made string splits *much* faster

[dhiltonp]

This is in a simple case, not with regexes. I suspect there is a similar optimization available in the neighboring code, but I wanted to get some feedback on this change first.

Before: BenchmarkSplit-8 20000 77864 ns/op
After: BenchmarkSplit-8 200000 8017 ns/op

Test run with go test -bench=BenchmarkString_split

edit:
changed comment to reflect move of test to strings_test.go.

[stevenh]

Can you include the benchmark as part of the PR too please.

[dhiltonp]

Done.

[dhiltonp]

I've done a little benchmarking of the regex sections of string.split().

As is, the same performance improvement cannot be directly applied, as golang's default regex library doesn't include undefined matches in this case: A<B>bold</B>and<CODE>coded</CODE>".split(/<(\/)?([^<>]+)>/).

However, assuming the replacement regex library (as in issues #205 and #215) performs splits resulting in undefined matches correctly, we can remove about 50 LoC from split and gain a 33% speed up in regex splits.

before: BenchmarkString_splitWithRegex-8 30000 40421 ns/op
after: BenchmarkString_splitWithRegex-8 50000 27063 ns/op

---

I'm currently adding benchmarks for related string operations, which will be useful for determining regressions introduced by #205.

[dhiltonp]

Any additional input?

[deoxxa]

Looks great, thanks! Sorry for the delay!