For pd64, I would just remove the precision (.6) and go normal %g, leaving it up to the system. I don't see the need for making it configurable globally or on a per-object basis, which is adding more complexity. Keep the "data pure" so people can clip, filter, apply an epsilon check for small numbers, etc. ;)
If you want to show *everything* losslessly, apparently %.9g for float and %.17g for double, although that seems overdone to me. There are also the FLT_DECIMAL_DIG and DBL_DECIMAL_DIG defines in float.h for this. :)
Ref: https://stackoverflow.com/a/21162120
I prefer the simplicity of letting %g of just do it's thing in this case, and keep %.6g as it is for pd32.
On Nov 26, 2024, at 2:20 AM, pd-dev-request@lists.iem.at wrote:
Message: 3 Date: Mon, 25 Nov 2024 22:20:32 -0300 From: Alexandre Torres Porres <porres@gmail.com mailto:porres@gmail.com> Subject: [PD-dev] Re: why nor increase the '%g' precision for pd64? To: Giulio Moro <giuliomoro@yahoo.it mailto:giuliomoro@yahoo.it> Cc: pd-dev <pd-dev@lists.iem.at mailto:pd-dev@lists.iem.at> Message-ID: <CAEAsFmg=F9zWi0970c83sP4dMYGoOF794w-5DbFTSVcUy0fiSA@mail.gmail.com mailto:CAEAsFmg=F9zWi0970c83sP4dMYGoOF794w-5DbFTSVcUy0fiSA@mail.gmail.com> Content-Type: multipart/alternative; boundary="00000000000076eed00627c6a88b"
The thing is that a higher precision in single precision will give you many float imprecision errors and it's also annoying.
I guess using %g is a good compromise for single precision in messages, objects (as arguments) and comments. Maybe being able to set with more versatility would be nice in an object like [print], or as a new parameter for the number boxes. Or I can try and create externals that deal with this for power users, I don't know...
But at least a hardcoded higher precision makes sense in general for pd64. I don't know yet which one would be good and others here may know better. And I think that a highger precision by default would deal with your issue Giulio, right?
cheers
-------- Dan Wilcox danomatika.com http://danomatika.com/ robotcowboy.com http://robotcowboy.com/