Ah, I see. I also found the issue was brought up previously: https://www.mail-archive.com/picolisp@software-lab.de/msg10799.html
I agree with keeping the spec small (in regards to the most recent discussion), but possibly another token to identify bignum returns? -Kevin On Thu, Dec 9, 2021 at 1:29 PM Alexander Burger <a...@software-lab.de> wrote: > Hi Kevin, > > > found some of the intermediate > > values (the determinate) were smaller than 1e6 (my scl is 6) and so my > > thought is to increase the scale to increase the precision. > > The problem are not the small numbers, but the larger ones. > > I think it is a design flaw in the float/double interface in PicoLisp, as > it > uses short numbers to pass floats and doubles to and from native. > > If you have scale 16 and parse 123.0, you get 61 bits: > > : (scl 16) > -> 16 > : (length (bin 123.0)) > -> 61 > > but a short num can be maximally 60 bits. > > A *small* number is no problem: > > : (length (bin 0.000123)) > -> 41 > > The conversion functions in @src/lib.c (mentioned in my previous mail) > should > indeed overflow to bignums. In fact, pil32 did it this way (it had no > short nums > but only bignums). > > ☺/ A!ex > > -- > UNSUBSCRIBE: mailto:picolisp@software-lab.de?subject=Unsubscribe >
