Saxon

Legacy ID: #3603766 Legacy Poster: Michael Kay (mhkay)

Please see http://sourceforge.net/tracker/index.php?func=detail&aid=1439685&group_id=29872&atid=397617 There will always be several different strings that convert to the same floating point number, so when you convert back to a string there can never be a guarantee that you will get the string you started with, though it should always be "equivalent". If you care about precision, I would strongly suggest using double rather than float. Having said that, the new conversion routine introduced in 8.7 has not proved to be a success.

Legacy ID: #3605065 Legacy Poster: Francis Upton (francisu)

Thanks Mike. This use of FP is for simple decimal monetary amounts which don't exceed 2 decimal places, so one hopes that the conversion would be fairly precise. When using relatively small quantities (such as this), is the use of "double" an improvement over "float"? Should I never use float? I took your file and that fixed the problem.

Legacy ID: #3605096 Legacy Poster: Michael Kay (mhkay)

For monetary amounts I would recommend decimal. Float might have enough digits for your numbers, but arithmetic rounding errors accumulate very rapidly in computations, for example if you take an average or a total of a set of numbers.

Legacy ID: #3605103 Legacy Poster: Michael Kay (mhkay)

And yes, my advice is never to use float. I really can't see the need for 32-bit floating point numbers in the 21st century.