Twayne wrote:
In news:[email protected],
Don Daugherty <[email protected]> typed:
On 3/3/2010 1:36 PM, AG wrote:
JOE Conner wrote:
On 3/3/2010 10:52 AM, AG wrote:
11.73 + 7 + 15.75 + 24.68 + 17.5 + 21.18 + 20.65 + 17.85
+ 19.25 + 23.1 + 12.6 + 10.5 + 23.8
TOOLS -> OPTIONS -> OPENOFFICE.ORG CALC -> CALCULATE ->
Right Panel, do you have PRECISION AS SHOWN checked on
one of the machines? Joe Conner, Poulsbo, WA USA

------------------------------------------------------------------------

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
Joe

Cheers for the rapid response.  Neither of our machines
had that option enabled.  However, even after enabling the
"precision as shown" the error persists on my partner's
version of Calc. Thanks

AG

From my reading of Help, you wouldn't want that option
enabled anyway since it would tend to result in more
roundoff's.  I'm thinking that JOE wasn't recommending it
as a cure but as the possible problem.

I know when messing with Floating Point math in about any program, you can get results similar to the original one stated here. Excel seems to use double-precision no matter what you display and usually avoids this issue but it can still be made to happen even in Excel. I'ts very easy to duplicate in most development languages too if you aren't watching your precision. Apparently, and I don't know this for a fact, Calc only uses the displayed digits for its math from a couple quick tests I just did. Somehow it seems to be 4 places of precision for calculations unless you program it differently. Personally I haven't encountered such a problem in Calc, but I think that's because I have a habit of always using at least two positions more of precision than is necessary, three if they aren't ridiculous numbers. There is a diminishing point of no return when you for instance set pi to be 3.14 or 3.14xx or 3.14xxxxx, and so forth. It's a complex subject for program authors and one that results in many debates around some circles about which way to do them. Even my invoices, which need two decimal places of precision, are calculated on 5 decimal places, but ... use the standard currency formatting of only two decimal places. No one wants a bill for $133.3333. But when I'm multiplying $5.35 by 10,013 pcs, the differences between that result and using only two places amounts to a lot of money at the end of the year over all the invoices. It comes out more accurate for me and more accurate for the customer too.

Dunno if that makes any sense; it's the best I can do right now though so I hope it helps someone out.

HTH,

Twayne`

Hi Twayne

Thanks for your thoughts, most of which I think that I could follow. I take your point about the float and if this had been replicated on my machine as well, I would say that you are probably onto something that would bear further investigation - such as increasing the number of decimal places. But, with the same list of numbers and the same programme (albeit a newer version), I could not replicate the problem ... even though we calibrated the settings to match across the machines. As I said in my reply to Harold - we are stumped and I have begun to wonder if it may be a hardware issue to do with the system clock (her machine also loses time, which reinforces my suspicions). Without getting into too much detail, I am going to have to resurrect an older machine, ensure that it still works and do an installation and then test it on there. If I can eradicate the problem on the resurrected box, then it just means doing the full installation so she can dual boot and all of the rest of her preferred settings and write it off as a crap build on her current machine, but if I can't eradicate the problem ... . But one step at a time. If the ideas/ suggestions here dry up, then I'll bite the bullet and see if I can get another machine up and running and migrate her to that.

Thanks for your ideas.

AG

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to