Benchmarking Go errors

How does creation, serialization, and wrapping changes with different errors?

Notation

• fmt.Errorf you wrap like this fmt.Errorf("my message %s with data %f: %w", err)
• DelayedError you wrap underlying error into field but do not call e.Err.Error() immediately
• errors.New you create error with string passed directly into struct, no fmt calls made.

Results

1. fmt.Errorf wrapping calls Error immediately. Delaying Error call is not utilized
2. errors.New is very fast when no complex formatting (floats/reflection/iterables) is necessary
3. DelayedError is 10x faster when Error calls are not done, since Error calls are lazily delayed
4. Error call for fmt.Errorf and for DelayedError takes same time

Benchmarks

$ go test -bench=. -benchtime=5s -benchmem ./...
goos: darwin
goarch: amd64
pkg: github.com/nikolaydubina/go-bench-errors
cpu: VirtualApple @ 2.50GHz
BenchmarkWrap_DelayedError_float64_Message_2wrap-10     	14211424	       419 ns/op	     216 B/op	       8 allocs/op
BenchmarkWrap_DelayedError_float64_2wrap-10             	99927066	        58 ns/op	      96 B/op	       2 allocs/op
BenchmarkWrap_DelayedError_float64_Message_5wrap-10     	 5403166	      1097 ns/op	     736 B/op	      20 allocs/op
BenchmarkWrap_DelayedError_float64_5wrap-10             	38846422	       153 ns/op	     240 B/op	       5 allocs/op
BenchmarkWrap_DelayedError_float64_Message_10wrap-10    	 2672426	      2246 ns/op	    2072 B/op	      40 allocs/op
BenchmarkWrap_DelayedError_float64_10wrap-10            	19889925	       299 ns/op	     480 B/op	      10 allocs/op
BenchmarkWrap_DelayedError_float64_Message_50wrap-10    	  394473	     14857 ns/op	   35814 B/op	     200 allocs/op
BenchmarkWrap_DelayedError_float64_50wrap-10            	 4079131	      1467 ns/op	    2400 B/op	      50 allocs/op
BenchmarkWrap_fmt_Errorf_float64_Message_2wrap-10       	13853347	       430 ns/op	     176 B/op	       6 allocs/op
BenchmarkWrap_fmt_Errorf_float64_2wrap-10               	13966632	       429 ns/op	     176 B/op	       6 allocs/op
BenchmarkWrap_fmt_Errorf_float64_Message_5wrap-10       	 5036449	      1191 ns/op	     664 B/op	      15 allocs/op
BenchmarkWrap_fmt_Errorf_float64_5wrap-10               	 5040122	      1189 ns/op	     664 B/op	      15 allocs/op
BenchmarkWrap_fmt_Errorf_float64_Message_10wrap-10      	 2418132	      2485 ns/op	    2048 B/op	      30 allocs/op
BenchmarkWrap_fmt_Errorf_float64_10wrap-10              	 2414413	      2484 ns/op	    2048 B/op	      30 allocs/op
BenchmarkWrap_fmt_Errorf_float64_Message_50wrap-10      	  353810	     16856 ns/op	   38867 B/op	     150 allocs/op
BenchmarkWrap_fmt_Errorf_float64_50wrap-10              	  353611	     16808 ns/op	   38867 B/op	     150 allocs/op
BenchmarkSingle_fmt_Errorf_float64_Message-10           	27763376	       215 ns/op	      72 B/op	       3 allocs/op
BenchmarkSingle_fmt_Errorf_float64-10                   	30912550	       192 ns/op	      72 B/op	       3 allocs/op
BenchmarkSingle_fmt_Errorf_int_Message-10               	46369344	       129 ns/op	      51 B/op	       3 allocs/op
BenchmarkSingle_fmt_Errorf_int-10                       	56102511	       106 ns/op	      51 B/op	       3 allocs/op
BenchmarkSingle_errors_New_int_Message-10               	100000000	        53 ns/op	       4 B/op	       1 allocs/op
BenchmarkSingle_errors_New_int-10                       	163252323	        36 ns/op	       4 B/op	       1 allocs/op
PASS
ok  	github.com/nikolaydubina/go-bench-errors	149.863s