(j3.2006) (SC22WG5.4210) [ukfortran] RE: 43 Fortran compilers

[N.M. Maclaren]

On Mar 1 2010, Bill Long wrote:
>
>Indeed, M77.  I looked at the manual and found that, in 1980, M77 had 
>the radical extension of A .op. B where op was and, or, xor, ... and A 
>and B were numeric type variables, with the operations bitwise.  Only 30 
>years ago.  Maybe this idea needs a bit more time to mature. :)

Or rot down?

It was a major nightmare - and I don't mean just for portability.
Even with a single compiler for a single architecture, bit-flipping
on REAL and the sign bit of integers is beyond the ability of almost
all programmers.  If I recall, those machines weren't twos complement,
either, but I didn't use them directly.

Restricting it to non-negative integers has worked in many languages,
but with the ongoing problem of priority.  A == B.OR.C has been a major
gotcha in every language that I know of that has allowed it.  Even using
two operators for the two priorities (and, to some extent, functionality)
helps only a little.

And, yes, I do mean that porting CDC Fortran wasn't JUST a matter of
changing its syntax - it was amazing how often one discovered that the
original programmer had made an error with bit twiddling due to the
above gotchas.

Regards,
Nick Maclaren.