Think the core processing needs to stick with standard floating point (especially for auto-whitelist).
Perhaps just the report formatting of the output could be 'cleaned' up a little bit. eric Vivek Khera wrote: > > >>>>> "EW" == Eric Whiting <[EMAIL PROTECTED]> writes: > > EW> I'm seeing some floating point rounding errors (examples below). I think > EW> it isn't really an error -- just an artifact of IEEE floating point > EW> math. It might look better with some sort of sprintf formatting to > EW> something like %6.4f or similar. > > Any reason why we can't internally switch to integer arithmetic by > multiplying all the scores by 10 then dividing by 10 for the final > display/comparison? A sort of poor-man's fixed point computation? As > an aside, you get speed improvements (not that the arithmetic is the > major time usage here...) > > Either that or use Math::BigFloat for better precision and control > over rounding. > > _______________________________________________________________ > > Have big pipes? SourceForge.net is looking for download mirrors. We supply > the hardware. You get the recognition. Email Us: [EMAIL PROTECTED] > _______________________________________________ > Spamassassin-talk mailing list > [EMAIL PROTECTED] > https://lists.sourceforge.net/lists/listinfo/spamassassin-talk _______________________________________________________________ Have big pipes? SourceForge.net is looking for download mirrors. We supply the hardware. You get the recognition. Email Us: [EMAIL PROTECTED] _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
