@ktye in ngn/k's tests? yes, that's what it does

@ktye what restriction?

on the values. are tagged pointers involved or purely data?

@ktye it's purely data

x.c is the (de)serializer, it's very dumb

@ngn how do you format, e.g. when do you switch from decimal to exponential? it seems different from small or large numbers.

@ktye this is the function that formats doubles

it seems to be 1e16 and 1e-4. 1e16 i think is rather large.

yep, 0<x≤1e-4 and 1e16<x use scientific

@ktye now i remember, i chose that because 17 meaningful digits determine any double uniquely

@ktye what caused this sudden interest in fp roundtripping? :)

i use only e3 e6 e9... i think it's more readable if the numbers represent physical quantities where you are used to have units with prefixes. i havn't seen it used elsewhere like that.

@ngn nothing particular. kelas asked about it.

@ktye except in india :)

does libc do that, when you have some env variable set??

i don't know