I was casting about for 10mm high intensity LEDs and found a site that sells them (said LEDs are no longer stock items with many electronics vendors). So I added them to my shopping cart. See below screen capture. Looks as though the site's math routines need a little rounding feature added.

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

Actually, it could be the same scheme used in Superman 3. That fraction of a penny is being routed to some evil mastermind's bank account.

_________________
Cat; the other white meat.

Too funny.

Base-2 floating point numbers are pretty terrible for decimal financial calculations. After your computation you usually have some fiddly oddball number like 234.99999995 due to truncation errors. While approximately correct it needs to be massaged to eliminate errors. So you add back machine epsilon (e.g. 2^−24) and truncate at the last digit to eliminate those fiddly odd ball digits. So truncate(234.99999995 + 2^−24) would yield 235.0000000.

I seem to recall some programming languages having builtin functions to handle this in a machine independent manner.

Can you write them a check for payment??

_________________
Regards, KM
https://github.com/floobydust

Best way to round off to an integer is to add 0.5 and then truncate - no need to fiddle with very small numbers, or guess at the error.

To round to the nearest cent, multiply by 100 and then round off as above. Then shift the point by two places - don't divide!

As for the odd bill, best to overpay, or you'd be in debt for a fraction of a cent and might be pursued to the ends of the earth.

In finance they often keep track of interest rates in basis points which requires four digits of precision, plus a fifth digit to account for truncation errors. Various other financial instruments are also quoted with four decimal places of precision as well.

It makes a lot of sense, sometimes, to work with scaled integers.

I recall a friend of mine had some problem at work reconciling some fraction-of-a-yen discrepancy to do with interest paid - so, not just fractions, but exponentials too. AIUI, one reason the HP15C calculator is so popular in finance is that its results are taken as gospel - the HP15C's algorithms are the definition of what to calculate and how to round.

It's a mistake to think that what's going on in calculation is arithmetic. It's related, but calculations can behave very differently to the way numbers behave.

This is where BCD numbers come in handy.

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

The check's in the mail!

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

Pretty sure you can pay that kind of amount easily with Bitcoin...

Every time I hear about Bitcoin, I want to kick myself...when I think back to the first time I ever heard about Bitcoin and it was something crazy like hundreds of Bitcoins for a dollar or something like that. If I had just invested a few hundred at the time....ugh.

_________________
Cat; the other white meat.

But you would surely have sold at $10, if not at $3... there are infinitely many bets we didn't take which we now know would have been winners...

Looks to me like it's using the CBM Basic floating point routines.

Someone fixed that bug: http://sleepingelephant.com/ipw-web/bulletin/bb/viewtopic.php?f=2&t=6772