[Prev][Next][Index][Thread]

Re: notation for extending precedence



At 4:34 PM 9/16/96, T. V. Raman wrote:
>Eventually, if we did feel the need to allow author defined precedence
>enhancements this can be done cleanly by providing definition syntax of the
>form
>(define-operator :operator-name 'foo :precedence :same-as "+")
>;excuse the lisp:-)
>Instead of :same-as the author would also be able to say :less-than and
>:greater-than (I implemented the above in Aster and it worked reasonably well
>for a sufficiently large collection of hairy math. 


Do you have a suggested way of specifying the original (built-in)
list of operators (maybe 250 of them) with their associativities
and precedences (maybe 80 precedence levels, since the operators
come in groups of the same precedence), without each operator
having to give the name of the previous one, and in which a few
operators (though not most) have different left and right precedences,
with the out-of-order ones specified by reference to other operators?

If so, we could dispense with numerical precedence levels
even for the built-in list of operators.



Follow-Ups: