Subject | Re: [IBO] DML Caching |
---|---|
Author | Jason Wharton |
Post date | 2001-09-10T20:54:10Z |
When I looked at the extended type it seemed to have a 64bit mantissa which
means that it is capable of holding an int64 without any loss of accuracy.
The part I don't know is if the extended type was designed to favor strict
decimal accuracy wherever possible or if it is more concerned about
significant digit places up to 15 digits, which if that is the case isn't
going to cut it.
Anyone else have any insights here?
Thanks,
Jason Wharton
CPS - Mesa AZ
http://www.ibobjects.com
means that it is capable of holding an int64 without any loss of accuracy.
The part I don't know is if the extended type was designed to favor strict
decimal accuracy wherever possible or if it is more concerned about
significant digit places up to 15 digits, which if that is the case isn't
going to cut it.
Anyone else have any insights here?
Thanks,
Jason Wharton
CPS - Mesa AZ
http://www.ibobjects.com
----- Original Message -----
From: "Geoff Worboys" <geoff@...>
To: <IBObjects@yahoogroups.com>
Sent: Saturday, September 08, 2001 11:38 PM
Subject: Re: [IBO] DML Caching
> > Assigning it from AsInt64 in Delphi 6 didn't work. I
> > am using AsExtended and this seems to work fine.
>
> Do you know if Extended in variant is treated as 80bit double for all
> operations - or do they get downsized to 64bit doubles? If they can
> be relied on to be always 80bit then perhaps it would OK - since the
> significance is close to the same (but not exactly). However I
> suspect that problems are likely at the limits of int64.
>
> This was the reason I suggested using AsString (varString variant),
> since this should allow fully accurate representation - even though it
> is probably not as efficient.
>
> The only other alternative would be to find some alternative to using
> variant arrays when int64s may be involved.
>
> Geoff Worboys
> Telesis Computing
>
>
>
>
>
> Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
>
>