Workaround: use fmt.Sprintf("%#.*g", 6, 1.5) in place of strconv.FormatFloat(1.5, 'g', 6, 64).

I agree that the doc and the implementation are in contradiction, and either one or the other should be fixed.

I'm not sure what is the basis for removing trailing zeros.
The number of significant digits in a number is not part of its float representation, that's why we ask the caller about it.
On the contrary the number of significant digits in a number, is part of its string representation, and by removing trailing zeros we are altering it.

The relevant piece of code seems to be here: https://golang.org/src/strconv/ftoa.go?s=5694:5737#L206

If the current implementation is preserved, I propose to fix like this:

The precision prec controls the number of digits (excluding the exponent) printed by the 'e', 'E', 'f', 'g', and 'G' formats. For 'e', 'E', and 'f' it is the number of digits after the decimal point. For 'g' and 'G' it is the total number of digits maximum number of significant digits (trailing zeros are removed). The special precision -1 uses the smallest number of digits necessary such that ParseFloat will return f exactly.

Verbs %g and %G in the fmt package follow C behavior, which is fine, but again the doc is not clear enough:

For floating-point values, width sets the minimum width of the field and precision sets the number of places after the decimal, if appropriate, except that for %g/%G precision sets the total number of significant digits maximum number of significant digits (trailing zeros are removed). For example, given 12.345 the format %6.3f prints 12.345 while %.3g prints 12.3. The default precision for %e, %f and %#g is 6; for %g it is the smallest number of digits necessary to identify the value uniquely.

Yes, the doc for the # flag has the missing information, but you won't find it if you just look for "precision".

Thanks, and I agree in general.

I've always been mindful of communicating the correct number of significant digits. With fmt and %g we at least have the # flag, but it is very unfortunate that strconv.FormatFloat doesn't offer any option.

Ideally, strconv.FormatFloat should offer such an option (and maybe support for other flags as well) and the post-processing in fmt could be simplified, offering on-par features to the users of fmt and strconv.FormatFloat. (It may not be easy without modifying the interface of strconv.FormatFloat, though.) Without that, I would personally favor fixing the implementation, but that would mean a small change in the behavior of current programs using it, so we may have to fix the documentation instead.

I don't think we should change the code, so marking this as a documentation fix.

Hi! It seems that no one is working on this right now... have we agreed that this is just a doc change? May I do it?

Change https://golang.org/cl/127135 mentions this issue: strconv: clarify "g" and "G" precision in the docs