PC Specification:
AMD FX-9580 4.7Ghz, 32GB RAM 1066MHz
Go Version:
go version go1.6.2 windows/amd64
$ GOGC=100 go test -benchmem -bench=Simple$
BenchmarkSchemaSimple-8 1000000 1633 ns/op 152 B/op 6 allocs/op
BenchmarkFormamSimple-8 2000000 777 ns/op 40 B/op 3 allocs/op
BenchmarkCjToolkitFormSimple-8 1000000 1772 ns/op 272 B/op 3 allocs/op
$ GOGC=800 go test -benchmem -bench=Simple$
BenchmarkSchemaSimple-8 1000000 1016 ns/op 152 B/op 6 allocs/op
BenchmarkFormamSimple-8 2000000 666 ns/op 40 B/op 3 allocs/op
BenchmarkCjToolkitFormSimple-8 2000000 851 ns/op 272 B/op 3 allocs/op
$ GOGC=100 go test -benchmem -bench=SimpleWithUnusedData$
BenchmarkSchemaSimpleWithUnusedData-8 100000 11152 ns/op 1112 B/op 43 allocs/op
BenchmarkFormamSimpleWithUnusedData-8 1000000 1696 ns/op 146 B/op 6 allocs/op
BenchmarkCjToolkitFormSimpleWithUnusedData-8 1000000 1880 ns/op 272 B/op 3 allocs/op
$ GOGC=800 go test -benchmem -bench=SimpleWithUnusedData$
BenchmarkSchemaSimpleWithUnusedData-8 200000 7030 ns/op 1112 B/op 43 allocs/op
BenchmarkFormamSimpleWithUnusedData-8 1000000 1093 ns/op 145 B/op 6 allocs/op
BenchmarkCjToolkitFormSimpleWithUnusedData-8 2000000 854 ns/op 272 B/op 3 allocs/op
$ GOGC=100 go test -benchmem -bench=Complex$
BenchmarkSchemaComplex-8 1000000 4238 ns/op 360 B/op 18 allocs/op
BenchmarkFormamComplex-8 1000000 2590 ns/op 128 B/op 12 allocs/op
BenchmarkCjToolkitFormComplex-8 1000000 3161 ns/op 272 B/op 3 allocs/op
$ GOGC=800 go test -benchmem -bench=Complex$
BenchmarkSchemaComplex-8 1000000 2948 ns/op 360 B/op 18 allocs/op
BenchmarkFormamComplex-8 1000000 2165 ns/op 128 B/op 12 allocs/op
BenchmarkCjToolkitFormComplex-8 1000000 2075 ns/op 272 B/op 3 allocs/op
$ GOGC=100 go test -benchmem -bench=ComplexWithUnusedData$
BenchmarkSchemaComplexWithUnusedData-8 100000 14338 ns/op 1320 B/op 55 allocs/op
BenchmarkFormamComplexWithUnusedData-8 1000000 2086 ns/op 177 B/op 10 allocs/op
BenchmarkCjToolkitFormComplexWithUnusedData-8 1000000 3160 ns/op 272 B/op 3 allocs/op
$ GOGC=800 go test -benchmem -bench=ComplexWithUnusedData$
BenchmarkSchemaComplexWithUnusedData-8 200000 9550 ns/op 1320 B/op 55 allocs/op
BenchmarkFormamComplexWithUnusedData-8 1000000 1910 ns/op 192 B/op 11 allocs/op
BenchmarkCjToolkitFormComplexWithUnusedData-8 1000000 2159 ns/op 272 B/op 3 allocs/op
As the benchmark above shows, CJToolkit Form 3.0 performs reasonably well against Gorilla Schema and Formam
In most cases it's outperformed Schema and in some cases it can outperform Formam, especially if you tune up the Garbage Collection to 800 (default is 100) than CJToolkit Form get faster most likely because it's consistently 3 allocs/op.
Introducing unused data to url.Values (or even HTTP POST) will cause a negative impact to Schema, thus memory allocation goes up dramatically, therefore performance goes down dramatically and this is quite scary. But only has little to no impact with CJToolkit Form and Formam.
CJToolkit Form has got a bigger codebase than the other two, but the codes are highly maintainable, testable and make no use of reflect, while the other two uses reflect. I firmly believe that a robust, easy-to-test and modular designed codebase is much better than having a quick and dirty codebase with the abuse of reflect.
I choosed not to use reflect because I just don't like the idea of using that for something as simple as map[string][]string.