math/big: Float.String inconsistent with format %v

[rogpeppe]

go version devel +926994fd7c Thu Nov 26 21:10:09 2020 +0000 linux/amd64

I was surprised today when I discovered that big.Float.String truncates to 10 decimal digits of precision, which is surely not a very useful default for a "big" float. I would expect the result to use as many digits of precision as necessary, as if it was using Float.Text('g', -1).

Another remark: most types that implement fmt.Stringer work with the %s verb. Perhaps big.Float could do that too, making %s equivalent to %v.

The following code prints three different things. I'd expect it to print the same thing three times:

x := big.NewFloat(4.2949672961e+09)

fmt.Printf("%v\n", x)
fmt.Printf("%s\n", x.String())
fmt.Printf("%s\n", x)

[cagedmantis]

/cc @griesemer @FiloSottile

[dmitshur]

@griesemer Do you expect you'll have time to investigate and fix this for Go 1.17? We're three weeks into the freeze now and it doesn't seem someone is working on this, so I'll tentatively move this to Backlog, but please feel free to move it back otherwise.

[griesemer]

Probably not hard to fix but will take some time to ensure we're not breaking something else. Won't get to this for 1.17 due to higher priority items.