core/state: optimize some internals during encoding

[holiman]

I noticed this wile investigating the state a bit. What we currently do before RLP-encoding a slot value is to call TrimLeft:

func TrimLeft(s []byte, cutset string) []byte {
	return TrimLeftFunc(s, makeCutsetFunc(cutset))
}

... which dynamically creates a cutSetFunction .
Then it calls TrimLeftFunc, which:

// TrimLeftFunc treats s as UTF-8-encoded bytes and returns a subslice of s by slicing off
// all leading UTF-8-encoded code points c that satisfy f(c).
func TrimLeftFunc(s []byte, f func(r rune) bool) []byte {

It interprets the data as UTF-8 encoded data, and parses runes.

In our case, it's the wrong abstraction, what we want to do here is simply trim leading zeroes, which has nothing whatsoever to do with UTF-8 and runes.

This PR replaces that with a simple for-loop, more or less.

benchmark results:

[user@work state]$ go test . -bench BenchmarkCut*
OK: 5 passed
goos: linux
goarch: amd64
pkg: github.com/ethereum/go-ethereum/core/state
BenchmarkCutOriginal-6     	10000000	       146 ns/op
BenchmarkCutsetterFn-6     	20000000	        86.3 ns/op
BenchmarkCutCustomTrim-6   	50000000	        28.9 ns/op
PASS
ok  	github.com/ethereum/go-ethereum/core/state	11.941s

It's probably a drop in the ocean though...

[karalabe]

return s[idx:] always?

[karalabe]

SGTM

[karalabe]

Linter fails, pls gofmt

[holiman]

done, let's see what the linter says next...