code review 5435089: strconv: fix documentation

strconv: fix documentation

Also: minor performance fix for large precision results.

benchmark                                   old ns/op    new ns/op    delta
strconv_test.BenchmarkFormatFloatDecimal         2734         2734   +0.00%
strconv_test.BenchmarkFormatFloat                3141         3139   -0.06%
strconv_test.BenchmarkFormatFloatExp             8970         8989   +0.21%
strconv_test.BenchmarkFormatFloatBig             3228         3208   -0.62%

Fixes issue 2535.

[gri, 2011-12-07 22:26:41]

Hello golang-dev@googlegroups.com,

I'd like you to review this change to
https://go.googlecode.com/hg/

[rsc, 2011-12-07 22:29:33]

LGTM

http://codereview.appspot.com/5435089/diff/4001/src/pkg/strconv/atoi.go
File src/pkg/strconv/atoi.go (right):

http://codereview.appspot.com/5435089/diff/4001/src/pkg/strconv/atoi.go#newco...
src/pkg/strconv/atoi.go:177: // Atoi is shorthand for int(ParseInt(s, 10, 0)).
Please undo.  I think the complaint in the issue tracker is overly pedantic.
No one is confused about the current text, and this text isn't any more
correct (it's a conversion of a 2-result function).

http://codereview.appspot.com/5435089/diff/4001/src/pkg/strconv/ftoa.go
File src/pkg/strconv/ftoa.go (right):

http://codereview.appspot.com/5435089/diff/4001/src/pkg/strconv/ftoa.go#newco...
src/pkg/strconv/ftoa.go:49: return string(genericFtoa(make([]byte, 0,
max(prec+4, 16)), f, fmt, prec, bitSize))
In general a float64 number needs 17 digits plus the
surrounding syntax to be written precisely.  So 24 would
probably be a better default choice than 16.

[gri, 2011-12-07 22:41:17]

PTAL.

http://codereview.appspot.com/5435089/diff/4001/src/pkg/strconv/atoi.go
File src/pkg/strconv/atoi.go (right):

http://codereview.appspot.com/5435089/diff/4001/src/pkg/strconv/atoi.go#newco...
src/pkg/strconv/atoi.go:177: // Atoi is shorthand for int(ParseInt(s, 10, 0)).
On 2011/12/07 22:29:33, rsc wrote:
> Please undo.  I think the complaint in the issue tracker is overly pedantic.
> No one is confused about the current text, and this text isn't any more
> correct (it's a conversion of a 2-result function).

Done.

http://codereview.appspot.com/5435089/diff/4001/src/pkg/strconv/ftoa.go
File src/pkg/strconv/ftoa.go (right):

http://codereview.appspot.com/5435089/diff/4001/src/pkg/strconv/ftoa.go#newco...
src/pkg/strconv/ftoa.go:49: return string(genericFtoa(make([]byte, 0,
max(prec+4, 16)), f, fmt, prec, bitSize))
On 2011/12/07 22:29:33, rsc wrote:
> In general a float64 number needs 17 digits plus the
> surrounding syntax to be written precisely.  So 24 would
> probably be a better default choice than 16.

It's a wash (actually everything gets slightly slower) except for big floats:

benchmark                                   old ns/op    new ns/op    delta
strconv_test.BenchmarkFormatFloatDecimal         2734         2746   +0.44%
strconv_test.BenchmarkFormatFloat                3141         3147   +0.19%
strconv_test.BenchmarkFormatFloatExp             8970         9017   +0.52%
strconv_test.BenchmarkFormatFloatBig             3228         3064   -5.08%

[gri, 2011-12-07 22:45:48]

*** Submitted as http://code.google.com/p/go/source/detail?r=65b2fe2d08bc ***

strconv: fix documentation

Also: minor performance fix for large precision results.

benchmark                                   old ns/op    new ns/op    delta
strconv_test.BenchmarkFormatFloatDecimal         2734         2734   +0.00%
strconv_test.BenchmarkFormatFloat                3141         3139   -0.06%
strconv_test.BenchmarkFormatFloatExp             8970         8989   +0.21%
strconv_test.BenchmarkFormatFloatBig             3228         3208   -0.62%

Fixes issue 2535.

R=rsc
CC=golang-dev
http://codereview.appspot.com/5435089

[rsc, 2011-12-07 22:46:00]

On Wed, Dec 7, 2011 at 17:41,  <gri@golang.org> wrote:
> It's a wash (actually everything gets slightly slower) except for big
> floats:
>
>
> benchmark                                   old ns/op    new ns/op
> delta
> strconv_test.BenchmarkFormatFloatDecimal         2734         2746
> +0.44%
> strconv_test.BenchmarkFormatFloat                3141         3147
> +0.19%
> strconv_test.BenchmarkFormatFloatExp             8970         9017
> +0.52%
> strconv_test.BenchmarkFormatFloatBig             3228         3064
> -5.08%

Seems like a win to me.