<p>Many of you have probably heard of this question or been assigned it in school or an interview. It seems simple:</p>
<p>for i := 0; i < len(str) / 2; i++ {
if str[i] != str[len(str) - 1 - i] {
return false
}
}</p>
<pre><code>return true
</code></pre>
<p>Except it isn't outside of ASCII land. Code points vary in byte length. Graphemes vary in code point length.</p>
<p>You can get something messy looking working for single code point Unicode subsets with unicode.DecodeLastRuneInString.
You can get something the kind-of-sort-of working with the extended text library's Form iterator, but it isn't in-place.</p>
<p>One component of what makes this problem so difficult is how difficult it is to get the number of characters in a Unicode string.</p>
<p>Please advise.</p>
<p>Edit: I am trying to build an in-place solution.</p>
<hr/>**评论：**<br/><br/>skarlso: <pre><p>Why not just reverse it, and compare? Runes go a long way too.</p></pre>venju: <pre><p>Characters may be multiple runes long (see grapheme clusters), and reversing at the rune level can make two identical grapheme clusters differ.</p></pre>skarlso: <pre><p>That's why I'm saying runes. Runes should keep the multibyte character like '你' in tack, I think. Though I have to cook up a short script to be absolutely sure about that. :)</p></pre>skarlso: <pre><p>Yep, this works absolutely fine:</p>
<p><a href="https://play.golang.org/p/_MD94Z3qhB" rel="nofollow">https://play.golang.org/p/_MD94Z3qhB</a> Playground to prove it.</p></pre>Fwippy: <pre><p>That doesn't work for multi-rune characters: <a href="https://play.golang.org/p/-S3T_Qqh12" rel="nofollow">https://play.golang.org/p/-S3T_Qqh12</a></p>
<p>Further reading: <a href="https://blog.golang.org/normalization" rel="nofollow">https://blog.golang.org/normalization</a></p>
<blockquote>
<p>Theoretically, there is no bound to the number of runes that can make up a Unicode character.</p>
</blockquote></pre>skarlso: <pre><p>Fair enough. Nice.</p></pre>chewxy: <pre><pre><code>asRunes := []rune(str)
l := len(asRunes)
</code></pre></pre>robpike: <pre><p>That (converting to runes) is also wrong in the presence of combining characters.</p>
<p>Human language is messy, that's all there is to it. So handling it correctly is in general also messy, and very difficult.</p>
<p>To answer the original question with a question: What is a character? The answer is nothing like as simple as you'd like. See <a href="https://blog.golang.org/strings" rel="nofollow">https://blog.golang.org/strings</a> for some background.</p></pre>skarlso: <pre><p>Hello Mr. Pike. I thought that using runes would solve this issue. Also, going with the rune reverse algo which is in the bare string package, would that solve the problem? I have a short playground here to prove it: <a href="https://play.golang.org/p/_MD94Z3qhB" rel="nofollow">https://play.golang.org/p/_MD94Z3qhB</a> which correctly preserve a multibyte character.</p></pre>TheMerovius: <pre><p>Again, the question is, "what is a character". Example:
<a href="https://play.golang.org/p/-tKQxgiezM" rel="nofollow">https://play.golang.org/p/-tKQxgiezM</a></p></pre>skarlso: <pre><p>Good point!</p></pre>chewxy: <pre><p>Agreed</p></pre>cjbprime: <pre><p>I guess I don't see the problem. Why would the divide-by-2 compare ever fail on realistic input? Yes, you could have different size characters, but if you do, I think you just guaranteed that it's not a palindrome?</p></pre>venju: <pre><p>Reversing the string at the byte level will make two identical multibyte characters differ from one another.</p></pre>
