Output full numeric precision for floating point numbers

[traversc]

This is a question/feature request, as I don't think it exists as an option.

Example:

x <- -75.1637420654296875
fwrite(data.table(a = x), file="/tmp/x.txt")
xi <- fread("/tmp/x.txt")$a

identical(x, xi)
 # FALSE
x-xi
# 1.421085e-14

x is an exact floating point value (that is, fully representable by IEEE standards) according to: https://www.h-schmidt.net/FloatConverter/IEEE754.html

I would like to write and read that value to its full precision.

Thank you

[MichaelChirico]

have you seen this Q&A? https://stackoverflow.com/q/3978266/3576984

…

On Tue, Jun 12, 2018, 1:01 AM Travers ***@***.***> wrote: This is a question/feature request, as I don't think it exists as an option. Example: x <- -75.1637420654296875 fwrite(data.table(a = x), file="/tmp/x.txt") xi <- fread("/tmp/x.txt")$a identical(x, xi) # FALSE x-xi # 1.421085e-14 x is an exact floating point value (that is, fully representable by IEEE standards) according to: https://www.h-schmidt.net/FloatConverter/IEEE754.html I would like to write and read that value to its full precision. Thank you — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#2929>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AHQQdaDQ16XePrbls5aYJ1yr5MRiIQX_ks5t7qJpgaJpZM4UjDj8> .

[MichaelChirico]

Since here, we match the output of write.csv for floating point numbers: 6c1ed96 On Tue, Jun 12, 2018, 7:48 AM Michael Chirico <michaelchirico4@gmail.com> wrote:

…

have you seen this Q&A? https://stackoverflow.com/q/3978266/3576984 On Tue, Jun 12, 2018, 1:01 AM Travers ***@***.***> wrote: > This is a question/feature request, as I don't think it exists as an > option. > > Example: > > x <- -75.1637420654296875 > fwrite(data.table(a = x), file="/tmp/x.txt") > xi <- fread("/tmp/x.txt")$a > > identical(x, xi) > # FALSE > x-xi > # 1.421085e-14 > > x is an exact floating point value (that is, fully representable by IEEE > standards) according to: > https://www.h-schmidt.net/FloatConverter/IEEE754.html > > I would like to write and read that value to its full precision. > > Thank you > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > <#2929>, or mute the > thread > <https://github.com/notifications/unsubscribe-auth/AHQQdaDQ16XePrbls5aYJ1yr5MRiIQX_ks5t7qJpgaJpZM4UjDj8> > . >

[traversc]

Thanks for the links Michael. Would it be possible in the future include an option to write full precision, or otherwise to include a format specifier e.g. "%10.5f" as in numpy.savetxt?

[MichaelChirico]

actually there's #2020 as well. an option may be helpful.

[MichaelChirico]

I guess this is subsumed by #2020. Test version is at #3716 if you'd like to have a go. Please follow up in #2020/#3716 or feel free to re-open if you think there's anything left uncovered here... thanks!