Years ago I was writing perl scripts and regexp became a default string manipulation approach for me. I even had a book dedicated to regexp intricacies. A decade and few languages later, regexp is still a tempting tool. Going with go, one can assume that statically compiled language with regexp matcher pre-compiled, should be blazing fast. Isn’t it? Lets test. Testing framework is a first class citizen in go not just for fun, but for profit as well.

Will test on a quite simple schenario of checking if string is empty or end with a question mark. Full code can be found at github.com/ataraskov.

Regexp function to test:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12

func WhatIsItRegexp(input string) string {
	isQuestion := regexp.MustCompile(`\?$`)
	isEmpty := regexp.MustCompile(`^\s*$`)
	switch {
	case isQuestion.MatchString(input):
		return question
	case isEmpty.MatchString(input):
		return empty
	default:
		return other
	}
}

Strings function to test:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10

func WhatIsItStrings(input string) string {
  switch {
    case strings.HasSuffix(input, "?"):
		return question
	case strings.TrimSpace(input) == "":
		return empty
	default:
		return other
	}
}

Benchmark results:

BenchmarkRegexp-16      120072   9792 ns/op   11030 B/op   177 allocs/op
BenchmarkStrings-16   71365939     16 ns/op       0 B/op     0 allocs/op

  The winner is obvious. Strings approach is more readable, faster, and cheaper. Though discarding regexp is not an option either. But going with regexp should be a last resort, IMHO ;)

Fri Jun 16, 2023 / 232 words / Golang Regexp Benchmark