Optimized code (without convertions)

func Ellipsis(str string, length int) string {
	str = strings.TrimSpace(str)

	const ellipsis = "..."

	cutPosition := 0
	for i := range str {
		if length == len(ellipsis) {
			cutPosition = i
		}

		if length--; length < 0 {
			return strings.TrimSpace(str[:cutPosition]) + ellipsis
		}
	}

	return str
}

Nice optimization! Using range to iterate runes avoids the []rune(str) allocation entirely. I've adopted your approach with one tweak: added an early return for the length < 3 edge case before the loop to keep the intent explicit, rather than relying on cutPosition staying at 0. Updated the PR -- thanks for the suggestion!

In my opinion

An early return that almost completely duplicates the main algorithm
for the sake of an extremely rare use case seems unnecessary.

I even abandoned the check if cutPosition == 0 {}

however, at least reuse the const ellipsis = "..."

Fair point -- the early return does duplicate logic for a rare edge case. I've simplified it: removed the early return and just use the ellipsis const for the fallback. Cleaner this way, thanks for the feedback.

Hi @samber, just a gentle ping on this one — I've incorporated the optimized rune iteration approach that @d-enk suggested and simplified the edge case handling. The fix prevents Unicode truncation in Ellipsis by operating on runes instead of bytes. Would love your review!

Thanks guys, for the work on this 👏

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.30%. Comparing base (d269b33) to head (ad37e77).
⚠️ Report is 9 commits behind head on master.

@@           Coverage Diff           @@
##           master     #796   +/-   ##
=======================================
  Coverage   94.29%   94.30%           
=======================================
  Files          18       18           
  Lines        2857     2860    +3     
=======================================
+ Hits         2694     2697    +3     
  Misses        149      149           
  Partials       14       14           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

I had surprise by using runes

Here the code won't do what is expected with Unicode character made of multiple Unicode char

https://go.dev/play/p/s-pb7mlpQVj

As mentioned here:

This bug could be fixed by counting the string grapheme cluster at github.com/rivo/uniseg .

If this policy is acceptable, I'll work on a PR for the fix!

Originally posted by @yasuaki640 in #520 (comment)

Using @rivo uniseg could be a solution. But as stated before adding a dependency is not recorded by @samber, so here checking what uniseg code do abd replicate it could be a solution

Another option would be to add github.com/rivo/uniseg in the comment of the helper for advanced cases.

@ccoVeille i'm going to add your example as a commented test case. If you want to make a PR to try to fix it, please do !