strings: add CutPrefix and CutSuffix

[cespare]

Inspired by #40135, I have another proposal for an addition to the strings package (and by extension, I suppose, the bytes package).

The strings functions I most frequently wish for are variants of TrimPrefix and TrimSuffix which report whether they did any trimming.

// TrimmedPrefix returns s without the provided leading prefix string
// and reports whether it found the prefix.
// If s doesn't start with prefix, TrimmedPrefix returns s, false.
// If prefix is the empty string, TrimmedPrefix returns s, true.
func TrimmedPrefix(s, prefix string) (trimmed string, ok bool)

Lacking these functions, Go authors resort to fragile code, often writing a prefix/suffix literal twice or hard-coding its length:

if strings.HasPrefix(s, "myprefix") {
	s = strings.TrimPrefix(s, "myprefix")
	...
}

if strings.HasPrefix(s, "myprefix") {
	s = s[len("myprefix"):]
	...
}

if strings.HasPrefix(s, "myprefix") {
	s = s[8:]
	...
}

if t := strings.TrimPrefix(s, "myprefix"); s != t {
	// had prefix
	s = t
	...
}

Of course, a function like TrimmedPrefix is easy to write and can exist outside of the standard library. At my company we have these functions in an internal string helper package and they see regular use. But certainly I wouldn't pull in a dependency for such a tiny helper function and so when I'm working on my own projects I generally just use the above workarounds.

---

Here are some examples from the Go source tree along with how they could be altered to use TrimmedPrefix/TrimmedSuffix:

• src/cmd/go/internal/modload/load.go:

  if strings.HasPrefix(suffix, "/vendor/") {
  	return strings.TrimPrefix(suffix, "/vendor/")
  }
  

  becomes

  if v, ok := strings.TrimmedPrefix(suffix, "/vendor/"); ok {
  	return v
  }
  

• src/cmd/go/proxy_test.go:

  if !strings.HasSuffix(name, ".txt") {
  	continue
  }
  name = strings.TrimSuffix(name, ".txt")
  

  becomes

  name, ok := strings.TrimmedSuffix(name, ".txt")
  if !ok {
  	continue
  }
  

• src/testing/benchmark.go:

  if strings.HasSuffix(s, "x") {
  	n, err := strconv.ParseInt(s[:len(s)-1], 10, 0)
  	...
  }
  

  becomes

  if s, ok := strings.TrimmedSuffix(s, "x"); ok {
  	n, err := strconv.ParseInt(s, 10, 0)
  	...
  }
  

• src/testing/fstest/mapfs.go:

  if strings.HasPrefix(fname, prefix) {
  	felem := fname[len(prefix):]
  	...
  }
  

  becomes

  if felem, ok := strings.TrimmedPrefix(fname, prefix); ok {
  	...
  }
  

• src/mime/mediatype.go:

  if !strings.HasPrefix(rest, ";") {
  	return "", "", v
  }
  
  rest = rest[1:] // consume semicolon
  

  becomes

  rest, ok := strings.TrimmedPrefix(rest, ";")
  if !ok {
  	return "", "", v
  }
  

• test/run.go:

  if strings.HasPrefix(line, "//") {
  	line = line[2:]
  } else {
  	continue
  }
  

  becomes

  line, ok := strings.TrimmedPrefix(line, "//")
  if !ok {
  	continue
  }
  

• test/run.go:

  if strings.HasPrefix(m, "LINE+") {
  	delta, _ := strconv.Atoi(m[5:])
  	n += delta
  	...
  }
  

  becomes

  if d, ok := strings.TrimmedPrefix(m, "LINE+"); ok {
  	delta, _ := strconv.Atoi(d)
  	n += delta
  	...
  }
  

• src/runtime/testdata/testprog/traceback_ancestors.go:

  if strings.HasPrefix(tb[pos:], "goroutine ") {
  	id := tb[pos+len("goroutine "):]
  	...
  }
  

  becomes

  if id, ok := strings.TrimmedPrefix(tb[pos:], "goroutine "); ok {
  	...
  }
  

---

Update 2022-03-26: Changed proposed names to TrimmedPrefix/TrimmedSuffix (per @ianlancetaylor's suggestion).

[ianlancetaylor]

TrimmedPrefix and TrimmedSuffix ?

[mvdan]

I find if t := strings.TrimPrefix(s, "myprefix"); s != t { reasonable, but that might just be because I've gotten used to the pattern. One argument in favor of it not being descriptive/intuitive enough is that it's not completely winning over alternatives like HasPrefix+TrimPrefix.

[jimmyfrasche]

CutPrefix and CutSuffix so they group with the proposed Cut in docs?

[rsc]

It's unclear whether we need these, but at the very least we'd need better names.
Adding to the minutes in hopes that someone will come up with better names.

[narqo]

(With all respect) It feels the proposal lacks the use cases where the new functions required. Knowing that the current implementation of both strings.TrimPrefix and strings.TrimSuffix is the following

func TrimPrefix(s, prefix string) string {
	if HasPrefix(s, prefix) {
		return s[len(prefix):]
	}
	return s

}

can't one always use TrimPrefix for all the samples from the issue's description, dropping Has-checks? I.e.

- if strings.HasPrefix(suffix, "/vendor/") {
- 	return strings.TrimPrefix(suffix, "/vendor/")
- }
+ return strings.TrimPrefix(suffix, "/vendor/")

[cespare]

@narqo no. The places where you use TrimSuffix2 do something different if the test string contains the prefix/suffix.

For example, you suggested

- if strings.HasPrefix(suffix, "/vendor/") {
- 	return strings.TrimPrefix(suffix, "/vendor/")
- }
+ return strings.TrimPrefix(suffix, "/vendor/")

but those are not the same thing, because your version returns if suffix does not start with "/vendor/" and the original does not.

I listed the filenames that all my examples come from. You can go check out that example in context and you'll see it looks like this:

if strings.HasPrefix(suffix, "/vendor/") {
	return strings.TrimPrefix(suffix, "/vendor/")
}
return targetPrefix + suffix

[rsc]

Putting on hold with Cut.

[cespare]

I adopted @ianlancetaylor's suggested names.

In the intervening time since filing this proposal, I haven't thought of a better name. But I've used these functions constantly, and I've ended up copy-pasting them into many of my projects.

[cespare]

@rsc Cut has been proposed, accepted, and implemented. Can we un-hold this proposal now?