proposal: fmt: append decimals to whole-number floats when using '%#v'

[dave]

Please answer these questions before submitting your issue. Thanks!

What version of Go are you using (go version)?

1.10.3

Does this issue reproduce with the latest release?

Yes

What operating system and processor architecture are you using (go env)?

darwin/amd64

What did you do?

fmt.Printf("%#v", float64(0.0))
Expected: 0.0
Actual: 0
https://play.golang.org/p/KkcUz1voJG4

As per the description in e97f407 I would expect to see the decimal point added for all floats. The %#v format string is intended to return valid Go syntax. In this case, the returned Go syntax represents an integer instead of a float.

It looks like the commit linked above forgot that the sharp flag is replaced with sharpV in the case of the v verb here...

There's more discussion of this here: dave/jennifer#39
Attn: @martisch

[dave]

Hey @martisch I'd love to have a go at fixing this myself... Would that be ok with you?

[martisch]

Sure, happy if you send a fix with test case, if this is indeed agreed upon to need a fix (https://golang.org/doc/contribute.html).

Looking at it again It is not 100% clear to me if we should fix this and documentation breaking backwards compatibility or remove the mention of v in the code that handles # for floats.

Note that the behavior of printing 0 also seems to be in go1.4 and possibly earlier so not introduced by e97f407. #v is special and the documentation says "always print a decimal point for %e, %E, %f, %F, %g and %G;" so my commit message was inconsistent in that regard but from the code it should have applied to #v too.

Go language spec says "A floating-point literal is a decimal representation of a floating-point constant. It has an integer part, a decimal point, a fractional part, and an exponent part." (where some parts can be omitted) and fmt documentation says "%#v a Go-syntax representation of the value". Interestingly it only says of the value and not a literal of the type, so it would seem to me 0 is ok according to the documentation and does not force a 0.0 or 0. literal.

@robpike for further opinion if #v should always have a dot or leave current formatting to not break existing uses of #v (since at least go1.4).

I lean to keeping output as is and fix the code to not handle the unreachable case of verb v in conjunction with f.sharp.

[dave]

Yep I agree it could break backwards compatibility. It would definitely break the work-around I made for the bug here.

I'll leave you guys to make the call. Happy to fix if needed...

[robpike]

I think it's a bug and should be fixed.

[gopherbot]

Change https://golang.org/cl/123956 mentions this issue: fmt: fix incorrect format of whole-number floats when using %#v

[dave]

So there's a couple of corner cases that I've made assumptions:

Should 1.0 print 1.0 or 1.?
Although 1. is more compact, I think 1.0 looks like more natural Go syntax...

Should 1e6 print 1e+06 or 1.0e+06?
Since 1e+06 (which is the current output) will already be interpreted as a float, I don't think we want to change it.

Here's a PR: #26383

[dave]

Hey is there anything I need to do to keep this from going stale?

[bcmills]

As noted in #27634, a breaking change to fmt should go through the proposal process, bug or otherwise.

[bcmills]

See also #27634 (comment) for an argument that the existing behavior was correct (but admittedly not the clearest possible choice).

[gopherbot]

Change https://golang.org/cl/142597 mentions this issue: Revert "fmt: fix incorrect format of whole-number floats when using %#v"

[rsc]

Because this broke things, we rolled it back. We're going to leave things as is. Declining the proposal.