Update SMALLFLOAT Larger Than -1.175494351E-38 Works?
Rick Spagna asked why Informix accepts a SMALLFLOAT like -5.6051938E-45, below the documented minimum magnitude, but rejects 3.4028234E+39. Andreas Legner could only insert about -1.1754943E-38 on Linux 14.10.FC10 but found that on Windows the minimum absolute value can be much smaller. Rick asked whether a bug would be opened; no answer in the thread.
Auto-generated by Claude from the posts below — may be imperfect; read the full thread.
According to multiple sources, the ranges of a FLOAT are: Positive numbers: Min Value: 1.175494351E-38 and Max Value: 3.402823466E+38 Negative Numbers: Min value: -3.402823466E+38 and Max Value: -1.175494351E-38 Given that, why can I successfully insert into an Informix SMALLFLOAT a value of -5.6051938E-45, which is greater than the maximum value? Yet, I can not insert a SMALLFLOAT value of 3.4028234E+39, which is greater than the maximum value: I've experimented a little by decreasing the exponent for negative numbers and when the exponent gets to -46, the update fails. Sure seems to me that this is a bug in Informix. Appreciate any input! ------------------------------ Thanks, Rick Spagna Optim Development Director UNICOM Global ------------------------------
Looks like the highest negative number (closest to zero) I can insert into a smallfloat, using dbaccess, is -1.175494281E-38.
I'm using 14.10.FC10 on Linux - what's yours?
------------------------------
Andreas Legner
------------------------------
Hi Andreas,
I'm also 14.10.FC10, but on Windows:
E:\\Programs\\informix>oninit -V
IBM Informix Dynamic Server Version 14.10.FC10W2DE
Thu Feb 1 12:57:41 CST 2024
E:\\Programs\\informix>
Can you try on Windows and see what you get?
Thanks,
Rick
------------------------------
Rick Spagna
------------------------------
Oh, yeah, tried it, and explored a little further: looks like specifically on Windows that minimum (absolute) value can be a LOT smaller than on other platforms, both for SMALLFLOAT and FLOAT. Strange since neither float nor double data types seem to support this... ------------------------------ Andreas Legner ------------------------------
Is a bug going to be opened on this so it can be corrected? ------------------------------ Rick Spagna ------------------------------