Testing whether characters of a string are unique

That's pretty good. It's short. The early returns communicate intent well, and since the function is short, they don't get lost. Formatting is good, and the variable names, although very short, are common abbreveviations that work well here.

I'd consider making the argument const, since it's not modified by the function.

I've been out of C++ a long time, so I might be missing something here, but I'd consider auto c:s instead of auto &c:s. I can't see a reason to work with a reference when a character is cheap to copy.

One theoretical place that this code might fail is when the string contains multibyte characters.

For example UTF-8 encoding is being used and string passed to the function is "€Abبώиב¥£€¢₡₢₣₤₥₦§₧₨₩₪₫₭₮漢Ä©óíßä" although no character is being repeated, the function will return true.

But handling this case is not straight forward at all, as standard library doesn't provide any way to iterate over actual characters in a string.

@Wayne Conrad has made some good points. I just have a few additional ones:

• I'm not sure about checking against a maximum value, but I'll address it anyway.

  In case you'd like to change the default 256, consider having a template parameter. That way, the value can be changed at just one location. It'll also give more meaning to that value.

  template <std::size_t maxChars = 256>
  bool unique_chars(std::string &s) {
      if (s.length() > maxChars) return false;
      std::bitset<maxChars> bs;
  
      // ...
  }
  

• To slightly add on to Wayne's point about const, make sure the argument is const& so that no unnecessary copying is done.

• Consider renaming the function to something like areUniqueChars(). Otherwise, it may sound like it's returning the number of unique chars.

• In C++, prefer to put unary operators (such as & and *) next to the type:

  std::string& s;
  

  for (auto& c : s) {}
  

Critiques (even pedantic ones) are welcome.

There's no 'functional specification' included with your question, there is just the code. You should include a functional specification, a.k.a. "the problem" to be solved, or "the requirements". Without one, we can see what the code does, but cannot assess whether what it does matches what it is required .

If the (missing) functional specification said, "string", then perhaps it doesn't mean only std::string ... for example, it might also mean std::wstring.

If so then your unique_chars should be a template function (in the same way that std::basic_string is a template class).

If you were working with very wide multi-byte characters (e.g. std::char32_t) you would prefer std::set or std::unordered_set (a hash set) instead of a std::bitset.