it, although if you and Andy are against it, I wouldn't be upset about axing it.