Making Value opaque means we can drop the interface kludges
in favor of a significantly simpler and faster representation.
v.Kind() will be a prime candidate for inlining too.
On a Thinkpad X201s using -benchtime 10:
benchmark old ns/op new ns/op delta
json.BenchmarkCodeEncoder 284391780 157415960 -44.65%
json.BenchmarkCodeMarshal 286979140 158992020 -44.60%
json.BenchmarkCodeDecoder 717175800 388288220 -45.86%
json.BenchmarkCodeUnmarshal 734470500 404548520 -44.92%
json.BenchmarkCodeUnmarshalReuse 707172280 385258720 -45.52%
json.BenchmarkSkipValue 24630036 18557062 -24.66%
benchmark old MB/s new MB/s speedup
json.BenchmarkCodeEncoder 6.82 12.33 1.81x
json.BenchmarkCodeMarshal 6.76 12.20 1.80x
json.BenchmarkCodeDecoder 2.71 5.00 1.85x
json.BenchmarkCodeUnmarshal 2.64 4.80 1.82x
json.BenchmarkCodeUnmarshalReuse 2.74 5.04 1.84x
json.BenchmarkSkipValue 77.92 103.42 1.33x
I cannot explain why BenchmarkSkipValue gets faster.
Maybe it is one of those code alignment things.
R=iant, r, gri, r
CC=golang-dev
https://golang.org/cl/5373101
Fixes use of c after Dial failure (causes crash).
May fix Dial failure by listening to 127.0.0.1:0
instead of 0.0.0.0:0 (tests should only listen on
localhost).
R=golang-dev, gri
CC=golang-dev
https://golang.org/cl/5395052
I've modified Plan 9's yacc to work with
the grammar in go.y. These are the only
changes necessary on the Go side.
R=rsc
CC=golang-dev
https://golang.org/cl/5375104
The wrong length was being sent, and two parameters
were also transposed. Made the record type be a type
and made the constants typed, to prevent that sort
of bug in the future.
Fixes#2469
R=golang-dev, edsrzf
CC=golang-dev
https://golang.org/cl/5394046
text/template does this (in an entirely different way), so
make html/template do the same. Before this fix, the template
{{.}} given a pointer to a string prints its address instead of its
value.
R=mikesamuel, r
CC=golang-dev
https://golang.org/cl/5370098
Ensure that empty NameLists always return
a zero length []string, not nil.
In practice NameLists are only used in a few
message types and always consumed by a for
range function so the difference between nil
and []string{} is not significant.
Also, add exp/ssh to pkg/Makefile as suggested
by rsc.
R=rsc, agl
CC=golang-dev
https://golang.org/cl/5400042
Move scanner allocation out of loop.
It's the only allocation in the test so it dominates
when it triggers a garbage collection.
R=golang-dev, r
CC=golang-dev
https://golang.org/cl/5369117
Errors in the code under test go to standard output.
Errors in testing or its usage go to standard error.
R=r
CC=golang-dev
https://golang.org/cl/5374090
I've been using this since April and posted it on the
mailing list, but it seems worth having in the repository.
Not sure about the location.
R=golang-dev, r, r
CC=golang-dev
https://golang.org/cl/5371100
An experiment: allow structs to be copied even if they
contain unexported fields. This gives packages the
ability to return opaque values in their APIs, like reflect
does for reflect.Value but without the kludgy hacks reflect
resorts to.
In general, we trust programmers not to do silly things
like *x = *y on a package's struct pointers, just as we trust
programmers not to do unicode.Letter = unicode.Digit,
but packages that want a harder guarantee can introduce
an extra level of indirection, like in the changes to os.File
in this CL or by using an interface type.
All in one CL so that it can be rolled back more easily if
we decide this is a bad idea.
Originally discussed in March 2011.
https://groups.google.com/group/golang-dev/t/3f5d30938c7c45ef
R=golang-dev, adg, dvyukov, r, bradfitz, jan.mercl, gri
CC=golang-dev
https://golang.org/cl/5372095
No need for creating a new nat each time.
Per Roger Peppe's suggestion; assuming
nat(nil) produces better code than nat{}.
R=rsc
CC=golang-dev
https://golang.org/cl/5375092
