[Info-vax] New VSI Roadmap (yipee!)

[David Froble]

John Reagan wrote:
> On Tuesday, March 3, 2015 at 10:07:10 AM UTC-5, David Froble wrote:
> 
>> As for those worrying about absolute precision with currency, when the 
>> CEO of a distribution company wants to see the monthly bookings and 
>> billings, he may want to see it rounded to thousands, and most 
>> definitely not pennys.
> 
> But I suspect the accountants or the class-action lawyers will care.
> 
>> All I did was two things, first mentioned that we have been very 
>> successful in using FP for currency, and second asked for a rounding 
>> function as part of a language.
>>
> 
> Since you've done this for a long time, you've went from VAX to Alpha
> in the that timeframe.  Did the switch from D to G (which drops 3
> bits of mantissa) make a difference?

Actually, this began back on RSTS/E and Basic Plus.  Long time ago.

The switch from D to G was interesting.  While we use D-Float, I'm aware 
that calculations are done in G-Float.  I've seen some minor problems. 
But not in the currency calculations.  With currency, after EVERY 
calculation, a round is performed, to inhibit the drift.  If you didn't 
dod this, there would be drift, and sometimes you'd be off a penny. 
Now, the accounting types would question this, but for other things it 
just wasn't an issue, and a non-issue once the practice of rounding 
after every calculation was implemented.

But yeah, FP can be fun.

> If you want to count pennies, use X_float.  You get 33 digits of
> pennies.  Using X_float to count integral things is fine.  You have
> to provide your own scaling of course.

For companies with say, $100 million a year in sales, you'd never have a 
problem with the 15 digits of precision in D-Float.

> Back to the rounding function, I can see why YOU want such a thing
> given the routine call overhead in BASIC, but lets see if we can fix
> the routine call overhead issue instead of adding builtin after
> builtin after builtin.  Maybe some keyword on the subroutine or such?

I cannot remember the details, but back in the dim past Randy Park 
developer a report writer utility he called XENTIS, and he ran into the 
overhead of the subroutine calls.  He found some compiler switch, which 
i cannot remember nor find, that bypassed some of the set-up and 
break-down of calling subroutines.  I seem to recall that it had to do 
with error trapping.

But yes, what you suggest would be a good thing.  Modular programming 
techniques lead to calling subroutines and external functions, and 
improving that would be helpful.

How do other languages handle such?