improve: GetAllCharsets and GetCharsetInfoByID

[hawkingrei]

What problem does this PR solve?

pingcap/tidb#9635
some of the functions may just return constants. There is no need to do compute every time if we save a map/slice and initialize them. Such as: GetAllCharsets and GetCharsetInfoByID

What is changed and how it works?

1. GetAllCharsets is inited at the init function, and GetAllCharsets can return the inited array directly.
2. GetCharsetInfoByID use a map to fetch Collation effectively, init the map in the init function too.

Check List

• Unit test

Code changes

• improve GetAllCharsets and GetCharsetInfoByID

Related changes

• Need to update go.mod in tidb

[shenli]

@hawkingrei Thanks!
Is there any benchmark result to show the improvement?

[hawkingrei]

Benchmark GetAllCharsets

Before improve:

go test -bench Benchmark -run GetAllCharsets ./... -count 5

goos: darwin
goarch: amd64
pkg: github.com/pingcap/parser/charset
BenchmarkGetAllCharsets-4   	 3000000	       405 ns/op
BenchmarkGetAllCharsets-4   	 3000000	       405 ns/op
BenchmarkGetAllCharsets-4   	 3000000	       397 ns/op
BenchmarkGetAllCharsets-4   	 3000000	       447 ns/op
BenchmarkGetAllCharsets-4   	 3000000	       443 ns/op
PASS
ok  	github.com/pingcap/parser/charset	8.616s
PASS

After improve

go test -bench Benchmark -run GetAllCharsets ./... -count 5

goos: darwin
goarch: amd64
pkg: github.com/pingcap/parser/charset
BenchmarkGetAllCharsets-4   	2000000000	         0.38 ns/op
BenchmarkGetAllCharsets-4   	2000000000	         0.37 ns/op
BenchmarkGetAllCharsets-4   	2000000000	         0.37 ns/op
BenchmarkGetAllCharsets-4   	2000000000	         0.36 ns/op
BenchmarkGetAllCharsets-4   	2000000000	         0.36 ns/op
PASS

Here is benchmark code.

func BenchmarkGetAllCharsets(b *testing.B) {
	b.ResetTimer()
	for i := 0; i < b.N; i++ {
		GetAllCharsets()
	}
}

[hawkingrei]

Benchmark GetAllCharsets

Before improve:

go test -bench Benchmark -run GetGetCharsetInfoByID ./... -count 2

goos: darwin
goarch: amd64
pkg: github.com/pingcap/parser/charset
BenchmarkGetCharsetInfoByID-4   	30000000	        64.4 ns/op
BenchmarkGetCharsetInfoByID-4   	30000000	        72.4 ns/op
PASS
ok  	github.com/pingcap/parser/charset	4.208s

After improve

go test -bench Benchmark -run GetGetCharsetInfoByID ./... -count 2

goos: darwin
goarch: amd64
pkg: github.com/pingcap/parser/charset
BenchmarkGetCharsetInfoByID-4   	100000000	        20.6 ns/op
BenchmarkGetCharsetInfoByID-4   	100000000	        18.5 ns/op
PASS

Here is benchmark code.

func BenchmarkGetCharsetInfoByID(b *testing.B) {
	index := rand.Intn(248)
	b.ResetTimer()
	for i := 0; i < b.N; i++ {
		GetCharsetInfoByID(index)
	}
}

[hawkingrei]

@shenli it looks good

[tiancaiamao]

LGTM

[zimulala]

LGTM