Benchmarks to compare the different string concatenation methods in Go. For all the details, see stringconcat_test.go.
There is also a blog post detailing the methodology and results.
In summary, for very few concatenations of short strings (fewer than hundred, length shorter than 10) using naive string appending is just fine. For more heavy-duty cases where efficiency is important, bytes.Buffer is the best choice out of the methods evaluated. strings.Join is a good choice when you already have a string slice that just needs to be concatenated into one string.
Here are how the methods tested stack up:
And here are the raw results (also including a benchmark for byte slices):
BenchmarkNaiveConcat10-3 2000000 1192 ns/op 360 B/op 11 allocs/op
BenchmarkNaiveConcat100-3 50000 34117 ns/op 26408 B/op 101 allocs/op
BenchmarkNaiveConcat1000-3 500 2641900 ns/op 2694414 B/op 1004 allocs/op
BenchmarkNaiveConcat10000-3 10 188733914 ns/op 271630262 B/op 10339 allocs/op
BenchmarkByteSlice10-3 2000000 717 ns/op 208 B/op 7 allocs/op
BenchmarkByteSlice100-3 300000 4000 ns/op 1552 B/op 10 allocs/op
BenchmarkByteSlice1000-3 30000 53874 ns/op 26128 B/op 16 allocs/op
BenchmarkByteSlice10000-3 2000 701131 ns/op 283668 B/op 24 allocs/op
BenchmarkByteSliceSize10-3 3000000 536 ns/op 200 B/op 4 allocs/op
BenchmarkByteSliceSize100-3 300000 3396 ns/op 1560 B/op 4 allocs/op
BenchmarkByteSliceSize1000-3 30000 40858 ns/op 15896 B/op 4 allocs/op
BenchmarkByteSliceSize10000-3 2000 575724 ns/op 163866 B/op 4 allocs/op
BenchmarkJoin10-3 1000000 1724 ns/op 648 B/op 9 allocs/op
BenchmarkJoin100-3 200000 9183 ns/op 5128 B/op 12 allocs/op
BenchmarkJoin1000-3 20000 82654 ns/op 43528 B/op 15 allocs/op
BenchmarkJoin10000-3 1000 1540246 ns/op 941844 B/op 24 allocs/op
BenchmarkJoinSize10-3 2000000 887 ns/op 312 B/op 5 allocs/op
BenchmarkJoinSize100-3 200000 5981 ns/op 2712 B/op 5 allocs/op
BenchmarkJoinSize1000-3 20000 66210 ns/op 27160 B/op 5 allocs/op
BenchmarkJoinSize10000-3 2000 814003 ns/op 278555 B/op 5 allocs/op
BenchmarkBufferString10-3 1000000 1354 ns/op 400 B/op 8 allocs/op
BenchmarkBufferString100-3 200000 6410 ns/op 2368 B/op 11 allocs/op
BenchmarkBufferString1000-3 20000 56510 ns/op 18880 B/op 14 allocs/op
BenchmarkBufferString10000-3 2000 699816 ns/op 170178 B/op 17 allocs/op
BenchmarkBufferSize10-3 2000000 766 ns/op 312 B/op 5 allocs/op
BenchmarkBufferSize100-3 300000 4548 ns/op 1672 B/op 5 allocs/op
BenchmarkBufferSize1000-3 30000 50437 ns/op 16008 B/op 5 allocs/op
BenchmarkBufferSize10000-3 2000 670182 ns/op 163978 B/op 5 allocs/op