fmt: sub-`10` exponents in formatted floats are zero-padded to at least two digits

[hudlow]

Go version

go version go1.23.4 darwin/arm64

Output of go env in your module/workspace:

GO111MODULE=''
GOARCH='arm64'
GOBIN=''
GOCACHE='/Users/zoomzoom/Library/Caches/go-build'
GOENV='/Users/zoomzoom/Library/Application Support/go/env'
GOEXE=''
GOEXPERIMENT=''
GOFLAGS=''
GOHOSTARCH='arm64'
GOHOSTOS='darwin'
GOINSECURE=''
GOMODCACHE='/Users/zoomzoom/go/pkg/mod'
GONOPROXY=''
GONOSUMDB=''
GOOS='darwin'
GOPATH='/Users/zoomzoom/go'
GOPRIVATE=''
GOPROXY='https://proxy.golang.org,direct'
GOROOT='/Users/zoomzoom/brew/Cellar/go/1.23.4/libexec'
GOSUMDB='sum.golang.org'
GOTMPDIR=''
GOTOOLCHAIN='local'
GOTOOLDIR='/Users/zoomzoom/brew/Cellar/go/1.23.4/libexec/pkg/tool/darwin_arm64'
GOVCS=''
GOVERSION='go1.23.4'
GODEBUG=''
GOTELEMETRY='local'
GOTELEMETRYDIR='/Users/zoomzoom/Library/Application Support/go/telemetry'
GCCGO='gccgo'
GOARM64='v8.0'
AR='ar'
CC='cc'
CXX='c++'
CGO_ENABLED='1'
GOMOD='/dev/null'
GOWORK=''
CGO_CFLAGS='-O2 -g'
CGO_CPPFLAGS=''
CGO_CXXFLAGS='-O2 -g'
CGO_FFLAGS='-O2 -g'
CGO_LDFLAGS='-O2 -g'
PKG_CONFIG='pkg-config'
GOGCCFLAGS='-fPIC -arch arm64 -pthread -fno-caret-diagnostics -Qunused-arguments -fmessage-length=0 -ffile-prefix-map=/tmp/go-build3247814003=/tmp/go-build -gno-record-gcc-switches -fno-common'

What did you do?

Using %e, %E, %x, or %X to format a floating-point value that ends up with an exponent between 0 and 9.

https://go.dev/play/p/oaSxLmPizik

fmt.Printf("%e, %E, %x, %X, %b\n", 1e0, 1e0, 1e0, 1e0, 5e+15)
fmt.Printf("%e, %E, %x, %X, %b\n", 1e1, 1e1, 1e1, 1e1, 1e+16)
fmt.Printf("%e, %E, %x, %X, %b\n", 1e10, 1e10, 1e10, 1e10, 1e20)
fmt.Printf("%e, %E, %x, %X, %b\n", 1e100, 1e100, 1e100, 1e100, 1e100)

What did you see happen?

The exponent value is zero-padded to two digits. Interestingly, %b does not exhibit this behavior.

1.000000e+00, 1.000000E+00, 0x1p+00, 0X1P+00, 5000000000000000p+0
1.000000e+01, 1.000000E+01, 0x1.4p+03, 0X1.4P+03, 5000000000000000p+1
1.000000e+10, 1.000000E+10, 0x1.2a05f2p+33, 0X1.2A05F2P+33, 6103515625000000p+14
1.000000e+100, 1.000000E+100, 0x1.249ad2594c37dp+332, 0X1.249AD2594C37DP+332, 5147557589468029p+280

What did you expect to see?

The exponent value is not zero-padded.

I realize it may be a bit cheeky to file this as a bug, given that the behavior seems deliberately coded. But, I couldn't find any documentation saying that it is correct behavior, and it doesn't seem to make much sense given that it does not result in a uniform length for exponents since they can also be three digits long.

I am also genuinely curious if it is intentional that %b behaves differently in this respect.

[gabyhelp]

Related Issues

• fmt: %#g behavior not consistent #36562 (closed)
• cmd/gofmt: change in exponent literal notation from 1.12 to 1.13 #32580 (closed)
• fmt: the "alternate form" (the # flag) for floating point is not supported #18857 (closed)
• fmt: %#v of integer value in float64 (fraction = 0) incorrectly has ".0" suffix #27598 (closed)
• fmt: get wrong result when using "%.30f" in fmt.Sprintf #47931 (closed)
• x/text/number, x/text/message: number formatting order of magnitude error #31254
• strconv: ParseFloat seems not working correctly with a large exponent #42436
• fmt: padding flag is not accepted with minus flag #51419 (closed)

Related Code Changes

• fmt: do not remove trailing zeros for %g and %G with #(sharp) flag

Related Discussions

• behavior of %f

_{(Emoji vote if this was helpful or unhelpful; more detailed feedback welcome in this discussion.)}

[robpike]

It's because that's what C does for the %e and %E formats:

% cat x.c
#include <stdio.h>

int main(void) {
   printf("%e\n", 13.0);
   return 0;
}
% gcc x.c
% a.out
1.300000e+01
% 

I suspect the behavior in Go for %b is more an accident then anything deliberate but probably won't change.

[meem]

FWIW, the C behavior (which does seem bizarre) is documented in the standard describing %e and %E (emphasis added):

A double argument representing a floating-point number is converted in the
style [−]d.ddd e±dd, where there is one digit (which is nonzero if the
argument is nonzero) before the decimal-point character and the number of
digits after it is equal to the precision; if the precision is missing, it is taken as
6; if the precision is zero and the # flag is not specified, no decimal-point
character appears. The value is rounded to the appropriate number of digits.
The E conversion specifier produces a number with E instead of e
introducing the exponent. The exponent always contains at least two digits,
and only as many more digits as necessary to represent the exponent. If the
value is zero, the exponent is zero.

[hudlow]

@robpike

It's because that's what C does for the %e and %E formats

I was holding out some hope this could be fixed since it's previously been stated that fmt is not beholden to C compatibility. But I assume at this point that changing this would run afoul of Hyrum's Law, and this should be considered a documentation bug?

@meem

FWIW, the C behavior (which does seem bizarre) is documented in the standard

I sort of have to imagine that this traces its lineage to a float formatter that didn't consider doubles? But ostensibly that would have been pre-C.

[robpike]

No, it's not beholden but it's also designed to be like it, especially when dealing with the same things. And because of the compatibility promise, the behavior is very unlikely to change. It's just too minor a thing to break compatibility with the past (and with C, to be honest).

[gopherbot]

Change https://go.dev/cl/636915 mentions this issue: fmt, strconv: document that exponent is always two digits

[hudlow]

Thank you @ianlancetaylor!